rotordynamic faults 2013

13
8/20/2019 Rotordynamic Faults 2013 http://slidepdf.com/reader/full/rotordynamic-faults-2013 1/13 Hindawi Publishing Corporation International Journal o Rotating Machinery Volume , Article ID ,  pages http://dx.doi.org/.// Review Article Rotordynamic Faults: Recent Advances in Diagnosis and Prognosis Ryan Walker, Sureshkumar Perinpanayagam, and Ian K. Jennions Integrated Vehicle Health Management Centre, University o Craneld, Bedordshire MK AL, UK Correspondence should be addressed to Ryan Walker; [email protected] Received October ; Revised January ; Accepted January Academic Editor: Arthur Lees Copyright © Ryan Walker et al. Tis is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. Diagnosis and condition monitoring in rotating machinery has been a subject o intense research or the last century. Recent developments indicate the drive towards integration o diagnosis and prognosis algorithms in uture integrated vehicle health management (IVHM) systems. With this in mind, this paper concentrates on highlighting some o the latest research on common aultsinrotatingmachines.Eightkey aultshavebeen described;theselectedaultsincludeunbalance, misalignment,rub/looseness, uid-induced instability, bearing ailure, shaf cracks, blade cracks, and shaf bow. Each o these aults has been detailed with regard to sensors, ault identication techniques, localization, prognosis, and modeling. Te intent o the paper is to highlight the latest technologies pioneering the drive towards next-generation IVHM systems or rotating machinery. 1. Introduction Te topic o diagnosing and prognosing aults in rotating machinery is an ongoing subject o research, with many developments published in a range o conerences and jour- nals annually. Tis research has the potential to become even more relevant in the coming years due to the rise o IVHM, inwhichthedrivetowardscondition-basedmaintenanceand wholevehiclemonitoringplaysavitalrole.Tispaperintends to survey some o the recent developments in the eld, with the aim o summarizing some o the more promising studiesandtrendswithrelevancetoutureIVHMsystemsor rotating machinery. Modern day rotating machines operate with a high level o reliability, and yet the drive or ever increased operation and decreased unscheduled maintenance is providing addi- tional challenges or industry. Te airline industry provides a currentexample othisdesire,withairlinespushingmanuac- turerstoenableshorterturnaroundtimesandtokeep aircraf intheairlonger,increasingcostbenet. Despite thehighlevel o reliability, the rotordynamic aults detailed in this paper remain aspects which require consideration in this drive or increased reliability and improved maintenance procedures []. In order to ully understand and summarize the trends and developments in this area, several hundred recent conerence and journal papers have been studied. Overall trends have beenhighlightedanddiscussedalongsidespecic papers o relevance. It is intended that the work should pro-  vide a broad reerence and summary or working engineers on some o the latest developments in rotordynamic ault diagnosis and prognosis, with specic application to papers o industrial relevance or the drive towards uture IVHM systems. In order to ully study the diagnosis and prognosis o rotordynamic aults, it has been deemed necessary to break down the topic o rotordynamic aults into sections dened as ollows: Sensors: sensors commonly used or diagnosis o spe- cic aults. Fault Identication: diagnosis and root cause detec- tion. Localisation : locating a specic ault within a complex system. Prognosis: prognosis o components and remaining useul lie.  Modelling : simulation o rotordynamic aults.

Upload: pastcal

Post on 07-Aug-2018

226 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Rotordynamic Faults 2013

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 113

Hindawi Publishing CorporationInternational Journal o Rotating Machinery Volume 983090983088983089983091 Article ID 983096983093983094983096983094983093 983089983090 pageshttpdxdoiorg983089983088983089983089983093983093983090983088983089983091983096983093983094983096983094983093

Review ArticleRotordynamic Faults Recent Advances inDiagnosis and Prognosis

Ryan Walker Sureshkumar Perinpanayagam and Ian K Jennions

Integrated Vehicle Health Management Centre University o Cran1047297eld Bedordshire MK983092983091 983088AL UK

Correspondence should be addressed to Ryan Walker rbwalkercran1047297eldacuk

Received 983096 October 983090983088983089983090 Revised 983089983093 January 983090983088983089983091 Accepted 983090983089 January 983090983088983089983091

Academic Editor Arthur Lees

Copyright copy 983090983088983089983091 Ryan Walker et al Tis is an open access article distributed under the Creative Commons Attribution Licensewhich permits unrestricted use distribution and reproduction in any medium provided the original work is properly cited

Diagnosis and condition monitoring in rotating machinery has been a subject o intense research or the last century Recentdevelopments indicate the drive towards integration o diagnosis and prognosis algorithms in uture integrated vehicle healthmanagement (IVHM) systems With this in mind this paper concentrates on highlighting some o the latest research on commonaults in rotating machines Eightkey aults havebeen described the selected aults include unbalance misalignment rublooseness1047298uid-induced instability bearing ailure shaf cracks blade cracks and shaf bow Each o these aults has been detailed with regardto sensors ault identi1047297cation techniques localization prognosis and modeling Te intent o the paper is to highlight the latesttechnologies pioneering the drive towards next-generation IVHM systems or rotating machinery

1 Introduction

Te topic o diagnosing and prognosing aults in rotatingmachinery is an ongoing subject o research with many developments published in a range o conerences and jour-nals annually Tis research has the potential to become evenmore relevant in the coming years due to the rise o IVHMin which the drive towards condition-based maintenance andwhole vehicle monitoring plays a vital role Tis paper intendsto survey some o the recent developments in the 1047297eldwith the aim o summarizing some o the more promising

studies and trendswith relevance to uture IVHM systems orrotating machinery

Modern day rotating machines operate with a high levelo reliability and yet the drive or ever increased operationand decreased unscheduled maintenance is providing addi-tional challenges or industry Te airline industry provides acurrent example o this desire with airlines pushing manuac-turers to enable shorter turnaround times and to keep aircrafin the air longer increasing cost bene1047297t Despite the highlevelo reliability the rotordynamic aults detailed in this paperremain aspects which require consideration in this drive orincreased reliability and improved maintenance procedures[983089]

In order to ully understand and summarize the trendsand developments in this area several hundred recentconerence and journal papers have been studied Overalltrends have been highlighted and discussed alongside speci1047297cpapers o relevance It is intended that the work should pro-

vide a broad reerence and summary or working engineerson some o the latest developments in rotordynamic aultdiagnosis and prognosis with speci1047297c application to paperso industrial relevance or the drive towards uture IVHMsystems

In order to ully study the diagnosis and prognosis o

rotordynamic aults it has been deemed necessary to break down the topic o rotordynamic aults into sections de1047297nedas ollows

Sensors sensors commonly used or diagnosis o spe-ci1047297c aults

Fault Identi1047297cation diagnosis and root cause detec-tion

Localisation locating a speci1047297c ault within a complexsystem

Prognosis prognosis o components and remaininguseul lie

Modelling simulation o rotordynamic aults

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 213

983090 International Journal o Rotating Machinery

Trough the study o the topics listed it is useul to placethe research conducted in this paper into context with regardto real-world applications Further to this it is intended toidentiy potential areas where more research is required inorder to push some o the recent technologies highlighted orthis study into industry

Anderson [983090] provides a summary o maintenance timebreakdown or a collection o military aircraf Tis indicatesthat as much as 983092983092 o on-aircraf maintenance time (whichin turn accounts or 983097983088 o total maintenance operations) isconsumed with inspection alone Te techniques addressedin this paper enable the maintenance to be more inormedand targeted with inventory ready when needed providinga signi1047297cant contribution to reducing maintenance time andcost

As the topic o rotordynamic aults is very large areao research the scope o this paper has to be limited Techoice o aults has been made afer considering the worksby Muszynska [983091] and Bently [983092] both o whom consider theundamentals o common aults in much detail Out o thewide range o possible rotordynamic aults eight have there-ore been selected Due to the general reliability o the currentgeneration o gas turbines aults alling outside o the eightlisted have been classed as ldquouncommonrdquo or the purposes o this study Tis decision was made by assessing the severity o each ault dependences on other aults and the level o research dedicated to diagnosis o each ault

Te scope o this paper is thereore con1047297ned to the ol-lowing aults

(i) unbalance

(ii) misalignment

(iii) rub and looseness

(iv) 1047298uid-induced instability

(v) bearing aults

(vi) shaf cracks

(vii) blade cracks

(viii) rotor bow

Each o these aults is varied and some are more commonthan others Te consensus rom the works reviewed withinthis paper is that unbalance is the most commonly occur-ring in part due to the 1047297ne tolerances applied to modernmachinery and also due to the links between unbalance andother aults Misalignment can also be identi1047297ed as another

common ault Te particular importance o these two aultsis highlighted by Domes [983093] who discusses common aultsrom the perspective o Rolls RoyceBearing ailure is perhapsthe most diverse ault here and has the highest concentrationo research in various areas Te decision has also been madeto separate aults occurring on the shafs o a system to thoseoccurring on rotors as these can be classi1047297ed as distinctaultsmdashparticularly with regards to localization and mainte-nance Despite these conclusions quantiying the aults withregards to the rate o occurrence was not possible due to thelack o commercially available data and each o the eightaults hasthereore beeninvestigatedto an equal level o detailin this paper

It should be noted that although the aults have beenseparated into eight categories they are by no means mutually exclusive Dependencies exist between many o the aults Acommon example o this is the interaction between unbal-ance and misalignment (outlined by Bently [983092]) Fault chainscan become even more complex misalignments leading to an

unbalance which causes a rub or an example Some workscontained herein detail single ault analysis and a ew reerto two or more aults Such ault dependencies are one o thelimiting actors when moving technologies rom the lab intoindustry An example o this would be a system capable o diagnosing unbalance aults without consideration or mis-alignment as a root cause may render such a system ineffec-tive when applied to a real-world scenario

Te selected papers have been urther re1047297ned based uponrelevance to aeroengine gas turbines and publication date(with works rom the last three years preerred)

As research into the diagnosis and prognosis o aults inrotating machines is a developing topic the application o such technologies in industry has not yet reached a stagewhere there are common solutions or even a set o estab-lished procedures to ollow Tis holds particularly true withregards to the broader 1047297elds o IVHM aking the state-o-the-art Joint Strike Fighter (JSF) program as one example o the current level o technology in circulation although theautonomic logistics proposed by Hess et al [983094] or the vehicleis undeniably impressive it is still a long way rom imple-menting some o the recent technologies claimed to besuccessul in a lab-based environment

Although as mentioned an established and commonly applied set o procedures and standards or IVHM has notyet emerged with regards to rotating machinery a numbero standards and procedures can be reerred to with regardsto detailing and quantiying the aults Tese include suchaspects as covered by ISO 983090983097983093983091983089983097983097983097 which details correctbalancing procedures and levels o mechanical vibrationAnother example would be in the UK Ministry o DeencemdashMilitary Aviation Authority (983090983088983089983088) JAP 983089983088983088A-983088983089mdashMilitary aviation engineering policy and regulation where a mentionis made o procedures or debris and vibration monitoringTis indicates a gradual move towards common ground orprocedures and standards or IVHM systems Despite thisresearch remains varied and separating the research withpotential or moving beyond the lab-based environment intoindustry can be difficult to identiy at 1047297rst glance

Te discussion contained herein thereore takes into

account the above detailed considerations and constraintswith the aims o the paper as ollows

(i) Highlight key examples o the latest research in eightcommon rotordynamic aults

(ii) Summarise current trends identi1047297ed rom the study

(iii) Detail the current state o the art and the uturework required or next generation IVHM systems orrotating machinery

2 Rotordynamic Faults

Te ollowing section details the aorementioned commonrotordynamic aults with regard to recent physics-based

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 313

International Journal o Rotating Machinery 983091

simulation work and traditional data-driven methodsResearch in this area o rotordynamics is particularly wide-spread and so this paper outlines only a ew recent areaso research Although not being o mainstream interest alist o commonly used sensors has been included able 983089represents a summary o the inormation collected and

detailed as a part o this paperAs discussed a wide selection o papers have beenreviewed or consideration in this paper o which it is notpossible to describe them all in detail In an attempt tohighlight some o the common themes across many o worksreviewed able 983089 has been created able 983089 by no meansdetails all relevant methods and approaches currently beinginvestigated however it details themes which were oundto be common across many areas o current research It isintended or the detailed table to compliment the importantworks described in more detail in the ollowing section o thepaper separated by ault type

983090983089 Unbalance Tis is one o the most common rotordy-namic aults [983092] every rotating machine has an inherent

degree o unbalance Unbalance as a ault can thereorebe de1047297ned as unbalance outside o a given tolerance levelA recent piece o research which demonstrates the ongoingdevelopment o data-driven techniques is that by Ganeriwalaet al [983095] who tested a technique or measuring operatingde1047298ection shapes (ODS) in order to detect unbalance casesTese studies were conducted on a machine ault simulator (aphysical simulator o common aults in rotating machines)mdashsuch simulators have the advantage o recreating aulty con-ditions quickly and easily enabling a new dimension to data-driven diagnostic techniques It is however worth mention-ing that data or these experiments were collected using 983089983092accelerometers which are easy to apply to such a simulatorbut it may be much more difficult to con1047297gure these many sensors on a complex system Despite this the paper achieveditsaim in proving thehypothesis ldquowhen an operating machinebecomes unbalanced its ODS will changerdquo

Regarding physics-based simulation o unbalance as aault the work by Sudhakar and Sekhar [983096] described amodel-based method or ault identi1047297cation using a minimalsensor suite Tis is achieved through the analysis o trans-

verse vibrations at a single location Troughout the paperthree different approaches are studiedmdashleast squares mini-mization equivalent loads minimization and vibrationmini-mization A reduced error is ound to occur using a proposed

modi1047297cation to the typical equivalent loads minimizationapproach Te work is o note due to the authors requirementto both identiy and locate (localize) unbalance

Localization and prognosis o unbalance pose a uniqueseto challenges andas a result research is still somewhat lim-ited in these areas One recent work which claims to localiseunbalance accurately is by Yang and Hsu [983097] the authorsuse trending data and reasoning systems to locate unbalanceand shaf bow across a system Quick diagnosis is achievedby avoiding the study o all ault combinations and theauthors claim the ability to localise shaf bow and unbalancein large rotating machines Te techniques rely extensively onprevious searches and are limited to large stable machines

operating at a speci1047297c RPM Remaining useul lie o unbal-ance is difficult to predict due to complicating actors Asdiscussed the ault dependencies are such that in many casesthe unbalance itsel will not be the ailing actor instead anunbalance may cause a rub which then leads to damagewhichcan be prognosesSuch combinations o aults and underlying

causes or unbalance lead to the need or remaining useullie predictions to be made based upon the exact nature o aspeci1047297c ault

983090983090 Misalignment Tis is another common ault which canpotentially in1047298ict considerable damage in rotating machinesAs with unbalance misalignment in a whole system can becomplicated by secondary aults (eg a misalignment whichcauses a rub) El-Shaei et al [983089983088] are an example o theongoing research in this area in this case a unique combina-tion o angular misalignment and oil whipwhirl is detailedTe authors describe how small degrees o misalignment canbe utilized in order to prevent the onset o 1047298uid-induced

instabilities tested through the use o a lab-based test rigSuch research presents a new dimension in looking atcommon rotordynamic aults with aspects that could beapplied to uture design or IVHM systems

El-Shaei et al [983089983089] demonstrate interesting research intomisalignment rom the perspective o physics-based simula-tion Te authors construct mathematical models o a simplerotorsystem witha misaligned coupling andcollect harmonicresponse data rom this to assess the severity o differentmisalignment cases Such models are useul throughout thelie o rotating machinesmdashrom design to implementationalthough again successul validation with experimentally obtained data is key Te authors highlight the act thatalthough misalignment is a prevalent and serious ault nocomprehensive research has been perormed or treating thisproblem Te methodology applied includes deriving theenergy expressions applying the Ritz series method con-structing the equations o motion and then using the har-monic balance method to look or multiharmonic responsesTe paper demonstrates the ongoing research to understandand model the undamentals o such aults in order thatimproveddiagnosisand prognosis methods can be built uponsuch knowledge

As with unbalance localization and prognosis o mis-alignment is a complex topic to research Studies such as El-Shaei et al [983089983089] can make accurate predictions or misalign-ment in a simple system with one coupling However real

systems (eg aircraf gas turbines) have many potential loca-tions o misalignment Tis is an area where ew researchershave made an impact Remaining useul lie predictions ormisalignment is complicated or the same reasons as withunbalance Villa et al [983089983090] discuss statistical diagnosis o misalignment aults with reerence to prognosis Te authorsuse the example o a wind turbine or their studies butstress the applicability to other systems Differentiation withunbalance aultsis also covered (as these two aultsare closely linked) Unlike the work by El-Shaei et al [983089983089] emphasis isgiven to the machine in question operating over a wide rangeo operating speeds and conditions Tis is achieved throughthe use o an angular resampling method Prognosisis tackled

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 413

983092 International Journal o Rotating Machinery

983137983138983148983141 983089 Rotordynamic ldquoault tablerdquo

Faults Sensors Fault identi1047297cation

techniques Localisation o ault Prognosis

Work o note in aultmodelling

UnbalanceAccelerometer

Velocity transducer

Proximity transducer

Operating de1047298ectionshapesime-requency

analysis and many others

Neural-network modelling and

time-requency analysis

FCM-Markov model

Unbalance in ull-engineFEA modelEarly detection inaeroengines

High radialloads andmisalignment

AccelerometerTermocoupleDisplacement

transducerAngular sensor

Bearing temperatureobservation advancedSofware-driven1047297ltering methods

Various vibrationanalysis techniques

Prognosticmodelling usingsimulated vibrationdata

Modelling therelationship betweenmisalignment and otheraultsHarmonic balancemethod

Rub andlooseness

AccelerometerVelocity transducer

Optical sensor

Single valuedecompositionime-requency analysis

Least-squares 1047297ttingapproach

Adaptive waveletdenoising orprognostics

Rob and looseness usingFEA models Wholeengine modellingtechniques

Fluid-inducedinstability

Proximity probesKeyphase transducer

Full spectrum plotsLarge amplitudesubsynchronous vibrationDetection

Blade tip clearanceand loadingdistribution

Cross-coupledstiffness anddirect-dampingstudies

Nonlinear modellingaccuracy o model basedDiagnostics numericalmodelling

Bearing ailure

AccelerometerVelocity transducer

Displacementtransducer

Proximity probes

Postailure analysis oruture ailure diagnosisDynamictime-requency analysis

Statistical change inthe bispectraldomain

Numericalmodelling o aeroengine bearingsModel-basedprediction o remaining useul lieOn-boardprognostics

High and low 1047297delity dynamic systemmodellingComparisons o algorithms tounderstand bearing vibration phenomena

Shaf cracks

Keyphase transducer

Proximity probes

Higher order spectra

Neural networksGrid support vectors

Forced response

measurements

Nonintrusive

torsion vibrationmonitoring

Statistical analysis FEA

modelling

Blade cracks Keyphase transducer

Proximity probes

Hilbert-HuangtransormationModel-baseddiagnosticsMorletwaveletquantumtechniques

Novel bearingstructures

Active magneticactuator

Global and localasymmetry modellingFinite element modelling

Rotor Bow Keyphase transducer

Proximity probesSeismic transducer

Examination o ldquoslow rollrdquo dataModel-baseddiagnostics

Partial mathematicalmodelling

Statistical symptomsanalysis

Advantages o aultmodelling oridenti1047297cationNonlinear modellingtechniques

through the use o a statistical diagnosis algorithm based onthe signi1047297cance level o the aults in question

983090983091 Rub and Looseness Rub is always a secondary ault (iea product o another ault such as looseness) and can leadto atigue and wear Rub and looseness can create com-plex vibration signals which are difficult to diagnose usingtraditional methods Modeling and simulation o rub andlooseness aults have been considered in several recentworks

Tis includes Ngolah et al [983089983091] which detail the moni-toring and diagnosis o common aults (including rub and

looseness) based upon a three-layer Arti1047297cial Neural Net-work (ANN) A series o 983089983088 key perormance indicators wereidenti1047297ed and used as training Te authors test the systemin a lab environment but stress the applicability to industrialapplications Te research indicates one o the latest methodso research which enables the implementation o diagnosistechniques It is a useul tool or rub and looseness studiesas it incorporates a variety o aults which could ldquounderlierdquosuch a ault Despite this the research relies on clear eaturesor each ault which can be much easier to identiy in a labenvironment as opposed to ldquonoisyrdquo industrial applications

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 513

International Journal o Rotating Machinery 983093

Lu et al [983089983092] have perormed several studies into rub andloosenessincluding thisexamplein which a 1047298exibly mountedshaf has an induced rub due to a range o contact ringsTe study ocusses on the potentially dangerous effects o rubin causing excessive nonsynchronous and chaotic vibrationsTe links with unbalance and misalignment are discussed

and detailed Te considerations or real-world cases in thedescribed research are considerable as part o the drivetowards ull understanding o the nonlinear effects o rub andlooseness

Localization o rub and looseness across whole systemsis relatively lightly studied in the literature Many works(including those already cited) look at single or dual-rotorsystems where localization o such aults is not an issue Inan industrial setting complex systems may comprise many rotors in several compressor and turbine stages signi1047297cantly complicating diagnosis o such aults Research into progno-sis o rotor-stator rubs lies mostly within the domain o data-driven techniques Modeling and simulation research can beused to support data-driven techniques or prognosis andcondition-based monitoring Han et al [983089983093] is an example o this the authors use 1047297nite element modeling to construct adual rotor model Various types o rub-impact are then stud-ied Such studies can provide a wide range o inormationwhich can then be combined with data obtained rom livesystems potentially with seeded aults in order to constructaccurate remaining useul lie predictions As pointed out by the authors one key advantage o simulation is the ability to study more complex systems with a higher number o rotors which is used throughout this research Tis presentsa different approach to identiying eatures or identi1047297cationo rub and looseness

983090983092 Fluid-Induced Instability Fluid-induced instabilities(ofen reerred to as whip and whirl) are potentially very serious aults which can result in wear atigue and extensivedamage to machine components Such instabilities can beound in interstage seals 1047298uid lubricated bearings andblade-tip clearances Research into simulating and modeling1047298uid-induced instability has produced several works o interest to ault diagnosis o rotating machines in the lastew years De Castro et al [983089983094] are a good example wherenonlinear mathematical models are prepared or a rotor-bearing system Te models are then used to predictinstability thresholds Te authors consider a test case againsta power plant turbine and a test rig thereore validating the

simulations Te case o unbalance aults causing whip andwhirl phenomena is also considered Te main conclusionthereore drawn rom the work is that the authors concludedthat nonlinear hydrodynamic journal bearing models enablesufficiently accurate simulations or predicting instability thresholds Fan et al [983089983095] represent an example o work romthe perspective o aeroengine turbines In this case startupconditions are studied using a ull Hilbert spectrum Te aimbehind the paper is to accurately predict the point at whichwhip and whirl occur thus enabling this to be avoided at thedesign stage Such 1047297ndings could potentially also be used toidentiy whip and whirl as the case o a ault afer a period o wear in the operating machine

Prognosing 1047298uid-induced instability is a relatively lightly researched topic Fluid instabilities can be covered as part o extensive research into remaining useul lie o bearings Tepotential exists or modeling and simulation techniques suchas those detailed above to become a part o prognosis or1047298uid-induced instabilities due to the act that it can be very

difficult to seed such aults into live systems or testing andevaluation As with other aults detailed in this paper many studies have been perormed with the aim o describing 1047298uid-induced instabilities based on the measurement or simulationo single (or occasionally dual) rotor setups Physics-basedsimulation with the aim o localising 1047298uid instability aultsacross a whole system can be limited by the complexity o both the ault and the system hence the simpli1047297cation tosingle rotor-stator bearing systems

983090983093 Bearing Failure An area where data-driven techniquesare still providing the basis o much research in the 1047297eld o rotordynamics is that o bearing ailure Te title ldquobearingailurerdquo can cover a wide range o potential issues whichcontinue to be studied in detail Faults can occur in all kindso engine bearingsmdashthe inner andouter case the cage andtherolling elements 1047298uid-induced instabilities (addressed in aseparate section) lubrication and the complexities o activemagnetic bearings to name some examples All types o bearing relevant to rotating machinery are the subject o ongoing research and this subject has the potential to ormseveral separate papers As a brie highlight some recentexamples are discussed as ollows

Data-driven techniques have enabled accurate bearingdiagnostics and prognostics to be described or a rangeo rotordynamic systems Despite the prevalence o data-driven research in this area research rom a physics-basedsimulation perspective has also recently produced someinteresting papers o relevance to condition monitoring andhealth management o rotating machinery

Tis includes Kappaganthu and Nataraj [983089983096] in whichrolling element bearings have been studied through the useo nonlinear models Te included nonlinearity in this caseis clearance and the model is then used in order to study chaotic motions in particular the regions o chaotic responseTe research orms a part o an ongoing drive to develop anaccurate model-based diagnostic technique or rolling ele-ment bearings taking into account clearance nonlinearitiesand chaotic responses

Guptaetal[983089983097] demonstrate another example o thelatestresearch into instability and chaos in rolling element bearingsthrough high-1047297delity simulations Tis detailed and complexstudy involves the application o a novel scheme to analyzethe quasiperiodic response o the system combined with anonautonomous ldquoshootingrdquo method Tis workhighlights thelevel o detail to which nonlinearities and complex nonlinearmotions in bearings are beginning to be understood andaccurately modeled Again such work presents the potentialor design o IVHM in uture evolutions o the research

As so much research hasbeen perormed (and is ongoing)into bearing aults across a wide variety o mechanicalsystems both prognostics and localization o bearing aults

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 613

983094 International Journal o Rotating Machinery

have been researched in somewhat more detail than someo the other aults detailed here Despite this much work still needs to be perormed in order to translate some o this core research into industrial applications Research suchas that detailed above has made signi1047297cant advances intodetermining bearing ailure as the root cause o malunction

Detecting which bearing is ailing across a complex systemhas received somewhat less research Bearing prognostics isanother area with much ongoing research being perormedmdashboth in the simulation and data-driven domains o givean example Hong et al [983090983088] combined grade lie andextensive mathematical modeling techniques in order to pro-duce prognostic models or aeroengine bearings Te resultsare described by the authors as ldquopractical and veri1047297ablerdquoAlthough a number o similar recent studies existthis work iso note or the extent o the studies perormed which includebearing test stand run-to-ailure validation Te lab resultsappear impressive this research has yet to be applied andtested in real lie applicationsmdashindicating that despite thenumber o parameters considered it is still not possible tomodel naturally occurring phenomena sufficiently

A large body o work in this area also exists romBorghesani et al [983090983089] where several examples o data driventechniques can be seen applied to a wide variety o bearingtypes An example o recent developments involves the use o ldquocepstrum prewhiteningrdquo in order to remove sufficient noiseor accurate bearing diagnosis and prognosis Tis work isparticularly noteworthy due to the emphasis on real-worldapplications where the traditional lab-based techniques o order tracking and synchronous averaging do not providesufficient noise removal or harsh industrial environmentsTe addition o such techniques is a crucial step in developingthe current generation o diagnosis and prognosis algorithmsor use in uture IVHM systems

983090983094 Shaf Cracks Another potentially serious ault in rotat-ing machinery is shaf cracks and so early detection o any such ault is highly important Methods o crack ormationandpropagation canbe diverse andrange rom highand low-cycle atigue to stress corrosion Simulation and modelingo shaf cracks can have signi1047297cant advantages over data-driven methods Perhaps the most obvious advantage is therelative simplicity o inserting a ault into orexample a 1047297niteelement model as opposed to seeding a ault in a workingindustrial machine As such research into shaf cracks has

been progressing steadily with the corresponding increasesin computing power

A clear synergy between data-driven and physics-basedsimulation research can be implied by a number o recentworks o research An example o recent advances rom a datacollection perspective is Li et al [983090983090] which details statisticalmodels based on historical data or condition monitoringpurposes Tis unique work uses the human auditory systemas inspiration or enriching methods o mechanical aults andeatures extraction Te results indicated by the paper areperhaps not as convincing as some other methods discussedin this paper it describes an interesting ldquooutside the boxrdquomethod o tackling common problems

From a modeling perspective Bachschmid et al [983090983091]cover a wide range o vibration phenomena in order todevelop a model-based identi1047297cation and severity procedureTis work is noted or its thoroughness in modeling pro-cedure including accurate modeling o the crack breathingmechanism A combination o high and low 1047297delity models

is validated through experimental study and ldquoexcellentrdquoaccuracy is claimed by the authors in detecting crack positionand depth through the use o the proposed model-baseddiagnostics

Te nature o shaf cracks has resulted in a wide variety o research beingperormed intoboth localization and prognos-tics o these aults (indeed the two topics can be consideredrelated) Recent examples o work in this area include Inoueetal[983090983092] who detail crack localization using orced responsemodeling In this case a test rig was constructed consistingo a circular shaf supported by two bearings Frequency-domain data was used to create a localization algorithmdesigned in combination with an FE model Although thisresearch provides good accuracy o localization in a labenvironment it remains untested in a more complex system(eg a ull gas turbine)

Sonnichsen [983090983093] describes 1047297nite element modeling o crack propagation with validation against experimentalresults provided to demonstrate the validity o such modelingtechniques Te paper concentrates on natural requenciesand resonance curves Whilst the authors claim improve-ments in the ability to understand such aults again thesystem in question is quite simplemdashand such FEA models aredifficult to scale up to ull size applications

983090983095 Blade Cracks Blade cracks i allowed to develop canresult in serious consequences Cracks can orm due to highcentriugal stresses across operational cycles (in the case o an aircraf gas turbine or example start up and take off through landing and taxi) As excessive crack growth canlead to catastrophic rotorblade ailure early detection andprognosis o such aults are essential As with shaf cracksphysics-driven simulation o blade cracks is an area o sig-ni1047297cant research Tis varies rom high-1047297delity 1047297nite elementmodels to low-1047297delity system and mathematical models Terecent work demonstrated by Green and Casey [983090983094]is agoodexample o recent mathematical modeling rom a diagnosisperspective In this paper the authors concentrate on early detection using global and local asymmetry crack models 983090X

harmonic components are identi1047297ed as key areas or the early detection o blade cracks however again this paper suffersrom being applied and tested on a relatively simple systemwhich may not scale up to a ull size turbine

Sonnichsen [983090983093] demonstrated high-1047297delity modelingthe authors used FEA to model crack growth making com-parisons and validating against an experimental rig Tiswork is particularly interesting as it outlines the advantagesand drawbacks with the latest state-o-the-art modelingtechniques

Localization and prognosis o blade cracks have alsobene1047297tted rom recent advances in simulation and modelingSawicki et al [983090983095] contains details o work on a novel active

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 713

International Journal o Rotating Machinery 983095

magnetic bearing system or use in the early detection local-ization and prognosis o blade cracks Again the emphasis ison early detection with the bearings used to excite the systemin order to obtain optimum response vibrations or analysisTe authors admit the approach hassome merit in diagnosingblade cracks however it is in the early stages o development

and work are ongoing FEA has also been used extensively to support blade crack prognostic tools Xiang et al [983090983096] arean extensive example o recent work In this case a numbero advanced FEA methods are applied to produce accurateFEA solutionsmdashthese include surace-1047297tting techniques andthe contour-plotting method Te authors experimentally

validate their work and suggest that it can be applied toprognosis and quantitative diagnosis o blade cracks Despitethe claimed advances again the scalability o such research toull size turbines is an issuemdashparticularly with regard to thecomplexity o the FEA models

983090983096 Rotor Bow Rotor bows can be a primary source o unwanted vibration in gas turbines Te main cause o arotor bow (rotor bows do not include bows due to gravity) isthermal differences in a system caused by operating con-ditions It is noted by Domes [983093] that this nonsymmetricalthermal distribution can cause excessive unbalance to theextent where a gas turbine will not start correctly Such rotorbows are common on start up or shut down and are ofenaccounted or in operational procedures However i thermalldquohot spotsrdquo exceed a given tolerance level they can causepermanent unbalances due to rotor de1047298ections Such rotorbows can lead to other aults includingrubbing andloosenesswhich complicate isolation and localization

raditional data-driven techniques or detecting rotorbows involve combinations o slow roll and vibration data[983090983097] More recently mathematical modeling techniques suchas that detailed by Meagher et al [983091983088] have been used inorder to diagnose residual rotor bows and differentiate theseaults rom other sources o unbalance Te authors o thispaper build upon established methods or the models and areunique in that they concentrate on response at the bearingpoints Tis is perhaps more useul orindustrial applicationsas vibration inormation is more readily available at bearingpoints rather than intrusive proximity probes which are ofenused in lab work

Te little work that exists on attempting to localise rotorbows across complex systems tends to be data-driven in

nature see Gaka and abaszewski [983091983089] where the authorsused statistical symptoms based on known data as a methodo diagnosing and prognosing a number o aults includingrotor bows and unbalance Tis paper attempts to address theirregularities and 1047298uctuations that occur over a long servicelie In order to achieve this a modi1047297ed energy processormodel is created using data drawn rom large steam turbinesover long periods o lie

Prognosing rotor bows is a complex subject As rotorbows are ofen caused by temperature de1047298ections makingpredictions or remaining useul lie and potential utureproblems lies not only in the realm o mechanical rotor-dynamics but also to some extent in thermodynamics Te

recent work detailed by Sinha [983091983090] is o note or detailingdiagnosis and quanti1047297cation o various rotordynamic aultsand describing the advantages o mathematical modelingover traditional vibration-based approaches Te topic o scalability regarding FEA models is discussed including anargument or the use o partial (simpli1047297ed) mathematical

models or large complex systemsAnother two works which are o interest with regard to

modeling o rotor bows include Shen et al [983091983091] where theauthors modeled a rotor-bearing system with a permanentrotor bow looking at the impact o secondary aults such asrub Te study o ault combinations in this paper is useulor ault differentiation studies however the authors study a permanent initial rotor bow Tis thereore does not takeinto account developing or worsening aults and the different

vibration phenomena that are observed as such aults aredeveloping

Lees et al [983091983092] describe the importance o model-basedault identi1047297cation techniques and outline recent research in

the area providing a good reerence paper or more researchon this speci1047297c ault

3 Discussion

983091983089 Sensor Suites Te subject o sensor suites is o greatimportance with regard to uture industrial applications Onone hand advanced and complex sensor suites generally enable improved ault localization and diagnosis howeverthe added complexity and cost has resulted in many o thesesystems being omitted rom the latest generations o rotatingmachinery in industrial applications From an industrialperspective the cost bene1047297t o additional sensors needs to

be signi1047297cant in order to justiy this approach In response anumber o works listed including [983096] as a good example takethe approach o achieving the same objectives using a greatly reduced sensor suite In many ways the required sensor suiteor many o the currently researched ault diagnosis andprognosis techniques provides an indication o the ability o the techniques to be used practically in industrial situationsA complex suite may return a very high success rate on alab-based rig however the impracticalities o mounting suchsuites on a real-world application negate the advantages

Further to this the consideration orsensor position mustbe considered Te use o proximity probes and keyphasortransducers has very clear advantages in a number o situ-

ations however the intrusive nature o the sensor preventsapplication in a number o complex systems With this inmind a number o the more promising studies listed rely onthe simple suiteo an accelerometer placed in positions whichare relatively remote rom the sources o vibration potentially with a noisy transer path

983091983090 Diagnosis It can be seen rom the research outlined inthe previous section that the diagnosis o aults in rotatingmachinery is a subject o much ongoing research Tisinvolves the improvement and development o traditional

vibration monitoring techniques development o new data-driven technologies and novel research into physics-based

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 813

983096 International Journal o Rotating Machinery

simulation and modeling In many cases these topics o research are dependent on one another or reasons o vali-dation veri1047297cation and speed o analysis In several casesit can be seen that multiple aults have been modeled orthe purposes o identi1047297cation and isolation However nostudies have yet been perormed which deal with all o the

aorementioned aults All o these aults are intrinsically related to one another Complex combinations o aults havebegun to be analyzed with the emphasis on developing new diagnosis techniques Physics-based modeling has proven toprovide signi1047297cant advances with regard to speci1047297c aultsnotably shaf and blade cracks where techniques such as FEAenable much easier aster and cheaper test data then seedingaults into live systems

It can be noted however that both high and low-1047297delity modeling techniques are being applied to cutting edgeresearch or all o the listed rotordynamic aults (and othersnot detailed in this paper) In addition to the advantages inthe speed o obtaining test results physics-based simulationis providing another dimension to data-driven techniquesSystem models are being used as part o logic and reasoningsuites in the identi1047297cation and differentiation o variousaults High-1047297delity models enable simulations o ault com-binations which are not possible practically or are pro-hibitively expensive in live systems

Nevertheless results obtained rom modeling studies stillneed to be validated against proven data-driven techniquesbeore implementation in industrial applications is possibleIt is also o note rom the literature reviewed or this paperthat almost all modeling or the diagnosis o aults involvesextensive simpli1047297cation ofen reducing potentially complexsystems down to one or two discshafbearing modelsAdapting the claimed results rom such research to systemswith many discsbearingsshafs is important in improvingexisting rotating machinery diagnostic techniques

Te inclusion o nonlinear effects is notable amongstmany authorsmdashas the understanding o the complex aultsand interactions continues to progress Such research opensthe potential or promising new avenues into areas o designor IVHM

Finally it is o note that despite the wide variety o advanced research here a consensus or the most efficienttechniques and algorithms has yet to emerge in order or the1047297nal adaption to industrial application to be achieved How-ever the consideration or industrial application taken by many o the authors detailed in this paper provides promise

or next generation IVHM systems

983091983091 Localisation Fault localization in rotating machinery isan important topic o research or uture condition-basedmonitoring systems Knowing not only what type o aulthas occurred but also where in the system is an importantconsideration which can in1047298uence maintenance proceduresin complex machinery It is worth noting rom the literaturesurveyed or this paper that many studies have ocusedon diagnosis and prognosis o single rotorbearing systemsOfen or legitimate reasonsmdashsimpli1047297cation or computingspeed or example Few studies however have taken intoaccount the localization o aultsacross whole systems Oneo

the more extensive examples highlighted in this paper (Hanet al [983089983093]) provides extensive analytical studies relevant tolocalization on a duel disk setup utilizing Hilbert-Huangransorms however in this example comprehensive valida-tion is lacking thus limiting the potential applicability o theresearch

Tis problem is not limited to modeling and simulation-based research Many newly developed data-driven tech-niques or diagnostics and prognostics claim good results by heavily instrumenting speci1047297c components o a test systemIn many industrial cases this is not possible practical or cost-effective

o give some examples a keyphasor transducer can beparticularly useul in diagnosing aults such as rotor bowHowever this equipment requires the ability to cut a key way or measurements to be perormed Optical sensors haverecently been applied to detect rotor unbalance yet such atechnology would be difficult to implement in a system withseveral rows o rotors (as in a gas turbine)

Whilst new diagnosis and prognosis techniques or aultsin rotating machinery are being continuously researchedthe lack o corresponding studies into localization can beconsidered one o the many challenges in promoting recentcore research into live industrial applications

983091983092 Fault Differentiation One promising development inrecent modeling and simulation o rotordynamic aults isthat o ault differentiation Several researchers have movedon rom studies on individual aults in order to concentrateon combinations o aults As stated rotordynamic aultssuch as those listed in this paper are linked to each otherSuch studies thereore concentrate on such topics as a

misalignment causing an unbalance or a looseness causinga rub Findings rom reports such as these are important inunderstanding complex anomalies In industrial applicationssimply detecting and rectiying an unbalance does not pro-

vide a satisactory solution i the root cause o the ault isa misalignment Tis is a complex topic as several aults canexhibit similar vibration characteristics making traditionaldetection techniques inaccurate in some cases

Making reerence to some o the aorementioned stud-ies those such as Bachschmid et al [983090983091] and Sonnichsen[983090983093] indicate some initial studies including several aultsTe results o such studies indicate a high success rate indifferentiating the aults in the given conditions however

a number o limitations still exist Studies such as theseconsider unbalancein the aultchains Tisis typically ldquostaticrdquoor simple unbalance applied to a single shafrotormdashin reality unbalance aults can be more complex than this In additionthe simple systems used to study and differentiate aultsare very different rom the complexities o a ull turbineHowever urther work may enable such research to ldquoscale-uprdquo to this stage with the aid o urther validation and

veri1047297cation

983091983093 Prognosis It can be seen rom the examples o recentresearch described or common aults that prognostic tech-niques are a topic where much work is being perormed

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 913

International Journal o Rotating Machinery 983097

Some o this work would be very difficult or impossibleto implement in an industrial situation others providepromising results o use to uture work in the area Predictingthe remaining useul lie o components is critical in thedevelopment o condition-based monitoring strategies orindustrial implementation

It can be noted that prognostic studies or certain rotor-dynamic aults are considerably more advanced than orothers Te most obvious examples o the aults detailed inthis paper are bearing aults shaf and blade cracks Teseare also areas where physics-driven simulations have hadan important impact Te ability to design any ault type(or combinations o aults) into such simulations (to be itslow-1047297delity mathematical models or high-1047297delity 1047297nite ele-ment analysis) has provided researchers with many differentavenues to explore Studies into the prognostics o otheraults can be complicated by various actors An unbalanceor example has had relatively little research perormed intoprognosis Te act that unbalance is ofen the cause o another underlying ault is one reason or the difficulty inresearching prognosis in detail or this ault Te results o unbalance are also very dependent on the severity o the aultand the system in which it occurs

It is worth noting that even or the aults where prognosisresearch is more advanced (eg shaf cracks) true develop-ment o the ault is not always studied Fan et al [983089983095] orexample study sets o discrete cracks in order to ldquoprognoserdquocrack growth In act thisstudy is more akinto detecting aultseverity with inerred effects on crack growth In contrast Liet al [983090983090] consider the continual development o such crackshowever the authors do admit this work is in its inancySuch points limit the ability or true prognosis to be achievedwithout extensive operational data

It is also worth noting that combinations o aults havebeen simulated extensively or the purpose o diagnosticshowever ew studies exist combining aultsor the purpose o prognostics Such studies are however a logical progressionrom some o those already perormed It may be that suchprognostic studies build upon some o the condition-basedmaintenance (CBM) or diagnostics High-level studies suchas Jaw and Merrill [983091983093] indicate how prognosis and diagnosistechniques can be combined into a CBM system Tis study is designed or a military aeroengine indicating the desire orsuch systems to be implementedmdashalthough the architecturedescribed has the capability o including various algorithmsoraultprognosis the amount o consideration orprognosis

o ault combinations is unclear

983091983094 Modelling Te large number o works reviewed or thispaper which includes modeling techniques in order to aiddiagnosis and prognosis indicates the possibilities providedby modern computer power and sofware developmentBroadly the modeling studied in the aorementioned papersbreaks down into two categories mathematical modeling and1047297nite element analysis (FEA) Te mathematical models suchasBorghesani etal [983090983089] ofen provide a theoretical basis uponwhich data driven or FEA studies can build and validateMathematical models despite being a traditional approach to

rotordynamic problems thereore continue to be developedto tackle more advanced problems

FEA techniques offer a constantly expanding area o simulation to explore Tey are used or research as diverseas crack propagation to unbalance localization [983091983094] Moderncodes such as NASRAN and Ansys enable rotordynamics to

be studied beyond the traditional stress and modal analysisTe ability to study the effects o rotation (eg throughCampbell diagrams and transient analysis) continues to driveresearch and innovation in this area Te studies discussedin this paper indicate the power o FEA with some highly accurate simulations having been perormed However thereis still a limitation o computer powermdashapplying some o the FEA techniques to whole engine models would proveto be too computationally expensive to be viable As aresult alternate methods such as model order reductionand system level modeling (and model-based reasoning orimplementation) are still required in order to make some o the FEA studies viable in an industrial environment

Nonlinearities continue to orm an important part o many recent studies with the complexity and detail o thespeci1047297ed aults being continually expanded Te depth o modeling has enabled in many cases accurate validationagainst experimental approaches Despite this completeunderstanding o the vibrational phenomena o rotatingmachinery is not currently possible despite the drive towardsthis end

983091983095 Rotordynamics and IVHM Te bulk o current researchinto rotordynamics rom the point o view o prognostichealth management (PHM) can be roughly divided intotwo types initial single-ault diagnosisprognosis techniquesand studies into the general requirements and limitationso current systems along with current and uture trendsAn example o the latter is Pusey [983091983095] who provides agood summary overview o current diagnosis and prognosistechniques with regard to condition-based maintenance

As a result o this split a clear gap exists between thecore research being perormed into rotordynamics rom acondition-based maintenance perspective and the identi1047297edneeds o industry aking a 1047298edgling piece o research andapplying it to a commercially-ready system (eg a gas turbineengine or an aircraf) is a long and complex task It isnevertheless worth noting that technologies or automatically detecting an unbalance or misalignment in a gas turbine were

developed over 983089983088 years beore the latest commercial aircrafwere conceptualized and yet these planes are still limited inthis capacity Tis highlights the need or work which linksthe undamental research into individual ault diagnosis toldquolive systemsrdquo in use in industry

Physics-based simulation and modeling o rotordynamicparts is a well-researched 1047297eld Such modeling has been usedas the basis o diagnosis and prognosis o aults by many researchers several recent examples have been outlined inthis paper Occasional pieces o work have been perormedinto modeling multiple aults such as Jain and Kundra [983091983096]who usea system model oronline identi1047297cation o unbalanceand cracks Beyond this however very limited research

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 1013

983089983088 International Journal o Rotating Machinery

Business casecost bene1047297tanalysis

Live system

RubCracksBearingfailure

Rotor bow Unbalance

Fault differentiationLocalisation

Root cause analysis

Remaining useful life

Data driven analysis and physics-based simulation

Mis-alignment

Experimental

1047297eld data

F983145983143983157983154983141 983089 Physics-based simulationmdashrom research to industry

exists Te demands o PHM techniques in industry are suchthat any systemmust notonly be capable o detecting multipleaults but must also be capable o detecting these aultsacrossa range o different systems Other considerations includethe aorementioned ability to differentiate between multipleaults Processing also needs to be taken into account as theobjective o these systems is to implement efficient conditionmonitoring and condition-based maintenance procedures I processing data is a long power-hungry process then thisaim cannot be achieved

Figure 983089 details a potential ramework required in order

to push core research such as that detailed in this reporttowards industrial applications Many studies now exist onindividual rotordynamic aults across a wide range o condi-tions and applications Some studies have taken this urtherwith advanced prognostic models and diagnosis o dual aults(primary cause and secondary effect) Future research inthe area o rotordynamics rom a PHM perspective couldpotentially provide the bridge between these studies and livesystems by combining physics-based simulations with data-driven techniques and validating against experimental data

Although quantiying the success o the research studiedisdifficultit ispossibleto de1047297ne the key areas in whicha tech-nique must excel in order to be considered viable Te work

byWheeleretal[983091983097] discussesin detailmetrics ordiagnosticand prognostic analysis as does that by Vachtsevanos [983092983088]and Saxena et al [983092983089] Te conclusions drawn rom thesepapers and applied in practice to research like that covered by this paper indicate that the ollowing metrics are importantwhen considering the potential o a given technique to diag-nose aults coverage alse positive rate and alse negativerate In the case o prognosis probabilities and lead timeto ailure are other important considerations Tis criteriaenables research to be assessed in terms o its suitability orindustrial applications Unortunately inormation on these

metrics is not made readily available by the authors o mostpapers

In terms o evaluating the effectivenesso theresearch dis-cussed in this paper there are difficulties in recommending aspeci1047297c technique over others or general application Mostpapers reviewed or this document (not just those reerencedand discussed in detail) take a technique (new or evolved)

validate or a given speci1047297c system and report success o theresearch Despite this some conclusions can be drawn romassessing common techniques applied across different studiesand different aults Although it is not possible to de1047297ne themost common methods or diagnosing andprognosing aultsin terms o numbers (as it was not possible to cover all recent

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 1113

International Journal o Rotating Machinery 983089983089

rotordynamics research or this paper) as perceived by theauthors the ollowing techniques have eatured prominently in the reviewed research

(i) Sensors accelerometers

(ii) Teoretical studies mathematical modeling

(iii) Physics-based simulation FEA(iv) Data-driven joint timerequency domain analysis

(v) Implementation neural networking

Tese techniques appear to be among the most promisingcurrently under development as they tend to eature numer-ous timesamongst some o the work with wider scope acrossault types and with the most comprehensive validationTere are o course many subsections to these techniqueshowever it shows one general direction o research and theclear possibilities posed in these areas

Future developments in the 1047297eld o IVHM or rotat-ing machinery may incorporate these techniques alongsideextensive use o nonlinear modeling and multiple aultinteractions Te 1047297eld o design or IVHM has only recently emerged however the potential exists or speci1047297c nonlinear-ities to be designed into a system in order to enable accuratediagnosis andprognosis o aults Te development o currentalgorithms to include diagnosis localization and prognosiso a range o aults will provide a signi1047297cant advancement oruture generations o IVHM systems Tis combined withcost-effective sensor suites indicates the potential or evolu-tions o some o the research detailed hereto orm part onextgeneration IVHM suites or rotating machinery

4 Conclusion

Tis paper has reviewed some o the latest research around anumber o rotordynamic aults namely unbalance misalign-ment rub and looseness 1047298uid-induced instability bearingaults shaf cracks blade cracks and rotor bow Each aultwas reviewed rom the perspective o sensors diagnosisprognosis localization and modeling

Key examples o recent work into the eight describedaults have been detailed through works by a number o eminent authors Additional work has been summarized andormatted or easy reerence Some current trends amongstthe recent body o work include developments in the vastarea o modeling nonlinearities combinations o high andlow 1047297delity modeling and synergy between data driven and

physics-based simulation approachesDespite the large volume o promising research reviewed

urther development in a number o areas is required in orderto produce effective next generation IVHM systems As suchuture developments may include usion o andor devel-opment o current algorithms to encompass all eight aultsdetailed consideration o prognosis diagnosis and localiza-tion achieved using a reduced cost-effective sensor suite

References

[983089] I K Jennions Integrated Vehicle Health Management Perspec-tives on an Emerging Field 983090983088983089983089

[983090] G Anderson ldquoProviding best-value IVHM solutions or agingaircrafrdquo in Proceedings o the 983097th Joint FAADoDNASA Aging Aircraf Conerence vol 983089 pp 983089ndash983089983089 983090983088983088983093

[983091] A Muszynska Rotordynamics aylor amp Francis 983089st ed edition983090983088983088983093

[983092] D E Bently Fundimentals o Rotating Machinery Diagnostics

ASME Press 983089st edition 983090983088983088983090[983093] B Domes ldquoVibration phenomena in aero-enginesrdquo in Proceed-ings o the 983097th International Conerence on Vibrations in Rotating Machinery Exeter (IMechE rsquo983088983096) vol 983089 pp 983089983093ndash983091983090 September983090983088983088983096

[983094] A Hess G Calvello and Dabney ldquoPHM a keyenabler or theJSF autonomie logistics support conceptrdquo in Proceedings o theIEEE Aerospace Conerence Proceedings pp 983091983093983092983091ndash983092983093983092983097 March983090983088983088983092

[983095] S N Ganeriwala B Schwarz and M H Richardson ldquoOperat-ing de1047298ection shapes detect unbalance in rotating equipmentrdquoSound and Vibration vol 983092983091 no 983093 pp 983089983089ndash983089983091 983090983088983088983097

[983096] G N D S Sudhakar and A S Sekhar ldquoIdenti1047297cation o unbal-ance in a rotor bearing systemrdquo Journal o Sound and Vibration

vol 983091983091983088 no 983089983088 pp 983090983090983097983097ndash983090983091983089983091 983090983088983089983089[983097] Yang and M W Hsu ldquoAn efficient diagnosis technique or

variations o shaf-bow and unbalancerdquo in Proceedings o the ASME International Design Engineering echnical Conerencesand Computers and Inormation in Engineering Conerence(DEC rsquo983088983097) pp 983093983095ndash983094983094 San Diego Cali USA September 983090983088983088983097

[983089983088] A El-Shaei S H aw1047297ck and M O A Mokhtar ldquoExperi-mental investigation o the effect o angular misalignemt onthe instability o plain journal bearingsrdquo in Proceedings o the ASMESLE InternationalJointribology Conerence(IJC rsquo983088983097)pp 983089983095983089ndash983089983095983091 Memphis enn USA October 983090983088983088983097

[983089983089] A El-Shaei S H aw1047297ck and M O A Mokhtar ldquoMisalign-ment modeling in rotating systemsrdquo in Proceedings o the ASMEurbo Expo H Bahaloo A Ebrahimi and M Samadi Eds vol

983089983095983089983089 pp 983097983095983091ndash983097983095983097 Orlando Fla USA June 983090983088983088983097

[983089983090] L F Villaet al ldquoStatistical diagnosisbased on vibration analysisor gear test-bench under non-stationary conditions o speedand loadrdquo Mechanical Systems and Signal Processing vol 983090983097 pp983092983091983094ndash983092983092983094 983090983088983089983090

[983089983091] C F Ngolah et al ldquoIntelligent ault recognition and diagnosisor rotating machines using neural networksrdquo Sofware Scienceand Computational Intelligence vol 983091 no 983092 983089983095 pages 983090983088983089983089

[983089983092] Y Lu Z Ren H Chen N Song and B Wen ldquoStudy on loose-ness and impactmdashrub coupling aults o a vertical dual-disk cantilever rotor- Bearing systemrdquo Key Engineering Materials vol 983091983093983091ndash983091983093983096 no 983092 pp 983090983092983095983097ndash983090983092983096983090 983090983088983088983095

[983089983093] Q Han Z Zhang and B Wen ldquoPeriodic motions o a dual-

disc rotor system with rub-impact at 1047297xed limiterrdquo Proceedingso the Institution o Mechanical Engineers vol 983090983090983090 no 983089983088 pp983089983097983091983093ndash983089983097983092983094 983090983088983088983096

[983089983094] H F de Castro K L Cavalca and R Nordmann ldquoWhirland whip instabilities in rotor-bearing system considering anonlinear orce modelrdquo Journal o Sound and Vibration vol 983091983089983095no 983089-983090 pp 983090983095983091ndash983090983097983091 983090983088983088983096

[983089983095] C C Fan et al ldquoMechanical systems and signal processingrdquoStudy o Start-Up Vibration Response or Oil Whirl and Dry Whip vol 983090983093 pp 983091983089983088983090ndash983091983089983089983093 983090983088983089983089

[983089983096] K Kappaganthu and C Nataraj ldquoNonlinear modeling andanalysis o a rolling element bearing with a clearancerdquo Commu-nications in Nonlinear Science and Numerical Simulation vol983089983094no 983089983088 pp 983092983089983091983092ndash983092983089983092983093 983090983088983089983089

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 1213

983089983090 International Journal o Rotating Machinery

[983089983097] C Gupta K Gupta and D K Sehgal ldquoInstability and chaoso a 1047298exible rotor ball bearing system an investigation on thein1047298uence o rotating imbalance and bearing clearancerdquo Journal o Engineering or Gas urbines and Power vol 983089983091983091 no 983096ArticleID 983088983096983090983093983088983089 983089983089 pages 983090983088983089983089

[983090983088] J Hong X Miao L Han and Y Ma ldquoPrognostics model orpredicting aero-engine bearing grade-lierdquo in Proceedings o the ASME urbo Expo vol 983089 pp 983094983091983097ndash983094983092983095 Orlando Fla USA June983090983088983088983097

[983090983089] P Borghesani P Pennacchi R B Randall N Sawalhi and RRicci ldquoApplication o cepstrum pre-whitening or the diagnosiso bearing aults under variable speed conditionsrdquo Mechanical Systems and Signal Processing 983090983088983089983090

[983090983090] Y Li J Zhang L Dai Z Zhang and J Liu ldquoAuditory-model-based eature extraction method or mechanical aults diagno-sisrdquo ChineseJournal o MechanicalEngineering (EnglishEdition) vol 983090983091 no 983091 pp 983091983097983089ndash983091983097983095 983090983088983089983088

[983090983091] N Bachschmid P Pennacchi and E anzi ldquoCracked rotatingshafs typical behaviors modeling and diagnosisrdquo IUAM Symposium on Emerging rends in Rotor Dynamics vol 983089983088983089983089 pp

983092983092983089ndash983092983093983092 983090983088983089983089[983090983092] Inoue N Nagata and Y Ishida ldquoFem modelling and

experimental veri1047297cation o a rotor system with a open crackrdquoin Proceedings o the ASME International Design Engineering echnical Conerences and Computers and Inormation in Engi-neering Conerence (DEC rsquo983088983097) pp 983089983089983089983091ndash983089983089983090983090 San Diego CaliUSA September 983090983088983088983097

[983090983093] H E Sonnichsen ldquoReal-time detection o developing cracksin jet engine rotorsrdquo in Proceedings o the IEEE AerospaceConerence pp 983089983095983091ndash983089983096983092 March 983090983088983088983088

[983090983094] I Green and C Casey ldquoCrack detection in a rotor dynamicsystem by vibration monitoringmdashpart I analysisrdquo Journal o Engineering or Gas urbines and Power vol 983089983090983095 no 983090 pp 983092983090983093ndash983092983091983094 983090983088983088983093

[983090983095] J Sawicki M I Friswell A H Pesch and A WroblewskildquoCondition monitoring o rotor usingactive magnetic actuatorrdquoin Proceedings o the ASME urbo Expo pp 983089983090983093983095ndash983089983090983094983093 BerlinGermany June 983090983088983088983096

[983090983096] J Xiang X Chen Q Mo and Z He ldquoIdenti1047297cation o crack ina rotor system based on wavelet 1047297nite element methodrdquo FiniteElements in Analysis and Design vol 983092983091 no 983089983092 pp 983089983088983094983096ndash983089983088983096983089983090983088983088983095

[983090983097] M G MaalouldquoSlow speed vibrationsignal analysisi you canrsquotdo it slow you canrsquot do it astrdquo in Proceedings o the ASME urboExpo pp 983093983093983097ndash983093983094983095 Montreal Canada May 983090983088983088983095

[983091983088] J Meagher X Wu and C Lencioni ldquoResponse o a warped1047298exible rotor with a 1047298uid bearingrdquo International Journal o

Rotating Machinery vol 983090983088983088983096 Article ID 983089983092983095983094983093983091 983097 pages 983090983088983088983096

[983091983089] Gaka andM abaszewski ldquoAn applicationo statistical symp-toms in machine condition diagnosticsrdquo Mechanical Systemsand Signal Processing vol 983090983093 no 983089 pp 983090983093983091ndash983090983094983093 983090983088983089983089

[983091983090] J K Sinha ldquoRecent trands in ault quanti1047297cation in rotatingmachinesrdquo Advances in Vibration EngIneerIng vol 983096 no 983089 pp983095983097ndash983096983093 983090983088983088983097

[983091983091] X Shen JJia andM Zhao ldquoNonlinear analysis o a rub-impactrotor-bearing system with initial permanent rotor bowrdquo Archiveo Applied Mechanics vol 983095983096 no 983091 pp 983090983090983093ndash983090983092983088 983090983088983088983096

[983091983092] A W Lees J K Sinha and M I Friswell ldquoModel-based identi-1047297cation o rotating machinesrdquo Mechanical Systems and Signal Processing vol 983090983091 no 983094 pp 983089983096983096983092ndash983089983096983097983091 983090983088983088983097

[983091983093] L C Jaw and W Merrill ldquoCBM+ research environmentmdashacilitating technology development experimentation andmaturationrdquo in Proceedings o the IEEE Aerospace Conerence(AC rsquo983088983096) March 983090983088983088983096

[983091983094] R Walker S Perinpanayagam and I K Jennions ldquoSimulatingUnbalance or Future IVHM Applicationsrdquo in Proceedings o theSociety or Experimental Mechanics (IMAC rsquo983089983090) vol 983090983097 pp 983089983092983089ndash983089983092983096 May 983090983088983089983090

[983091983095] H C Pusey ldquourbomachinery condition monitoring and ail-ure prognosisrdquo Sound and Vibration vol 983092983089 no 983091 pp 983089983088ndash983089983093983090983088983088983095

[983091983096] J R Jain and K Kundra ldquoModel based online diagnosiso unbalance and transverse atigue crack in rotor systemsrdquo Mechanics Research Communications vol 983091983089 no 983093 pp 983093983093983095ndash983093983094983096983090983088983088983092

[983091983097] K R Wheeler Kurtoglu and S D Poll ldquoA survey o healthmanagement user objectives related to diagnostic and prognos-tic metricsrdquo in Proceedings o the 983090983097th Computers and Inor-mation Engineering Conerence vol 983090 pp 983089983090983096983095ndash983089983090983097983096 983090983088983088983097

[983092983088] G Vachtsevanos ldquoPerormance metrics or ault prognosis o

complex systemsrdquo in IEEE Systems Readiness echnology Con- erence (AUOESCON rsquo983088983091) pp 983091983092983089ndash983091983092983093 September 983090983088983088983091

[983092983089] A Saxena J Celaya E Balaban et al ldquoMetrics or evaluatingperormance o prognostic techniquesrdquo in Proceedings o theInternational Conerence on Prognostics and Health Manage-ment (PHM rsquo983088983096) pp 983089ndash983089983095 October 983090983088983088983096

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 1313

Submit your manuscripts at

httpwwwhindawicom

Page 2: Rotordynamic Faults 2013

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 213

983090 International Journal o Rotating Machinery

Trough the study o the topics listed it is useul to placethe research conducted in this paper into context with regardto real-world applications Further to this it is intended toidentiy potential areas where more research is required inorder to push some o the recent technologies highlighted orthis study into industry

Anderson [983090] provides a summary o maintenance timebreakdown or a collection o military aircraf Tis indicatesthat as much as 983092983092 o on-aircraf maintenance time (whichin turn accounts or 983097983088 o total maintenance operations) isconsumed with inspection alone Te techniques addressedin this paper enable the maintenance to be more inormedand targeted with inventory ready when needed providinga signi1047297cant contribution to reducing maintenance time andcost

As the topic o rotordynamic aults is very large areao research the scope o this paper has to be limited Techoice o aults has been made afer considering the worksby Muszynska [983091] and Bently [983092] both o whom consider theundamentals o common aults in much detail Out o thewide range o possible rotordynamic aults eight have there-ore been selected Due to the general reliability o the currentgeneration o gas turbines aults alling outside o the eightlisted have been classed as ldquouncommonrdquo or the purposes o this study Tis decision was made by assessing the severity o each ault dependences on other aults and the level o research dedicated to diagnosis o each ault

Te scope o this paper is thereore con1047297ned to the ol-lowing aults

(i) unbalance

(ii) misalignment

(iii) rub and looseness

(iv) 1047298uid-induced instability

(v) bearing aults

(vi) shaf cracks

(vii) blade cracks

(viii) rotor bow

Each o these aults is varied and some are more commonthan others Te consensus rom the works reviewed withinthis paper is that unbalance is the most commonly occur-ring in part due to the 1047297ne tolerances applied to modernmachinery and also due to the links between unbalance andother aults Misalignment can also be identi1047297ed as another

common ault Te particular importance o these two aultsis highlighted by Domes [983093] who discusses common aultsrom the perspective o Rolls RoyceBearing ailure is perhapsthe most diverse ault here and has the highest concentrationo research in various areas Te decision has also been madeto separate aults occurring on the shafs o a system to thoseoccurring on rotors as these can be classi1047297ed as distinctaultsmdashparticularly with regards to localization and mainte-nance Despite these conclusions quantiying the aults withregards to the rate o occurrence was not possible due to thelack o commercially available data and each o the eightaults hasthereore beeninvestigatedto an equal level o detailin this paper

It should be noted that although the aults have beenseparated into eight categories they are by no means mutually exclusive Dependencies exist between many o the aults Acommon example o this is the interaction between unbal-ance and misalignment (outlined by Bently [983092]) Fault chainscan become even more complex misalignments leading to an

unbalance which causes a rub or an example Some workscontained herein detail single ault analysis and a ew reerto two or more aults Such ault dependencies are one o thelimiting actors when moving technologies rom the lab intoindustry An example o this would be a system capable o diagnosing unbalance aults without consideration or mis-alignment as a root cause may render such a system ineffec-tive when applied to a real-world scenario

Te selected papers have been urther re1047297ned based uponrelevance to aeroengine gas turbines and publication date(with works rom the last three years preerred)

As research into the diagnosis and prognosis o aults inrotating machines is a developing topic the application o such technologies in industry has not yet reached a stagewhere there are common solutions or even a set o estab-lished procedures to ollow Tis holds particularly true withregards to the broader 1047297elds o IVHM aking the state-o-the-art Joint Strike Fighter (JSF) program as one example o the current level o technology in circulation although theautonomic logistics proposed by Hess et al [983094] or the vehicleis undeniably impressive it is still a long way rom imple-menting some o the recent technologies claimed to besuccessul in a lab-based environment

Although as mentioned an established and commonly applied set o procedures and standards or IVHM has notyet emerged with regards to rotating machinery a numbero standards and procedures can be reerred to with regardsto detailing and quantiying the aults Tese include suchaspects as covered by ISO 983090983097983093983091983089983097983097983097 which details correctbalancing procedures and levels o mechanical vibrationAnother example would be in the UK Ministry o DeencemdashMilitary Aviation Authority (983090983088983089983088) JAP 983089983088983088A-983088983089mdashMilitary aviation engineering policy and regulation where a mentionis made o procedures or debris and vibration monitoringTis indicates a gradual move towards common ground orprocedures and standards or IVHM systems Despite thisresearch remains varied and separating the research withpotential or moving beyond the lab-based environment intoindustry can be difficult to identiy at 1047297rst glance

Te discussion contained herein thereore takes into

account the above detailed considerations and constraintswith the aims o the paper as ollows

(i) Highlight key examples o the latest research in eightcommon rotordynamic aults

(ii) Summarise current trends identi1047297ed rom the study

(iii) Detail the current state o the art and the uturework required or next generation IVHM systems orrotating machinery

2 Rotordynamic Faults

Te ollowing section details the aorementioned commonrotordynamic aults with regard to recent physics-based

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 313

International Journal o Rotating Machinery 983091

simulation work and traditional data-driven methodsResearch in this area o rotordynamics is particularly wide-spread and so this paper outlines only a ew recent areaso research Although not being o mainstream interest alist o commonly used sensors has been included able 983089represents a summary o the inormation collected and

detailed as a part o this paperAs discussed a wide selection o papers have beenreviewed or consideration in this paper o which it is notpossible to describe them all in detail In an attempt tohighlight some o the common themes across many o worksreviewed able 983089 has been created able 983089 by no meansdetails all relevant methods and approaches currently beinginvestigated however it details themes which were oundto be common across many areas o current research It isintended or the detailed table to compliment the importantworks described in more detail in the ollowing section o thepaper separated by ault type

983090983089 Unbalance Tis is one o the most common rotordy-namic aults [983092] every rotating machine has an inherent

degree o unbalance Unbalance as a ault can thereorebe de1047297ned as unbalance outside o a given tolerance levelA recent piece o research which demonstrates the ongoingdevelopment o data-driven techniques is that by Ganeriwalaet al [983095] who tested a technique or measuring operatingde1047298ection shapes (ODS) in order to detect unbalance casesTese studies were conducted on a machine ault simulator (aphysical simulator o common aults in rotating machines)mdashsuch simulators have the advantage o recreating aulty con-ditions quickly and easily enabling a new dimension to data-driven diagnostic techniques It is however worth mention-ing that data or these experiments were collected using 983089983092accelerometers which are easy to apply to such a simulatorbut it may be much more difficult to con1047297gure these many sensors on a complex system Despite this the paper achieveditsaim in proving thehypothesis ldquowhen an operating machinebecomes unbalanced its ODS will changerdquo

Regarding physics-based simulation o unbalance as aault the work by Sudhakar and Sekhar [983096] described amodel-based method or ault identi1047297cation using a minimalsensor suite Tis is achieved through the analysis o trans-

verse vibrations at a single location Troughout the paperthree different approaches are studiedmdashleast squares mini-mization equivalent loads minimization and vibrationmini-mization A reduced error is ound to occur using a proposed

modi1047297cation to the typical equivalent loads minimizationapproach Te work is o note due to the authors requirementto both identiy and locate (localize) unbalance

Localization and prognosis o unbalance pose a uniqueseto challenges andas a result research is still somewhat lim-ited in these areas One recent work which claims to localiseunbalance accurately is by Yang and Hsu [983097] the authorsuse trending data and reasoning systems to locate unbalanceand shaf bow across a system Quick diagnosis is achievedby avoiding the study o all ault combinations and theauthors claim the ability to localise shaf bow and unbalancein large rotating machines Te techniques rely extensively onprevious searches and are limited to large stable machines

operating at a speci1047297c RPM Remaining useul lie o unbal-ance is difficult to predict due to complicating actors Asdiscussed the ault dependencies are such that in many casesthe unbalance itsel will not be the ailing actor instead anunbalance may cause a rub which then leads to damagewhichcan be prognosesSuch combinations o aults and underlying

causes or unbalance lead to the need or remaining useullie predictions to be made based upon the exact nature o aspeci1047297c ault

983090983090 Misalignment Tis is another common ault which canpotentially in1047298ict considerable damage in rotating machinesAs with unbalance misalignment in a whole system can becomplicated by secondary aults (eg a misalignment whichcauses a rub) El-Shaei et al [983089983088] are an example o theongoing research in this area in this case a unique combina-tion o angular misalignment and oil whipwhirl is detailedTe authors describe how small degrees o misalignment canbe utilized in order to prevent the onset o 1047298uid-induced

instabilities tested through the use o a lab-based test rigSuch research presents a new dimension in looking atcommon rotordynamic aults with aspects that could beapplied to uture design or IVHM systems

El-Shaei et al [983089983089] demonstrate interesting research intomisalignment rom the perspective o physics-based simula-tion Te authors construct mathematical models o a simplerotorsystem witha misaligned coupling andcollect harmonicresponse data rom this to assess the severity o differentmisalignment cases Such models are useul throughout thelie o rotating machinesmdashrom design to implementationalthough again successul validation with experimentally obtained data is key Te authors highlight the act thatalthough misalignment is a prevalent and serious ault nocomprehensive research has been perormed or treating thisproblem Te methodology applied includes deriving theenergy expressions applying the Ritz series method con-structing the equations o motion and then using the har-monic balance method to look or multiharmonic responsesTe paper demonstrates the ongoing research to understandand model the undamentals o such aults in order thatimproveddiagnosisand prognosis methods can be built uponsuch knowledge

As with unbalance localization and prognosis o mis-alignment is a complex topic to research Studies such as El-Shaei et al [983089983089] can make accurate predictions or misalign-ment in a simple system with one coupling However real

systems (eg aircraf gas turbines) have many potential loca-tions o misalignment Tis is an area where ew researchershave made an impact Remaining useul lie predictions ormisalignment is complicated or the same reasons as withunbalance Villa et al [983089983090] discuss statistical diagnosis o misalignment aults with reerence to prognosis Te authorsuse the example o a wind turbine or their studies butstress the applicability to other systems Differentiation withunbalance aultsis also covered (as these two aultsare closely linked) Unlike the work by El-Shaei et al [983089983089] emphasis isgiven to the machine in question operating over a wide rangeo operating speeds and conditions Tis is achieved throughthe use o an angular resampling method Prognosisis tackled

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 413

983092 International Journal o Rotating Machinery

983137983138983148983141 983089 Rotordynamic ldquoault tablerdquo

Faults Sensors Fault identi1047297cation

techniques Localisation o ault Prognosis

Work o note in aultmodelling

UnbalanceAccelerometer

Velocity transducer

Proximity transducer

Operating de1047298ectionshapesime-requency

analysis and many others

Neural-network modelling and

time-requency analysis

FCM-Markov model

Unbalance in ull-engineFEA modelEarly detection inaeroengines

High radialloads andmisalignment

AccelerometerTermocoupleDisplacement

transducerAngular sensor

Bearing temperatureobservation advancedSofware-driven1047297ltering methods

Various vibrationanalysis techniques

Prognosticmodelling usingsimulated vibrationdata

Modelling therelationship betweenmisalignment and otheraultsHarmonic balancemethod

Rub andlooseness

AccelerometerVelocity transducer

Optical sensor

Single valuedecompositionime-requency analysis

Least-squares 1047297ttingapproach

Adaptive waveletdenoising orprognostics

Rob and looseness usingFEA models Wholeengine modellingtechniques

Fluid-inducedinstability

Proximity probesKeyphase transducer

Full spectrum plotsLarge amplitudesubsynchronous vibrationDetection

Blade tip clearanceand loadingdistribution

Cross-coupledstiffness anddirect-dampingstudies

Nonlinear modellingaccuracy o model basedDiagnostics numericalmodelling

Bearing ailure

AccelerometerVelocity transducer

Displacementtransducer

Proximity probes

Postailure analysis oruture ailure diagnosisDynamictime-requency analysis

Statistical change inthe bispectraldomain

Numericalmodelling o aeroengine bearingsModel-basedprediction o remaining useul lieOn-boardprognostics

High and low 1047297delity dynamic systemmodellingComparisons o algorithms tounderstand bearing vibration phenomena

Shaf cracks

Keyphase transducer

Proximity probes

Higher order spectra

Neural networksGrid support vectors

Forced response

measurements

Nonintrusive

torsion vibrationmonitoring

Statistical analysis FEA

modelling

Blade cracks Keyphase transducer

Proximity probes

Hilbert-HuangtransormationModel-baseddiagnosticsMorletwaveletquantumtechniques

Novel bearingstructures

Active magneticactuator

Global and localasymmetry modellingFinite element modelling

Rotor Bow Keyphase transducer

Proximity probesSeismic transducer

Examination o ldquoslow rollrdquo dataModel-baseddiagnostics

Partial mathematicalmodelling

Statistical symptomsanalysis

Advantages o aultmodelling oridenti1047297cationNonlinear modellingtechniques

through the use o a statistical diagnosis algorithm based onthe signi1047297cance level o the aults in question

983090983091 Rub and Looseness Rub is always a secondary ault (iea product o another ault such as looseness) and can leadto atigue and wear Rub and looseness can create com-plex vibration signals which are difficult to diagnose usingtraditional methods Modeling and simulation o rub andlooseness aults have been considered in several recentworks

Tis includes Ngolah et al [983089983091] which detail the moni-toring and diagnosis o common aults (including rub and

looseness) based upon a three-layer Arti1047297cial Neural Net-work (ANN) A series o 983089983088 key perormance indicators wereidenti1047297ed and used as training Te authors test the systemin a lab environment but stress the applicability to industrialapplications Te research indicates one o the latest methodso research which enables the implementation o diagnosistechniques It is a useul tool or rub and looseness studiesas it incorporates a variety o aults which could ldquounderlierdquosuch a ault Despite this the research relies on clear eaturesor each ault which can be much easier to identiy in a labenvironment as opposed to ldquonoisyrdquo industrial applications

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 513

International Journal o Rotating Machinery 983093

Lu et al [983089983092] have perormed several studies into rub andloosenessincluding thisexamplein which a 1047298exibly mountedshaf has an induced rub due to a range o contact ringsTe study ocusses on the potentially dangerous effects o rubin causing excessive nonsynchronous and chaotic vibrationsTe links with unbalance and misalignment are discussed

and detailed Te considerations or real-world cases in thedescribed research are considerable as part o the drivetowards ull understanding o the nonlinear effects o rub andlooseness

Localization o rub and looseness across whole systemsis relatively lightly studied in the literature Many works(including those already cited) look at single or dual-rotorsystems where localization o such aults is not an issue Inan industrial setting complex systems may comprise many rotors in several compressor and turbine stages signi1047297cantly complicating diagnosis o such aults Research into progno-sis o rotor-stator rubs lies mostly within the domain o data-driven techniques Modeling and simulation research can beused to support data-driven techniques or prognosis andcondition-based monitoring Han et al [983089983093] is an example o this the authors use 1047297nite element modeling to construct adual rotor model Various types o rub-impact are then stud-ied Such studies can provide a wide range o inormationwhich can then be combined with data obtained rom livesystems potentially with seeded aults in order to constructaccurate remaining useul lie predictions As pointed out by the authors one key advantage o simulation is the ability to study more complex systems with a higher number o rotors which is used throughout this research Tis presentsa different approach to identiying eatures or identi1047297cationo rub and looseness

983090983092 Fluid-Induced Instability Fluid-induced instabilities(ofen reerred to as whip and whirl) are potentially very serious aults which can result in wear atigue and extensivedamage to machine components Such instabilities can beound in interstage seals 1047298uid lubricated bearings andblade-tip clearances Research into simulating and modeling1047298uid-induced instability has produced several works o interest to ault diagnosis o rotating machines in the lastew years De Castro et al [983089983094] are a good example wherenonlinear mathematical models are prepared or a rotor-bearing system Te models are then used to predictinstability thresholds Te authors consider a test case againsta power plant turbine and a test rig thereore validating the

simulations Te case o unbalance aults causing whip andwhirl phenomena is also considered Te main conclusionthereore drawn rom the work is that the authors concludedthat nonlinear hydrodynamic journal bearing models enablesufficiently accurate simulations or predicting instability thresholds Fan et al [983089983095] represent an example o work romthe perspective o aeroengine turbines In this case startupconditions are studied using a ull Hilbert spectrum Te aimbehind the paper is to accurately predict the point at whichwhip and whirl occur thus enabling this to be avoided at thedesign stage Such 1047297ndings could potentially also be used toidentiy whip and whirl as the case o a ault afer a period o wear in the operating machine

Prognosing 1047298uid-induced instability is a relatively lightly researched topic Fluid instabilities can be covered as part o extensive research into remaining useul lie o bearings Tepotential exists or modeling and simulation techniques suchas those detailed above to become a part o prognosis or1047298uid-induced instabilities due to the act that it can be very

difficult to seed such aults into live systems or testing andevaluation As with other aults detailed in this paper many studies have been perormed with the aim o describing 1047298uid-induced instabilities based on the measurement or simulationo single (or occasionally dual) rotor setups Physics-basedsimulation with the aim o localising 1047298uid instability aultsacross a whole system can be limited by the complexity o both the ault and the system hence the simpli1047297cation tosingle rotor-stator bearing systems

983090983093 Bearing Failure An area where data-driven techniquesare still providing the basis o much research in the 1047297eld o rotordynamics is that o bearing ailure Te title ldquobearingailurerdquo can cover a wide range o potential issues whichcontinue to be studied in detail Faults can occur in all kindso engine bearingsmdashthe inner andouter case the cage andtherolling elements 1047298uid-induced instabilities (addressed in aseparate section) lubrication and the complexities o activemagnetic bearings to name some examples All types o bearing relevant to rotating machinery are the subject o ongoing research and this subject has the potential to ormseveral separate papers As a brie highlight some recentexamples are discussed as ollows

Data-driven techniques have enabled accurate bearingdiagnostics and prognostics to be described or a rangeo rotordynamic systems Despite the prevalence o data-driven research in this area research rom a physics-basedsimulation perspective has also recently produced someinteresting papers o relevance to condition monitoring andhealth management o rotating machinery

Tis includes Kappaganthu and Nataraj [983089983096] in whichrolling element bearings have been studied through the useo nonlinear models Te included nonlinearity in this caseis clearance and the model is then used in order to study chaotic motions in particular the regions o chaotic responseTe research orms a part o an ongoing drive to develop anaccurate model-based diagnostic technique or rolling ele-ment bearings taking into account clearance nonlinearitiesand chaotic responses

Guptaetal[983089983097] demonstrate another example o thelatestresearch into instability and chaos in rolling element bearingsthrough high-1047297delity simulations Tis detailed and complexstudy involves the application o a novel scheme to analyzethe quasiperiodic response o the system combined with anonautonomous ldquoshootingrdquo method Tis workhighlights thelevel o detail to which nonlinearities and complex nonlinearmotions in bearings are beginning to be understood andaccurately modeled Again such work presents the potentialor design o IVHM in uture evolutions o the research

As so much research hasbeen perormed (and is ongoing)into bearing aults across a wide variety o mechanicalsystems both prognostics and localization o bearing aults

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 613

983094 International Journal o Rotating Machinery

have been researched in somewhat more detail than someo the other aults detailed here Despite this much work still needs to be perormed in order to translate some o this core research into industrial applications Research suchas that detailed above has made signi1047297cant advances intodetermining bearing ailure as the root cause o malunction

Detecting which bearing is ailing across a complex systemhas received somewhat less research Bearing prognostics isanother area with much ongoing research being perormedmdashboth in the simulation and data-driven domains o givean example Hong et al [983090983088] combined grade lie andextensive mathematical modeling techniques in order to pro-duce prognostic models or aeroengine bearings Te resultsare described by the authors as ldquopractical and veri1047297ablerdquoAlthough a number o similar recent studies existthis work iso note or the extent o the studies perormed which includebearing test stand run-to-ailure validation Te lab resultsappear impressive this research has yet to be applied andtested in real lie applicationsmdashindicating that despite thenumber o parameters considered it is still not possible tomodel naturally occurring phenomena sufficiently

A large body o work in this area also exists romBorghesani et al [983090983089] where several examples o data driventechniques can be seen applied to a wide variety o bearingtypes An example o recent developments involves the use o ldquocepstrum prewhiteningrdquo in order to remove sufficient noiseor accurate bearing diagnosis and prognosis Tis work isparticularly noteworthy due to the emphasis on real-worldapplications where the traditional lab-based techniques o order tracking and synchronous averaging do not providesufficient noise removal or harsh industrial environmentsTe addition o such techniques is a crucial step in developingthe current generation o diagnosis and prognosis algorithmsor use in uture IVHM systems

983090983094 Shaf Cracks Another potentially serious ault in rotat-ing machinery is shaf cracks and so early detection o any such ault is highly important Methods o crack ormationandpropagation canbe diverse andrange rom highand low-cycle atigue to stress corrosion Simulation and modelingo shaf cracks can have signi1047297cant advantages over data-driven methods Perhaps the most obvious advantage is therelative simplicity o inserting a ault into orexample a 1047297niteelement model as opposed to seeding a ault in a workingindustrial machine As such research into shaf cracks has

been progressing steadily with the corresponding increasesin computing power

A clear synergy between data-driven and physics-basedsimulation research can be implied by a number o recentworks o research An example o recent advances rom a datacollection perspective is Li et al [983090983090] which details statisticalmodels based on historical data or condition monitoringpurposes Tis unique work uses the human auditory systemas inspiration or enriching methods o mechanical aults andeatures extraction Te results indicated by the paper areperhaps not as convincing as some other methods discussedin this paper it describes an interesting ldquooutside the boxrdquomethod o tackling common problems

From a modeling perspective Bachschmid et al [983090983091]cover a wide range o vibration phenomena in order todevelop a model-based identi1047297cation and severity procedureTis work is noted or its thoroughness in modeling pro-cedure including accurate modeling o the crack breathingmechanism A combination o high and low 1047297delity models

is validated through experimental study and ldquoexcellentrdquoaccuracy is claimed by the authors in detecting crack positionand depth through the use o the proposed model-baseddiagnostics

Te nature o shaf cracks has resulted in a wide variety o research beingperormed intoboth localization and prognos-tics o these aults (indeed the two topics can be consideredrelated) Recent examples o work in this area include Inoueetal[983090983092] who detail crack localization using orced responsemodeling In this case a test rig was constructed consistingo a circular shaf supported by two bearings Frequency-domain data was used to create a localization algorithmdesigned in combination with an FE model Although thisresearch provides good accuracy o localization in a labenvironment it remains untested in a more complex system(eg a ull gas turbine)

Sonnichsen [983090983093] describes 1047297nite element modeling o crack propagation with validation against experimentalresults provided to demonstrate the validity o such modelingtechniques Te paper concentrates on natural requenciesand resonance curves Whilst the authors claim improve-ments in the ability to understand such aults again thesystem in question is quite simplemdashand such FEA models aredifficult to scale up to ull size applications

983090983095 Blade Cracks Blade cracks i allowed to develop canresult in serious consequences Cracks can orm due to highcentriugal stresses across operational cycles (in the case o an aircraf gas turbine or example start up and take off through landing and taxi) As excessive crack growth canlead to catastrophic rotorblade ailure early detection andprognosis o such aults are essential As with shaf cracksphysics-driven simulation o blade cracks is an area o sig-ni1047297cant research Tis varies rom high-1047297delity 1047297nite elementmodels to low-1047297delity system and mathematical models Terecent work demonstrated by Green and Casey [983090983094]is agoodexample o recent mathematical modeling rom a diagnosisperspective In this paper the authors concentrate on early detection using global and local asymmetry crack models 983090X

harmonic components are identi1047297ed as key areas or the early detection o blade cracks however again this paper suffersrom being applied and tested on a relatively simple systemwhich may not scale up to a ull size turbine

Sonnichsen [983090983093] demonstrated high-1047297delity modelingthe authors used FEA to model crack growth making com-parisons and validating against an experimental rig Tiswork is particularly interesting as it outlines the advantagesand drawbacks with the latest state-o-the-art modelingtechniques

Localization and prognosis o blade cracks have alsobene1047297tted rom recent advances in simulation and modelingSawicki et al [983090983095] contains details o work on a novel active

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 713

International Journal o Rotating Machinery 983095

magnetic bearing system or use in the early detection local-ization and prognosis o blade cracks Again the emphasis ison early detection with the bearings used to excite the systemin order to obtain optimum response vibrations or analysisTe authors admit the approach hassome merit in diagnosingblade cracks however it is in the early stages o development

and work are ongoing FEA has also been used extensively to support blade crack prognostic tools Xiang et al [983090983096] arean extensive example o recent work In this case a numbero advanced FEA methods are applied to produce accurateFEA solutionsmdashthese include surace-1047297tting techniques andthe contour-plotting method Te authors experimentally

validate their work and suggest that it can be applied toprognosis and quantitative diagnosis o blade cracks Despitethe claimed advances again the scalability o such research toull size turbines is an issuemdashparticularly with regard to thecomplexity o the FEA models

983090983096 Rotor Bow Rotor bows can be a primary source o unwanted vibration in gas turbines Te main cause o arotor bow (rotor bows do not include bows due to gravity) isthermal differences in a system caused by operating con-ditions It is noted by Domes [983093] that this nonsymmetricalthermal distribution can cause excessive unbalance to theextent where a gas turbine will not start correctly Such rotorbows are common on start up or shut down and are ofenaccounted or in operational procedures However i thermalldquohot spotsrdquo exceed a given tolerance level they can causepermanent unbalances due to rotor de1047298ections Such rotorbows can lead to other aults includingrubbing andloosenesswhich complicate isolation and localization

raditional data-driven techniques or detecting rotorbows involve combinations o slow roll and vibration data[983090983097] More recently mathematical modeling techniques suchas that detailed by Meagher et al [983091983088] have been used inorder to diagnose residual rotor bows and differentiate theseaults rom other sources o unbalance Te authors o thispaper build upon established methods or the models and areunique in that they concentrate on response at the bearingpoints Tis is perhaps more useul orindustrial applicationsas vibration inormation is more readily available at bearingpoints rather than intrusive proximity probes which are ofenused in lab work

Te little work that exists on attempting to localise rotorbows across complex systems tends to be data-driven in

nature see Gaka and abaszewski [983091983089] where the authorsused statistical symptoms based on known data as a methodo diagnosing and prognosing a number o aults includingrotor bows and unbalance Tis paper attempts to address theirregularities and 1047298uctuations that occur over a long servicelie In order to achieve this a modi1047297ed energy processormodel is created using data drawn rom large steam turbinesover long periods o lie

Prognosing rotor bows is a complex subject As rotorbows are ofen caused by temperature de1047298ections makingpredictions or remaining useul lie and potential utureproblems lies not only in the realm o mechanical rotor-dynamics but also to some extent in thermodynamics Te

recent work detailed by Sinha [983091983090] is o note or detailingdiagnosis and quanti1047297cation o various rotordynamic aultsand describing the advantages o mathematical modelingover traditional vibration-based approaches Te topic o scalability regarding FEA models is discussed including anargument or the use o partial (simpli1047297ed) mathematical

models or large complex systemsAnother two works which are o interest with regard to

modeling o rotor bows include Shen et al [983091983091] where theauthors modeled a rotor-bearing system with a permanentrotor bow looking at the impact o secondary aults such asrub Te study o ault combinations in this paper is useulor ault differentiation studies however the authors study a permanent initial rotor bow Tis thereore does not takeinto account developing or worsening aults and the different

vibration phenomena that are observed as such aults aredeveloping

Lees et al [983091983092] describe the importance o model-basedault identi1047297cation techniques and outline recent research in

the area providing a good reerence paper or more researchon this speci1047297c ault

3 Discussion

983091983089 Sensor Suites Te subject o sensor suites is o greatimportance with regard to uture industrial applications Onone hand advanced and complex sensor suites generally enable improved ault localization and diagnosis howeverthe added complexity and cost has resulted in many o thesesystems being omitted rom the latest generations o rotatingmachinery in industrial applications From an industrialperspective the cost bene1047297t o additional sensors needs to

be signi1047297cant in order to justiy this approach In response anumber o works listed including [983096] as a good example takethe approach o achieving the same objectives using a greatly reduced sensor suite In many ways the required sensor suiteor many o the currently researched ault diagnosis andprognosis techniques provides an indication o the ability o the techniques to be used practically in industrial situationsA complex suite may return a very high success rate on alab-based rig however the impracticalities o mounting suchsuites on a real-world application negate the advantages

Further to this the consideration orsensor position mustbe considered Te use o proximity probes and keyphasortransducers has very clear advantages in a number o situ-

ations however the intrusive nature o the sensor preventsapplication in a number o complex systems With this inmind a number o the more promising studies listed rely onthe simple suiteo an accelerometer placed in positions whichare relatively remote rom the sources o vibration potentially with a noisy transer path

983091983090 Diagnosis It can be seen rom the research outlined inthe previous section that the diagnosis o aults in rotatingmachinery is a subject o much ongoing research Tisinvolves the improvement and development o traditional

vibration monitoring techniques development o new data-driven technologies and novel research into physics-based

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 813

983096 International Journal o Rotating Machinery

simulation and modeling In many cases these topics o research are dependent on one another or reasons o vali-dation veri1047297cation and speed o analysis In several casesit can be seen that multiple aults have been modeled orthe purposes o identi1047297cation and isolation However nostudies have yet been perormed which deal with all o the

aorementioned aults All o these aults are intrinsically related to one another Complex combinations o aults havebegun to be analyzed with the emphasis on developing new diagnosis techniques Physics-based modeling has proven toprovide signi1047297cant advances with regard to speci1047297c aultsnotably shaf and blade cracks where techniques such as FEAenable much easier aster and cheaper test data then seedingaults into live systems

It can be noted however that both high and low-1047297delity modeling techniques are being applied to cutting edgeresearch or all o the listed rotordynamic aults (and othersnot detailed in this paper) In addition to the advantages inthe speed o obtaining test results physics-based simulationis providing another dimension to data-driven techniquesSystem models are being used as part o logic and reasoningsuites in the identi1047297cation and differentiation o variousaults High-1047297delity models enable simulations o ault com-binations which are not possible practically or are pro-hibitively expensive in live systems

Nevertheless results obtained rom modeling studies stillneed to be validated against proven data-driven techniquesbeore implementation in industrial applications is possibleIt is also o note rom the literature reviewed or this paperthat almost all modeling or the diagnosis o aults involvesextensive simpli1047297cation ofen reducing potentially complexsystems down to one or two discshafbearing modelsAdapting the claimed results rom such research to systemswith many discsbearingsshafs is important in improvingexisting rotating machinery diagnostic techniques

Te inclusion o nonlinear effects is notable amongstmany authorsmdashas the understanding o the complex aultsand interactions continues to progress Such research opensthe potential or promising new avenues into areas o designor IVHM

Finally it is o note that despite the wide variety o advanced research here a consensus or the most efficienttechniques and algorithms has yet to emerge in order or the1047297nal adaption to industrial application to be achieved How-ever the consideration or industrial application taken by many o the authors detailed in this paper provides promise

or next generation IVHM systems

983091983091 Localisation Fault localization in rotating machinery isan important topic o research or uture condition-basedmonitoring systems Knowing not only what type o aulthas occurred but also where in the system is an importantconsideration which can in1047298uence maintenance proceduresin complex machinery It is worth noting rom the literaturesurveyed or this paper that many studies have ocusedon diagnosis and prognosis o single rotorbearing systemsOfen or legitimate reasonsmdashsimpli1047297cation or computingspeed or example Few studies however have taken intoaccount the localization o aultsacross whole systems Oneo

the more extensive examples highlighted in this paper (Hanet al [983089983093]) provides extensive analytical studies relevant tolocalization on a duel disk setup utilizing Hilbert-Huangransorms however in this example comprehensive valida-tion is lacking thus limiting the potential applicability o theresearch

Tis problem is not limited to modeling and simulation-based research Many newly developed data-driven tech-niques or diagnostics and prognostics claim good results by heavily instrumenting speci1047297c components o a test systemIn many industrial cases this is not possible practical or cost-effective

o give some examples a keyphasor transducer can beparticularly useul in diagnosing aults such as rotor bowHowever this equipment requires the ability to cut a key way or measurements to be perormed Optical sensors haverecently been applied to detect rotor unbalance yet such atechnology would be difficult to implement in a system withseveral rows o rotors (as in a gas turbine)

Whilst new diagnosis and prognosis techniques or aultsin rotating machinery are being continuously researchedthe lack o corresponding studies into localization can beconsidered one o the many challenges in promoting recentcore research into live industrial applications

983091983092 Fault Differentiation One promising development inrecent modeling and simulation o rotordynamic aults isthat o ault differentiation Several researchers have movedon rom studies on individual aults in order to concentrateon combinations o aults As stated rotordynamic aultssuch as those listed in this paper are linked to each otherSuch studies thereore concentrate on such topics as a

misalignment causing an unbalance or a looseness causinga rub Findings rom reports such as these are important inunderstanding complex anomalies In industrial applicationssimply detecting and rectiying an unbalance does not pro-

vide a satisactory solution i the root cause o the ault isa misalignment Tis is a complex topic as several aults canexhibit similar vibration characteristics making traditionaldetection techniques inaccurate in some cases

Making reerence to some o the aorementioned stud-ies those such as Bachschmid et al [983090983091] and Sonnichsen[983090983093] indicate some initial studies including several aultsTe results o such studies indicate a high success rate indifferentiating the aults in the given conditions however

a number o limitations still exist Studies such as theseconsider unbalancein the aultchains Tisis typically ldquostaticrdquoor simple unbalance applied to a single shafrotormdashin reality unbalance aults can be more complex than this In additionthe simple systems used to study and differentiate aultsare very different rom the complexities o a ull turbineHowever urther work may enable such research to ldquoscale-uprdquo to this stage with the aid o urther validation and

veri1047297cation

983091983093 Prognosis It can be seen rom the examples o recentresearch described or common aults that prognostic tech-niques are a topic where much work is being perormed

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 913

International Journal o Rotating Machinery 983097

Some o this work would be very difficult or impossibleto implement in an industrial situation others providepromising results o use to uture work in the area Predictingthe remaining useul lie o components is critical in thedevelopment o condition-based monitoring strategies orindustrial implementation

It can be noted that prognostic studies or certain rotor-dynamic aults are considerably more advanced than orothers Te most obvious examples o the aults detailed inthis paper are bearing aults shaf and blade cracks Teseare also areas where physics-driven simulations have hadan important impact Te ability to design any ault type(or combinations o aults) into such simulations (to be itslow-1047297delity mathematical models or high-1047297delity 1047297nite ele-ment analysis) has provided researchers with many differentavenues to explore Studies into the prognostics o otheraults can be complicated by various actors An unbalanceor example has had relatively little research perormed intoprognosis Te act that unbalance is ofen the cause o another underlying ault is one reason or the difficulty inresearching prognosis in detail or this ault Te results o unbalance are also very dependent on the severity o the aultand the system in which it occurs

It is worth noting that even or the aults where prognosisresearch is more advanced (eg shaf cracks) true develop-ment o the ault is not always studied Fan et al [983089983095] orexample study sets o discrete cracks in order to ldquoprognoserdquocrack growth In act thisstudy is more akinto detecting aultseverity with inerred effects on crack growth In contrast Liet al [983090983090] consider the continual development o such crackshowever the authors do admit this work is in its inancySuch points limit the ability or true prognosis to be achievedwithout extensive operational data

It is also worth noting that combinations o aults havebeen simulated extensively or the purpose o diagnosticshowever ew studies exist combining aultsor the purpose o prognostics Such studies are however a logical progressionrom some o those already perormed It may be that suchprognostic studies build upon some o the condition-basedmaintenance (CBM) or diagnostics High-level studies suchas Jaw and Merrill [983091983093] indicate how prognosis and diagnosistechniques can be combined into a CBM system Tis study is designed or a military aeroengine indicating the desire orsuch systems to be implementedmdashalthough the architecturedescribed has the capability o including various algorithmsoraultprognosis the amount o consideration orprognosis

o ault combinations is unclear

983091983094 Modelling Te large number o works reviewed or thispaper which includes modeling techniques in order to aiddiagnosis and prognosis indicates the possibilities providedby modern computer power and sofware developmentBroadly the modeling studied in the aorementioned papersbreaks down into two categories mathematical modeling and1047297nite element analysis (FEA) Te mathematical models suchasBorghesani etal [983090983089] ofen provide a theoretical basis uponwhich data driven or FEA studies can build and validateMathematical models despite being a traditional approach to

rotordynamic problems thereore continue to be developedto tackle more advanced problems

FEA techniques offer a constantly expanding area o simulation to explore Tey are used or research as diverseas crack propagation to unbalance localization [983091983094] Moderncodes such as NASRAN and Ansys enable rotordynamics to

be studied beyond the traditional stress and modal analysisTe ability to study the effects o rotation (eg throughCampbell diagrams and transient analysis) continues to driveresearch and innovation in this area Te studies discussedin this paper indicate the power o FEA with some highly accurate simulations having been perormed However thereis still a limitation o computer powermdashapplying some o the FEA techniques to whole engine models would proveto be too computationally expensive to be viable As aresult alternate methods such as model order reductionand system level modeling (and model-based reasoning orimplementation) are still required in order to make some o the FEA studies viable in an industrial environment

Nonlinearities continue to orm an important part o many recent studies with the complexity and detail o thespeci1047297ed aults being continually expanded Te depth o modeling has enabled in many cases accurate validationagainst experimental approaches Despite this completeunderstanding o the vibrational phenomena o rotatingmachinery is not currently possible despite the drive towardsthis end

983091983095 Rotordynamics and IVHM Te bulk o current researchinto rotordynamics rom the point o view o prognostichealth management (PHM) can be roughly divided intotwo types initial single-ault diagnosisprognosis techniquesand studies into the general requirements and limitationso current systems along with current and uture trendsAn example o the latter is Pusey [983091983095] who provides agood summary overview o current diagnosis and prognosistechniques with regard to condition-based maintenance

As a result o this split a clear gap exists between thecore research being perormed into rotordynamics rom acondition-based maintenance perspective and the identi1047297edneeds o industry aking a 1047298edgling piece o research andapplying it to a commercially-ready system (eg a gas turbineengine or an aircraf) is a long and complex task It isnevertheless worth noting that technologies or automatically detecting an unbalance or misalignment in a gas turbine were

developed over 983089983088 years beore the latest commercial aircrafwere conceptualized and yet these planes are still limited inthis capacity Tis highlights the need or work which linksthe undamental research into individual ault diagnosis toldquolive systemsrdquo in use in industry

Physics-based simulation and modeling o rotordynamicparts is a well-researched 1047297eld Such modeling has been usedas the basis o diagnosis and prognosis o aults by many researchers several recent examples have been outlined inthis paper Occasional pieces o work have been perormedinto modeling multiple aults such as Jain and Kundra [983091983096]who usea system model oronline identi1047297cation o unbalanceand cracks Beyond this however very limited research

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 1013

983089983088 International Journal o Rotating Machinery

Business casecost bene1047297tanalysis

Live system

RubCracksBearingfailure

Rotor bow Unbalance

Fault differentiationLocalisation

Root cause analysis

Remaining useful life

Data driven analysis and physics-based simulation

Mis-alignment

Experimental

1047297eld data

F983145983143983157983154983141 983089 Physics-based simulationmdashrom research to industry

exists Te demands o PHM techniques in industry are suchthat any systemmust notonly be capable o detecting multipleaults but must also be capable o detecting these aultsacrossa range o different systems Other considerations includethe aorementioned ability to differentiate between multipleaults Processing also needs to be taken into account as theobjective o these systems is to implement efficient conditionmonitoring and condition-based maintenance procedures I processing data is a long power-hungry process then thisaim cannot be achieved

Figure 983089 details a potential ramework required in order

to push core research such as that detailed in this reporttowards industrial applications Many studies now exist onindividual rotordynamic aults across a wide range o condi-tions and applications Some studies have taken this urtherwith advanced prognostic models and diagnosis o dual aults(primary cause and secondary effect) Future research inthe area o rotordynamics rom a PHM perspective couldpotentially provide the bridge between these studies and livesystems by combining physics-based simulations with data-driven techniques and validating against experimental data

Although quantiying the success o the research studiedisdifficultit ispossibleto de1047297ne the key areas in whicha tech-nique must excel in order to be considered viable Te work

byWheeleretal[983091983097] discussesin detailmetrics ordiagnosticand prognostic analysis as does that by Vachtsevanos [983092983088]and Saxena et al [983092983089] Te conclusions drawn rom thesepapers and applied in practice to research like that covered by this paper indicate that the ollowing metrics are importantwhen considering the potential o a given technique to diag-nose aults coverage alse positive rate and alse negativerate In the case o prognosis probabilities and lead timeto ailure are other important considerations Tis criteriaenables research to be assessed in terms o its suitability orindustrial applications Unortunately inormation on these

metrics is not made readily available by the authors o mostpapers

In terms o evaluating the effectivenesso theresearch dis-cussed in this paper there are difficulties in recommending aspeci1047297c technique over others or general application Mostpapers reviewed or this document (not just those reerencedand discussed in detail) take a technique (new or evolved)

validate or a given speci1047297c system and report success o theresearch Despite this some conclusions can be drawn romassessing common techniques applied across different studiesand different aults Although it is not possible to de1047297ne themost common methods or diagnosing andprognosing aultsin terms o numbers (as it was not possible to cover all recent

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 1113

International Journal o Rotating Machinery 983089983089

rotordynamics research or this paper) as perceived by theauthors the ollowing techniques have eatured prominently in the reviewed research

(i) Sensors accelerometers

(ii) Teoretical studies mathematical modeling

(iii) Physics-based simulation FEA(iv) Data-driven joint timerequency domain analysis

(v) Implementation neural networking

Tese techniques appear to be among the most promisingcurrently under development as they tend to eature numer-ous timesamongst some o the work with wider scope acrossault types and with the most comprehensive validationTere are o course many subsections to these techniqueshowever it shows one general direction o research and theclear possibilities posed in these areas

Future developments in the 1047297eld o IVHM or rotat-ing machinery may incorporate these techniques alongsideextensive use o nonlinear modeling and multiple aultinteractions Te 1047297eld o design or IVHM has only recently emerged however the potential exists or speci1047297c nonlinear-ities to be designed into a system in order to enable accuratediagnosis andprognosis o aults Te development o currentalgorithms to include diagnosis localization and prognosiso a range o aults will provide a signi1047297cant advancement oruture generations o IVHM systems Tis combined withcost-effective sensor suites indicates the potential or evolu-tions o some o the research detailed hereto orm part onextgeneration IVHM suites or rotating machinery

4 Conclusion

Tis paper has reviewed some o the latest research around anumber o rotordynamic aults namely unbalance misalign-ment rub and looseness 1047298uid-induced instability bearingaults shaf cracks blade cracks and rotor bow Each aultwas reviewed rom the perspective o sensors diagnosisprognosis localization and modeling

Key examples o recent work into the eight describedaults have been detailed through works by a number o eminent authors Additional work has been summarized andormatted or easy reerence Some current trends amongstthe recent body o work include developments in the vastarea o modeling nonlinearities combinations o high andlow 1047297delity modeling and synergy between data driven and

physics-based simulation approachesDespite the large volume o promising research reviewed

urther development in a number o areas is required in orderto produce effective next generation IVHM systems As suchuture developments may include usion o andor devel-opment o current algorithms to encompass all eight aultsdetailed consideration o prognosis diagnosis and localiza-tion achieved using a reduced cost-effective sensor suite

References

[983089] I K Jennions Integrated Vehicle Health Management Perspec-tives on an Emerging Field 983090983088983089983089

[983090] G Anderson ldquoProviding best-value IVHM solutions or agingaircrafrdquo in Proceedings o the 983097th Joint FAADoDNASA Aging Aircraf Conerence vol 983089 pp 983089ndash983089983089 983090983088983088983093

[983091] A Muszynska Rotordynamics aylor amp Francis 983089st ed edition983090983088983088983093

[983092] D E Bently Fundimentals o Rotating Machinery Diagnostics

ASME Press 983089st edition 983090983088983088983090[983093] B Domes ldquoVibration phenomena in aero-enginesrdquo in Proceed-ings o the 983097th International Conerence on Vibrations in Rotating Machinery Exeter (IMechE rsquo983088983096) vol 983089 pp 983089983093ndash983091983090 September983090983088983088983096

[983094] A Hess G Calvello and Dabney ldquoPHM a keyenabler or theJSF autonomie logistics support conceptrdquo in Proceedings o theIEEE Aerospace Conerence Proceedings pp 983091983093983092983091ndash983092983093983092983097 March983090983088983088983092

[983095] S N Ganeriwala B Schwarz and M H Richardson ldquoOperat-ing de1047298ection shapes detect unbalance in rotating equipmentrdquoSound and Vibration vol 983092983091 no 983093 pp 983089983089ndash983089983091 983090983088983088983097

[983096] G N D S Sudhakar and A S Sekhar ldquoIdenti1047297cation o unbal-ance in a rotor bearing systemrdquo Journal o Sound and Vibration

vol 983091983091983088 no 983089983088 pp 983090983090983097983097ndash983090983091983089983091 983090983088983089983089[983097] Yang and M W Hsu ldquoAn efficient diagnosis technique or

variations o shaf-bow and unbalancerdquo in Proceedings o the ASME International Design Engineering echnical Conerencesand Computers and Inormation in Engineering Conerence(DEC rsquo983088983097) pp 983093983095ndash983094983094 San Diego Cali USA September 983090983088983088983097

[983089983088] A El-Shaei S H aw1047297ck and M O A Mokhtar ldquoExperi-mental investigation o the effect o angular misalignemt onthe instability o plain journal bearingsrdquo in Proceedings o the ASMESLE InternationalJointribology Conerence(IJC rsquo983088983097)pp 983089983095983089ndash983089983095983091 Memphis enn USA October 983090983088983088983097

[983089983089] A El-Shaei S H aw1047297ck and M O A Mokhtar ldquoMisalign-ment modeling in rotating systemsrdquo in Proceedings o the ASMEurbo Expo H Bahaloo A Ebrahimi and M Samadi Eds vol

983089983095983089983089 pp 983097983095983091ndash983097983095983097 Orlando Fla USA June 983090983088983088983097

[983089983090] L F Villaet al ldquoStatistical diagnosisbased on vibration analysisor gear test-bench under non-stationary conditions o speedand loadrdquo Mechanical Systems and Signal Processing vol 983090983097 pp983092983091983094ndash983092983092983094 983090983088983089983090

[983089983091] C F Ngolah et al ldquoIntelligent ault recognition and diagnosisor rotating machines using neural networksrdquo Sofware Scienceand Computational Intelligence vol 983091 no 983092 983089983095 pages 983090983088983089983089

[983089983092] Y Lu Z Ren H Chen N Song and B Wen ldquoStudy on loose-ness and impactmdashrub coupling aults o a vertical dual-disk cantilever rotor- Bearing systemrdquo Key Engineering Materials vol 983091983093983091ndash983091983093983096 no 983092 pp 983090983092983095983097ndash983090983092983096983090 983090983088983088983095

[983089983093] Q Han Z Zhang and B Wen ldquoPeriodic motions o a dual-

disc rotor system with rub-impact at 1047297xed limiterrdquo Proceedingso the Institution o Mechanical Engineers vol 983090983090983090 no 983089983088 pp983089983097983091983093ndash983089983097983092983094 983090983088983088983096

[983089983094] H F de Castro K L Cavalca and R Nordmann ldquoWhirland whip instabilities in rotor-bearing system considering anonlinear orce modelrdquo Journal o Sound and Vibration vol 983091983089983095no 983089-983090 pp 983090983095983091ndash983090983097983091 983090983088983088983096

[983089983095] C C Fan et al ldquoMechanical systems and signal processingrdquoStudy o Start-Up Vibration Response or Oil Whirl and Dry Whip vol 983090983093 pp 983091983089983088983090ndash983091983089983089983093 983090983088983089983089

[983089983096] K Kappaganthu and C Nataraj ldquoNonlinear modeling andanalysis o a rolling element bearing with a clearancerdquo Commu-nications in Nonlinear Science and Numerical Simulation vol983089983094no 983089983088 pp 983092983089983091983092ndash983092983089983092983093 983090983088983089983089

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 1213

983089983090 International Journal o Rotating Machinery

[983089983097] C Gupta K Gupta and D K Sehgal ldquoInstability and chaoso a 1047298exible rotor ball bearing system an investigation on thein1047298uence o rotating imbalance and bearing clearancerdquo Journal o Engineering or Gas urbines and Power vol 983089983091983091 no 983096ArticleID 983088983096983090983093983088983089 983089983089 pages 983090983088983089983089

[983090983088] J Hong X Miao L Han and Y Ma ldquoPrognostics model orpredicting aero-engine bearing grade-lierdquo in Proceedings o the ASME urbo Expo vol 983089 pp 983094983091983097ndash983094983092983095 Orlando Fla USA June983090983088983088983097

[983090983089] P Borghesani P Pennacchi R B Randall N Sawalhi and RRicci ldquoApplication o cepstrum pre-whitening or the diagnosiso bearing aults under variable speed conditionsrdquo Mechanical Systems and Signal Processing 983090983088983089983090

[983090983090] Y Li J Zhang L Dai Z Zhang and J Liu ldquoAuditory-model-based eature extraction method or mechanical aults diagno-sisrdquo ChineseJournal o MechanicalEngineering (EnglishEdition) vol 983090983091 no 983091 pp 983091983097983089ndash983091983097983095 983090983088983089983088

[983090983091] N Bachschmid P Pennacchi and E anzi ldquoCracked rotatingshafs typical behaviors modeling and diagnosisrdquo IUAM Symposium on Emerging rends in Rotor Dynamics vol 983089983088983089983089 pp

983092983092983089ndash983092983093983092 983090983088983089983089[983090983092] Inoue N Nagata and Y Ishida ldquoFem modelling and

experimental veri1047297cation o a rotor system with a open crackrdquoin Proceedings o the ASME International Design Engineering echnical Conerences and Computers and Inormation in Engi-neering Conerence (DEC rsquo983088983097) pp 983089983089983089983091ndash983089983089983090983090 San Diego CaliUSA September 983090983088983088983097

[983090983093] H E Sonnichsen ldquoReal-time detection o developing cracksin jet engine rotorsrdquo in Proceedings o the IEEE AerospaceConerence pp 983089983095983091ndash983089983096983092 March 983090983088983088983088

[983090983094] I Green and C Casey ldquoCrack detection in a rotor dynamicsystem by vibration monitoringmdashpart I analysisrdquo Journal o Engineering or Gas urbines and Power vol 983089983090983095 no 983090 pp 983092983090983093ndash983092983091983094 983090983088983088983093

[983090983095] J Sawicki M I Friswell A H Pesch and A WroblewskildquoCondition monitoring o rotor usingactive magnetic actuatorrdquoin Proceedings o the ASME urbo Expo pp 983089983090983093983095ndash983089983090983094983093 BerlinGermany June 983090983088983088983096

[983090983096] J Xiang X Chen Q Mo and Z He ldquoIdenti1047297cation o crack ina rotor system based on wavelet 1047297nite element methodrdquo FiniteElements in Analysis and Design vol 983092983091 no 983089983092 pp 983089983088983094983096ndash983089983088983096983089983090983088983088983095

[983090983097] M G MaalouldquoSlow speed vibrationsignal analysisi you canrsquotdo it slow you canrsquot do it astrdquo in Proceedings o the ASME urboExpo pp 983093983093983097ndash983093983094983095 Montreal Canada May 983090983088983088983095

[983091983088] J Meagher X Wu and C Lencioni ldquoResponse o a warped1047298exible rotor with a 1047298uid bearingrdquo International Journal o

Rotating Machinery vol 983090983088983088983096 Article ID 983089983092983095983094983093983091 983097 pages 983090983088983088983096

[983091983089] Gaka andM abaszewski ldquoAn applicationo statistical symp-toms in machine condition diagnosticsrdquo Mechanical Systemsand Signal Processing vol 983090983093 no 983089 pp 983090983093983091ndash983090983094983093 983090983088983089983089

[983091983090] J K Sinha ldquoRecent trands in ault quanti1047297cation in rotatingmachinesrdquo Advances in Vibration EngIneerIng vol 983096 no 983089 pp983095983097ndash983096983093 983090983088983088983097

[983091983091] X Shen JJia andM Zhao ldquoNonlinear analysis o a rub-impactrotor-bearing system with initial permanent rotor bowrdquo Archiveo Applied Mechanics vol 983095983096 no 983091 pp 983090983090983093ndash983090983092983088 983090983088983088983096

[983091983092] A W Lees J K Sinha and M I Friswell ldquoModel-based identi-1047297cation o rotating machinesrdquo Mechanical Systems and Signal Processing vol 983090983091 no 983094 pp 983089983096983096983092ndash983089983096983097983091 983090983088983088983097

[983091983093] L C Jaw and W Merrill ldquoCBM+ research environmentmdashacilitating technology development experimentation andmaturationrdquo in Proceedings o the IEEE Aerospace Conerence(AC rsquo983088983096) March 983090983088983088983096

[983091983094] R Walker S Perinpanayagam and I K Jennions ldquoSimulatingUnbalance or Future IVHM Applicationsrdquo in Proceedings o theSociety or Experimental Mechanics (IMAC rsquo983089983090) vol 983090983097 pp 983089983092983089ndash983089983092983096 May 983090983088983089983090

[983091983095] H C Pusey ldquourbomachinery condition monitoring and ail-ure prognosisrdquo Sound and Vibration vol 983092983089 no 983091 pp 983089983088ndash983089983093983090983088983088983095

[983091983096] J R Jain and K Kundra ldquoModel based online diagnosiso unbalance and transverse atigue crack in rotor systemsrdquo Mechanics Research Communications vol 983091983089 no 983093 pp 983093983093983095ndash983093983094983096983090983088983088983092

[983091983097] K R Wheeler Kurtoglu and S D Poll ldquoA survey o healthmanagement user objectives related to diagnostic and prognos-tic metricsrdquo in Proceedings o the 983090983097th Computers and Inor-mation Engineering Conerence vol 983090 pp 983089983090983096983095ndash983089983090983097983096 983090983088983088983097

[983092983088] G Vachtsevanos ldquoPerormance metrics or ault prognosis o

complex systemsrdquo in IEEE Systems Readiness echnology Con- erence (AUOESCON rsquo983088983091) pp 983091983092983089ndash983091983092983093 September 983090983088983088983091

[983092983089] A Saxena J Celaya E Balaban et al ldquoMetrics or evaluatingperormance o prognostic techniquesrdquo in Proceedings o theInternational Conerence on Prognostics and Health Manage-ment (PHM rsquo983088983096) pp 983089ndash983089983095 October 983090983088983088983096

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 1313

Submit your manuscripts at

httpwwwhindawicom

Page 3: Rotordynamic Faults 2013

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 313

International Journal o Rotating Machinery 983091

simulation work and traditional data-driven methodsResearch in this area o rotordynamics is particularly wide-spread and so this paper outlines only a ew recent areaso research Although not being o mainstream interest alist o commonly used sensors has been included able 983089represents a summary o the inormation collected and

detailed as a part o this paperAs discussed a wide selection o papers have beenreviewed or consideration in this paper o which it is notpossible to describe them all in detail In an attempt tohighlight some o the common themes across many o worksreviewed able 983089 has been created able 983089 by no meansdetails all relevant methods and approaches currently beinginvestigated however it details themes which were oundto be common across many areas o current research It isintended or the detailed table to compliment the importantworks described in more detail in the ollowing section o thepaper separated by ault type

983090983089 Unbalance Tis is one o the most common rotordy-namic aults [983092] every rotating machine has an inherent

degree o unbalance Unbalance as a ault can thereorebe de1047297ned as unbalance outside o a given tolerance levelA recent piece o research which demonstrates the ongoingdevelopment o data-driven techniques is that by Ganeriwalaet al [983095] who tested a technique or measuring operatingde1047298ection shapes (ODS) in order to detect unbalance casesTese studies were conducted on a machine ault simulator (aphysical simulator o common aults in rotating machines)mdashsuch simulators have the advantage o recreating aulty con-ditions quickly and easily enabling a new dimension to data-driven diagnostic techniques It is however worth mention-ing that data or these experiments were collected using 983089983092accelerometers which are easy to apply to such a simulatorbut it may be much more difficult to con1047297gure these many sensors on a complex system Despite this the paper achieveditsaim in proving thehypothesis ldquowhen an operating machinebecomes unbalanced its ODS will changerdquo

Regarding physics-based simulation o unbalance as aault the work by Sudhakar and Sekhar [983096] described amodel-based method or ault identi1047297cation using a minimalsensor suite Tis is achieved through the analysis o trans-

verse vibrations at a single location Troughout the paperthree different approaches are studiedmdashleast squares mini-mization equivalent loads minimization and vibrationmini-mization A reduced error is ound to occur using a proposed

modi1047297cation to the typical equivalent loads minimizationapproach Te work is o note due to the authors requirementto both identiy and locate (localize) unbalance

Localization and prognosis o unbalance pose a uniqueseto challenges andas a result research is still somewhat lim-ited in these areas One recent work which claims to localiseunbalance accurately is by Yang and Hsu [983097] the authorsuse trending data and reasoning systems to locate unbalanceand shaf bow across a system Quick diagnosis is achievedby avoiding the study o all ault combinations and theauthors claim the ability to localise shaf bow and unbalancein large rotating machines Te techniques rely extensively onprevious searches and are limited to large stable machines

operating at a speci1047297c RPM Remaining useul lie o unbal-ance is difficult to predict due to complicating actors Asdiscussed the ault dependencies are such that in many casesthe unbalance itsel will not be the ailing actor instead anunbalance may cause a rub which then leads to damagewhichcan be prognosesSuch combinations o aults and underlying

causes or unbalance lead to the need or remaining useullie predictions to be made based upon the exact nature o aspeci1047297c ault

983090983090 Misalignment Tis is another common ault which canpotentially in1047298ict considerable damage in rotating machinesAs with unbalance misalignment in a whole system can becomplicated by secondary aults (eg a misalignment whichcauses a rub) El-Shaei et al [983089983088] are an example o theongoing research in this area in this case a unique combina-tion o angular misalignment and oil whipwhirl is detailedTe authors describe how small degrees o misalignment canbe utilized in order to prevent the onset o 1047298uid-induced

instabilities tested through the use o a lab-based test rigSuch research presents a new dimension in looking atcommon rotordynamic aults with aspects that could beapplied to uture design or IVHM systems

El-Shaei et al [983089983089] demonstrate interesting research intomisalignment rom the perspective o physics-based simula-tion Te authors construct mathematical models o a simplerotorsystem witha misaligned coupling andcollect harmonicresponse data rom this to assess the severity o differentmisalignment cases Such models are useul throughout thelie o rotating machinesmdashrom design to implementationalthough again successul validation with experimentally obtained data is key Te authors highlight the act thatalthough misalignment is a prevalent and serious ault nocomprehensive research has been perormed or treating thisproblem Te methodology applied includes deriving theenergy expressions applying the Ritz series method con-structing the equations o motion and then using the har-monic balance method to look or multiharmonic responsesTe paper demonstrates the ongoing research to understandand model the undamentals o such aults in order thatimproveddiagnosisand prognosis methods can be built uponsuch knowledge

As with unbalance localization and prognosis o mis-alignment is a complex topic to research Studies such as El-Shaei et al [983089983089] can make accurate predictions or misalign-ment in a simple system with one coupling However real

systems (eg aircraf gas turbines) have many potential loca-tions o misalignment Tis is an area where ew researchershave made an impact Remaining useul lie predictions ormisalignment is complicated or the same reasons as withunbalance Villa et al [983089983090] discuss statistical diagnosis o misalignment aults with reerence to prognosis Te authorsuse the example o a wind turbine or their studies butstress the applicability to other systems Differentiation withunbalance aultsis also covered (as these two aultsare closely linked) Unlike the work by El-Shaei et al [983089983089] emphasis isgiven to the machine in question operating over a wide rangeo operating speeds and conditions Tis is achieved throughthe use o an angular resampling method Prognosisis tackled

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 413

983092 International Journal o Rotating Machinery

983137983138983148983141 983089 Rotordynamic ldquoault tablerdquo

Faults Sensors Fault identi1047297cation

techniques Localisation o ault Prognosis

Work o note in aultmodelling

UnbalanceAccelerometer

Velocity transducer

Proximity transducer

Operating de1047298ectionshapesime-requency

analysis and many others

Neural-network modelling and

time-requency analysis

FCM-Markov model

Unbalance in ull-engineFEA modelEarly detection inaeroengines

High radialloads andmisalignment

AccelerometerTermocoupleDisplacement

transducerAngular sensor

Bearing temperatureobservation advancedSofware-driven1047297ltering methods

Various vibrationanalysis techniques

Prognosticmodelling usingsimulated vibrationdata

Modelling therelationship betweenmisalignment and otheraultsHarmonic balancemethod

Rub andlooseness

AccelerometerVelocity transducer

Optical sensor

Single valuedecompositionime-requency analysis

Least-squares 1047297ttingapproach

Adaptive waveletdenoising orprognostics

Rob and looseness usingFEA models Wholeengine modellingtechniques

Fluid-inducedinstability

Proximity probesKeyphase transducer

Full spectrum plotsLarge amplitudesubsynchronous vibrationDetection

Blade tip clearanceand loadingdistribution

Cross-coupledstiffness anddirect-dampingstudies

Nonlinear modellingaccuracy o model basedDiagnostics numericalmodelling

Bearing ailure

AccelerometerVelocity transducer

Displacementtransducer

Proximity probes

Postailure analysis oruture ailure diagnosisDynamictime-requency analysis

Statistical change inthe bispectraldomain

Numericalmodelling o aeroengine bearingsModel-basedprediction o remaining useul lieOn-boardprognostics

High and low 1047297delity dynamic systemmodellingComparisons o algorithms tounderstand bearing vibration phenomena

Shaf cracks

Keyphase transducer

Proximity probes

Higher order spectra

Neural networksGrid support vectors

Forced response

measurements

Nonintrusive

torsion vibrationmonitoring

Statistical analysis FEA

modelling

Blade cracks Keyphase transducer

Proximity probes

Hilbert-HuangtransormationModel-baseddiagnosticsMorletwaveletquantumtechniques

Novel bearingstructures

Active magneticactuator

Global and localasymmetry modellingFinite element modelling

Rotor Bow Keyphase transducer

Proximity probesSeismic transducer

Examination o ldquoslow rollrdquo dataModel-baseddiagnostics

Partial mathematicalmodelling

Statistical symptomsanalysis

Advantages o aultmodelling oridenti1047297cationNonlinear modellingtechniques

through the use o a statistical diagnosis algorithm based onthe signi1047297cance level o the aults in question

983090983091 Rub and Looseness Rub is always a secondary ault (iea product o another ault such as looseness) and can leadto atigue and wear Rub and looseness can create com-plex vibration signals which are difficult to diagnose usingtraditional methods Modeling and simulation o rub andlooseness aults have been considered in several recentworks

Tis includes Ngolah et al [983089983091] which detail the moni-toring and diagnosis o common aults (including rub and

looseness) based upon a three-layer Arti1047297cial Neural Net-work (ANN) A series o 983089983088 key perormance indicators wereidenti1047297ed and used as training Te authors test the systemin a lab environment but stress the applicability to industrialapplications Te research indicates one o the latest methodso research which enables the implementation o diagnosistechniques It is a useul tool or rub and looseness studiesas it incorporates a variety o aults which could ldquounderlierdquosuch a ault Despite this the research relies on clear eaturesor each ault which can be much easier to identiy in a labenvironment as opposed to ldquonoisyrdquo industrial applications

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 513

International Journal o Rotating Machinery 983093

Lu et al [983089983092] have perormed several studies into rub andloosenessincluding thisexamplein which a 1047298exibly mountedshaf has an induced rub due to a range o contact ringsTe study ocusses on the potentially dangerous effects o rubin causing excessive nonsynchronous and chaotic vibrationsTe links with unbalance and misalignment are discussed

and detailed Te considerations or real-world cases in thedescribed research are considerable as part o the drivetowards ull understanding o the nonlinear effects o rub andlooseness

Localization o rub and looseness across whole systemsis relatively lightly studied in the literature Many works(including those already cited) look at single or dual-rotorsystems where localization o such aults is not an issue Inan industrial setting complex systems may comprise many rotors in several compressor and turbine stages signi1047297cantly complicating diagnosis o such aults Research into progno-sis o rotor-stator rubs lies mostly within the domain o data-driven techniques Modeling and simulation research can beused to support data-driven techniques or prognosis andcondition-based monitoring Han et al [983089983093] is an example o this the authors use 1047297nite element modeling to construct adual rotor model Various types o rub-impact are then stud-ied Such studies can provide a wide range o inormationwhich can then be combined with data obtained rom livesystems potentially with seeded aults in order to constructaccurate remaining useul lie predictions As pointed out by the authors one key advantage o simulation is the ability to study more complex systems with a higher number o rotors which is used throughout this research Tis presentsa different approach to identiying eatures or identi1047297cationo rub and looseness

983090983092 Fluid-Induced Instability Fluid-induced instabilities(ofen reerred to as whip and whirl) are potentially very serious aults which can result in wear atigue and extensivedamage to machine components Such instabilities can beound in interstage seals 1047298uid lubricated bearings andblade-tip clearances Research into simulating and modeling1047298uid-induced instability has produced several works o interest to ault diagnosis o rotating machines in the lastew years De Castro et al [983089983094] are a good example wherenonlinear mathematical models are prepared or a rotor-bearing system Te models are then used to predictinstability thresholds Te authors consider a test case againsta power plant turbine and a test rig thereore validating the

simulations Te case o unbalance aults causing whip andwhirl phenomena is also considered Te main conclusionthereore drawn rom the work is that the authors concludedthat nonlinear hydrodynamic journal bearing models enablesufficiently accurate simulations or predicting instability thresholds Fan et al [983089983095] represent an example o work romthe perspective o aeroengine turbines In this case startupconditions are studied using a ull Hilbert spectrum Te aimbehind the paper is to accurately predict the point at whichwhip and whirl occur thus enabling this to be avoided at thedesign stage Such 1047297ndings could potentially also be used toidentiy whip and whirl as the case o a ault afer a period o wear in the operating machine

Prognosing 1047298uid-induced instability is a relatively lightly researched topic Fluid instabilities can be covered as part o extensive research into remaining useul lie o bearings Tepotential exists or modeling and simulation techniques suchas those detailed above to become a part o prognosis or1047298uid-induced instabilities due to the act that it can be very

difficult to seed such aults into live systems or testing andevaluation As with other aults detailed in this paper many studies have been perormed with the aim o describing 1047298uid-induced instabilities based on the measurement or simulationo single (or occasionally dual) rotor setups Physics-basedsimulation with the aim o localising 1047298uid instability aultsacross a whole system can be limited by the complexity o both the ault and the system hence the simpli1047297cation tosingle rotor-stator bearing systems

983090983093 Bearing Failure An area where data-driven techniquesare still providing the basis o much research in the 1047297eld o rotordynamics is that o bearing ailure Te title ldquobearingailurerdquo can cover a wide range o potential issues whichcontinue to be studied in detail Faults can occur in all kindso engine bearingsmdashthe inner andouter case the cage andtherolling elements 1047298uid-induced instabilities (addressed in aseparate section) lubrication and the complexities o activemagnetic bearings to name some examples All types o bearing relevant to rotating machinery are the subject o ongoing research and this subject has the potential to ormseveral separate papers As a brie highlight some recentexamples are discussed as ollows

Data-driven techniques have enabled accurate bearingdiagnostics and prognostics to be described or a rangeo rotordynamic systems Despite the prevalence o data-driven research in this area research rom a physics-basedsimulation perspective has also recently produced someinteresting papers o relevance to condition monitoring andhealth management o rotating machinery

Tis includes Kappaganthu and Nataraj [983089983096] in whichrolling element bearings have been studied through the useo nonlinear models Te included nonlinearity in this caseis clearance and the model is then used in order to study chaotic motions in particular the regions o chaotic responseTe research orms a part o an ongoing drive to develop anaccurate model-based diagnostic technique or rolling ele-ment bearings taking into account clearance nonlinearitiesand chaotic responses

Guptaetal[983089983097] demonstrate another example o thelatestresearch into instability and chaos in rolling element bearingsthrough high-1047297delity simulations Tis detailed and complexstudy involves the application o a novel scheme to analyzethe quasiperiodic response o the system combined with anonautonomous ldquoshootingrdquo method Tis workhighlights thelevel o detail to which nonlinearities and complex nonlinearmotions in bearings are beginning to be understood andaccurately modeled Again such work presents the potentialor design o IVHM in uture evolutions o the research

As so much research hasbeen perormed (and is ongoing)into bearing aults across a wide variety o mechanicalsystems both prognostics and localization o bearing aults

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 613

983094 International Journal o Rotating Machinery

have been researched in somewhat more detail than someo the other aults detailed here Despite this much work still needs to be perormed in order to translate some o this core research into industrial applications Research suchas that detailed above has made signi1047297cant advances intodetermining bearing ailure as the root cause o malunction

Detecting which bearing is ailing across a complex systemhas received somewhat less research Bearing prognostics isanother area with much ongoing research being perormedmdashboth in the simulation and data-driven domains o givean example Hong et al [983090983088] combined grade lie andextensive mathematical modeling techniques in order to pro-duce prognostic models or aeroengine bearings Te resultsare described by the authors as ldquopractical and veri1047297ablerdquoAlthough a number o similar recent studies existthis work iso note or the extent o the studies perormed which includebearing test stand run-to-ailure validation Te lab resultsappear impressive this research has yet to be applied andtested in real lie applicationsmdashindicating that despite thenumber o parameters considered it is still not possible tomodel naturally occurring phenomena sufficiently

A large body o work in this area also exists romBorghesani et al [983090983089] where several examples o data driventechniques can be seen applied to a wide variety o bearingtypes An example o recent developments involves the use o ldquocepstrum prewhiteningrdquo in order to remove sufficient noiseor accurate bearing diagnosis and prognosis Tis work isparticularly noteworthy due to the emphasis on real-worldapplications where the traditional lab-based techniques o order tracking and synchronous averaging do not providesufficient noise removal or harsh industrial environmentsTe addition o such techniques is a crucial step in developingthe current generation o diagnosis and prognosis algorithmsor use in uture IVHM systems

983090983094 Shaf Cracks Another potentially serious ault in rotat-ing machinery is shaf cracks and so early detection o any such ault is highly important Methods o crack ormationandpropagation canbe diverse andrange rom highand low-cycle atigue to stress corrosion Simulation and modelingo shaf cracks can have signi1047297cant advantages over data-driven methods Perhaps the most obvious advantage is therelative simplicity o inserting a ault into orexample a 1047297niteelement model as opposed to seeding a ault in a workingindustrial machine As such research into shaf cracks has

been progressing steadily with the corresponding increasesin computing power

A clear synergy between data-driven and physics-basedsimulation research can be implied by a number o recentworks o research An example o recent advances rom a datacollection perspective is Li et al [983090983090] which details statisticalmodels based on historical data or condition monitoringpurposes Tis unique work uses the human auditory systemas inspiration or enriching methods o mechanical aults andeatures extraction Te results indicated by the paper areperhaps not as convincing as some other methods discussedin this paper it describes an interesting ldquooutside the boxrdquomethod o tackling common problems

From a modeling perspective Bachschmid et al [983090983091]cover a wide range o vibration phenomena in order todevelop a model-based identi1047297cation and severity procedureTis work is noted or its thoroughness in modeling pro-cedure including accurate modeling o the crack breathingmechanism A combination o high and low 1047297delity models

is validated through experimental study and ldquoexcellentrdquoaccuracy is claimed by the authors in detecting crack positionand depth through the use o the proposed model-baseddiagnostics

Te nature o shaf cracks has resulted in a wide variety o research beingperormed intoboth localization and prognos-tics o these aults (indeed the two topics can be consideredrelated) Recent examples o work in this area include Inoueetal[983090983092] who detail crack localization using orced responsemodeling In this case a test rig was constructed consistingo a circular shaf supported by two bearings Frequency-domain data was used to create a localization algorithmdesigned in combination with an FE model Although thisresearch provides good accuracy o localization in a labenvironment it remains untested in a more complex system(eg a ull gas turbine)

Sonnichsen [983090983093] describes 1047297nite element modeling o crack propagation with validation against experimentalresults provided to demonstrate the validity o such modelingtechniques Te paper concentrates on natural requenciesand resonance curves Whilst the authors claim improve-ments in the ability to understand such aults again thesystem in question is quite simplemdashand such FEA models aredifficult to scale up to ull size applications

983090983095 Blade Cracks Blade cracks i allowed to develop canresult in serious consequences Cracks can orm due to highcentriugal stresses across operational cycles (in the case o an aircraf gas turbine or example start up and take off through landing and taxi) As excessive crack growth canlead to catastrophic rotorblade ailure early detection andprognosis o such aults are essential As with shaf cracksphysics-driven simulation o blade cracks is an area o sig-ni1047297cant research Tis varies rom high-1047297delity 1047297nite elementmodels to low-1047297delity system and mathematical models Terecent work demonstrated by Green and Casey [983090983094]is agoodexample o recent mathematical modeling rom a diagnosisperspective In this paper the authors concentrate on early detection using global and local asymmetry crack models 983090X

harmonic components are identi1047297ed as key areas or the early detection o blade cracks however again this paper suffersrom being applied and tested on a relatively simple systemwhich may not scale up to a ull size turbine

Sonnichsen [983090983093] demonstrated high-1047297delity modelingthe authors used FEA to model crack growth making com-parisons and validating against an experimental rig Tiswork is particularly interesting as it outlines the advantagesand drawbacks with the latest state-o-the-art modelingtechniques

Localization and prognosis o blade cracks have alsobene1047297tted rom recent advances in simulation and modelingSawicki et al [983090983095] contains details o work on a novel active

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 713

International Journal o Rotating Machinery 983095

magnetic bearing system or use in the early detection local-ization and prognosis o blade cracks Again the emphasis ison early detection with the bearings used to excite the systemin order to obtain optimum response vibrations or analysisTe authors admit the approach hassome merit in diagnosingblade cracks however it is in the early stages o development

and work are ongoing FEA has also been used extensively to support blade crack prognostic tools Xiang et al [983090983096] arean extensive example o recent work In this case a numbero advanced FEA methods are applied to produce accurateFEA solutionsmdashthese include surace-1047297tting techniques andthe contour-plotting method Te authors experimentally

validate their work and suggest that it can be applied toprognosis and quantitative diagnosis o blade cracks Despitethe claimed advances again the scalability o such research toull size turbines is an issuemdashparticularly with regard to thecomplexity o the FEA models

983090983096 Rotor Bow Rotor bows can be a primary source o unwanted vibration in gas turbines Te main cause o arotor bow (rotor bows do not include bows due to gravity) isthermal differences in a system caused by operating con-ditions It is noted by Domes [983093] that this nonsymmetricalthermal distribution can cause excessive unbalance to theextent where a gas turbine will not start correctly Such rotorbows are common on start up or shut down and are ofenaccounted or in operational procedures However i thermalldquohot spotsrdquo exceed a given tolerance level they can causepermanent unbalances due to rotor de1047298ections Such rotorbows can lead to other aults includingrubbing andloosenesswhich complicate isolation and localization

raditional data-driven techniques or detecting rotorbows involve combinations o slow roll and vibration data[983090983097] More recently mathematical modeling techniques suchas that detailed by Meagher et al [983091983088] have been used inorder to diagnose residual rotor bows and differentiate theseaults rom other sources o unbalance Te authors o thispaper build upon established methods or the models and areunique in that they concentrate on response at the bearingpoints Tis is perhaps more useul orindustrial applicationsas vibration inormation is more readily available at bearingpoints rather than intrusive proximity probes which are ofenused in lab work

Te little work that exists on attempting to localise rotorbows across complex systems tends to be data-driven in

nature see Gaka and abaszewski [983091983089] where the authorsused statistical symptoms based on known data as a methodo diagnosing and prognosing a number o aults includingrotor bows and unbalance Tis paper attempts to address theirregularities and 1047298uctuations that occur over a long servicelie In order to achieve this a modi1047297ed energy processormodel is created using data drawn rom large steam turbinesover long periods o lie

Prognosing rotor bows is a complex subject As rotorbows are ofen caused by temperature de1047298ections makingpredictions or remaining useul lie and potential utureproblems lies not only in the realm o mechanical rotor-dynamics but also to some extent in thermodynamics Te

recent work detailed by Sinha [983091983090] is o note or detailingdiagnosis and quanti1047297cation o various rotordynamic aultsand describing the advantages o mathematical modelingover traditional vibration-based approaches Te topic o scalability regarding FEA models is discussed including anargument or the use o partial (simpli1047297ed) mathematical

models or large complex systemsAnother two works which are o interest with regard to

modeling o rotor bows include Shen et al [983091983091] where theauthors modeled a rotor-bearing system with a permanentrotor bow looking at the impact o secondary aults such asrub Te study o ault combinations in this paper is useulor ault differentiation studies however the authors study a permanent initial rotor bow Tis thereore does not takeinto account developing or worsening aults and the different

vibration phenomena that are observed as such aults aredeveloping

Lees et al [983091983092] describe the importance o model-basedault identi1047297cation techniques and outline recent research in

the area providing a good reerence paper or more researchon this speci1047297c ault

3 Discussion

983091983089 Sensor Suites Te subject o sensor suites is o greatimportance with regard to uture industrial applications Onone hand advanced and complex sensor suites generally enable improved ault localization and diagnosis howeverthe added complexity and cost has resulted in many o thesesystems being omitted rom the latest generations o rotatingmachinery in industrial applications From an industrialperspective the cost bene1047297t o additional sensors needs to

be signi1047297cant in order to justiy this approach In response anumber o works listed including [983096] as a good example takethe approach o achieving the same objectives using a greatly reduced sensor suite In many ways the required sensor suiteor many o the currently researched ault diagnosis andprognosis techniques provides an indication o the ability o the techniques to be used practically in industrial situationsA complex suite may return a very high success rate on alab-based rig however the impracticalities o mounting suchsuites on a real-world application negate the advantages

Further to this the consideration orsensor position mustbe considered Te use o proximity probes and keyphasortransducers has very clear advantages in a number o situ-

ations however the intrusive nature o the sensor preventsapplication in a number o complex systems With this inmind a number o the more promising studies listed rely onthe simple suiteo an accelerometer placed in positions whichare relatively remote rom the sources o vibration potentially with a noisy transer path

983091983090 Diagnosis It can be seen rom the research outlined inthe previous section that the diagnosis o aults in rotatingmachinery is a subject o much ongoing research Tisinvolves the improvement and development o traditional

vibration monitoring techniques development o new data-driven technologies and novel research into physics-based

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 813

983096 International Journal o Rotating Machinery

simulation and modeling In many cases these topics o research are dependent on one another or reasons o vali-dation veri1047297cation and speed o analysis In several casesit can be seen that multiple aults have been modeled orthe purposes o identi1047297cation and isolation However nostudies have yet been perormed which deal with all o the

aorementioned aults All o these aults are intrinsically related to one another Complex combinations o aults havebegun to be analyzed with the emphasis on developing new diagnosis techniques Physics-based modeling has proven toprovide signi1047297cant advances with regard to speci1047297c aultsnotably shaf and blade cracks where techniques such as FEAenable much easier aster and cheaper test data then seedingaults into live systems

It can be noted however that both high and low-1047297delity modeling techniques are being applied to cutting edgeresearch or all o the listed rotordynamic aults (and othersnot detailed in this paper) In addition to the advantages inthe speed o obtaining test results physics-based simulationis providing another dimension to data-driven techniquesSystem models are being used as part o logic and reasoningsuites in the identi1047297cation and differentiation o variousaults High-1047297delity models enable simulations o ault com-binations which are not possible practically or are pro-hibitively expensive in live systems

Nevertheless results obtained rom modeling studies stillneed to be validated against proven data-driven techniquesbeore implementation in industrial applications is possibleIt is also o note rom the literature reviewed or this paperthat almost all modeling or the diagnosis o aults involvesextensive simpli1047297cation ofen reducing potentially complexsystems down to one or two discshafbearing modelsAdapting the claimed results rom such research to systemswith many discsbearingsshafs is important in improvingexisting rotating machinery diagnostic techniques

Te inclusion o nonlinear effects is notable amongstmany authorsmdashas the understanding o the complex aultsand interactions continues to progress Such research opensthe potential or promising new avenues into areas o designor IVHM

Finally it is o note that despite the wide variety o advanced research here a consensus or the most efficienttechniques and algorithms has yet to emerge in order or the1047297nal adaption to industrial application to be achieved How-ever the consideration or industrial application taken by many o the authors detailed in this paper provides promise

or next generation IVHM systems

983091983091 Localisation Fault localization in rotating machinery isan important topic o research or uture condition-basedmonitoring systems Knowing not only what type o aulthas occurred but also where in the system is an importantconsideration which can in1047298uence maintenance proceduresin complex machinery It is worth noting rom the literaturesurveyed or this paper that many studies have ocusedon diagnosis and prognosis o single rotorbearing systemsOfen or legitimate reasonsmdashsimpli1047297cation or computingspeed or example Few studies however have taken intoaccount the localization o aultsacross whole systems Oneo

the more extensive examples highlighted in this paper (Hanet al [983089983093]) provides extensive analytical studies relevant tolocalization on a duel disk setup utilizing Hilbert-Huangransorms however in this example comprehensive valida-tion is lacking thus limiting the potential applicability o theresearch

Tis problem is not limited to modeling and simulation-based research Many newly developed data-driven tech-niques or diagnostics and prognostics claim good results by heavily instrumenting speci1047297c components o a test systemIn many industrial cases this is not possible practical or cost-effective

o give some examples a keyphasor transducer can beparticularly useul in diagnosing aults such as rotor bowHowever this equipment requires the ability to cut a key way or measurements to be perormed Optical sensors haverecently been applied to detect rotor unbalance yet such atechnology would be difficult to implement in a system withseveral rows o rotors (as in a gas turbine)

Whilst new diagnosis and prognosis techniques or aultsin rotating machinery are being continuously researchedthe lack o corresponding studies into localization can beconsidered one o the many challenges in promoting recentcore research into live industrial applications

983091983092 Fault Differentiation One promising development inrecent modeling and simulation o rotordynamic aults isthat o ault differentiation Several researchers have movedon rom studies on individual aults in order to concentrateon combinations o aults As stated rotordynamic aultssuch as those listed in this paper are linked to each otherSuch studies thereore concentrate on such topics as a

misalignment causing an unbalance or a looseness causinga rub Findings rom reports such as these are important inunderstanding complex anomalies In industrial applicationssimply detecting and rectiying an unbalance does not pro-

vide a satisactory solution i the root cause o the ault isa misalignment Tis is a complex topic as several aults canexhibit similar vibration characteristics making traditionaldetection techniques inaccurate in some cases

Making reerence to some o the aorementioned stud-ies those such as Bachschmid et al [983090983091] and Sonnichsen[983090983093] indicate some initial studies including several aultsTe results o such studies indicate a high success rate indifferentiating the aults in the given conditions however

a number o limitations still exist Studies such as theseconsider unbalancein the aultchains Tisis typically ldquostaticrdquoor simple unbalance applied to a single shafrotormdashin reality unbalance aults can be more complex than this In additionthe simple systems used to study and differentiate aultsare very different rom the complexities o a ull turbineHowever urther work may enable such research to ldquoscale-uprdquo to this stage with the aid o urther validation and

veri1047297cation

983091983093 Prognosis It can be seen rom the examples o recentresearch described or common aults that prognostic tech-niques are a topic where much work is being perormed

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 913

International Journal o Rotating Machinery 983097

Some o this work would be very difficult or impossibleto implement in an industrial situation others providepromising results o use to uture work in the area Predictingthe remaining useul lie o components is critical in thedevelopment o condition-based monitoring strategies orindustrial implementation

It can be noted that prognostic studies or certain rotor-dynamic aults are considerably more advanced than orothers Te most obvious examples o the aults detailed inthis paper are bearing aults shaf and blade cracks Teseare also areas where physics-driven simulations have hadan important impact Te ability to design any ault type(or combinations o aults) into such simulations (to be itslow-1047297delity mathematical models or high-1047297delity 1047297nite ele-ment analysis) has provided researchers with many differentavenues to explore Studies into the prognostics o otheraults can be complicated by various actors An unbalanceor example has had relatively little research perormed intoprognosis Te act that unbalance is ofen the cause o another underlying ault is one reason or the difficulty inresearching prognosis in detail or this ault Te results o unbalance are also very dependent on the severity o the aultand the system in which it occurs

It is worth noting that even or the aults where prognosisresearch is more advanced (eg shaf cracks) true develop-ment o the ault is not always studied Fan et al [983089983095] orexample study sets o discrete cracks in order to ldquoprognoserdquocrack growth In act thisstudy is more akinto detecting aultseverity with inerred effects on crack growth In contrast Liet al [983090983090] consider the continual development o such crackshowever the authors do admit this work is in its inancySuch points limit the ability or true prognosis to be achievedwithout extensive operational data

It is also worth noting that combinations o aults havebeen simulated extensively or the purpose o diagnosticshowever ew studies exist combining aultsor the purpose o prognostics Such studies are however a logical progressionrom some o those already perormed It may be that suchprognostic studies build upon some o the condition-basedmaintenance (CBM) or diagnostics High-level studies suchas Jaw and Merrill [983091983093] indicate how prognosis and diagnosistechniques can be combined into a CBM system Tis study is designed or a military aeroengine indicating the desire orsuch systems to be implementedmdashalthough the architecturedescribed has the capability o including various algorithmsoraultprognosis the amount o consideration orprognosis

o ault combinations is unclear

983091983094 Modelling Te large number o works reviewed or thispaper which includes modeling techniques in order to aiddiagnosis and prognosis indicates the possibilities providedby modern computer power and sofware developmentBroadly the modeling studied in the aorementioned papersbreaks down into two categories mathematical modeling and1047297nite element analysis (FEA) Te mathematical models suchasBorghesani etal [983090983089] ofen provide a theoretical basis uponwhich data driven or FEA studies can build and validateMathematical models despite being a traditional approach to

rotordynamic problems thereore continue to be developedto tackle more advanced problems

FEA techniques offer a constantly expanding area o simulation to explore Tey are used or research as diverseas crack propagation to unbalance localization [983091983094] Moderncodes such as NASRAN and Ansys enable rotordynamics to

be studied beyond the traditional stress and modal analysisTe ability to study the effects o rotation (eg throughCampbell diagrams and transient analysis) continues to driveresearch and innovation in this area Te studies discussedin this paper indicate the power o FEA with some highly accurate simulations having been perormed However thereis still a limitation o computer powermdashapplying some o the FEA techniques to whole engine models would proveto be too computationally expensive to be viable As aresult alternate methods such as model order reductionand system level modeling (and model-based reasoning orimplementation) are still required in order to make some o the FEA studies viable in an industrial environment

Nonlinearities continue to orm an important part o many recent studies with the complexity and detail o thespeci1047297ed aults being continually expanded Te depth o modeling has enabled in many cases accurate validationagainst experimental approaches Despite this completeunderstanding o the vibrational phenomena o rotatingmachinery is not currently possible despite the drive towardsthis end

983091983095 Rotordynamics and IVHM Te bulk o current researchinto rotordynamics rom the point o view o prognostichealth management (PHM) can be roughly divided intotwo types initial single-ault diagnosisprognosis techniquesand studies into the general requirements and limitationso current systems along with current and uture trendsAn example o the latter is Pusey [983091983095] who provides agood summary overview o current diagnosis and prognosistechniques with regard to condition-based maintenance

As a result o this split a clear gap exists between thecore research being perormed into rotordynamics rom acondition-based maintenance perspective and the identi1047297edneeds o industry aking a 1047298edgling piece o research andapplying it to a commercially-ready system (eg a gas turbineengine or an aircraf) is a long and complex task It isnevertheless worth noting that technologies or automatically detecting an unbalance or misalignment in a gas turbine were

developed over 983089983088 years beore the latest commercial aircrafwere conceptualized and yet these planes are still limited inthis capacity Tis highlights the need or work which linksthe undamental research into individual ault diagnosis toldquolive systemsrdquo in use in industry

Physics-based simulation and modeling o rotordynamicparts is a well-researched 1047297eld Such modeling has been usedas the basis o diagnosis and prognosis o aults by many researchers several recent examples have been outlined inthis paper Occasional pieces o work have been perormedinto modeling multiple aults such as Jain and Kundra [983091983096]who usea system model oronline identi1047297cation o unbalanceand cracks Beyond this however very limited research

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 1013

983089983088 International Journal o Rotating Machinery

Business casecost bene1047297tanalysis

Live system

RubCracksBearingfailure

Rotor bow Unbalance

Fault differentiationLocalisation

Root cause analysis

Remaining useful life

Data driven analysis and physics-based simulation

Mis-alignment

Experimental

1047297eld data

F983145983143983157983154983141 983089 Physics-based simulationmdashrom research to industry

exists Te demands o PHM techniques in industry are suchthat any systemmust notonly be capable o detecting multipleaults but must also be capable o detecting these aultsacrossa range o different systems Other considerations includethe aorementioned ability to differentiate between multipleaults Processing also needs to be taken into account as theobjective o these systems is to implement efficient conditionmonitoring and condition-based maintenance procedures I processing data is a long power-hungry process then thisaim cannot be achieved

Figure 983089 details a potential ramework required in order

to push core research such as that detailed in this reporttowards industrial applications Many studies now exist onindividual rotordynamic aults across a wide range o condi-tions and applications Some studies have taken this urtherwith advanced prognostic models and diagnosis o dual aults(primary cause and secondary effect) Future research inthe area o rotordynamics rom a PHM perspective couldpotentially provide the bridge between these studies and livesystems by combining physics-based simulations with data-driven techniques and validating against experimental data

Although quantiying the success o the research studiedisdifficultit ispossibleto de1047297ne the key areas in whicha tech-nique must excel in order to be considered viable Te work

byWheeleretal[983091983097] discussesin detailmetrics ordiagnosticand prognostic analysis as does that by Vachtsevanos [983092983088]and Saxena et al [983092983089] Te conclusions drawn rom thesepapers and applied in practice to research like that covered by this paper indicate that the ollowing metrics are importantwhen considering the potential o a given technique to diag-nose aults coverage alse positive rate and alse negativerate In the case o prognosis probabilities and lead timeto ailure are other important considerations Tis criteriaenables research to be assessed in terms o its suitability orindustrial applications Unortunately inormation on these

metrics is not made readily available by the authors o mostpapers

In terms o evaluating the effectivenesso theresearch dis-cussed in this paper there are difficulties in recommending aspeci1047297c technique over others or general application Mostpapers reviewed or this document (not just those reerencedand discussed in detail) take a technique (new or evolved)

validate or a given speci1047297c system and report success o theresearch Despite this some conclusions can be drawn romassessing common techniques applied across different studiesand different aults Although it is not possible to de1047297ne themost common methods or diagnosing andprognosing aultsin terms o numbers (as it was not possible to cover all recent

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 1113

International Journal o Rotating Machinery 983089983089

rotordynamics research or this paper) as perceived by theauthors the ollowing techniques have eatured prominently in the reviewed research

(i) Sensors accelerometers

(ii) Teoretical studies mathematical modeling

(iii) Physics-based simulation FEA(iv) Data-driven joint timerequency domain analysis

(v) Implementation neural networking

Tese techniques appear to be among the most promisingcurrently under development as they tend to eature numer-ous timesamongst some o the work with wider scope acrossault types and with the most comprehensive validationTere are o course many subsections to these techniqueshowever it shows one general direction o research and theclear possibilities posed in these areas

Future developments in the 1047297eld o IVHM or rotat-ing machinery may incorporate these techniques alongsideextensive use o nonlinear modeling and multiple aultinteractions Te 1047297eld o design or IVHM has only recently emerged however the potential exists or speci1047297c nonlinear-ities to be designed into a system in order to enable accuratediagnosis andprognosis o aults Te development o currentalgorithms to include diagnosis localization and prognosiso a range o aults will provide a signi1047297cant advancement oruture generations o IVHM systems Tis combined withcost-effective sensor suites indicates the potential or evolu-tions o some o the research detailed hereto orm part onextgeneration IVHM suites or rotating machinery

4 Conclusion

Tis paper has reviewed some o the latest research around anumber o rotordynamic aults namely unbalance misalign-ment rub and looseness 1047298uid-induced instability bearingaults shaf cracks blade cracks and rotor bow Each aultwas reviewed rom the perspective o sensors diagnosisprognosis localization and modeling

Key examples o recent work into the eight describedaults have been detailed through works by a number o eminent authors Additional work has been summarized andormatted or easy reerence Some current trends amongstthe recent body o work include developments in the vastarea o modeling nonlinearities combinations o high andlow 1047297delity modeling and synergy between data driven and

physics-based simulation approachesDespite the large volume o promising research reviewed

urther development in a number o areas is required in orderto produce effective next generation IVHM systems As suchuture developments may include usion o andor devel-opment o current algorithms to encompass all eight aultsdetailed consideration o prognosis diagnosis and localiza-tion achieved using a reduced cost-effective sensor suite

References

[983089] I K Jennions Integrated Vehicle Health Management Perspec-tives on an Emerging Field 983090983088983089983089

[983090] G Anderson ldquoProviding best-value IVHM solutions or agingaircrafrdquo in Proceedings o the 983097th Joint FAADoDNASA Aging Aircraf Conerence vol 983089 pp 983089ndash983089983089 983090983088983088983093

[983091] A Muszynska Rotordynamics aylor amp Francis 983089st ed edition983090983088983088983093

[983092] D E Bently Fundimentals o Rotating Machinery Diagnostics

ASME Press 983089st edition 983090983088983088983090[983093] B Domes ldquoVibration phenomena in aero-enginesrdquo in Proceed-ings o the 983097th International Conerence on Vibrations in Rotating Machinery Exeter (IMechE rsquo983088983096) vol 983089 pp 983089983093ndash983091983090 September983090983088983088983096

[983094] A Hess G Calvello and Dabney ldquoPHM a keyenabler or theJSF autonomie logistics support conceptrdquo in Proceedings o theIEEE Aerospace Conerence Proceedings pp 983091983093983092983091ndash983092983093983092983097 March983090983088983088983092

[983095] S N Ganeriwala B Schwarz and M H Richardson ldquoOperat-ing de1047298ection shapes detect unbalance in rotating equipmentrdquoSound and Vibration vol 983092983091 no 983093 pp 983089983089ndash983089983091 983090983088983088983097

[983096] G N D S Sudhakar and A S Sekhar ldquoIdenti1047297cation o unbal-ance in a rotor bearing systemrdquo Journal o Sound and Vibration

vol 983091983091983088 no 983089983088 pp 983090983090983097983097ndash983090983091983089983091 983090983088983089983089[983097] Yang and M W Hsu ldquoAn efficient diagnosis technique or

variations o shaf-bow and unbalancerdquo in Proceedings o the ASME International Design Engineering echnical Conerencesand Computers and Inormation in Engineering Conerence(DEC rsquo983088983097) pp 983093983095ndash983094983094 San Diego Cali USA September 983090983088983088983097

[983089983088] A El-Shaei S H aw1047297ck and M O A Mokhtar ldquoExperi-mental investigation o the effect o angular misalignemt onthe instability o plain journal bearingsrdquo in Proceedings o the ASMESLE InternationalJointribology Conerence(IJC rsquo983088983097)pp 983089983095983089ndash983089983095983091 Memphis enn USA October 983090983088983088983097

[983089983089] A El-Shaei S H aw1047297ck and M O A Mokhtar ldquoMisalign-ment modeling in rotating systemsrdquo in Proceedings o the ASMEurbo Expo H Bahaloo A Ebrahimi and M Samadi Eds vol

983089983095983089983089 pp 983097983095983091ndash983097983095983097 Orlando Fla USA June 983090983088983088983097

[983089983090] L F Villaet al ldquoStatistical diagnosisbased on vibration analysisor gear test-bench under non-stationary conditions o speedand loadrdquo Mechanical Systems and Signal Processing vol 983090983097 pp983092983091983094ndash983092983092983094 983090983088983089983090

[983089983091] C F Ngolah et al ldquoIntelligent ault recognition and diagnosisor rotating machines using neural networksrdquo Sofware Scienceand Computational Intelligence vol 983091 no 983092 983089983095 pages 983090983088983089983089

[983089983092] Y Lu Z Ren H Chen N Song and B Wen ldquoStudy on loose-ness and impactmdashrub coupling aults o a vertical dual-disk cantilever rotor- Bearing systemrdquo Key Engineering Materials vol 983091983093983091ndash983091983093983096 no 983092 pp 983090983092983095983097ndash983090983092983096983090 983090983088983088983095

[983089983093] Q Han Z Zhang and B Wen ldquoPeriodic motions o a dual-

disc rotor system with rub-impact at 1047297xed limiterrdquo Proceedingso the Institution o Mechanical Engineers vol 983090983090983090 no 983089983088 pp983089983097983091983093ndash983089983097983092983094 983090983088983088983096

[983089983094] H F de Castro K L Cavalca and R Nordmann ldquoWhirland whip instabilities in rotor-bearing system considering anonlinear orce modelrdquo Journal o Sound and Vibration vol 983091983089983095no 983089-983090 pp 983090983095983091ndash983090983097983091 983090983088983088983096

[983089983095] C C Fan et al ldquoMechanical systems and signal processingrdquoStudy o Start-Up Vibration Response or Oil Whirl and Dry Whip vol 983090983093 pp 983091983089983088983090ndash983091983089983089983093 983090983088983089983089

[983089983096] K Kappaganthu and C Nataraj ldquoNonlinear modeling andanalysis o a rolling element bearing with a clearancerdquo Commu-nications in Nonlinear Science and Numerical Simulation vol983089983094no 983089983088 pp 983092983089983091983092ndash983092983089983092983093 983090983088983089983089

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 1213

983089983090 International Journal o Rotating Machinery

[983089983097] C Gupta K Gupta and D K Sehgal ldquoInstability and chaoso a 1047298exible rotor ball bearing system an investigation on thein1047298uence o rotating imbalance and bearing clearancerdquo Journal o Engineering or Gas urbines and Power vol 983089983091983091 no 983096ArticleID 983088983096983090983093983088983089 983089983089 pages 983090983088983089983089

[983090983088] J Hong X Miao L Han and Y Ma ldquoPrognostics model orpredicting aero-engine bearing grade-lierdquo in Proceedings o the ASME urbo Expo vol 983089 pp 983094983091983097ndash983094983092983095 Orlando Fla USA June983090983088983088983097

[983090983089] P Borghesani P Pennacchi R B Randall N Sawalhi and RRicci ldquoApplication o cepstrum pre-whitening or the diagnosiso bearing aults under variable speed conditionsrdquo Mechanical Systems and Signal Processing 983090983088983089983090

[983090983090] Y Li J Zhang L Dai Z Zhang and J Liu ldquoAuditory-model-based eature extraction method or mechanical aults diagno-sisrdquo ChineseJournal o MechanicalEngineering (EnglishEdition) vol 983090983091 no 983091 pp 983091983097983089ndash983091983097983095 983090983088983089983088

[983090983091] N Bachschmid P Pennacchi and E anzi ldquoCracked rotatingshafs typical behaviors modeling and diagnosisrdquo IUAM Symposium on Emerging rends in Rotor Dynamics vol 983089983088983089983089 pp

983092983092983089ndash983092983093983092 983090983088983089983089[983090983092] Inoue N Nagata and Y Ishida ldquoFem modelling and

experimental veri1047297cation o a rotor system with a open crackrdquoin Proceedings o the ASME International Design Engineering echnical Conerences and Computers and Inormation in Engi-neering Conerence (DEC rsquo983088983097) pp 983089983089983089983091ndash983089983089983090983090 San Diego CaliUSA September 983090983088983088983097

[983090983093] H E Sonnichsen ldquoReal-time detection o developing cracksin jet engine rotorsrdquo in Proceedings o the IEEE AerospaceConerence pp 983089983095983091ndash983089983096983092 March 983090983088983088983088

[983090983094] I Green and C Casey ldquoCrack detection in a rotor dynamicsystem by vibration monitoringmdashpart I analysisrdquo Journal o Engineering or Gas urbines and Power vol 983089983090983095 no 983090 pp 983092983090983093ndash983092983091983094 983090983088983088983093

[983090983095] J Sawicki M I Friswell A H Pesch and A WroblewskildquoCondition monitoring o rotor usingactive magnetic actuatorrdquoin Proceedings o the ASME urbo Expo pp 983089983090983093983095ndash983089983090983094983093 BerlinGermany June 983090983088983088983096

[983090983096] J Xiang X Chen Q Mo and Z He ldquoIdenti1047297cation o crack ina rotor system based on wavelet 1047297nite element methodrdquo FiniteElements in Analysis and Design vol 983092983091 no 983089983092 pp 983089983088983094983096ndash983089983088983096983089983090983088983088983095

[983090983097] M G MaalouldquoSlow speed vibrationsignal analysisi you canrsquotdo it slow you canrsquot do it astrdquo in Proceedings o the ASME urboExpo pp 983093983093983097ndash983093983094983095 Montreal Canada May 983090983088983088983095

[983091983088] J Meagher X Wu and C Lencioni ldquoResponse o a warped1047298exible rotor with a 1047298uid bearingrdquo International Journal o

Rotating Machinery vol 983090983088983088983096 Article ID 983089983092983095983094983093983091 983097 pages 983090983088983088983096

[983091983089] Gaka andM abaszewski ldquoAn applicationo statistical symp-toms in machine condition diagnosticsrdquo Mechanical Systemsand Signal Processing vol 983090983093 no 983089 pp 983090983093983091ndash983090983094983093 983090983088983089983089

[983091983090] J K Sinha ldquoRecent trands in ault quanti1047297cation in rotatingmachinesrdquo Advances in Vibration EngIneerIng vol 983096 no 983089 pp983095983097ndash983096983093 983090983088983088983097

[983091983091] X Shen JJia andM Zhao ldquoNonlinear analysis o a rub-impactrotor-bearing system with initial permanent rotor bowrdquo Archiveo Applied Mechanics vol 983095983096 no 983091 pp 983090983090983093ndash983090983092983088 983090983088983088983096

[983091983092] A W Lees J K Sinha and M I Friswell ldquoModel-based identi-1047297cation o rotating machinesrdquo Mechanical Systems and Signal Processing vol 983090983091 no 983094 pp 983089983096983096983092ndash983089983096983097983091 983090983088983088983097

[983091983093] L C Jaw and W Merrill ldquoCBM+ research environmentmdashacilitating technology development experimentation andmaturationrdquo in Proceedings o the IEEE Aerospace Conerence(AC rsquo983088983096) March 983090983088983088983096

[983091983094] R Walker S Perinpanayagam and I K Jennions ldquoSimulatingUnbalance or Future IVHM Applicationsrdquo in Proceedings o theSociety or Experimental Mechanics (IMAC rsquo983089983090) vol 983090983097 pp 983089983092983089ndash983089983092983096 May 983090983088983089983090

[983091983095] H C Pusey ldquourbomachinery condition monitoring and ail-ure prognosisrdquo Sound and Vibration vol 983092983089 no 983091 pp 983089983088ndash983089983093983090983088983088983095

[983091983096] J R Jain and K Kundra ldquoModel based online diagnosiso unbalance and transverse atigue crack in rotor systemsrdquo Mechanics Research Communications vol 983091983089 no 983093 pp 983093983093983095ndash983093983094983096983090983088983088983092

[983091983097] K R Wheeler Kurtoglu and S D Poll ldquoA survey o healthmanagement user objectives related to diagnostic and prognos-tic metricsrdquo in Proceedings o the 983090983097th Computers and Inor-mation Engineering Conerence vol 983090 pp 983089983090983096983095ndash983089983090983097983096 983090983088983088983097

[983092983088] G Vachtsevanos ldquoPerormance metrics or ault prognosis o

complex systemsrdquo in IEEE Systems Readiness echnology Con- erence (AUOESCON rsquo983088983091) pp 983091983092983089ndash983091983092983093 September 983090983088983088983091

[983092983089] A Saxena J Celaya E Balaban et al ldquoMetrics or evaluatingperormance o prognostic techniquesrdquo in Proceedings o theInternational Conerence on Prognostics and Health Manage-ment (PHM rsquo983088983096) pp 983089ndash983089983095 October 983090983088983088983096

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 1313

Submit your manuscripts at

httpwwwhindawicom

Page 4: Rotordynamic Faults 2013

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 413

983092 International Journal o Rotating Machinery

983137983138983148983141 983089 Rotordynamic ldquoault tablerdquo

Faults Sensors Fault identi1047297cation

techniques Localisation o ault Prognosis

Work o note in aultmodelling

UnbalanceAccelerometer

Velocity transducer

Proximity transducer

Operating de1047298ectionshapesime-requency

analysis and many others

Neural-network modelling and

time-requency analysis

FCM-Markov model

Unbalance in ull-engineFEA modelEarly detection inaeroengines

High radialloads andmisalignment

AccelerometerTermocoupleDisplacement

transducerAngular sensor

Bearing temperatureobservation advancedSofware-driven1047297ltering methods

Various vibrationanalysis techniques

Prognosticmodelling usingsimulated vibrationdata

Modelling therelationship betweenmisalignment and otheraultsHarmonic balancemethod

Rub andlooseness

AccelerometerVelocity transducer

Optical sensor

Single valuedecompositionime-requency analysis

Least-squares 1047297ttingapproach

Adaptive waveletdenoising orprognostics

Rob and looseness usingFEA models Wholeengine modellingtechniques

Fluid-inducedinstability

Proximity probesKeyphase transducer

Full spectrum plotsLarge amplitudesubsynchronous vibrationDetection

Blade tip clearanceand loadingdistribution

Cross-coupledstiffness anddirect-dampingstudies

Nonlinear modellingaccuracy o model basedDiagnostics numericalmodelling

Bearing ailure

AccelerometerVelocity transducer

Displacementtransducer

Proximity probes

Postailure analysis oruture ailure diagnosisDynamictime-requency analysis

Statistical change inthe bispectraldomain

Numericalmodelling o aeroengine bearingsModel-basedprediction o remaining useul lieOn-boardprognostics

High and low 1047297delity dynamic systemmodellingComparisons o algorithms tounderstand bearing vibration phenomena

Shaf cracks

Keyphase transducer

Proximity probes

Higher order spectra

Neural networksGrid support vectors

Forced response

measurements

Nonintrusive

torsion vibrationmonitoring

Statistical analysis FEA

modelling

Blade cracks Keyphase transducer

Proximity probes

Hilbert-HuangtransormationModel-baseddiagnosticsMorletwaveletquantumtechniques

Novel bearingstructures

Active magneticactuator

Global and localasymmetry modellingFinite element modelling

Rotor Bow Keyphase transducer

Proximity probesSeismic transducer

Examination o ldquoslow rollrdquo dataModel-baseddiagnostics

Partial mathematicalmodelling

Statistical symptomsanalysis

Advantages o aultmodelling oridenti1047297cationNonlinear modellingtechniques

through the use o a statistical diagnosis algorithm based onthe signi1047297cance level o the aults in question

983090983091 Rub and Looseness Rub is always a secondary ault (iea product o another ault such as looseness) and can leadto atigue and wear Rub and looseness can create com-plex vibration signals which are difficult to diagnose usingtraditional methods Modeling and simulation o rub andlooseness aults have been considered in several recentworks

Tis includes Ngolah et al [983089983091] which detail the moni-toring and diagnosis o common aults (including rub and

looseness) based upon a three-layer Arti1047297cial Neural Net-work (ANN) A series o 983089983088 key perormance indicators wereidenti1047297ed and used as training Te authors test the systemin a lab environment but stress the applicability to industrialapplications Te research indicates one o the latest methodso research which enables the implementation o diagnosistechniques It is a useul tool or rub and looseness studiesas it incorporates a variety o aults which could ldquounderlierdquosuch a ault Despite this the research relies on clear eaturesor each ault which can be much easier to identiy in a labenvironment as opposed to ldquonoisyrdquo industrial applications

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 513

International Journal o Rotating Machinery 983093

Lu et al [983089983092] have perormed several studies into rub andloosenessincluding thisexamplein which a 1047298exibly mountedshaf has an induced rub due to a range o contact ringsTe study ocusses on the potentially dangerous effects o rubin causing excessive nonsynchronous and chaotic vibrationsTe links with unbalance and misalignment are discussed

and detailed Te considerations or real-world cases in thedescribed research are considerable as part o the drivetowards ull understanding o the nonlinear effects o rub andlooseness

Localization o rub and looseness across whole systemsis relatively lightly studied in the literature Many works(including those already cited) look at single or dual-rotorsystems where localization o such aults is not an issue Inan industrial setting complex systems may comprise many rotors in several compressor and turbine stages signi1047297cantly complicating diagnosis o such aults Research into progno-sis o rotor-stator rubs lies mostly within the domain o data-driven techniques Modeling and simulation research can beused to support data-driven techniques or prognosis andcondition-based monitoring Han et al [983089983093] is an example o this the authors use 1047297nite element modeling to construct adual rotor model Various types o rub-impact are then stud-ied Such studies can provide a wide range o inormationwhich can then be combined with data obtained rom livesystems potentially with seeded aults in order to constructaccurate remaining useul lie predictions As pointed out by the authors one key advantage o simulation is the ability to study more complex systems with a higher number o rotors which is used throughout this research Tis presentsa different approach to identiying eatures or identi1047297cationo rub and looseness

983090983092 Fluid-Induced Instability Fluid-induced instabilities(ofen reerred to as whip and whirl) are potentially very serious aults which can result in wear atigue and extensivedamage to machine components Such instabilities can beound in interstage seals 1047298uid lubricated bearings andblade-tip clearances Research into simulating and modeling1047298uid-induced instability has produced several works o interest to ault diagnosis o rotating machines in the lastew years De Castro et al [983089983094] are a good example wherenonlinear mathematical models are prepared or a rotor-bearing system Te models are then used to predictinstability thresholds Te authors consider a test case againsta power plant turbine and a test rig thereore validating the

simulations Te case o unbalance aults causing whip andwhirl phenomena is also considered Te main conclusionthereore drawn rom the work is that the authors concludedthat nonlinear hydrodynamic journal bearing models enablesufficiently accurate simulations or predicting instability thresholds Fan et al [983089983095] represent an example o work romthe perspective o aeroengine turbines In this case startupconditions are studied using a ull Hilbert spectrum Te aimbehind the paper is to accurately predict the point at whichwhip and whirl occur thus enabling this to be avoided at thedesign stage Such 1047297ndings could potentially also be used toidentiy whip and whirl as the case o a ault afer a period o wear in the operating machine

Prognosing 1047298uid-induced instability is a relatively lightly researched topic Fluid instabilities can be covered as part o extensive research into remaining useul lie o bearings Tepotential exists or modeling and simulation techniques suchas those detailed above to become a part o prognosis or1047298uid-induced instabilities due to the act that it can be very

difficult to seed such aults into live systems or testing andevaluation As with other aults detailed in this paper many studies have been perormed with the aim o describing 1047298uid-induced instabilities based on the measurement or simulationo single (or occasionally dual) rotor setups Physics-basedsimulation with the aim o localising 1047298uid instability aultsacross a whole system can be limited by the complexity o both the ault and the system hence the simpli1047297cation tosingle rotor-stator bearing systems

983090983093 Bearing Failure An area where data-driven techniquesare still providing the basis o much research in the 1047297eld o rotordynamics is that o bearing ailure Te title ldquobearingailurerdquo can cover a wide range o potential issues whichcontinue to be studied in detail Faults can occur in all kindso engine bearingsmdashthe inner andouter case the cage andtherolling elements 1047298uid-induced instabilities (addressed in aseparate section) lubrication and the complexities o activemagnetic bearings to name some examples All types o bearing relevant to rotating machinery are the subject o ongoing research and this subject has the potential to ormseveral separate papers As a brie highlight some recentexamples are discussed as ollows

Data-driven techniques have enabled accurate bearingdiagnostics and prognostics to be described or a rangeo rotordynamic systems Despite the prevalence o data-driven research in this area research rom a physics-basedsimulation perspective has also recently produced someinteresting papers o relevance to condition monitoring andhealth management o rotating machinery

Tis includes Kappaganthu and Nataraj [983089983096] in whichrolling element bearings have been studied through the useo nonlinear models Te included nonlinearity in this caseis clearance and the model is then used in order to study chaotic motions in particular the regions o chaotic responseTe research orms a part o an ongoing drive to develop anaccurate model-based diagnostic technique or rolling ele-ment bearings taking into account clearance nonlinearitiesand chaotic responses

Guptaetal[983089983097] demonstrate another example o thelatestresearch into instability and chaos in rolling element bearingsthrough high-1047297delity simulations Tis detailed and complexstudy involves the application o a novel scheme to analyzethe quasiperiodic response o the system combined with anonautonomous ldquoshootingrdquo method Tis workhighlights thelevel o detail to which nonlinearities and complex nonlinearmotions in bearings are beginning to be understood andaccurately modeled Again such work presents the potentialor design o IVHM in uture evolutions o the research

As so much research hasbeen perormed (and is ongoing)into bearing aults across a wide variety o mechanicalsystems both prognostics and localization o bearing aults

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 613

983094 International Journal o Rotating Machinery

have been researched in somewhat more detail than someo the other aults detailed here Despite this much work still needs to be perormed in order to translate some o this core research into industrial applications Research suchas that detailed above has made signi1047297cant advances intodetermining bearing ailure as the root cause o malunction

Detecting which bearing is ailing across a complex systemhas received somewhat less research Bearing prognostics isanother area with much ongoing research being perormedmdashboth in the simulation and data-driven domains o givean example Hong et al [983090983088] combined grade lie andextensive mathematical modeling techniques in order to pro-duce prognostic models or aeroengine bearings Te resultsare described by the authors as ldquopractical and veri1047297ablerdquoAlthough a number o similar recent studies existthis work iso note or the extent o the studies perormed which includebearing test stand run-to-ailure validation Te lab resultsappear impressive this research has yet to be applied andtested in real lie applicationsmdashindicating that despite thenumber o parameters considered it is still not possible tomodel naturally occurring phenomena sufficiently

A large body o work in this area also exists romBorghesani et al [983090983089] where several examples o data driventechniques can be seen applied to a wide variety o bearingtypes An example o recent developments involves the use o ldquocepstrum prewhiteningrdquo in order to remove sufficient noiseor accurate bearing diagnosis and prognosis Tis work isparticularly noteworthy due to the emphasis on real-worldapplications where the traditional lab-based techniques o order tracking and synchronous averaging do not providesufficient noise removal or harsh industrial environmentsTe addition o such techniques is a crucial step in developingthe current generation o diagnosis and prognosis algorithmsor use in uture IVHM systems

983090983094 Shaf Cracks Another potentially serious ault in rotat-ing machinery is shaf cracks and so early detection o any such ault is highly important Methods o crack ormationandpropagation canbe diverse andrange rom highand low-cycle atigue to stress corrosion Simulation and modelingo shaf cracks can have signi1047297cant advantages over data-driven methods Perhaps the most obvious advantage is therelative simplicity o inserting a ault into orexample a 1047297niteelement model as opposed to seeding a ault in a workingindustrial machine As such research into shaf cracks has

been progressing steadily with the corresponding increasesin computing power

A clear synergy between data-driven and physics-basedsimulation research can be implied by a number o recentworks o research An example o recent advances rom a datacollection perspective is Li et al [983090983090] which details statisticalmodels based on historical data or condition monitoringpurposes Tis unique work uses the human auditory systemas inspiration or enriching methods o mechanical aults andeatures extraction Te results indicated by the paper areperhaps not as convincing as some other methods discussedin this paper it describes an interesting ldquooutside the boxrdquomethod o tackling common problems

From a modeling perspective Bachschmid et al [983090983091]cover a wide range o vibration phenomena in order todevelop a model-based identi1047297cation and severity procedureTis work is noted or its thoroughness in modeling pro-cedure including accurate modeling o the crack breathingmechanism A combination o high and low 1047297delity models

is validated through experimental study and ldquoexcellentrdquoaccuracy is claimed by the authors in detecting crack positionand depth through the use o the proposed model-baseddiagnostics

Te nature o shaf cracks has resulted in a wide variety o research beingperormed intoboth localization and prognos-tics o these aults (indeed the two topics can be consideredrelated) Recent examples o work in this area include Inoueetal[983090983092] who detail crack localization using orced responsemodeling In this case a test rig was constructed consistingo a circular shaf supported by two bearings Frequency-domain data was used to create a localization algorithmdesigned in combination with an FE model Although thisresearch provides good accuracy o localization in a labenvironment it remains untested in a more complex system(eg a ull gas turbine)

Sonnichsen [983090983093] describes 1047297nite element modeling o crack propagation with validation against experimentalresults provided to demonstrate the validity o such modelingtechniques Te paper concentrates on natural requenciesand resonance curves Whilst the authors claim improve-ments in the ability to understand such aults again thesystem in question is quite simplemdashand such FEA models aredifficult to scale up to ull size applications

983090983095 Blade Cracks Blade cracks i allowed to develop canresult in serious consequences Cracks can orm due to highcentriugal stresses across operational cycles (in the case o an aircraf gas turbine or example start up and take off through landing and taxi) As excessive crack growth canlead to catastrophic rotorblade ailure early detection andprognosis o such aults are essential As with shaf cracksphysics-driven simulation o blade cracks is an area o sig-ni1047297cant research Tis varies rom high-1047297delity 1047297nite elementmodels to low-1047297delity system and mathematical models Terecent work demonstrated by Green and Casey [983090983094]is agoodexample o recent mathematical modeling rom a diagnosisperspective In this paper the authors concentrate on early detection using global and local asymmetry crack models 983090X

harmonic components are identi1047297ed as key areas or the early detection o blade cracks however again this paper suffersrom being applied and tested on a relatively simple systemwhich may not scale up to a ull size turbine

Sonnichsen [983090983093] demonstrated high-1047297delity modelingthe authors used FEA to model crack growth making com-parisons and validating against an experimental rig Tiswork is particularly interesting as it outlines the advantagesand drawbacks with the latest state-o-the-art modelingtechniques

Localization and prognosis o blade cracks have alsobene1047297tted rom recent advances in simulation and modelingSawicki et al [983090983095] contains details o work on a novel active

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 713

International Journal o Rotating Machinery 983095

magnetic bearing system or use in the early detection local-ization and prognosis o blade cracks Again the emphasis ison early detection with the bearings used to excite the systemin order to obtain optimum response vibrations or analysisTe authors admit the approach hassome merit in diagnosingblade cracks however it is in the early stages o development

and work are ongoing FEA has also been used extensively to support blade crack prognostic tools Xiang et al [983090983096] arean extensive example o recent work In this case a numbero advanced FEA methods are applied to produce accurateFEA solutionsmdashthese include surace-1047297tting techniques andthe contour-plotting method Te authors experimentally

validate their work and suggest that it can be applied toprognosis and quantitative diagnosis o blade cracks Despitethe claimed advances again the scalability o such research toull size turbines is an issuemdashparticularly with regard to thecomplexity o the FEA models

983090983096 Rotor Bow Rotor bows can be a primary source o unwanted vibration in gas turbines Te main cause o arotor bow (rotor bows do not include bows due to gravity) isthermal differences in a system caused by operating con-ditions It is noted by Domes [983093] that this nonsymmetricalthermal distribution can cause excessive unbalance to theextent where a gas turbine will not start correctly Such rotorbows are common on start up or shut down and are ofenaccounted or in operational procedures However i thermalldquohot spotsrdquo exceed a given tolerance level they can causepermanent unbalances due to rotor de1047298ections Such rotorbows can lead to other aults includingrubbing andloosenesswhich complicate isolation and localization

raditional data-driven techniques or detecting rotorbows involve combinations o slow roll and vibration data[983090983097] More recently mathematical modeling techniques suchas that detailed by Meagher et al [983091983088] have been used inorder to diagnose residual rotor bows and differentiate theseaults rom other sources o unbalance Te authors o thispaper build upon established methods or the models and areunique in that they concentrate on response at the bearingpoints Tis is perhaps more useul orindustrial applicationsas vibration inormation is more readily available at bearingpoints rather than intrusive proximity probes which are ofenused in lab work

Te little work that exists on attempting to localise rotorbows across complex systems tends to be data-driven in

nature see Gaka and abaszewski [983091983089] where the authorsused statistical symptoms based on known data as a methodo diagnosing and prognosing a number o aults includingrotor bows and unbalance Tis paper attempts to address theirregularities and 1047298uctuations that occur over a long servicelie In order to achieve this a modi1047297ed energy processormodel is created using data drawn rom large steam turbinesover long periods o lie

Prognosing rotor bows is a complex subject As rotorbows are ofen caused by temperature de1047298ections makingpredictions or remaining useul lie and potential utureproblems lies not only in the realm o mechanical rotor-dynamics but also to some extent in thermodynamics Te

recent work detailed by Sinha [983091983090] is o note or detailingdiagnosis and quanti1047297cation o various rotordynamic aultsand describing the advantages o mathematical modelingover traditional vibration-based approaches Te topic o scalability regarding FEA models is discussed including anargument or the use o partial (simpli1047297ed) mathematical

models or large complex systemsAnother two works which are o interest with regard to

modeling o rotor bows include Shen et al [983091983091] where theauthors modeled a rotor-bearing system with a permanentrotor bow looking at the impact o secondary aults such asrub Te study o ault combinations in this paper is useulor ault differentiation studies however the authors study a permanent initial rotor bow Tis thereore does not takeinto account developing or worsening aults and the different

vibration phenomena that are observed as such aults aredeveloping

Lees et al [983091983092] describe the importance o model-basedault identi1047297cation techniques and outline recent research in

the area providing a good reerence paper or more researchon this speci1047297c ault

3 Discussion

983091983089 Sensor Suites Te subject o sensor suites is o greatimportance with regard to uture industrial applications Onone hand advanced and complex sensor suites generally enable improved ault localization and diagnosis howeverthe added complexity and cost has resulted in many o thesesystems being omitted rom the latest generations o rotatingmachinery in industrial applications From an industrialperspective the cost bene1047297t o additional sensors needs to

be signi1047297cant in order to justiy this approach In response anumber o works listed including [983096] as a good example takethe approach o achieving the same objectives using a greatly reduced sensor suite In many ways the required sensor suiteor many o the currently researched ault diagnosis andprognosis techniques provides an indication o the ability o the techniques to be used practically in industrial situationsA complex suite may return a very high success rate on alab-based rig however the impracticalities o mounting suchsuites on a real-world application negate the advantages

Further to this the consideration orsensor position mustbe considered Te use o proximity probes and keyphasortransducers has very clear advantages in a number o situ-

ations however the intrusive nature o the sensor preventsapplication in a number o complex systems With this inmind a number o the more promising studies listed rely onthe simple suiteo an accelerometer placed in positions whichare relatively remote rom the sources o vibration potentially with a noisy transer path

983091983090 Diagnosis It can be seen rom the research outlined inthe previous section that the diagnosis o aults in rotatingmachinery is a subject o much ongoing research Tisinvolves the improvement and development o traditional

vibration monitoring techniques development o new data-driven technologies and novel research into physics-based

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 813

983096 International Journal o Rotating Machinery

simulation and modeling In many cases these topics o research are dependent on one another or reasons o vali-dation veri1047297cation and speed o analysis In several casesit can be seen that multiple aults have been modeled orthe purposes o identi1047297cation and isolation However nostudies have yet been perormed which deal with all o the

aorementioned aults All o these aults are intrinsically related to one another Complex combinations o aults havebegun to be analyzed with the emphasis on developing new diagnosis techniques Physics-based modeling has proven toprovide signi1047297cant advances with regard to speci1047297c aultsnotably shaf and blade cracks where techniques such as FEAenable much easier aster and cheaper test data then seedingaults into live systems

It can be noted however that both high and low-1047297delity modeling techniques are being applied to cutting edgeresearch or all o the listed rotordynamic aults (and othersnot detailed in this paper) In addition to the advantages inthe speed o obtaining test results physics-based simulationis providing another dimension to data-driven techniquesSystem models are being used as part o logic and reasoningsuites in the identi1047297cation and differentiation o variousaults High-1047297delity models enable simulations o ault com-binations which are not possible practically or are pro-hibitively expensive in live systems

Nevertheless results obtained rom modeling studies stillneed to be validated against proven data-driven techniquesbeore implementation in industrial applications is possibleIt is also o note rom the literature reviewed or this paperthat almost all modeling or the diagnosis o aults involvesextensive simpli1047297cation ofen reducing potentially complexsystems down to one or two discshafbearing modelsAdapting the claimed results rom such research to systemswith many discsbearingsshafs is important in improvingexisting rotating machinery diagnostic techniques

Te inclusion o nonlinear effects is notable amongstmany authorsmdashas the understanding o the complex aultsand interactions continues to progress Such research opensthe potential or promising new avenues into areas o designor IVHM

Finally it is o note that despite the wide variety o advanced research here a consensus or the most efficienttechniques and algorithms has yet to emerge in order or the1047297nal adaption to industrial application to be achieved How-ever the consideration or industrial application taken by many o the authors detailed in this paper provides promise

or next generation IVHM systems

983091983091 Localisation Fault localization in rotating machinery isan important topic o research or uture condition-basedmonitoring systems Knowing not only what type o aulthas occurred but also where in the system is an importantconsideration which can in1047298uence maintenance proceduresin complex machinery It is worth noting rom the literaturesurveyed or this paper that many studies have ocusedon diagnosis and prognosis o single rotorbearing systemsOfen or legitimate reasonsmdashsimpli1047297cation or computingspeed or example Few studies however have taken intoaccount the localization o aultsacross whole systems Oneo

the more extensive examples highlighted in this paper (Hanet al [983089983093]) provides extensive analytical studies relevant tolocalization on a duel disk setup utilizing Hilbert-Huangransorms however in this example comprehensive valida-tion is lacking thus limiting the potential applicability o theresearch

Tis problem is not limited to modeling and simulation-based research Many newly developed data-driven tech-niques or diagnostics and prognostics claim good results by heavily instrumenting speci1047297c components o a test systemIn many industrial cases this is not possible practical or cost-effective

o give some examples a keyphasor transducer can beparticularly useul in diagnosing aults such as rotor bowHowever this equipment requires the ability to cut a key way or measurements to be perormed Optical sensors haverecently been applied to detect rotor unbalance yet such atechnology would be difficult to implement in a system withseveral rows o rotors (as in a gas turbine)

Whilst new diagnosis and prognosis techniques or aultsin rotating machinery are being continuously researchedthe lack o corresponding studies into localization can beconsidered one o the many challenges in promoting recentcore research into live industrial applications

983091983092 Fault Differentiation One promising development inrecent modeling and simulation o rotordynamic aults isthat o ault differentiation Several researchers have movedon rom studies on individual aults in order to concentrateon combinations o aults As stated rotordynamic aultssuch as those listed in this paper are linked to each otherSuch studies thereore concentrate on such topics as a

misalignment causing an unbalance or a looseness causinga rub Findings rom reports such as these are important inunderstanding complex anomalies In industrial applicationssimply detecting and rectiying an unbalance does not pro-

vide a satisactory solution i the root cause o the ault isa misalignment Tis is a complex topic as several aults canexhibit similar vibration characteristics making traditionaldetection techniques inaccurate in some cases

Making reerence to some o the aorementioned stud-ies those such as Bachschmid et al [983090983091] and Sonnichsen[983090983093] indicate some initial studies including several aultsTe results o such studies indicate a high success rate indifferentiating the aults in the given conditions however

a number o limitations still exist Studies such as theseconsider unbalancein the aultchains Tisis typically ldquostaticrdquoor simple unbalance applied to a single shafrotormdashin reality unbalance aults can be more complex than this In additionthe simple systems used to study and differentiate aultsare very different rom the complexities o a ull turbineHowever urther work may enable such research to ldquoscale-uprdquo to this stage with the aid o urther validation and

veri1047297cation

983091983093 Prognosis It can be seen rom the examples o recentresearch described or common aults that prognostic tech-niques are a topic where much work is being perormed

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 913

International Journal o Rotating Machinery 983097

Some o this work would be very difficult or impossibleto implement in an industrial situation others providepromising results o use to uture work in the area Predictingthe remaining useul lie o components is critical in thedevelopment o condition-based monitoring strategies orindustrial implementation

It can be noted that prognostic studies or certain rotor-dynamic aults are considerably more advanced than orothers Te most obvious examples o the aults detailed inthis paper are bearing aults shaf and blade cracks Teseare also areas where physics-driven simulations have hadan important impact Te ability to design any ault type(or combinations o aults) into such simulations (to be itslow-1047297delity mathematical models or high-1047297delity 1047297nite ele-ment analysis) has provided researchers with many differentavenues to explore Studies into the prognostics o otheraults can be complicated by various actors An unbalanceor example has had relatively little research perormed intoprognosis Te act that unbalance is ofen the cause o another underlying ault is one reason or the difficulty inresearching prognosis in detail or this ault Te results o unbalance are also very dependent on the severity o the aultand the system in which it occurs

It is worth noting that even or the aults where prognosisresearch is more advanced (eg shaf cracks) true develop-ment o the ault is not always studied Fan et al [983089983095] orexample study sets o discrete cracks in order to ldquoprognoserdquocrack growth In act thisstudy is more akinto detecting aultseverity with inerred effects on crack growth In contrast Liet al [983090983090] consider the continual development o such crackshowever the authors do admit this work is in its inancySuch points limit the ability or true prognosis to be achievedwithout extensive operational data

It is also worth noting that combinations o aults havebeen simulated extensively or the purpose o diagnosticshowever ew studies exist combining aultsor the purpose o prognostics Such studies are however a logical progressionrom some o those already perormed It may be that suchprognostic studies build upon some o the condition-basedmaintenance (CBM) or diagnostics High-level studies suchas Jaw and Merrill [983091983093] indicate how prognosis and diagnosistechniques can be combined into a CBM system Tis study is designed or a military aeroengine indicating the desire orsuch systems to be implementedmdashalthough the architecturedescribed has the capability o including various algorithmsoraultprognosis the amount o consideration orprognosis

o ault combinations is unclear

983091983094 Modelling Te large number o works reviewed or thispaper which includes modeling techniques in order to aiddiagnosis and prognosis indicates the possibilities providedby modern computer power and sofware developmentBroadly the modeling studied in the aorementioned papersbreaks down into two categories mathematical modeling and1047297nite element analysis (FEA) Te mathematical models suchasBorghesani etal [983090983089] ofen provide a theoretical basis uponwhich data driven or FEA studies can build and validateMathematical models despite being a traditional approach to

rotordynamic problems thereore continue to be developedto tackle more advanced problems

FEA techniques offer a constantly expanding area o simulation to explore Tey are used or research as diverseas crack propagation to unbalance localization [983091983094] Moderncodes such as NASRAN and Ansys enable rotordynamics to

be studied beyond the traditional stress and modal analysisTe ability to study the effects o rotation (eg throughCampbell diagrams and transient analysis) continues to driveresearch and innovation in this area Te studies discussedin this paper indicate the power o FEA with some highly accurate simulations having been perormed However thereis still a limitation o computer powermdashapplying some o the FEA techniques to whole engine models would proveto be too computationally expensive to be viable As aresult alternate methods such as model order reductionand system level modeling (and model-based reasoning orimplementation) are still required in order to make some o the FEA studies viable in an industrial environment

Nonlinearities continue to orm an important part o many recent studies with the complexity and detail o thespeci1047297ed aults being continually expanded Te depth o modeling has enabled in many cases accurate validationagainst experimental approaches Despite this completeunderstanding o the vibrational phenomena o rotatingmachinery is not currently possible despite the drive towardsthis end

983091983095 Rotordynamics and IVHM Te bulk o current researchinto rotordynamics rom the point o view o prognostichealth management (PHM) can be roughly divided intotwo types initial single-ault diagnosisprognosis techniquesand studies into the general requirements and limitationso current systems along with current and uture trendsAn example o the latter is Pusey [983091983095] who provides agood summary overview o current diagnosis and prognosistechniques with regard to condition-based maintenance

As a result o this split a clear gap exists between thecore research being perormed into rotordynamics rom acondition-based maintenance perspective and the identi1047297edneeds o industry aking a 1047298edgling piece o research andapplying it to a commercially-ready system (eg a gas turbineengine or an aircraf) is a long and complex task It isnevertheless worth noting that technologies or automatically detecting an unbalance or misalignment in a gas turbine were

developed over 983089983088 years beore the latest commercial aircrafwere conceptualized and yet these planes are still limited inthis capacity Tis highlights the need or work which linksthe undamental research into individual ault diagnosis toldquolive systemsrdquo in use in industry

Physics-based simulation and modeling o rotordynamicparts is a well-researched 1047297eld Such modeling has been usedas the basis o diagnosis and prognosis o aults by many researchers several recent examples have been outlined inthis paper Occasional pieces o work have been perormedinto modeling multiple aults such as Jain and Kundra [983091983096]who usea system model oronline identi1047297cation o unbalanceand cracks Beyond this however very limited research

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 1013

983089983088 International Journal o Rotating Machinery

Business casecost bene1047297tanalysis

Live system

RubCracksBearingfailure

Rotor bow Unbalance

Fault differentiationLocalisation

Root cause analysis

Remaining useful life

Data driven analysis and physics-based simulation

Mis-alignment

Experimental

1047297eld data

F983145983143983157983154983141 983089 Physics-based simulationmdashrom research to industry

exists Te demands o PHM techniques in industry are suchthat any systemmust notonly be capable o detecting multipleaults but must also be capable o detecting these aultsacrossa range o different systems Other considerations includethe aorementioned ability to differentiate between multipleaults Processing also needs to be taken into account as theobjective o these systems is to implement efficient conditionmonitoring and condition-based maintenance procedures I processing data is a long power-hungry process then thisaim cannot be achieved

Figure 983089 details a potential ramework required in order

to push core research such as that detailed in this reporttowards industrial applications Many studies now exist onindividual rotordynamic aults across a wide range o condi-tions and applications Some studies have taken this urtherwith advanced prognostic models and diagnosis o dual aults(primary cause and secondary effect) Future research inthe area o rotordynamics rom a PHM perspective couldpotentially provide the bridge between these studies and livesystems by combining physics-based simulations with data-driven techniques and validating against experimental data

Although quantiying the success o the research studiedisdifficultit ispossibleto de1047297ne the key areas in whicha tech-nique must excel in order to be considered viable Te work

byWheeleretal[983091983097] discussesin detailmetrics ordiagnosticand prognostic analysis as does that by Vachtsevanos [983092983088]and Saxena et al [983092983089] Te conclusions drawn rom thesepapers and applied in practice to research like that covered by this paper indicate that the ollowing metrics are importantwhen considering the potential o a given technique to diag-nose aults coverage alse positive rate and alse negativerate In the case o prognosis probabilities and lead timeto ailure are other important considerations Tis criteriaenables research to be assessed in terms o its suitability orindustrial applications Unortunately inormation on these

metrics is not made readily available by the authors o mostpapers

In terms o evaluating the effectivenesso theresearch dis-cussed in this paper there are difficulties in recommending aspeci1047297c technique over others or general application Mostpapers reviewed or this document (not just those reerencedand discussed in detail) take a technique (new or evolved)

validate or a given speci1047297c system and report success o theresearch Despite this some conclusions can be drawn romassessing common techniques applied across different studiesand different aults Although it is not possible to de1047297ne themost common methods or diagnosing andprognosing aultsin terms o numbers (as it was not possible to cover all recent

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 1113

International Journal o Rotating Machinery 983089983089

rotordynamics research or this paper) as perceived by theauthors the ollowing techniques have eatured prominently in the reviewed research

(i) Sensors accelerometers

(ii) Teoretical studies mathematical modeling

(iii) Physics-based simulation FEA(iv) Data-driven joint timerequency domain analysis

(v) Implementation neural networking

Tese techniques appear to be among the most promisingcurrently under development as they tend to eature numer-ous timesamongst some o the work with wider scope acrossault types and with the most comprehensive validationTere are o course many subsections to these techniqueshowever it shows one general direction o research and theclear possibilities posed in these areas

Future developments in the 1047297eld o IVHM or rotat-ing machinery may incorporate these techniques alongsideextensive use o nonlinear modeling and multiple aultinteractions Te 1047297eld o design or IVHM has only recently emerged however the potential exists or speci1047297c nonlinear-ities to be designed into a system in order to enable accuratediagnosis andprognosis o aults Te development o currentalgorithms to include diagnosis localization and prognosiso a range o aults will provide a signi1047297cant advancement oruture generations o IVHM systems Tis combined withcost-effective sensor suites indicates the potential or evolu-tions o some o the research detailed hereto orm part onextgeneration IVHM suites or rotating machinery

4 Conclusion

Tis paper has reviewed some o the latest research around anumber o rotordynamic aults namely unbalance misalign-ment rub and looseness 1047298uid-induced instability bearingaults shaf cracks blade cracks and rotor bow Each aultwas reviewed rom the perspective o sensors diagnosisprognosis localization and modeling

Key examples o recent work into the eight describedaults have been detailed through works by a number o eminent authors Additional work has been summarized andormatted or easy reerence Some current trends amongstthe recent body o work include developments in the vastarea o modeling nonlinearities combinations o high andlow 1047297delity modeling and synergy between data driven and

physics-based simulation approachesDespite the large volume o promising research reviewed

urther development in a number o areas is required in orderto produce effective next generation IVHM systems As suchuture developments may include usion o andor devel-opment o current algorithms to encompass all eight aultsdetailed consideration o prognosis diagnosis and localiza-tion achieved using a reduced cost-effective sensor suite

References

[983089] I K Jennions Integrated Vehicle Health Management Perspec-tives on an Emerging Field 983090983088983089983089

[983090] G Anderson ldquoProviding best-value IVHM solutions or agingaircrafrdquo in Proceedings o the 983097th Joint FAADoDNASA Aging Aircraf Conerence vol 983089 pp 983089ndash983089983089 983090983088983088983093

[983091] A Muszynska Rotordynamics aylor amp Francis 983089st ed edition983090983088983088983093

[983092] D E Bently Fundimentals o Rotating Machinery Diagnostics

ASME Press 983089st edition 983090983088983088983090[983093] B Domes ldquoVibration phenomena in aero-enginesrdquo in Proceed-ings o the 983097th International Conerence on Vibrations in Rotating Machinery Exeter (IMechE rsquo983088983096) vol 983089 pp 983089983093ndash983091983090 September983090983088983088983096

[983094] A Hess G Calvello and Dabney ldquoPHM a keyenabler or theJSF autonomie logistics support conceptrdquo in Proceedings o theIEEE Aerospace Conerence Proceedings pp 983091983093983092983091ndash983092983093983092983097 March983090983088983088983092

[983095] S N Ganeriwala B Schwarz and M H Richardson ldquoOperat-ing de1047298ection shapes detect unbalance in rotating equipmentrdquoSound and Vibration vol 983092983091 no 983093 pp 983089983089ndash983089983091 983090983088983088983097

[983096] G N D S Sudhakar and A S Sekhar ldquoIdenti1047297cation o unbal-ance in a rotor bearing systemrdquo Journal o Sound and Vibration

vol 983091983091983088 no 983089983088 pp 983090983090983097983097ndash983090983091983089983091 983090983088983089983089[983097] Yang and M W Hsu ldquoAn efficient diagnosis technique or

variations o shaf-bow and unbalancerdquo in Proceedings o the ASME International Design Engineering echnical Conerencesand Computers and Inormation in Engineering Conerence(DEC rsquo983088983097) pp 983093983095ndash983094983094 San Diego Cali USA September 983090983088983088983097

[983089983088] A El-Shaei S H aw1047297ck and M O A Mokhtar ldquoExperi-mental investigation o the effect o angular misalignemt onthe instability o plain journal bearingsrdquo in Proceedings o the ASMESLE InternationalJointribology Conerence(IJC rsquo983088983097)pp 983089983095983089ndash983089983095983091 Memphis enn USA October 983090983088983088983097

[983089983089] A El-Shaei S H aw1047297ck and M O A Mokhtar ldquoMisalign-ment modeling in rotating systemsrdquo in Proceedings o the ASMEurbo Expo H Bahaloo A Ebrahimi and M Samadi Eds vol

983089983095983089983089 pp 983097983095983091ndash983097983095983097 Orlando Fla USA June 983090983088983088983097

[983089983090] L F Villaet al ldquoStatistical diagnosisbased on vibration analysisor gear test-bench under non-stationary conditions o speedand loadrdquo Mechanical Systems and Signal Processing vol 983090983097 pp983092983091983094ndash983092983092983094 983090983088983089983090

[983089983091] C F Ngolah et al ldquoIntelligent ault recognition and diagnosisor rotating machines using neural networksrdquo Sofware Scienceand Computational Intelligence vol 983091 no 983092 983089983095 pages 983090983088983089983089

[983089983092] Y Lu Z Ren H Chen N Song and B Wen ldquoStudy on loose-ness and impactmdashrub coupling aults o a vertical dual-disk cantilever rotor- Bearing systemrdquo Key Engineering Materials vol 983091983093983091ndash983091983093983096 no 983092 pp 983090983092983095983097ndash983090983092983096983090 983090983088983088983095

[983089983093] Q Han Z Zhang and B Wen ldquoPeriodic motions o a dual-

disc rotor system with rub-impact at 1047297xed limiterrdquo Proceedingso the Institution o Mechanical Engineers vol 983090983090983090 no 983089983088 pp983089983097983091983093ndash983089983097983092983094 983090983088983088983096

[983089983094] H F de Castro K L Cavalca and R Nordmann ldquoWhirland whip instabilities in rotor-bearing system considering anonlinear orce modelrdquo Journal o Sound and Vibration vol 983091983089983095no 983089-983090 pp 983090983095983091ndash983090983097983091 983090983088983088983096

[983089983095] C C Fan et al ldquoMechanical systems and signal processingrdquoStudy o Start-Up Vibration Response or Oil Whirl and Dry Whip vol 983090983093 pp 983091983089983088983090ndash983091983089983089983093 983090983088983089983089

[983089983096] K Kappaganthu and C Nataraj ldquoNonlinear modeling andanalysis o a rolling element bearing with a clearancerdquo Commu-nications in Nonlinear Science and Numerical Simulation vol983089983094no 983089983088 pp 983092983089983091983092ndash983092983089983092983093 983090983088983089983089

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 1213

983089983090 International Journal o Rotating Machinery

[983089983097] C Gupta K Gupta and D K Sehgal ldquoInstability and chaoso a 1047298exible rotor ball bearing system an investigation on thein1047298uence o rotating imbalance and bearing clearancerdquo Journal o Engineering or Gas urbines and Power vol 983089983091983091 no 983096ArticleID 983088983096983090983093983088983089 983089983089 pages 983090983088983089983089

[983090983088] J Hong X Miao L Han and Y Ma ldquoPrognostics model orpredicting aero-engine bearing grade-lierdquo in Proceedings o the ASME urbo Expo vol 983089 pp 983094983091983097ndash983094983092983095 Orlando Fla USA June983090983088983088983097

[983090983089] P Borghesani P Pennacchi R B Randall N Sawalhi and RRicci ldquoApplication o cepstrum pre-whitening or the diagnosiso bearing aults under variable speed conditionsrdquo Mechanical Systems and Signal Processing 983090983088983089983090

[983090983090] Y Li J Zhang L Dai Z Zhang and J Liu ldquoAuditory-model-based eature extraction method or mechanical aults diagno-sisrdquo ChineseJournal o MechanicalEngineering (EnglishEdition) vol 983090983091 no 983091 pp 983091983097983089ndash983091983097983095 983090983088983089983088

[983090983091] N Bachschmid P Pennacchi and E anzi ldquoCracked rotatingshafs typical behaviors modeling and diagnosisrdquo IUAM Symposium on Emerging rends in Rotor Dynamics vol 983089983088983089983089 pp

983092983092983089ndash983092983093983092 983090983088983089983089[983090983092] Inoue N Nagata and Y Ishida ldquoFem modelling and

experimental veri1047297cation o a rotor system with a open crackrdquoin Proceedings o the ASME International Design Engineering echnical Conerences and Computers and Inormation in Engi-neering Conerence (DEC rsquo983088983097) pp 983089983089983089983091ndash983089983089983090983090 San Diego CaliUSA September 983090983088983088983097

[983090983093] H E Sonnichsen ldquoReal-time detection o developing cracksin jet engine rotorsrdquo in Proceedings o the IEEE AerospaceConerence pp 983089983095983091ndash983089983096983092 March 983090983088983088983088

[983090983094] I Green and C Casey ldquoCrack detection in a rotor dynamicsystem by vibration monitoringmdashpart I analysisrdquo Journal o Engineering or Gas urbines and Power vol 983089983090983095 no 983090 pp 983092983090983093ndash983092983091983094 983090983088983088983093

[983090983095] J Sawicki M I Friswell A H Pesch and A WroblewskildquoCondition monitoring o rotor usingactive magnetic actuatorrdquoin Proceedings o the ASME urbo Expo pp 983089983090983093983095ndash983089983090983094983093 BerlinGermany June 983090983088983088983096

[983090983096] J Xiang X Chen Q Mo and Z He ldquoIdenti1047297cation o crack ina rotor system based on wavelet 1047297nite element methodrdquo FiniteElements in Analysis and Design vol 983092983091 no 983089983092 pp 983089983088983094983096ndash983089983088983096983089983090983088983088983095

[983090983097] M G MaalouldquoSlow speed vibrationsignal analysisi you canrsquotdo it slow you canrsquot do it astrdquo in Proceedings o the ASME urboExpo pp 983093983093983097ndash983093983094983095 Montreal Canada May 983090983088983088983095

[983091983088] J Meagher X Wu and C Lencioni ldquoResponse o a warped1047298exible rotor with a 1047298uid bearingrdquo International Journal o

Rotating Machinery vol 983090983088983088983096 Article ID 983089983092983095983094983093983091 983097 pages 983090983088983088983096

[983091983089] Gaka andM abaszewski ldquoAn applicationo statistical symp-toms in machine condition diagnosticsrdquo Mechanical Systemsand Signal Processing vol 983090983093 no 983089 pp 983090983093983091ndash983090983094983093 983090983088983089983089

[983091983090] J K Sinha ldquoRecent trands in ault quanti1047297cation in rotatingmachinesrdquo Advances in Vibration EngIneerIng vol 983096 no 983089 pp983095983097ndash983096983093 983090983088983088983097

[983091983091] X Shen JJia andM Zhao ldquoNonlinear analysis o a rub-impactrotor-bearing system with initial permanent rotor bowrdquo Archiveo Applied Mechanics vol 983095983096 no 983091 pp 983090983090983093ndash983090983092983088 983090983088983088983096

[983091983092] A W Lees J K Sinha and M I Friswell ldquoModel-based identi-1047297cation o rotating machinesrdquo Mechanical Systems and Signal Processing vol 983090983091 no 983094 pp 983089983096983096983092ndash983089983096983097983091 983090983088983088983097

[983091983093] L C Jaw and W Merrill ldquoCBM+ research environmentmdashacilitating technology development experimentation andmaturationrdquo in Proceedings o the IEEE Aerospace Conerence(AC rsquo983088983096) March 983090983088983088983096

[983091983094] R Walker S Perinpanayagam and I K Jennions ldquoSimulatingUnbalance or Future IVHM Applicationsrdquo in Proceedings o theSociety or Experimental Mechanics (IMAC rsquo983089983090) vol 983090983097 pp 983089983092983089ndash983089983092983096 May 983090983088983089983090

[983091983095] H C Pusey ldquourbomachinery condition monitoring and ail-ure prognosisrdquo Sound and Vibration vol 983092983089 no 983091 pp 983089983088ndash983089983093983090983088983088983095

[983091983096] J R Jain and K Kundra ldquoModel based online diagnosiso unbalance and transverse atigue crack in rotor systemsrdquo Mechanics Research Communications vol 983091983089 no 983093 pp 983093983093983095ndash983093983094983096983090983088983088983092

[983091983097] K R Wheeler Kurtoglu and S D Poll ldquoA survey o healthmanagement user objectives related to diagnostic and prognos-tic metricsrdquo in Proceedings o the 983090983097th Computers and Inor-mation Engineering Conerence vol 983090 pp 983089983090983096983095ndash983089983090983097983096 983090983088983088983097

[983092983088] G Vachtsevanos ldquoPerormance metrics or ault prognosis o

complex systemsrdquo in IEEE Systems Readiness echnology Con- erence (AUOESCON rsquo983088983091) pp 983091983092983089ndash983091983092983093 September 983090983088983088983091

[983092983089] A Saxena J Celaya E Balaban et al ldquoMetrics or evaluatingperormance o prognostic techniquesrdquo in Proceedings o theInternational Conerence on Prognostics and Health Manage-ment (PHM rsquo983088983096) pp 983089ndash983089983095 October 983090983088983088983096

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 1313

Submit your manuscripts at

httpwwwhindawicom

Page 5: Rotordynamic Faults 2013

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 513

International Journal o Rotating Machinery 983093

Lu et al [983089983092] have perormed several studies into rub andloosenessincluding thisexamplein which a 1047298exibly mountedshaf has an induced rub due to a range o contact ringsTe study ocusses on the potentially dangerous effects o rubin causing excessive nonsynchronous and chaotic vibrationsTe links with unbalance and misalignment are discussed

and detailed Te considerations or real-world cases in thedescribed research are considerable as part o the drivetowards ull understanding o the nonlinear effects o rub andlooseness

Localization o rub and looseness across whole systemsis relatively lightly studied in the literature Many works(including those already cited) look at single or dual-rotorsystems where localization o such aults is not an issue Inan industrial setting complex systems may comprise many rotors in several compressor and turbine stages signi1047297cantly complicating diagnosis o such aults Research into progno-sis o rotor-stator rubs lies mostly within the domain o data-driven techniques Modeling and simulation research can beused to support data-driven techniques or prognosis andcondition-based monitoring Han et al [983089983093] is an example o this the authors use 1047297nite element modeling to construct adual rotor model Various types o rub-impact are then stud-ied Such studies can provide a wide range o inormationwhich can then be combined with data obtained rom livesystems potentially with seeded aults in order to constructaccurate remaining useul lie predictions As pointed out by the authors one key advantage o simulation is the ability to study more complex systems with a higher number o rotors which is used throughout this research Tis presentsa different approach to identiying eatures or identi1047297cationo rub and looseness

983090983092 Fluid-Induced Instability Fluid-induced instabilities(ofen reerred to as whip and whirl) are potentially very serious aults which can result in wear atigue and extensivedamage to machine components Such instabilities can beound in interstage seals 1047298uid lubricated bearings andblade-tip clearances Research into simulating and modeling1047298uid-induced instability has produced several works o interest to ault diagnosis o rotating machines in the lastew years De Castro et al [983089983094] are a good example wherenonlinear mathematical models are prepared or a rotor-bearing system Te models are then used to predictinstability thresholds Te authors consider a test case againsta power plant turbine and a test rig thereore validating the

simulations Te case o unbalance aults causing whip andwhirl phenomena is also considered Te main conclusionthereore drawn rom the work is that the authors concludedthat nonlinear hydrodynamic journal bearing models enablesufficiently accurate simulations or predicting instability thresholds Fan et al [983089983095] represent an example o work romthe perspective o aeroengine turbines In this case startupconditions are studied using a ull Hilbert spectrum Te aimbehind the paper is to accurately predict the point at whichwhip and whirl occur thus enabling this to be avoided at thedesign stage Such 1047297ndings could potentially also be used toidentiy whip and whirl as the case o a ault afer a period o wear in the operating machine

Prognosing 1047298uid-induced instability is a relatively lightly researched topic Fluid instabilities can be covered as part o extensive research into remaining useul lie o bearings Tepotential exists or modeling and simulation techniques suchas those detailed above to become a part o prognosis or1047298uid-induced instabilities due to the act that it can be very

difficult to seed such aults into live systems or testing andevaluation As with other aults detailed in this paper many studies have been perormed with the aim o describing 1047298uid-induced instabilities based on the measurement or simulationo single (or occasionally dual) rotor setups Physics-basedsimulation with the aim o localising 1047298uid instability aultsacross a whole system can be limited by the complexity o both the ault and the system hence the simpli1047297cation tosingle rotor-stator bearing systems

983090983093 Bearing Failure An area where data-driven techniquesare still providing the basis o much research in the 1047297eld o rotordynamics is that o bearing ailure Te title ldquobearingailurerdquo can cover a wide range o potential issues whichcontinue to be studied in detail Faults can occur in all kindso engine bearingsmdashthe inner andouter case the cage andtherolling elements 1047298uid-induced instabilities (addressed in aseparate section) lubrication and the complexities o activemagnetic bearings to name some examples All types o bearing relevant to rotating machinery are the subject o ongoing research and this subject has the potential to ormseveral separate papers As a brie highlight some recentexamples are discussed as ollows

Data-driven techniques have enabled accurate bearingdiagnostics and prognostics to be described or a rangeo rotordynamic systems Despite the prevalence o data-driven research in this area research rom a physics-basedsimulation perspective has also recently produced someinteresting papers o relevance to condition monitoring andhealth management o rotating machinery

Tis includes Kappaganthu and Nataraj [983089983096] in whichrolling element bearings have been studied through the useo nonlinear models Te included nonlinearity in this caseis clearance and the model is then used in order to study chaotic motions in particular the regions o chaotic responseTe research orms a part o an ongoing drive to develop anaccurate model-based diagnostic technique or rolling ele-ment bearings taking into account clearance nonlinearitiesand chaotic responses

Guptaetal[983089983097] demonstrate another example o thelatestresearch into instability and chaos in rolling element bearingsthrough high-1047297delity simulations Tis detailed and complexstudy involves the application o a novel scheme to analyzethe quasiperiodic response o the system combined with anonautonomous ldquoshootingrdquo method Tis workhighlights thelevel o detail to which nonlinearities and complex nonlinearmotions in bearings are beginning to be understood andaccurately modeled Again such work presents the potentialor design o IVHM in uture evolutions o the research

As so much research hasbeen perormed (and is ongoing)into bearing aults across a wide variety o mechanicalsystems both prognostics and localization o bearing aults

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 613

983094 International Journal o Rotating Machinery

have been researched in somewhat more detail than someo the other aults detailed here Despite this much work still needs to be perormed in order to translate some o this core research into industrial applications Research suchas that detailed above has made signi1047297cant advances intodetermining bearing ailure as the root cause o malunction

Detecting which bearing is ailing across a complex systemhas received somewhat less research Bearing prognostics isanother area with much ongoing research being perormedmdashboth in the simulation and data-driven domains o givean example Hong et al [983090983088] combined grade lie andextensive mathematical modeling techniques in order to pro-duce prognostic models or aeroengine bearings Te resultsare described by the authors as ldquopractical and veri1047297ablerdquoAlthough a number o similar recent studies existthis work iso note or the extent o the studies perormed which includebearing test stand run-to-ailure validation Te lab resultsappear impressive this research has yet to be applied andtested in real lie applicationsmdashindicating that despite thenumber o parameters considered it is still not possible tomodel naturally occurring phenomena sufficiently

A large body o work in this area also exists romBorghesani et al [983090983089] where several examples o data driventechniques can be seen applied to a wide variety o bearingtypes An example o recent developments involves the use o ldquocepstrum prewhiteningrdquo in order to remove sufficient noiseor accurate bearing diagnosis and prognosis Tis work isparticularly noteworthy due to the emphasis on real-worldapplications where the traditional lab-based techniques o order tracking and synchronous averaging do not providesufficient noise removal or harsh industrial environmentsTe addition o such techniques is a crucial step in developingthe current generation o diagnosis and prognosis algorithmsor use in uture IVHM systems

983090983094 Shaf Cracks Another potentially serious ault in rotat-ing machinery is shaf cracks and so early detection o any such ault is highly important Methods o crack ormationandpropagation canbe diverse andrange rom highand low-cycle atigue to stress corrosion Simulation and modelingo shaf cracks can have signi1047297cant advantages over data-driven methods Perhaps the most obvious advantage is therelative simplicity o inserting a ault into orexample a 1047297niteelement model as opposed to seeding a ault in a workingindustrial machine As such research into shaf cracks has

been progressing steadily with the corresponding increasesin computing power

A clear synergy between data-driven and physics-basedsimulation research can be implied by a number o recentworks o research An example o recent advances rom a datacollection perspective is Li et al [983090983090] which details statisticalmodels based on historical data or condition monitoringpurposes Tis unique work uses the human auditory systemas inspiration or enriching methods o mechanical aults andeatures extraction Te results indicated by the paper areperhaps not as convincing as some other methods discussedin this paper it describes an interesting ldquooutside the boxrdquomethod o tackling common problems

From a modeling perspective Bachschmid et al [983090983091]cover a wide range o vibration phenomena in order todevelop a model-based identi1047297cation and severity procedureTis work is noted or its thoroughness in modeling pro-cedure including accurate modeling o the crack breathingmechanism A combination o high and low 1047297delity models

is validated through experimental study and ldquoexcellentrdquoaccuracy is claimed by the authors in detecting crack positionand depth through the use o the proposed model-baseddiagnostics

Te nature o shaf cracks has resulted in a wide variety o research beingperormed intoboth localization and prognos-tics o these aults (indeed the two topics can be consideredrelated) Recent examples o work in this area include Inoueetal[983090983092] who detail crack localization using orced responsemodeling In this case a test rig was constructed consistingo a circular shaf supported by two bearings Frequency-domain data was used to create a localization algorithmdesigned in combination with an FE model Although thisresearch provides good accuracy o localization in a labenvironment it remains untested in a more complex system(eg a ull gas turbine)

Sonnichsen [983090983093] describes 1047297nite element modeling o crack propagation with validation against experimentalresults provided to demonstrate the validity o such modelingtechniques Te paper concentrates on natural requenciesand resonance curves Whilst the authors claim improve-ments in the ability to understand such aults again thesystem in question is quite simplemdashand such FEA models aredifficult to scale up to ull size applications

983090983095 Blade Cracks Blade cracks i allowed to develop canresult in serious consequences Cracks can orm due to highcentriugal stresses across operational cycles (in the case o an aircraf gas turbine or example start up and take off through landing and taxi) As excessive crack growth canlead to catastrophic rotorblade ailure early detection andprognosis o such aults are essential As with shaf cracksphysics-driven simulation o blade cracks is an area o sig-ni1047297cant research Tis varies rom high-1047297delity 1047297nite elementmodels to low-1047297delity system and mathematical models Terecent work demonstrated by Green and Casey [983090983094]is agoodexample o recent mathematical modeling rom a diagnosisperspective In this paper the authors concentrate on early detection using global and local asymmetry crack models 983090X

harmonic components are identi1047297ed as key areas or the early detection o blade cracks however again this paper suffersrom being applied and tested on a relatively simple systemwhich may not scale up to a ull size turbine

Sonnichsen [983090983093] demonstrated high-1047297delity modelingthe authors used FEA to model crack growth making com-parisons and validating against an experimental rig Tiswork is particularly interesting as it outlines the advantagesand drawbacks with the latest state-o-the-art modelingtechniques

Localization and prognosis o blade cracks have alsobene1047297tted rom recent advances in simulation and modelingSawicki et al [983090983095] contains details o work on a novel active

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 713

International Journal o Rotating Machinery 983095

magnetic bearing system or use in the early detection local-ization and prognosis o blade cracks Again the emphasis ison early detection with the bearings used to excite the systemin order to obtain optimum response vibrations or analysisTe authors admit the approach hassome merit in diagnosingblade cracks however it is in the early stages o development

and work are ongoing FEA has also been used extensively to support blade crack prognostic tools Xiang et al [983090983096] arean extensive example o recent work In this case a numbero advanced FEA methods are applied to produce accurateFEA solutionsmdashthese include surace-1047297tting techniques andthe contour-plotting method Te authors experimentally

validate their work and suggest that it can be applied toprognosis and quantitative diagnosis o blade cracks Despitethe claimed advances again the scalability o such research toull size turbines is an issuemdashparticularly with regard to thecomplexity o the FEA models

983090983096 Rotor Bow Rotor bows can be a primary source o unwanted vibration in gas turbines Te main cause o arotor bow (rotor bows do not include bows due to gravity) isthermal differences in a system caused by operating con-ditions It is noted by Domes [983093] that this nonsymmetricalthermal distribution can cause excessive unbalance to theextent where a gas turbine will not start correctly Such rotorbows are common on start up or shut down and are ofenaccounted or in operational procedures However i thermalldquohot spotsrdquo exceed a given tolerance level they can causepermanent unbalances due to rotor de1047298ections Such rotorbows can lead to other aults includingrubbing andloosenesswhich complicate isolation and localization

raditional data-driven techniques or detecting rotorbows involve combinations o slow roll and vibration data[983090983097] More recently mathematical modeling techniques suchas that detailed by Meagher et al [983091983088] have been used inorder to diagnose residual rotor bows and differentiate theseaults rom other sources o unbalance Te authors o thispaper build upon established methods or the models and areunique in that they concentrate on response at the bearingpoints Tis is perhaps more useul orindustrial applicationsas vibration inormation is more readily available at bearingpoints rather than intrusive proximity probes which are ofenused in lab work

Te little work that exists on attempting to localise rotorbows across complex systems tends to be data-driven in

nature see Gaka and abaszewski [983091983089] where the authorsused statistical symptoms based on known data as a methodo diagnosing and prognosing a number o aults includingrotor bows and unbalance Tis paper attempts to address theirregularities and 1047298uctuations that occur over a long servicelie In order to achieve this a modi1047297ed energy processormodel is created using data drawn rom large steam turbinesover long periods o lie

Prognosing rotor bows is a complex subject As rotorbows are ofen caused by temperature de1047298ections makingpredictions or remaining useul lie and potential utureproblems lies not only in the realm o mechanical rotor-dynamics but also to some extent in thermodynamics Te

recent work detailed by Sinha [983091983090] is o note or detailingdiagnosis and quanti1047297cation o various rotordynamic aultsand describing the advantages o mathematical modelingover traditional vibration-based approaches Te topic o scalability regarding FEA models is discussed including anargument or the use o partial (simpli1047297ed) mathematical

models or large complex systemsAnother two works which are o interest with regard to

modeling o rotor bows include Shen et al [983091983091] where theauthors modeled a rotor-bearing system with a permanentrotor bow looking at the impact o secondary aults such asrub Te study o ault combinations in this paper is useulor ault differentiation studies however the authors study a permanent initial rotor bow Tis thereore does not takeinto account developing or worsening aults and the different

vibration phenomena that are observed as such aults aredeveloping

Lees et al [983091983092] describe the importance o model-basedault identi1047297cation techniques and outline recent research in

the area providing a good reerence paper or more researchon this speci1047297c ault

3 Discussion

983091983089 Sensor Suites Te subject o sensor suites is o greatimportance with regard to uture industrial applications Onone hand advanced and complex sensor suites generally enable improved ault localization and diagnosis howeverthe added complexity and cost has resulted in many o thesesystems being omitted rom the latest generations o rotatingmachinery in industrial applications From an industrialperspective the cost bene1047297t o additional sensors needs to

be signi1047297cant in order to justiy this approach In response anumber o works listed including [983096] as a good example takethe approach o achieving the same objectives using a greatly reduced sensor suite In many ways the required sensor suiteor many o the currently researched ault diagnosis andprognosis techniques provides an indication o the ability o the techniques to be used practically in industrial situationsA complex suite may return a very high success rate on alab-based rig however the impracticalities o mounting suchsuites on a real-world application negate the advantages

Further to this the consideration orsensor position mustbe considered Te use o proximity probes and keyphasortransducers has very clear advantages in a number o situ-

ations however the intrusive nature o the sensor preventsapplication in a number o complex systems With this inmind a number o the more promising studies listed rely onthe simple suiteo an accelerometer placed in positions whichare relatively remote rom the sources o vibration potentially with a noisy transer path

983091983090 Diagnosis It can be seen rom the research outlined inthe previous section that the diagnosis o aults in rotatingmachinery is a subject o much ongoing research Tisinvolves the improvement and development o traditional

vibration monitoring techniques development o new data-driven technologies and novel research into physics-based

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 813

983096 International Journal o Rotating Machinery

simulation and modeling In many cases these topics o research are dependent on one another or reasons o vali-dation veri1047297cation and speed o analysis In several casesit can be seen that multiple aults have been modeled orthe purposes o identi1047297cation and isolation However nostudies have yet been perormed which deal with all o the

aorementioned aults All o these aults are intrinsically related to one another Complex combinations o aults havebegun to be analyzed with the emphasis on developing new diagnosis techniques Physics-based modeling has proven toprovide signi1047297cant advances with regard to speci1047297c aultsnotably shaf and blade cracks where techniques such as FEAenable much easier aster and cheaper test data then seedingaults into live systems

It can be noted however that both high and low-1047297delity modeling techniques are being applied to cutting edgeresearch or all o the listed rotordynamic aults (and othersnot detailed in this paper) In addition to the advantages inthe speed o obtaining test results physics-based simulationis providing another dimension to data-driven techniquesSystem models are being used as part o logic and reasoningsuites in the identi1047297cation and differentiation o variousaults High-1047297delity models enable simulations o ault com-binations which are not possible practically or are pro-hibitively expensive in live systems

Nevertheless results obtained rom modeling studies stillneed to be validated against proven data-driven techniquesbeore implementation in industrial applications is possibleIt is also o note rom the literature reviewed or this paperthat almost all modeling or the diagnosis o aults involvesextensive simpli1047297cation ofen reducing potentially complexsystems down to one or two discshafbearing modelsAdapting the claimed results rom such research to systemswith many discsbearingsshafs is important in improvingexisting rotating machinery diagnostic techniques

Te inclusion o nonlinear effects is notable amongstmany authorsmdashas the understanding o the complex aultsand interactions continues to progress Such research opensthe potential or promising new avenues into areas o designor IVHM

Finally it is o note that despite the wide variety o advanced research here a consensus or the most efficienttechniques and algorithms has yet to emerge in order or the1047297nal adaption to industrial application to be achieved How-ever the consideration or industrial application taken by many o the authors detailed in this paper provides promise

or next generation IVHM systems

983091983091 Localisation Fault localization in rotating machinery isan important topic o research or uture condition-basedmonitoring systems Knowing not only what type o aulthas occurred but also where in the system is an importantconsideration which can in1047298uence maintenance proceduresin complex machinery It is worth noting rom the literaturesurveyed or this paper that many studies have ocusedon diagnosis and prognosis o single rotorbearing systemsOfen or legitimate reasonsmdashsimpli1047297cation or computingspeed or example Few studies however have taken intoaccount the localization o aultsacross whole systems Oneo

the more extensive examples highlighted in this paper (Hanet al [983089983093]) provides extensive analytical studies relevant tolocalization on a duel disk setup utilizing Hilbert-Huangransorms however in this example comprehensive valida-tion is lacking thus limiting the potential applicability o theresearch

Tis problem is not limited to modeling and simulation-based research Many newly developed data-driven tech-niques or diagnostics and prognostics claim good results by heavily instrumenting speci1047297c components o a test systemIn many industrial cases this is not possible practical or cost-effective

o give some examples a keyphasor transducer can beparticularly useul in diagnosing aults such as rotor bowHowever this equipment requires the ability to cut a key way or measurements to be perormed Optical sensors haverecently been applied to detect rotor unbalance yet such atechnology would be difficult to implement in a system withseveral rows o rotors (as in a gas turbine)

Whilst new diagnosis and prognosis techniques or aultsin rotating machinery are being continuously researchedthe lack o corresponding studies into localization can beconsidered one o the many challenges in promoting recentcore research into live industrial applications

983091983092 Fault Differentiation One promising development inrecent modeling and simulation o rotordynamic aults isthat o ault differentiation Several researchers have movedon rom studies on individual aults in order to concentrateon combinations o aults As stated rotordynamic aultssuch as those listed in this paper are linked to each otherSuch studies thereore concentrate on such topics as a

misalignment causing an unbalance or a looseness causinga rub Findings rom reports such as these are important inunderstanding complex anomalies In industrial applicationssimply detecting and rectiying an unbalance does not pro-

vide a satisactory solution i the root cause o the ault isa misalignment Tis is a complex topic as several aults canexhibit similar vibration characteristics making traditionaldetection techniques inaccurate in some cases

Making reerence to some o the aorementioned stud-ies those such as Bachschmid et al [983090983091] and Sonnichsen[983090983093] indicate some initial studies including several aultsTe results o such studies indicate a high success rate indifferentiating the aults in the given conditions however

a number o limitations still exist Studies such as theseconsider unbalancein the aultchains Tisis typically ldquostaticrdquoor simple unbalance applied to a single shafrotormdashin reality unbalance aults can be more complex than this In additionthe simple systems used to study and differentiate aultsare very different rom the complexities o a ull turbineHowever urther work may enable such research to ldquoscale-uprdquo to this stage with the aid o urther validation and

veri1047297cation

983091983093 Prognosis It can be seen rom the examples o recentresearch described or common aults that prognostic tech-niques are a topic where much work is being perormed

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 913

International Journal o Rotating Machinery 983097

Some o this work would be very difficult or impossibleto implement in an industrial situation others providepromising results o use to uture work in the area Predictingthe remaining useul lie o components is critical in thedevelopment o condition-based monitoring strategies orindustrial implementation

It can be noted that prognostic studies or certain rotor-dynamic aults are considerably more advanced than orothers Te most obvious examples o the aults detailed inthis paper are bearing aults shaf and blade cracks Teseare also areas where physics-driven simulations have hadan important impact Te ability to design any ault type(or combinations o aults) into such simulations (to be itslow-1047297delity mathematical models or high-1047297delity 1047297nite ele-ment analysis) has provided researchers with many differentavenues to explore Studies into the prognostics o otheraults can be complicated by various actors An unbalanceor example has had relatively little research perormed intoprognosis Te act that unbalance is ofen the cause o another underlying ault is one reason or the difficulty inresearching prognosis in detail or this ault Te results o unbalance are also very dependent on the severity o the aultand the system in which it occurs

It is worth noting that even or the aults where prognosisresearch is more advanced (eg shaf cracks) true develop-ment o the ault is not always studied Fan et al [983089983095] orexample study sets o discrete cracks in order to ldquoprognoserdquocrack growth In act thisstudy is more akinto detecting aultseverity with inerred effects on crack growth In contrast Liet al [983090983090] consider the continual development o such crackshowever the authors do admit this work is in its inancySuch points limit the ability or true prognosis to be achievedwithout extensive operational data

It is also worth noting that combinations o aults havebeen simulated extensively or the purpose o diagnosticshowever ew studies exist combining aultsor the purpose o prognostics Such studies are however a logical progressionrom some o those already perormed It may be that suchprognostic studies build upon some o the condition-basedmaintenance (CBM) or diagnostics High-level studies suchas Jaw and Merrill [983091983093] indicate how prognosis and diagnosistechniques can be combined into a CBM system Tis study is designed or a military aeroengine indicating the desire orsuch systems to be implementedmdashalthough the architecturedescribed has the capability o including various algorithmsoraultprognosis the amount o consideration orprognosis

o ault combinations is unclear

983091983094 Modelling Te large number o works reviewed or thispaper which includes modeling techniques in order to aiddiagnosis and prognosis indicates the possibilities providedby modern computer power and sofware developmentBroadly the modeling studied in the aorementioned papersbreaks down into two categories mathematical modeling and1047297nite element analysis (FEA) Te mathematical models suchasBorghesani etal [983090983089] ofen provide a theoretical basis uponwhich data driven or FEA studies can build and validateMathematical models despite being a traditional approach to

rotordynamic problems thereore continue to be developedto tackle more advanced problems

FEA techniques offer a constantly expanding area o simulation to explore Tey are used or research as diverseas crack propagation to unbalance localization [983091983094] Moderncodes such as NASRAN and Ansys enable rotordynamics to

be studied beyond the traditional stress and modal analysisTe ability to study the effects o rotation (eg throughCampbell diagrams and transient analysis) continues to driveresearch and innovation in this area Te studies discussedin this paper indicate the power o FEA with some highly accurate simulations having been perormed However thereis still a limitation o computer powermdashapplying some o the FEA techniques to whole engine models would proveto be too computationally expensive to be viable As aresult alternate methods such as model order reductionand system level modeling (and model-based reasoning orimplementation) are still required in order to make some o the FEA studies viable in an industrial environment

Nonlinearities continue to orm an important part o many recent studies with the complexity and detail o thespeci1047297ed aults being continually expanded Te depth o modeling has enabled in many cases accurate validationagainst experimental approaches Despite this completeunderstanding o the vibrational phenomena o rotatingmachinery is not currently possible despite the drive towardsthis end

983091983095 Rotordynamics and IVHM Te bulk o current researchinto rotordynamics rom the point o view o prognostichealth management (PHM) can be roughly divided intotwo types initial single-ault diagnosisprognosis techniquesand studies into the general requirements and limitationso current systems along with current and uture trendsAn example o the latter is Pusey [983091983095] who provides agood summary overview o current diagnosis and prognosistechniques with regard to condition-based maintenance

As a result o this split a clear gap exists between thecore research being perormed into rotordynamics rom acondition-based maintenance perspective and the identi1047297edneeds o industry aking a 1047298edgling piece o research andapplying it to a commercially-ready system (eg a gas turbineengine or an aircraf) is a long and complex task It isnevertheless worth noting that technologies or automatically detecting an unbalance or misalignment in a gas turbine were

developed over 983089983088 years beore the latest commercial aircrafwere conceptualized and yet these planes are still limited inthis capacity Tis highlights the need or work which linksthe undamental research into individual ault diagnosis toldquolive systemsrdquo in use in industry

Physics-based simulation and modeling o rotordynamicparts is a well-researched 1047297eld Such modeling has been usedas the basis o diagnosis and prognosis o aults by many researchers several recent examples have been outlined inthis paper Occasional pieces o work have been perormedinto modeling multiple aults such as Jain and Kundra [983091983096]who usea system model oronline identi1047297cation o unbalanceand cracks Beyond this however very limited research

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 1013

983089983088 International Journal o Rotating Machinery

Business casecost bene1047297tanalysis

Live system

RubCracksBearingfailure

Rotor bow Unbalance

Fault differentiationLocalisation

Root cause analysis

Remaining useful life

Data driven analysis and physics-based simulation

Mis-alignment

Experimental

1047297eld data

F983145983143983157983154983141 983089 Physics-based simulationmdashrom research to industry

exists Te demands o PHM techniques in industry are suchthat any systemmust notonly be capable o detecting multipleaults but must also be capable o detecting these aultsacrossa range o different systems Other considerations includethe aorementioned ability to differentiate between multipleaults Processing also needs to be taken into account as theobjective o these systems is to implement efficient conditionmonitoring and condition-based maintenance procedures I processing data is a long power-hungry process then thisaim cannot be achieved

Figure 983089 details a potential ramework required in order

to push core research such as that detailed in this reporttowards industrial applications Many studies now exist onindividual rotordynamic aults across a wide range o condi-tions and applications Some studies have taken this urtherwith advanced prognostic models and diagnosis o dual aults(primary cause and secondary effect) Future research inthe area o rotordynamics rom a PHM perspective couldpotentially provide the bridge between these studies and livesystems by combining physics-based simulations with data-driven techniques and validating against experimental data

Although quantiying the success o the research studiedisdifficultit ispossibleto de1047297ne the key areas in whicha tech-nique must excel in order to be considered viable Te work

byWheeleretal[983091983097] discussesin detailmetrics ordiagnosticand prognostic analysis as does that by Vachtsevanos [983092983088]and Saxena et al [983092983089] Te conclusions drawn rom thesepapers and applied in practice to research like that covered by this paper indicate that the ollowing metrics are importantwhen considering the potential o a given technique to diag-nose aults coverage alse positive rate and alse negativerate In the case o prognosis probabilities and lead timeto ailure are other important considerations Tis criteriaenables research to be assessed in terms o its suitability orindustrial applications Unortunately inormation on these

metrics is not made readily available by the authors o mostpapers

In terms o evaluating the effectivenesso theresearch dis-cussed in this paper there are difficulties in recommending aspeci1047297c technique over others or general application Mostpapers reviewed or this document (not just those reerencedand discussed in detail) take a technique (new or evolved)

validate or a given speci1047297c system and report success o theresearch Despite this some conclusions can be drawn romassessing common techniques applied across different studiesand different aults Although it is not possible to de1047297ne themost common methods or diagnosing andprognosing aultsin terms o numbers (as it was not possible to cover all recent

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 1113

International Journal o Rotating Machinery 983089983089

rotordynamics research or this paper) as perceived by theauthors the ollowing techniques have eatured prominently in the reviewed research

(i) Sensors accelerometers

(ii) Teoretical studies mathematical modeling

(iii) Physics-based simulation FEA(iv) Data-driven joint timerequency domain analysis

(v) Implementation neural networking

Tese techniques appear to be among the most promisingcurrently under development as they tend to eature numer-ous timesamongst some o the work with wider scope acrossault types and with the most comprehensive validationTere are o course many subsections to these techniqueshowever it shows one general direction o research and theclear possibilities posed in these areas

Future developments in the 1047297eld o IVHM or rotat-ing machinery may incorporate these techniques alongsideextensive use o nonlinear modeling and multiple aultinteractions Te 1047297eld o design or IVHM has only recently emerged however the potential exists or speci1047297c nonlinear-ities to be designed into a system in order to enable accuratediagnosis andprognosis o aults Te development o currentalgorithms to include diagnosis localization and prognosiso a range o aults will provide a signi1047297cant advancement oruture generations o IVHM systems Tis combined withcost-effective sensor suites indicates the potential or evolu-tions o some o the research detailed hereto orm part onextgeneration IVHM suites or rotating machinery

4 Conclusion

Tis paper has reviewed some o the latest research around anumber o rotordynamic aults namely unbalance misalign-ment rub and looseness 1047298uid-induced instability bearingaults shaf cracks blade cracks and rotor bow Each aultwas reviewed rom the perspective o sensors diagnosisprognosis localization and modeling

Key examples o recent work into the eight describedaults have been detailed through works by a number o eminent authors Additional work has been summarized andormatted or easy reerence Some current trends amongstthe recent body o work include developments in the vastarea o modeling nonlinearities combinations o high andlow 1047297delity modeling and synergy between data driven and

physics-based simulation approachesDespite the large volume o promising research reviewed

urther development in a number o areas is required in orderto produce effective next generation IVHM systems As suchuture developments may include usion o andor devel-opment o current algorithms to encompass all eight aultsdetailed consideration o prognosis diagnosis and localiza-tion achieved using a reduced cost-effective sensor suite

References

[983089] I K Jennions Integrated Vehicle Health Management Perspec-tives on an Emerging Field 983090983088983089983089

[983090] G Anderson ldquoProviding best-value IVHM solutions or agingaircrafrdquo in Proceedings o the 983097th Joint FAADoDNASA Aging Aircraf Conerence vol 983089 pp 983089ndash983089983089 983090983088983088983093

[983091] A Muszynska Rotordynamics aylor amp Francis 983089st ed edition983090983088983088983093

[983092] D E Bently Fundimentals o Rotating Machinery Diagnostics

ASME Press 983089st edition 983090983088983088983090[983093] B Domes ldquoVibration phenomena in aero-enginesrdquo in Proceed-ings o the 983097th International Conerence on Vibrations in Rotating Machinery Exeter (IMechE rsquo983088983096) vol 983089 pp 983089983093ndash983091983090 September983090983088983088983096

[983094] A Hess G Calvello and Dabney ldquoPHM a keyenabler or theJSF autonomie logistics support conceptrdquo in Proceedings o theIEEE Aerospace Conerence Proceedings pp 983091983093983092983091ndash983092983093983092983097 March983090983088983088983092

[983095] S N Ganeriwala B Schwarz and M H Richardson ldquoOperat-ing de1047298ection shapes detect unbalance in rotating equipmentrdquoSound and Vibration vol 983092983091 no 983093 pp 983089983089ndash983089983091 983090983088983088983097

[983096] G N D S Sudhakar and A S Sekhar ldquoIdenti1047297cation o unbal-ance in a rotor bearing systemrdquo Journal o Sound and Vibration

vol 983091983091983088 no 983089983088 pp 983090983090983097983097ndash983090983091983089983091 983090983088983089983089[983097] Yang and M W Hsu ldquoAn efficient diagnosis technique or

variations o shaf-bow and unbalancerdquo in Proceedings o the ASME International Design Engineering echnical Conerencesand Computers and Inormation in Engineering Conerence(DEC rsquo983088983097) pp 983093983095ndash983094983094 San Diego Cali USA September 983090983088983088983097

[983089983088] A El-Shaei S H aw1047297ck and M O A Mokhtar ldquoExperi-mental investigation o the effect o angular misalignemt onthe instability o plain journal bearingsrdquo in Proceedings o the ASMESLE InternationalJointribology Conerence(IJC rsquo983088983097)pp 983089983095983089ndash983089983095983091 Memphis enn USA October 983090983088983088983097

[983089983089] A El-Shaei S H aw1047297ck and M O A Mokhtar ldquoMisalign-ment modeling in rotating systemsrdquo in Proceedings o the ASMEurbo Expo H Bahaloo A Ebrahimi and M Samadi Eds vol

983089983095983089983089 pp 983097983095983091ndash983097983095983097 Orlando Fla USA June 983090983088983088983097

[983089983090] L F Villaet al ldquoStatistical diagnosisbased on vibration analysisor gear test-bench under non-stationary conditions o speedand loadrdquo Mechanical Systems and Signal Processing vol 983090983097 pp983092983091983094ndash983092983092983094 983090983088983089983090

[983089983091] C F Ngolah et al ldquoIntelligent ault recognition and diagnosisor rotating machines using neural networksrdquo Sofware Scienceand Computational Intelligence vol 983091 no 983092 983089983095 pages 983090983088983089983089

[983089983092] Y Lu Z Ren H Chen N Song and B Wen ldquoStudy on loose-ness and impactmdashrub coupling aults o a vertical dual-disk cantilever rotor- Bearing systemrdquo Key Engineering Materials vol 983091983093983091ndash983091983093983096 no 983092 pp 983090983092983095983097ndash983090983092983096983090 983090983088983088983095

[983089983093] Q Han Z Zhang and B Wen ldquoPeriodic motions o a dual-

disc rotor system with rub-impact at 1047297xed limiterrdquo Proceedingso the Institution o Mechanical Engineers vol 983090983090983090 no 983089983088 pp983089983097983091983093ndash983089983097983092983094 983090983088983088983096

[983089983094] H F de Castro K L Cavalca and R Nordmann ldquoWhirland whip instabilities in rotor-bearing system considering anonlinear orce modelrdquo Journal o Sound and Vibration vol 983091983089983095no 983089-983090 pp 983090983095983091ndash983090983097983091 983090983088983088983096

[983089983095] C C Fan et al ldquoMechanical systems and signal processingrdquoStudy o Start-Up Vibration Response or Oil Whirl and Dry Whip vol 983090983093 pp 983091983089983088983090ndash983091983089983089983093 983090983088983089983089

[983089983096] K Kappaganthu and C Nataraj ldquoNonlinear modeling andanalysis o a rolling element bearing with a clearancerdquo Commu-nications in Nonlinear Science and Numerical Simulation vol983089983094no 983089983088 pp 983092983089983091983092ndash983092983089983092983093 983090983088983089983089

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 1213

983089983090 International Journal o Rotating Machinery

[983089983097] C Gupta K Gupta and D K Sehgal ldquoInstability and chaoso a 1047298exible rotor ball bearing system an investigation on thein1047298uence o rotating imbalance and bearing clearancerdquo Journal o Engineering or Gas urbines and Power vol 983089983091983091 no 983096ArticleID 983088983096983090983093983088983089 983089983089 pages 983090983088983089983089

[983090983088] J Hong X Miao L Han and Y Ma ldquoPrognostics model orpredicting aero-engine bearing grade-lierdquo in Proceedings o the ASME urbo Expo vol 983089 pp 983094983091983097ndash983094983092983095 Orlando Fla USA June983090983088983088983097

[983090983089] P Borghesani P Pennacchi R B Randall N Sawalhi and RRicci ldquoApplication o cepstrum pre-whitening or the diagnosiso bearing aults under variable speed conditionsrdquo Mechanical Systems and Signal Processing 983090983088983089983090

[983090983090] Y Li J Zhang L Dai Z Zhang and J Liu ldquoAuditory-model-based eature extraction method or mechanical aults diagno-sisrdquo ChineseJournal o MechanicalEngineering (EnglishEdition) vol 983090983091 no 983091 pp 983091983097983089ndash983091983097983095 983090983088983089983088

[983090983091] N Bachschmid P Pennacchi and E anzi ldquoCracked rotatingshafs typical behaviors modeling and diagnosisrdquo IUAM Symposium on Emerging rends in Rotor Dynamics vol 983089983088983089983089 pp

983092983092983089ndash983092983093983092 983090983088983089983089[983090983092] Inoue N Nagata and Y Ishida ldquoFem modelling and

experimental veri1047297cation o a rotor system with a open crackrdquoin Proceedings o the ASME International Design Engineering echnical Conerences and Computers and Inormation in Engi-neering Conerence (DEC rsquo983088983097) pp 983089983089983089983091ndash983089983089983090983090 San Diego CaliUSA September 983090983088983088983097

[983090983093] H E Sonnichsen ldquoReal-time detection o developing cracksin jet engine rotorsrdquo in Proceedings o the IEEE AerospaceConerence pp 983089983095983091ndash983089983096983092 March 983090983088983088983088

[983090983094] I Green and C Casey ldquoCrack detection in a rotor dynamicsystem by vibration monitoringmdashpart I analysisrdquo Journal o Engineering or Gas urbines and Power vol 983089983090983095 no 983090 pp 983092983090983093ndash983092983091983094 983090983088983088983093

[983090983095] J Sawicki M I Friswell A H Pesch and A WroblewskildquoCondition monitoring o rotor usingactive magnetic actuatorrdquoin Proceedings o the ASME urbo Expo pp 983089983090983093983095ndash983089983090983094983093 BerlinGermany June 983090983088983088983096

[983090983096] J Xiang X Chen Q Mo and Z He ldquoIdenti1047297cation o crack ina rotor system based on wavelet 1047297nite element methodrdquo FiniteElements in Analysis and Design vol 983092983091 no 983089983092 pp 983089983088983094983096ndash983089983088983096983089983090983088983088983095

[983090983097] M G MaalouldquoSlow speed vibrationsignal analysisi you canrsquotdo it slow you canrsquot do it astrdquo in Proceedings o the ASME urboExpo pp 983093983093983097ndash983093983094983095 Montreal Canada May 983090983088983088983095

[983091983088] J Meagher X Wu and C Lencioni ldquoResponse o a warped1047298exible rotor with a 1047298uid bearingrdquo International Journal o

Rotating Machinery vol 983090983088983088983096 Article ID 983089983092983095983094983093983091 983097 pages 983090983088983088983096

[983091983089] Gaka andM abaszewski ldquoAn applicationo statistical symp-toms in machine condition diagnosticsrdquo Mechanical Systemsand Signal Processing vol 983090983093 no 983089 pp 983090983093983091ndash983090983094983093 983090983088983089983089

[983091983090] J K Sinha ldquoRecent trands in ault quanti1047297cation in rotatingmachinesrdquo Advances in Vibration EngIneerIng vol 983096 no 983089 pp983095983097ndash983096983093 983090983088983088983097

[983091983091] X Shen JJia andM Zhao ldquoNonlinear analysis o a rub-impactrotor-bearing system with initial permanent rotor bowrdquo Archiveo Applied Mechanics vol 983095983096 no 983091 pp 983090983090983093ndash983090983092983088 983090983088983088983096

[983091983092] A W Lees J K Sinha and M I Friswell ldquoModel-based identi-1047297cation o rotating machinesrdquo Mechanical Systems and Signal Processing vol 983090983091 no 983094 pp 983089983096983096983092ndash983089983096983097983091 983090983088983088983097

[983091983093] L C Jaw and W Merrill ldquoCBM+ research environmentmdashacilitating technology development experimentation andmaturationrdquo in Proceedings o the IEEE Aerospace Conerence(AC rsquo983088983096) March 983090983088983088983096

[983091983094] R Walker S Perinpanayagam and I K Jennions ldquoSimulatingUnbalance or Future IVHM Applicationsrdquo in Proceedings o theSociety or Experimental Mechanics (IMAC rsquo983089983090) vol 983090983097 pp 983089983092983089ndash983089983092983096 May 983090983088983089983090

[983091983095] H C Pusey ldquourbomachinery condition monitoring and ail-ure prognosisrdquo Sound and Vibration vol 983092983089 no 983091 pp 983089983088ndash983089983093983090983088983088983095

[983091983096] J R Jain and K Kundra ldquoModel based online diagnosiso unbalance and transverse atigue crack in rotor systemsrdquo Mechanics Research Communications vol 983091983089 no 983093 pp 983093983093983095ndash983093983094983096983090983088983088983092

[983091983097] K R Wheeler Kurtoglu and S D Poll ldquoA survey o healthmanagement user objectives related to diagnostic and prognos-tic metricsrdquo in Proceedings o the 983090983097th Computers and Inor-mation Engineering Conerence vol 983090 pp 983089983090983096983095ndash983089983090983097983096 983090983088983088983097

[983092983088] G Vachtsevanos ldquoPerormance metrics or ault prognosis o

complex systemsrdquo in IEEE Systems Readiness echnology Con- erence (AUOESCON rsquo983088983091) pp 983091983092983089ndash983091983092983093 September 983090983088983088983091

[983092983089] A Saxena J Celaya E Balaban et al ldquoMetrics or evaluatingperormance o prognostic techniquesrdquo in Proceedings o theInternational Conerence on Prognostics and Health Manage-ment (PHM rsquo983088983096) pp 983089ndash983089983095 October 983090983088983088983096

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 1313

Submit your manuscripts at

httpwwwhindawicom

Page 6: Rotordynamic Faults 2013

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 613

983094 International Journal o Rotating Machinery

have been researched in somewhat more detail than someo the other aults detailed here Despite this much work still needs to be perormed in order to translate some o this core research into industrial applications Research suchas that detailed above has made signi1047297cant advances intodetermining bearing ailure as the root cause o malunction

Detecting which bearing is ailing across a complex systemhas received somewhat less research Bearing prognostics isanother area with much ongoing research being perormedmdashboth in the simulation and data-driven domains o givean example Hong et al [983090983088] combined grade lie andextensive mathematical modeling techniques in order to pro-duce prognostic models or aeroengine bearings Te resultsare described by the authors as ldquopractical and veri1047297ablerdquoAlthough a number o similar recent studies existthis work iso note or the extent o the studies perormed which includebearing test stand run-to-ailure validation Te lab resultsappear impressive this research has yet to be applied andtested in real lie applicationsmdashindicating that despite thenumber o parameters considered it is still not possible tomodel naturally occurring phenomena sufficiently

A large body o work in this area also exists romBorghesani et al [983090983089] where several examples o data driventechniques can be seen applied to a wide variety o bearingtypes An example o recent developments involves the use o ldquocepstrum prewhiteningrdquo in order to remove sufficient noiseor accurate bearing diagnosis and prognosis Tis work isparticularly noteworthy due to the emphasis on real-worldapplications where the traditional lab-based techniques o order tracking and synchronous averaging do not providesufficient noise removal or harsh industrial environmentsTe addition o such techniques is a crucial step in developingthe current generation o diagnosis and prognosis algorithmsor use in uture IVHM systems

983090983094 Shaf Cracks Another potentially serious ault in rotat-ing machinery is shaf cracks and so early detection o any such ault is highly important Methods o crack ormationandpropagation canbe diverse andrange rom highand low-cycle atigue to stress corrosion Simulation and modelingo shaf cracks can have signi1047297cant advantages over data-driven methods Perhaps the most obvious advantage is therelative simplicity o inserting a ault into orexample a 1047297niteelement model as opposed to seeding a ault in a workingindustrial machine As such research into shaf cracks has

been progressing steadily with the corresponding increasesin computing power

A clear synergy between data-driven and physics-basedsimulation research can be implied by a number o recentworks o research An example o recent advances rom a datacollection perspective is Li et al [983090983090] which details statisticalmodels based on historical data or condition monitoringpurposes Tis unique work uses the human auditory systemas inspiration or enriching methods o mechanical aults andeatures extraction Te results indicated by the paper areperhaps not as convincing as some other methods discussedin this paper it describes an interesting ldquooutside the boxrdquomethod o tackling common problems

From a modeling perspective Bachschmid et al [983090983091]cover a wide range o vibration phenomena in order todevelop a model-based identi1047297cation and severity procedureTis work is noted or its thoroughness in modeling pro-cedure including accurate modeling o the crack breathingmechanism A combination o high and low 1047297delity models

is validated through experimental study and ldquoexcellentrdquoaccuracy is claimed by the authors in detecting crack positionand depth through the use o the proposed model-baseddiagnostics

Te nature o shaf cracks has resulted in a wide variety o research beingperormed intoboth localization and prognos-tics o these aults (indeed the two topics can be consideredrelated) Recent examples o work in this area include Inoueetal[983090983092] who detail crack localization using orced responsemodeling In this case a test rig was constructed consistingo a circular shaf supported by two bearings Frequency-domain data was used to create a localization algorithmdesigned in combination with an FE model Although thisresearch provides good accuracy o localization in a labenvironment it remains untested in a more complex system(eg a ull gas turbine)

Sonnichsen [983090983093] describes 1047297nite element modeling o crack propagation with validation against experimentalresults provided to demonstrate the validity o such modelingtechniques Te paper concentrates on natural requenciesand resonance curves Whilst the authors claim improve-ments in the ability to understand such aults again thesystem in question is quite simplemdashand such FEA models aredifficult to scale up to ull size applications

983090983095 Blade Cracks Blade cracks i allowed to develop canresult in serious consequences Cracks can orm due to highcentriugal stresses across operational cycles (in the case o an aircraf gas turbine or example start up and take off through landing and taxi) As excessive crack growth canlead to catastrophic rotorblade ailure early detection andprognosis o such aults are essential As with shaf cracksphysics-driven simulation o blade cracks is an area o sig-ni1047297cant research Tis varies rom high-1047297delity 1047297nite elementmodels to low-1047297delity system and mathematical models Terecent work demonstrated by Green and Casey [983090983094]is agoodexample o recent mathematical modeling rom a diagnosisperspective In this paper the authors concentrate on early detection using global and local asymmetry crack models 983090X

harmonic components are identi1047297ed as key areas or the early detection o blade cracks however again this paper suffersrom being applied and tested on a relatively simple systemwhich may not scale up to a ull size turbine

Sonnichsen [983090983093] demonstrated high-1047297delity modelingthe authors used FEA to model crack growth making com-parisons and validating against an experimental rig Tiswork is particularly interesting as it outlines the advantagesand drawbacks with the latest state-o-the-art modelingtechniques

Localization and prognosis o blade cracks have alsobene1047297tted rom recent advances in simulation and modelingSawicki et al [983090983095] contains details o work on a novel active

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 713

International Journal o Rotating Machinery 983095

magnetic bearing system or use in the early detection local-ization and prognosis o blade cracks Again the emphasis ison early detection with the bearings used to excite the systemin order to obtain optimum response vibrations or analysisTe authors admit the approach hassome merit in diagnosingblade cracks however it is in the early stages o development

and work are ongoing FEA has also been used extensively to support blade crack prognostic tools Xiang et al [983090983096] arean extensive example o recent work In this case a numbero advanced FEA methods are applied to produce accurateFEA solutionsmdashthese include surace-1047297tting techniques andthe contour-plotting method Te authors experimentally

validate their work and suggest that it can be applied toprognosis and quantitative diagnosis o blade cracks Despitethe claimed advances again the scalability o such research toull size turbines is an issuemdashparticularly with regard to thecomplexity o the FEA models

983090983096 Rotor Bow Rotor bows can be a primary source o unwanted vibration in gas turbines Te main cause o arotor bow (rotor bows do not include bows due to gravity) isthermal differences in a system caused by operating con-ditions It is noted by Domes [983093] that this nonsymmetricalthermal distribution can cause excessive unbalance to theextent where a gas turbine will not start correctly Such rotorbows are common on start up or shut down and are ofenaccounted or in operational procedures However i thermalldquohot spotsrdquo exceed a given tolerance level they can causepermanent unbalances due to rotor de1047298ections Such rotorbows can lead to other aults includingrubbing andloosenesswhich complicate isolation and localization

raditional data-driven techniques or detecting rotorbows involve combinations o slow roll and vibration data[983090983097] More recently mathematical modeling techniques suchas that detailed by Meagher et al [983091983088] have been used inorder to diagnose residual rotor bows and differentiate theseaults rom other sources o unbalance Te authors o thispaper build upon established methods or the models and areunique in that they concentrate on response at the bearingpoints Tis is perhaps more useul orindustrial applicationsas vibration inormation is more readily available at bearingpoints rather than intrusive proximity probes which are ofenused in lab work

Te little work that exists on attempting to localise rotorbows across complex systems tends to be data-driven in

nature see Gaka and abaszewski [983091983089] where the authorsused statistical symptoms based on known data as a methodo diagnosing and prognosing a number o aults includingrotor bows and unbalance Tis paper attempts to address theirregularities and 1047298uctuations that occur over a long servicelie In order to achieve this a modi1047297ed energy processormodel is created using data drawn rom large steam turbinesover long periods o lie

Prognosing rotor bows is a complex subject As rotorbows are ofen caused by temperature de1047298ections makingpredictions or remaining useul lie and potential utureproblems lies not only in the realm o mechanical rotor-dynamics but also to some extent in thermodynamics Te

recent work detailed by Sinha [983091983090] is o note or detailingdiagnosis and quanti1047297cation o various rotordynamic aultsand describing the advantages o mathematical modelingover traditional vibration-based approaches Te topic o scalability regarding FEA models is discussed including anargument or the use o partial (simpli1047297ed) mathematical

models or large complex systemsAnother two works which are o interest with regard to

modeling o rotor bows include Shen et al [983091983091] where theauthors modeled a rotor-bearing system with a permanentrotor bow looking at the impact o secondary aults such asrub Te study o ault combinations in this paper is useulor ault differentiation studies however the authors study a permanent initial rotor bow Tis thereore does not takeinto account developing or worsening aults and the different

vibration phenomena that are observed as such aults aredeveloping

Lees et al [983091983092] describe the importance o model-basedault identi1047297cation techniques and outline recent research in

the area providing a good reerence paper or more researchon this speci1047297c ault

3 Discussion

983091983089 Sensor Suites Te subject o sensor suites is o greatimportance with regard to uture industrial applications Onone hand advanced and complex sensor suites generally enable improved ault localization and diagnosis howeverthe added complexity and cost has resulted in many o thesesystems being omitted rom the latest generations o rotatingmachinery in industrial applications From an industrialperspective the cost bene1047297t o additional sensors needs to

be signi1047297cant in order to justiy this approach In response anumber o works listed including [983096] as a good example takethe approach o achieving the same objectives using a greatly reduced sensor suite In many ways the required sensor suiteor many o the currently researched ault diagnosis andprognosis techniques provides an indication o the ability o the techniques to be used practically in industrial situationsA complex suite may return a very high success rate on alab-based rig however the impracticalities o mounting suchsuites on a real-world application negate the advantages

Further to this the consideration orsensor position mustbe considered Te use o proximity probes and keyphasortransducers has very clear advantages in a number o situ-

ations however the intrusive nature o the sensor preventsapplication in a number o complex systems With this inmind a number o the more promising studies listed rely onthe simple suiteo an accelerometer placed in positions whichare relatively remote rom the sources o vibration potentially with a noisy transer path

983091983090 Diagnosis It can be seen rom the research outlined inthe previous section that the diagnosis o aults in rotatingmachinery is a subject o much ongoing research Tisinvolves the improvement and development o traditional

vibration monitoring techniques development o new data-driven technologies and novel research into physics-based

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 813

983096 International Journal o Rotating Machinery

simulation and modeling In many cases these topics o research are dependent on one another or reasons o vali-dation veri1047297cation and speed o analysis In several casesit can be seen that multiple aults have been modeled orthe purposes o identi1047297cation and isolation However nostudies have yet been perormed which deal with all o the

aorementioned aults All o these aults are intrinsically related to one another Complex combinations o aults havebegun to be analyzed with the emphasis on developing new diagnosis techniques Physics-based modeling has proven toprovide signi1047297cant advances with regard to speci1047297c aultsnotably shaf and blade cracks where techniques such as FEAenable much easier aster and cheaper test data then seedingaults into live systems

It can be noted however that both high and low-1047297delity modeling techniques are being applied to cutting edgeresearch or all o the listed rotordynamic aults (and othersnot detailed in this paper) In addition to the advantages inthe speed o obtaining test results physics-based simulationis providing another dimension to data-driven techniquesSystem models are being used as part o logic and reasoningsuites in the identi1047297cation and differentiation o variousaults High-1047297delity models enable simulations o ault com-binations which are not possible practically or are pro-hibitively expensive in live systems

Nevertheless results obtained rom modeling studies stillneed to be validated against proven data-driven techniquesbeore implementation in industrial applications is possibleIt is also o note rom the literature reviewed or this paperthat almost all modeling or the diagnosis o aults involvesextensive simpli1047297cation ofen reducing potentially complexsystems down to one or two discshafbearing modelsAdapting the claimed results rom such research to systemswith many discsbearingsshafs is important in improvingexisting rotating machinery diagnostic techniques

Te inclusion o nonlinear effects is notable amongstmany authorsmdashas the understanding o the complex aultsand interactions continues to progress Such research opensthe potential or promising new avenues into areas o designor IVHM

Finally it is o note that despite the wide variety o advanced research here a consensus or the most efficienttechniques and algorithms has yet to emerge in order or the1047297nal adaption to industrial application to be achieved How-ever the consideration or industrial application taken by many o the authors detailed in this paper provides promise

or next generation IVHM systems

983091983091 Localisation Fault localization in rotating machinery isan important topic o research or uture condition-basedmonitoring systems Knowing not only what type o aulthas occurred but also where in the system is an importantconsideration which can in1047298uence maintenance proceduresin complex machinery It is worth noting rom the literaturesurveyed or this paper that many studies have ocusedon diagnosis and prognosis o single rotorbearing systemsOfen or legitimate reasonsmdashsimpli1047297cation or computingspeed or example Few studies however have taken intoaccount the localization o aultsacross whole systems Oneo

the more extensive examples highlighted in this paper (Hanet al [983089983093]) provides extensive analytical studies relevant tolocalization on a duel disk setup utilizing Hilbert-Huangransorms however in this example comprehensive valida-tion is lacking thus limiting the potential applicability o theresearch

Tis problem is not limited to modeling and simulation-based research Many newly developed data-driven tech-niques or diagnostics and prognostics claim good results by heavily instrumenting speci1047297c components o a test systemIn many industrial cases this is not possible practical or cost-effective

o give some examples a keyphasor transducer can beparticularly useul in diagnosing aults such as rotor bowHowever this equipment requires the ability to cut a key way or measurements to be perormed Optical sensors haverecently been applied to detect rotor unbalance yet such atechnology would be difficult to implement in a system withseveral rows o rotors (as in a gas turbine)

Whilst new diagnosis and prognosis techniques or aultsin rotating machinery are being continuously researchedthe lack o corresponding studies into localization can beconsidered one o the many challenges in promoting recentcore research into live industrial applications

983091983092 Fault Differentiation One promising development inrecent modeling and simulation o rotordynamic aults isthat o ault differentiation Several researchers have movedon rom studies on individual aults in order to concentrateon combinations o aults As stated rotordynamic aultssuch as those listed in this paper are linked to each otherSuch studies thereore concentrate on such topics as a

misalignment causing an unbalance or a looseness causinga rub Findings rom reports such as these are important inunderstanding complex anomalies In industrial applicationssimply detecting and rectiying an unbalance does not pro-

vide a satisactory solution i the root cause o the ault isa misalignment Tis is a complex topic as several aults canexhibit similar vibration characteristics making traditionaldetection techniques inaccurate in some cases

Making reerence to some o the aorementioned stud-ies those such as Bachschmid et al [983090983091] and Sonnichsen[983090983093] indicate some initial studies including several aultsTe results o such studies indicate a high success rate indifferentiating the aults in the given conditions however

a number o limitations still exist Studies such as theseconsider unbalancein the aultchains Tisis typically ldquostaticrdquoor simple unbalance applied to a single shafrotormdashin reality unbalance aults can be more complex than this In additionthe simple systems used to study and differentiate aultsare very different rom the complexities o a ull turbineHowever urther work may enable such research to ldquoscale-uprdquo to this stage with the aid o urther validation and

veri1047297cation

983091983093 Prognosis It can be seen rom the examples o recentresearch described or common aults that prognostic tech-niques are a topic where much work is being perormed

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 913

International Journal o Rotating Machinery 983097

Some o this work would be very difficult or impossibleto implement in an industrial situation others providepromising results o use to uture work in the area Predictingthe remaining useul lie o components is critical in thedevelopment o condition-based monitoring strategies orindustrial implementation

It can be noted that prognostic studies or certain rotor-dynamic aults are considerably more advanced than orothers Te most obvious examples o the aults detailed inthis paper are bearing aults shaf and blade cracks Teseare also areas where physics-driven simulations have hadan important impact Te ability to design any ault type(or combinations o aults) into such simulations (to be itslow-1047297delity mathematical models or high-1047297delity 1047297nite ele-ment analysis) has provided researchers with many differentavenues to explore Studies into the prognostics o otheraults can be complicated by various actors An unbalanceor example has had relatively little research perormed intoprognosis Te act that unbalance is ofen the cause o another underlying ault is one reason or the difficulty inresearching prognosis in detail or this ault Te results o unbalance are also very dependent on the severity o the aultand the system in which it occurs

It is worth noting that even or the aults where prognosisresearch is more advanced (eg shaf cracks) true develop-ment o the ault is not always studied Fan et al [983089983095] orexample study sets o discrete cracks in order to ldquoprognoserdquocrack growth In act thisstudy is more akinto detecting aultseverity with inerred effects on crack growth In contrast Liet al [983090983090] consider the continual development o such crackshowever the authors do admit this work is in its inancySuch points limit the ability or true prognosis to be achievedwithout extensive operational data

It is also worth noting that combinations o aults havebeen simulated extensively or the purpose o diagnosticshowever ew studies exist combining aultsor the purpose o prognostics Such studies are however a logical progressionrom some o those already perormed It may be that suchprognostic studies build upon some o the condition-basedmaintenance (CBM) or diagnostics High-level studies suchas Jaw and Merrill [983091983093] indicate how prognosis and diagnosistechniques can be combined into a CBM system Tis study is designed or a military aeroengine indicating the desire orsuch systems to be implementedmdashalthough the architecturedescribed has the capability o including various algorithmsoraultprognosis the amount o consideration orprognosis

o ault combinations is unclear

983091983094 Modelling Te large number o works reviewed or thispaper which includes modeling techniques in order to aiddiagnosis and prognosis indicates the possibilities providedby modern computer power and sofware developmentBroadly the modeling studied in the aorementioned papersbreaks down into two categories mathematical modeling and1047297nite element analysis (FEA) Te mathematical models suchasBorghesani etal [983090983089] ofen provide a theoretical basis uponwhich data driven or FEA studies can build and validateMathematical models despite being a traditional approach to

rotordynamic problems thereore continue to be developedto tackle more advanced problems

FEA techniques offer a constantly expanding area o simulation to explore Tey are used or research as diverseas crack propagation to unbalance localization [983091983094] Moderncodes such as NASRAN and Ansys enable rotordynamics to

be studied beyond the traditional stress and modal analysisTe ability to study the effects o rotation (eg throughCampbell diagrams and transient analysis) continues to driveresearch and innovation in this area Te studies discussedin this paper indicate the power o FEA with some highly accurate simulations having been perormed However thereis still a limitation o computer powermdashapplying some o the FEA techniques to whole engine models would proveto be too computationally expensive to be viable As aresult alternate methods such as model order reductionand system level modeling (and model-based reasoning orimplementation) are still required in order to make some o the FEA studies viable in an industrial environment

Nonlinearities continue to orm an important part o many recent studies with the complexity and detail o thespeci1047297ed aults being continually expanded Te depth o modeling has enabled in many cases accurate validationagainst experimental approaches Despite this completeunderstanding o the vibrational phenomena o rotatingmachinery is not currently possible despite the drive towardsthis end

983091983095 Rotordynamics and IVHM Te bulk o current researchinto rotordynamics rom the point o view o prognostichealth management (PHM) can be roughly divided intotwo types initial single-ault diagnosisprognosis techniquesand studies into the general requirements and limitationso current systems along with current and uture trendsAn example o the latter is Pusey [983091983095] who provides agood summary overview o current diagnosis and prognosistechniques with regard to condition-based maintenance

As a result o this split a clear gap exists between thecore research being perormed into rotordynamics rom acondition-based maintenance perspective and the identi1047297edneeds o industry aking a 1047298edgling piece o research andapplying it to a commercially-ready system (eg a gas turbineengine or an aircraf) is a long and complex task It isnevertheless worth noting that technologies or automatically detecting an unbalance or misalignment in a gas turbine were

developed over 983089983088 years beore the latest commercial aircrafwere conceptualized and yet these planes are still limited inthis capacity Tis highlights the need or work which linksthe undamental research into individual ault diagnosis toldquolive systemsrdquo in use in industry

Physics-based simulation and modeling o rotordynamicparts is a well-researched 1047297eld Such modeling has been usedas the basis o diagnosis and prognosis o aults by many researchers several recent examples have been outlined inthis paper Occasional pieces o work have been perormedinto modeling multiple aults such as Jain and Kundra [983091983096]who usea system model oronline identi1047297cation o unbalanceand cracks Beyond this however very limited research

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 1013

983089983088 International Journal o Rotating Machinery

Business casecost bene1047297tanalysis

Live system

RubCracksBearingfailure

Rotor bow Unbalance

Fault differentiationLocalisation

Root cause analysis

Remaining useful life

Data driven analysis and physics-based simulation

Mis-alignment

Experimental

1047297eld data

F983145983143983157983154983141 983089 Physics-based simulationmdashrom research to industry

exists Te demands o PHM techniques in industry are suchthat any systemmust notonly be capable o detecting multipleaults but must also be capable o detecting these aultsacrossa range o different systems Other considerations includethe aorementioned ability to differentiate between multipleaults Processing also needs to be taken into account as theobjective o these systems is to implement efficient conditionmonitoring and condition-based maintenance procedures I processing data is a long power-hungry process then thisaim cannot be achieved

Figure 983089 details a potential ramework required in order

to push core research such as that detailed in this reporttowards industrial applications Many studies now exist onindividual rotordynamic aults across a wide range o condi-tions and applications Some studies have taken this urtherwith advanced prognostic models and diagnosis o dual aults(primary cause and secondary effect) Future research inthe area o rotordynamics rom a PHM perspective couldpotentially provide the bridge between these studies and livesystems by combining physics-based simulations with data-driven techniques and validating against experimental data

Although quantiying the success o the research studiedisdifficultit ispossibleto de1047297ne the key areas in whicha tech-nique must excel in order to be considered viable Te work

byWheeleretal[983091983097] discussesin detailmetrics ordiagnosticand prognostic analysis as does that by Vachtsevanos [983092983088]and Saxena et al [983092983089] Te conclusions drawn rom thesepapers and applied in practice to research like that covered by this paper indicate that the ollowing metrics are importantwhen considering the potential o a given technique to diag-nose aults coverage alse positive rate and alse negativerate In the case o prognosis probabilities and lead timeto ailure are other important considerations Tis criteriaenables research to be assessed in terms o its suitability orindustrial applications Unortunately inormation on these

metrics is not made readily available by the authors o mostpapers

In terms o evaluating the effectivenesso theresearch dis-cussed in this paper there are difficulties in recommending aspeci1047297c technique over others or general application Mostpapers reviewed or this document (not just those reerencedand discussed in detail) take a technique (new or evolved)

validate or a given speci1047297c system and report success o theresearch Despite this some conclusions can be drawn romassessing common techniques applied across different studiesand different aults Although it is not possible to de1047297ne themost common methods or diagnosing andprognosing aultsin terms o numbers (as it was not possible to cover all recent

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 1113

International Journal o Rotating Machinery 983089983089

rotordynamics research or this paper) as perceived by theauthors the ollowing techniques have eatured prominently in the reviewed research

(i) Sensors accelerometers

(ii) Teoretical studies mathematical modeling

(iii) Physics-based simulation FEA(iv) Data-driven joint timerequency domain analysis

(v) Implementation neural networking

Tese techniques appear to be among the most promisingcurrently under development as they tend to eature numer-ous timesamongst some o the work with wider scope acrossault types and with the most comprehensive validationTere are o course many subsections to these techniqueshowever it shows one general direction o research and theclear possibilities posed in these areas

Future developments in the 1047297eld o IVHM or rotat-ing machinery may incorporate these techniques alongsideextensive use o nonlinear modeling and multiple aultinteractions Te 1047297eld o design or IVHM has only recently emerged however the potential exists or speci1047297c nonlinear-ities to be designed into a system in order to enable accuratediagnosis andprognosis o aults Te development o currentalgorithms to include diagnosis localization and prognosiso a range o aults will provide a signi1047297cant advancement oruture generations o IVHM systems Tis combined withcost-effective sensor suites indicates the potential or evolu-tions o some o the research detailed hereto orm part onextgeneration IVHM suites or rotating machinery

4 Conclusion

Tis paper has reviewed some o the latest research around anumber o rotordynamic aults namely unbalance misalign-ment rub and looseness 1047298uid-induced instability bearingaults shaf cracks blade cracks and rotor bow Each aultwas reviewed rom the perspective o sensors diagnosisprognosis localization and modeling

Key examples o recent work into the eight describedaults have been detailed through works by a number o eminent authors Additional work has been summarized andormatted or easy reerence Some current trends amongstthe recent body o work include developments in the vastarea o modeling nonlinearities combinations o high andlow 1047297delity modeling and synergy between data driven and

physics-based simulation approachesDespite the large volume o promising research reviewed

urther development in a number o areas is required in orderto produce effective next generation IVHM systems As suchuture developments may include usion o andor devel-opment o current algorithms to encompass all eight aultsdetailed consideration o prognosis diagnosis and localiza-tion achieved using a reduced cost-effective sensor suite

References

[983089] I K Jennions Integrated Vehicle Health Management Perspec-tives on an Emerging Field 983090983088983089983089

[983090] G Anderson ldquoProviding best-value IVHM solutions or agingaircrafrdquo in Proceedings o the 983097th Joint FAADoDNASA Aging Aircraf Conerence vol 983089 pp 983089ndash983089983089 983090983088983088983093

[983091] A Muszynska Rotordynamics aylor amp Francis 983089st ed edition983090983088983088983093

[983092] D E Bently Fundimentals o Rotating Machinery Diagnostics

ASME Press 983089st edition 983090983088983088983090[983093] B Domes ldquoVibration phenomena in aero-enginesrdquo in Proceed-ings o the 983097th International Conerence on Vibrations in Rotating Machinery Exeter (IMechE rsquo983088983096) vol 983089 pp 983089983093ndash983091983090 September983090983088983088983096

[983094] A Hess G Calvello and Dabney ldquoPHM a keyenabler or theJSF autonomie logistics support conceptrdquo in Proceedings o theIEEE Aerospace Conerence Proceedings pp 983091983093983092983091ndash983092983093983092983097 March983090983088983088983092

[983095] S N Ganeriwala B Schwarz and M H Richardson ldquoOperat-ing de1047298ection shapes detect unbalance in rotating equipmentrdquoSound and Vibration vol 983092983091 no 983093 pp 983089983089ndash983089983091 983090983088983088983097

[983096] G N D S Sudhakar and A S Sekhar ldquoIdenti1047297cation o unbal-ance in a rotor bearing systemrdquo Journal o Sound and Vibration

vol 983091983091983088 no 983089983088 pp 983090983090983097983097ndash983090983091983089983091 983090983088983089983089[983097] Yang and M W Hsu ldquoAn efficient diagnosis technique or

variations o shaf-bow and unbalancerdquo in Proceedings o the ASME International Design Engineering echnical Conerencesand Computers and Inormation in Engineering Conerence(DEC rsquo983088983097) pp 983093983095ndash983094983094 San Diego Cali USA September 983090983088983088983097

[983089983088] A El-Shaei S H aw1047297ck and M O A Mokhtar ldquoExperi-mental investigation o the effect o angular misalignemt onthe instability o plain journal bearingsrdquo in Proceedings o the ASMESLE InternationalJointribology Conerence(IJC rsquo983088983097)pp 983089983095983089ndash983089983095983091 Memphis enn USA October 983090983088983088983097

[983089983089] A El-Shaei S H aw1047297ck and M O A Mokhtar ldquoMisalign-ment modeling in rotating systemsrdquo in Proceedings o the ASMEurbo Expo H Bahaloo A Ebrahimi and M Samadi Eds vol

983089983095983089983089 pp 983097983095983091ndash983097983095983097 Orlando Fla USA June 983090983088983088983097

[983089983090] L F Villaet al ldquoStatistical diagnosisbased on vibration analysisor gear test-bench under non-stationary conditions o speedand loadrdquo Mechanical Systems and Signal Processing vol 983090983097 pp983092983091983094ndash983092983092983094 983090983088983089983090

[983089983091] C F Ngolah et al ldquoIntelligent ault recognition and diagnosisor rotating machines using neural networksrdquo Sofware Scienceand Computational Intelligence vol 983091 no 983092 983089983095 pages 983090983088983089983089

[983089983092] Y Lu Z Ren H Chen N Song and B Wen ldquoStudy on loose-ness and impactmdashrub coupling aults o a vertical dual-disk cantilever rotor- Bearing systemrdquo Key Engineering Materials vol 983091983093983091ndash983091983093983096 no 983092 pp 983090983092983095983097ndash983090983092983096983090 983090983088983088983095

[983089983093] Q Han Z Zhang and B Wen ldquoPeriodic motions o a dual-

disc rotor system with rub-impact at 1047297xed limiterrdquo Proceedingso the Institution o Mechanical Engineers vol 983090983090983090 no 983089983088 pp983089983097983091983093ndash983089983097983092983094 983090983088983088983096

[983089983094] H F de Castro K L Cavalca and R Nordmann ldquoWhirland whip instabilities in rotor-bearing system considering anonlinear orce modelrdquo Journal o Sound and Vibration vol 983091983089983095no 983089-983090 pp 983090983095983091ndash983090983097983091 983090983088983088983096

[983089983095] C C Fan et al ldquoMechanical systems and signal processingrdquoStudy o Start-Up Vibration Response or Oil Whirl and Dry Whip vol 983090983093 pp 983091983089983088983090ndash983091983089983089983093 983090983088983089983089

[983089983096] K Kappaganthu and C Nataraj ldquoNonlinear modeling andanalysis o a rolling element bearing with a clearancerdquo Commu-nications in Nonlinear Science and Numerical Simulation vol983089983094no 983089983088 pp 983092983089983091983092ndash983092983089983092983093 983090983088983089983089

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 1213

983089983090 International Journal o Rotating Machinery

[983089983097] C Gupta K Gupta and D K Sehgal ldquoInstability and chaoso a 1047298exible rotor ball bearing system an investigation on thein1047298uence o rotating imbalance and bearing clearancerdquo Journal o Engineering or Gas urbines and Power vol 983089983091983091 no 983096ArticleID 983088983096983090983093983088983089 983089983089 pages 983090983088983089983089

[983090983088] J Hong X Miao L Han and Y Ma ldquoPrognostics model orpredicting aero-engine bearing grade-lierdquo in Proceedings o the ASME urbo Expo vol 983089 pp 983094983091983097ndash983094983092983095 Orlando Fla USA June983090983088983088983097

[983090983089] P Borghesani P Pennacchi R B Randall N Sawalhi and RRicci ldquoApplication o cepstrum pre-whitening or the diagnosiso bearing aults under variable speed conditionsrdquo Mechanical Systems and Signal Processing 983090983088983089983090

[983090983090] Y Li J Zhang L Dai Z Zhang and J Liu ldquoAuditory-model-based eature extraction method or mechanical aults diagno-sisrdquo ChineseJournal o MechanicalEngineering (EnglishEdition) vol 983090983091 no 983091 pp 983091983097983089ndash983091983097983095 983090983088983089983088

[983090983091] N Bachschmid P Pennacchi and E anzi ldquoCracked rotatingshafs typical behaviors modeling and diagnosisrdquo IUAM Symposium on Emerging rends in Rotor Dynamics vol 983089983088983089983089 pp

983092983092983089ndash983092983093983092 983090983088983089983089[983090983092] Inoue N Nagata and Y Ishida ldquoFem modelling and

experimental veri1047297cation o a rotor system with a open crackrdquoin Proceedings o the ASME International Design Engineering echnical Conerences and Computers and Inormation in Engi-neering Conerence (DEC rsquo983088983097) pp 983089983089983089983091ndash983089983089983090983090 San Diego CaliUSA September 983090983088983088983097

[983090983093] H E Sonnichsen ldquoReal-time detection o developing cracksin jet engine rotorsrdquo in Proceedings o the IEEE AerospaceConerence pp 983089983095983091ndash983089983096983092 March 983090983088983088983088

[983090983094] I Green and C Casey ldquoCrack detection in a rotor dynamicsystem by vibration monitoringmdashpart I analysisrdquo Journal o Engineering or Gas urbines and Power vol 983089983090983095 no 983090 pp 983092983090983093ndash983092983091983094 983090983088983088983093

[983090983095] J Sawicki M I Friswell A H Pesch and A WroblewskildquoCondition monitoring o rotor usingactive magnetic actuatorrdquoin Proceedings o the ASME urbo Expo pp 983089983090983093983095ndash983089983090983094983093 BerlinGermany June 983090983088983088983096

[983090983096] J Xiang X Chen Q Mo and Z He ldquoIdenti1047297cation o crack ina rotor system based on wavelet 1047297nite element methodrdquo FiniteElements in Analysis and Design vol 983092983091 no 983089983092 pp 983089983088983094983096ndash983089983088983096983089983090983088983088983095

[983090983097] M G MaalouldquoSlow speed vibrationsignal analysisi you canrsquotdo it slow you canrsquot do it astrdquo in Proceedings o the ASME urboExpo pp 983093983093983097ndash983093983094983095 Montreal Canada May 983090983088983088983095

[983091983088] J Meagher X Wu and C Lencioni ldquoResponse o a warped1047298exible rotor with a 1047298uid bearingrdquo International Journal o

Rotating Machinery vol 983090983088983088983096 Article ID 983089983092983095983094983093983091 983097 pages 983090983088983088983096

[983091983089] Gaka andM abaszewski ldquoAn applicationo statistical symp-toms in machine condition diagnosticsrdquo Mechanical Systemsand Signal Processing vol 983090983093 no 983089 pp 983090983093983091ndash983090983094983093 983090983088983089983089

[983091983090] J K Sinha ldquoRecent trands in ault quanti1047297cation in rotatingmachinesrdquo Advances in Vibration EngIneerIng vol 983096 no 983089 pp983095983097ndash983096983093 983090983088983088983097

[983091983091] X Shen JJia andM Zhao ldquoNonlinear analysis o a rub-impactrotor-bearing system with initial permanent rotor bowrdquo Archiveo Applied Mechanics vol 983095983096 no 983091 pp 983090983090983093ndash983090983092983088 983090983088983088983096

[983091983092] A W Lees J K Sinha and M I Friswell ldquoModel-based identi-1047297cation o rotating machinesrdquo Mechanical Systems and Signal Processing vol 983090983091 no 983094 pp 983089983096983096983092ndash983089983096983097983091 983090983088983088983097

[983091983093] L C Jaw and W Merrill ldquoCBM+ research environmentmdashacilitating technology development experimentation andmaturationrdquo in Proceedings o the IEEE Aerospace Conerence(AC rsquo983088983096) March 983090983088983088983096

[983091983094] R Walker S Perinpanayagam and I K Jennions ldquoSimulatingUnbalance or Future IVHM Applicationsrdquo in Proceedings o theSociety or Experimental Mechanics (IMAC rsquo983089983090) vol 983090983097 pp 983089983092983089ndash983089983092983096 May 983090983088983089983090

[983091983095] H C Pusey ldquourbomachinery condition monitoring and ail-ure prognosisrdquo Sound and Vibration vol 983092983089 no 983091 pp 983089983088ndash983089983093983090983088983088983095

[983091983096] J R Jain and K Kundra ldquoModel based online diagnosiso unbalance and transverse atigue crack in rotor systemsrdquo Mechanics Research Communications vol 983091983089 no 983093 pp 983093983093983095ndash983093983094983096983090983088983088983092

[983091983097] K R Wheeler Kurtoglu and S D Poll ldquoA survey o healthmanagement user objectives related to diagnostic and prognos-tic metricsrdquo in Proceedings o the 983090983097th Computers and Inor-mation Engineering Conerence vol 983090 pp 983089983090983096983095ndash983089983090983097983096 983090983088983088983097

[983092983088] G Vachtsevanos ldquoPerormance metrics or ault prognosis o

complex systemsrdquo in IEEE Systems Readiness echnology Con- erence (AUOESCON rsquo983088983091) pp 983091983092983089ndash983091983092983093 September 983090983088983088983091

[983092983089] A Saxena J Celaya E Balaban et al ldquoMetrics or evaluatingperormance o prognostic techniquesrdquo in Proceedings o theInternational Conerence on Prognostics and Health Manage-ment (PHM rsquo983088983096) pp 983089ndash983089983095 October 983090983088983088983096

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 1313

Submit your manuscripts at

httpwwwhindawicom

Page 7: Rotordynamic Faults 2013

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 713

International Journal o Rotating Machinery 983095

magnetic bearing system or use in the early detection local-ization and prognosis o blade cracks Again the emphasis ison early detection with the bearings used to excite the systemin order to obtain optimum response vibrations or analysisTe authors admit the approach hassome merit in diagnosingblade cracks however it is in the early stages o development

and work are ongoing FEA has also been used extensively to support blade crack prognostic tools Xiang et al [983090983096] arean extensive example o recent work In this case a numbero advanced FEA methods are applied to produce accurateFEA solutionsmdashthese include surace-1047297tting techniques andthe contour-plotting method Te authors experimentally

validate their work and suggest that it can be applied toprognosis and quantitative diagnosis o blade cracks Despitethe claimed advances again the scalability o such research toull size turbines is an issuemdashparticularly with regard to thecomplexity o the FEA models

983090983096 Rotor Bow Rotor bows can be a primary source o unwanted vibration in gas turbines Te main cause o arotor bow (rotor bows do not include bows due to gravity) isthermal differences in a system caused by operating con-ditions It is noted by Domes [983093] that this nonsymmetricalthermal distribution can cause excessive unbalance to theextent where a gas turbine will not start correctly Such rotorbows are common on start up or shut down and are ofenaccounted or in operational procedures However i thermalldquohot spotsrdquo exceed a given tolerance level they can causepermanent unbalances due to rotor de1047298ections Such rotorbows can lead to other aults includingrubbing andloosenesswhich complicate isolation and localization

raditional data-driven techniques or detecting rotorbows involve combinations o slow roll and vibration data[983090983097] More recently mathematical modeling techniques suchas that detailed by Meagher et al [983091983088] have been used inorder to diagnose residual rotor bows and differentiate theseaults rom other sources o unbalance Te authors o thispaper build upon established methods or the models and areunique in that they concentrate on response at the bearingpoints Tis is perhaps more useul orindustrial applicationsas vibration inormation is more readily available at bearingpoints rather than intrusive proximity probes which are ofenused in lab work

Te little work that exists on attempting to localise rotorbows across complex systems tends to be data-driven in

nature see Gaka and abaszewski [983091983089] where the authorsused statistical symptoms based on known data as a methodo diagnosing and prognosing a number o aults includingrotor bows and unbalance Tis paper attempts to address theirregularities and 1047298uctuations that occur over a long servicelie In order to achieve this a modi1047297ed energy processormodel is created using data drawn rom large steam turbinesover long periods o lie

Prognosing rotor bows is a complex subject As rotorbows are ofen caused by temperature de1047298ections makingpredictions or remaining useul lie and potential utureproblems lies not only in the realm o mechanical rotor-dynamics but also to some extent in thermodynamics Te

recent work detailed by Sinha [983091983090] is o note or detailingdiagnosis and quanti1047297cation o various rotordynamic aultsand describing the advantages o mathematical modelingover traditional vibration-based approaches Te topic o scalability regarding FEA models is discussed including anargument or the use o partial (simpli1047297ed) mathematical

models or large complex systemsAnother two works which are o interest with regard to

modeling o rotor bows include Shen et al [983091983091] where theauthors modeled a rotor-bearing system with a permanentrotor bow looking at the impact o secondary aults such asrub Te study o ault combinations in this paper is useulor ault differentiation studies however the authors study a permanent initial rotor bow Tis thereore does not takeinto account developing or worsening aults and the different

vibration phenomena that are observed as such aults aredeveloping

Lees et al [983091983092] describe the importance o model-basedault identi1047297cation techniques and outline recent research in

the area providing a good reerence paper or more researchon this speci1047297c ault

3 Discussion

983091983089 Sensor Suites Te subject o sensor suites is o greatimportance with regard to uture industrial applications Onone hand advanced and complex sensor suites generally enable improved ault localization and diagnosis howeverthe added complexity and cost has resulted in many o thesesystems being omitted rom the latest generations o rotatingmachinery in industrial applications From an industrialperspective the cost bene1047297t o additional sensors needs to

be signi1047297cant in order to justiy this approach In response anumber o works listed including [983096] as a good example takethe approach o achieving the same objectives using a greatly reduced sensor suite In many ways the required sensor suiteor many o the currently researched ault diagnosis andprognosis techniques provides an indication o the ability o the techniques to be used practically in industrial situationsA complex suite may return a very high success rate on alab-based rig however the impracticalities o mounting suchsuites on a real-world application negate the advantages

Further to this the consideration orsensor position mustbe considered Te use o proximity probes and keyphasortransducers has very clear advantages in a number o situ-

ations however the intrusive nature o the sensor preventsapplication in a number o complex systems With this inmind a number o the more promising studies listed rely onthe simple suiteo an accelerometer placed in positions whichare relatively remote rom the sources o vibration potentially with a noisy transer path

983091983090 Diagnosis It can be seen rom the research outlined inthe previous section that the diagnosis o aults in rotatingmachinery is a subject o much ongoing research Tisinvolves the improvement and development o traditional

vibration monitoring techniques development o new data-driven technologies and novel research into physics-based

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 813

983096 International Journal o Rotating Machinery

simulation and modeling In many cases these topics o research are dependent on one another or reasons o vali-dation veri1047297cation and speed o analysis In several casesit can be seen that multiple aults have been modeled orthe purposes o identi1047297cation and isolation However nostudies have yet been perormed which deal with all o the

aorementioned aults All o these aults are intrinsically related to one another Complex combinations o aults havebegun to be analyzed with the emphasis on developing new diagnosis techniques Physics-based modeling has proven toprovide signi1047297cant advances with regard to speci1047297c aultsnotably shaf and blade cracks where techniques such as FEAenable much easier aster and cheaper test data then seedingaults into live systems

It can be noted however that both high and low-1047297delity modeling techniques are being applied to cutting edgeresearch or all o the listed rotordynamic aults (and othersnot detailed in this paper) In addition to the advantages inthe speed o obtaining test results physics-based simulationis providing another dimension to data-driven techniquesSystem models are being used as part o logic and reasoningsuites in the identi1047297cation and differentiation o variousaults High-1047297delity models enable simulations o ault com-binations which are not possible practically or are pro-hibitively expensive in live systems

Nevertheless results obtained rom modeling studies stillneed to be validated against proven data-driven techniquesbeore implementation in industrial applications is possibleIt is also o note rom the literature reviewed or this paperthat almost all modeling or the diagnosis o aults involvesextensive simpli1047297cation ofen reducing potentially complexsystems down to one or two discshafbearing modelsAdapting the claimed results rom such research to systemswith many discsbearingsshafs is important in improvingexisting rotating machinery diagnostic techniques

Te inclusion o nonlinear effects is notable amongstmany authorsmdashas the understanding o the complex aultsand interactions continues to progress Such research opensthe potential or promising new avenues into areas o designor IVHM

Finally it is o note that despite the wide variety o advanced research here a consensus or the most efficienttechniques and algorithms has yet to emerge in order or the1047297nal adaption to industrial application to be achieved How-ever the consideration or industrial application taken by many o the authors detailed in this paper provides promise

or next generation IVHM systems

983091983091 Localisation Fault localization in rotating machinery isan important topic o research or uture condition-basedmonitoring systems Knowing not only what type o aulthas occurred but also where in the system is an importantconsideration which can in1047298uence maintenance proceduresin complex machinery It is worth noting rom the literaturesurveyed or this paper that many studies have ocusedon diagnosis and prognosis o single rotorbearing systemsOfen or legitimate reasonsmdashsimpli1047297cation or computingspeed or example Few studies however have taken intoaccount the localization o aultsacross whole systems Oneo

the more extensive examples highlighted in this paper (Hanet al [983089983093]) provides extensive analytical studies relevant tolocalization on a duel disk setup utilizing Hilbert-Huangransorms however in this example comprehensive valida-tion is lacking thus limiting the potential applicability o theresearch

Tis problem is not limited to modeling and simulation-based research Many newly developed data-driven tech-niques or diagnostics and prognostics claim good results by heavily instrumenting speci1047297c components o a test systemIn many industrial cases this is not possible practical or cost-effective

o give some examples a keyphasor transducer can beparticularly useul in diagnosing aults such as rotor bowHowever this equipment requires the ability to cut a key way or measurements to be perormed Optical sensors haverecently been applied to detect rotor unbalance yet such atechnology would be difficult to implement in a system withseveral rows o rotors (as in a gas turbine)

Whilst new diagnosis and prognosis techniques or aultsin rotating machinery are being continuously researchedthe lack o corresponding studies into localization can beconsidered one o the many challenges in promoting recentcore research into live industrial applications

983091983092 Fault Differentiation One promising development inrecent modeling and simulation o rotordynamic aults isthat o ault differentiation Several researchers have movedon rom studies on individual aults in order to concentrateon combinations o aults As stated rotordynamic aultssuch as those listed in this paper are linked to each otherSuch studies thereore concentrate on such topics as a

misalignment causing an unbalance or a looseness causinga rub Findings rom reports such as these are important inunderstanding complex anomalies In industrial applicationssimply detecting and rectiying an unbalance does not pro-

vide a satisactory solution i the root cause o the ault isa misalignment Tis is a complex topic as several aults canexhibit similar vibration characteristics making traditionaldetection techniques inaccurate in some cases

Making reerence to some o the aorementioned stud-ies those such as Bachschmid et al [983090983091] and Sonnichsen[983090983093] indicate some initial studies including several aultsTe results o such studies indicate a high success rate indifferentiating the aults in the given conditions however

a number o limitations still exist Studies such as theseconsider unbalancein the aultchains Tisis typically ldquostaticrdquoor simple unbalance applied to a single shafrotormdashin reality unbalance aults can be more complex than this In additionthe simple systems used to study and differentiate aultsare very different rom the complexities o a ull turbineHowever urther work may enable such research to ldquoscale-uprdquo to this stage with the aid o urther validation and

veri1047297cation

983091983093 Prognosis It can be seen rom the examples o recentresearch described or common aults that prognostic tech-niques are a topic where much work is being perormed

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 913

International Journal o Rotating Machinery 983097

Some o this work would be very difficult or impossibleto implement in an industrial situation others providepromising results o use to uture work in the area Predictingthe remaining useul lie o components is critical in thedevelopment o condition-based monitoring strategies orindustrial implementation

It can be noted that prognostic studies or certain rotor-dynamic aults are considerably more advanced than orothers Te most obvious examples o the aults detailed inthis paper are bearing aults shaf and blade cracks Teseare also areas where physics-driven simulations have hadan important impact Te ability to design any ault type(or combinations o aults) into such simulations (to be itslow-1047297delity mathematical models or high-1047297delity 1047297nite ele-ment analysis) has provided researchers with many differentavenues to explore Studies into the prognostics o otheraults can be complicated by various actors An unbalanceor example has had relatively little research perormed intoprognosis Te act that unbalance is ofen the cause o another underlying ault is one reason or the difficulty inresearching prognosis in detail or this ault Te results o unbalance are also very dependent on the severity o the aultand the system in which it occurs

It is worth noting that even or the aults where prognosisresearch is more advanced (eg shaf cracks) true develop-ment o the ault is not always studied Fan et al [983089983095] orexample study sets o discrete cracks in order to ldquoprognoserdquocrack growth In act thisstudy is more akinto detecting aultseverity with inerred effects on crack growth In contrast Liet al [983090983090] consider the continual development o such crackshowever the authors do admit this work is in its inancySuch points limit the ability or true prognosis to be achievedwithout extensive operational data

It is also worth noting that combinations o aults havebeen simulated extensively or the purpose o diagnosticshowever ew studies exist combining aultsor the purpose o prognostics Such studies are however a logical progressionrom some o those already perormed It may be that suchprognostic studies build upon some o the condition-basedmaintenance (CBM) or diagnostics High-level studies suchas Jaw and Merrill [983091983093] indicate how prognosis and diagnosistechniques can be combined into a CBM system Tis study is designed or a military aeroengine indicating the desire orsuch systems to be implementedmdashalthough the architecturedescribed has the capability o including various algorithmsoraultprognosis the amount o consideration orprognosis

o ault combinations is unclear

983091983094 Modelling Te large number o works reviewed or thispaper which includes modeling techniques in order to aiddiagnosis and prognosis indicates the possibilities providedby modern computer power and sofware developmentBroadly the modeling studied in the aorementioned papersbreaks down into two categories mathematical modeling and1047297nite element analysis (FEA) Te mathematical models suchasBorghesani etal [983090983089] ofen provide a theoretical basis uponwhich data driven or FEA studies can build and validateMathematical models despite being a traditional approach to

rotordynamic problems thereore continue to be developedto tackle more advanced problems

FEA techniques offer a constantly expanding area o simulation to explore Tey are used or research as diverseas crack propagation to unbalance localization [983091983094] Moderncodes such as NASRAN and Ansys enable rotordynamics to

be studied beyond the traditional stress and modal analysisTe ability to study the effects o rotation (eg throughCampbell diagrams and transient analysis) continues to driveresearch and innovation in this area Te studies discussedin this paper indicate the power o FEA with some highly accurate simulations having been perormed However thereis still a limitation o computer powermdashapplying some o the FEA techniques to whole engine models would proveto be too computationally expensive to be viable As aresult alternate methods such as model order reductionand system level modeling (and model-based reasoning orimplementation) are still required in order to make some o the FEA studies viable in an industrial environment

Nonlinearities continue to orm an important part o many recent studies with the complexity and detail o thespeci1047297ed aults being continually expanded Te depth o modeling has enabled in many cases accurate validationagainst experimental approaches Despite this completeunderstanding o the vibrational phenomena o rotatingmachinery is not currently possible despite the drive towardsthis end

983091983095 Rotordynamics and IVHM Te bulk o current researchinto rotordynamics rom the point o view o prognostichealth management (PHM) can be roughly divided intotwo types initial single-ault diagnosisprognosis techniquesand studies into the general requirements and limitationso current systems along with current and uture trendsAn example o the latter is Pusey [983091983095] who provides agood summary overview o current diagnosis and prognosistechniques with regard to condition-based maintenance

As a result o this split a clear gap exists between thecore research being perormed into rotordynamics rom acondition-based maintenance perspective and the identi1047297edneeds o industry aking a 1047298edgling piece o research andapplying it to a commercially-ready system (eg a gas turbineengine or an aircraf) is a long and complex task It isnevertheless worth noting that technologies or automatically detecting an unbalance or misalignment in a gas turbine were

developed over 983089983088 years beore the latest commercial aircrafwere conceptualized and yet these planes are still limited inthis capacity Tis highlights the need or work which linksthe undamental research into individual ault diagnosis toldquolive systemsrdquo in use in industry

Physics-based simulation and modeling o rotordynamicparts is a well-researched 1047297eld Such modeling has been usedas the basis o diagnosis and prognosis o aults by many researchers several recent examples have been outlined inthis paper Occasional pieces o work have been perormedinto modeling multiple aults such as Jain and Kundra [983091983096]who usea system model oronline identi1047297cation o unbalanceand cracks Beyond this however very limited research

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 1013

983089983088 International Journal o Rotating Machinery

Business casecost bene1047297tanalysis

Live system

RubCracksBearingfailure

Rotor bow Unbalance

Fault differentiationLocalisation

Root cause analysis

Remaining useful life

Data driven analysis and physics-based simulation

Mis-alignment

Experimental

1047297eld data

F983145983143983157983154983141 983089 Physics-based simulationmdashrom research to industry

exists Te demands o PHM techniques in industry are suchthat any systemmust notonly be capable o detecting multipleaults but must also be capable o detecting these aultsacrossa range o different systems Other considerations includethe aorementioned ability to differentiate between multipleaults Processing also needs to be taken into account as theobjective o these systems is to implement efficient conditionmonitoring and condition-based maintenance procedures I processing data is a long power-hungry process then thisaim cannot be achieved

Figure 983089 details a potential ramework required in order

to push core research such as that detailed in this reporttowards industrial applications Many studies now exist onindividual rotordynamic aults across a wide range o condi-tions and applications Some studies have taken this urtherwith advanced prognostic models and diagnosis o dual aults(primary cause and secondary effect) Future research inthe area o rotordynamics rom a PHM perspective couldpotentially provide the bridge between these studies and livesystems by combining physics-based simulations with data-driven techniques and validating against experimental data

Although quantiying the success o the research studiedisdifficultit ispossibleto de1047297ne the key areas in whicha tech-nique must excel in order to be considered viable Te work

byWheeleretal[983091983097] discussesin detailmetrics ordiagnosticand prognostic analysis as does that by Vachtsevanos [983092983088]and Saxena et al [983092983089] Te conclusions drawn rom thesepapers and applied in practice to research like that covered by this paper indicate that the ollowing metrics are importantwhen considering the potential o a given technique to diag-nose aults coverage alse positive rate and alse negativerate In the case o prognosis probabilities and lead timeto ailure are other important considerations Tis criteriaenables research to be assessed in terms o its suitability orindustrial applications Unortunately inormation on these

metrics is not made readily available by the authors o mostpapers

In terms o evaluating the effectivenesso theresearch dis-cussed in this paper there are difficulties in recommending aspeci1047297c technique over others or general application Mostpapers reviewed or this document (not just those reerencedand discussed in detail) take a technique (new or evolved)

validate or a given speci1047297c system and report success o theresearch Despite this some conclusions can be drawn romassessing common techniques applied across different studiesand different aults Although it is not possible to de1047297ne themost common methods or diagnosing andprognosing aultsin terms o numbers (as it was not possible to cover all recent

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 1113

International Journal o Rotating Machinery 983089983089

rotordynamics research or this paper) as perceived by theauthors the ollowing techniques have eatured prominently in the reviewed research

(i) Sensors accelerometers

(ii) Teoretical studies mathematical modeling

(iii) Physics-based simulation FEA(iv) Data-driven joint timerequency domain analysis

(v) Implementation neural networking

Tese techniques appear to be among the most promisingcurrently under development as they tend to eature numer-ous timesamongst some o the work with wider scope acrossault types and with the most comprehensive validationTere are o course many subsections to these techniqueshowever it shows one general direction o research and theclear possibilities posed in these areas

Future developments in the 1047297eld o IVHM or rotat-ing machinery may incorporate these techniques alongsideextensive use o nonlinear modeling and multiple aultinteractions Te 1047297eld o design or IVHM has only recently emerged however the potential exists or speci1047297c nonlinear-ities to be designed into a system in order to enable accuratediagnosis andprognosis o aults Te development o currentalgorithms to include diagnosis localization and prognosiso a range o aults will provide a signi1047297cant advancement oruture generations o IVHM systems Tis combined withcost-effective sensor suites indicates the potential or evolu-tions o some o the research detailed hereto orm part onextgeneration IVHM suites or rotating machinery

4 Conclusion

Tis paper has reviewed some o the latest research around anumber o rotordynamic aults namely unbalance misalign-ment rub and looseness 1047298uid-induced instability bearingaults shaf cracks blade cracks and rotor bow Each aultwas reviewed rom the perspective o sensors diagnosisprognosis localization and modeling

Key examples o recent work into the eight describedaults have been detailed through works by a number o eminent authors Additional work has been summarized andormatted or easy reerence Some current trends amongstthe recent body o work include developments in the vastarea o modeling nonlinearities combinations o high andlow 1047297delity modeling and synergy between data driven and

physics-based simulation approachesDespite the large volume o promising research reviewed

urther development in a number o areas is required in orderto produce effective next generation IVHM systems As suchuture developments may include usion o andor devel-opment o current algorithms to encompass all eight aultsdetailed consideration o prognosis diagnosis and localiza-tion achieved using a reduced cost-effective sensor suite

References

[983089] I K Jennions Integrated Vehicle Health Management Perspec-tives on an Emerging Field 983090983088983089983089

[983090] G Anderson ldquoProviding best-value IVHM solutions or agingaircrafrdquo in Proceedings o the 983097th Joint FAADoDNASA Aging Aircraf Conerence vol 983089 pp 983089ndash983089983089 983090983088983088983093

[983091] A Muszynska Rotordynamics aylor amp Francis 983089st ed edition983090983088983088983093

[983092] D E Bently Fundimentals o Rotating Machinery Diagnostics

ASME Press 983089st edition 983090983088983088983090[983093] B Domes ldquoVibration phenomena in aero-enginesrdquo in Proceed-ings o the 983097th International Conerence on Vibrations in Rotating Machinery Exeter (IMechE rsquo983088983096) vol 983089 pp 983089983093ndash983091983090 September983090983088983088983096

[983094] A Hess G Calvello and Dabney ldquoPHM a keyenabler or theJSF autonomie logistics support conceptrdquo in Proceedings o theIEEE Aerospace Conerence Proceedings pp 983091983093983092983091ndash983092983093983092983097 March983090983088983088983092

[983095] S N Ganeriwala B Schwarz and M H Richardson ldquoOperat-ing de1047298ection shapes detect unbalance in rotating equipmentrdquoSound and Vibration vol 983092983091 no 983093 pp 983089983089ndash983089983091 983090983088983088983097

[983096] G N D S Sudhakar and A S Sekhar ldquoIdenti1047297cation o unbal-ance in a rotor bearing systemrdquo Journal o Sound and Vibration

vol 983091983091983088 no 983089983088 pp 983090983090983097983097ndash983090983091983089983091 983090983088983089983089[983097] Yang and M W Hsu ldquoAn efficient diagnosis technique or

variations o shaf-bow and unbalancerdquo in Proceedings o the ASME International Design Engineering echnical Conerencesand Computers and Inormation in Engineering Conerence(DEC rsquo983088983097) pp 983093983095ndash983094983094 San Diego Cali USA September 983090983088983088983097

[983089983088] A El-Shaei S H aw1047297ck and M O A Mokhtar ldquoExperi-mental investigation o the effect o angular misalignemt onthe instability o plain journal bearingsrdquo in Proceedings o the ASMESLE InternationalJointribology Conerence(IJC rsquo983088983097)pp 983089983095983089ndash983089983095983091 Memphis enn USA October 983090983088983088983097

[983089983089] A El-Shaei S H aw1047297ck and M O A Mokhtar ldquoMisalign-ment modeling in rotating systemsrdquo in Proceedings o the ASMEurbo Expo H Bahaloo A Ebrahimi and M Samadi Eds vol

983089983095983089983089 pp 983097983095983091ndash983097983095983097 Orlando Fla USA June 983090983088983088983097

[983089983090] L F Villaet al ldquoStatistical diagnosisbased on vibration analysisor gear test-bench under non-stationary conditions o speedand loadrdquo Mechanical Systems and Signal Processing vol 983090983097 pp983092983091983094ndash983092983092983094 983090983088983089983090

[983089983091] C F Ngolah et al ldquoIntelligent ault recognition and diagnosisor rotating machines using neural networksrdquo Sofware Scienceand Computational Intelligence vol 983091 no 983092 983089983095 pages 983090983088983089983089

[983089983092] Y Lu Z Ren H Chen N Song and B Wen ldquoStudy on loose-ness and impactmdashrub coupling aults o a vertical dual-disk cantilever rotor- Bearing systemrdquo Key Engineering Materials vol 983091983093983091ndash983091983093983096 no 983092 pp 983090983092983095983097ndash983090983092983096983090 983090983088983088983095

[983089983093] Q Han Z Zhang and B Wen ldquoPeriodic motions o a dual-

disc rotor system with rub-impact at 1047297xed limiterrdquo Proceedingso the Institution o Mechanical Engineers vol 983090983090983090 no 983089983088 pp983089983097983091983093ndash983089983097983092983094 983090983088983088983096

[983089983094] H F de Castro K L Cavalca and R Nordmann ldquoWhirland whip instabilities in rotor-bearing system considering anonlinear orce modelrdquo Journal o Sound and Vibration vol 983091983089983095no 983089-983090 pp 983090983095983091ndash983090983097983091 983090983088983088983096

[983089983095] C C Fan et al ldquoMechanical systems and signal processingrdquoStudy o Start-Up Vibration Response or Oil Whirl and Dry Whip vol 983090983093 pp 983091983089983088983090ndash983091983089983089983093 983090983088983089983089

[983089983096] K Kappaganthu and C Nataraj ldquoNonlinear modeling andanalysis o a rolling element bearing with a clearancerdquo Commu-nications in Nonlinear Science and Numerical Simulation vol983089983094no 983089983088 pp 983092983089983091983092ndash983092983089983092983093 983090983088983089983089

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 1213

983089983090 International Journal o Rotating Machinery

[983089983097] C Gupta K Gupta and D K Sehgal ldquoInstability and chaoso a 1047298exible rotor ball bearing system an investigation on thein1047298uence o rotating imbalance and bearing clearancerdquo Journal o Engineering or Gas urbines and Power vol 983089983091983091 no 983096ArticleID 983088983096983090983093983088983089 983089983089 pages 983090983088983089983089

[983090983088] J Hong X Miao L Han and Y Ma ldquoPrognostics model orpredicting aero-engine bearing grade-lierdquo in Proceedings o the ASME urbo Expo vol 983089 pp 983094983091983097ndash983094983092983095 Orlando Fla USA June983090983088983088983097

[983090983089] P Borghesani P Pennacchi R B Randall N Sawalhi and RRicci ldquoApplication o cepstrum pre-whitening or the diagnosiso bearing aults under variable speed conditionsrdquo Mechanical Systems and Signal Processing 983090983088983089983090

[983090983090] Y Li J Zhang L Dai Z Zhang and J Liu ldquoAuditory-model-based eature extraction method or mechanical aults diagno-sisrdquo ChineseJournal o MechanicalEngineering (EnglishEdition) vol 983090983091 no 983091 pp 983091983097983089ndash983091983097983095 983090983088983089983088

[983090983091] N Bachschmid P Pennacchi and E anzi ldquoCracked rotatingshafs typical behaviors modeling and diagnosisrdquo IUAM Symposium on Emerging rends in Rotor Dynamics vol 983089983088983089983089 pp

983092983092983089ndash983092983093983092 983090983088983089983089[983090983092] Inoue N Nagata and Y Ishida ldquoFem modelling and

experimental veri1047297cation o a rotor system with a open crackrdquoin Proceedings o the ASME International Design Engineering echnical Conerences and Computers and Inormation in Engi-neering Conerence (DEC rsquo983088983097) pp 983089983089983089983091ndash983089983089983090983090 San Diego CaliUSA September 983090983088983088983097

[983090983093] H E Sonnichsen ldquoReal-time detection o developing cracksin jet engine rotorsrdquo in Proceedings o the IEEE AerospaceConerence pp 983089983095983091ndash983089983096983092 March 983090983088983088983088

[983090983094] I Green and C Casey ldquoCrack detection in a rotor dynamicsystem by vibration monitoringmdashpart I analysisrdquo Journal o Engineering or Gas urbines and Power vol 983089983090983095 no 983090 pp 983092983090983093ndash983092983091983094 983090983088983088983093

[983090983095] J Sawicki M I Friswell A H Pesch and A WroblewskildquoCondition monitoring o rotor usingactive magnetic actuatorrdquoin Proceedings o the ASME urbo Expo pp 983089983090983093983095ndash983089983090983094983093 BerlinGermany June 983090983088983088983096

[983090983096] J Xiang X Chen Q Mo and Z He ldquoIdenti1047297cation o crack ina rotor system based on wavelet 1047297nite element methodrdquo FiniteElements in Analysis and Design vol 983092983091 no 983089983092 pp 983089983088983094983096ndash983089983088983096983089983090983088983088983095

[983090983097] M G MaalouldquoSlow speed vibrationsignal analysisi you canrsquotdo it slow you canrsquot do it astrdquo in Proceedings o the ASME urboExpo pp 983093983093983097ndash983093983094983095 Montreal Canada May 983090983088983088983095

[983091983088] J Meagher X Wu and C Lencioni ldquoResponse o a warped1047298exible rotor with a 1047298uid bearingrdquo International Journal o

Rotating Machinery vol 983090983088983088983096 Article ID 983089983092983095983094983093983091 983097 pages 983090983088983088983096

[983091983089] Gaka andM abaszewski ldquoAn applicationo statistical symp-toms in machine condition diagnosticsrdquo Mechanical Systemsand Signal Processing vol 983090983093 no 983089 pp 983090983093983091ndash983090983094983093 983090983088983089983089

[983091983090] J K Sinha ldquoRecent trands in ault quanti1047297cation in rotatingmachinesrdquo Advances in Vibration EngIneerIng vol 983096 no 983089 pp983095983097ndash983096983093 983090983088983088983097

[983091983091] X Shen JJia andM Zhao ldquoNonlinear analysis o a rub-impactrotor-bearing system with initial permanent rotor bowrdquo Archiveo Applied Mechanics vol 983095983096 no 983091 pp 983090983090983093ndash983090983092983088 983090983088983088983096

[983091983092] A W Lees J K Sinha and M I Friswell ldquoModel-based identi-1047297cation o rotating machinesrdquo Mechanical Systems and Signal Processing vol 983090983091 no 983094 pp 983089983096983096983092ndash983089983096983097983091 983090983088983088983097

[983091983093] L C Jaw and W Merrill ldquoCBM+ research environmentmdashacilitating technology development experimentation andmaturationrdquo in Proceedings o the IEEE Aerospace Conerence(AC rsquo983088983096) March 983090983088983088983096

[983091983094] R Walker S Perinpanayagam and I K Jennions ldquoSimulatingUnbalance or Future IVHM Applicationsrdquo in Proceedings o theSociety or Experimental Mechanics (IMAC rsquo983089983090) vol 983090983097 pp 983089983092983089ndash983089983092983096 May 983090983088983089983090

[983091983095] H C Pusey ldquourbomachinery condition monitoring and ail-ure prognosisrdquo Sound and Vibration vol 983092983089 no 983091 pp 983089983088ndash983089983093983090983088983088983095

[983091983096] J R Jain and K Kundra ldquoModel based online diagnosiso unbalance and transverse atigue crack in rotor systemsrdquo Mechanics Research Communications vol 983091983089 no 983093 pp 983093983093983095ndash983093983094983096983090983088983088983092

[983091983097] K R Wheeler Kurtoglu and S D Poll ldquoA survey o healthmanagement user objectives related to diagnostic and prognos-tic metricsrdquo in Proceedings o the 983090983097th Computers and Inor-mation Engineering Conerence vol 983090 pp 983089983090983096983095ndash983089983090983097983096 983090983088983088983097

[983092983088] G Vachtsevanos ldquoPerormance metrics or ault prognosis o

complex systemsrdquo in IEEE Systems Readiness echnology Con- erence (AUOESCON rsquo983088983091) pp 983091983092983089ndash983091983092983093 September 983090983088983088983091

[983092983089] A Saxena J Celaya E Balaban et al ldquoMetrics or evaluatingperormance o prognostic techniquesrdquo in Proceedings o theInternational Conerence on Prognostics and Health Manage-ment (PHM rsquo983088983096) pp 983089ndash983089983095 October 983090983088983088983096

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 1313

Submit your manuscripts at

httpwwwhindawicom

Page 8: Rotordynamic Faults 2013

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 813

983096 International Journal o Rotating Machinery

simulation and modeling In many cases these topics o research are dependent on one another or reasons o vali-dation veri1047297cation and speed o analysis In several casesit can be seen that multiple aults have been modeled orthe purposes o identi1047297cation and isolation However nostudies have yet been perormed which deal with all o the

aorementioned aults All o these aults are intrinsically related to one another Complex combinations o aults havebegun to be analyzed with the emphasis on developing new diagnosis techniques Physics-based modeling has proven toprovide signi1047297cant advances with regard to speci1047297c aultsnotably shaf and blade cracks where techniques such as FEAenable much easier aster and cheaper test data then seedingaults into live systems

It can be noted however that both high and low-1047297delity modeling techniques are being applied to cutting edgeresearch or all o the listed rotordynamic aults (and othersnot detailed in this paper) In addition to the advantages inthe speed o obtaining test results physics-based simulationis providing another dimension to data-driven techniquesSystem models are being used as part o logic and reasoningsuites in the identi1047297cation and differentiation o variousaults High-1047297delity models enable simulations o ault com-binations which are not possible practically or are pro-hibitively expensive in live systems

Nevertheless results obtained rom modeling studies stillneed to be validated against proven data-driven techniquesbeore implementation in industrial applications is possibleIt is also o note rom the literature reviewed or this paperthat almost all modeling or the diagnosis o aults involvesextensive simpli1047297cation ofen reducing potentially complexsystems down to one or two discshafbearing modelsAdapting the claimed results rom such research to systemswith many discsbearingsshafs is important in improvingexisting rotating machinery diagnostic techniques

Te inclusion o nonlinear effects is notable amongstmany authorsmdashas the understanding o the complex aultsand interactions continues to progress Such research opensthe potential or promising new avenues into areas o designor IVHM

Finally it is o note that despite the wide variety o advanced research here a consensus or the most efficienttechniques and algorithms has yet to emerge in order or the1047297nal adaption to industrial application to be achieved How-ever the consideration or industrial application taken by many o the authors detailed in this paper provides promise

or next generation IVHM systems

983091983091 Localisation Fault localization in rotating machinery isan important topic o research or uture condition-basedmonitoring systems Knowing not only what type o aulthas occurred but also where in the system is an importantconsideration which can in1047298uence maintenance proceduresin complex machinery It is worth noting rom the literaturesurveyed or this paper that many studies have ocusedon diagnosis and prognosis o single rotorbearing systemsOfen or legitimate reasonsmdashsimpli1047297cation or computingspeed or example Few studies however have taken intoaccount the localization o aultsacross whole systems Oneo

the more extensive examples highlighted in this paper (Hanet al [983089983093]) provides extensive analytical studies relevant tolocalization on a duel disk setup utilizing Hilbert-Huangransorms however in this example comprehensive valida-tion is lacking thus limiting the potential applicability o theresearch

Tis problem is not limited to modeling and simulation-based research Many newly developed data-driven tech-niques or diagnostics and prognostics claim good results by heavily instrumenting speci1047297c components o a test systemIn many industrial cases this is not possible practical or cost-effective

o give some examples a keyphasor transducer can beparticularly useul in diagnosing aults such as rotor bowHowever this equipment requires the ability to cut a key way or measurements to be perormed Optical sensors haverecently been applied to detect rotor unbalance yet such atechnology would be difficult to implement in a system withseveral rows o rotors (as in a gas turbine)

Whilst new diagnosis and prognosis techniques or aultsin rotating machinery are being continuously researchedthe lack o corresponding studies into localization can beconsidered one o the many challenges in promoting recentcore research into live industrial applications

983091983092 Fault Differentiation One promising development inrecent modeling and simulation o rotordynamic aults isthat o ault differentiation Several researchers have movedon rom studies on individual aults in order to concentrateon combinations o aults As stated rotordynamic aultssuch as those listed in this paper are linked to each otherSuch studies thereore concentrate on such topics as a

misalignment causing an unbalance or a looseness causinga rub Findings rom reports such as these are important inunderstanding complex anomalies In industrial applicationssimply detecting and rectiying an unbalance does not pro-

vide a satisactory solution i the root cause o the ault isa misalignment Tis is a complex topic as several aults canexhibit similar vibration characteristics making traditionaldetection techniques inaccurate in some cases

Making reerence to some o the aorementioned stud-ies those such as Bachschmid et al [983090983091] and Sonnichsen[983090983093] indicate some initial studies including several aultsTe results o such studies indicate a high success rate indifferentiating the aults in the given conditions however

a number o limitations still exist Studies such as theseconsider unbalancein the aultchains Tisis typically ldquostaticrdquoor simple unbalance applied to a single shafrotormdashin reality unbalance aults can be more complex than this In additionthe simple systems used to study and differentiate aultsare very different rom the complexities o a ull turbineHowever urther work may enable such research to ldquoscale-uprdquo to this stage with the aid o urther validation and

veri1047297cation

983091983093 Prognosis It can be seen rom the examples o recentresearch described or common aults that prognostic tech-niques are a topic where much work is being perormed

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 913

International Journal o Rotating Machinery 983097

Some o this work would be very difficult or impossibleto implement in an industrial situation others providepromising results o use to uture work in the area Predictingthe remaining useul lie o components is critical in thedevelopment o condition-based monitoring strategies orindustrial implementation

It can be noted that prognostic studies or certain rotor-dynamic aults are considerably more advanced than orothers Te most obvious examples o the aults detailed inthis paper are bearing aults shaf and blade cracks Teseare also areas where physics-driven simulations have hadan important impact Te ability to design any ault type(or combinations o aults) into such simulations (to be itslow-1047297delity mathematical models or high-1047297delity 1047297nite ele-ment analysis) has provided researchers with many differentavenues to explore Studies into the prognostics o otheraults can be complicated by various actors An unbalanceor example has had relatively little research perormed intoprognosis Te act that unbalance is ofen the cause o another underlying ault is one reason or the difficulty inresearching prognosis in detail or this ault Te results o unbalance are also very dependent on the severity o the aultand the system in which it occurs

It is worth noting that even or the aults where prognosisresearch is more advanced (eg shaf cracks) true develop-ment o the ault is not always studied Fan et al [983089983095] orexample study sets o discrete cracks in order to ldquoprognoserdquocrack growth In act thisstudy is more akinto detecting aultseverity with inerred effects on crack growth In contrast Liet al [983090983090] consider the continual development o such crackshowever the authors do admit this work is in its inancySuch points limit the ability or true prognosis to be achievedwithout extensive operational data

It is also worth noting that combinations o aults havebeen simulated extensively or the purpose o diagnosticshowever ew studies exist combining aultsor the purpose o prognostics Such studies are however a logical progressionrom some o those already perormed It may be that suchprognostic studies build upon some o the condition-basedmaintenance (CBM) or diagnostics High-level studies suchas Jaw and Merrill [983091983093] indicate how prognosis and diagnosistechniques can be combined into a CBM system Tis study is designed or a military aeroengine indicating the desire orsuch systems to be implementedmdashalthough the architecturedescribed has the capability o including various algorithmsoraultprognosis the amount o consideration orprognosis

o ault combinations is unclear

983091983094 Modelling Te large number o works reviewed or thispaper which includes modeling techniques in order to aiddiagnosis and prognosis indicates the possibilities providedby modern computer power and sofware developmentBroadly the modeling studied in the aorementioned papersbreaks down into two categories mathematical modeling and1047297nite element analysis (FEA) Te mathematical models suchasBorghesani etal [983090983089] ofen provide a theoretical basis uponwhich data driven or FEA studies can build and validateMathematical models despite being a traditional approach to

rotordynamic problems thereore continue to be developedto tackle more advanced problems

FEA techniques offer a constantly expanding area o simulation to explore Tey are used or research as diverseas crack propagation to unbalance localization [983091983094] Moderncodes such as NASRAN and Ansys enable rotordynamics to

be studied beyond the traditional stress and modal analysisTe ability to study the effects o rotation (eg throughCampbell diagrams and transient analysis) continues to driveresearch and innovation in this area Te studies discussedin this paper indicate the power o FEA with some highly accurate simulations having been perormed However thereis still a limitation o computer powermdashapplying some o the FEA techniques to whole engine models would proveto be too computationally expensive to be viable As aresult alternate methods such as model order reductionand system level modeling (and model-based reasoning orimplementation) are still required in order to make some o the FEA studies viable in an industrial environment

Nonlinearities continue to orm an important part o many recent studies with the complexity and detail o thespeci1047297ed aults being continually expanded Te depth o modeling has enabled in many cases accurate validationagainst experimental approaches Despite this completeunderstanding o the vibrational phenomena o rotatingmachinery is not currently possible despite the drive towardsthis end

983091983095 Rotordynamics and IVHM Te bulk o current researchinto rotordynamics rom the point o view o prognostichealth management (PHM) can be roughly divided intotwo types initial single-ault diagnosisprognosis techniquesand studies into the general requirements and limitationso current systems along with current and uture trendsAn example o the latter is Pusey [983091983095] who provides agood summary overview o current diagnosis and prognosistechniques with regard to condition-based maintenance

As a result o this split a clear gap exists between thecore research being perormed into rotordynamics rom acondition-based maintenance perspective and the identi1047297edneeds o industry aking a 1047298edgling piece o research andapplying it to a commercially-ready system (eg a gas turbineengine or an aircraf) is a long and complex task It isnevertheless worth noting that technologies or automatically detecting an unbalance or misalignment in a gas turbine were

developed over 983089983088 years beore the latest commercial aircrafwere conceptualized and yet these planes are still limited inthis capacity Tis highlights the need or work which linksthe undamental research into individual ault diagnosis toldquolive systemsrdquo in use in industry

Physics-based simulation and modeling o rotordynamicparts is a well-researched 1047297eld Such modeling has been usedas the basis o diagnosis and prognosis o aults by many researchers several recent examples have been outlined inthis paper Occasional pieces o work have been perormedinto modeling multiple aults such as Jain and Kundra [983091983096]who usea system model oronline identi1047297cation o unbalanceand cracks Beyond this however very limited research

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 1013

983089983088 International Journal o Rotating Machinery

Business casecost bene1047297tanalysis

Live system

RubCracksBearingfailure

Rotor bow Unbalance

Fault differentiationLocalisation

Root cause analysis

Remaining useful life

Data driven analysis and physics-based simulation

Mis-alignment

Experimental

1047297eld data

F983145983143983157983154983141 983089 Physics-based simulationmdashrom research to industry

exists Te demands o PHM techniques in industry are suchthat any systemmust notonly be capable o detecting multipleaults but must also be capable o detecting these aultsacrossa range o different systems Other considerations includethe aorementioned ability to differentiate between multipleaults Processing also needs to be taken into account as theobjective o these systems is to implement efficient conditionmonitoring and condition-based maintenance procedures I processing data is a long power-hungry process then thisaim cannot be achieved

Figure 983089 details a potential ramework required in order

to push core research such as that detailed in this reporttowards industrial applications Many studies now exist onindividual rotordynamic aults across a wide range o condi-tions and applications Some studies have taken this urtherwith advanced prognostic models and diagnosis o dual aults(primary cause and secondary effect) Future research inthe area o rotordynamics rom a PHM perspective couldpotentially provide the bridge between these studies and livesystems by combining physics-based simulations with data-driven techniques and validating against experimental data

Although quantiying the success o the research studiedisdifficultit ispossibleto de1047297ne the key areas in whicha tech-nique must excel in order to be considered viable Te work

byWheeleretal[983091983097] discussesin detailmetrics ordiagnosticand prognostic analysis as does that by Vachtsevanos [983092983088]and Saxena et al [983092983089] Te conclusions drawn rom thesepapers and applied in practice to research like that covered by this paper indicate that the ollowing metrics are importantwhen considering the potential o a given technique to diag-nose aults coverage alse positive rate and alse negativerate In the case o prognosis probabilities and lead timeto ailure are other important considerations Tis criteriaenables research to be assessed in terms o its suitability orindustrial applications Unortunately inormation on these

metrics is not made readily available by the authors o mostpapers

In terms o evaluating the effectivenesso theresearch dis-cussed in this paper there are difficulties in recommending aspeci1047297c technique over others or general application Mostpapers reviewed or this document (not just those reerencedand discussed in detail) take a technique (new or evolved)

validate or a given speci1047297c system and report success o theresearch Despite this some conclusions can be drawn romassessing common techniques applied across different studiesand different aults Although it is not possible to de1047297ne themost common methods or diagnosing andprognosing aultsin terms o numbers (as it was not possible to cover all recent

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 1113

International Journal o Rotating Machinery 983089983089

rotordynamics research or this paper) as perceived by theauthors the ollowing techniques have eatured prominently in the reviewed research

(i) Sensors accelerometers

(ii) Teoretical studies mathematical modeling

(iii) Physics-based simulation FEA(iv) Data-driven joint timerequency domain analysis

(v) Implementation neural networking

Tese techniques appear to be among the most promisingcurrently under development as they tend to eature numer-ous timesamongst some o the work with wider scope acrossault types and with the most comprehensive validationTere are o course many subsections to these techniqueshowever it shows one general direction o research and theclear possibilities posed in these areas

Future developments in the 1047297eld o IVHM or rotat-ing machinery may incorporate these techniques alongsideextensive use o nonlinear modeling and multiple aultinteractions Te 1047297eld o design or IVHM has only recently emerged however the potential exists or speci1047297c nonlinear-ities to be designed into a system in order to enable accuratediagnosis andprognosis o aults Te development o currentalgorithms to include diagnosis localization and prognosiso a range o aults will provide a signi1047297cant advancement oruture generations o IVHM systems Tis combined withcost-effective sensor suites indicates the potential or evolu-tions o some o the research detailed hereto orm part onextgeneration IVHM suites or rotating machinery

4 Conclusion

Tis paper has reviewed some o the latest research around anumber o rotordynamic aults namely unbalance misalign-ment rub and looseness 1047298uid-induced instability bearingaults shaf cracks blade cracks and rotor bow Each aultwas reviewed rom the perspective o sensors diagnosisprognosis localization and modeling

Key examples o recent work into the eight describedaults have been detailed through works by a number o eminent authors Additional work has been summarized andormatted or easy reerence Some current trends amongstthe recent body o work include developments in the vastarea o modeling nonlinearities combinations o high andlow 1047297delity modeling and synergy between data driven and

physics-based simulation approachesDespite the large volume o promising research reviewed

urther development in a number o areas is required in orderto produce effective next generation IVHM systems As suchuture developments may include usion o andor devel-opment o current algorithms to encompass all eight aultsdetailed consideration o prognosis diagnosis and localiza-tion achieved using a reduced cost-effective sensor suite

References

[983089] I K Jennions Integrated Vehicle Health Management Perspec-tives on an Emerging Field 983090983088983089983089

[983090] G Anderson ldquoProviding best-value IVHM solutions or agingaircrafrdquo in Proceedings o the 983097th Joint FAADoDNASA Aging Aircraf Conerence vol 983089 pp 983089ndash983089983089 983090983088983088983093

[983091] A Muszynska Rotordynamics aylor amp Francis 983089st ed edition983090983088983088983093

[983092] D E Bently Fundimentals o Rotating Machinery Diagnostics

ASME Press 983089st edition 983090983088983088983090[983093] B Domes ldquoVibration phenomena in aero-enginesrdquo in Proceed-ings o the 983097th International Conerence on Vibrations in Rotating Machinery Exeter (IMechE rsquo983088983096) vol 983089 pp 983089983093ndash983091983090 September983090983088983088983096

[983094] A Hess G Calvello and Dabney ldquoPHM a keyenabler or theJSF autonomie logistics support conceptrdquo in Proceedings o theIEEE Aerospace Conerence Proceedings pp 983091983093983092983091ndash983092983093983092983097 March983090983088983088983092

[983095] S N Ganeriwala B Schwarz and M H Richardson ldquoOperat-ing de1047298ection shapes detect unbalance in rotating equipmentrdquoSound and Vibration vol 983092983091 no 983093 pp 983089983089ndash983089983091 983090983088983088983097

[983096] G N D S Sudhakar and A S Sekhar ldquoIdenti1047297cation o unbal-ance in a rotor bearing systemrdquo Journal o Sound and Vibration

vol 983091983091983088 no 983089983088 pp 983090983090983097983097ndash983090983091983089983091 983090983088983089983089[983097] Yang and M W Hsu ldquoAn efficient diagnosis technique or

variations o shaf-bow and unbalancerdquo in Proceedings o the ASME International Design Engineering echnical Conerencesand Computers and Inormation in Engineering Conerence(DEC rsquo983088983097) pp 983093983095ndash983094983094 San Diego Cali USA September 983090983088983088983097

[983089983088] A El-Shaei S H aw1047297ck and M O A Mokhtar ldquoExperi-mental investigation o the effect o angular misalignemt onthe instability o plain journal bearingsrdquo in Proceedings o the ASMESLE InternationalJointribology Conerence(IJC rsquo983088983097)pp 983089983095983089ndash983089983095983091 Memphis enn USA October 983090983088983088983097

[983089983089] A El-Shaei S H aw1047297ck and M O A Mokhtar ldquoMisalign-ment modeling in rotating systemsrdquo in Proceedings o the ASMEurbo Expo H Bahaloo A Ebrahimi and M Samadi Eds vol

983089983095983089983089 pp 983097983095983091ndash983097983095983097 Orlando Fla USA June 983090983088983088983097

[983089983090] L F Villaet al ldquoStatistical diagnosisbased on vibration analysisor gear test-bench under non-stationary conditions o speedand loadrdquo Mechanical Systems and Signal Processing vol 983090983097 pp983092983091983094ndash983092983092983094 983090983088983089983090

[983089983091] C F Ngolah et al ldquoIntelligent ault recognition and diagnosisor rotating machines using neural networksrdquo Sofware Scienceand Computational Intelligence vol 983091 no 983092 983089983095 pages 983090983088983089983089

[983089983092] Y Lu Z Ren H Chen N Song and B Wen ldquoStudy on loose-ness and impactmdashrub coupling aults o a vertical dual-disk cantilever rotor- Bearing systemrdquo Key Engineering Materials vol 983091983093983091ndash983091983093983096 no 983092 pp 983090983092983095983097ndash983090983092983096983090 983090983088983088983095

[983089983093] Q Han Z Zhang and B Wen ldquoPeriodic motions o a dual-

disc rotor system with rub-impact at 1047297xed limiterrdquo Proceedingso the Institution o Mechanical Engineers vol 983090983090983090 no 983089983088 pp983089983097983091983093ndash983089983097983092983094 983090983088983088983096

[983089983094] H F de Castro K L Cavalca and R Nordmann ldquoWhirland whip instabilities in rotor-bearing system considering anonlinear orce modelrdquo Journal o Sound and Vibration vol 983091983089983095no 983089-983090 pp 983090983095983091ndash983090983097983091 983090983088983088983096

[983089983095] C C Fan et al ldquoMechanical systems and signal processingrdquoStudy o Start-Up Vibration Response or Oil Whirl and Dry Whip vol 983090983093 pp 983091983089983088983090ndash983091983089983089983093 983090983088983089983089

[983089983096] K Kappaganthu and C Nataraj ldquoNonlinear modeling andanalysis o a rolling element bearing with a clearancerdquo Commu-nications in Nonlinear Science and Numerical Simulation vol983089983094no 983089983088 pp 983092983089983091983092ndash983092983089983092983093 983090983088983089983089

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 1213

983089983090 International Journal o Rotating Machinery

[983089983097] C Gupta K Gupta and D K Sehgal ldquoInstability and chaoso a 1047298exible rotor ball bearing system an investigation on thein1047298uence o rotating imbalance and bearing clearancerdquo Journal o Engineering or Gas urbines and Power vol 983089983091983091 no 983096ArticleID 983088983096983090983093983088983089 983089983089 pages 983090983088983089983089

[983090983088] J Hong X Miao L Han and Y Ma ldquoPrognostics model orpredicting aero-engine bearing grade-lierdquo in Proceedings o the ASME urbo Expo vol 983089 pp 983094983091983097ndash983094983092983095 Orlando Fla USA June983090983088983088983097

[983090983089] P Borghesani P Pennacchi R B Randall N Sawalhi and RRicci ldquoApplication o cepstrum pre-whitening or the diagnosiso bearing aults under variable speed conditionsrdquo Mechanical Systems and Signal Processing 983090983088983089983090

[983090983090] Y Li J Zhang L Dai Z Zhang and J Liu ldquoAuditory-model-based eature extraction method or mechanical aults diagno-sisrdquo ChineseJournal o MechanicalEngineering (EnglishEdition) vol 983090983091 no 983091 pp 983091983097983089ndash983091983097983095 983090983088983089983088

[983090983091] N Bachschmid P Pennacchi and E anzi ldquoCracked rotatingshafs typical behaviors modeling and diagnosisrdquo IUAM Symposium on Emerging rends in Rotor Dynamics vol 983089983088983089983089 pp

983092983092983089ndash983092983093983092 983090983088983089983089[983090983092] Inoue N Nagata and Y Ishida ldquoFem modelling and

experimental veri1047297cation o a rotor system with a open crackrdquoin Proceedings o the ASME International Design Engineering echnical Conerences and Computers and Inormation in Engi-neering Conerence (DEC rsquo983088983097) pp 983089983089983089983091ndash983089983089983090983090 San Diego CaliUSA September 983090983088983088983097

[983090983093] H E Sonnichsen ldquoReal-time detection o developing cracksin jet engine rotorsrdquo in Proceedings o the IEEE AerospaceConerence pp 983089983095983091ndash983089983096983092 March 983090983088983088983088

[983090983094] I Green and C Casey ldquoCrack detection in a rotor dynamicsystem by vibration monitoringmdashpart I analysisrdquo Journal o Engineering or Gas urbines and Power vol 983089983090983095 no 983090 pp 983092983090983093ndash983092983091983094 983090983088983088983093

[983090983095] J Sawicki M I Friswell A H Pesch and A WroblewskildquoCondition monitoring o rotor usingactive magnetic actuatorrdquoin Proceedings o the ASME urbo Expo pp 983089983090983093983095ndash983089983090983094983093 BerlinGermany June 983090983088983088983096

[983090983096] J Xiang X Chen Q Mo and Z He ldquoIdenti1047297cation o crack ina rotor system based on wavelet 1047297nite element methodrdquo FiniteElements in Analysis and Design vol 983092983091 no 983089983092 pp 983089983088983094983096ndash983089983088983096983089983090983088983088983095

[983090983097] M G MaalouldquoSlow speed vibrationsignal analysisi you canrsquotdo it slow you canrsquot do it astrdquo in Proceedings o the ASME urboExpo pp 983093983093983097ndash983093983094983095 Montreal Canada May 983090983088983088983095

[983091983088] J Meagher X Wu and C Lencioni ldquoResponse o a warped1047298exible rotor with a 1047298uid bearingrdquo International Journal o

Rotating Machinery vol 983090983088983088983096 Article ID 983089983092983095983094983093983091 983097 pages 983090983088983088983096

[983091983089] Gaka andM abaszewski ldquoAn applicationo statistical symp-toms in machine condition diagnosticsrdquo Mechanical Systemsand Signal Processing vol 983090983093 no 983089 pp 983090983093983091ndash983090983094983093 983090983088983089983089

[983091983090] J K Sinha ldquoRecent trands in ault quanti1047297cation in rotatingmachinesrdquo Advances in Vibration EngIneerIng vol 983096 no 983089 pp983095983097ndash983096983093 983090983088983088983097

[983091983091] X Shen JJia andM Zhao ldquoNonlinear analysis o a rub-impactrotor-bearing system with initial permanent rotor bowrdquo Archiveo Applied Mechanics vol 983095983096 no 983091 pp 983090983090983093ndash983090983092983088 983090983088983088983096

[983091983092] A W Lees J K Sinha and M I Friswell ldquoModel-based identi-1047297cation o rotating machinesrdquo Mechanical Systems and Signal Processing vol 983090983091 no 983094 pp 983089983096983096983092ndash983089983096983097983091 983090983088983088983097

[983091983093] L C Jaw and W Merrill ldquoCBM+ research environmentmdashacilitating technology development experimentation andmaturationrdquo in Proceedings o the IEEE Aerospace Conerence(AC rsquo983088983096) March 983090983088983088983096

[983091983094] R Walker S Perinpanayagam and I K Jennions ldquoSimulatingUnbalance or Future IVHM Applicationsrdquo in Proceedings o theSociety or Experimental Mechanics (IMAC rsquo983089983090) vol 983090983097 pp 983089983092983089ndash983089983092983096 May 983090983088983089983090

[983091983095] H C Pusey ldquourbomachinery condition monitoring and ail-ure prognosisrdquo Sound and Vibration vol 983092983089 no 983091 pp 983089983088ndash983089983093983090983088983088983095

[983091983096] J R Jain and K Kundra ldquoModel based online diagnosiso unbalance and transverse atigue crack in rotor systemsrdquo Mechanics Research Communications vol 983091983089 no 983093 pp 983093983093983095ndash983093983094983096983090983088983088983092

[983091983097] K R Wheeler Kurtoglu and S D Poll ldquoA survey o healthmanagement user objectives related to diagnostic and prognos-tic metricsrdquo in Proceedings o the 983090983097th Computers and Inor-mation Engineering Conerence vol 983090 pp 983089983090983096983095ndash983089983090983097983096 983090983088983088983097

[983092983088] G Vachtsevanos ldquoPerormance metrics or ault prognosis o

complex systemsrdquo in IEEE Systems Readiness echnology Con- erence (AUOESCON rsquo983088983091) pp 983091983092983089ndash983091983092983093 September 983090983088983088983091

[983092983089] A Saxena J Celaya E Balaban et al ldquoMetrics or evaluatingperormance o prognostic techniquesrdquo in Proceedings o theInternational Conerence on Prognostics and Health Manage-ment (PHM rsquo983088983096) pp 983089ndash983089983095 October 983090983088983088983096

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 1313

Submit your manuscripts at

httpwwwhindawicom

Page 9: Rotordynamic Faults 2013

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 913

International Journal o Rotating Machinery 983097

Some o this work would be very difficult or impossibleto implement in an industrial situation others providepromising results o use to uture work in the area Predictingthe remaining useul lie o components is critical in thedevelopment o condition-based monitoring strategies orindustrial implementation

It can be noted that prognostic studies or certain rotor-dynamic aults are considerably more advanced than orothers Te most obvious examples o the aults detailed inthis paper are bearing aults shaf and blade cracks Teseare also areas where physics-driven simulations have hadan important impact Te ability to design any ault type(or combinations o aults) into such simulations (to be itslow-1047297delity mathematical models or high-1047297delity 1047297nite ele-ment analysis) has provided researchers with many differentavenues to explore Studies into the prognostics o otheraults can be complicated by various actors An unbalanceor example has had relatively little research perormed intoprognosis Te act that unbalance is ofen the cause o another underlying ault is one reason or the difficulty inresearching prognosis in detail or this ault Te results o unbalance are also very dependent on the severity o the aultand the system in which it occurs

It is worth noting that even or the aults where prognosisresearch is more advanced (eg shaf cracks) true develop-ment o the ault is not always studied Fan et al [983089983095] orexample study sets o discrete cracks in order to ldquoprognoserdquocrack growth In act thisstudy is more akinto detecting aultseverity with inerred effects on crack growth In contrast Liet al [983090983090] consider the continual development o such crackshowever the authors do admit this work is in its inancySuch points limit the ability or true prognosis to be achievedwithout extensive operational data

It is also worth noting that combinations o aults havebeen simulated extensively or the purpose o diagnosticshowever ew studies exist combining aultsor the purpose o prognostics Such studies are however a logical progressionrom some o those already perormed It may be that suchprognostic studies build upon some o the condition-basedmaintenance (CBM) or diagnostics High-level studies suchas Jaw and Merrill [983091983093] indicate how prognosis and diagnosistechniques can be combined into a CBM system Tis study is designed or a military aeroengine indicating the desire orsuch systems to be implementedmdashalthough the architecturedescribed has the capability o including various algorithmsoraultprognosis the amount o consideration orprognosis

o ault combinations is unclear

983091983094 Modelling Te large number o works reviewed or thispaper which includes modeling techniques in order to aiddiagnosis and prognosis indicates the possibilities providedby modern computer power and sofware developmentBroadly the modeling studied in the aorementioned papersbreaks down into two categories mathematical modeling and1047297nite element analysis (FEA) Te mathematical models suchasBorghesani etal [983090983089] ofen provide a theoretical basis uponwhich data driven or FEA studies can build and validateMathematical models despite being a traditional approach to

rotordynamic problems thereore continue to be developedto tackle more advanced problems

FEA techniques offer a constantly expanding area o simulation to explore Tey are used or research as diverseas crack propagation to unbalance localization [983091983094] Moderncodes such as NASRAN and Ansys enable rotordynamics to

be studied beyond the traditional stress and modal analysisTe ability to study the effects o rotation (eg throughCampbell diagrams and transient analysis) continues to driveresearch and innovation in this area Te studies discussedin this paper indicate the power o FEA with some highly accurate simulations having been perormed However thereis still a limitation o computer powermdashapplying some o the FEA techniques to whole engine models would proveto be too computationally expensive to be viable As aresult alternate methods such as model order reductionand system level modeling (and model-based reasoning orimplementation) are still required in order to make some o the FEA studies viable in an industrial environment

Nonlinearities continue to orm an important part o many recent studies with the complexity and detail o thespeci1047297ed aults being continually expanded Te depth o modeling has enabled in many cases accurate validationagainst experimental approaches Despite this completeunderstanding o the vibrational phenomena o rotatingmachinery is not currently possible despite the drive towardsthis end

983091983095 Rotordynamics and IVHM Te bulk o current researchinto rotordynamics rom the point o view o prognostichealth management (PHM) can be roughly divided intotwo types initial single-ault diagnosisprognosis techniquesand studies into the general requirements and limitationso current systems along with current and uture trendsAn example o the latter is Pusey [983091983095] who provides agood summary overview o current diagnosis and prognosistechniques with regard to condition-based maintenance

As a result o this split a clear gap exists between thecore research being perormed into rotordynamics rom acondition-based maintenance perspective and the identi1047297edneeds o industry aking a 1047298edgling piece o research andapplying it to a commercially-ready system (eg a gas turbineengine or an aircraf) is a long and complex task It isnevertheless worth noting that technologies or automatically detecting an unbalance or misalignment in a gas turbine were

developed over 983089983088 years beore the latest commercial aircrafwere conceptualized and yet these planes are still limited inthis capacity Tis highlights the need or work which linksthe undamental research into individual ault diagnosis toldquolive systemsrdquo in use in industry

Physics-based simulation and modeling o rotordynamicparts is a well-researched 1047297eld Such modeling has been usedas the basis o diagnosis and prognosis o aults by many researchers several recent examples have been outlined inthis paper Occasional pieces o work have been perormedinto modeling multiple aults such as Jain and Kundra [983091983096]who usea system model oronline identi1047297cation o unbalanceand cracks Beyond this however very limited research

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 1013

983089983088 International Journal o Rotating Machinery

Business casecost bene1047297tanalysis

Live system

RubCracksBearingfailure

Rotor bow Unbalance

Fault differentiationLocalisation

Root cause analysis

Remaining useful life

Data driven analysis and physics-based simulation

Mis-alignment

Experimental

1047297eld data

F983145983143983157983154983141 983089 Physics-based simulationmdashrom research to industry

exists Te demands o PHM techniques in industry are suchthat any systemmust notonly be capable o detecting multipleaults but must also be capable o detecting these aultsacrossa range o different systems Other considerations includethe aorementioned ability to differentiate between multipleaults Processing also needs to be taken into account as theobjective o these systems is to implement efficient conditionmonitoring and condition-based maintenance procedures I processing data is a long power-hungry process then thisaim cannot be achieved

Figure 983089 details a potential ramework required in order

to push core research such as that detailed in this reporttowards industrial applications Many studies now exist onindividual rotordynamic aults across a wide range o condi-tions and applications Some studies have taken this urtherwith advanced prognostic models and diagnosis o dual aults(primary cause and secondary effect) Future research inthe area o rotordynamics rom a PHM perspective couldpotentially provide the bridge between these studies and livesystems by combining physics-based simulations with data-driven techniques and validating against experimental data

Although quantiying the success o the research studiedisdifficultit ispossibleto de1047297ne the key areas in whicha tech-nique must excel in order to be considered viable Te work

byWheeleretal[983091983097] discussesin detailmetrics ordiagnosticand prognostic analysis as does that by Vachtsevanos [983092983088]and Saxena et al [983092983089] Te conclusions drawn rom thesepapers and applied in practice to research like that covered by this paper indicate that the ollowing metrics are importantwhen considering the potential o a given technique to diag-nose aults coverage alse positive rate and alse negativerate In the case o prognosis probabilities and lead timeto ailure are other important considerations Tis criteriaenables research to be assessed in terms o its suitability orindustrial applications Unortunately inormation on these

metrics is not made readily available by the authors o mostpapers

In terms o evaluating the effectivenesso theresearch dis-cussed in this paper there are difficulties in recommending aspeci1047297c technique over others or general application Mostpapers reviewed or this document (not just those reerencedand discussed in detail) take a technique (new or evolved)

validate or a given speci1047297c system and report success o theresearch Despite this some conclusions can be drawn romassessing common techniques applied across different studiesand different aults Although it is not possible to de1047297ne themost common methods or diagnosing andprognosing aultsin terms o numbers (as it was not possible to cover all recent

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 1113

International Journal o Rotating Machinery 983089983089

rotordynamics research or this paper) as perceived by theauthors the ollowing techniques have eatured prominently in the reviewed research

(i) Sensors accelerometers

(ii) Teoretical studies mathematical modeling

(iii) Physics-based simulation FEA(iv) Data-driven joint timerequency domain analysis

(v) Implementation neural networking

Tese techniques appear to be among the most promisingcurrently under development as they tend to eature numer-ous timesamongst some o the work with wider scope acrossault types and with the most comprehensive validationTere are o course many subsections to these techniqueshowever it shows one general direction o research and theclear possibilities posed in these areas

Future developments in the 1047297eld o IVHM or rotat-ing machinery may incorporate these techniques alongsideextensive use o nonlinear modeling and multiple aultinteractions Te 1047297eld o design or IVHM has only recently emerged however the potential exists or speci1047297c nonlinear-ities to be designed into a system in order to enable accuratediagnosis andprognosis o aults Te development o currentalgorithms to include diagnosis localization and prognosiso a range o aults will provide a signi1047297cant advancement oruture generations o IVHM systems Tis combined withcost-effective sensor suites indicates the potential or evolu-tions o some o the research detailed hereto orm part onextgeneration IVHM suites or rotating machinery

4 Conclusion

Tis paper has reviewed some o the latest research around anumber o rotordynamic aults namely unbalance misalign-ment rub and looseness 1047298uid-induced instability bearingaults shaf cracks blade cracks and rotor bow Each aultwas reviewed rom the perspective o sensors diagnosisprognosis localization and modeling

Key examples o recent work into the eight describedaults have been detailed through works by a number o eminent authors Additional work has been summarized andormatted or easy reerence Some current trends amongstthe recent body o work include developments in the vastarea o modeling nonlinearities combinations o high andlow 1047297delity modeling and synergy between data driven and

physics-based simulation approachesDespite the large volume o promising research reviewed

urther development in a number o areas is required in orderto produce effective next generation IVHM systems As suchuture developments may include usion o andor devel-opment o current algorithms to encompass all eight aultsdetailed consideration o prognosis diagnosis and localiza-tion achieved using a reduced cost-effective sensor suite

References

[983089] I K Jennions Integrated Vehicle Health Management Perspec-tives on an Emerging Field 983090983088983089983089

[983090] G Anderson ldquoProviding best-value IVHM solutions or agingaircrafrdquo in Proceedings o the 983097th Joint FAADoDNASA Aging Aircraf Conerence vol 983089 pp 983089ndash983089983089 983090983088983088983093

[983091] A Muszynska Rotordynamics aylor amp Francis 983089st ed edition983090983088983088983093

[983092] D E Bently Fundimentals o Rotating Machinery Diagnostics

ASME Press 983089st edition 983090983088983088983090[983093] B Domes ldquoVibration phenomena in aero-enginesrdquo in Proceed-ings o the 983097th International Conerence on Vibrations in Rotating Machinery Exeter (IMechE rsquo983088983096) vol 983089 pp 983089983093ndash983091983090 September983090983088983088983096

[983094] A Hess G Calvello and Dabney ldquoPHM a keyenabler or theJSF autonomie logistics support conceptrdquo in Proceedings o theIEEE Aerospace Conerence Proceedings pp 983091983093983092983091ndash983092983093983092983097 March983090983088983088983092

[983095] S N Ganeriwala B Schwarz and M H Richardson ldquoOperat-ing de1047298ection shapes detect unbalance in rotating equipmentrdquoSound and Vibration vol 983092983091 no 983093 pp 983089983089ndash983089983091 983090983088983088983097

[983096] G N D S Sudhakar and A S Sekhar ldquoIdenti1047297cation o unbal-ance in a rotor bearing systemrdquo Journal o Sound and Vibration

vol 983091983091983088 no 983089983088 pp 983090983090983097983097ndash983090983091983089983091 983090983088983089983089[983097] Yang and M W Hsu ldquoAn efficient diagnosis technique or

variations o shaf-bow and unbalancerdquo in Proceedings o the ASME International Design Engineering echnical Conerencesand Computers and Inormation in Engineering Conerence(DEC rsquo983088983097) pp 983093983095ndash983094983094 San Diego Cali USA September 983090983088983088983097

[983089983088] A El-Shaei S H aw1047297ck and M O A Mokhtar ldquoExperi-mental investigation o the effect o angular misalignemt onthe instability o plain journal bearingsrdquo in Proceedings o the ASMESLE InternationalJointribology Conerence(IJC rsquo983088983097)pp 983089983095983089ndash983089983095983091 Memphis enn USA October 983090983088983088983097

[983089983089] A El-Shaei S H aw1047297ck and M O A Mokhtar ldquoMisalign-ment modeling in rotating systemsrdquo in Proceedings o the ASMEurbo Expo H Bahaloo A Ebrahimi and M Samadi Eds vol

983089983095983089983089 pp 983097983095983091ndash983097983095983097 Orlando Fla USA June 983090983088983088983097

[983089983090] L F Villaet al ldquoStatistical diagnosisbased on vibration analysisor gear test-bench under non-stationary conditions o speedand loadrdquo Mechanical Systems and Signal Processing vol 983090983097 pp983092983091983094ndash983092983092983094 983090983088983089983090

[983089983091] C F Ngolah et al ldquoIntelligent ault recognition and diagnosisor rotating machines using neural networksrdquo Sofware Scienceand Computational Intelligence vol 983091 no 983092 983089983095 pages 983090983088983089983089

[983089983092] Y Lu Z Ren H Chen N Song and B Wen ldquoStudy on loose-ness and impactmdashrub coupling aults o a vertical dual-disk cantilever rotor- Bearing systemrdquo Key Engineering Materials vol 983091983093983091ndash983091983093983096 no 983092 pp 983090983092983095983097ndash983090983092983096983090 983090983088983088983095

[983089983093] Q Han Z Zhang and B Wen ldquoPeriodic motions o a dual-

disc rotor system with rub-impact at 1047297xed limiterrdquo Proceedingso the Institution o Mechanical Engineers vol 983090983090983090 no 983089983088 pp983089983097983091983093ndash983089983097983092983094 983090983088983088983096

[983089983094] H F de Castro K L Cavalca and R Nordmann ldquoWhirland whip instabilities in rotor-bearing system considering anonlinear orce modelrdquo Journal o Sound and Vibration vol 983091983089983095no 983089-983090 pp 983090983095983091ndash983090983097983091 983090983088983088983096

[983089983095] C C Fan et al ldquoMechanical systems and signal processingrdquoStudy o Start-Up Vibration Response or Oil Whirl and Dry Whip vol 983090983093 pp 983091983089983088983090ndash983091983089983089983093 983090983088983089983089

[983089983096] K Kappaganthu and C Nataraj ldquoNonlinear modeling andanalysis o a rolling element bearing with a clearancerdquo Commu-nications in Nonlinear Science and Numerical Simulation vol983089983094no 983089983088 pp 983092983089983091983092ndash983092983089983092983093 983090983088983089983089

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 1213

983089983090 International Journal o Rotating Machinery

[983089983097] C Gupta K Gupta and D K Sehgal ldquoInstability and chaoso a 1047298exible rotor ball bearing system an investigation on thein1047298uence o rotating imbalance and bearing clearancerdquo Journal o Engineering or Gas urbines and Power vol 983089983091983091 no 983096ArticleID 983088983096983090983093983088983089 983089983089 pages 983090983088983089983089

[983090983088] J Hong X Miao L Han and Y Ma ldquoPrognostics model orpredicting aero-engine bearing grade-lierdquo in Proceedings o the ASME urbo Expo vol 983089 pp 983094983091983097ndash983094983092983095 Orlando Fla USA June983090983088983088983097

[983090983089] P Borghesani P Pennacchi R B Randall N Sawalhi and RRicci ldquoApplication o cepstrum pre-whitening or the diagnosiso bearing aults under variable speed conditionsrdquo Mechanical Systems and Signal Processing 983090983088983089983090

[983090983090] Y Li J Zhang L Dai Z Zhang and J Liu ldquoAuditory-model-based eature extraction method or mechanical aults diagno-sisrdquo ChineseJournal o MechanicalEngineering (EnglishEdition) vol 983090983091 no 983091 pp 983091983097983089ndash983091983097983095 983090983088983089983088

[983090983091] N Bachschmid P Pennacchi and E anzi ldquoCracked rotatingshafs typical behaviors modeling and diagnosisrdquo IUAM Symposium on Emerging rends in Rotor Dynamics vol 983089983088983089983089 pp

983092983092983089ndash983092983093983092 983090983088983089983089[983090983092] Inoue N Nagata and Y Ishida ldquoFem modelling and

experimental veri1047297cation o a rotor system with a open crackrdquoin Proceedings o the ASME International Design Engineering echnical Conerences and Computers and Inormation in Engi-neering Conerence (DEC rsquo983088983097) pp 983089983089983089983091ndash983089983089983090983090 San Diego CaliUSA September 983090983088983088983097

[983090983093] H E Sonnichsen ldquoReal-time detection o developing cracksin jet engine rotorsrdquo in Proceedings o the IEEE AerospaceConerence pp 983089983095983091ndash983089983096983092 March 983090983088983088983088

[983090983094] I Green and C Casey ldquoCrack detection in a rotor dynamicsystem by vibration monitoringmdashpart I analysisrdquo Journal o Engineering or Gas urbines and Power vol 983089983090983095 no 983090 pp 983092983090983093ndash983092983091983094 983090983088983088983093

[983090983095] J Sawicki M I Friswell A H Pesch and A WroblewskildquoCondition monitoring o rotor usingactive magnetic actuatorrdquoin Proceedings o the ASME urbo Expo pp 983089983090983093983095ndash983089983090983094983093 BerlinGermany June 983090983088983088983096

[983090983096] J Xiang X Chen Q Mo and Z He ldquoIdenti1047297cation o crack ina rotor system based on wavelet 1047297nite element methodrdquo FiniteElements in Analysis and Design vol 983092983091 no 983089983092 pp 983089983088983094983096ndash983089983088983096983089983090983088983088983095

[983090983097] M G MaalouldquoSlow speed vibrationsignal analysisi you canrsquotdo it slow you canrsquot do it astrdquo in Proceedings o the ASME urboExpo pp 983093983093983097ndash983093983094983095 Montreal Canada May 983090983088983088983095

[983091983088] J Meagher X Wu and C Lencioni ldquoResponse o a warped1047298exible rotor with a 1047298uid bearingrdquo International Journal o

Rotating Machinery vol 983090983088983088983096 Article ID 983089983092983095983094983093983091 983097 pages 983090983088983088983096

[983091983089] Gaka andM abaszewski ldquoAn applicationo statistical symp-toms in machine condition diagnosticsrdquo Mechanical Systemsand Signal Processing vol 983090983093 no 983089 pp 983090983093983091ndash983090983094983093 983090983088983089983089

[983091983090] J K Sinha ldquoRecent trands in ault quanti1047297cation in rotatingmachinesrdquo Advances in Vibration EngIneerIng vol 983096 no 983089 pp983095983097ndash983096983093 983090983088983088983097

[983091983091] X Shen JJia andM Zhao ldquoNonlinear analysis o a rub-impactrotor-bearing system with initial permanent rotor bowrdquo Archiveo Applied Mechanics vol 983095983096 no 983091 pp 983090983090983093ndash983090983092983088 983090983088983088983096

[983091983092] A W Lees J K Sinha and M I Friswell ldquoModel-based identi-1047297cation o rotating machinesrdquo Mechanical Systems and Signal Processing vol 983090983091 no 983094 pp 983089983096983096983092ndash983089983096983097983091 983090983088983088983097

[983091983093] L C Jaw and W Merrill ldquoCBM+ research environmentmdashacilitating technology development experimentation andmaturationrdquo in Proceedings o the IEEE Aerospace Conerence(AC rsquo983088983096) March 983090983088983088983096

[983091983094] R Walker S Perinpanayagam and I K Jennions ldquoSimulatingUnbalance or Future IVHM Applicationsrdquo in Proceedings o theSociety or Experimental Mechanics (IMAC rsquo983089983090) vol 983090983097 pp 983089983092983089ndash983089983092983096 May 983090983088983089983090

[983091983095] H C Pusey ldquourbomachinery condition monitoring and ail-ure prognosisrdquo Sound and Vibration vol 983092983089 no 983091 pp 983089983088ndash983089983093983090983088983088983095

[983091983096] J R Jain and K Kundra ldquoModel based online diagnosiso unbalance and transverse atigue crack in rotor systemsrdquo Mechanics Research Communications vol 983091983089 no 983093 pp 983093983093983095ndash983093983094983096983090983088983088983092

[983091983097] K R Wheeler Kurtoglu and S D Poll ldquoA survey o healthmanagement user objectives related to diagnostic and prognos-tic metricsrdquo in Proceedings o the 983090983097th Computers and Inor-mation Engineering Conerence vol 983090 pp 983089983090983096983095ndash983089983090983097983096 983090983088983088983097

[983092983088] G Vachtsevanos ldquoPerormance metrics or ault prognosis o

complex systemsrdquo in IEEE Systems Readiness echnology Con- erence (AUOESCON rsquo983088983091) pp 983091983092983089ndash983091983092983093 September 983090983088983088983091

[983092983089] A Saxena J Celaya E Balaban et al ldquoMetrics or evaluatingperormance o prognostic techniquesrdquo in Proceedings o theInternational Conerence on Prognostics and Health Manage-ment (PHM rsquo983088983096) pp 983089ndash983089983095 October 983090983088983088983096

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 1313

Submit your manuscripts at

httpwwwhindawicom

Page 10: Rotordynamic Faults 2013

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 1013

983089983088 International Journal o Rotating Machinery

Business casecost bene1047297tanalysis

Live system

RubCracksBearingfailure

Rotor bow Unbalance

Fault differentiationLocalisation

Root cause analysis

Remaining useful life

Data driven analysis and physics-based simulation

Mis-alignment

Experimental

1047297eld data

F983145983143983157983154983141 983089 Physics-based simulationmdashrom research to industry

exists Te demands o PHM techniques in industry are suchthat any systemmust notonly be capable o detecting multipleaults but must also be capable o detecting these aultsacrossa range o different systems Other considerations includethe aorementioned ability to differentiate between multipleaults Processing also needs to be taken into account as theobjective o these systems is to implement efficient conditionmonitoring and condition-based maintenance procedures I processing data is a long power-hungry process then thisaim cannot be achieved

Figure 983089 details a potential ramework required in order

to push core research such as that detailed in this reporttowards industrial applications Many studies now exist onindividual rotordynamic aults across a wide range o condi-tions and applications Some studies have taken this urtherwith advanced prognostic models and diagnosis o dual aults(primary cause and secondary effect) Future research inthe area o rotordynamics rom a PHM perspective couldpotentially provide the bridge between these studies and livesystems by combining physics-based simulations with data-driven techniques and validating against experimental data

Although quantiying the success o the research studiedisdifficultit ispossibleto de1047297ne the key areas in whicha tech-nique must excel in order to be considered viable Te work

byWheeleretal[983091983097] discussesin detailmetrics ordiagnosticand prognostic analysis as does that by Vachtsevanos [983092983088]and Saxena et al [983092983089] Te conclusions drawn rom thesepapers and applied in practice to research like that covered by this paper indicate that the ollowing metrics are importantwhen considering the potential o a given technique to diag-nose aults coverage alse positive rate and alse negativerate In the case o prognosis probabilities and lead timeto ailure are other important considerations Tis criteriaenables research to be assessed in terms o its suitability orindustrial applications Unortunately inormation on these

metrics is not made readily available by the authors o mostpapers

In terms o evaluating the effectivenesso theresearch dis-cussed in this paper there are difficulties in recommending aspeci1047297c technique over others or general application Mostpapers reviewed or this document (not just those reerencedand discussed in detail) take a technique (new or evolved)

validate or a given speci1047297c system and report success o theresearch Despite this some conclusions can be drawn romassessing common techniques applied across different studiesand different aults Although it is not possible to de1047297ne themost common methods or diagnosing andprognosing aultsin terms o numbers (as it was not possible to cover all recent

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 1113

International Journal o Rotating Machinery 983089983089

rotordynamics research or this paper) as perceived by theauthors the ollowing techniques have eatured prominently in the reviewed research

(i) Sensors accelerometers

(ii) Teoretical studies mathematical modeling

(iii) Physics-based simulation FEA(iv) Data-driven joint timerequency domain analysis

(v) Implementation neural networking

Tese techniques appear to be among the most promisingcurrently under development as they tend to eature numer-ous timesamongst some o the work with wider scope acrossault types and with the most comprehensive validationTere are o course many subsections to these techniqueshowever it shows one general direction o research and theclear possibilities posed in these areas

Future developments in the 1047297eld o IVHM or rotat-ing machinery may incorporate these techniques alongsideextensive use o nonlinear modeling and multiple aultinteractions Te 1047297eld o design or IVHM has only recently emerged however the potential exists or speci1047297c nonlinear-ities to be designed into a system in order to enable accuratediagnosis andprognosis o aults Te development o currentalgorithms to include diagnosis localization and prognosiso a range o aults will provide a signi1047297cant advancement oruture generations o IVHM systems Tis combined withcost-effective sensor suites indicates the potential or evolu-tions o some o the research detailed hereto orm part onextgeneration IVHM suites or rotating machinery

4 Conclusion

Tis paper has reviewed some o the latest research around anumber o rotordynamic aults namely unbalance misalign-ment rub and looseness 1047298uid-induced instability bearingaults shaf cracks blade cracks and rotor bow Each aultwas reviewed rom the perspective o sensors diagnosisprognosis localization and modeling

Key examples o recent work into the eight describedaults have been detailed through works by a number o eminent authors Additional work has been summarized andormatted or easy reerence Some current trends amongstthe recent body o work include developments in the vastarea o modeling nonlinearities combinations o high andlow 1047297delity modeling and synergy between data driven and

physics-based simulation approachesDespite the large volume o promising research reviewed

urther development in a number o areas is required in orderto produce effective next generation IVHM systems As suchuture developments may include usion o andor devel-opment o current algorithms to encompass all eight aultsdetailed consideration o prognosis diagnosis and localiza-tion achieved using a reduced cost-effective sensor suite

References

[983089] I K Jennions Integrated Vehicle Health Management Perspec-tives on an Emerging Field 983090983088983089983089

[983090] G Anderson ldquoProviding best-value IVHM solutions or agingaircrafrdquo in Proceedings o the 983097th Joint FAADoDNASA Aging Aircraf Conerence vol 983089 pp 983089ndash983089983089 983090983088983088983093

[983091] A Muszynska Rotordynamics aylor amp Francis 983089st ed edition983090983088983088983093

[983092] D E Bently Fundimentals o Rotating Machinery Diagnostics

ASME Press 983089st edition 983090983088983088983090[983093] B Domes ldquoVibration phenomena in aero-enginesrdquo in Proceed-ings o the 983097th International Conerence on Vibrations in Rotating Machinery Exeter (IMechE rsquo983088983096) vol 983089 pp 983089983093ndash983091983090 September983090983088983088983096

[983094] A Hess G Calvello and Dabney ldquoPHM a keyenabler or theJSF autonomie logistics support conceptrdquo in Proceedings o theIEEE Aerospace Conerence Proceedings pp 983091983093983092983091ndash983092983093983092983097 March983090983088983088983092

[983095] S N Ganeriwala B Schwarz and M H Richardson ldquoOperat-ing de1047298ection shapes detect unbalance in rotating equipmentrdquoSound and Vibration vol 983092983091 no 983093 pp 983089983089ndash983089983091 983090983088983088983097

[983096] G N D S Sudhakar and A S Sekhar ldquoIdenti1047297cation o unbal-ance in a rotor bearing systemrdquo Journal o Sound and Vibration

vol 983091983091983088 no 983089983088 pp 983090983090983097983097ndash983090983091983089983091 983090983088983089983089[983097] Yang and M W Hsu ldquoAn efficient diagnosis technique or

variations o shaf-bow and unbalancerdquo in Proceedings o the ASME International Design Engineering echnical Conerencesand Computers and Inormation in Engineering Conerence(DEC rsquo983088983097) pp 983093983095ndash983094983094 San Diego Cali USA September 983090983088983088983097

[983089983088] A El-Shaei S H aw1047297ck and M O A Mokhtar ldquoExperi-mental investigation o the effect o angular misalignemt onthe instability o plain journal bearingsrdquo in Proceedings o the ASMESLE InternationalJointribology Conerence(IJC rsquo983088983097)pp 983089983095983089ndash983089983095983091 Memphis enn USA October 983090983088983088983097

[983089983089] A El-Shaei S H aw1047297ck and M O A Mokhtar ldquoMisalign-ment modeling in rotating systemsrdquo in Proceedings o the ASMEurbo Expo H Bahaloo A Ebrahimi and M Samadi Eds vol

983089983095983089983089 pp 983097983095983091ndash983097983095983097 Orlando Fla USA June 983090983088983088983097

[983089983090] L F Villaet al ldquoStatistical diagnosisbased on vibration analysisor gear test-bench under non-stationary conditions o speedand loadrdquo Mechanical Systems and Signal Processing vol 983090983097 pp983092983091983094ndash983092983092983094 983090983088983089983090

[983089983091] C F Ngolah et al ldquoIntelligent ault recognition and diagnosisor rotating machines using neural networksrdquo Sofware Scienceand Computational Intelligence vol 983091 no 983092 983089983095 pages 983090983088983089983089

[983089983092] Y Lu Z Ren H Chen N Song and B Wen ldquoStudy on loose-ness and impactmdashrub coupling aults o a vertical dual-disk cantilever rotor- Bearing systemrdquo Key Engineering Materials vol 983091983093983091ndash983091983093983096 no 983092 pp 983090983092983095983097ndash983090983092983096983090 983090983088983088983095

[983089983093] Q Han Z Zhang and B Wen ldquoPeriodic motions o a dual-

disc rotor system with rub-impact at 1047297xed limiterrdquo Proceedingso the Institution o Mechanical Engineers vol 983090983090983090 no 983089983088 pp983089983097983091983093ndash983089983097983092983094 983090983088983088983096

[983089983094] H F de Castro K L Cavalca and R Nordmann ldquoWhirland whip instabilities in rotor-bearing system considering anonlinear orce modelrdquo Journal o Sound and Vibration vol 983091983089983095no 983089-983090 pp 983090983095983091ndash983090983097983091 983090983088983088983096

[983089983095] C C Fan et al ldquoMechanical systems and signal processingrdquoStudy o Start-Up Vibration Response or Oil Whirl and Dry Whip vol 983090983093 pp 983091983089983088983090ndash983091983089983089983093 983090983088983089983089

[983089983096] K Kappaganthu and C Nataraj ldquoNonlinear modeling andanalysis o a rolling element bearing with a clearancerdquo Commu-nications in Nonlinear Science and Numerical Simulation vol983089983094no 983089983088 pp 983092983089983091983092ndash983092983089983092983093 983090983088983089983089

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 1213

983089983090 International Journal o Rotating Machinery

[983089983097] C Gupta K Gupta and D K Sehgal ldquoInstability and chaoso a 1047298exible rotor ball bearing system an investigation on thein1047298uence o rotating imbalance and bearing clearancerdquo Journal o Engineering or Gas urbines and Power vol 983089983091983091 no 983096ArticleID 983088983096983090983093983088983089 983089983089 pages 983090983088983089983089

[983090983088] J Hong X Miao L Han and Y Ma ldquoPrognostics model orpredicting aero-engine bearing grade-lierdquo in Proceedings o the ASME urbo Expo vol 983089 pp 983094983091983097ndash983094983092983095 Orlando Fla USA June983090983088983088983097

[983090983089] P Borghesani P Pennacchi R B Randall N Sawalhi and RRicci ldquoApplication o cepstrum pre-whitening or the diagnosiso bearing aults under variable speed conditionsrdquo Mechanical Systems and Signal Processing 983090983088983089983090

[983090983090] Y Li J Zhang L Dai Z Zhang and J Liu ldquoAuditory-model-based eature extraction method or mechanical aults diagno-sisrdquo ChineseJournal o MechanicalEngineering (EnglishEdition) vol 983090983091 no 983091 pp 983091983097983089ndash983091983097983095 983090983088983089983088

[983090983091] N Bachschmid P Pennacchi and E anzi ldquoCracked rotatingshafs typical behaviors modeling and diagnosisrdquo IUAM Symposium on Emerging rends in Rotor Dynamics vol 983089983088983089983089 pp

983092983092983089ndash983092983093983092 983090983088983089983089[983090983092] Inoue N Nagata and Y Ishida ldquoFem modelling and

experimental veri1047297cation o a rotor system with a open crackrdquoin Proceedings o the ASME International Design Engineering echnical Conerences and Computers and Inormation in Engi-neering Conerence (DEC rsquo983088983097) pp 983089983089983089983091ndash983089983089983090983090 San Diego CaliUSA September 983090983088983088983097

[983090983093] H E Sonnichsen ldquoReal-time detection o developing cracksin jet engine rotorsrdquo in Proceedings o the IEEE AerospaceConerence pp 983089983095983091ndash983089983096983092 March 983090983088983088983088

[983090983094] I Green and C Casey ldquoCrack detection in a rotor dynamicsystem by vibration monitoringmdashpart I analysisrdquo Journal o Engineering or Gas urbines and Power vol 983089983090983095 no 983090 pp 983092983090983093ndash983092983091983094 983090983088983088983093

[983090983095] J Sawicki M I Friswell A H Pesch and A WroblewskildquoCondition monitoring o rotor usingactive magnetic actuatorrdquoin Proceedings o the ASME urbo Expo pp 983089983090983093983095ndash983089983090983094983093 BerlinGermany June 983090983088983088983096

[983090983096] J Xiang X Chen Q Mo and Z He ldquoIdenti1047297cation o crack ina rotor system based on wavelet 1047297nite element methodrdquo FiniteElements in Analysis and Design vol 983092983091 no 983089983092 pp 983089983088983094983096ndash983089983088983096983089983090983088983088983095

[983090983097] M G MaalouldquoSlow speed vibrationsignal analysisi you canrsquotdo it slow you canrsquot do it astrdquo in Proceedings o the ASME urboExpo pp 983093983093983097ndash983093983094983095 Montreal Canada May 983090983088983088983095

[983091983088] J Meagher X Wu and C Lencioni ldquoResponse o a warped1047298exible rotor with a 1047298uid bearingrdquo International Journal o

Rotating Machinery vol 983090983088983088983096 Article ID 983089983092983095983094983093983091 983097 pages 983090983088983088983096

[983091983089] Gaka andM abaszewski ldquoAn applicationo statistical symp-toms in machine condition diagnosticsrdquo Mechanical Systemsand Signal Processing vol 983090983093 no 983089 pp 983090983093983091ndash983090983094983093 983090983088983089983089

[983091983090] J K Sinha ldquoRecent trands in ault quanti1047297cation in rotatingmachinesrdquo Advances in Vibration EngIneerIng vol 983096 no 983089 pp983095983097ndash983096983093 983090983088983088983097

[983091983091] X Shen JJia andM Zhao ldquoNonlinear analysis o a rub-impactrotor-bearing system with initial permanent rotor bowrdquo Archiveo Applied Mechanics vol 983095983096 no 983091 pp 983090983090983093ndash983090983092983088 983090983088983088983096

[983091983092] A W Lees J K Sinha and M I Friswell ldquoModel-based identi-1047297cation o rotating machinesrdquo Mechanical Systems and Signal Processing vol 983090983091 no 983094 pp 983089983096983096983092ndash983089983096983097983091 983090983088983088983097

[983091983093] L C Jaw and W Merrill ldquoCBM+ research environmentmdashacilitating technology development experimentation andmaturationrdquo in Proceedings o the IEEE Aerospace Conerence(AC rsquo983088983096) March 983090983088983088983096

[983091983094] R Walker S Perinpanayagam and I K Jennions ldquoSimulatingUnbalance or Future IVHM Applicationsrdquo in Proceedings o theSociety or Experimental Mechanics (IMAC rsquo983089983090) vol 983090983097 pp 983089983092983089ndash983089983092983096 May 983090983088983089983090

[983091983095] H C Pusey ldquourbomachinery condition monitoring and ail-ure prognosisrdquo Sound and Vibration vol 983092983089 no 983091 pp 983089983088ndash983089983093983090983088983088983095

[983091983096] J R Jain and K Kundra ldquoModel based online diagnosiso unbalance and transverse atigue crack in rotor systemsrdquo Mechanics Research Communications vol 983091983089 no 983093 pp 983093983093983095ndash983093983094983096983090983088983088983092

[983091983097] K R Wheeler Kurtoglu and S D Poll ldquoA survey o healthmanagement user objectives related to diagnostic and prognos-tic metricsrdquo in Proceedings o the 983090983097th Computers and Inor-mation Engineering Conerence vol 983090 pp 983089983090983096983095ndash983089983090983097983096 983090983088983088983097

[983092983088] G Vachtsevanos ldquoPerormance metrics or ault prognosis o

complex systemsrdquo in IEEE Systems Readiness echnology Con- erence (AUOESCON rsquo983088983091) pp 983091983092983089ndash983091983092983093 September 983090983088983088983091

[983092983089] A Saxena J Celaya E Balaban et al ldquoMetrics or evaluatingperormance o prognostic techniquesrdquo in Proceedings o theInternational Conerence on Prognostics and Health Manage-ment (PHM rsquo983088983096) pp 983089ndash983089983095 October 983090983088983088983096

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 1313

Submit your manuscripts at

httpwwwhindawicom

Page 11: Rotordynamic Faults 2013

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 1113

International Journal o Rotating Machinery 983089983089

rotordynamics research or this paper) as perceived by theauthors the ollowing techniques have eatured prominently in the reviewed research

(i) Sensors accelerometers

(ii) Teoretical studies mathematical modeling

(iii) Physics-based simulation FEA(iv) Data-driven joint timerequency domain analysis

(v) Implementation neural networking

Tese techniques appear to be among the most promisingcurrently under development as they tend to eature numer-ous timesamongst some o the work with wider scope acrossault types and with the most comprehensive validationTere are o course many subsections to these techniqueshowever it shows one general direction o research and theclear possibilities posed in these areas

Future developments in the 1047297eld o IVHM or rotat-ing machinery may incorporate these techniques alongsideextensive use o nonlinear modeling and multiple aultinteractions Te 1047297eld o design or IVHM has only recently emerged however the potential exists or speci1047297c nonlinear-ities to be designed into a system in order to enable accuratediagnosis andprognosis o aults Te development o currentalgorithms to include diagnosis localization and prognosiso a range o aults will provide a signi1047297cant advancement oruture generations o IVHM systems Tis combined withcost-effective sensor suites indicates the potential or evolu-tions o some o the research detailed hereto orm part onextgeneration IVHM suites or rotating machinery

4 Conclusion

Tis paper has reviewed some o the latest research around anumber o rotordynamic aults namely unbalance misalign-ment rub and looseness 1047298uid-induced instability bearingaults shaf cracks blade cracks and rotor bow Each aultwas reviewed rom the perspective o sensors diagnosisprognosis localization and modeling

Key examples o recent work into the eight describedaults have been detailed through works by a number o eminent authors Additional work has been summarized andormatted or easy reerence Some current trends amongstthe recent body o work include developments in the vastarea o modeling nonlinearities combinations o high andlow 1047297delity modeling and synergy between data driven and

physics-based simulation approachesDespite the large volume o promising research reviewed

urther development in a number o areas is required in orderto produce effective next generation IVHM systems As suchuture developments may include usion o andor devel-opment o current algorithms to encompass all eight aultsdetailed consideration o prognosis diagnosis and localiza-tion achieved using a reduced cost-effective sensor suite

References

[983089] I K Jennions Integrated Vehicle Health Management Perspec-tives on an Emerging Field 983090983088983089983089

[983090] G Anderson ldquoProviding best-value IVHM solutions or agingaircrafrdquo in Proceedings o the 983097th Joint FAADoDNASA Aging Aircraf Conerence vol 983089 pp 983089ndash983089983089 983090983088983088983093

[983091] A Muszynska Rotordynamics aylor amp Francis 983089st ed edition983090983088983088983093

[983092] D E Bently Fundimentals o Rotating Machinery Diagnostics

ASME Press 983089st edition 983090983088983088983090[983093] B Domes ldquoVibration phenomena in aero-enginesrdquo in Proceed-ings o the 983097th International Conerence on Vibrations in Rotating Machinery Exeter (IMechE rsquo983088983096) vol 983089 pp 983089983093ndash983091983090 September983090983088983088983096

[983094] A Hess G Calvello and Dabney ldquoPHM a keyenabler or theJSF autonomie logistics support conceptrdquo in Proceedings o theIEEE Aerospace Conerence Proceedings pp 983091983093983092983091ndash983092983093983092983097 March983090983088983088983092

[983095] S N Ganeriwala B Schwarz and M H Richardson ldquoOperat-ing de1047298ection shapes detect unbalance in rotating equipmentrdquoSound and Vibration vol 983092983091 no 983093 pp 983089983089ndash983089983091 983090983088983088983097

[983096] G N D S Sudhakar and A S Sekhar ldquoIdenti1047297cation o unbal-ance in a rotor bearing systemrdquo Journal o Sound and Vibration

vol 983091983091983088 no 983089983088 pp 983090983090983097983097ndash983090983091983089983091 983090983088983089983089[983097] Yang and M W Hsu ldquoAn efficient diagnosis technique or

variations o shaf-bow and unbalancerdquo in Proceedings o the ASME International Design Engineering echnical Conerencesand Computers and Inormation in Engineering Conerence(DEC rsquo983088983097) pp 983093983095ndash983094983094 San Diego Cali USA September 983090983088983088983097

[983089983088] A El-Shaei S H aw1047297ck and M O A Mokhtar ldquoExperi-mental investigation o the effect o angular misalignemt onthe instability o plain journal bearingsrdquo in Proceedings o the ASMESLE InternationalJointribology Conerence(IJC rsquo983088983097)pp 983089983095983089ndash983089983095983091 Memphis enn USA October 983090983088983088983097

[983089983089] A El-Shaei S H aw1047297ck and M O A Mokhtar ldquoMisalign-ment modeling in rotating systemsrdquo in Proceedings o the ASMEurbo Expo H Bahaloo A Ebrahimi and M Samadi Eds vol

983089983095983089983089 pp 983097983095983091ndash983097983095983097 Orlando Fla USA June 983090983088983088983097

[983089983090] L F Villaet al ldquoStatistical diagnosisbased on vibration analysisor gear test-bench under non-stationary conditions o speedand loadrdquo Mechanical Systems and Signal Processing vol 983090983097 pp983092983091983094ndash983092983092983094 983090983088983089983090

[983089983091] C F Ngolah et al ldquoIntelligent ault recognition and diagnosisor rotating machines using neural networksrdquo Sofware Scienceand Computational Intelligence vol 983091 no 983092 983089983095 pages 983090983088983089983089

[983089983092] Y Lu Z Ren H Chen N Song and B Wen ldquoStudy on loose-ness and impactmdashrub coupling aults o a vertical dual-disk cantilever rotor- Bearing systemrdquo Key Engineering Materials vol 983091983093983091ndash983091983093983096 no 983092 pp 983090983092983095983097ndash983090983092983096983090 983090983088983088983095

[983089983093] Q Han Z Zhang and B Wen ldquoPeriodic motions o a dual-

disc rotor system with rub-impact at 1047297xed limiterrdquo Proceedingso the Institution o Mechanical Engineers vol 983090983090983090 no 983089983088 pp983089983097983091983093ndash983089983097983092983094 983090983088983088983096

[983089983094] H F de Castro K L Cavalca and R Nordmann ldquoWhirland whip instabilities in rotor-bearing system considering anonlinear orce modelrdquo Journal o Sound and Vibration vol 983091983089983095no 983089-983090 pp 983090983095983091ndash983090983097983091 983090983088983088983096

[983089983095] C C Fan et al ldquoMechanical systems and signal processingrdquoStudy o Start-Up Vibration Response or Oil Whirl and Dry Whip vol 983090983093 pp 983091983089983088983090ndash983091983089983089983093 983090983088983089983089

[983089983096] K Kappaganthu and C Nataraj ldquoNonlinear modeling andanalysis o a rolling element bearing with a clearancerdquo Commu-nications in Nonlinear Science and Numerical Simulation vol983089983094no 983089983088 pp 983092983089983091983092ndash983092983089983092983093 983090983088983089983089

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 1213

983089983090 International Journal o Rotating Machinery

[983089983097] C Gupta K Gupta and D K Sehgal ldquoInstability and chaoso a 1047298exible rotor ball bearing system an investigation on thein1047298uence o rotating imbalance and bearing clearancerdquo Journal o Engineering or Gas urbines and Power vol 983089983091983091 no 983096ArticleID 983088983096983090983093983088983089 983089983089 pages 983090983088983089983089

[983090983088] J Hong X Miao L Han and Y Ma ldquoPrognostics model orpredicting aero-engine bearing grade-lierdquo in Proceedings o the ASME urbo Expo vol 983089 pp 983094983091983097ndash983094983092983095 Orlando Fla USA June983090983088983088983097

[983090983089] P Borghesani P Pennacchi R B Randall N Sawalhi and RRicci ldquoApplication o cepstrum pre-whitening or the diagnosiso bearing aults under variable speed conditionsrdquo Mechanical Systems and Signal Processing 983090983088983089983090

[983090983090] Y Li J Zhang L Dai Z Zhang and J Liu ldquoAuditory-model-based eature extraction method or mechanical aults diagno-sisrdquo ChineseJournal o MechanicalEngineering (EnglishEdition) vol 983090983091 no 983091 pp 983091983097983089ndash983091983097983095 983090983088983089983088

[983090983091] N Bachschmid P Pennacchi and E anzi ldquoCracked rotatingshafs typical behaviors modeling and diagnosisrdquo IUAM Symposium on Emerging rends in Rotor Dynamics vol 983089983088983089983089 pp

983092983092983089ndash983092983093983092 983090983088983089983089[983090983092] Inoue N Nagata and Y Ishida ldquoFem modelling and

experimental veri1047297cation o a rotor system with a open crackrdquoin Proceedings o the ASME International Design Engineering echnical Conerences and Computers and Inormation in Engi-neering Conerence (DEC rsquo983088983097) pp 983089983089983089983091ndash983089983089983090983090 San Diego CaliUSA September 983090983088983088983097

[983090983093] H E Sonnichsen ldquoReal-time detection o developing cracksin jet engine rotorsrdquo in Proceedings o the IEEE AerospaceConerence pp 983089983095983091ndash983089983096983092 March 983090983088983088983088

[983090983094] I Green and C Casey ldquoCrack detection in a rotor dynamicsystem by vibration monitoringmdashpart I analysisrdquo Journal o Engineering or Gas urbines and Power vol 983089983090983095 no 983090 pp 983092983090983093ndash983092983091983094 983090983088983088983093

[983090983095] J Sawicki M I Friswell A H Pesch and A WroblewskildquoCondition monitoring o rotor usingactive magnetic actuatorrdquoin Proceedings o the ASME urbo Expo pp 983089983090983093983095ndash983089983090983094983093 BerlinGermany June 983090983088983088983096

[983090983096] J Xiang X Chen Q Mo and Z He ldquoIdenti1047297cation o crack ina rotor system based on wavelet 1047297nite element methodrdquo FiniteElements in Analysis and Design vol 983092983091 no 983089983092 pp 983089983088983094983096ndash983089983088983096983089983090983088983088983095

[983090983097] M G MaalouldquoSlow speed vibrationsignal analysisi you canrsquotdo it slow you canrsquot do it astrdquo in Proceedings o the ASME urboExpo pp 983093983093983097ndash983093983094983095 Montreal Canada May 983090983088983088983095

[983091983088] J Meagher X Wu and C Lencioni ldquoResponse o a warped1047298exible rotor with a 1047298uid bearingrdquo International Journal o

Rotating Machinery vol 983090983088983088983096 Article ID 983089983092983095983094983093983091 983097 pages 983090983088983088983096

[983091983089] Gaka andM abaszewski ldquoAn applicationo statistical symp-toms in machine condition diagnosticsrdquo Mechanical Systemsand Signal Processing vol 983090983093 no 983089 pp 983090983093983091ndash983090983094983093 983090983088983089983089

[983091983090] J K Sinha ldquoRecent trands in ault quanti1047297cation in rotatingmachinesrdquo Advances in Vibration EngIneerIng vol 983096 no 983089 pp983095983097ndash983096983093 983090983088983088983097

[983091983091] X Shen JJia andM Zhao ldquoNonlinear analysis o a rub-impactrotor-bearing system with initial permanent rotor bowrdquo Archiveo Applied Mechanics vol 983095983096 no 983091 pp 983090983090983093ndash983090983092983088 983090983088983088983096

[983091983092] A W Lees J K Sinha and M I Friswell ldquoModel-based identi-1047297cation o rotating machinesrdquo Mechanical Systems and Signal Processing vol 983090983091 no 983094 pp 983089983096983096983092ndash983089983096983097983091 983090983088983088983097

[983091983093] L C Jaw and W Merrill ldquoCBM+ research environmentmdashacilitating technology development experimentation andmaturationrdquo in Proceedings o the IEEE Aerospace Conerence(AC rsquo983088983096) March 983090983088983088983096

[983091983094] R Walker S Perinpanayagam and I K Jennions ldquoSimulatingUnbalance or Future IVHM Applicationsrdquo in Proceedings o theSociety or Experimental Mechanics (IMAC rsquo983089983090) vol 983090983097 pp 983089983092983089ndash983089983092983096 May 983090983088983089983090

[983091983095] H C Pusey ldquourbomachinery condition monitoring and ail-ure prognosisrdquo Sound and Vibration vol 983092983089 no 983091 pp 983089983088ndash983089983093983090983088983088983095

[983091983096] J R Jain and K Kundra ldquoModel based online diagnosiso unbalance and transverse atigue crack in rotor systemsrdquo Mechanics Research Communications vol 983091983089 no 983093 pp 983093983093983095ndash983093983094983096983090983088983088983092

[983091983097] K R Wheeler Kurtoglu and S D Poll ldquoA survey o healthmanagement user objectives related to diagnostic and prognos-tic metricsrdquo in Proceedings o the 983090983097th Computers and Inor-mation Engineering Conerence vol 983090 pp 983089983090983096983095ndash983089983090983097983096 983090983088983088983097

[983092983088] G Vachtsevanos ldquoPerormance metrics or ault prognosis o

complex systemsrdquo in IEEE Systems Readiness echnology Con- erence (AUOESCON rsquo983088983091) pp 983091983092983089ndash983091983092983093 September 983090983088983088983091

[983092983089] A Saxena J Celaya E Balaban et al ldquoMetrics or evaluatingperormance o prognostic techniquesrdquo in Proceedings o theInternational Conerence on Prognostics and Health Manage-ment (PHM rsquo983088983096) pp 983089ndash983089983095 October 983090983088983088983096

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 1313

Submit your manuscripts at

httpwwwhindawicom

Page 12: Rotordynamic Faults 2013

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 1213

983089983090 International Journal o Rotating Machinery

[983089983097] C Gupta K Gupta and D K Sehgal ldquoInstability and chaoso a 1047298exible rotor ball bearing system an investigation on thein1047298uence o rotating imbalance and bearing clearancerdquo Journal o Engineering or Gas urbines and Power vol 983089983091983091 no 983096ArticleID 983088983096983090983093983088983089 983089983089 pages 983090983088983089983089

[983090983088] J Hong X Miao L Han and Y Ma ldquoPrognostics model orpredicting aero-engine bearing grade-lierdquo in Proceedings o the ASME urbo Expo vol 983089 pp 983094983091983097ndash983094983092983095 Orlando Fla USA June983090983088983088983097

[983090983089] P Borghesani P Pennacchi R B Randall N Sawalhi and RRicci ldquoApplication o cepstrum pre-whitening or the diagnosiso bearing aults under variable speed conditionsrdquo Mechanical Systems and Signal Processing 983090983088983089983090

[983090983090] Y Li J Zhang L Dai Z Zhang and J Liu ldquoAuditory-model-based eature extraction method or mechanical aults diagno-sisrdquo ChineseJournal o MechanicalEngineering (EnglishEdition) vol 983090983091 no 983091 pp 983091983097983089ndash983091983097983095 983090983088983089983088

[983090983091] N Bachschmid P Pennacchi and E anzi ldquoCracked rotatingshafs typical behaviors modeling and diagnosisrdquo IUAM Symposium on Emerging rends in Rotor Dynamics vol 983089983088983089983089 pp

983092983092983089ndash983092983093983092 983090983088983089983089[983090983092] Inoue N Nagata and Y Ishida ldquoFem modelling and

experimental veri1047297cation o a rotor system with a open crackrdquoin Proceedings o the ASME International Design Engineering echnical Conerences and Computers and Inormation in Engi-neering Conerence (DEC rsquo983088983097) pp 983089983089983089983091ndash983089983089983090983090 San Diego CaliUSA September 983090983088983088983097

[983090983093] H E Sonnichsen ldquoReal-time detection o developing cracksin jet engine rotorsrdquo in Proceedings o the IEEE AerospaceConerence pp 983089983095983091ndash983089983096983092 March 983090983088983088983088

[983090983094] I Green and C Casey ldquoCrack detection in a rotor dynamicsystem by vibration monitoringmdashpart I analysisrdquo Journal o Engineering or Gas urbines and Power vol 983089983090983095 no 983090 pp 983092983090983093ndash983092983091983094 983090983088983088983093

[983090983095] J Sawicki M I Friswell A H Pesch and A WroblewskildquoCondition monitoring o rotor usingactive magnetic actuatorrdquoin Proceedings o the ASME urbo Expo pp 983089983090983093983095ndash983089983090983094983093 BerlinGermany June 983090983088983088983096

[983090983096] J Xiang X Chen Q Mo and Z He ldquoIdenti1047297cation o crack ina rotor system based on wavelet 1047297nite element methodrdquo FiniteElements in Analysis and Design vol 983092983091 no 983089983092 pp 983089983088983094983096ndash983089983088983096983089983090983088983088983095

[983090983097] M G MaalouldquoSlow speed vibrationsignal analysisi you canrsquotdo it slow you canrsquot do it astrdquo in Proceedings o the ASME urboExpo pp 983093983093983097ndash983093983094983095 Montreal Canada May 983090983088983088983095

[983091983088] J Meagher X Wu and C Lencioni ldquoResponse o a warped1047298exible rotor with a 1047298uid bearingrdquo International Journal o

Rotating Machinery vol 983090983088983088983096 Article ID 983089983092983095983094983093983091 983097 pages 983090983088983088983096

[983091983089] Gaka andM abaszewski ldquoAn applicationo statistical symp-toms in machine condition diagnosticsrdquo Mechanical Systemsand Signal Processing vol 983090983093 no 983089 pp 983090983093983091ndash983090983094983093 983090983088983089983089

[983091983090] J K Sinha ldquoRecent trands in ault quanti1047297cation in rotatingmachinesrdquo Advances in Vibration EngIneerIng vol 983096 no 983089 pp983095983097ndash983096983093 983090983088983088983097

[983091983091] X Shen JJia andM Zhao ldquoNonlinear analysis o a rub-impactrotor-bearing system with initial permanent rotor bowrdquo Archiveo Applied Mechanics vol 983095983096 no 983091 pp 983090983090983093ndash983090983092983088 983090983088983088983096

[983091983092] A W Lees J K Sinha and M I Friswell ldquoModel-based identi-1047297cation o rotating machinesrdquo Mechanical Systems and Signal Processing vol 983090983091 no 983094 pp 983089983096983096983092ndash983089983096983097983091 983090983088983088983097

[983091983093] L C Jaw and W Merrill ldquoCBM+ research environmentmdashacilitating technology development experimentation andmaturationrdquo in Proceedings o the IEEE Aerospace Conerence(AC rsquo983088983096) March 983090983088983088983096

[983091983094] R Walker S Perinpanayagam and I K Jennions ldquoSimulatingUnbalance or Future IVHM Applicationsrdquo in Proceedings o theSociety or Experimental Mechanics (IMAC rsquo983089983090) vol 983090983097 pp 983089983092983089ndash983089983092983096 May 983090983088983089983090

[983091983095] H C Pusey ldquourbomachinery condition monitoring and ail-ure prognosisrdquo Sound and Vibration vol 983092983089 no 983091 pp 983089983088ndash983089983093983090983088983088983095

[983091983096] J R Jain and K Kundra ldquoModel based online diagnosiso unbalance and transverse atigue crack in rotor systemsrdquo Mechanics Research Communications vol 983091983089 no 983093 pp 983093983093983095ndash983093983094983096983090983088983088983092

[983091983097] K R Wheeler Kurtoglu and S D Poll ldquoA survey o healthmanagement user objectives related to diagnostic and prognos-tic metricsrdquo in Proceedings o the 983090983097th Computers and Inor-mation Engineering Conerence vol 983090 pp 983089983090983096983095ndash983089983090983097983096 983090983088983088983097

[983092983088] G Vachtsevanos ldquoPerormance metrics or ault prognosis o

complex systemsrdquo in IEEE Systems Readiness echnology Con- erence (AUOESCON rsquo983088983091) pp 983091983092983089ndash983091983092983093 September 983090983088983088983091

[983092983089] A Saxena J Celaya E Balaban et al ldquoMetrics or evaluatingperormance o prognostic techniquesrdquo in Proceedings o theInternational Conerence on Prognostics and Health Manage-ment (PHM rsquo983088983096) pp 983089ndash983089983095 October 983090983088983088983096

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 1313

Submit your manuscripts at

httpwwwhindawicom

Page 13: Rotordynamic Faults 2013

8202019 Rotordynamic Faults 2013

httpslidepdfcomreaderfullrotordynamic-faults-2013 1313

Submit your manuscripts at

httpwwwhindawicom