1

1 Introduction

During the past few years there has been an increasedinterest of the aircraft community on design loads foraircraft. Consequently there was a workshop in 1996SC73 on “Loads and Requirements for Military Aircraft”(AGARD Report 815). Elastic effects on design loadswere presented at a Workshop: “Static AeroelasticEffects on High Performance Aircraft.”

Also an Agadogragh was written on Gust Loads:AGARDograph 317: “Manual on the Flight of FlexibleAircraft in Turbulence.” All these topics are covered inthis manual.

With the increased use of active control systems onaircraft, there is currently a strong need to revisit someconcepts used for conventional aircraft and to identify thecorrection to be brought forward to existing procedures tocompute the several loads affecting a military aircraft andthe effect of the active control system. Special attentionhas been given to cover these items.

This report contains the following:

Maneuver Loads

Under this topic, design loads derivation covers thefollowing aspects:

• Aerodynamic/inertia loads• Aeroservoelastic effects• Effects of control system failure on design envelope• Dynamic loads

Gust loads

Although not a major concern for fighter aircraft, gustloads play an important role on aircraft that are designedunder civil requirements. A complete description of themethods used is presented along with recommendationson their use. The effect of control system failure isdescribed for the case of gust alleviation systems inAppendix A.

Aircraft/Landing Gear Loads

The specification of a landing gear as a system is shownin the Appendix B.

Limit Loads Concept

Limit load concepts and design loads criteria are exploredfor actively controlled aircraft.

CONCLUSIONS

In this manual several approaches are presented how tocalculate design loads for existing and future aircraft.There is a description of requirements included withsome historical background.

It very soon becomes clear that for fly by wire, agile,inherently unstable aircraft, these requirements as far asmanoeuvres are concerned are obsolete.

Therefore, an approach as described for the Eurofighter,where flight parameters are restricted and care freehandling of the aircraft is provided, is a possible solution.

Gust loads are also presented with some very interestingcomparisons of methods dealing with non-linear aircraft.

There is also an extensive compendium of dynamic loadswhich may be designing the aircraft structure.

A more global approach is also shown which tries toavoid insufficiencies of classical load regulations.

It is hoped that this manual can be helpful for aircraftdesigners to produce realistic flight loads which willresult in optimum weight structures.

2 Loads Requirements Review

The design of modern fighter aircraft is becoming anincreasingly complex process, and the establishment ofdesign criteria is an extremely important element in thatprocess. The Structures and Materials Panel of AGARDhave noted with concern that the existing designmaneuver load regulations in the NATO nations a ) arenot uniform in content and b) do not generally reflect theactual service experience of the aircraft.

Therefore an AGARD manual was prepared which triesto put together the latest requirement and methods whichhave been used for the design of recent modern airplanes.As an introduction to the present situations twocontributions to military requirements are given. The firstone gives a suggestion how maneuver loads criteria couldbe developed for modern agile aircraft.

In the second one the changes in the USAF StructuralLoad Requirements are presented which show theevolution of general load criteria valid for every aircraftto a specific document which is part of the overallspecification.

Similarly a specification for undercarriage is shown inthe Appendix B. The third set of specifications is for civilairplanes and is laid down in JAR25 (not included in thisreport).

2

2.1 The development of maneuver loadcriteria for agile aircraft

Max HacklingerMunich, FRG

AGARD Report 746, May 1987

2.1.1 IntroductionThe flight maneuver loads are major design criteria foragile aircraft (aerobatics, trainer, fighter aircraft), becauselarge portions of their airframe are sized by these loads.They also belong traditionally to the most elusiveengineering criteria and so far engineers never succeededin precisely predicting what pilots will eventually do withtheir machines. One extreme solution to this problemwould be to put so much strength into the structure thatthe aerodynamic and pilot tolerance capabilities can befully exploited by maneuvering without failure. This ismore or less the case with aerobatics aircraft, but modernfighters would grow far too heavy by this rule.To keep things lucid in this overview, I shall try togeneralize or simplify the Problems but retain theessential interrelations. Fig. 1 serves to illustrate this:

Figure 1

Box 1 contains the pilot's sensomotoric capabilities, thatis, his production of time, force and frequency dependentinputs into the aircraft controls.Box 2 resembles the complete flight control systemfunction from the sensors down to powered actuators. Ithas to satisfy not only aircraft stability but alsoman-machine stability criteria among others.Box 3 stands for the airframe with its aerodynamic andstructural capabilities to produce and withstand maneuverloads.Box 4 contains the physiological limitations of the pilot -his tolerance of high g, angular acceleration etc. Box 4acts as a single limiting function on box 3 and can betreated independently, but all other boxes are stronglycoupled with multiple feedback paths.

In the course of an aircraft development programme, box4 is given a priori, and apart from special training effects,box 1 is also given at the start in average form. Box 3 isfrozen relatively early by definition of the aircraftconfiguration and so is the architecture of box 2. But thenfor a long period of simulation and flight testing thefunctions of 2 are optimized, not only for the cleanaircraft but for a variety of external stores. To a lesserdegree corrections are also possible in this period for box3. This optimization process concerns both handlingqualities and maneuver loads, but the approaches aredifferent. The handling specialist has to analyze thewhole spectrum of possible flight maneuvers with mainemphasis an stability and achievement of performance.Design load investigations are a search for maximal andan experienced loads analyst can narrow down the vastspectrum of possible flight cases to relatively few whichbecome load critical. However, this process is becomingincreasingly difficult with modern active control systemsand the control system departments have to live with anew burden - the responsibility for causing exotic loads.

As a basis for a return to safe ground when the followingdiscussions of advanced maneuver systems leads us toofar astray, the next chapter gives a summary of thepresent status of maneuver load regulations for agileaircraft.

2.1.2 Status of present Criteria

The easiest way of obtaining maneuver loads is to assumeabrupt control surface movement to the stops, limitedonly by pilot or actuator force, and to derive the resultingairloads without aircraft motion analysis. This cheapmethod is still in use for certification of some civilaircraft but all the military regulations now requiresequences of pilot control inputs to initiate load criticalmaneuvers. The following regulations will besummarized here:

pilot steeringcapability

flight controlsystem capability

airframe capabilityaero & structure

pilottolerance

2

3 4

1

stability criteria (PIO etc.)

structural coupling, stability

limitingfunction

manoeuvre flown

sen

sor

dat

a

feed

bac

k vi

a se

nso

ry c

ues

3

• MIL-A-008861 A (USAF) 1971 for the US AirForce

• MIL-A-8861 B (AS) 1986 for the US Navy

• DEF-STAN 00-970 1983 for the UK

• AIR 2004 E 1979 for France.

The US situation at the moment is curious. (A) used to bethe main US specification for flight loads over manyyears. It has been replaced for the Air Force in 1985 byMIL-A-87221 (USAF), but this new specification is onlya frame without the essential quantitative material and assuch no great help for the designer. The US Navy on theother hand, who traditionally used to have their own anddifferent specification, have now adopted the old USAFSpec. (A) and updated and amplified it for application tomodern control system technology, including direct forcecontrol, thrust vectoring etc. Thus (B) seems to be themost up-to-date specification available now. Althoughmodern fighter tactics use combined control inputs inseveral axes, for a starting basis we prefer to treat themseparately as pitching, rolling and yawing maneuvers.

2.1.2.1 Pitching manoeurves

US Air Force

Fig. 2 shows the longitudinal control inputs for a checkedmaneuver required in (A) to rapidly achieve high loadfactors. Table 1 gives the corresponding boundaryconditions. Case (a) requires to pull maximum positive gby a triangular control input; if the maximum is notachievable by this, then the pilot shall pull to the stopsand hold for such time that max. g is attained. Case (b) issimilar to (a) but control displacement and holding time t3

shall be just sufficient to achieve max. g at the end of thechecking movement. Case (c) is similar to (b) but withcontrol movement not only back to zero but 1/2 of thepositive amplitude into the negative direction.

Fig. 2 Stick Inputs for pitching cases of 8861A

Limit load factor

Basic designmass

Allmasses

Max designmass

Air

craf

t cla

ss

Max Minat VH

Min atVL

Max Minat VH

t 1 [

sec]

A,F,T1)

8.0 -3.0 -1.0 4.0 -2.0 0.2

A,F,T2)

6.5 -3.0 -1.0 4.0 -2.0 0.2

O 6.0 -3.0 -1.0 3.0 -1.0 0.3

U 4.0 -2.0 0 2.5 -1.0 0.3

1) subsonic2) supersonic

Table 1: Symmetrical maneuver parameters of 8861 A

These theoretical maneuvers are certainly not exactlywhat pilots will do with modern fighters, but as long aswe can not use the vast amount of combat simulationresults as an all embracing envelope for flight loads, theyprovide at least a design basis – and they havehistorically produced reasonable maneuver loads,particularly tail loads.

US Navy:

(B) has adopted these 3 cases with slightly changedboundary conditions, see Table 2,

Limit load factor

Basic designmass

Allmasses

Max designmass

Air

craf

t cla

ss

Max Minat VH

Min atVL

Max Minat VH

t 1 [

sec]

F, A 7.5 -3.0 -1.0 5.5 -2.0 0.2

T 7.5 -3.0 -1.0 4.0 -2.0 0.2

O 6.5 -3.0 0 3.0 -1.0 0.3

U 4.0 -2.0 0 2.5 -1.0 0.3

Table 2: Symmetrical maneuver parameters of 8861 B

(d) maximum control authority in the negative directionshall be applied until maximum stabilizer or wing loadhas been attained. This can mean more than –δ/2 in case(c).

4

(e) is a special case for “computer control”, fly -by-wire,active control, stability augmentation, the direct liftcontrol, or other types of control system where the pilotcontrol inputs do not directly its establish control surfaceposition" which we shall call here generically ACTsystems. This case requires that aircraft strength shallalso be sufficient to cover modifications of cases (a) to(c) caused by ACT systems partially failed (transients,changed gains etc.), a requirement which is easier statedthan proven.

UK

In the UK, pitching maneuvers have traditionally beencovered by airplane response calculations after theCzaykowski method which assumed an exponentialfunction for elevator movement and no checking. Thiswas an expedient way to obtain tail loads but the new UKspecification (C) advises that pilot control inputs shouldbe used now. It does not specify any details of these.

France

The French specification (D) is very similar to case (a) of(A), with two differences: it has other load factors, seeTable 3, and it allows a slower stick return to neutral intime t2; for servo controls t1 = t2 shall be derived frommaximum control surface rate under zero load. It doesnot require checking into the negative region as (A) and(B). (see Fig. 3)

Limit load factorAircraftclass

Max min

T1

[sec]

T2

[sec]

III n1* -0.4 n1 0.2 0.3

II 4.0 -1.6 0.2 0.3

I 2.5 -1.0 0.3 0.3

Table3: Symmetrical maneuver parameters of AIR 2004E* n1 defined in the aircraft specification

Fig. 3 Control Inputs of AIR 2004 E

2.1.2.2 Rolling maneuvers (with pitching)

US Air Force

The rolling cases of (A) assume rapid control inputs andreversal (checked maneuvers), see Fig. 4. With 267 Nforce the stick shall be moved sideways in 0.1 sec, helduntil the specified bank angle is attained and thenreverted to neutral in 0.1 sec. If a roll rate greater than270°/s would result, control position may be lessened tojust achieve this value, but the roll rates shall never belower than those necessary to achieve the time to bankcriteria in the handling qualities specification (T360 = 2.8sec gives Pmax ≈150°/sec).

Fast 180° rolls are required starting from level flight with-1 to + 1g.Fast 360° rolls are required starting from n=1.Rolling pull out is required to start from steady levelturns with load factors from 1 to 8 n1 ( for a typical 8 gairplane this is 1 to 6.4 g).

By application of rapid lateral control (Fig. 4) the aircraftshall be rolled through twice the initial bank angle. In ourtypical example this would be a bank angle change of162°. Longitudinal control may be used to preventexceeding 0.8 n1 during maneuver.

Fig. 4 Stick Input of rolling cases of 8861 A

US Navy

The US Navy has in (B) adopted the rolling criteria of(A) but with significant additions: for ACT aircraft thePilot force is replaced by "maximum control authority".The reference to roll performance requirements isremoved - probably because this criterion used to be lessstringent than the 270 °/sec in most cases. Important isthe explicit reference to external store configurations; therolling cases of (A) have often been met in the cleanconfiguration only. But most important is the addition ofa new case for ACT aircraft. It states that the aircraftshall be designed for maximum abrupt pilot inputs in allthree axes. But it also states that these inputs shall in nocase lead to higher rates and load factors than theconventional cases.This paragraph is remarkable in several respects. Itdescribes a control system which would digest thewildest pilots Inputs into control outputs which aretailored to just achieve the old load maximum. It showsclearly the dilemma of the rule maker in the face of rapidtechnical development. This is the dream of the now

5

much advertised carefree (foolproof) handling system, Inreality control systems are primarily optimized for actualmaneuver performance and not for achievement of sometheoretical load cases. On the positive side this criterionrecognizes the need to retain some reference to provenmaneuver design load practice.Another addition in (B) is the requirement that thestructure shall also be designed to withstand thedemonstration requirements of MIL-D-87088 (AS),which apparently is not obvious.

UK

In the UK a wider envelope of initial conditions isrequired for the rolling cases, including a negative g rollreversal: -1.5 to 7.2 g. For the maximum roll rate severallimits are given: at least 1 1/3 of the roll performance

criteria in the handling specification which amounts toabout 200 °/sec; 200 °/sec for ground attack and 250°/sec for aerial combat maneuvers. The control input timehistory is roughly as in (A).

France

The French specification also requires negative initialconditions for the rolling cases: -1.6 to 6.4 g. (D) has control inputs similar to (A), butwith t1 = 0.2 and t3 = 0.3 or maximum servo capability.The roll limits are more severe, i.e., a full 360° roll andpmax ���������� ���������������������� ������� !���� �

that US pilots tend to avoid negative g maneuvers incontrast to their European colleagues:Table 4 summarizes the rolling parameters for a typical8 g airplane.

( A ) ( B ) ( C ) ( D )

MIL-A-8861 A MIL-A-8861-B DEF STAN 970 AIR 2004 E

180° roll –1 to +1 g360° roll at 1grolling pull outfrom 1 to 6.4 g,t1 = t2 = 0.1 sec,pmax = 270°/sec

Same as A plus ACS foolproof ness with maximumcontrol authority plusdemonstrationrequirements

Rolling pull out from –1.5 to 7.2 g,pmax = 1.33 p handling��������

Ground attack 200°/secAerial combat 250°/secNo t1, but maximum servocapability

360 ° roll, pmax = 360°/secrolling pull out from –1.6to 6.4 gt1 = 0.2 sect2 = 0.3 secor max servo capabilityunder zero load andt1 = t2

Table 4: Comparison of rolling parameters (8g airplane)

2.1.2.3 Yawing Maneuvers

Fig. 5 Rudder Inputs of 8861 A

US Air Force

Apart from the usual engine failures cases, (A) specifieslow and high speed rudder reversal.Fig. 5a shows the rudder for maneuvers from straight andlevel flight. At low speed 1334 N pedal force arerequired, at high speed 800 N.Fig. 5b shows the rudder input for the reversal case; frommaximum steady sideslip a fast recovery to zero yawshall be made.

US Navy

(B) has adopted these design cases and amplified themwith three new ones:

• for aircraft with direct side force control, strengthshall be provided for abrupt application of controlauthority up to a maximum side load factor of ny =3.

• for aircraft with lateral thrust vectoring capability,all maneuvers specified in the handling and stabilitycriteria shall also be covered in the loads analysis.

6

• it is general practice that evasive maneuvers such asjinking, missile break etc. shall be considered in theloads analysis.

UK

(C) requires a rudder kick with 667 N pedal force ormaximum output of the control system at all speeds. Italso requires the traditional British fishtail maneuver:starting from straight level flight, the rudder is movedsinusoidal for 1 1/2 periods of the Dutch Roll frequencywith an amplitude corresponding to 445 N pedal force or2/3 of the actuator maximum.

France

(D) has a rudder reversal case very similar to Fig. 5 b anda rudder kick without reversal, but both slightly slowerthan (A) due to t1 = 0,3 sec.Spinning is somewhat marginal for our theme of pilotcontrolled maneuvers but it deserves mentioning that itcan cause rather high loads. (B) has now increased theyawing velocity of agile aircraft with fuselage mountedengines from the 200 °/sec in (A) to 286 °/sec. This is asevere requirement for long fuselages.

The following figures show typical load maneuversresulting from application of the current US Mil-Specs. toan aircraft with moderate amount of ACT (Tornado).

Fig. 6 gives time histories of response quantities in arapid pitching maneuver with the control input specifiedin Fig. 2, case (a). displacement �max and holding time arejust sufficient to achieve nz max'

Fig. 6 Tornado rapid pitch, case(a) M=0.9, 1000ft, fullCSAS

Fig. 7 is a time history of response quantities resultingfrom the control input of case c in Fig. 2 which is criticalfor taileron bending moment BM.

Fig. 7 Tornado rapid pitch, case (c), M = 0.92, 22500 ft,full CSAS

Fig. 8 corresponds to the rolling pull out maneuver withinitial load factor 0,8 nl. This is another critical case fortaileron loads.

Fig.8 Tornado rolling pull out M=0.92, 19100ft, fullCSAS

2.1.3 The influence of piloting technique

Having set the scene of present structural maneuvercriteria, the next step is to review how realistic they are ina changed tactical environment with different pilotingtechniques. Mohrman has given a good account of thesechanges in [1], describing engagement rolls, turn reversalwith push down to increase roll rate, jinking maneuversetc. From the fact that these maneuvers are only weaklycorrelated with the specification maneuvers one might betempted to conclude that the old specifications should beabandoned altogether in favor of realistic simulation ofcombat maneuvers. Before deciding upon radical cuthowever, several arguments need to be considered.

7

Even for the old-fashioned aircraft without ACT thespecified control inputs were never fully representative ofactual pilot handling. They came closest for a controlsystem with a solid stick directly connected to tailsurfaces without sophisticated tabs, but they were onlyengineering simplifications of nature - like a ( 1 - cos )gust which does exist nowhere but is used to producereasonable loads.

Pilots are quite inventive in finding new techniques forcombat maneuvering - in fact this is part of the selectionprocess (survival of the fittest). For this reason and due tochanged tactical scenarios, most aircraft later in theirservice life are used differently from the way projected atthe design stage. If a sophisticated simulated combatmaneuver is used to derive critical design loads this casemay be overtaken by evolution after a few years inservice. ACT gives the possibility of late adjustments ofthe limiting functions, ideally by software changes only,but this is equally true for an aircraft designed to the oldcriteria.

Perhaps the major difference between the old criteria andthe new piloting techniques lies in the longer sequencesof combined maneuvers and not so much in the shortelementary inputs (stick to the stops, maximum pilotforce).

If so, it would be easier to adapt an aircraft designed tothe old criteria to changed operational practice than onewith sizing load cases derived from specific complexsimulated maneuvers.

An important difference to the old criteria exists in theabsolute level of maneuver loads. Improved g-suits,increased aircraft performance and improved controlsystems with load limitation - all these factors have ledpilots to pull limit loads more often and for longerduration. There is also indication for an increasedapplication of negative g in jinking maneuvers. Thisgeneral tendency goes so far that high performanceaircraft are now more frequently crashed due to pilotincapacitation (GLC).

The increased overall load level certainly necessitatesadjustment of the old fatigue strength criteria (e.g.MIL-8866); whether it also requires expansion of thedesign g-envelope, is debatable. Following the rationalewhich has been the basis of our airworthiness criteria formany years now, it would be sound engineering practiceto increase design strength if the overall load level hasstatistically increased. Other people argue however, thatthe load limiting capability of ACT does not only justifystaying with the old design loads, but even reducing thefactor of safety.

Whilst designers are confronted with a very real increasein the overall level of the symmetrical load cases, thesituation is more obscure with the unsymmetrical loads.Due to various scheduled interconnects between rudder,taileron, aileron or spoilers, the pilot now is rarely awareof the effect his commands have on the aircraft controlsurfaces. The only real limitation of unsymmetricalmaneuvers is probably the pilot's tolerance to lateralacceleration which is far less than in the verticaldirection. Turning to Fig. 1 again, this control function isexecuted via the feedback path between boxes 3 and 1.

At this point it is well to remember that the results of anyground based simulation are severely limited by theabsence of realistic motion cues to the pilot - neverthelessthese simulations have become an indispensabledevelopment tool.

2.1.4 The influence of advanced controlsystems

The cockpit environment has drastically changed inrecent years with the rapid development of flight controlsystems. For many decades pilots had to move largecontrols against inertia and air forces to keep theirmachines under control. Most of the aircraft in servicenow have still control movement but artificial feel toprovide some indication of the flight conditions. Nowsidestick controllers are being introduced which are verysensitive and require almost no motion. Although man isbasically a motion sensitive animal, pilots seem to haveadapted to this type of control. But from our viewpoint ofaircraft loads, we should keep in mind that many naturallimitations which used to prevent the pilot fromcommanding critical flight situations, do not exist withACT-aircraft. The conventional type of control isessentially a low pass filter. With sidestick controllersmany high frequency inputs, some of them unintentional,can make the FCS nervous.

Several loading cases in the existing criteria are based onmaximum pilot forces. The attempt in (B) to replace thisfor ACT-aircraft by "maximum pilot authority" is notconvincing. What is this pilot authority? The phrase"maximum deflection of motivators" in (C) does notresolve the problem either. This is just another casewhere we have lost an engineering yardstick which usedto work well in the past.

More important than changes at the input side arechanges in the main FCS functions. Traditionally, flightcontrol systems have been optimized for handlingqualities, with a few loads related functions like roll ratelimitation incorporated separately. So the problem was toprovide maximum maneuverability with sufficient flightstability to prevent loss of control. This task requires highauthority and strong control outputs. Now ACT systemshave a new basic function, load limitation, which requireslow authority and mild control outputs. Thus FCSoptimization has become a much more demanding task tounite two conflicting targets.

The FCS-certification effort has also increased drasticallywith automatic load limitation since the FCS is now adirect component of the proof of structural integrity.Where it was previously efficient to show thatconsecutive failures in the FCS led to degraded handlingbut still preserved a minimum get-you-home capability,the load limiting function of the FCS is directly safetycritical and must therefore satisfy severe criteria forfailure rates, redundancy etc.. To a degree this is reflectedin (B) by the requirement that the loading cases shall alsoinclude different failure states of the FCS. The associatedproblems are severe and can only be touched upon:Sensor redundancy, -disparity, software qualification,load distribution and a. o.

8

It is clear that proof of airworthiness of ACT aircraftwould be incomplete with consideration of thedeterministic loads cases only the ACT part needs to betreated statistically and this can be a cumbersome journeythrough the woods of failure trees. Quantitative guidancecan be taken from [2]

The overall failure rates given there are still applicable tonew designs.

Let us return now to the "carefree handling" conceptwhich appears to offer great possibilities for loads controland which Air Staffs are all too ready to specify becauseit would reduce pilots workload significantly and freethem for tactical tasks. In our context of maneuver loadssuch a control system ideally would limit all flight loadsto the design values so that neither pilot nor designerneed to worry about exceeding the structural capability ofthe airframe. This requires a large number of reliableinputs - air data, flight path coordinates, but alsocontinuous compete knowledge of the aircraft massstatus, including external stores partially released (speedlimits would probably still have to be observed by thepilot).

The central problem of such a system however, is the factthat good handling qualities and reliable load limitationhave conflicting tendencies in the FCS optimization. Soat best, a compromise can be achieved where due to theload limiting functions the handling envelopes arereduced, particularly in the upper left hand corner.

Load distribution is another complicating factor for anACT aircraft the same flight condition can often beachieved with a variety of aircraft configurations,depending an foreplane position, maneuver flapscheduling and perhaps vectored thrust. Assessment ofthose cases is even more difficult because airloaddistribution is already a great problem on modern agileaircraft due to non - linearities, elastic structure, fuselagelift, dynamic lift etc.

It appears unlikely that we shall see comprehensivecarefree handling control systems in operational usewhich would also effect complete load limitation. Morerealistic is the selection of a few single parameters suchas symmetric g, roll rate and perhaps sideslip which arecontrolled automatically. After all, who wants a formula1 racing car with a carefree handling control system?

One of the great benefits of ACT is its flexibility. Wherepreviously adjustment of the handling characteristicsduring development was very limited to changes ofsprings, bobweights and control surface tabs, it is nowpossible to tailor handling qualities over a wide rangeduring flight testing without large hardware changes.Also greater changes in operational usage can beaccommodated later on by ACT. This has consequencesfor the loads; they are subject to larger changes duringthe aircraft life. On the other hand development ofmodern aircraft takes so long that the basic configurationmust be frozen long before the final loads situation isknown with confidence.

In consequence, the certification process needs to bechanged too. It is futile from the start trying to findstructural maneuver load criteria which cover alleventualities. What we can do is to keep our feet anproven ground initially, that is to use the updated

conventional criteria for the basic design. Then, for along period of simulation and flight testing, adjustmentsare made whenever weak areas are discovered. Thisrequires an integrated approach by the FCS and loadsdepartments. The certification process must recognizethis by not aiming at the usual final operational clearance,but over many years providing preliminary clearanceswhich reflect the temporary state of knowledge abouttested maneuver loads and the related build standard ofthe FCS.

In summary, the maneuver loads part of aircraft designhas evolved from a relatively clean-cut, predeterminedanalysis to a long iterative process which graduallyutilizes flight test information to expand the flightenvelopes; a process which is also much more demandingbecause it involves the reliability of the FCS in provingstructural integrity.

2.1.5 ConclusionDesign maneuver load regulations in the NATO nationshave evolved from crude assumptions of single controlsurface movement to relatively complicated series ofPilot inputs in all three axes. These inputs need to bestandardized to permit the assessment of structural loadswith reasonable effort, but with the advent of activecontrol technology the hiatus between standardizedcontrol inputs for load assessment and actual pilotpractice with agile aircraft is rapidly increasing. Asolution of this dilemma may be to design flight controlsystems such that they provide "carefree handling", thatis a system which even for the wildest pilot inputs doesnot lead to structural damage. But this solution has alsodisadvantages:a) structural designers lose the wealth of experiencecontained in previous design practice and with it theirbasis for initial dimensioning of the airframe. This affectsa large portion of the aircraft mass and later re-designmay be impossible.b) Structural safety becomes crucially dependent an thefunctioning of black boxes and their connections. As longas we have no technically feasible direct load sensing andcontrolling system, a compromise is proposed: Use thebest combination of the old criteria for initial design butallow for a long development period flight control systemadjustments of load critical functions to fully exploit themaneuver capability of the aircraft without structuraldamage. This will require a flexible system of operationalclearances where the user can not have a completedefinition of the maneuver capabilities at the start of aprogram.

We have no consistent set of airworthiness criteria whichfully covers maneuver loads of agile aircraft.

Attempts to update the existing criteria to embrace thevast possibilities of ACT are only partially successful.

Proof of airworthiness of aircraft with ACT has becomemore demanding since the load influencing functions ofthe FCS are directly safety critical and must be analyzedfor failure to the same quantitative criteria as the structureitself.

The existing criteria can and should still be used forinitial design to define the airframe. Certification needs to

9

become adaptive to reflect a long period of testing andFCS changes .

2.1.6 References:

( A ) MIL-A-008861 A (USAF) 31.03.1971Airplane Strength and Rigidity, Flight Loads

( B ) MIL-A-8861 D (AS) 07.02.1986Airplane Strength and Rigidity, Flight Loads

( C ) DEF STAN 00-970 October 1985Design and Airworthiness Requirements forService Aircraft, Volume 1 Airplanes,Part 2 Structural Strength and Design for Flight

( D ) AIR 2004 E Resistance des Avion 08.03.1979

[ 1 ] Mohrman, R.:Selecting Design Cases for Future AircraftAGARD-Report 730, 1986

[ 2 ] Hacklinger , M.:Airworthiness Criteria for Operational ActiveControl Systems.Paper for DGLR panel Aeroelastics andStructural Dynamics 1979 (translation)

2.2 Changes in USAF Structural LoadsRequirements

Daniel Sheets and Robert GeramiLoads and Dynamic Branch

Aeronautical System DivisionASD/ENFL, Wright Patterson Air force Base OH, 45433-

6503, USAAGARD Report 746 , May 1987

The new General Specification for Aircraft Structures,MIL-A-87221 (USAF), does not establish the traditional,fixed requirements, but instead it presents the currenttailored approach to establishing structural loadsrequirements. In most cases the previous specificationsset arbitrary load levels and conditions to be used inaircraft design. These requirements were based uponhistorical experience, without consideration of futurepotential needs or capabilities brought about bytechnology advances. Instead, the new philosophyrequires that loading conditions be established rationallyfor each weapon system based on anticipated usage.Also, compliance with each condition must be verified byanalysis, model test, or full scale measurement.

2.2.1 Introduction

During the late 1970s, several conditions came togetherthat caused the US Air Force to develop new aircraftstructural specifications. While the USAF has always hada policy of reviewing, revising, and upgrading existingspecifications, there were factors favoring a new

approach. The contracting and legal authorities believedthat the existing system of many layers of specificationsneeded to be simplified. Also, rapidly advancingstructural technologies, coupled with new realms ofperformance and control capabilities, demanded that thestructural specifications address much wider range ofconditions while using an ever widening mix oftechnologies. The new military specification for aircraftstructures, MIL-A-87221 (USAF), is a major deviationfrom past requirement practices. It establishes weaponsystem uniquely tailored structural performance andverification requirements for airframes based on anin-depth consideration of operational needs andanticipated usage. In the past, specifications set arbitraryconditions, levels, and values to be used in the design ofbroad categories of aircraft.

Various sources have alleged that design requirementshave not kept pace with current usage practices;especially in the area of flight combat maneuvers. Theseallegations ignore the new requirement philosophy andare wrong for several reasons. The specification,MIL-A-87221 (USAF), does not preclude theconsideration of any type of loading situation. The newspecification actually requires the consideration of anyloading condition that can be identified for eitheranalysis, model testing, or full scale measurement.Therefore, if a loading condition is overlooked, the faultis not with MIL-A-87221 since it is not a set of rigid,pre-determined requirements.

Thus, this new approach does place a greater reliance onthe designer's insight and ability to correctly anticipatethe actual service loads. The term designer represents abroad spectrum of individuals associated with the USAF,System Contractor, and not just from the System ProjectOffice which manages system development for theUSAF. Anyone attempting to use the specification mustunderstand that this one document covers all types ofaircraft; from light observation, to the largest transport, tothe fastest fighters, to any of the most advanced flightvehicles. Therefore, any application of this newspecification must be tailored to the specific type ofaircraft under design. It should also be understood that notwo aircraft designs, even of the same general type, willhave identical anticipated usage. Therefore, not only mustthe detail design specification be tailored to a specifictype or category of aircraft, but it must also reflect thespecific anticipated usage of the aircraft being designedand performance capabilities brought about bytechnology improvements in aerodynamics, controlsystem integration, materials, and human factors.

2.2.2 Structural Loading Condition

The general organization of MIL-A-87221 is shown infigure 1. Structural loading requirements are developedthrough the application of section 3.4 of the appendix.The verification of these requirements is established bythe use of section 4.4, also of the appendix. Thisprocedure when incorporated into the new specificationgives the user the best features of both a checklistapproach and total design freedom. The loadingrequirement section 3.4, is divided into flight and groundconditions as shown in figure 2. The flight and ground

10

conditions are divided into subsections as shown infigures 2a and 2b respectively. Each of the manysubsections contain various specific load sources whichthe designer can either accept or modify as appropriate.During aircraft design, particular care must be exercisedin defining both the structural loading conditions and theassociate distributions used to design the airframe, whichin turn directly influences the performance and reliabilityof the aircraft. No single section of the specification canbe addressed independently. All requirements pertainingto all technologies must be considered as one unifiedentity. Both flight and ground operating conditions mustbe based on the anticipated usage, unique to a specificaircraft design effort. These conditions reflect theoperational usage from which design loads shall evolve.

Even though this new approach gives the designerconsiderable flexibility, the designer is not abandoned toestablishing all requirements without guidance orassistance. In both the requirement and verificationsections, numerous possibilities are presented forconsideration. The applicability or non-applicability ofBach suggested requirement or verification can beindicated by inserting either "APP" or "N/A" in a blankprovided with Bach one. For those that are consideredapplicable, either the requirement or verificationprocedure is then fully defined. Additionally, uniquerequirements can be added as a direct product of thetailoring process.

2.2.3 Flight Loading Conditions

The flight conditions (subsection of 3.4) consists ofthirteen categories, from the Standard symmetricalmaneuvers, to missile evasion, to the all inclusive"Other" category which is the one that both frees thedesigner from rigid requirements and simultaneouslyburdens him with the need to better define anticipatedusage. The maneuver load category suggests a minimumof five sub-categories for consideration. There is, ofcourse, the usual symmetric maneuver envelope, figure 3.However, due to current usage, various maneuvers suchas extreme yaw, jinking, or missile lock evasion aresuggested for design consideration. Any maneuver whichis possible for an anticipated aircraft and its usage, mustbe considered for design purposes.

Other changes can be found in the area of turbulenceanalysis. Historically, gust loading conditions have beenanalyzed by a discrete approach. However, the currentprocedure is to employ an exceedence distributioncalculation. In order to establish the exceedencedistribution, various parameters are needed. Fortunately,the new specification does suggest values for these terms;figure 4 is an example from the specification. Also,historically, maneuver and gust loading were consideredindependent and non-concurrent of each other except foraircraft engaged in low altitude missions. However,MIL-A-87221 actually suggests the designer rationallyconsider various conditions where gust and maneuverloads are combined because they concurrently affect theaircraft.

A very different type of load condition occurs duringin-flight refueling. While some services use the probe anddrogue system, a few others use the flying boomapproach; a few use both types of in-flight refueling

systems. This specification provides guidance in boththese areas to establish appropriate design conditions.

Since the very beginning of aircraft pressurization,specifications have addressed its loading effects.However, this new specification addresses pressurizationin a more inclusive manner then in the past. Usually,pressurization concerns have been focused an cockpits orcrew compartments. In contrast, the new specificationaddresses all portions of the aircraft structure subject to apressure differential. The requirements to considerpressurization even apply to such areas as fuel tanks,avionics bays, or photographic compartments. The broadapplication of this section of the specification requiresconstant and capable vigilance by the designer to includeall pertinent structure.

Since this specification does not presume to directlyaddress all possible loading phenomena, a specialcategory is reserved for any unique situations. Thiscategory is called "other" and is available so the designercan completely define all anticipated aircraft flightloading conditions. The important aspect of this categoryis that the designer is free to include any flight loadingcondition derived from operational requirements that canbe appropriately defined for analysis

2.2.4 Ground Loading Conditions

While aircraft ground operations are not as glamorous asflight performance, they can be the source of significantloading conditions. Unlike flight conditions, there havebeen very few changes to ground operating conditions inrecent years. In some cases the loading levels have beendecreased due to improved civil engineering capabilities;improved runways, taxiways, ramps, etc. Ground loadingconditions include all ground operations (taxi, landing,braking, etc.) and maintenance operations (towing,jacking, hoisting, etc.).

2.2.4.1 Ground Operations

Since the earliest days of aircraft, ground operations havechanged very little. Most of these changes have been inthe area of load magnitude, not in the type or source ofload. Before takeoff, an aircraft normally needs to taxi,turn, pivot, and brake. Various combinations of theseoperations must be considered in order to fully analyzerealistic ground operations. The resultant loads are highlydependent on the operating conditions, which are in turndependent on the aircraft type and anticipated mission.

2.2.4.2 Takeoff and Landing.

Usually takeoffs and landings are performed on hardsmooth surfaces which are of more than adequate length.However, in some situations the surface is not ofadequate length, hardness, or smoothness. Therefore,takeoff specifications must either anticipate all possiblesituations or allow the designer to establish specifictakeoff and landing requirements for each system. Forexample, consideration is given to rough semi-prepared

11

and unprepared surfaces. Even rocket and catapultassisted launch is included in the specification. However,the designer is free to consider devices such as ski-jumps,if they are appropriate to the aircraft and missionsinvolved. Since takeoffs are addressed; so too arelandings. Various surfaces, arrestment devices anddeceleration procedures are included for consideration aspossible load producing conditions. The designer andeventual user must work together to correctly establishlanding requirements, since they can vary greatlydepending on the final usage of the aircraft.

2.2.4.3 Towing

Since the beginning of aviation, it has been necessary totow aircraft. While the designer is free to define his owntowing conditions and associated loads, he must also toverify the legitimacy of these conditions. In this categorythe new specification comes close to the previous AirForce criteria specifications by providing the valuesgiven in figures 5 and 6. One should remember that thesetowing conditions are very much result of years ofempirical experience. Justifying and verifying newtowing load conditions could be a very difficult task.

2.2.4.4 Crashes

Unfortunately not all flights are successful; some end incrashes. Different types of aircraft require various typesof design considerations for crash loads, depending antheir inherent dangers due to mission and generalconfiguration. For example, fighters pose crash problemswith respect to seats, fuel tanks, or cockpit equipment,but definitely not litters or bunks. However, the design ofa transport would most assuredly involve crash loadconsiderations for cargo, litters, bunks, or even temporaryfuel tanks in the cargo compartment. The newspecification suggests various combinations of on-boardequipment. These suggested values, figure 7, are verysimilar to the historic ones which in the past were firmrequirements. Today a designer can use factors other thanthe suggested ones, as long as the alternate load factorscan be substantiated.

2.2.4.5 Maintenance

Even daily maintenance actions can impose variousloading conditions on aircraft. Many maintenanceoperations require towing, jacking, or hoisting whichsubject the aircraft to abnormal and unusual loadingcombinations that must be considered during aircraftdesign. General data is supplied for these conditions, seefigure 8. However, following the tailoring inMIL-A-87221 (USAF)., the designer is free to define anylevel of maintenance induced loading which can besubstantiated.

2.2.4.6 CONCLUSIONS

The new specification, MIL-A-87221, will allow designrequirements to be more closely tailored to theanticipated use of the aircraft. In this way the finalproduct will be more efficient, with less wasted,unneeded, and unused capabilities. This will lead in turnto reduce costs of ownership for Air Force weaponsystems. This specification has been applied to thedefinition of requirements for the Advanced TacticalFighter. This process is now taking place.

12

13

14

15

16

17

18

3 Maneuver Loads

Design maneuver load regulations in the NATO nationshave evolved from crude assumptions of single controlsurface movement to relatively complicated series ofpilot inputs in all three axes. These inputs need to bestandardized to permit the assessment of structural loadswith reasonable effort, but with the advent of activecontrol technology the hiatus between standardizedcontrol inputs for load assessment and actual pilotpractice with agile aircraft is rapidly increasing.

The flight maneuver loads are major design criteria foragile aircraft (aerobatics, trainer, fighter aircraft), becauselarge portions of their airframe are sized by these loads.They also belong traditionally to the most elusiveengineering criteria and so far engineers have neversucceeded in precisely predicting what pilots willeventually do with their machines. One extreme solutionto this problem would be to put so much strength into thestructure that the aerodynamic and pilot tolerancecapabilities can be fully exploited by maneuveringwithout failure. This is more or less the case withaerobatics aircraft. But modern fighters would grow fartoo heavy by this rule.

So the history of maneuver load criteria reflects acontinuous struggle to find a reasonable compromisebetween criteria which do not unduly penalize totalaircraft performance by overweight and a tolerablenumber of accidents caused by structural failure.

Several approaches are presented in the next sectionswhich have been used for the design of the most recentfighter airplanes.

3.1 Classical Approach

3.1.1 Definitions

Loads External Loads on the structure

Limit Load• Military Specification (MIL-Spec.):Maximum loads which can result from authorized flightand ground use of the aircraft including certainmaintenance and system failuresRequirement: The cumulative effects of elastic, permanentor thermal deformations resulting from limit loads shallnot inhibit or degrade the mechanical operations of theairplane.

• Civil Requirements (FAR,JAR):Maximum loads to be expected in service.Requirement: Without detrimental permanent deformationof the structure. The deformation may not interfere withsafe operation.

Ultimate Load• Military Specification:Limit Load multiplied by a factor of safety.Requirement: No structural failure shall occur

• Civil Requirements:Limit Load multiplied by a factor of safety.

Requirement: No failure of the structure for at least 3seconds.

Factor of Safety• Military Specification:The Factor of Safety shall be 1.5.

• Civil Requirements:A Factor of Safety of 1.5 must be applied to the prescribedLimit Load, which are considered external loads on thestructure.

General Definition:Safety Factors are used in aircraft structural design toprevent failures when the structure is subjected to variousindeterminate uncertainties which could not be properlyaccessed by the technological means, such as:

• the possible occurrences, during flight or groundoperations, of load levels higher than the limit load

• uncertainties in the theoretical or experimentaldeterminations of stresses

• scatter in the properties of structural materials, andinaccuracies in workmanship and production

• deterioration of materials during the operational lifeof the aircraft.

Static LoadsAirframe static loads are considered to be those loads thatchange only with flight condition: i. e. airspeed, altitude,(angle of incidence, sideslip, rotation rates, ..) etc. with aloads / loads-parameter oscillating below 2 Hz. Theseloads can be considered to be in a steady non oscillatingstate (rigid body motion).

Dynamic LoadsDynamic loads are considered to be those loads whicharise from various oscillating elastic or aeroelasticexcitation which frequencies above 2 Hz. The loads are tobe determined by dynamic loads approaches, dependingon the sources of excitation and would include:• Atmospheric turbulence / Gusts• Buffet / Buffeting / Buzz• Stores Release and Jettison• Missile Firing• Hammershock• Ground Operations• Birdstrike• etc.

Maximum Load = Maximum external Load (general used as classical definition)

• resulting from authorized flight use (Mil.Specification)

• expected in service (FAR/ JAR – Requirement)

• derived by the Maximum Load Concept Approach

• limited by the Flight Control System, applyingFlight Parameter Envelope Approach

• derived from operational flight monitoring applyingOperational Flight Parameter Approach

19

• derived from load spectra (cumulative occurrencesof loads) applying Extreme Value Distribution

Maximum Load = the structure is capable to support(used in More Global Approach)

• Maximum load case which produces the maximumvalue of at least 1 failure strength criterion,integrating Load Severity Indicators.

3.1.2 Limit Load Concept

Strength requirements are specified in terms of

•••• Limit Loads• Military Specifications:

MIL-A-8860 (ASG),MIL-A-008860 A (USAF),AFGS-87221 A

is the maximum load normally authorized foroperations.

• Federal Aviation Regulations:Part 23,Part 25

is the maximum load to be expected in service.

•••• Ultimate Loadsis limit loads multiplied by prescribed factors of safety.

The basic premise of the Limit Load Concept is to definethat load, or set of loads, which the structure should becapable of withstanding without permanent deformation,interference or malfunctions of devices, degradation ofperformance, or other detrimental effects.

At any load up to limit loads, the deformation may notinterfere with safe operation. The structure must be ableto support ultimate loads without failure for at least 3seconds. The limit loads, to be used in the design of theairframe subject to a deterministic design criteria, shall bethe most critical combination of loads which can resultfrom authorized ground and flight use of the aircraft.

3.1.2.1 Conventional Aircraft

A limit load or limit load factor which establishes astrength level for design of the airplane and componentsis the maximum load factor normally authorized foroperations.

The determination of the limit loads is largely specifiedin the regulations (MIL, FAR, Def., etc) and isindependently of the missions / maneuvers actuallyperformed in operation. Worst case conditions are usuallyselected as a conservative approach.

Safety factors were introduced into the design of thestructure to take care of uncertainties which could not be

properly assessed by the technological means of thattime, such as:

• the possible occurrence of load levels higherthan the limit load

• uncertainties in the theoretical or experimentaldetermination of stresses

• scatter in the properties of structural materials,and inaccuracies in workmanship andproduction

• deterioration of the strength of materials duringthe operational life of the aircraft

3.1.2.2 Actively Controlled Aircraft

For actively controlled aircraft the limit loads are to bedetermined taking into account the flight control system(fly by wire, load alleviation) for:

• normal operating conditions, without systemfailures

• conditions due to possible system failures

The resulting loads have to be considered for designrespectively proof of the structure.

For civil aircraft required by recent regulations (FAR,JAR):

• for normal operating systemsas limit loads, ultimate loads applying theprescribed safety factor (1.5)

• for failure conditionsthe safety factor is determined by the failureprobability distinctive:

• active failure ( at time of failure )

• passive failure ( after failure for continuation offlight )

The purpose for the integration of an active controlsystem is to enhance maneuver performance while noteroding structural reliability, safety, and service life.The application is described in Ref. (1)

Reasons for applying other ApproachesFor conventionally controlled aircraft the regulationsgives unequivocal deterministic criteria for thedetermination of the most critical combination of loads.

e.g. for flight maneuvers, the regulations (Mil-A-8861)prescribe the time history of the control surfacedeflections and numerically define several essentialmaneuver – load parameter for the determination ofdesign load level.

20

Obviously with the introduction of active controltechnology, as well as care free maneuvering features,recent specifications no longer define the control surfacedeflections but rather provide the cockpit displacementsof the controls in the cockpit (Mil-A-8861).This means that existing design load regulations andspecifications based on conventional aircraftconfigurations, structural design concepts and controlsystems technologies, may not be adequate to giveunequivocal criteria for the determinations of designloads and ensure the structural integrity of future aircraftusing novel control methods.To cope with using the limit load concept for activelycontrolled aircraft several approaches have beenapplied:

• Maximum load conceptBackground and suggested models are described in3.2.1.

An example of application:

• The flight control system for a naturally unstableaircraft is designed with the feature to feed inmaneuver parameter boundaries ( load factors, rates,accelerations ) in such a way that limit design loadsare not exceeded.

This approach could lead to a reduction of the safetyfactor for flight maneuver loads keeping the structuralsafety at least as for conventional aircraft e.g. from 1.5 to1.4 for EFA.The application is described in Ref. (2).

Flight Parameter Envelope ApproachThe loads process is described in 3.2.5

Probabilistic determination of limit load

Operational Flight Parameter ApproachThe procedure is described in 3.2.2

3.1.2.3 References

[ 1 ] H.-M. Besch, H.-G. Giesseler, J. SchullerAGARD Report 815,Impact of Electronic Flight Control System (EFCS)Failure Cases on Structural Design Loads

[ 2 ] Sensburg O., Bartsch O., Bergmann H.Journal of Aircraft, Vol.24, No.11, Nov. 1987Reduction of the Ultimate Factor by applying aMaximum Load Concept.

3.1.3 Safety Factors Review

3.1.3.1 History

The present - day safety factor for aircraft structures, asapplied to manned aircraft, dates back 70 years. During

the last 30 years considerable progress has been made inthe fields of structural materials, semi finished productsand testing methods. Furthermore advances inaerodynamic and aeroelasticity, combined withdevelopments in electronic data processing, facilitate amore precise prediction of structural loads and structuralanalysis.

A reappraisal of the safety factor would therefore seem tobe in order, not with the intention of lowering the level ofsafety, but with the aim for examining the various safetyrequirements in the light of present knowledge. This,together with the fact that there exists a lack of a rationalbasis for the factors of safety concept presently appliedto the design of air vehicles, brought up a discussion ofchanging the structural safety concept and the factorsinvolved within AGARD-SMP in 1977. The Structuraland Materials Panel formed an ad hoc Group to conductthis discussion. Three pilot papers contained in Ref.(1)addressed the different aspects to be envisaged, and showup inconsistencies of the present concept as well asmeans and methods for permissible changes.The result of the discussion following the presentationsbefore the Sub - Committee was, that it would not beappropriate at the present time to change the concept, butit was found worthwhile to have a collection andevaluation of all those factors concerning structural safetyincluding the philosophies which back up the applicationof these factors.The Sub - Committee found it most suitable to collect allpertinent data and back up information by means of aquestionnaire, which was drafted by two coordinators(one for North America, one for Europe) and reviewed bythe members of the Sub - Committee.This questionnaire was distributed to the addressedAirworthiness Authorities of the NATO - Nations with arequest for cooperation. The replies of the questionnairewere summarized and evaluated by the coordinators andpresented before the Sub - Committee. The answersgiven, including the results of personal discussionsbetween coordinators and nominated representatives ofthe authorities, are condensed published in Ref.(2).From the evaluation it may be concluded that there existsa considerable amount of agreement with respect to theFactors of Safety and their application. On the otherhand, some disagreements and interpretations haveresulted. Thus, this report forms a basis for discussing thedisagreements in order to achieve a higher degree ofconformity between the authorities of NATO - Countrieswith a regard to structural safety and reliability.

At that time the present concept and the Factors of Safetywere in general regarded as satisfactory with the intentionto review the Safety Concept till such time as moreknowledge and experience in application of newtechnologies are available;e.g.• Improvement of knowledge about flight and ground

loads occurring in service (operational loads) toknow the margin between the design conditions andthe operational conditions.

21

• Introduction of new technologies, which are notincluded in the scope of the existing designrequirements

• active control

• behavior of new materials ( composites )

3.1.3.2 REFERENCES

[1] AGARD - Report No. 661Factors of Safety , Historical Development , State of theArt and Future Outlook.

[2] AGARD - Report No. 667Factors of Safety , Related to Structural Integrity .A Review of Data from Military AirworthinessAuthorities.

3.1.3.3 Possible Methods for Splitting of SafetyFactors

In the mean time significant progress and experiences inload determination for conventional aircraft and foractively controlled aircraft have been made as well asdeterminations of load conditions have been applied forcases which are not covered by the several existingairworthiness regulations; e.g. as special conditions.Therefore it is time to take up the review of the SafetyFactor Concept. Factors of safety can be rationalized bysplitting into Loads (FSl) and structural / materialuncertainties (FSs).The present - day safety factor covers the uncertainties asa global factor mainly applied for

• possible exeedances of loads in relation to thedesign loads

• uncertainties in structural analysis

without realizing the particular uncertainties of loads andstructural analysis separately i.e. the global factor isapplied as the same value for both. This application of thesame factor of safety for loads determination and forstructural analysis can lead to an apparent margin ofsafety which is higher or lower than the global factor isintended to cover.By splitting the factor into two parts, as suggested by theStudy Group Structures of AECMA (see chapter 3.2.1.1)for loads and for structural analysis, a clear relation of thesafety margin is determined.

• FSl for loads uncertainties

• FSs for structure uncertainties

The product of both factors is known, keeping theapproved total factor of 1.5 .

FS = FSl x FSs = 1.25 x 1.20 = 1.50

Another suggestion from US ( D. Gibson) is to divide theFactor of Safety into three terms

o U1 uncertainty related to loads computationo U2 ” ” to operational environmento U3 ” ” to structural analysis

In this proposal U1 and U3 are the same as FSl and FSs.U2 for predicting the actual operational environmentmight be applied using deterministic criteria. Theproposed values for all terms are 1.15.

e.g. U1 x U2 x U3 = 1.15 x 1.15 x 1.15 = 1.52

For aircraft which apparently will not be able to exceeddesign loads during operations e.g.

• applying operational maneuver models for derivingor updating of design loads (see chapter 3.2.4)

• applying flight parameters envelope approach forlimiting specified response parameters (see chapter3.2.5 )

The value of U2 might be 1.0 resulting in a final Factorof Safety

FS = 1.15 x 1.15 = 1.32

3.2 Non Classical Approach

3.2.1 Maximum Load Concept

3.2.1.1 Background

The Airworthiness Committee of the international CivilAviation Organization (ICAO) discussed, among otherthings, the subject of maximum load concept in theperiod from1957 to 1970. It was decided in Montreal inlate 1970 not to pursue this concept for the time being asa possible basis for airworthiness regulations. Severalproposals however, were made to improve structuralsafety. This subject was also discussed by the StudyGroup Structures of the AECMA (AssociationEuropienne des Constructeurs de Material Aerospatial) inthe context of the Joint Airworthiness Requirement(JAR). These deliberations led to the suggestion to splitthe proven safety factor of 1.5 into two parts, in a rationalfashion, one for uncertainties in the loading(determination of loads), the other for uncertainties instrength analysis including scatter of material propertiesand inaccuracies in construction.

22

Allowable loads are defined as those load values that willonly be exceeded by expected loads with a prescribedsmall probability. These loads are then referred to asmaximum loads.Gust or landing loads are strongly influenced by randomphysical or human characteristics. But also in these casessafety could be much better defined by extrapolation ofloads from statistical data, rather than the application of asafety factor of 1.5 for all cases. Furthermore, loads thatare limited naturally by the ability of the aircraft toproduce them, or by internal aircraft systems, (loadalleviation, flight control systems) could be regarded asmaximum loads to which a safety factor need not beapplied. The determination of maximum loads with asmall probability of being exceeded is entirely possiblefor modern fighters which are limited in their maneuvers,or for control configured vehicles (CCV) which are inany case equipped with an active flight control system(fly–by–wire). As a principle the prescribed designboundaries and the corresponding safety factor should notbe applied separately, i.e. the entire design philosophyshould be considered. Therefore a mixed application ofvarious regulations to a single project is not advisable.Up to now the safety factor has been reduced in only afew cases. Within the pertinent regulations only the caseof the American MIL-A-8860 (ASG) issue is known,where no safety margin is required for the undercarriageand its supporting structure.It may be supposed that with the consent of theappropriate authorities the safety factor or the load levelcould be reduced in the following cases:

• in emergencies, such as emergency landings into anarresting net or cable

• for transient phenomena (hammer shock pressure inaircraft inlets)

• where actuators are power-limited and large loadscannot be produced

3.2.1.2 Suggested Models

The following models are proposed for the application ofthe Maximum Load Concept.

Semi-statistical / semi deterministicIn the past operational loads were predominantly checkedby measurement of the main load parameters, in the formof cumulative frequencies or load - parameter - spectra(Ref. 1).They are:

• the normal load factor, in flight and on the ground

• the angle of sideslip and/or the transverse load factor

• the rolling velocity in flight

• the bank angle during landing

On the basis of these load - parameter - spectra aprobability of occurrence of the main load parameters isdefined for each type of mission and maneuver, and themaximum value of the main load parameter can bedetermined from this.

If, for instance, an aircraft is designed for air-to-aircombat, a maximum load factor of 9.0 may be derivedfrom the statistical cumulative frequency distribution forevery tenth aircraft after 4000 flight hours. This value istaken to be maximum main load parameter. For this loadparameter the loads produced by the maneuvers specifiedin the pertinent regulations are determined by means of adeterministic calculation such that the maximum value ofthe main load parameter is just attained, but notexceeded. An example is the loads as a function of timeproduced by the actuation of cockpit controls accordingto MIL-A-008861.A recent approach for active controlled aircraft has beenapplied to the European Fighter (EFA) for thedetermination of the design loads, called Flight ParameterEnvelope Approach. ( Description see 3.2.5 )

Semi-statistical / semi empiricalIt has been known for years that VG and VGHmeasurements do not suffice for the definition of criteriafor structural design.In order to obtain statistically supported design criteria, aspecial NACA Sub-Committee on Aircraft Loadsrecommended (1954) to expand statistical load programsto the extend that they included measurements of timehistories of eight parameters, three linear accelerations (x,y, z,), three angular accelerations (p, q, r,), airspeed (V)and altitude (H).The first measurement of this kind where made with theF 105 D Fighter with the aim to develop a maneuver loadconcept which was to predict design loads (Ref. 2). Alldata were processed to calculate time histories of loads,with peaks called “observed loads”. The dataoscillogramms were examined in order to define 23recognizable types of maneuver. Assuming that for everytype of maneuver the same sequence of aircraft motionoccurs with the exception of differences in amplitude andduration, the measured parameters were normalized withrespect to amplitude and time.Finally, to determine the loads, the normalizedparameters were denormalized in order to get the loadpeak distribution for the wing, the fuselage, and theempennage. The good agreement between the observedand predicted load peak distribution demonstrated thefeasibility of the maneuver model technique for the F-105D aircraft. The F-106 Fighter was selected to demonstratethis model, thereby determining the model’s usefulnesson another aircraft. The detailed results of 3770 flight testhours made it possible to apply the maneuver modeltechnique i.e. the empirical calculation of componentloads as compared to F-106 design loads (Ref. 3).The results in the form of cumulative occurrence of theloads for wing, elevon, and vertical tail made it possibleto determine the design load for a given cumulativeoccurrence.

23

A recent approach has been elaborated in the WorkingGroup 27 of AGARD-SMP called Operational ManeuverModel. The demonstration of the feasibility is reported inAGARD Advisory Report 340 Evaluation of Loads fromoperational Flight Maneuvers (Ref. 4).(Description see 3.2.2 Operational Flight ParameterApproach)

Statistical: Extreme Value DistributionAs a rule, load spectra are produced with the objective ofdetermining magnitude and frequency of operationalloads. These, in turn, are used in fatigue tests todetermine the corresponding fatigue life of structure.Loads spectra like these are derived from relatively shorttime records, compared to the actual operational life time;they do not contain those maximum values that might beexpected to occur during the entire operational life of thestructure, i.e. a knowledge of which is necessary for thedesign.

Determination of Extreme Value DistributionIn cases where the range, the maximum value, and scatterof the spectrum may be safely assumed, an extreme -value distribution can be established, describing extremevalues of loads / load parameters by its magnitude andrelated probability of exceedences (suggested by Prof. O.Buxbaum, ( Ref. 5 )). By means of extreme loaddistributions the derivation of extreme loads is feasiblefor determinate probabilities of exceedences, and therebythe design load can be determined.

Examples of applications

• Maximum rolling moments on horizontal tailderived from in - flight measurement with C160Military Transport Aircraft, AGARD Report No.661, page 9

Fig. 1 shows the extreme – value distribution

• Maximum loads on vertical tail derived from in -flight measurements with F-106 Fighter AircraftAIAA - Paper No. 70-948, page 8

Fig. 2 shows the cumulative occurrences

3.2.1.3 References

[1] J. Taylor, Manual of Aircraft Loads,AGARDograph 83 (1965)

[2] Larry E. Clay and Heber L. Short,Statistical predicting Maneuver Loads from eight-channelFlight DataReport No. TL 166-68-1 (1/1968) NASA CR-100152

[3] James D. Jost and Guin S. Johnson,Structural Design Loads for Strength Fatigue computedwith a multi-variable Load Environment ModelAIAA - Paper No. 70 - 948

[4] AGARD ADVISORY REPORT 340Evaluation of Loads from Operational Flight ManeuversFinal Working Group Report of Structures and MaterialsPanel Working Group 27

[5] O. Buxbaum,Verfahren zur Ermittlung von Bemessungslastenschwingbruchgefährdeter Bauteile aus Extremwerten vonHäufigkeitsverteilungenLBF - Bericht Nr. FB - 75 (1967)

24

FIG. 1 EXTREME – VALUE DISTRIBUTION

25

FIG. 2 CUMULATIVE OCCURRANCES OF VERTCAL STABILIZER LOADS

26

3.2.2 Operational Flight ParameterApproach

3.2.2.1 Introduction

The determination of the design maneuver loads islargely specified in regulations independently of themaneuvers or missions actually performed in operation.For conventionally controlled aircraft the regulations givethe time history of the control surface deflections andnumerically define several essential maneuver – loadparameters for the determination of the design load level.Obviously with the introduction of the fly-by-wire and/oractive control technology, as well as care freemaneuvering features, recent specifications no longerdefine the control surface deflections but rather providethe cockpit displacements of the controls in the cockpit.This means that existing design load regulations andspecifications based on conventional aircraftconfigurations, structural design concepts and controlsystem technologies, may not be adequate to ensure thestructural integrity of future military aircraftconfigurations using novel control methods, structuralconcepts and combat tactics.

In service, maneuvers, especially combat maneuvers, areflown in accordance with practiced rules that lead tospecified motions of the aircraft in the sky. An evaluationof operational flight maneuvers has been made forseveral aircraft types flown by the USAF, CF and GAFwith the aim of deriving operational loads by applyingparameters measured in operational flights.

This approach is based on the assumption that maneuverstrained and flown by the NATO Air Forces can bestandardized.The standardized maneuver time history is thereplacement as a quasi unit maneuver, for all operationalmaneuvers of the same type.The Standardized Maneuver is obtained by normalizationof parameter amplitudes and maneuver time to make theparameters independent of mass configurations, intensityof the maneuver, flight condition, flight control system,and of the aircraft type.The goal is to find a standardized time history for eachtype of maneuver, which is independent of the extremevalues of the relevant parameters and aircraft type.

One promising approach is to derive design loads from acareful analysis of operational maneuvers by currentfighters to extract critical parameters and their range ofvalues. To investigate this approach, Working Group 27“Evaluation of Loads from Operational Flight Maneuver”was formed, AGARD involvement was particularlyrelevant since it allowed the expansion of the types ofaircraft and the control systems considered in the study.The Working Group formulated a set of activities thataddressed the fundamental premises of a method togenerate operational loads from flight parameters bydetermination of Standard Maneuvers independent of theaircraft type and the control system.

The flow chart in Figure 1 presents the general data flowand indicates the major phases of the procedure.These operational loads can be statistically evaluated foruse in static design and for fracture assessment.

In the first part of the procedure the verification of theOperational Maneuver Parameter Time Histories isdescribed in boxes with black frames, Fig 3.2.3.The steps of the verification are:

• Recording and Evaluation of Operational Parameters• Identification of the Maneuver Types• Normalization of the Parameters• Determination of the Standard Maneuver Types

In the second part the Derivation of Operational FlightLoads is described in boxes with red frames in 3.2.4applying the Maneuver Model in the steps:

• Selection of the Standard Maneuver Type to beconsidered

• Definition of the Boundary Condition as designcriteria

• Calculation of the Control Deflections necessary toperform the Operational Maneuver

• Response Calculation and Verification of theparameter time history

• Determination of Structural Loads

The evaluation of this procedure done by the WorkingGroup (WG 27) has demonstrated the feasibility ofdetermining loads from operational flight maneuvers(Ref. 1)

This Operational Flight Maneuver Approach can be usedfor:

• The judgment of the operational load level foraircraft already designed with regard to the designlevel (static and fatigue) as specified in theregulations.

• That means the margin between design loads and theextreme operational loads is known.

• The determination of the load level for static andfatigue design due to operation for new aircraft to bedeveloped.

3.2.2.2 References

(1) AGARD ADVISORY REPORT 340Structures and Materials Panel, Working Group 27on Evaluation of Loads from Operational FlightManeuvers.

(2) AGARD REPORT 815Loads and Requirements for Military Aircraft, Page3 –1, and Page 4 – 1

27

Fig. 1: Procedure Overview

3.2.3 Determination and Verification ofOperational Maneuver Parameters andTime Histories

3.2.3.1 Verification Performed

Based on the hypothesis that all operational maneuversperformed in service can be verified as standardmaneuvers ( normalized parameter time histories for eachindependent maneuver type ) the determination ofoperational loads is feasible applying the OperationalFlight Parameter Approach. The verification of thisapproach to generate operational loads from flightparameters by determination of a set Standard Maneuversconsisting of normalized operational parameter timehistories is described.The Standard Maneuver procedure is shown in figure 2 asa flow chart.For each type of Standard Maneuvers the normalizedmotion parameters are to be validated independent ofaircraft type, mass configuration and flight controlsystem.

For the evaluation of operational parameters, thefollowing data were made available and have beenjudged as applicable.

• Flight test data by GAF Test Center for specificoperational maneuvers on three aircraft ( Alpha Jet,F – 4 F, Tornado)

• Data from simulations by GAF for specificoperational maneuvers recorded on Dual FlightSimulator for two aircraft ( F – 4, JF – 90 )

• Service data by USAF recorded on the F-16(selected subset from over 300 sorties from 97aircraft )

• Service data by CF recorded on the CF-18 fleetmonitoring) (selected subset of CF-18 fleetmonitoring )

Taking all data available, which have been found to besuitable for separation into maneuver types, the data baseis about 13 maneuver types.For two maneuver types, High - g – turn and Barrel roll,more than 60 maneuvers for each maneuver type havebeen considered as applicable for evaluation.

Recorded Operational Parameters

C B

Operational ParametersTime Histories

Standard Maneuver Type A

C B

NormalizationProcess

Maneuver Type A

C B

Boundary Conditions

Maneuver Type A

Aircraft Basic Data

M A N E U V ER M O D E L

Structural LoadsStatic Design and / or Fatigue

Maneuver Identification

Flight-Test-Data Service-DataSimulation-Data

28

The actively controlled aircraft ( Tornado, F-16, CF-18 )fit in the same scatter band as the conventional controlledaircraft. This means the hypothesis that the operationalmaneuvers are performed in the same way, i.e.performing the same normalized parameter time history,can be considered as confirmed.

The result is, that the Operational Standard Maneuverindependent of the aircraft type is applicable as unit inputfor calculation of the movement of a specific aircraft byreconstitution of the real aircraft configuration and flightcondition.

3.2.3.2 OPERATIONAL PARAMETERS

The number of parameters defining the aircraft motionshould be chosen in such a way that recording andevaluation cause minimal expense. This can be achievedby using parameters available from existing systems ofthe aircraft. Each aircraft motion must be represented bya data set of relevant parameter time histories.

The following operational parameters are necessary:

Ma Mach-numberAlt Altitude

n(x) Longitudinal Load Factorn(y) Lateral Load Factorn(z) Normal Load Factor

p Roll Rateq Pitch Rater Yaw Rate

t Maneuver Time

the Eulerian Angles, if available:

φ Bank Angleθ Pitch AltitudeΨ Heading

and additional parameters only for the verificationprocess:

α(alpha) Angle of Attackβ(beta) Angle of Sideslip

ξ(xi) Aileron / Flaperon Deflectionη(eta) Elevator Deflectionζ(zeta) Rudder Deflection

3.2.3.3 STANDARD MANEUVER PROCEDURE

Provided the operational parameter time histories of thebasic parameter are available in correct units, thisprocedure includes several steps:

(1) Maneuver type identification

(2) Normalization of relevant parameter time historiesfor a number of identified maneuvers of the samemaneuver type for comparison

(3) Determination of the mean values for each relevantparameter time history of the same maneuver type

(4) Idealization and tuning of the parameter timehistories

(5) Determination of the standard maneuver timehistories

The result of this procedure is a data set of standardizedparameter time histories. The parameters are roll rate,pitch rate and yaw rate of the selected maneuver type.See Figure 2.

29

FIG 2: Standard Maneuver Procedure

30

3.2.3.4 MANEUVER IDENTIFICATION

The goal of the maneuver identification is to select therelevant maneuver segments from the recordedoperational data base. A maneuver is identified bycomparing the observed data with the predefinedmaneuver characteristics as described in the ManeuverType Description of selected maneuvers:

TurnN(z) ≤ 2, p ≥ ± 20°/ sec, φ ≈ 40 ÷ 90°

Roll steady to bank angle, pull, the bank angle is held aslong as desired, opposite roll back to level

Roll rates of opposite sign before and after g peak.

High g TurnN(z) > 2Turn Maneuver

BreakN(z) > 3

High g Turn Maneuver with g peak during initialmaneuver time.

ScissorsA series of High g Turn Maneuvers

Roll ReversalN(z) >2, p >±20°/sec, φ ≈ 20 ÷ 90°

Roll steady to bank angle, directly opposite roll back tolevel.

High g Rolls / Barrel RollsN(z) > 1.5, p > ± 20°/sec, φ (max) ≈ 360°

Roll steady in one directionBarrel Roll over top θ rise to a peak value . Barrel rollunderneath θ descend to a negative peak value.

Pull sym.N(z) > 1.5 ∆ φ < 10°

From ≈ 1g to ≈ 1g

The maneuver identification parameters are mainly loadfactor n(z), roll rate p and bank angle φ.

First:The data are checked for completeness and suitability forseparating them into missions and maneuver types.

Second:The start and end time of each maneuver type areidentified when the roll rate is near zero and the g isapproximately 1.The bank angle also indicates the type of maneuver, i. e.full roll φ ≈ 360°, half roll φ ≈ 180°, turn < 90°

Figure3 :Identified Time Histories of Correlated Operational Parameters

31

FIG 4: Unified Roll Directions

FIG 5:Normalizsation of Parameters

32

Figure 3 shows as an example for the identification of aHigh g Turn Maneuver. In this case the roll rate traceprimarily defines the maneuver length.The pilot first rolls the aircraft in the direction of the turnand finally rolls it back to the wings level position. Inparallel, the g rises to a peak value. The peak is held aslong as desired. The g drops down from its peak as theaircraft is rolled back to the wings level.The start and the end of the maneuver are determined asfollows: the maneuver starts when the first negative /positive deflection of the roll rate trace starts and themaneuver finishes after recovering i.e. the oppositedeflection of the trace, decreased to zero.

The Eulerian angles φ, θ, Ψ,give the aircraft orientationwith respect to the earth’s coordinate system.

The bank angle values indicate the type of maneuver asdefined in Maneuver Type Description.

All recorded parameters are time related.

3.2.3.5 NORMALIZATION

Normalization is necessary because several maneuvers ofthe same type are different in roll direction , amplitude ofmotion and in maneuver time. For the calculation ofloads from operational maneuvers it not important toseparate the maneuver types into different roll directions.

Therefore, maneuvers of the same type are transformedinto a unified roll direction. See Figure 4.

For a requisite comparison, a two – dimensionalnormalization is necessary.Figure 5 illustrates the basic procedure of normalization.The ordinate presents one of the parameters of motion :y= n(y), n(z), p, ........for several maneuvers of the sametype : y(1), y(2), ........y(n).These parameters are normalized by relating them to themaximum values (absolute derivation from zero) whichhave occurred. This means the maximum value of eachnormalized parameter becomes in this case:

Y= y(1)max = y(2)max = + 1.0

The time is presented by the abscissa t , where by themaneuver executing time is marked by t(1), t(2), .......t(n)for several maneuvers.The normalization is accomplished in that way that:

• firstly, the maneuver time is chosen as the value 1.0 i. e.t(1)= t(2) = T = 1.0• secondly, the extreme values of the relevant parametersis chosen at the same normalized time.

The time scale normalization factor for all correlatedparameters: n(y),n(z),p, q, r, φ, θ, Ψ, within, foreexample, a High g Turn was derived from the roll ratetrace. See Figure 6

FIG 6: Correlated Parameters

33

FIG 7: Normalized Roll Rate Trace

FIG 8: Time Ratio

34

Fig.9 Shifted Roll Rate Traces

Fig. 10 Comparison of Normalized Rate Traces

35

In the normalized time scale, T=0 corresponds to the timewhen the roll rate trace first goes negative or positive(start of the maneuver ), and T=1 corresponds to the timewhen the roll rate trace is back to zero after the oppositeroll rate peak (finish of the maneuver). Figure 7 showsthe normalized roll rate trace (positive roll direction).

This normalization procedure is dependent on theaccurate maneuver start value. (p≈0)

In several cases the start values of the available timeslices are very poor. One reason is the low sample rate ofe.g. 1 or 2/sec. Recordings from flight tests are sampled24 times per second.An other reason is the selected parameter thresholdvalues of the data reduction and maneuver identificationprocess, combined with a low sample rate.For these cases an upgraded normalization procedure,derived from the basic procedure, is used.

The estimated time of a High g Turn t(m) had a veryhigh correlation with the difference between the time ofthe first and the second roll rate peak. See Figure 8. Thistime ratio is very important for the normalizationprocedure

The time transformation from real time into normalizedtime requires several steps:

1. Determination of time ratio. The time ratio is definedby t`(1)= dt/t(m)2. Harmonization For the comparison of the parametertraces, a harmonization of the maneuver time ratio isnecessary.

sfn = scale factor

3. Shifting A new interpolation of a similar number oftime steps for each of the correlated parameters for allmaneuver of the same type is necessary Then the roll ratetraces were shifted in a way, that all selected first peakscoincided at the same time step.

All correlated parameters are shifted parallel in thesimilar way.

Figure 9 presents the comparison of the shifted roll ratetraces versus normalized time for the selected High gTurn maneuvers.

The amplitudes of the traced are normalized individually.Each value of the trace is divided by its absolutedeviation value from zero, therefore, all normalizedamplitudes will fall between ±1.0.Figure 10 shows the result of the “peak to peak”normalization procedure.

The application of the two-dimensional normalizationprocedure is very helpful for the comparison of maneuvertime histories. In this normalized form, all parameter timehistories are independent of the aircraft type.

3.2.3.6 MEAN VALUES

After normalization of the maneuver time, for all selectedmaneuvers of the same type, the typical values of therelevant parameters – in this case the peaks of the rollrate – coincide at the same normalized time. Eachparameter time history contains the similar number oftime steps, independent of is individual maneuver length.This is the basis for calculating the arithmetic meanvalues for each of the time steps.

Figure 9 presents the comparison of the non- normalizedroll rate traces versus normalized time for the selectedHigh g Turn maneuvers. The roll rate is a good examplefor all relevant parameters.Note: The amplitudes for the mean value calculation arenot normalized.

The mean value is defined by:

y j

y j

nm

i

i

n

( )( )

= =∑

1

n = number of maneuver of the same type

j = time step

yi (j) = relevant parameter

ym (j) = mean value

The mean values of all parameters have been formed incombination by smoothing of the time history.For plot comparison, a normalization of the amplitude isnecessary.

3.2.3.7 IDEALIZATION

The mean value traces represent a good estimation of therelationship between the selected parameters during amaneuver (e. g . High g Turn ).For the compensation of any minor errors by the meanvalue calculation and for reasons of compatibility, themean values have to be idealized and tuned.The interpretation of “idealized and tuned” as follows:To cover the most extreme peaks of the control surfacedeflections possible, the most extreme accelerations inroll (p), pitch (q), and yaw (r ) are used.These values are obtained by linearization of theacceleration time history in a way that the same responseof the aircraft is obtained.

For the idealization, the calculation is performed in threesteps.

nn sftsftsftsft ∗′==∗′=∗′=∗′ .....332211

36

�yy

x=

∆∆

In the first step, the following parameters werecalculated:

The three angular accelerations p, q and r bydifferentiating the three angular rates p (roll), q (pitch)and r (yaw) with respect to maneuver time.

The differentiation was given by

In the second step, the acceleration traces p, q, r, werereplaced by linearized traces

With respect to the zeros of the traces and extremevalues of p, q, r and thecorresponding extreme values of roll -, pitch- and yawrate.

Figure 11 presents the comparison of derived rollacceleration trace and idealized trace versus maneuvertime for a High g Turn Maneuver.

In the third step, the three angular rates (roll, pitch,yaw) were recalculatedBy integrating the idealized values of the three angularaccelerations (p, q, r).

FIG 11 : Idealization Traces

FIG 12: Standard Maneuver

For the reasons of compatibility, the idealized data haveto be tuned, that means the relation between the three

Eulerian angles Φ, Θ Ψ and the angular rates p, q, r isverified with the equations:

The result is the standardized maneuver.

Figure 12 presents the idealized and tuned “standardized”traces of the three angular rates for a High g Turnmaneuver.For each type of standardized maneuver the normalizedmotion parameters are independent of aircraft type, massconfiguration and flight control system.

3.2.4 Flight Loads derived from OperationalManeuvers

The determination of operational loads is considered asfeasible applying an Operational Maneuver Model. Theessential input for the Operational Maneuver Model is aset of Operational Standard Maneuvers consisting ofnormalized operational parameter time histories, asdetermined in 3.2.3.The operational loads can be determined by introducingaircraft basic data, flight condition and boundaryconditions for the maneuver to be considered.

3.2.4.1 Application of the Operational ManeuverModel

The application of the Operational Maneuver Model isfeasible for the determination of loads in general.

� for Extreme Operational Loads / Limit Loads takinginto account the boundary conditions for design

� for Fatigue Loads by building a usage spectrummade up of reconstituted Operational StandardManeuvers

� for Loads related to recorded parameters taking intoaccount the recorded parameters directly withoutapplication of standardization procedure(normalization, mean values, tuning, idealization)and without tailoring by boundary conditions

Aircraft Basic Data

Aircraft basic data is the main input for the OperationalManeuver Model and is required to perform thereconstitution from the standardized maneuvers.

q = ∗ + ∗ ∗� cos � sin cosΘ Φ Ψ Φ Θ

p = − ∗� � sinΦ Ψ Θ

r = − ∗ + ∗ ∗� sin � cos cosΘ Φ Ψ Φ Θ

37

For calculation of the control deflections necessary togenerate the operational parameter time history, thefollowing data are needed:

• Aircraft configuration• geometric data• operational mass• inertia properties

• Aerodynamic data set for the aircraft Cl, Cm= f(α),Cy, Cl, Cn = f(α,β)

• Flight Control System data• for conventionally controlled aircraft: mechanical

gearing / limits

• for active controlled aircraft: flight control law(EFCS)

• Engine data- thrust

• Flight Condition- airspeed, Ma– altitude

3.2.4.2 For calculation of structural loads onaircraft components the following data areneeded:

- aerodynamic data set for the components to beconsidered (wing. tailplane)- mass data for the components to be considered

3.2.4.3 Boundary Conditions as Design Criteria

Boundary Conditions have to be considered as the maininput for defining the load level.This is necessary for the determination of the extremeoperational maneuvers and consequently for theverification of design loads.

The boundary parameters to be defined for an operationalmaneuver are:

→→→→ Design Maneuverso the shortest maneuver time (Tman = minimum)o realizable by the control system and the

aerodynamic limitso the maximum vertical load factor ( nz )o the maximum lateral load factor ( ny )o the maximum bank angle (φ) for the maneuver

to be considered

These boundary condition parameters can be derivedfrom spectra of main load parameters by applyingextreme value distributions, an example is shown inFigure 13.If no spectra are available the main load parametersstated in the Design Requirements ( MIL – Spec. ) can beapplied.

→→→→ Fatigue ManeuversAll the main load parameters can be taken from relatedspectra available.

The procedure of Operational Maneuver Model isshown in Figure 14 as a flow chart.

Using the Standardized Operational Parameters thereconstitution into real time is performed.For these operational parameters time histories in realtime the control deflections necessary to generate theoperational maneuver can be determined as follows:→ roll control ξ by applying roll and yaw equations

→ pitch control η using the pitch equation, taking intoaccount the symmetrical aileron deflections

→ yaw control ζ by applying sideslip and yaw equations

Using these control deflections the response calculation isdone for real time conditions, but for the purpose ofchecking the results with respect to the standardizedmaneuvers, the response parameters are normalized.In a comparison of the parameters between input andoutput, the standardization is checked. In case ofconfirmation of the conformity of the main responseparameters with the standardized parameters, the outputparameters are considered to be verified. These verifieddata represent the model parameters for load calculation.The calculation of the Operational Loads is performed inthe conventional manner applying the verified modelparameters in particular the control deflectionsdetermined for the Operational Maneuver to beconsidered.

38

FIG 13 : Boundary Conditions for Design Maneuvers

39

FIG 14 : Procedure of the Operational Maneuver Model

40

3.2.5 Flight Parameter Envelopes Approach

Abstract

This part of the manual will explain in detail the FlightParameter Envelope Approach:

A new method to determine the critical flight design loadsfor a modern control configured fighter aircraft. The wayfrom the initial design phase up to the Final OperationalClearance (FOC) will be examined.

The Flight Parameter Envelope Approach has to be seenin conjunction with the new design tools (i.e. LoadsModel) and the modern digital Flight Control Systemswith carefree handling and load limiting procedures. Thedefinition of Flight Parameter Envelopes will then beuseful and feasible if computer tools are available to doextensive load investigations for the total aircraft underbalanced aircraft conditions and if the FCS will limit theaircraft responses (carefree handling) and with it theaircraft loads (load limiting system).

The definition of Flight Parameter Envelopes may be aproblem for new aeroplanes where in the beginning ofthe aircraft development only limited information aboutthe aircraft responses from previous or similar aircraft isavailable. New techniques, such as thrust vectoring forhigh angle of attack maneuvering in combination withhigher dynamic pressures may cause new problems. Butthe poststall flight conditions up to now known are onlyloads critical locally because the dynamic pressures inthe flown poststall regime is low.

However for aircraft like the Eurofighter generation thedefinition of Flight Parameter Envelopes is a useful andfeasible approach to determine the critical flight designloads and to overcome the additional problem thatMilitary Specifications became more and more obsoletefor aircraft design.

List of Symbols

A/C AircraftALE Allowable Loads Envelope

CFC Carbon Fibre Composites

DOF Degree of FreedomFCS Flight Control System

FOC Final Operational Clearance

HISSS Aerodynamic Program - HigherOrder Panel Sub- and SupersonicSingularity Method

IFTC Initial Flight Training Clearance

MAST Major Airframe Static Test

MAFT Major Airframe Fatigue Test

MLA Maneuver Load Alleviation

RF Structural Reserve Factor

flimit Limit Load Factor

fult. Ultimate Load Factor

Fx, Fy, Fz Forces

Mx, My, Mz Moments

c. g. center of gravity

qdyn dynamic pressure

nx, ny, nz load factors

p roll velocity

q pitch velocity

r yaw velocity

pdot roll acceleration

qdot pitch acceleration

rdot yaw acceleration

α angle of attack

β sideslip angle

β∗ qdyn product of sideslip angle anddynamic pressure

ηF/P foreplane deflection angle

ηT/E trailing edge deflection angle

δR rudder deflection angle

3.2.5.1 Introduction

When starting with feasibility studies for a new fighteraircraft in the beginning of the eighties indications froman aircraft designed in the early seventies were confirmedthat a change of the applications of MilitarySpecifications for the aircraft design would be necessary.This was also valid for the evaluation of aircraft designloads (e.g. MIL-A-08861A).

The increase in new technologies e.g.

increase of computer capacity

digital flight control systems (FCS)

new materials – e.g. Carbon Fibre Composites (CFC)

better and more efficient design tools – e.g. StructuralOptimization Tool, Loads Model, etc.

led to a change of the design and performancerequirements for a new fighter generation.

The high work load of the pilots should be reduced incontrast to the increase of the tasks of the aircraft such asperformance, agility, etc.. The consequence was to design

an aerodynamic unstable aircraft - increase of agilitywith a digital Flight Control System (FCS)

The requirement to reduce the workload of the pilot couldbe fulfilled by a carefree handling and automatic loadlimiting procedure in the FCS control laws. With it thecontrol function of the pilot for the instrument panel inthe cockpit is reduced to a minimum and eyes out of thecockpit whilst maneuvering is possible.

To overcome the new situation for calculation of criticaldesign loads for modern fighter aircraft the so calledFlight Parameter Envelope Approach was developed andwill be described here for an aerodynamically unstableaircraft with foreplanes (see Fig. 1) featuring:

41

• artificial longitudinal stability

• extensive control augmentation throughout the flightenvelope

• carefree maneuver capability with automatic loadprotection achieved by careful control of maneuverresponse parameters

Fig. 1 - “Demonstrator Aircraft” for Flight ParameterEnvelope Approach

The main problem is to realize an agile and carefree loadlimiting FCS. Therefore a robust structural design of theairframe is necessary including an appropriate growthpotential for possible changes of the FCS control lawscovering aircraft role changes which may influence thedesign loads and with it the aircraft structure. To makesure that the airframe and the FCS are harmonized:aircraft structure and FCS control laws have to bedeveloped concurrently.

In comparison to earlier aircraft like Tornado the designloads for the new FCS controlled fighter aircraft have tobe defined without a detailed knowledge of the finalstandard of the FCS because

a very limited understanding of the FCS- control lawsis available in the initial design phase.

This problem can be solved by the definition of newStructural Design Criteria where among other designconditions the principal flight maneuver requirements forthe aircraft have to be defined. In this case the FCSdependent loads critical Flight Parameter Envelopes (s.Fig. 2) are defined by:

translatory accelerations (ny, nz)

rotational velocities (p, r)

rotational accelerations (pdot, qdot, rdot)

sideslip conditions (β∗ qdyn)

etc.

To take into consideration all requirements of thedifferent aircraft design disciplines the Flight ParameterEnvelopes have to be defined in not only consideringFCS but also

Flightmechanics

Aerodynamics

Structural Dynamics

Loads

Fig. 2 – Loads Critical Flight Parameter Envelopes for the Loads Model – Interdependence between the Flight ParameterEnvelopes and Critical Design Load Cases for Main A/C- Components

42

The calculation of aircraft design loads will be done witha modern computer tool the so called Loads Model andthe Flight Parameter Envelopes are a part of this tool.

3.2.5.2 The Flight Parameter Envelope Approachand the Loads Model

Both the FCS dependent Flight Parameter Envelopes(Fig. 2) and the Loads Model (Fig. 3) result in a highlyefficient computer tool for aircraft design loadcalculations:

- the maneuver requirements of the aircraft controlledby the FCS are indirectly defined by the FlightParameter Envelopes and the Loads Model containsall the important aircraft mass and aerodynamicinformation’s which have to be known to calculatethe critical design loads for the aircraft

3.2.5.3 Description of the Loads Model

The today’s computer capacities allow extensive loadinvestigations considering:

- all mass information’s (masses, c.g.’s, moments-of-inertia, mass distributions) for the total aircraft andspecific aircraft components

- the corresponding aerodynamic information(aerodynamic pressures, aerodynamic coefficients/derivatives) for the total aircraft and the definedaircraft components for different Mach numbers

- the static aeroelastic input (flexibility factors andincrements for total aircraft and aircraft

components) to correct the rigid aerodynamics(aerodynamic pressures, aerodynamic coefficients/derivatives) for defined Mach numbers.

The mass- and aerodynamic data have to be stated fordifferent loads critical aircraft configurations.

The idea of the Loads Model is to calculate the criticalaircraft component design loads (aircraft componentloads envelopes) to get balanced load cases for the totalaircraft. That means the total sum of the aircraftcomponent forces and moments is zero (equilibrium) foreach load case:

Σ Fx,y,z = 0 Σ Mx,y,z = 0

These balanced load cases (Fig. 4) are the basis for thecalculation of nodal point loads for the total aircraftFinite Element Model (FE-Model) and for the stressanalysis.

Simplified the Loads Model is a combination of big input

and output data files and a number of computer programs(Fig. 3). The input data sets contain all information whichis necessary for load calculations while the output datasets contain the results of the load calculations as loadcase conditions, forces, moments, aircraft componentload envelopes, etc..

The computer programs of the Loads Model can beclassified into two different groups

- programs to establish and to handle the required datasets

- programs to compute the critical aircraft componentloads (balanced load cases, loads envelopes)

Fig. 3 – Loads Model - Overall View

43

Fig. 4 – Total Aircraft – Balanced Load Case

To use the Loads Model efficiently the structural designrules including the flight maneuver requirements have tobe defined for the new aircraft. This will be done in theSDC.

3.2.5.4 Structural Design Criteria (SDC)

Because more and more the Military Specifications (e.g.MIL-A-08861A) are getting obsolete for the design ofmodern fighter aircraft it becomes important to define thenew structural design rules in the Structural DesignCriteria.

The following conditions have to be defined in theStructural Design Criteria:

Design Flight Envelope- Mach/altitude

nz-max./min. vs. Mach

flimit, fult. - limit/ultimate load factor

Loads critical aircraft configurations with andwithout stores – key configurations

Aircraft design masses:

Basic Flight Design Mass, Maximum Design Mass,Minimum Flying Mass, Landing Design Mass, etc.

Gust conditions:

gust design speeds in combination with aircraftspeeds, gust lengths

Temperatures:

maximum recovery temperature

maximum stagnation temperature

Ground Loads Criteria:

sink rate, crosswind, arresting, repaired runway, etc.

Departure and Spin

Hammershock conditions

Bird strike conditions

Static aeroelastic requirements

Flutter/divergence requirements

Fatigue conditions:

safe life or fail save philosophy

g-spectrum, scatter factor, aircraft service life, etc.

etc.

Additional to the above described design conditions also

the principal flight manoeuvre requirements for theaircraft

have to be defined.

3.2.5.5 Flight Parameter Envelopes for StructuralDesign

The application of the single axis pitch, roll or yawmaneuvers (MIL-A-08861A) is no longer sufficient forthe definition of design loads (Fig. 5 and Fig. 6).

0.0 0.5 1.0 1.5 2.0 2.5

-15

-10

-5

0

5

10

15

Time (s)

Y T

itle

Vertical Load Factor Angle-of-Attack Pitch Rate Taileron Pilot Input

Fig. 5 – MIL - Pull-Push Maneuver

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0

-6

-4

-2

0

2

4

6

8

Time (s)

Vertical Load Factor Angle-of-Sideslip Lateral Load Factor Taileron Roll Rate Roll Acceleration Yaw Rate Pilot Input

Fig. 6 – MIL - Rolling Pull Out Maneuver

The carefree maneuver capability with automatic loadprotection allows the superposition of combined pilotcontrol inputs in roll, pitch and yaw and with it numerousdifferent operational maneuvers which have to be takenunder consideration to find the critical design loads.Some typical pilot stick inputs for flight clearancemaneuvers are shown in Fig. 7.

44

Fig. 7 – Typical Pilot Stick Input

The following Flight Parameter Envelopes have to bedefined (s. Fig. 2):

nz = f(qdot)

ny = f(rdot)

nz = f(p, pdot, r, rdot, ny, β*qdyn)

p, r vs. pdot, rdot

As it can be seen mainly the inertia dominated parametersas the translatory accelerations (nz, ny) and the rotationalvelocities (p, r) and rotational accelerations (pdot, qdot, rdot)have to be defined while only one aerodynamicparameter is β∗ qdyn (sideslip angle ∗ dynamic pressure).The sideslip angle β is well controllable by the FCS andwith it the product β∗ qdyn. β∗ qdyn can be defined underconsideration of the gust requirements for the aircraft.

Important for the definition of the Flight ParameterEnvelopes for the structural design of an aircraft are alsothe possible tolerances of the flight parameters (s. Fig. 8).These have to be defined

Fig. 8 – Flight Control System Design - Tolerance of Flight Parameter

- For example:

to define nzmax./min. for the most important FlightParameter Envelopes

nz = f(qdot)

nz = f(p, pdot, r, rdot, ny, β*qdyn)

it should be known how exact the FCS controls thevertical load factor nz (s. Fig. 8):

nz = nz max./min. ± ∆nz

If in this case the defined tolerances are to small anincrease of the nz overswing (±∆nz) may causeproblems, because the load limiting procedure ofthe FCS can become uncertain therefore or on theother hand an increase of the critical aircraft loadshas to be accepted for which the aircraft structurehas to be checked for.

These Flight Parameter Envelopes will be used now todetermine the design load and the load envelopes for theaircraft main components – see Para. 3.2.5.8.

The interdependence between the Flight ParameterEnvelopes and critical design load cases for the differentaircraft components can be seen on Fig. 2.

3.2.5.6 Total Aircraft and ComponentAerodynamics

To get “balanced load cases” the total aircraftaerodynamic as well as the corresponding componentaerodynamic is integrated in the Loads Model regardingall loads critical aerodynamic influences. The result mustfulfil the condition:

- sum of component aerodynamics = total aircraftaerodynamics

The following aerodynamic data sets are part of theLoads Model:

- aerodynamic pressures of the total aircraft for allaerodynamic influences (α, β, control surfacedeflections, p, q, r, etc.) for different Mach numbers

- the corresponding aerodynamiccoefficients/derivatives of the aircraft components -result of aerodynamic pressure integration – for alldefined monitor stations (Fig. 9)

- the corresponding aerodynamiccoefficients/derivatives of the total aircraft – sum ofcomponent coefficient/ derivatives

- the static aeroelastic corrections of the aerodynamicpressures for all aerodynamic influences as

α, β, control deflections, p, q, r, etc.

and the aerodynamic pressures of aeroelastic inertiaeffects and the corresponding integration results(coefficients/derivatives) for

nz, ny, pdot, qdot, rdot

together with the correction factors and incrementsfor the aerodynamic coefficients/derivatives for theaircraft components and the total aircraft

- the corrected flexible aerodynamic pressuresincluding the corresponding flexible total aircraftaerodynamics and the flexible aircraft componentaerodynamics

45

The main programs for establishing the requiredaerodynamic data sets and for data set handling are:

- a theoretical aerodynamic program (e.g. the DasaHISSS program – higher order panel method) tocalculate the rigid aerodynamic pressures for theabove described loads relevant aerodynamicinfluences.

In Fig. 10 it is shown how starting from a CATIAmodel the HISS panel model will be derived.

- a correlation and integration program to compareand correct the theoretical total aircraft aerodynamicresults up to first total aircraft wind tunnelmeasurements and with it to correct the aerodynamic

Fig. 10 – HISSS Panel Model of “DemonstratorAircraft” – Calculation of Aerodynamic Pressures

for Total Aircraft

- pressures and the aerodynamic coefficients/derivatives for the aircraft and the aircraftcomponents

- a static aeroelastic program to calculate theaeroelastic pressure increments for the correction ofthe rigid pressure distributions and to calculate thecorrection factors and increments for theaerodynamic coefficients/derivatives for the aircraftcomponents and the total aircraft to establish theflexible aerodynamic data set.

an aerodynamic pressure summation program tosummarize the aerodynamic pressures due to

α, β, control deflections, p, q, r, etc.

for the selected critical load cases to calculate theaerodynamic nodal point loads for the FE- Model.

3.2.5.7 Total Aircraft- and Component Masses

For the calculation of “balanced load cases” the massconditions for the defined design masses (Basic FlightDesign Mass, Maximum Design Mass, Minimum FlyingMass, Landing Design Mass, etc.) for the total aircraft as

aircraft mass

aircraft c.g.

aircraft moments of inertia

as well as the corresponding component mass conditionshave to be integrated into the Loads Model.

- Sum of component masses = total aircraft mass

Fig. 9 - Load Monitor Stations for “Demonstrator Aircraft” and Corresponding Main Loads Components

46

The following mass data sets are part of the LoadsModel:

- the aircraft component masses, component c.g.’s andmoments of inertia including the correspondinginternal fuel states and external stores (Fig. 9 – A/CMonitor Stations)

- the total aircraft mass, c.g., moments of inertiaincluding the internal fuel states and external storesas sum of the above described aircraft componentmasses

3.2.5.8 Aircraft Loads Monitoring

The calculation of critical design load cases (loadsmonitoring) for the aircraft components (monitorstations) can be started when the required input data setsfor the Loads Model are established. The outcome of theaircraft loads monitoring are Loads Envelopes (Fig. 11)for the defined monitor stations.

The computer program which will be used for thecalculation of critical load cases under consideration ofthe defined Flight Parameter Envelopes is the so called“Balance Program”. The loads analysis for the monitorstations (Fig. 9) will be performed by means of userdefined dynamic equilibrium points (time steps of a timedependent flight simulation):

- The user has to define for each load case thefollowing flight parameters

Mach number, altitude, nz, ny, p, pdot, q, qdot, r, rdot

respecting the Flight Parameter Envelopes (Fig. 2)and as a special case for this “demonstrator” aircraft

the foreplane deflection (ηF/P) and trailing edgedeflection (ηT/E-sym.)

under consideration of the foreplane schedule

- The Balance Program will define the remainingones:

α, β, η-T/E-sym. or η-F/P, η-T/E-unsym., δ-R

and nx and the thrust level

if required. In a second step the corresponding air-,inertia- and net- loads for all monitor stations arecomputed for the selection of critical design loadsto establish the loads envelopes for the definedaircraft components

To be sure that the defined requirements will be fulfilledthe program also checks

- the derived control surface deflection anglescompared to the max. deflection angles

- the derived hinge moments for the control surfacescompared to the max. defined hinge moments ifnecessary

- the user defined flight parameters compared to theFlight Parameter Envelopes

It seems to be useful to establish a program for loadscalculations which can be used for different degrees offreedom (DOF):

- 6 DOF – balance of Fx, Fy, Fz, Mx, My, Mz

- 5 DOF - without Fx balance (tangential force)

- 3 DOF – balance of Fx, Fz, My for pure symmetricconditions

- 2 DOF – balance of Fz, My for pure symmetricconditions without Fx balance

It should also be possible later on in the aircraft clearancephase when the carefree handling and load limiting FCSis available to use a flight simulation program to do timedependent loads critical flight simulations and tocalculate the corresponding flight load time histories (air-, inertia-, net- loads for all time steps) for the aircraftmonitor stations with the Loads Model.

Fig. 11 – Example of Loads Envelopes for Monitor Stations – Design Load Cases

47

To fulfil the above described additional program checkfunctions the following margins have to be defined:

- max. deflection angles for control surfaces versusMach number

- max. allowable hinge moments for the controlsurfaces respective max. normal forces if necessary -as result of structural optimization of wing, fin andforeplane

- engine thrust conditions if necessary

- Maneuver Load Alleviation (MLA) concept if theFCS will have a MLA procedure – to reduce thewing bending moment – respective the other in Para.3.2.5.13 described load reducing FCS rules

- as a special case for this “demonstrator” aircraft theforeplane trim schedule including possibletolerances because the foreplane and the trailingedge flaps will be used for symmetric flight control

3.2.5.9 Loads Process, Aircraft Design andClearance Phases

After the feasibility studies respective definition phasethe normal development process of an aircraft structurehas three phases:

- Design Phase

- Check Stress Phase

- Structural Clearance Phase

For these three development phases the accuracy of theinput data (aircraft masses, aerodynamic, etc.) for theLoads Model differs and with it the accuracy of the loadcalculations. But as explained before the standard of theinput data for the Loads Model is relatively high even atthe beginning of the aircraft development due to moderncomputer tools (i.e. theoretical aerodynamic programs)and the possible crossreading to other similar aircraft.

But more important is that with the Flight ParameterEnvelopes the principal flight maneuver requirements forthe aircraft can be defined very early and with it theinteraction of FCS and the aircraft loads. During thedevelopment of the aircraft structure the Flight ParameterEnvelopes have to be checked in line with the FCSdevelopment.

3.2.5.10 Design Phase

Before starting loads calculations with the 1st flexibleLoads Model in the Design Phase the in Para. 3.2.3.8described prerequisites have to be settled additional to theFlight Parameter Envelopes to be sure that the loads arethe critical ones and are not maximized:

- A structural optimization has to be done and with itan optimization of the control surface efficienciesunder consideration of aeroelastic influences, failureconditions and deflection rates (Fig. 12). Based onthese optimization studies the critical hingemoments respective normal forces for the controlsurfaces can be defined. The result of optimization is“configuration freeze”.

- The max. deflection angles versus Mach number andthe maneuver conditions for the control surfaceshave to be defined – for example the foreplane trimschedule.

- A maneuver load alleviation (MLA) concept shouldbe defined if necessary under consideration of

the required reduction of wing root bendingmoment for high g conditions

the trailing edge split flap schedule as functionof g respective α

the foreplane trim schedule.

Fig. 12 – Flexible Loads Model -Static Aeroelastic Influences

If all these prerequisites are defined and integrated in theLoads Model the load investigation can start.

During the Design Phase the Loads Model consists oftheoretical linear aerodynamics compared with firstwindtunnel test results and corrected if necessary. Theflexible aerodynamic data set includes all important staticaeroelastic corrections for selected Mach/altitude points(Fig. 13).

Fig. 13 – Flight Envelope Mach-Altitude Points forFlexible Loads Model – Flexible Aerodynamic Data Set

The main benefit to do the load investigations with thefirst flexible Loads Model is

- the loads for the aircraft components can becalculated for total aircraft balanced conditions fordifferent aerodynamic configurations (with andwithout stores) and different aircraft masses (fuel,external stores) under consideration of the FCSrequirements (Flight Parameter Envelopes).

48

3.2.5.11 Check Stress Phase

The Check Stress Phase is the second development phase.The design loads have to be checked and updated withthe updated Loads Model for the design of the productionaircraft structure:

- the panel model for the theoretical aerodynamiccalculations has to be updated (configurationchanges, external stores, etc.)

- the new theoretical linear aerodynamic has to beupdated by comparing and correcting it to the latestwindtunnel tests (configuration changes, additionalstore configurations, mass flow, etc.)

- first windtunnel based store aerodynamic incrementscan be available (store balances) and can beincluded in the Loads Model

- the static aeroelastic corrections have to be updatedby using the updated structure (FE- Model) and theupdated aerodynamic pressures

- the aircraft masses have to be updated forproduction aircraft standard

- the foreplane trim schedule and the tolerances forthe trim schedule have to be updated

- the MLA concept has to be checked and updated ifnecessary

- the max. hinge moments for the control surfaceshave to be checked and updated if necessary

- if required additional monitor stations have to beincluded in the Loads Model

- the Flight Parameter Envelopes have to be checkedand updated in line with the FCS development. Thatmeans in detail that the flight control laws have tobe reviewed during all design phases to check theirfunction as a load limiting system. For example thedefined tolerances of the Flight ParameterEnvelopes have to be checked, e.g. the nz tolerances:

nz max./min. ± ∆nz

as explained in Para. 3.2.5.5.

As for the Design Phase the load calculations have to bedone by using the Balance Program and the updatedFlight Parameter Envelopes. The up to now availableFCS has only a check function because the carefreehandling and load limiting procedures are not finallyagreed (preliminary carefree handling). The loadinvestigation should be expanded and additionalMach/altitude points should be considered.

The revised aircraft component design load cases(balanced load cases, load envelopes) from the CheckStress Phase are the basis for the stress analysis for theproduction aircraft and with it for the structural clearanceactivities in the Clearance Phase.

3.2.5.12 Structural Clearance Phase

The aircraft clearance will be done in different steps fromthe first flight clearance for the prototypes up to theInitial Flight Training Clearance (IFTC) and the FinalOperational Clearance (FOC - 100% load level) for theproduction aircraft.

The aircraft structure has to be cleared for the conditionsdefined in the Structural Design Criteria as there are:

design flight envelope (Ma/altitude)

critical aircraft configurations

limit/ultimate load factor

aircraft design masses

nz-max./min. vs. Mach

etc.

Fig. 14 – Allowable Load Envelope for AircraftClearance Phases – Structural Reserve Factors < 1.0 are

considered

For the clearance of the aircraft structure so calledAllowable Loads Envelopes (ALE) will be used. TheALE’s (Fig. 14) contain the structural information of theprototypes respective of the production aircraft. TheALE’s have to be defined by the stress office based onthe design load envelopes of the aircraft components andunder consideration of the results from the stress analysisand structural tests. To be on the severe side during theclearance activities (flight test) only structural ReserveFactors (RF) < 1.0 have to be considered in the ALE’s.

The prerequisites to increase the clearance level are :

- Major Airframe Static Test (MAST) to limit,ultimate, failure load condition and other aircraftcomponent tests - to check the aircraft structure

- FCS updates – from preliminary carefree handling tofull carefree handling to check the load limitingprocedure of the FCS

- Validation of the Loads Model via the Flight LoadSurvey to update the data basis for loads monitoringand to proof also the load limiting procedure of theFCS

The first Loads Model for the structural clearance of theaircraft consists of non-linear aerodynamic data based onwind tunnel pressure plotting measurements. Thevalidation of this non-linear Loads Model will be done bythe Flight Load Survey. The Flight Load Survey will beperformed for selected primary aircraft configurations(clean aircraft and external store configurations). Duringthe Flight Load Survey aerodynamic pressures of thesurfaces (wing, foreplane, fin) and the fuselage will be

49

measured (Fig. 15). The integrated pressures(aerodynamic coefficients for the total aircraft and foraircraft components) will be correlated against the loadpredictions from the non-linear Loads Model. The LoadsModel will be than corrected where significantdiscrepancies exist. Finally the flight validated LoadsModel for the primary aircraft configurations is availableand should be used for the Final Operational Clearance(FOC) – 100 % load level and production FCS.

During the Structural Clearance Phase at all clearancelevels the confidence that the load level will not beexceeded has to be shown by the loads monitoring ofloads critical flight simulations using the current FCS andthe validated Loads Model. Some typical pilot stickinputs for the flight simulations (flight clearancemaneuvers) are shown on Fig. 7.

The loads from the simulated flight maneuvers have to becompared to the Allowable Loads Envelopes for eachmonitor station. If the loads monitoring shows that theloads are inside the ALE’s the clearance step is fulfilled.If not:

- the areas have to be defined where control lawchanges are required to maintain acceptable loads

or

- modifications may be necessary to improve theaircraft structure for higher loads

3.2.5.13 Load Optimized Maneuvers

In the past the aircraft were optimized mainly toaerodynamic performance conditions (drag, etc.) and thedesign loads were the result of the aerodynamicconfiguration, the aircraft mass conditions and theapplication of single axis pitch, roll or yaw maneuvers(e.g. MIL-A-08861A).

A new possibility for the latest high performance fighteraircraft generation like Eurofighter are load optimizedmaneuvers because the FCS can be used in some casesfor load reduction under the consideration that the aircraftperformance is not prejudiced.

Three examples for load optimized maneuvers controlledby the FCS are given below:

1. Load optimized foreplane/trailing edge deflectionschedule as a special case for the “demonstrator”aircraft described in this paper:

a) reduction of front fuselage loads

The front fuselage loads are normallydominated by the inertia loads. To reduce thefront fuselage loads (Fz -normal force and My -vertical bending moment) the foreplane has tobe deflected in that way that the aerodynamicforeplane loads are acting against the frontfuselage inertia loads (s. Fig.16 ). In this casethe aircraft has to be controlled by the trailingedge flaps.

b) reduction of trailing edge flap loads - e.g.hinge moments.

For low g conditions (1g) where the maximumroll performance of the aircraft is required thetrailing edge flaps can be zero loaded for theaircraft trim conditions by trimming the aircraftonly with the foreplane. The trailing edge flapitself has to be deflected in that way that the αinfluence on the flap will be compensated:ηT/E-symm(nz=1.0)= f(α, Mach, A/C-cg)With it the flap hinge moments can be reducedand the roll efficiency of the aircraft can beincreased in some cases.

Fig. 15 – Flight Load Survey - Pressure Transducers at the Prototype of “Demonstrator Aircraft”

50

Fig. 16 – Front Fuselage - Load Reduction LoadOptimized Foreplane/Trailing Edge Schedule

Procedure a) may be used only for the front fuselageloads critical flight conditions as high g’s turns at lowaircraft masses (minimum flying mass) where the normalaerodynamic discharge for the front fuselage is aminimum and with it the net load is a maximum. In thiscase the trailing edge flap loading is relatively lowcompared to the maximum aircraft rolling conditions andcan be used therefore for exclusive aircraft control in thepitch axis. In all other cases the aircraft performance willbe more important.Procedure b) is a possible solution for hinge momentreduction if the control surface loads are increasing andthe size of the flap actuators cannot be changed.

2. Maneuver Load Alleviation - MLA (differentialtrailing edge flap deflection of i/b- o/b- flap):

the shift of the aerodynamic center of pressuretowards the wing root reduces the wing rootbending moment and with it the wingattachment load conditionsIn this case the i/b- flap has to be deflecteddownwards to increase the wing lift in theinboard wing area while the o/b- flap has to bedeflected upwards to reduce the lift in theoutboard wing area under the condition that thetotal wing lift has not to be changed (s. Fig. 17).This differential trailing edge flap deflectionhas to be superimposed to the full span trailingedge flap trim condition. The small effect onthe aircraft trim conditions by using the MLA-system has to be corrected by a full spantrailing edge deflection itself or by theforeplane.

Fig. 17 – Maneuver Load Alleviation (MLA) Change ofWing Lift Distribution and Shift of Center of Pressure

The MLA- system could be important at highg’s and high dynamic pressure in the lower α-region (elliptical wing lift distribution, linearaerodynamics).At higher α there may be a natural shift of thecenter of pressure to the wing root because thewing lift distribution becomes more and more atriangle due to non linear aerodynamics. (s. Fig.18).

Fig. 18 – Spanwise Normal Force Distribution NaturalShift of Center of Pressure to the Wing Root

The MLA- system can be important for the critical wingup bending conditions at max. g’s for the static designrespective the most critical g’s (mean proportional g’s)for fatigue design because the aerodynamic design oftendidn’t allow to increase the lever arm of the wing rootattachment to carry over the wing bending moment by acouple of forces (s. Fig. 19).

3. Prevention of overswing of control surfaces(deflection angles):

to prevent load peaks on the control surfacesduring rapid aircraft maneuvers (e.g. rapidrolling) an overswing of the control surfacesshould be avoided. An example for the trailingedge flap is shown on Fig. 20. In this case theoverswing of the flap is optimized by a smallchange of the T90 condition and with it the flaploads (hinge moments) are reducedsignificantly.

Fig. 19 – Wing Root – Carry Over of Wing BendingMoment

The above described maneuvers can be defined for thecritical static design loads as well as for fatigue loadswhich becomes more and more important for thestructural design of the aircraft.

51

In all these cases it must decided whether the loadoptimized maneuvers sacrifice aircraft performance orwhether the benefit (i.e. mass saving) is big enough tocompensate the loss of performance!

Fig. 20 – Dynamic Overswing of Trailing Edge Flaps – Change of T-90 Conditions

One way to assess this question is to evaluate requiredoperational maneuvers with respect to extreme or fatiguemaneuvers as evaluated by the former AGARD-WG 27(AGARD AR 340). For further information see Chapter3.2.2 – Operational Flight Parameter Approach.

On the other hand the β∗ qdyn requirement defined in theflight parameter envelopes (s. Fig. 2) is also a loadlimiting condition controlled by the FCS as explained inPara. 3.2.5.5. With it the Fin loads and the side force andside bending moment of the rear and front fuselage canbe limited.

3.2.5.14 Ultimate Load Factors

Historically a reduction of the ultimate load factor fult.

was done several times down to fult.=1.5 now which wasfor a long time seen as the lowest possible limit.

The situation was changed for FCS controlled aircraftwith carefree handling and load limiting procedures.

Based on the assumption that the aerodynamic and inertiaflight loads for the aircraft are limited by the FCS bycontrolling the important flight parameters

β, p and nz respective α

directly the ultimate load factor can be reduced forexample from

fult.=1.5 to fult.=1.4

(as agreed with the British-, German-, Italian- andSpanish- authorities for the Eurofighter)

But as explained in Para. 3.2.5.12 an extensive FlightLoad Survey has to be done to verify the load limitingprocedure of the FCS and to proof the reduction of theultimate load factor.

For FCS independent loads (e.g. landing gear loads,Hammershock pressures, etc.) the ultimate load factorwill still be 1.5.

For further information about the ultimate load factor seeChapter 3.1.3 – Safety Factor Review.

3.2.5.15 Conclusion

The calculation of aircraft loads under consideration ofFlight Parameter Envelopes is useful and practicable formodern high performance fighter aircraft with a carefreehandling and load limiting FCS.

As demonstrated for the Eurofighter:

- the integrated design of FCS and aircraft structure ispossible

- the carefree handling and load limiting procedure ofthe FCS is working

- the defined design loads by using the FlightParameter Envelopes are acceptable and leading to arobust but not to conservative design of the aircraftstructure - compared to the loads evaluated with theFCS (time dependent flight load simulations) lateron in the A/C- Clearance Phase the design loads arewell

- the reduction of the ultimate load factor from f-ult =1.5 to f-ult = 1.4 based on the FCS- load limitingfunction is useful and leads to a lighter aircraftstructure

On the other hand the enormous increase in systemcomplexity for a modern high performance fighteraircraft with a carefree handling and load limiting FCSleads to extensive investigations:

- the flight control laws have to be reviewed during alldesign phases to check their function as a loadlimiting system

- the necessary careful and accurate loadinvestigations during all design phases are veryextensive

- an extensive Flight Load Survey has to be done forLoads Model validation and with it to proof the loadlimiting procedure of the FCS and additional ifnecessary to proof the reduction of the ultimate loadfactor

- the ALE concept has to be verified by detailed stressanalysis, static test and possible restrengthening ofthe aircraft structure

As explained above the permanent monitoring of thestructural design parameters as Flight ParameterEnvelopes, ALE’s, etc., is indispensable to minimize therisk of a non optimal structural design of the aircraft.

Therefore it should be emphasized once more thatvarious disciplines as Loads, Aeroelastics,Flightmechanics, Flight Control, Stress, Aerodynamics,Flight Test have to cooperate in a very close manner, theso called concurrent aircraft engineering.

52

3.3 Dynamic Loads

3.3.1 IntroductionThe intention of this chapter is to discuss the predictionof unsteady loads arising as a result of pilot actions (asopposed to atmospheric turbulence, say). Gusts andground loads are treated in separate chapters. Loads dueto buffet and buffeting, hammershock, gunfire and storeejection/release loads are mentioned. The aim is met bybriefly describing the background, prediction processesand calculation methods, and certification issues.Consideration of the latter is essential, even at the designstage. In addition, the likely way forward for this“technology” is noted. A table is provided as a guide forconsideration of dynamic loading sources and theireffects on an airframe.

In addition, examples of dynamic load analyses andtesting for validation purposes are given in section 3.4,whilst birdstrike is discussed in 3.5. The latter does notstrictly come within the terms of this chapter, but isclassified under ‘threats’. However, it is such asignificant source of aircraft in-service incidents, andhence a driver of future designs, that it is included here.

In the course of the item, reference is made to somespecific papers and work known to the author. However,it should be noted that hundreds of technical papersrelating to the overall subject are available world-wide.Since there are several approaches documented, thischapter does not make prescriptive statements regardingthe “correct” approach. Rather, readers are encouragedto adopt information and data applicable and appropriateto their own specific technical challenges. The aim is toraise awareness, not define methods in detail.

The airframe static load can be thought of as one thatchanges only with flight condition e.g. airspeed, angle ofincidence, altitude etc. For the purposes of this report,the airframe dynamic load component can be consideredto be the oscillating part of the load which has afrequency in the range 2 - 100Hz. This is not a hard andfast rule. However, loads oscillating below 2Hz can beconsidered to be due to 'rigid body' motion. Above100Hz, the load is unlikely to be adversely affecting amajor structural item, more likely to be a localized effecte.g. an acoustic, stores or equipment environmentaleffect.

There are many sources of dynamic loads on a militarycombat aircraft. Traditionally, combat aircraft were notdesigned and optimized to the degree that is expectedtoday. Dynamic effects were therefore included in theearly design phases of an aircraft project by applying afactor to the static design loads (which were usuallymaneuver defined for combat aircraft). The pessimismthat this introduced could be tolerated and covered themajority of dynamic loading effects. It was only whenstructural or equipment problems emerged during projectdevelopment, or even in-service, that dynamic loads wereconsidered in more detail. This situation wascompounded by an absence of advanced unsteadyresponse prediction tools.

The performance of modern military combat aircraft hasincreased, taking the airframe into situations where the

airflow over the structure becomes separated andoscillatory. The unsteady environment to which amodern airframe is subjected has therefore becomeincreasingly harsh. At the same time, a requirementexists to reduce the factors applied to the design loads todrive down structural mass. The need to predict theunsteady load component more accurately, to ensuresafety, has therefore become correspondingly moreimportant. To that end, modern military combat aircraftare designed to withstand the worst static and dynamicload cases which they are likely to encounter in-service.This has led to some regions of modern combat aircraftstructures being designed by dynamic load cases.

3.3.2 Types of Dynamically Acting Loads

3.3.2.1 Buzz

Buzz is a single degree of freedom flutter wherebylimited amplitude oscillations of surface panels or controlsurfaces occur due to a loss in aerodynamic damping andmay involve the local resonance of such surfaces. Thisloss is attributed to boundary layer and shock waveinduced instabilities in the surrounding flow field.Examples of such instabilities include oscillations ofshock waves over a control surface and separated flowcaused by an upstream shock wave.

Although the limited amplitudes of oscillation associatedwith buzz phenomena do not cause catastrophic structuralfailure, as can happen with a two (or more) degree offreedom flutter, structural fatigue can arise. Commonsolutions to reduce the adverse effects of buzzphenomena include manipulation of the flow field (e.g.using vortex generators) to reduce instabilities andstiffening of the control surface hinges to reduce freeplay.

3.3.2.2 Buffet and Buffeting

Buffet is an excitation caused by the separation of airflow over a surface. This can be separation in anunsteady manner causing excitation of the surface fromwhich it is separating, or separation from upstreamcomponents such that the resulting unsteady flowimpinges upon a downstream surface. This is worse athigh angles of attack. Buffeting is the associatedairframe structural response. Buffet and buffeting arephenomena that are unavoidable in highly maneuverablecombat aircraft.

For many years fighter aircraft have had to penetrate intothe buffeting region of the flight envelope in order to gainmaximum turn performance. With conventional controlsystems, the buffet onset was in many ways a usefulfeature because it provided the pilot with a clear warningthat he was approaching the limits of aircraftcontrollability. Increasing buffet penetration, for instanceby increasing angle of attack, is also accompanied byrelated characteristics such as wing-rock and nose slice.

With the advent of complex, active flight controlsystems, modern aircraft can remain controllable wellbeyond traditional boundaries, and even into post-stallconditions. This has implications upon structural designdue to the potentially greater time spent in unsteady flow

53

conditions (fatigue implications) and the large magnitudeof these unsteady loading actions (strength).Consequently, the ability to predict these flows hasassumed a far greater importance in aircraft design.

Another consequence of active flight control systems isthe potential for affecting the structural response underunsteady loading conditions. If the system interpretsstructural response as aircraft response and tries tocorrect it by driving the controls, then there is a potentialfor increasing the loads on the structure. This area ofexpertise is known as Aero-servo-elasticity (ASE) orStructural Coupling. A well-designed flight controlsystem (FCS) will not exhibit such adversecharacteristics. It is not a design driver when assessingloads, but an awareness of the total system (aircraft +FCS) characteristics is required for flight clearance work.

Ways of using active control for reducing structuralresponse to unsteady loading, like buffet, are underconsideration. A view of this is given in reference 1.

The above is applicable to combat aircraft. However,buffet also occurs due to impingement of vortical andwake flow on downstream surfaces, separated flow overcontrol surfaces, and flow interaction between adjacentstores (or engines), their pylons and other airframestructure, to name a few generic examples. These are notrestricted to highly maneuverable aircraft. Indeed,straight and level flight at transonic conditions, on anyclass of aircraft, can lead to complex shock-boundarylayer interactions, which induce separated flow and hencebuffet, i.e. a forced response.

Further ’buffet inducers’ include excrescence andcavities. Examples of the former include blade aerials,chaff/flare dispensers, auxiliary cooling system intakesand exhausts. Flow separation occurs from these unlessthey are carefully designed, and faired-in specifically toavoid this phenomenon. The result is unsteady pressurefluctuations on surrounding, external paneling andsurfaces. The risk here is that surface panel modalfrequencies can be excited which can lead to rapidfatiguing of the affected structure.

Flow spillage from cavities can have similar effects. Thecavities can be those occurring when the landing gear isdeployed, or when internally carried weapons arereleased. The latter is likely to be much more of aproblem due to the wider range of flight conditions atwhich it may occur.

Further, there is much potential for adversely affectingthe internal and back-up structure of the weapons bay dueto acoustic effects. Similarly, stores and equipmentinstalled in the bay will have difficult environmental

clearance issues to overcome. Control of such acousticenvironments is a major study area.

3.3.2.3 Hammershock

Hammershock (H/S) is an event whereby an aircraftengine surges, sending a pressure pulse upstream,opposing the direction of airflow that would exist duringnormal engine operation. This results in a loss of engineperformance, the possibility of a flame-out and/orpermanent engine damage.

H/S events can occur anywhere within a combat aircraftflight envelope but are more significant at the envelopeextremities. They have many causes. These include:

• over-fuelling;• bird strike;• foreign object ingestion and• disturbed intake airflow (e.g. wake ingestion).

A single surge may occur or a series of pressure pulsesmay be generated if the surge becomes 'locked-in' i.e.conditions are such that repeated surges occur.

The pressure pulse created impinges on the engine intakeand on the forward fuselage. Both of these items musthave sufficient strength to withstand a H/S event. This isparticularly critical for aircraft which have foreplaneslocated in the path of the pulse. The concern here is thata locked-in surge may occur with a pulse frequency closeto a fundamental foreplane vibration mode. If an item ofstructure is excited at a frequency near one of its naturalvibration modes (i.e. a resonant frequency), the resultingamplitudes of vibration and hence load are large.

Realistic prediction of the excitation can be achieved bydeliberately surging an engine on the ground andmeasuring the resultant pressure pulse amplitudes in theintake duct, splitter plate/lip regions and forward of theintake. Account can then be taken of airspeed, altitudeetc. to derive excitation throughout the desired flightenvelope. Wind tunnel testing is an alternative approach,but scale effects are significant, and can lead to majorover-prediction if not accounted for adequately.H/S was considered during the development of EAP(shown in Figure 1). This resulted in the foreplanes beingmodified to prevent them 'tuning' with the predicted pulseH/S frequency. This proved to be overly cautious. Theactual pressure pulses dissipated more quickly than wasanticipated or had been measured in the wind-tunnel.This experience, of course, can be used on future aircraftprojects.

54

Figure 1 : EAP Technology Demonstrator

3.3.2.3.1 Influence on inlet duct design

Examples of load cases on the inlet duct includemaneuver ‘g’-loads, steady state pressures andhydrostatic pressures of neighboring fuel tanks. However,the pressure loads acting on the inlet duct caused by thepropagation of the high velocity pressure wave(s)associated with surge phenomena is the predominantdesign factor for combat aircraft.

The majority of modern combat aircraft utilizerectangular, or other non-circular, shaped inlets with agradual longitudinal change into a circular shape duct inorder to merge effectively with the engine face. The H/Sloads become critical for such variable duct geometry dueto complex load paths in the throat region and stressdistributions around the corners of, say, a rectangularinlet. The H/S loads associated with the circular ductsections produce hoop tension and are less critical.

From reference 2, two aspects of H/S phenomena whichare of importance to the dynamic response of the intakeduct structure are (i) magnitude of the pressure wave and(ii) the rise time to positive and negative peaks. It shouldbe noted that the negative peak is caused by the reflected

H/S pressure wave at the forward intake. Figure 2 showsa typical example of a H/S excitation time history inwhich the vertical axis represents the ratio of incrementalH/S pressure to maximum incremental H/S pressure andthe horizontal axis corresponds to the H/S pulse duration(τ).

The characteristics of H/S loading as described aboveleads to the consideration of dynamic magnification ofloads during duct design, especially when taking intoaccount of ‘locked in’ surges. This is due to the potentialof a pulse sequence having repetition frequencies whichcould coincide with the natural frequencies of the ductpaneling.

Conventional approaches of designing ducts to cope withH/S loads include increasing duct skin thickness andemploying additional ring stiffeners around the duct inbetween the frames. Furthermore, special attention ismade to the local design of frames and stiffeners in therectangular sections of the duct as well as axial fastenerand bond peel strengths which could result in localizedstructural strengthening. Approaches such as these serveto increase duct weight: an undesirable trend.

Rectangular inlet Circular duct at engine face

55

Figure 2 Characteristics of Hammershock loading

Another aspect of duct design in relation to H/Sphenomena is the attenuation of pressure waves.Attenuation is key to the reduction of pressure loadsacting throughout the duct, particularly in critical areassuch as frontal inlet region. Two processes (detaileddiscussion provided in reference 3) which can relievepressures are (i) airflow bleed through a bypass exitwhich reduces diffuser volume and (ii) ramp edgeleakage to the plenum allowing pressure transmissions atsonic velocity. However, trade-off studies must beconducted to determine the feasibility of duct weightreduction due to the alleviation of pressure loads, againstthe losses in intake efficiency during operation of thebleed / leakage processes, and the weight increases due toimplementation of the more complex mechanismsinvolved.

3.3.2.4 Gunfire

This is an obvious source of high energy, short durationdynamic loading. Attention is traditionally given todesigning structure to absorb recoil forces transmitted toit, whether from an internal or pod-mounted installation.Conventional metallic structure, with its joints andfastenings, tends to absorb energy (via damping andfriction) better than extensively bonded designs. Hence,transmission of loading is limited. With bondedstructures the recoil effects can affect a much larger partof the airframe. This gives the potential for tuning withmodal frequencies, and hence loading problems.

Muzzle/exhaust blast could increase this effect iftransmitted through a significant part of the airframe. Itcould be possible for some parts to be loaded by both therecoil forces and the blast effects. Even if this is not thecase, the blast effects on localized external structureshould be assessed. Again, tuning with panel modalfrequencies is a possibility given the current range of

gunfire rates. From the blast impingement point of view,pod mounted guns are usually better. Almost bydefinition, they are mounted such that the gun muzzlewill be further away from the aircraft. This would beexpected to allow some dissipation of the blast energybefore hitting the nearest parts of the airframe.

3.3.2.5 Store Release / Jettison / Missile Firing

Stores release can vary from jettison of fuel tanks tomissile firing activities. Stores release design cases arefew and far between, but the possibility must beconsidered. The effects of store release during extrememaneuvers must be assessed.

Excitation of the airframe arises from the 'kick' providedby the loss of mass during release, this effect beingdirectly in line with the mass of the store, and also fromthe ejector release units which push the store away fromthe aircraft. Unlike buffet, gunblast and H/S excitation,the point of application of a release ‘impulse’ to thestructure is more localized. However, the effect can bejust as global if significant transmission through theairframe is possible, as discussed in the previous sectionon gunfire.

Special design consideration must be given to 'ripple'store releases i.e. multiple stores released in rapidsuccession. This may be required to give a widemunitions coverage of the target or as part of anemergency stores jettison sequence. As with H/S events,the proximity of release 'pulses' could have an excitationfrequency close to a major airframe vibration mode. Theresult would be large structural oscillations. This implieslarge structural loads but would also affect 'dumb' storedelivery accuracy.

∆∆∆∆PHS / ∆∆∆∆PMAX HS

Time

1.00

0.00

-0.40

½ ττττ ττττ

56

3.3.3 Prediction Process & Methods

3.3.3.1 Loads Prediction and Simulation

The main emphasis here is about primary lifting surfacesundergoing general bending and torsional responses dueto a dynamic loading action, eg. buffet excitation.Localized loads use similar principles, but may not needa full aero-structural simulation. This depends upon theneeds of the technical problem being addressed.

There are 2 major approaches. The first is empirical, andassumes that the new design is similar in general natureto a previous project for which there exists an adequatedatabase of information.

The second approach can be classed as the theoreticalapproach although it does not yield an exact solution; theaccuracy being dependant upon the quality of the inputdata, and the inherent assumptions regarding linearity ofcharacteristics.

3.3.3.1.1 Empirical ApproachAn example of a successful use of an empirical approachis that of designing EAP to account for fin buffeting.Figure 3 illustrates how an initial prediction of structuralresponse can be carried out. From Tornado measuredcharacteristics, an estimate of EAP fin response wasmade. It assumes that the dominant parameters affectingthe fin response are wing sweep angle, incidence, anddynamic pressure.

Incidence (AoA)

EMPIRICISMFin vibration characteristics

TORNADO( =450)

TORNADO( =250)

EAP( =570)

FIGURE 3. Fin Vibration Characteristics

Actual numbers on the axes are removed to preserve theunclassified nature of this document. However, use ofthe original plot will lead to the response on the EAP finfor a given flight condition. Assuming a detailedknowledge of the fin structural characteristics, then theinternal structural loads can be derived. This wassuccessful because of the large amount of informationgenerated, and hence available, in the course of studyingfin buffeting on Tornado.

As stated before, there is a large amount of publiclyavailable information which could allow derivation ofempirical methods for other projects. The example givenwould not, of course, be applicable to twin fin designs, orif the new fin structure (and, hence, modal response) wasradically different.

3.3.3.1.2 Theoretical Approach

This approach requires a numerical model of the structure(inertia, damping and stiffness), numerical representationof the oscillatory aerodynamics (damping and stiffness)

and numerical representation of the forcing function (eg.buffet excitation).

The mathematical equation to be solved is of thefollowing form

Ax V Bx V Cx Dx Ex F tE E�� � � ( )+ + + + =σ 2

whereA = generalized inertia matrixB = generalized aerodynamic damping matrixC = generalized aerodynamic stiffness matrixD = generalized structural damping matrixE = generalized structural stiffness matrixVE = equivalent airspeedx = generalized co-ordinatesσ = relative air densityF(t)= generalized forcing function

Post-processing of the output from the response solutionleads to derivation of loads at defined points on thestructure. The process is shown diagrammatically infigure 4.

57

NASTRAN, or In-Company developed alternative, isused as the analytical tool for the calculation techniqueshown above.There are several points to note. In current practice, theunsteady aerodynamics and structural models are linearapproximations. Development of improved, advancedaerodynamic methods is discussed later. For early designinformation there is unlikely to be detailed structural andmass data available. In addition, the excitation functionmay well be derived from existing databases pendingavailability of wind tunnel test data.

For the detailed design and clearance phases of a projectthe response model is likely to be the same as that usedfor Flutter assessments. During the clearance phases of aproject, it should be possible to include a structural modelmatched to reflect GVT data. The excitation data willprobably be based on wind tunnel testing of the finalizedproject lines. However, it will still be subject to scalingfrom wind-tunnel to full scale, as well as normal wind

tunnel accuracies. This is for a rigid wind tunnel modeland is illustrated in figure 5.

An interesting, but less used variation of the above, is tocreate a dynamically scaled, flexible wind tunnel model.This involves scaling the full size structuralcharacteristics to the model, but does mean that thesurface forces and moments can be measured directly.There is still the problem of then re-scaling to full size inorder to derive the full scale loads.

The first approach is likely to be used earlier in thedesign cycle. Unless the new aircraft is a development ofan existing type, detailed structural information will notbe available for manufacture of the flexible wind tunnelmodel. The latter is also likely to be more expensivebecause, in addition to increased model manufacturingcosts, a dedicated set of test runs will be required. Therigid data can possibly be acquired on a ride-along basiswith other testing.

AlternativeAerodynamic

TheoreticalMethods

Buffet Exction

Figure 4 : Buffeting Response Calculation Process(Generalise from fig. 16 of Ref. 3)

Finite ElementModel Modal Vib.

Characteristics

Doublet-LatticeAerodynamics [Nastranbased or in-house]

Dynamic ResponseLoadsDerivation[Forces &Moments]

Figure 5 : Development of Buffet Excitation

Tunnel Test of RigidModel

Pressure Measurement Buffet ExcitationW/T to Full- SizeScaling

58

A useful guide to the ‘state-of-the-art’ for numericalaeroelastic simulation techniques is reference 4.

3.3.3.1.3 Hybrid W/T - CFD Techniques

Reference 5 is experimentally based and gives a goodsummary of the aerostructural buffet problem. As itpoints out, testing is expensive. Ideally, given theadvances in computing power in recent years, increasingmaturity of steady CFD techniques and acceleratinginterest in unsteady CFD, then it should be possible toreplace some of the wind tunnel testing essential toreference 5 and generally improve accuracy of theaerodynamic predictions.

Researchers are now beginning to develop theseapproaches. Until unsteady CFD techniques are moremature, a pragmatic approach is needed to allow theengineer (as opposed to the researcher) a means ofaddressing buffet and buffeting early in the designprocess. Hence, a combination of steady CFD analysiswith unsteady pressure measurements from wind tunneltesting is a realistic approach. There are still someproblems, most notably prediction of aerodynamicdamping levels during buffeting at higher incidences.

3.3.3.1.4 Superposition of Steady and UnsteadyLoading

The above treatment relates to derivation of the unsteadyexcitation. However, it is the total response, and henceloading, that we are interested in from the structuraldesign and clearance point of view.

An aircraft operating on the ground or in flightencounters two distinct types of loading - static anddynamic. Of course, the airframe structure itself cannotdistinguish between the two loads. It is subject to thecombination of them, the total load.

Design activities are affected by available predictiontools and techniques. It is common practice, for thepurposes of aircraft design and clearance activities, thatthe two ‘types’ of loads are calculated discretely. Theseare then combined to give total predicted load. Figure 6shows the principle diagrammatically.

It is important to ensure a coherent approach. There aredifferent ways of achieving the same result by assumingthat the principle of superposition holds (see tablebelow).

-3.5

-2.5

-1.5

-0.5

0.5

1.5

2.5

3.5

4.5

5.5

6.5

0

0.2

0.3

0.5

0.7

0.9

1.1

1.3

1.5

1.7

1.9

TIME

LO

AD

(kN

)

-3.5

-2.5

-1.5

-0.5

0.5

1.5

2.5

3.5

4.5

5.5

6.5

0

0.2

0.3

0.5

0.7

0.9

1.1

1.3

1.5

1.7

1.9

TIME

LO

AD

(kN

)

-3.5

-2.5

-1.5

-0.5

0.5

1.5

2.5

3.5

4.5

5.5

6.50

0.2

0.3

0.5

0.7

0.9

1.1

1.3

1.5

1.7

1.9

TIME

LO

AD

(kN

)DYNAMIC LOADS - A DEFINITION

Nominal frequency range 2 -100 Hz

Figure 6: Superposition of Steady and Unsteady Loads

Quasi-Steady Loads Simulation Methods Dynamics Simulation Methods1. Time varying throughout manoeuvre ie.

‘rigid body’ steady manoeuvre loadsIncremental loads due to unsteady effects on aflexible structure

2. Constant loads from starting point ofmanoeuvre

Incremental loads due to time varying ‘rigidbody’ motion + Incremental loads due tounsteady effects on a flexible structure

3. - Total loads due to time varying ‘rigid body’motion + loads due to unsteady effects on aflexible structure + FCS

Steady + Dynamic = Total

Nominal frequency range 2 - 100 Hz

59

These approaches are driven by pragmatic applications ofavailable methods and tools. It is a recognition that notall organizations have the latest available technology andcomputing power. Indeed, the third approach above isonly recently becoming more common as ‘tool sets’ anddesign processes become more integrated. For instance,formerly it might have been necessary to have separatemethods for development and analysis of structural,aerodynamic and FCS models. If consideration of other‘disciplines’ was necessary, each would probably modelthe others in its’ own home environment. This led to anumber of notionally similar numerical models beingdeveloped - each needing extensive quality assurance andchecking, and none of them fully compatible.As stated before, there is no definitive method. Readersmust judge the appropriate way forward for their ownparticular projects. However, it should be noted thatsome aspects of 1 and 2 above are favourable because thequasi-steady loads can be based upon more mature,speedier, theoretical methods (CFD) than unsteadyloading. In addition, for similar reasons there are likelyto be more extensive wind tunnel test data available.

3.3.4 Design Assumptions, Criteria andCertification

Reference 6, gives a very brief overview of importantdynamic loading phenomena that should be consideredduring the design of combat aircraft. It notes, however,that specific design and certification criteria/guidelinesare few.

This can lead to lengthy discussions with Customers andCertification Authorities about what should be addressedin design and certification of a given aircraft project.Experience has shown that an open-minded approach atthe design stage, which can include work that positivelyeliminates a phenomenon from consideration, will ensurea smoother progression, later in the project cycle, to flightclearance and qualification. In short, at present there areno hard rules governing consideration of dynamic loadingin structural design, other than that it should be taken intoaccount!

As engineers, we are bound to consider these loadingactions because they can be significant. This isillustrated by the technical papers covering fin and tailbuffeting on F-18, and similar aircraft, which arenumerous (e.g. references 5, 7, 8, 9 and 10 picked nearlyat random from a wide choice). Wing buffeting is a wellknown phenomenon, and also well documented. It isclear that buffeting must be examined in the early stagesof design for aircraft with significant maneuvercapability. The problem for other areas is deciding whatis an acceptably low risk for a given set of circumstances.Often, there are little data available which can beanalyzed effectively.

It is stressed that the reader must decide what isappropriate for his particular work. It must be clear whatthe latest design criteria are, and what is applicable to agiven project. If standards change through the life of anaircraft project, this can lead to a very complexdocumentation trail!

USE OF UNSTEADY CFD IN EXCITATIONPREDICTION

• Databases• Experimental W / T Flight

TRADITIONAL

• Steady CFD• W / T Unsteady Pressure Measurements

PRESENT

• Unsteady CFD For Magnitude and Frequency Content

( Excitation Response Structural Interaction)

FUTURE

Figure 7: Use of Unsteady CFD in Excitation Prediction

3.3.5 Developments

The above figure illustrates the changing approach to theuse of CFD in the prediction and simulation of dynamicloading phenomena. The overall thrust has been to beable to use CFD to replace/supplement wind tunnelmeasurements for prediction of buffet, and other,

unsteady excitation. In addition, use of CFD forimproved response aerodynamics (particularly damping )increasingly allows assessment of aerodynamically non-linear effects. Key to this capability on the response sideis the unsteady CFD/structural modeling interfacingmethods. This is available at research and academic

60

levels, but is not yet sufficiently robust or rapid forproduction application.

Reference 11 gives an outline of some work done in theUK to address the shorter term requirements of engineers.It reports on the combination of an extensive set of windtunnel tests with the aim of providing insight into theaerodynamic phenomena associated with novel wingplanforms. These planforms impact both steady andunsteady aerodynamics.

The wind tunnel tests have produced steady pressuredistributions, overall forces and moments, surface oilflow patterns and unsteady surface pressure frequencyspectra. The steady flow results have been comparedwith output from converged Reynolds Averaged Navier-Stokes (RANS) CFD solutions.

The work has enabled a design tool to be proposed foruse early in the design process. For an arbitrary wingplanform, at maneuvering conditions, steady CFD can beused to establish mean flow topology, including trackingof vortex shear layers. Empirical representations of thecharacteristic buffet frequencies can then identify thedominant frequencies of the dynamic loads. Whencoupled with relatively simple finite element models,predictions of buffeting response are expected to besufficiently accurate to enable meaningful evaluation andcomparison of different wing planforms.

3.3.6 SummaryThe above discussions are aimed at raising awareness ofdynamic loading effects, and their prediction, which isadvisable to consider at the design stage of an aircraftproject. Historically, this has not been so prevalent, butis necessary now due to the requirements to moreeffectively optimize structures, from both a strength andfatigue point of view. Indeed, active control of structuralresponse (due to buffeting, say) is under very energeticresearch and must now also be considered as a possibleoption at the design stage of an aircraft project.

Because of the immense breadth of the subject, there areno definitive statements here. Readers are required toformulate their own approach to their own particulartechnical challenges.

It is apparent that wind tunnel and CFD methods are vitalto future prediction techniques, particularly of non-linearaerodynamic effects. However, examination of non-linear structural effects (e.g. control surface backlashcharacteristics) as part of the overall aero-structuralsystem are dependant upon more robust and rapidtechniques for coupling CFD with a FEM than areavailable at present.

The table below is intended as an aide memoir. Itsummarizes different types of dynamic loading andwhich parts of an aircraft they affect. It includes gustsand ground operations for completeness, although theseare described in different chapters.

61

SOURCE OFLOADING

COMPONENTS AFFECTED TYPES OF AIRCRAFT /COMMENTS

ATMOSPHERICTURBULENCE / GUSTS

WINGFORE / TAIL PLANEFINFUSELAGECREWEQUIPMENTSTORES & PYLONSSENSORS & PROBES

HIGH SPEED AIRCRAFT WITHRELATIVELY LOW WINGLOADING

BUFFET / BUFFETING /BUZZ

WINGFORE / TAIL PLANEFINSTORES & PYLONSLOCALISED EFFECTSeg. Excrescences Panels Sensors & Probes Airbrake

ALL TYPES, BUTPARTICULARLY THOSE WITHSIGNIFICANT AoA ANDMANOEUVRING CAPABILITY

Bluff shaped excrescences mountedon large panels

STORES RELEASE &JETTISON

WINGFUSELAGEPYLONSATTACHMENTS &BACK-UP STRUCTURE

ALL TYPES

MISSILE FIRING As above +PLUME EFFECTS on Local panels Control surfaces Tailplane etc.

ALL TYPES

HAMMERSHOCK INTAKE & DUCTFOREPLANESFRONT FUSELAGESENSORS & PROBES

CANARD CONFIGURATIONSWITH CHIN INTAKES AFT OFFOREPLANES

GROUNDOPERATIONS

WINGFORE / TAIL PLANEFINFUSELAGECREWEQUIPMENTSTORES & PYLONSSENSORS & PROBES

ALL TYPES BUT WORSE FORCARRIER-BORNE & VSTOL

Any extreme action that can beachieved by the pilot

BIRDSTRIKE NOSE CONECOCKPIT / TRANSPARENCYFOREPLANEWING LEADING EDGEINLET FACEPlus any other forward facing sectionsof the airframe

ALL TYPES

Other hazards include airborne andground debris

3.3.7 AcknowledgementsThanks are due for the assistance of Mr. S Samarasekera,BAE SYSTEMS Aerodynamic Technology , and to Mr.C Bingham, BAE SYSTEMS Structural Technology.

3.3.8 References

1. PAPER PRESENTED AT RTO CONFERENCE –OTTAWA OCT 1999NASA LANGLEY RESEARCH CENTER’SContributions to international active buffetalleviation programs,R. W. MOSES, OCTOBER 1999

62

2. AGARD-R-815THE IMPACT OF DYNAMIC LOADS ON THEDESIGN OF MILITARY AIRCRAFTPapers presented at 83rd Meeting of the AGARDStructures and Materials Panel, held in Florence,Italy, 4-5 September 1996 Published February 1997

3. REVIEW OF HAMMERSHOCK PRESSURES INAIRCRAFT INLETSL C YOUNG and W D BEAULIEURockwell International, Los Angles, CaliforniaJANUARY 1975

4. AGARD-R-822Numerical Unsteady Aerodynamic and AeroelasticSimulationPapers presented at Workshop in Aalborg, Denmark,OCTOBER 1997 Published March 1998

5. AGARD-CP-483 paper 11PREDICTIONS OF F-111 TACT AIRCRAFTBUFFET RESPONSEAM CUNNINGHAM jr., CF COEAPRIL 1990

6. AGARD-R-815 paper 9DYNAMIC LOADING CONSIDERATIONS INDESIGN OF MODERN COMBAT AIRCRAFTR CHAPMAN SEPTEMBER 1996

7. AGARD-CP-483 paper 2A UNIFIED APPROACH TO BUFFETRESPONSE OF FIGHTER AIRCRAFTEMPENNAGEMA FERMAN et al APRIL 1990

8. AIAA paper 91-1049SOME BUFFET RESPONSECHARACTERISTICS OF A TWIN-VERTICAL-TAIL CONFIGURATIONSW MOSS et al APRIL 1991

9. AIAA paper 92-2127BUFFET LOAD MEASUREMENTS ON AN F/A-18 VERTICAL FIN AT HIGH-ANGLE-OF-ATTACKBHK LEE, FC TANG JANUARY 1992

10. AGARD-R-815 paper 6A COMPARISON OF PRESSUREMEASUREMENTS BETWEEN A FULL-SCALEAND 1/6 SCALE F/A-18 TWIN TAIL DURINGBUFFETRW MOSES, E PENDLETONSEPTEMBER 1996

11. BATH UNIVERSITY, UK Ph.D. ThesisAN INVESTIGATION OF BUFFET OVER LOWOBSERVABLE PLANFORMSM I WOODS 1999

3.4 Managing the Technical Risk – DynamicLoads in-flight Monitoring

The principle adopted throughout design and clearance ofcombat aircraft with respect to dynamic loads is one ofcaution, due to the known deficiencies in predictiontechniques. Each design could be over-engineered andevery clearance might be unduly restrictive if theapproximations remain un-quantified. To try to minimize

this risk, dynamic loading predictions are validatedagainst flight test measurements during envelopeexpansion flying within the development phase of theproject.

The flight test envelope expansion process for moderncombat aircraft is a rapid one. To be able to keep pacewith this programme whilst ensuring that in-flightdynamic loads are on the safe side of predictions, a highlevel of visibility of aircraft response amplitudes andtrends is required. In addition, for really rapid turn-around and test-conduct these data need to be presentedto the monitoring engineer in real time. In this way,should response trends appear to be worse or responseamplitudes greater than predictions, the testing can behalted, or modified, before safety is compromised.Further, due to the data visibility, in-depth evaluation ofany discrepancies can then be carried out post-flight moreeffectively.

Real-time unsteady response monitoring is achieved atBAE Szstems, Warton, via the 'Dynamic LoadsMonitoring System'. The low cost system described here,commissioned at BAE Systems, Warton, has been usedfor the EF2000 Project. It is currently undergoingmodernization.

3.4.1 Dynamic Loads Monitoring System

The Dynamic Loads Monitoring System comprises aseries of pen recorders which display up to 24 real-timeacceleration time-histories for various defined locationson the aircraft. Figure 1 shows a typical instrumentationlayout for vibration monitoring on a military aircraft(EAP). In addition, a VAX-based, in-house developedsoftware package displays the following in real-time:

• fin acceleration/dynamic pressure at a definedfin location vs. incidence angle. These data arecompared with a predicted fin buffet trendwhich takes into account, if required, airbrakeoperation;

• fin acceleration at a defined location vs.incidence angle. These data can be comparedwith a user-defined maximum allowableacceleration;

• wing accelerations for up to 3 defined winglocations. These data are compared with user-defined maximum allowable accelerations;

• wing acceleration/dynamic pressure at adefined wing location vs. incidence angle.These data are compared with a predicted wingbuffeting trend.

A typical example of the software output is shown infigure 2.

It is worth noting at this stage that airframe loads aremonitored, by implication, via acceleration levels i.e. it isassumed that, if unsteady acceleration predictions areconsistent with measurements, then the airframe dynamicloads will also match predictions. Two outputs aretherefore required from the load prediction modelsmentioned earlier. The first, for design and clearancepurposes, is actual loading information. The second, forloads monitoring purposes, is acceleration response data.

63

Strain-gauges could be used to measure load 'directly'.There are, however, a number of problems associatedwith their use, namely:

• suitable calibrations being available to convertgauge signal to load;

• reliability of the gauges and the signals thatthey produce;

• strain gauge signals vary with temperature;

• the gauge is measuring structural load in ahighly localized area, making prediction moredifficult to do accurately. Measuredaccelerations give a more global picture ofstructural response.

3.4.2 Dynamic Loading PhenomenaMonitored

In an ideal world, the dynamic loads engineer would beable to monitor all regions of an aircraft for all types ofunsteady phenomenon. This would, of course, bring withit the problem of how to display such a volume of data ina usable form. Unfortunately (or fortunately), there is alimit to the amount of instrumentation which can be fittedto a given test aircraft. Priorities must be decided as towhich dynamic loading effects are to be monitored, butnever to the detriment of flight safety. This decision maybe made easier if loading predictions for a given effectare small compared to available structural strength andcan therefore be safely disregarded.

The monitoring system at Warton is used to assess thedynamic response induced by:

• gust loading and flutter test induced dynamicloads via acceleration time-histories displayedon the pen recorders;

• fin and wing buffet loads via accelerationamplitudes and trends with incidence angle,displayed using the VAX-based monitoringsoftware.

3.4.3 Dynamic Loads Monitoring SystemImplementation

Figure 3 shows how the Dynamic Loads MonitoringSystem is implemented at Warton.

Accelerometer data from various locations on theairframe is transmitted to the Monitoring System (via aGround Station) at a rate of 512 samples per second.Using the Nyquist Theorem, this allows the monitoringengineer to observe vibration response having amaximum theoretical frequency of 256Hz. Thisfrequency range is sufficient for the dynamic phenomenabeing monitored, as defined earlier. In addition, aselection of aircraft data (Mach no., incidence angle,dynamic pressure and time) are transmitted to the systemat 32 samples per second.

The (digital) accelerometer data to be displayed using thepen recorders is converted to an analogue signal and isplotted throughout the flight. This provides a useful dataquality check in addition to displaying responseamplitudes. The pens used for this have a transfer

function such that signals with frequencies up to around80Hz are not attenuated.

The VAX-based software component of the monitoringsystem is only used for certain flight test points - thosewhere significant wing and/or fin buffet is likely to occure.g. wind-up turn maneuvers. The fin and wing buffetaccelerometer data are conditioned as follows:

• high and low-pass filtered to remove any DCsignal component and to include only theresponse frequencies of interest. This islimited to only those frequencies associatedwith the first few fundamental aircraft vibrationmodes (the modes most likely to causestructural damage in the case of buffetmonitoring).

• data 'drop-outs' are checked for and any data'spikes' are suppressed.

Buffet analysis is initiated and terminated by themonitoring engineer. Conditioned data is captured by thesystem over one second and the requisite analysisperformed to obtain zero-to-peak acceleration levels andzero-to-peak acceleration levels normalized by dynamicpressure. These data are then plotted to the monitorscreen (vs. incidence angle where applicable) using thelower rate aircraft data. This process is repeated until thesystem is commanded to stop. The plot presented to theuser is therefore continually updated as a given maneuverprogresses. This process is summarized in figure 4.

The data acquired during monitoring are saved to disk forpost-flight analysis, if required.

Figure 5 shows an example of the wing buffet dataavailable to a monitoring engineer during a wind-up-turn(WUT) maneuver. The acceleration time-history for awing parameter is shown (W3). It can be seen that as theWUT progresses, the vibration amplitude increases andthen attenuates as the turn is completed and straight andlevel flight resumed. Peak acceleration amplitudes forthis and two other accelerometers (W1, W2 and W3) areplotted for comparison with user-defined maximumallowable vibration levels at 1 second intervals. Inaddition, the trend of peak g/dynamic pressure is plottedagainst incidence angle for comparison with the predictedtrend.

Figure 5 shows that whilst an acceleration time-history isuseful as a data quality check, the software basedmonitoring system provides a quick way of verifying thatthe dynamic loading on the aircraft is within prescribedlimits. Simplification of the loads monitoring task iswelcome in the high-pressure flight test environment.

Figure 5 shows that, for this test point at least:

• wing buffet trend predictions are well matchedby flight measurements and

• amplitudes of vibration at the wingaccelerometer locations are well withinallowable limits.

As such, with respect to buffeting response, this test hasbeen flown safely. It should be noted that these resultsare for a single test point. To form any sensibleconclusions about the predictive techniques used, a moreextensive survey of results would have to be performed.

64

FIGURE 1 - Typical Accelerometer Layout on Military Aircraft (EAP)

Accelerometer Locations

65

F IG U R E 2 - M onitor ing S ystem E xam ple D ata P lots

PEAK G (%)

PEAK G (%)PEAK G / DYNAMIC PRESSURE

PEAK G / DYNAMIC PRESSURE

W 1

W 2

W 3

FIN

BU

FF

ET

WIN

G B

UF

FE

T

Pre

dic

ted

Tre

nd

Pre

dic

ted

Tre

nd

M 0

.5

M 0

.9

NC

IDE

NC

E A

NG

LE (

degs

)

NC

IDE

NC

E A

NG

LE (

degs

)

NC

IDE

NC

E A

NG

LE (

degs

)

66

FIGURE 3 - Dynamic Loads Monitoring System General Layout

Dat

a T

rans

mis

sion

Flig

ht T

est

Gro

und

Sta

tion

AC

CE

LER

OM

ET

ER

DA

TA

(51

2 S

/S)

Acc

eler

omet

ers

FLI

GH

T D

AT

A(3

2 S

/S)

Inci

denc

eM

ach

No.

Dyn

amic

Pre

ssur

eT

ime

PE

N R

EC

OR

DE

RS

SIG

NA

L C

ON

DIT

ION

ING

Filt

erin

gS

pike

Sup

pres

sion

Dro

p-O

ut C

heck

s

DIG

ITA

L T

O A

NA

LOG

UE

CO

NV

ER

SIO

N

DY

NA

MIC

LO

AD

S M

ON

ITO

RIN

G S

YS

TE

M

VA

X

SA

VE

67

FIGURE 4 - Calculation of Trends With Aircraft Incidence

AN

AL

YS

IS

AN

AL

YS

IS

AN

AL

YS

IS

AN

AL

YS

IS

AN

AL

YS

IS

AN

AL

YS

IS

AN

AL

YS

IS

AN

AL

YS

IS

AN

AL

YS

IS

cel. (g)

cidence,(degs)

ynamicressure, (kPa)

USER DEFINEDACQUISITIONSTART

TIME

gpeak gpeak gpeak

αmean αmean αmean

qmean qmean qmean

PLOTgpeak/qmean vs αmean&gpeak vs αmean&gpeak vs wing transducer

UPDATE PLOT UPDATE PLOT

1 Second

68

FIGURE 5 - Monitoring System Example Wing Buffeting Output

W3

TIM

E H

IST

OR

Y

Pre

dict

ed T

rend

M 0

.5

M 0

.9

G PEAK (%)

G PEAK / DYNAMIC PRESSURE

INC

IDE

NC

E A

NG

LE (

degs

)W 1

W 2W 3

Allo

wab

le V

ibra

tion

Am

plitu

de L

imits

3.5 Airframe Certification Against BirdstrikeThreats

The phenomena of birdstrikes requires seriousassessment during the design stages of an aircraft. Overthe last decade there has been an increase in fatalaccidents due to birdstrikes on military aircraft.Furthermore, it is the single greatest cause of militaryaircraft loss in peace time.

To certify the airframe against birdstrikes, resistance torepresentative impulse loads acting on all leading edgeand forward facing sections of the airframe must beconsidered early in the design phase. The design workwould involve predictions of stress levels associated withsuch loads in both the skin and sub–structure of thefrontal airframe region. To prevent stress levelsexceeding the allowable limit, high strain rateperformance, yield strength and fracture toughness maybe critical factors in determining material selection.

Furthermore, past testing has revealed that structuralcomponents with sharp leading edges (i.e. leading edgeradius less than bird diameter) leads to a significantincrease in the impact velocity required to causestructural damage, due to higher local stiffness levelsinherent in smaller radii. Therefore, design specificationsof leading edges for forward facing regions of theairframe can be influenced by birdstrike phenomena, inaddition to aerodynamic, structural and manufacturingaspects.

3.5.1 Certification via Empirical Testing

Chapter 209 of Ref. 1 specifies the minimumrequirements for the resistance of airframes to damagecaused by birdstrike ;

• A 1kg bird with an impact velocity of 480 knotsmust not penetrate the structure.

69

• A 1kg bird with an impact velocity of 366 knotsmust not cause structural damage.

The latter specification reflects the need to reduce thecost of repair after lower kinematics energy impacts.Currently, meeting this specification is an expensive andtime consuming procedure, primarily due to modelmanufacture and test set up costs.

The standard approach is to fire real (dead) birds usingcompressed air in a gas gun. The birds are fired atvarying projectile velocities (up to high subsonic MachNo.’s) onto the frontal area of the airframe, i.e. nosecone, transparency, intake lips, foreplane, wing leadingedges etc. Testing considers birdstrikes head on to theairframe and angles up to 15 - 17 degree azimuth fromthe nose direction. Maximum deflections of the structureare recorded and the impacted structure is inspected fordamage and evidence of penetration. This data may besupported by strain gauge information, high speedphotography and deflection time history data from lasermeasuring devices. Due to the difficulties involved infiring real birds, the inherent variability in the birdstructure, the difficulty in controlling the centre of gravitylocation and the bird orientation, tests are notoriouslyprone to high levels of variability.

Empirical design rules are available for metallicstructures however equivalent methods are not availablefor composites making the potential role of analysis moreimportant. A single test that fails the structure may notprovide much information for a successful redesign to beproduced, particularly in the light of other designconsiderations that may apply.

Current DevelopmentsIn an attempt to alleviate costs involved with standardbirdstrike testing, one approach that has been accepted inthe civil aerospace industry is to certify aircraft againstbirdstrikes using ‘generic analysis’ (Ref. 2). However, itmay be some time yet before military aircraft would beallowed to be certified in this way.

The idea behind the generic analysis approach is that ifyou have designed and tested a similar component before,and if the analytical method has proven accuracy,clearance of a new ‘generic’ component can be achievedby analysis alone. Generic analysis requirescomprehensive understanding of mechanical propertiesand failure modes of the airframe structure and birdbehaviour under impact. Bird impacts above a certainvelocity threshold has been shown to be essentiallyfluidic. The modelling of an event which incorporatesboth fluidic and structural behaviour, with strong

interaction, presents significant challenges to theavailable codes and analysis techniques.

Coupled Euler-Lagrange and ‘smooth particlehydrodynamic’ codes are now being developed that willsignificantly improve the modelling capability in thefuture. Current analytical techniques attempt to representthe bird behaviour in the best possible manor in aLagrangian approach.

The failure behaviour of structures under high velocityimpact and the representation of these events in the codesis also subject to on going research and development.This is particularly significant in the area of compositematerials where there are many complex failure modesand particular problems in including these effects into thecodes.

To address these issues and improve the analyticalcapability several working groups and research activitieshave been set up in industry. These include programs thathave established bird biometrics and flocking behaviour,investigated the use of more consistent artificial birds,investigated the high rate failure behaviour of compositesand assessed the on-going developments in the availablecodes.

The results of one (FE based) birdstrike prediction tool isshown in Figure 6 below. The figure shows a strain mapof a leading edge after impact and allows directcomparisons with strains measured from experiment.

Upon extensive validation of birdstrike FE predictiontools, some form of certification of airframes againstbirdstrikes by analysis could become feasible, although itis envisaged that empirical testing will never be fullyeradicated from a combat aircraft’s developmentalprogramme.

3.5.2 References

1. DEFENCE – STANDARD 00-970 MINISTRY OFDEFENCEDESIGN AND AIRWORTHINESSREQUIREMENTS FOR SERVICE AIRCRAFT VOLUME 1 – AEROPLANES, BOOK 1

2. S351 IMechE SEMINAR PAPER 1DEVELOPMENT OF A BIRDSTRIKECLEARANCE PHILOSOPHYC H EDGEPublished in ‘Foreign Object Impact and EnergyAbsorbing Structure’ MARCH 1998

70

Figure 6: Birdstrike FE-Prediction

4 Gust loads

4.1 Introduction

Aircraft are often subjected to abrupt movements of air inthe form of turbulence or gusts. These gusts can imposeconsiderable loads on aircraft. Gusts may come from alldirections. Vertical gusts load the wing, fuselage andhorizontal tail. In the case of horizontal gusts wedistinguish lateral or “side” gusts, loading the fuselage,vertical tail and pylons and longitudinal or “head-on”gusts which may cause important loads on flap structure.

For transport type aircraft, gust load cases are the mostcritical for strength design, and gust loads are the mainfatigue loading source for the major part of the structure.Combat type aircraft structures are generally manoeuvreload critical, but for specific parts of the structure likethin outer wing sections and pylons, gusts may determinecritical design load cases1. Since the recognition thatturbulence produced significant loads (around 1915) gustdesign criteria have been formulated, which have evolvedover the years and are still under development2,3.

All major current Airworthiness Codes include two setsof gust criteria, based on a “Discrete Gust” concept and a“Continuous Gust” concept. In the following, the mainaspects of these two concepts will be briefly explained.

4.1.1 Discrete Gusts

The basic loading mechanism of gusts is schematicallyillustrated in fig 4.1. An aircraft flying with speed Ventering an upward gust with velocity w experiences asudden change in angle of attack ∆α=w/V. This givesrise to an additional air load

V

wSCV

2

1L L

2

αρ=∆

It will be clear, however, that the abrupt or “sharp-edged”gust indicated in figure 4.1 is physically impossible; itimplies an instantaneous change in lift and a real gustmust have some distance over which its effect builds up.Additionally, due to so-called aerodynamic inertia, asudden change in angle of attack does not immediatelyresult in a proportional change in lift. Hence, the load feltby the structure is modified by this effect. The resultingload depends upon the size and shape of the gust and theresponse characteristics of the aircraft. Different“Discrete Gust” shapes have been assumed in gustcriteria, ranging from the simple “sharp-edged “ shapeshown in figure 4.1 (in the early twenties), through the“ramp type” gust used in e.g. the former BCARRequirements to the “1-cos” gust shape included inalmost all current airworthiness codes.

71

Essentially, the Discrete Gust Criterion consists of a“design gust” of specified shape and magnitude Uds

(which is a function of altitude). The design value Ydes ofany load quantity y is to be found by calculating the timeresponse y(t) to the gust, and taking the maximum of y(t)as Ydes. For many years, the main Airworthiness Codesincluded simplifying assumptions with regard to thelength of the gust ( e.g. a (1-cos)-gust of 25 wing cords)and allowed the assumption of an aircraft response inplunge only (“in the absence of a more rationalinvestigation”), resulting in very simple gust-responseexpressions as given e.g. by the well known “PrattFormula”3.

With the growing size and increasing flexibility ofaircraft these assumptions became more and moreunacceptable. Hence, the major Airworthiness Codescurrently demand for a full dynamic response calculation,including all rigid and all relevant elastic modes. As thelength of the gust has a direct effect on the structuralresponse, a range of gust lengths has to be considered.The one giving the highest design load (the “TunedDiscrete Gust”) must be assumed, up to a defined level ofseverity e.g. the minimum gust distance is specified.

4.1.2 Continuous Gusts

The discrete gust concept assumes an atmosphere whereseparate and independent “gust bumps” occur that mayhit the aircraft. Measurements in gusty conditions,however, revealed a pattern more resembling a process ofcontinuous turbulence. This notion led in the earlysixties to the development of a completely new gustconcept and a set of additional Design Criteria, known asthe “Continuous Turbulence Concept” and the PSD(Power Spectral Density) Gust Design Criteria.

In this concept, the loading action is described as acontinuous process of random turbulence. Over shorterperiods of time this process may be considered asstationary with Gaussian properties and standarddeviation σw. In the longer term, the standard deviationor gust intensity is not a constant, but varies randomlywith a given probability function. The turbulence ischaracterized by the “von Karman” type Power SpectralDensity function, describing how the energy in theprocess is distributed with frequency.

On the basis of this turbulence concept, two designmethods were developed referred to as the “MissionAnalysis” and the “Design Envelope” Concepts5. The“Mission Analysis Concept”, which is of a purelystatistical nature, has the virtue of elegance. It is,however, difficult to apply and may lead tounconservative predictions if the actual “Mission Profile”of an aircraft changes and starts to deviate from designassumptions. Hence, the criterion is seldom applied and itis expected that in the near future it will be deleted fromthe Airworthiness Codes.

The “Design Envelope” criterion shows a resemblance tothe Discrete Gust Criterion in that it also specifies a“design gust strength” Uσ as a function of altitude thedesign value Ydes of any load quantity y is found from

σ= U*AY rdes

The response parameter rA , which is actually theratio of the standard deviations of the load output y andgust input w in stationary Gaussian turbulence, may beconsidered as defining an “average weighted” response;

rA is calculated by integrating the product of loadtransfer function squared and the turbulence PSD

function over all gust frequencies. Thus, rA definesessentially an “average response”, taking into account forwhich frequencies the load is sensitive (as defined by thetransfer function) and also which gust frequencies (or“gust lengths”) occur in the atmosphere.

Comparing now the PSD- gust criterion and the Discretegust criterion, we notice the difference and the reasonwhy both criteria are included in our design procedures.The PSD criterion is based on a rational and consistentmodel of the atmospheric turbulence; it defines designloads that are based on an average response, consideringall possible gust lengths that prevail in randomturbulence. The Discrete Gust Criterion is typically a“worst case” criterion; the highest load resulting from adiscrete bump with most adverse length must be taken.The Discrete Gust cases are included and maintained inairworthiness codes to safeguard against sudden more orless “stand alone” gust outbursts that have been observedto occur in practice.

4.1.3 Gust Load Requirements

Gust load requirements have been, and are subject to, aprocess of continuous change due to the experiencegained from previous aircraft, changes in aircraft designphilosophy and advances in analysis techniques. Section4.2 gives an overview of the gust requirements in theprincipal civil and military requirements prevailing today.The military requirements tend to lag behind compared toFAR/JAR 25, due to a lack of available flight data as wellas the lower criticality of gust loads for military aircraft.In FAR 25 and JAR 25, major changes have beenincluded over the last few years with regard to thediscrete gust cases and a major change of the continuousgust criteria is in preparation. A relevant part of theassociated NPRM (Notice on Proposed Rule Making) isincluded in Paragraph 4.2.

These developments have prepared by the ARAC Loadsand Dynamic Handling Working Group, supported by theCommittee of International Gust Specialists.Airworthiness Requirements tend to be put in rathergeneral “legal” terms, which may be subject to differentinterpretation. Additional documents, describingacceptable means and methods to comply with therequirements may be very helpful. Such informationmay be contained in ACJ’s (Acceptable means ofCompliance to JAR) in the case of JAR requirements, orin Advisory Circulars in the case of FAR requirements.

Traditionally, the calculation of aircraft response hasbeen made assuming linearity. With the advent ofnonlinear active control systems, aircraft are becomingincreasingly nonlinear and the assumption of linearity isbecoming more and more unacceptable for accurate loadprediction. The calculation of the response to a discretegust for a nonlinear aircraft may be time-consuming butoffers no fundamental problem. Three deterministic typemethods are considered here: Matched Filter Theory, the

72

Noback (or IDPSD) method and the Spectral Gust(Brink-Spalink ) method.

The existing PSD gust design criteria, however, arefundamentally based on linear response behaviour.Current Airworthiness Codes do not contain explicit ruleshow to determine PSD-gust loads for non-linear aircraft,but the NPRM presented in paragraph 4.2 foresees in thisshortcoming. In case of significant non-linearities, oneapproach towards determining the PSD design loads is tocalculate the aircraft response in the time domain of theaircraft to a patch of stationary random turbulence withan rms. value equal to 0.4 times the design gust velocityUσ. This procedure is known as the “StochasticSimulation method”, is physically well founded,straightforward and relatively easy to apply but verycomputer time consuming and hence expensive. Thealternative Probability Exceedence Criterion (PEC)method is also considered. A further approach is theStatistical Discrete Gust method, which attempts tocombine both discrete and stochastic methodologies. Fulldetails about the methods can be found in AppendixA4.1. There is a need to assess, validate and comparethese methods before they can be accepted forCertification purposes.

Section 4.3 presents a comparison of the above methodsfor design load calculations using various aircraft modelswith different nonlinearities. Two different institutescarried out these calculations and comparative results aregiven. Concluding remarks are presented in section 4.4.

4.2 Overview of Gust Requirements

4.2.1 Draft NPRM on ContinuousTurbulence.

The Discrete Gust Criteria in FAR25 and JAR25 havebeen changed a few years ago, but the Continuous GustRequirements in these codes have not been changed sincethe late sixties.

A Draft NPRM (Notice on Proposed Rule Making) hasbeen prepared recently, proposing changes in the FAR25.It is expected that these proposed changes will also beadopted in the JAR 25 Code.

The proposed requirement includes a revision to the gustintensity model used in the design envelope method tocontinuous turbulence on the basis of more recentstatistical data (including CAADRP data). The missionanalysis method will be eliminated and a newrequirement included for considering combined verticaland lateral turbulence. Provisions for treating non-linearities will also be included.

A summary of the most relevant changes that areproposed for paragraph 25.341 are:

(b) Continuous Turbulence criteria: ………(3) The limit turbulence intensities Uσ, in feet

per second true airspeed required for compliance withparagraph are –

(i) At speed from VB to VC:Uσ =Uσref Fg

Where –Uσref is the turbulence intensity that varieslinearly with the altitude from 90 fps (TAS atsea level to 79 fps (TAS) at 24000 feet and isthen constant at 79 fps (TAS) up to an altitudeof 50000 feet.Fg is the flight profile alleviation factor definedin paragraph (a)(6) of this section;(ii) At speed VD: Uσ is equal to ½ the

values obtained under subparagraph(3)(i) of this paragraph.

(iii) At speeds between VC and VD: Uσ isequal to a value obtained by linearinterpolation.

(iv) At all speeds both positive andnegative continuous turbulence mustbe considered.

(4) When an automatic system affecting thedynamic response of the airplane isincluded in the analysis, the effects ofsystem non-linearities on loads must betaken into account in a realistic orconservative manner.

(5) If necessary for the assessment of loadson airplanes with significant non-linearities, it must be assumed that theturbulence field has a root-mean squarevelocity equal to 0.4 times the Uσ valuesspecified in subparagraph (3). The valueof limit load is that load with the sameprobability of exceedence in theturbulence field as a velocity of Uσ.

(6) The resultant combined stresses fromboth the vertical and lateral componentsof turbulence must be considered whensignificant. The stresses must bedetermined on the assumption that thevertical and lateral components areuncorrelated.

4.3 Comparison of Methods to calculatedContinuous Turbulence Design Loads forNon-Linear Aircraft

This section presents results of comparative studies toevaluate methods for the calculation of design loads. Thesimulations were carried out by the National AerospaceLaboratory NLR N and the University of ManchesterUK, using the same aircraft models. A number ofdifferent methods were considered:

Stochastic Methods• Stochastic Simulation (SS)• Probability of Exceedence Criterion (PEC)• Power Spectral Density (PSD) [only for the linear

cases]

Deterministic Methods• Matched Filter Based method (MFB), both 1-

dimensional and Multidimensional• Indirect Deterministic Power Spectral Density

Method (IDPSD)• Spectral Gust procedure (SG)

Stochastic-Deterministic Methods• Statistical Discrete Gust (SDG)

73

A brief description of these methods is given in AppendixA4.1.

The following nonlinear aircraft models were used:

- Noback model: 2 DOF large transport aircraft withload alleviation through ailerons.

- F100 model: medium-sized transport with "Fokker-100-like" characteristics with load alleviationthrough ailerons.

- A310 model: an A310 model with load alleviationthrough ailerons and spoilers.

A description of these models is given in Appendix A4.2.Nonlinearity is introduced in these models by limits onthe control surface deflections. The A310 model controlsurfaces can only deflect upward (max. 10 deg.) in thenonlinear version, so that a non-symmetrical nonlinearityis introduced. Analysis could be performed using eitherthe linear or non-linear versions of these models.

4.3.1 Analyses made by NLR

The NLR investigation4 compared the three Deterministicmethods with the Stochastic Simulation methods and thePSD technique for the linear cases. For linear aircraftmodels, these Deterministic PSD methods and StochasticSimulation result in design and correlated load values yd

and zc that are equal to the "standard" PSD loads:

. U A = z U A = y zyzcyd σσ ρ

For nonlinear aircraft models, the standard PSD methodcannot be applied, because the model transfer functionsare then dependent on the input signal. The StochasticSimulation method has been proposed for the definitionof design and correlated loads in nonlinear cases. Thismethod is based on the probability of exceedence of loadlevels. The Deterministic methods aim to comply withthis Stochastic Simulation procedure in nonlinearcalculations.

By showing results of calculations for these three aircraftmodels it was demonstrated that the Deterministic andthe Stochastic Simulation procedures effectively lead tocorrect PSD loads in linear cases. The results for threenonlinear aircraft models obtained with the Deterministicmethods are presented, and the degree of compliance ofthe Deterministic methods with Stochastic Simulationwas investigated.

In Appendix A4.1 it is explained that the Deterministicmethods follow a more or less similar scheme. Anessential part in the procedures is the so-called gust filter.The Power Spectral Density of the gust filter response toa pulse input should have the von Karman powerspectrum shape. The impulse response power spectrumcan be calculated directly from the frequency-domainrepresentation of the gust filter G(jf):

( ) ( ) ( )T

jfGjfGf

∗

Φ =

where T = length of impulse response.

The gust filter impulse response for the IDPSD filtergives by definition exactly the von Karman Spectrum.Comparing the original MFB gust filter ("NASA"), and anew MFB gust filter that has been taken from Hoblit5, itappears that the Hoblit filter clearly approaches the vonKarman PSD better than the original NASA filter. TheHoblit gust filter has therefore been implemented in thepresent MFB procedure, which resulted in correct PSDloads in linear cases, contrary to MFB with the originalNASA gust filter, where slight deviations from AUσ werefound.

The bar-charts in figures 4.2 - 4.7 show the results of thecalculations for the three aircraft models and fivecalculation methods. The notation in the axis labels ofthese figures is as follows:

y,des = design load value of load quantity y.y,cor z = correlated value of y if z has its

design value.nonlin = closed loop system, nonlinear

(limited) load alleviation.nolim = closed loop system, linear

(unlimited) load alleviation.nocon = open loop system (linear).Stoch. Simul. = Stochastic Simulation result.PSD = standard PSD result.POS = "positive" design load case (A310

model only).NEG = "negative" design load case (A310

model only).

Note that correlated load values in some cases are givenwith opposite sign, indicated by a minus sign in thelegend. The results for the linear and nonlinear versionsof the A310 model are given in separate figures, becausethere is a difference between "positive" and "negative"nonlinear design load cases, due to the fact that aileronsand spoilers can only deflect upward in the nonlinearversion of this model.

These bar charts demonstrate that the three Deterministicmethods comply with the standard PSD results in linearcases, so it may be concluded that all Deterministicprocedures lead to correct results for linear aircraftmodels. Figure 4.2 for the linear A310 model showsstandard PSD results and Deterministic PSD resultstogether with Stochastic Simulation results. It can be seenthat the Stochastic Simulation procedure gives designloads close to the standard PSD values, and correlatedloads may deviate a few percent (of the design loadvalue) from the theoretical value, see for instance thecorrelated bending for the uncontrolled A310 model.

In nonlinear conditions, where controller actions arelimited, the Stochastic and Deterministic methods lead todifferent results. MFB and IDPSD do not differ much,but the correlated load values are different in some cases.A second optimization loop could have been added toMFB/IDPSD, calculating outputs at e.g. four more k/Keq

values around the optimum found, and find a highermaximum output with somewhat different correlated loadvalues. An even more rigorous search routine, the "multi-dimensional search", might also be applied. As it isbelieved, on the basis of NASA investigations, that sucha routine would change the design conditions by a verysmall amount in respect to the one-dimensional search,such calculations were not performed.

74

MFB and IDPSD both approach the StochasticSimulation results reasonably in figure 4.3; only thecorrelated value of ∆n for the nonlinear F100 model isreally very incorrect (wrong sign) for both methods, seefigure 4.4. The corresponding MFB/IDPSD design levelsof the bending moment in figure 4.5 differ more than10 % from the Stochastic Simulation value. The SGprocedure design loads and correlated loads can bothdeviate appreciably from Stochastic Simulation results.Similar findings were obtained for the Noback model,figures 4.6-4.7, where the major differences occur in thecorrelated y values.

The ailerons and spoilers of the A310 model can onlydeflect upward in the nonlinear version, so that differentgust design loads will occur in positive and negativedirections. In the IDPSD and MFB procedures, negativegust cases are created by reversing the sign of the gustinputs to the "first system". In the SG procedure the signof a design load is determined, by calculating the sign of:

dtyy 0∫∞

where y is the load quantity response to an SG input.

It can be seen in figure 4.3 that the positive and negativedesign load cases of wing bending do not differsignificantly, but the negative torsion design load isconsiderably lower than the positive design load in theresults of Stochastic Simulation, MFB, and IDPSD. It is agood point for MFB and IDPSD that they appear torepresent this effect in the same way as the StochasticSimulation method.

With regard to the required computational times thefollowing observations could be made. The SG method isvery fast, because only four time responses arecalculated. The IDPSD method takes some morecalculation time than MFB, because the "first system"response in IDPSD is twice as long as in MFB. StochasticSimulation takes much more time than the other methods(14 times the MFB time), mainly due to the countingprocedures for finding design levels and correlated loads.

The following conclusions can be drawn from thiscomparison of Deterministic methods with the StochasticSimulation and "standard" PSD methods:

- With the Hoblit gust filter, MFB is equivalent toIDPSD and "standard" PSD in linear cases.

- The results of MFB and IDPSD are reasonablysimilar in nonlinear cases; correlated loads maydeviate somewhat.

- MFB and IDPSD reasonably approach StochasticSimulation results in nonlinear cases, but this is notenough for design load calculations.

- The SG method deviates significantly from the othermethods in nonlinear cases.

- Stochastic Simulation takes much more calculationtime than the Deterministic methods.

4.3.2 Analyses made by the University ofManchester

The following methods were investigated at theUniversity of Manchester:

- IDPSD: Indirect Deterministic Power SpectralDensity

- MFB 1-D: Matched Filter Based 1-Dimensional- MFB Multi-D: Matched Filter Based Multi-

Dimensional- PEC: Probability of Exceedence Criteria- SS: Stochastic Simulation- SDG: Statistical Discrete Gust

The description of the methods can be found in Appendix4.1. The methods were applied to the simple 2-dof andA310 aircraft. Since the absolutely correct design loadcannot be obtained for a nonlinear system, one of themethods was to be used as a benchmark. In this case, thebenchmark was chosen to be the Matched Filter Based 1-Dimensional search method. This choice was dictated bythe relative simplicity of the method and by the fact thatit is less computationally expensive than the othermethods. However, the term "benchmark" does not implythat the design loads predicted by the MDB 1-D methodare taken to be the best estimates.

The graphical comparisons between the methodspresented in this section are based on the followingfigures (unless otherwise stated).

- Figures 4.8 and 4.9 show a direct comparison ofmaximum and correlated loads obtained by themethods for the Noback aircraft model.

- Figures 4.10 and 4.11 show a direct comparison ofmaximum and correlated loads obtained by themethods for the A310 aircraft model.

- Figures 4.12 and 4.13 Load variation with time andcritical gust shape for Noback aircraft load 2 andA310 load 3

4.3.2.1 Stochastic Simulation Method

The figures show a very good agreement between resultsusing the SS method and those from the two deterministicmethods. Figure 4.12 shows the load variation with timeand the critical gust shape for the Noback aircraft aspredicted by the MFB, SS and IDPSD methods. It can beseen that, even thought there is some differences betweenthe three gust shapes, the load variations are in very goodagreement with each other. This phenomenon highlightsthe main difficulty in predicting gust loads and worst-case gusts for nonlinear aircraft i.e. that there is not onesingle solution.

The good agreement between the two deterministicmethods and the SSB however, heavily depends on the

choice of the value of the turbulence intensity, gσ . The

authors of reference 6 suggest that, in order to comparethe two methods, the value of the turbulence intensityused with the MFB scheme should be

σσ Ug =

where σU is the design gust velocity. For the SSB

method, the suggested value is

3/σσ Ug =

75

The turbulence intensity used during the course of thiswork was

5.2/σσ Ug =

This value was preferred4 to 3/σU because it agrees

more closely with the representative, wrσ , value at

normal civil aircraft cruising altitudes.

4.3.2.2 PEC method

The design and correlated loads obtained by the PECmethod are in considerable agreement with thoseobtained by the SSB method, which is logical since bothmethods are stochastic approaches applied to the samesimulated patches of turbulence.

The comments made in the previous paragraph aboutturbulence intensity also apply to the PEC approach.

4.3.2.3 SDG method

The SDG method is the approach that yields loads whichare in least agreement with those obtained from the othertechniques. For the Noback aircraft, the SDG yields themost conservative design load for load 1 and the leastconservative one for load 2. For the A310, the SDGestimate for load 3 is in good agreement with thoseobtained from the DPSD procedures but, for load 4 theSDG again provides the least conservative design loads.This discrepancy is caused by the fact that the SDGmethodology, being based on a search through families ofdiscrete gusts, is significantly different to the other fourmethodologies (see Appendix 4.1).

4.3.2.4 IDPSD method

The agreement between the IDPSD and the MFB 1-Dmethods is, generally, very good. For the particular caseof the worst-case gust for Load2 of the Noback aircraft(figure 4.12), the agreement breaks down to a certainextent. The figure shows that the gust shape estimatedusing the IDPSD lies between the SSB and MFB 1-Dgusts. Nevertheless the resulting maximum loads are stillcomparable.

Since both the Noback and MFB 1-D methods aredeterministic methods, estimating worst-case gusts thereis no problem with scaling the turbulence intensity valuein order to get agreement between the two methods.

4.3.2.5 MFB Multi-Dimensional Search

Table 4.1 shows a comparison of results from the 1-dimensional and the multi-dimensional MFB searches,obtained from the Noback and A310 models. The tableconfirms previous findings7,8 that the 1-dimensionalsearch provides a very good estimate of the design load.The design loads for the Noback model have beenimproved upon by the MFB M-D method by up to 6.8%.However, for the A310 model, the improvement is almost

negligible. The fact that the multi-dimensional search ismuch more computationally expensive but only delivers asmall improvement in the final result suggests that the 1-dimensional search is more suitable, especially in thecase of the gust-load prediction for a full aircraft, wherethe design loads need to be predicted at a very largenumber of stations over the whole aircraft.

% Improvement6.86.70.10.2

Load MFB 1-D MFB M-D

Noback Load 1 10.73 m/s2 11.46 m/s2

Noback Load 1 6.55 m/s2 7.02 m/s2

A310 load 2 2.8242x106 lb.ft 2.8261x106 lb.ft

A310 load 3 2.3736x105 lb.ft 2.3793x105 lb.ft

Table 4.1:Comparison of design loads by the MFB M-Dand MFB 1-D methods for the Noback and A310 models

4.3.2.6 Comparative Results

The IDPSD method tends to predict slightly moreconservative results than the MFB 1-D method. In thecase of the Noback model the IDPSD results are closestto those obtained from the MFB M-D method. Since theSSB and PEC are stochastic, their design load predictionschange slightly every time the calculations areperformed. Consequently, there is no definitive way ofdetermining whether these predictions are generally moreor less conservative than the results obtained with theother two methods.

Another important conclusion is that the design loadpredictions of the methods agree more closely with eachother than the correlated load predictions. In reference 4this phenomenon is also noted. Additionally, Vink4

shows the cause of the phenomenon to be that thetheoretical standard deviation of the design load willgenerally be smaller than the theoretical standarddeviation of the correlated loads.

In many cases the methods predict very different worst-case gust shapes but quite similar design loads. Table 4.1shows the worst-case gusts and resulting load variationscalculated from the SSB, MFB and IDPSD methods forthe A310 wing torsion load. It can be clearly seen thatthree considerably different worst-case gust shapes yieldvery similar load variations and, hence, maximum loads.Again, this phenomenon is caused by the nonlinearity ofthe aircraft under investigation.

Table 4.2 compares the computational expense of theSSB, MFB 1-D, PEC and IDPSD methods. Neither theCPU time nor the number of floating point operations(flops) figures are absolute. CPU time depends on thecomputer used, the software installed. The number offlops performed depends on the programming and on theroutine that counts the flops. Nevertheless there is a clearpattern to the results in the tables. The leastcomputationally expensive method is the MFB 1-D andthe most computationally expensive one is the SSB, with

76

the IDPSD and PEC methods lying somewhere inbetween. The CPU time and number of flops for themulti-dimensional MFB and SDG methods are labelled"variable" in the table since the method relies on adirected random search. Hence, the duration of thecalculations is different every time the procedure isapplied, but always much longer than the duration of anyof the other methods.

Method CPU timeIDPSD 24.45MFB 1-D 18.73MFB M-D Variable*PEC 100.93SDG Variable*SSB 274.85

Table 4.2:Comparison of computational expense of themethods (applied to the A310 model) * Variable timesare caused by optimization procedures

4.4 Conclusions & Recommendations

This report has provided a brief historical backgroundand an overview of the current state of the airworthinessregulations as regards to gust loadings. In the future,certification regarding the effects of non-linearities on thegust loading of aircraft will become increasinglyimportant. A number of the most promising gust loadprediction methods, including both stochastic anddeterministic techniques, have been described andcompared analytically.

The nature of non-linear systems means that the principleof superposition does not hold and large amount ofcomputation is required to determine the design gustloads. Even then, there is no guarantee that a maximumhas been achieved. The computation can be performedeither via a stochastic approach that considers a largeamount of turbulent data, or a deterministic procedurewhereby some type of search is undertaken to find themaximum loads.

Two comparative studies were carried out using threedifferent non-linear aircraft models. Gust loads obtainedusing the different methods were compared. It was foundthat most of the analysis techniques gave similarestimates, although some variation in results was foundusing the version of the Statistical Discrete Gust methodemployed for this work, and also the Spectral Gustmethod. There is not enough evidence however tocategorically say one method is better, or worse, than theothers. The deterministic methods require lesscomputation.

There is a requirement for the research community todevelop new analysis methods that are able to predictdesign gust loads without resorting to large amounts ofcomputation. The test cases used in this study should beemployed as benchmark test cases for future comparativework.

4.5 References

1 Various authors: Loads and Requirements forMilitary Aircraft. Papers presented at the 83rdMeeting of the AGARD SMP, Florence,September 1996. AGARD Report 815,February 1997.

2 Flomenhoft, H.I., ‘Brief History of Gust Modelsfor Aircraft Design’ J. Aircraft v31 n5 pp1225– 1227 1994.

3 Fuller. J.R., ‘Evolution of Airplane Gust LoadsDesign Requirements’ J Aircraft v32 n2 pp 235– 246. 1995.

4 Vink,W.J.; A stochastic simulation procedurecompared to deterministic methods for PSDgust design loads. NLR TP 98240, 1998.

5 Hoblit,F.M.; Gust loads on aircraft: conceptsand applications, AIAA,Inc.,1988

6 R.C. Scott, A.S. Pototzky, and B. Perry III,Matched-Filter and Stochastic-Simulation-Based methods of gust loads prediction,Journal of Aircraft, 32(5):1047--1055, 1995

7 P.J. Goggin, Comparison of stochastic anddeterministic nonlinear gust analysis methodsto meet continuous turbulence criteria. Report798, AGARD, May 1994.

8 R.C. Scott, A.S. Pototzky, and B. Perry III,Computation of maximized gust loads fornonlinear aircraft using Matched-Filter-Basedschemes. J.Aircraft,30(5):763--768, 1993.

9 R.Noback, S.D.G., P.S.D. and the nonlinearairplane, TP 88018 U, NLR, NationalAerospace Laboratory, Holland, 1988

10 J.G. Jones, Statistical-Discrete-Gust Methodfor predicting aircraft loads and dynamicresponse, Journal of Aircraft, 26(4):382-392,1989.

11 D.L. Hull, Design limit loads based uponstatistical discrete gust methodology, Report798, AGARD, May 1994.

12 G.W. Foster & J.G. Jones, Analysis ofatmospheric turbulence measurements byspectral and discrete-gust methods,Aeronautical Journal}, pp 162-176, 1989.

13 E.Aarst & J.Korst, Simulated annealing andBoltzman machines, John Wiley & Sons, 1989.

14 R.C. Scott, A.S. Pototzky, & B.Perry III,Similarity between methods based on matchedfilter theory and on stochastic simulation,AIAA-92-2369-CP, 1992.

15 A.S. Pototzky & T.A. Zeiler, Calculating time-correlated gust loads using matched filter andrandom process theories. Journal of Aircraft,28(5):346-352, 1991.

16 R.C. Scott, A.S. Pototzky, & B.Perry III,Computation of maximized gust loads fornonlinear aircraft using Matched-Filter-Basedschemes, Journal of Aircraft, 30(5):763-768,1993.

17 J.E. Cooper & G.Dimitriadis, Prediction ofmaximum loads due to turbulent gusts usingnonlinear system identification, In Proceedingsof the CEAS International Forum onAeroelasticity and Structural Dynamics,Volume II, pages 71-78, Rome, Italy, June1997

77

18 J.G. Jones, Formulation of Design Envelopecriterion in terms of Deterministic SpectralProcedure, J. Aircraft, 30(1):137-139, 1993.

19 G.Rosenberg, D.A.Cowling, & M.Hockenhull,The deterministic spectral procedure for gustresponse analysis of nonlinear aircraft models.Intl Forum on Aeroelasticity and StructuralDynamics. pp 339 –358. 1993

20 R.C. Scott, A.S. Pototzky, and B. Perry III,Maximized gust loads for a nonlinear airplane

using matched filter theory and constrainedoptimization. NASA TM 104138, 1991.

21 R.Noback, The Deterministic Power-Spectral-Density method for nonlinear systems, TP92342 U, NLR, National AerospaceLaboratory, Holland, 1992.

22 R.Noback. The Deterministic Power-Spectral-Density method for linear systems. TP 92062U, NLR, National Aerospace Laboratory,Holland, 1992.

V

V

w

∆α

∆α = w/V

∆L=1/2 ρ V2 S Clα ∆α

Figure 4.1. Basic Gust Loading Mechanism

78

Figure 4.2 Bending and Torsion Loads. Linear A310.

79

Figure 4.3 Bending and Torsion Loads. Non-Linear A310.

80

Figure 4.4 F-100 Design and Correlated Loads

Figure 4.5: F-100 Design and Correlated Loads

81

Figure 4.6 Noback Aircraft c/g Acceleration

Figure 4.7 Noback Model c/g Acceleration by Aileron

82

1 20

2

4

6

8

10

12

Design Load -Correlated Load

Cen

tre

of G

ravi

ty A

ccel

erat

ion

(m/s

2 )

IDPSDMFB 1-DMFB M-DPECSDGSSB

Figure 4.8: Results for Noback model, centre of gravity acceleration

1 20

1

2

3

4

5

6

7

8

Design Load -Correlated Load

CoG

Acc

eler

atio

n C

ause

d by

Aile

ron

Onl

y (m

/s2 ) IDPSD

MFB 1-DMFB M-DPECSDGSSB

Figure 4.9: Results for Noback model, centre of gravity acceleration caused by aileron only

83

1 20

0.5

1

1.5

2

2.5

3x 10

6

Design Load -Correlated Load

Win

g B

endi

ng (

lb.ft

)

IDPSDMFB 1-DMFB M-DPECSDGSSB

Figure 4.10: Results for A310 model, wing bending

Figure 4.11: Results for A310 model, wing torsion

1 20

0 . 5

1

1 . 5

2

2 . 5x 10

5

Des ign Load -Co r re la ted Load

Win

g To

rsio

n (lb

.ft)

I D P S D

M F B 1 - DM F B M - D

P E CS D GS S B

84

Figure 4.12: Comparison between SSB, MFB 1-D and IDPSD (labeled ‘nob’)for Noback a/c load 2 (design load and gust shape)

Figure 4.13: Comparison between SSB, MFB 1-D and IDPSD (labeled ‘nob’) for A310 wing torsion(design load and gust shape)

85

4.6 APPENDIX A4.1

Methods for design gust load prediction for nonlinearaircraft

This appendix gives a brief description of the methodsconsidered in this chapter. They have been categorized aseither Stochastic or Deterministic methods, althougharguably the Statistical Discrete Gust methods could bein their own section. Further details can be found in thereferences.

4.6.1 Stochastic Methods

4.6.1.1 Probability of Exceedence Criteria

The Probability of Exceedence Criteria (PEC) method9 isan extension of the Power Spectral Density method(PSD) for nonlinear aircraft. The PEC is stochastic andattempts to calculate design loads. The procedure is asfollows 7,9:

1. The flight conditions at which the design loads areto be evaluated are prescribed and values of Uσ andb2 are determined from the airworthinessrequirements. b2 is a coefficient used in theexpression for the probability that the load willexceed the design load - its variation with altitudecan be found in reference 5.

2. A representative value of the rms gust intensity,

wrσ , is computed using

( )2

/411 22

2

bUbwr

σσ++

=

3. An input white noise signal with wrσ is generated,

passed through a gust pre-filter and fed into thenonlinear aeroelastic model. The resulting load timehistory for load y is used to calculate the probabilitythat the design load will be exceeded in a turbulent

flow-field of intensity wrσ using

( )

wr

y2wr

2

2

wr

wrd

dydA2

yexpA

2

1

,yyP

d

σ

σ

−σπ

=σ>

∫∞

where wryA σσ /=

4. The design load is defined as the value of the loadfor which

=>

wr

wrd

UyyP

σσ σ

2erfc

2

1),(

where erfc is the error function complement.

Instead of calculating the probability distribution of loady, it is possible to obtain the design load by estimating thenumber of exceedences, N, of this load given by4

),yy(Pdt

TN wrd

tot σ>=

where Ttot is the total length of the simulation (inseconds) and dt is the time step. Then, the arraycontaining the load response y is sorted from higher tolower values and the design load is the Nth element of thesorted array. If N is not an integer, linear interpolationcan be used to obtain the design load.

σ

σ− σ

w22

2w

2

Uerfc

b2exp

This procedure only gives an estimate of the nonlineardesign load which may be substantially different to thereal value9. The estimate can be improved by repeating

the procedure for two values of wσ at which the value

of the following quantity is the same

Then, the design loads obtained for these two values ofgust intensity can be combined with the initial estimatesuch that

)(y25.0)(y25.0)(y5.0y 2wd1wdwrdd σ+σ+σ=

It has been suggested4 that, instead of three simulations

with three different values of wσ , only one simulation

with 5.2/σσ Uw = can be performed. The results

will be adequate in the altitude range of 22,000ft-35,000ft

since, in this range, the value of wrσ is very close to

5.2/σU . This latter approach is also adopted in the

present work since it is suggested that increasing the total

simulation length at one value of wσ improves the

quality of the design load predictions by a larger amountthan increasing the number of simulations at different

values of wσ .

The correlated loads can be obtained using

P(z > zc|y – yd) = 0.5

i.e. the probability of load z to be higher than thecorrelated load, zc , when load y assumes its design valueis 0.5. This is implemented by extracting the value of z atall the time instances were y=yd. The probabilitydistribution of these values is then calculated and thecorrelated load is obtained as the load whose probabilityis 0.5. As with the design load, the correlated loads canbe obtained using the number of exceedences instead ofthe probability distribution.

Since the PEC input to an aeroelastic model is stochasticturbulence, modelled as white noise, in order for themethod to work accurately, long simulation times areneeded so that the variance of the input is as close as

86

possible to wσ and its mean is almost zero. However,

the advantage that this method has over some of the lesscomputationally demanding discrete gust methods is thatthe airworthiness requirements concerned are moreuniformly defined 5.

4.6.1.2 Statistical Discrete Gust Method

The Statistical Discrete Gust Method (SDG) has beenintroduced as a method that employs a better descriptionof atmospheric turbulence than the Power SpectralDensity method for extreme gusts on linear aircraft 10,11.This description is based on families of discrete 1-cosineramp gusts. The present implementation of the SDGmethodology is based on a similar implementation9. Itshould be noted that the method was developed as anattempt to bridge the gap between continuous turbulenceand discrete gusts methodologies and is beingcontinuously refined, most recently with the use ofwavelets. The SDG calculates design loads.

Figure A4.1 shows a single discrete gust, as used by theSDG method. Initially, its velocity increases in a 1-cosinefashion until, at a distance H, it levels out to the value Uwhich is given by

3/10 HUU =

if H is less than L, the length-scale of turbulence, and3/1

0 LUU =

if LH ≥ . The value of U0 is decided by the

equivalence of the design value of gσ in the continuous

turbulence PSD analysis to the SDG analysis as11

4.100gU

σ=

where gσ is obtained from the airworthiness

requirements5.

For extreme turbulence the scaling of equation the gustvelocity equation changes to

6/10HUU =

This is how the SDG methodology bridges the gapbetween continuous turbulence and discrete gusts.Continuous turbulence is assumed to be self-similar,which is where the 1/3 scaling law comes from. Self-similarity can be modelled as a stretching transformation.In the time-domain, if the time axis is stretched by acertain amount, h, the dependent variable, say y(t), will

be stretched by λ−h . The similarity parameter λ can be

chosen such that the function )(htyh λ− is statistically

independent of h. This value for λ can be obtained byconsidering the spectrum, Φ(ω) of the process y(t), whenstretched by h, which in reference is shown12 to satisfy

( ) ( ) ( )ωωλ Φ=Φ+− hh /12

In the special case where the process y(t) is turbulent, theVon Karman spectrum applies, i.e.

( )

( ) ( )6/112

2

23322

6/52

211

339.11

339.13

81

2

339.11

1

+

+

=Φ=Φ

+

=Φ

V

L

V

L

L

V

L

L

g

g

ω

ω

πσωω

ωπ

σω

Simple algebra shows that the limit of both Φ11(ω) andΦ22(ω) as ω tends to infinity (which defines the inertialsubrange where self-similarity applies) is

3/5lim −∞→ =Φ ωω A

where A is a proportionality constant. Consequently

( ) 3/53/5

12 −−

+− =

ωωλ A

hAh

For this expression to be satisfied, h must vanish from theleft-hand-side, or

3

512 −=−− λ

Hence for continuous, self-similar turbulence, λ=1/3.

Discrete gusts are extreme events for which self-similarity breaks down. They are larger-scale and moreordered events than the background turbulence withinwhich they are contained. The similarity parameter forsuch events is given by10

3

3

3

1 D−−=λ

where D is termed the active volume of turbulence and

has values 32 ≤< D . For D=3 the standard self-similar value, λ=1/3, is obtained. For a value of D=2.5,the extreme turbulence similarity parameter is obtained,λ=1/6. Hence, with a simple change in the scaling law,the SDG method can be made also applicable to extremeturbulent events like discrete gusts.

At a particular value for the gust-length, H, the nonlinearaeroelastic system under consideration will exhibit amaximum load response. The maximum value of this

maximum response, 1γ is an estimate for the design

load, yd1. A second estimate is obtained using a pair ofgusts as shown in figure A4.2. Here, there are threeparameters that govern the gust shape, H1, H2 and thespacing between the two gusts, S. The values of these

parameters are varied until the maximum, 2γ , is

obtained. Another two estimates for the design load are

87

calculated using two pairs of gusts and four pairs ofgusts. Finally, four design loads are calculated using

0444

0333

0222

0111

Upy

Upy

Upy

Upy

d

d

d

d

γγγγ

====

with p1=1.0, p2=0.81, p3=0.57 and p4=0.40. For highlydamped systems the first two design values are moreimportant, for slightly damped ones the last two designvalues predominate.

For linear systems, estimating the maximum responsedue to SDG gusts is simple since superposition can beemployed. For nonlinear systems this estimation can onlybe performed by means of an optimization scheme,especially for the longer gust-shapes. The optimizationscheme chosen for this study was Simulated Annealing13.

4.6.1.3 Stochastic Simulation

The Stochastic Simulation method (SS) modelscontinuous turbulence as a white noise input with a VonKarman spectrum, in the same way as the PEC method.Hence, the SSB is stochastic and can calculate designloads, correlated loads and worst-case gusts, given atarget value for the design load. The procedure is asfollows14:

1. A Gaussian white noise signal with unity variance isgenerated and fed through a gust pre-filter, such as

( )

+

+

+

+

+

πσ

=

V

Ls0898.01

V

Ls823.01

V

Ls083.21

sV

L1298.01s

V

L618.21

V

L

sG

g

The output of the filter is a time history ofcontinuous turbulence data. The object is to identifysegments of this time history that lead up to peakloads.

2. A number of long time-domain simulations areperformed

3. The load time histories obtained from thesimulations are analysed. Instances in time areisolated where the load exhibits a peak near aprescribed value or within a specified range. Thenstandard durations of time data leading up to thepeak values are extracted, lined up in time andaveraged. The result is 'averaged-extracted' time-histories of the excitation waveform (input to thegust filter), gust profile (section of turbulence data)and load. These have been shown to be directlyequivalent to results obtained by the MFBmethods14, if the value of the turbulence intensity

gσ is selected appropriately.

To ensure that there is an adequate number of extractedsamples so that the final waveforms are as smooth aspossible, very long simulations are required (1000

seconds has been suggested14). Long simulation timesalso ensure that the white noise input has a variance veryclose to unity and a mean very close to zero. Finally, theextraction and averaging process must take placeseparately for positive and negative peak load values.

The stochastic simulation method, as outlined herecannot be used on its own since it requires a target load tobe specified, around which it will search for peaks in theload response. This target load value can be supplied byanother method. The authors of ref. 14 used the MFBmulti-dimensional search procedure to obtain the targetdesign load value and picked peaks in the SSB load

output within %8± of that value. Of course, the objectof their work was to show that the MFB results areequivalent to stochastic results. In a straightforwarddesign loads calculation it would be extremely wastefulto use two of the most computationally expensivemethods to produce the same results twice.

However, it is suggested here that the SSB method can beused to supplement results obtained by the Probability ofExceedence Criteria method. As mentioned earlier, thePEC method will only produce values for the design andcorrelated loads. It will not calculate time-variations ofthe loads or the gust velocity. The SSB, on the other handcan produce design and correlated load responses andcritical gust waveforms. Hence, the PEC method can beused to yield a target value for the design load to besubsequently used with the SSB method.

4.6.2 Deterministic Methods

Figure A4.3 and table A4.1 summarize the Deterministicprocedures. An input signal to the "first aircraft system",H1, is generated by feeding a pulse through a (vonKarman) gust filter G, with ,G(jf),=[Φn

ww(f)]2. The powerspectrum of the input to the first system will thus havethe shape of the von Karman spectrum. The pulsestrength k is variable in the MFB method, and constant inthe IDPSD (k=Uσ) and SG (k=Uσ√T, where T = length ofgust input) methods. It should be noted, that the gustfilter in the MFB method is only an approximation of thevon Karman spectrum, and in the version used in thisreport it is the Hoblit approximation .

The first aircraft system, H1, represents the non-linearaircraft equations of motion in MFB and SG. In IDPSD,H1 is a linearized version of the non-linear aircraft, byreplacing the non-linearity by a linear element with an"equivalent gain", Keq. Keq is a multiplication factor to theoriginal gain in the feedback loop, with 0⊆ Keq ⊆ 1

For nonlinear systems, the three Deterministic methods

apply different procedures:

- MFB varies the strength k of the input pulse to the

first gust filter.

- IDPSD varies the value of the equivalent gain that

represents the nonlinearity in the first system.

- SG varies the phase relation of the gust filter, which

is limited to only four different phase relations.

88

4.6.2.1 Matched Filter Based 1-Dimensionalsearch

Matched Filter Theory (MFT) was originally developedas a tool used in radar technology15. The main objectiveof the method is the design of a filter such that itsresponse to a known input signal is maximum at aspecific time, which makes it suitable for application togust response problems. The method can only be appliedto linear systems because it makes use of the principle ofsuperposition, which does not apply to nonlinear systems.However, by applying a search procedure, it can beadapted to provide results for nonlinear aircraft. Themethod is deterministic.

The technique is quite simple and consists of thefollowing steps15,16 :

1. A unit impulse of a certain strength Kg is applied tothe system.

2. The unit impulse passes through a pre-filterdescribing gust turbulence (usually the Von KarmanGust pre-filter).

3. The pre-filtered input is fed into the aircraft modeland the response of the various loads is obtained(e.g. wing root bending and torsional moments).

4. The response of the load whose design value is to beestimated is isolated, reversed in time, normalizedby its own energy and multiplied by Uσ, the designgust velocity (which is determined by airworthinessrequirements 5).

5. The resulting signal is the input that maximizes theresponse of the chosen load for this particularimpulse strength, Kg. It is then fed back into thesystem (first the Gust pre-filter, then the aircraftmodel) in order to obtain the response of the loadwhose design value is to be estimated and also theresponses of the other loads (which are termed thecorrelated loads).

6. The procedure is repeated from step 1 with adifferent Kg.

The characterization of the method as one-dimensionalrefers to the variation of Kg. The end result is a graph ofpeak load versus initial impulse strength. The maximumof this function is the design load and the gust input thatcauses it is termed the Matched Excitation Waveform. Itmust be mentioned at this point that the method does notguarantee that the maximum load for a nonlinear aircraftwill be obtained. As was found in refs. 7 and 17, thevariation of peak load with initial impulse strength forsome types of nonlinearities (e.g. freeplay and bilinearstiffness) does not display a global maximum (instead itslowly asymptotes to a certain value).

4.6.2.2 Deterministic Spectral Procedure

This method was first proposed by Jones18. In its mostgeneral form it is based on the assumption that thereexists a single deterministic input function that causes amaximum response in an aircraft load. It states that adesign load on an aircraft can be obtained by evaluatingthe load response to a family of deterministic gust inputswith a prescribed constraint. In practice, this implies asearch for the worst case gust, subject to the constraintthat the energy of the gusts investigated is constant. The

method is deterministic. The procedure consists of thefollowing steps:

1. A model input shape in the time-domain isgenerated.

2. The input shape is parameterized to produce a set ofdescribing coefficients

3. The coefficients are used to generate the inputwaveform

4. The energy of the input is constrained by dividingthe signal by its rms value

5. The constrained waveform is fed into a turbulencepre-filter and next through the nonlinear aircraftsystem

6. The aircraft load response is assessed. If it has notbeen maximized the coefficients that generate theinput are changed and the process is repeated fromstep 3.

This iterative procedure requires a constrainedoptimization scheme, to ensure that the maximum loadhas been obtained, and a model input shape. Theoptimization scheme proposed originally18 was simulatedannealing. Another approach16 is to convert theconstrained optimization problem to an unconstrainedone by means of the Kreisselmeier-Steinhauser function.

As for the generation of the initial input shape, twoapproaches have been proposed. In ref. 19 a white noisegust model is used. The problem with this approach isthat it is more difficult to parameterize a random signalthan a deterministic one. Alternatively 16, the MFB 1-dimensional search results are proposed as the input tothe DSP loop, which results in what is called the MFBmulti-dimensional search procedure.

The parameterization process is probably the most crucialaspect of the DSP method. Input waveforms have to bedescribed by a minimum number of coefficients tominimize computational cost but this description has tobe as accurate as possible. Again, two popular procedurescan be found in the literature. The first19 is to fit thewaveform by a number of half-sinusoid (or cosinusoid)functions. The other approach is to fit the waveformusing a set of Chebyshev polynomials16. In the samereference, a Fourier series approach was considered but itwas found to be much more computationally expensive.

The most common implementation of the DSP method isthe Multi-Dimensional Matched Filter Based methodwhich is described next.

4.6.2.3 Multi-Dimensional Matched Filter BasedMethod

The Multi-Dimensional Matched Filter Based (MFBMulti-D) method16,20 for gust load prediction fornonlinear aircraft is a practical application of theDeterministic Spectral Procedure. It was designed toprovide a more computationally efficient alternative tothe Stochastic Simulation Based approach. Reference 16shows how the method provides almost identical resultsto those obtained by use of the SSB but with lesscomputational effort. The method is deterministic.

89

The MFB Multi-D approach revolves around the fact thatthe usual design envelope analysis can be reformulated asan exactly equivalent time-domain worst-case analysis. Inother words, the search for a worst-case gust load in thepresence of a turbulence field of prescribed intensity isequivalent to the search for a design load19. Hence, thesimplest possible procedure for determining the worst-case load is to simulate very long patches of turbulenceand to look within the load response of the aeroelasticsystem in question for the design load. This is thestochastic simulation approach that requires significantamounts of computation.

The worst-case load problem can be simplified by notingthat the significant part of a long turbulent signal thatcauses the maximum load is short and can beapproximated as a discrete gust. Hence the MFB Multi-Dmethod searches for the single discrete worst-case gustwaveform thus avoiding the need for long simulationtimes.

The implementation of the method is as follows, alsodepicted graphically in figure A4.4:1. An initial guess for the worst-case gust waveform

(or matched excitation waveform) is obtained by useof the 1-dimensional MFB procedure.

2. The initial guess is parameterized. In the presentapplication the parameterization scheme used isChebyshev Polynomials.

3. The values of the various parameters are changedand the resulting waveform is fed into theaeroelastic system (including a turbulence pre-filteras described earlier).

4. The resulting maximum load is compared to theprevious value for the worst-case gust load and isaccepted or rejected according to some optimizationprocedure. The optimization procedure used for thepresent application is Simulated Annealing. Theprocedure is repeated, i.e. the parameters arechanged again resulting in a new gust waveformwhich is then used as an input to the system, untilthe worst-case gust load is obtained.

4.6.2.4 Indirect Deterministic Power SpectralDensity Method

The Indirect Deterministic Power Spectral Densitymethod (IDPSD)20,21, is derived from the DesignEnvelope Analysis5 of the continuous Power SpectralDensity method. For linear aircraft it yields design loadsequal to those obtained by the PSD method but using adeterministic input, in a similar way to the linear MFTmethod. For nonlinear systems it can be extrapolated to a1-dimensional search procedure, equivalent to the MFB1-D search but involving a linearized representation ofthe system. The method is deterministic.

The IDPSD procedure is very similar to the MFB 1-Dmethod with two main differences. Firstly, the IDPSDmethod uses a different gust filter and, secondly, theinitial excitation is applied to a linearised version of thesystem whose output is then reversed, normalized and fedinto the nonlinear system. Hence, the MFB 1-D methodconsists of a filtered impulse of variable strength fed intothe nonlinear system, the resulting gust waveform beingfed into the same system. In the IDPSD method, an initialinput of constant strength is fed into a linearised system,called the first system, whose nonlinear element has been

replaced by a variable gain. The resulting waveformforms the input to the nonlinear system, called the secondsystem. The search procedure consists of varying thelinear gain until the response of the second system ismaximized.

The input to the first system is given by )(tVUσ ,

where σU is the design gust velocity and V(t) is the

Fourier Transform of the two-sided Von Karman

Spectrum, ( )ωwwΦ , given by

( )22

2

339.11

339.13

81

+

+

=Φ

V

L

V

L

V

Lww

ω

ωω

where ω is the radial frequency, L is the turbulencelength-scale and V is the aircraft velocity22. This inputcan be alternatively defined as the Auto-Correlation

function pertaining to ( )ωwwΦ , i.e.

ωωπ

ω dtRtV tjww e)(

2

1)()( 22 ∫

∞

∞−

Φ==

The Von Karman Spectrum can be expressed in a morepractical form as the Auto-Correlation function of thefiltered MFB impulse,

2)(

)()()(

τ

ττ

g

gg

u

tuutV

+=

where ug is the MFB filtered impulse gust velocity, theoverbars denote averaging and τ is an integrationvariable. The solid line is the Fourier Transform resultand differs from the Auto-Correlation result (dotted line)in that it takes negative values away from the peak. As aconsequence the Auto-Correlation result was preferredfor the present work.

The IDPSD Method procedure is as follows:

1. )(tVUσ is formed, say using equation (6).

2. The input is fed into the linearized aircraft modelwith linear gain K and the response of the variousloads is obtained (e.g. wing root bending andtorsional moments).

3. The response of the load whose design value is to becalculated is isolated, convoluted by V(t),normalized by its own energy and multiplied by

σU , the design gust velocity.

4. The resulting signal is the input that maximizes theresponse of the chosen load for this particularlinearised gain, K. The signal is then fed into thenonlinear system in order to obtain the response ofthe load whose design value is to be calculated andalso the responses of the correlated loads.

5. The procedure is repeated from step 2 with adifferent K.

Reference 21 suggests that the values of the linearizedgain should be between 0 and 1.

90

Table A4.1 Elements of Deterministic Methods4

Element Matched filter(Scott e.a.)

IDPSD(Noback)

Spectral Gust(Brink-Spalink e.a.)

ImpulseStrength k

k variable k = Uσ k = Uσ*√T

GustPrefilter G(jf)

|G(jf)|≈ √Φn(f)One set ϕ(f)

|G(jf)| = √Φn(f)One set ϕ(f)=0For all f

|G(jf)| = √Φn(f)four sets ϕ(f)

AircraftSystem H1(y)

(Nonlinear)set of equationsfor output y

LinearizedEquations;Variable"equivalent gain"

Nonlinearset of equationsfor output y

Calculationy-norm:

∫

∫

∞

∞

∞

∞

)( )( =

)( =

*

-

2/1

2

-

2/1

norm

dfjfsjfs

dttsy

-----------------------------------------------------For linear system:

yyUk

kdfGGHHky

desnorm

y*11

+

-

2

2/1

norm

= = if

A = . . =

→

∗∞

∞∫

σ

"Criticalgustprofile" w(t)

For linear systemssame profile formatched filter and IDPSD

AircraftSystem H2(y)

Nonlinear setof equations

Ydes Variable kydes = [yt]max

Variable gainof H1(y)ydes = [yt]max

SG stopshere:Four valuesfor ynorm,

T

yy

(max) = norm

des

91

4.7 Appendix A4.2 Description of AircraftModels

Three symmetrical aircraft models have been consideredin this research. The first one is a simple model of alarge transport aircraft with two degrees of freedom, pitchand plunge, and a load alleviation system that feeds backthe centre of gravity acceleration to aileron deflection.The model is shown in figure A4.5. The functions C(s)and D(s) are the transformed Wagner - and Küssnerfunctions representing unsteady aerodynamic loads.Output y in the figure is the centre of gravityacceleration, and output z is the centre of gravityacceleration caused by aileron action only. This model iscalled the Noback-model in this report.

The second model represents an aircraft with "Fokker-100-like" characteristics. This model has the two rigiddegrees of freedom pitch and plunge, and ten symmetricflexible degrees of freedom. This flexibility isrepresented by the first ten natural modes of the aircraftstructure. Aerodynamic forces are calculated with striptheory, and unsteady aerodynamics is accounted for byWagner - and Küssner functions. The wing has 27 stripsand the tail 13; the fuselage is considered as one liftingsurface. The Wagner - and Küssner functions arecalculated at 3 locations on the wing and at 1 location onthe horizontal tail.

The gust penetration effect and the time delay of thedownwash angle at the tail with respect to the wing areincluded. Taking these two effects into account, makes itnecessary to apply time delays to the gust input, and tothe state variables (because the angle of incidence at thereference point on the wing is a function of all states)respectively. Especially the latter considerably increasesthe total number of system states.

A Load Alleviation System is implemented in the modelthat feeds back the load factor to a (symmetrical) ailerondeflection. Figure A4.6 shows the aircraft system withthe feedback loop to the aileron input. The configurationof the Fokker 100 model used in this report is:

ma/c = 40,000 kg Iy = 1.782 106 kgm2

V = 220 m/s, altitude = 7000 mcentre of gravity location at 25 % mean-aerodynamic-chord.

The third model has been distributed at the GustSpecialists Meeting of March 1995. It represents an A310aircraft, containing plunge, pitch, and 3 symmetricflexible degrees of freedom. Unsteady response isassumed instantaneous, and gust penetration is notrepresented. The aircraft with control system is depictedin figure A4.7. The centre of gravity acceleration is fedback to both the ailerons and the spoilers through afeedback gain of 30 degrees per g load factor. Aileronsand spoilers have the same authority: deflections between0 and 10 degrees. This means that the nonlinearity in thiscontrol system is "non-symmetric"; the control surfacescan only deflect upward. The load quantity outputs of thissystem are the increments of:- Engine lateral acceleration [g].- Wing bending moment [lb.ft].- Wing torque [lb.ft].- Load factor [g].

92

Figure A4.1 Single SDG Gust

Figure A4.2 Pair of Statistical Discrete Gusts

93

Figure A4.3 Process for Deterministic Methods

94

Figure A4.5 Noback Aircraft Model

Figure A1.4: Graphical description of MFB Multi-D procedure

95

a/c responsel o a d s r e s p o n s e s

0

t r im

u [ 3 ]

s e l e c t _ d n

[ t ,w]

i n p u t

0

e l e v a t o r

x ' = A x + B u y = C x + D u

S t a t e - S p a c e

M u x

M u x

1

T a . s + 1

A i l e r o n

-K

- 2 0 d e g / g- 1 0 < y < 1 0

Figure A4.6 Fokker-100 Model

u[4]

se lec t_dn

1/ .3048

m2 f t

p i / 1 8 0

deg2rad

statespace1

To Workspace1

outp

To Workspace

nfi lt(s)

dfi l t(s)

TFF gustf i l te r x ' = A x + B u

y = Cx+Du

State-Space

1

taus.s+1

Spo i le r

Sa tu ra t i on

M u x

Mux1

M u x

M u x

m a i n g a i n

G a i n

[T, inpt ]

FromWorkspace

1

taua.s+1

A i l e r o n

Figure A4.7 A310 Model

96

5 A More Global Approach

5.1 Why a more global approach

It comes from the necessity to get rid of insufficiencies ofclassical load regulations, the main lines of theseregulations being:

• Limit loads are defined as "maximum loads"expected in service.

• Regulations prescribe the set of loading conditions(ex.: manoeuvres), or directly the computationprocedure (gust, ground loads), to be considered forfinding these "maximum loads".

• Ultimate loads result from multiplication of limitloads by a prescribe safety factor.

The sources of difficulties are principally:

• The chronic lack of exhaustively of regulationloading conditions set up from flight experience ofpast programme.

Already with conventionally controlled aircraftmanufacturers had to add "company" design loadcases, for instance to cover countered maneuverswhere the pilot, remaining inside limit values of"official" load factors and control surface deflections,could easily make severe structural loading.Matters worsen when new technologies come, whichhas been met, in particular with:

− the design of fly by wire combat aircraft andthe associated concept of care free piloting,where "maximum loads" can be reached everyday as result of extremely complex and variousdynamic maneuvers, far from regulationmaneuvers.

− the design of re-entry vehicles with their "hotstructures", where limit conditions result fromcombinations of mechanical, thermal loads, andaging conditions, closely depending onstructural design.

• The need to clarify the meaning of the word"maximum loads" ; its have been often restricted toloading conditions corresponding to maximum valuesof "general load" components, notion becominginsufficient when "long beam theory" is not relevant(e.g. delta wings), where local structural failuremodes are not only led by "general loads".

Still more severe difficulties occur when thermalloads, or any physical or chemical environmentalconditions, or aging and fatigue effects, must beconsidered in addition to mechanical loads.

• The safety factors philosophy

− first it is a need to clarify the present safetyfactor rules when other physical effects(thermal, environmental, aging/fatigue, …) areadded to mechanical loads, where severalcomponents of safety factor must appear,corresponding to each physical effects.

− more fundamentally we have to open the debateof safety factor evolutions with innovation,with the progress both of design solutions andof analysis process, knowing that we are to dayunable to quantify, inside the present globalsafety factor, separated contributions of loads,manufacturing, strength, …, or of any otheruncertain elements.

Faced with these questions since the mid 70ies withMIRAGE 2000 programme and after with RAFALE,DASSAULT AVIATION have developed andexperienced the "more global approach", alreadypresented to AGARD SMP in 1984 and 1996 (ref. 1 and2) and reminded hereafter to be proposed now to theRTO community.

To note that this approach, including extensions tothermal loads, have been carried by ESA and CNES fordesign loads of HERMES space shuttle .

5.2 Limit Loads

5.2.1 Basic principles of the "more globalapproach" for limit loads

They are:

• To keep (even to reinforce) the limit load definitionof classical regulations:

Limit loads are the maximum loads expected inservice .

• To consider that it is not necessary to prescribe anyparticular set of loading conditions withinregulations.

"Maximum loads" must come from scenario analysesof missions/flight conditions/ environments, suited tothe designed product.In practice, this don't prevent aircraft designer frombuilding a set of

"reference design load cases",

under his responsibility and to demonstrate that these"reference design loads" envelop the maximum loadsexpected in service.

• To clearly define the meaning of the sentence :

"Maximum loads expected in service" ,

and to propose a practical process for theirdetermination (see hereafter).

5.2.2 "Maximum loads" through "Load SeverityIndicators"

The notion of "maximum Loads" has a meaning onlythrough the effects of loads induced on the structure:

A load case is referred to as a maximum load case assoon as it produces the maximum value of at least 1failure mode strength criterion.

97

Which need in theory :

• To identify of all structure failure modes liable tooccur under mechanical loading (local stress - orstrain - induced ruptures, local or general buckling,non-allowable overall deflections, …), and moregenerally under all other physical effects (thermal,aging, …).

• To allocate to each one of these failure modes of ascalar strength criterion calculable in function ofthe loading conditions and of the structure design.When necessary the strength criteria may take intoaccount thermomechanical and aging effects.

• To sweep all "expected" loading conditions (see6.2.3) calculating each of these strength criteria.

To reduce the effort of monitoring thousands of localstrength criteria, we have introduced the notion of :

"Load Severity Indicators".

Which are few tens to few hundreds of scalar indicatorsstanding in monotonic relation to a structure area strengthcriteria, whatever the loading.

As "load severity indicators" are generally chosen:

− components of stress or strain in pilot points,

− internal reactions (e.g. : loads on the wing or controlsurface attachment bearings),

− classical "general loads" components (shear force,bending moment …) on particular sections.*

Computation management will be simplified if theseverity indicators remain linear functions of the loads ;they can then be calculated at low cost in function offlight parameters, starting from a matrix of "loadseverity indicator operators" giving the relation withflight mechanics state vector, this table being built priorto maneuver computations.

The strain gauge distribution of flight test aircraft willattempt to reflect the choice of load severity indicators,thereby providing for calibration and validation of theoperators and thus, of the whole load computationprocess.

Once "load Severity Indicator operators" arebuilt/calibrated/validated, the computer cost of maximumload case selection comes cheap, corresponding to linearcombinations of "load severity indicator operators",downstream sweeping of:

• flight mechanics simulations, (numerical simulations/ real time flight simulator),

• environmental aircraft responses (gust, turbulence,…),

• ground load conditions,

• etc… ,

marking as limit load case conditions where maximal of"load severity indicators" are reached,

and/or :

checking that these maximal remain under the level of"reference design loads" chosen a priori.

5.2.3 "Maximum Loads Expected in Service"

That means that we have to sweep all possible scenario,during an aircraft life, of missions / maneuvers /environments /…, computing previous Load SeverityIndicators, and selecting, as design load cases, loadingconditions where load severity indicators are maximal.

When relevant, it can correspond to probabilisticanalyses in the spirit of Continuous Turbulenceregulations( e.g. FAR 25, appendix G)

− to determine from mission analysis limit value of"load severity indicators", corresponding to 1average exceeding per aircraft life .

− to ensure that the limit load set (or the "referencedesign load" set of the manufacturer) envelop theselimit values.

5.2.4 Application to design of "fly by wire"aircraft

It have been detailed in reference 2, the principle is tointegrate the designs of structure and of Flight ControlSystem via the following iterative process :

• Start from a first set of "reference design loads"

− from aircraft manufacturer experience

− reflecting flight quality requirements

• Design of airframe

− supported by F.E./Aeroelasticity analyses /optimizations

− delivering "load severity indicators" operatorsand their associated limit values

• Design of F.C.S.

− to maintain "load severity indicator" responsesbelow their limit values for all possible scenarioof missions / maneuvers / environments,

or

− to define new limit load cases (→ airframedesign iteration).

5.3 Ultimate load definition and Safety Factors formultiphysical effects

When limit loads contain only "mechanical effects" thedefinition could remain "as is" :

98

Ultimate loads result of multiplication of limit loadsby a prescribed safety factor.

When others physical effects (thermal, aging, …) occurin limit conditions, specific safety factors must beapplied successively and separately on each of theseeffects (the others remaining at their limit values) ; forinstance:

• on heat fluxes or on parts of heat fluxes or onresulting temperature fields.

• on life duration for fatigue/aging loads.

• for each kind of other physical/chemicalenvironmental conditions.

The nature and the levels of these specific safety factormust be adapted for each type of vehicle liable to meetthese special physical effects, levels could result fromprobabilistic considerations ( see § 6.4.2 ) .

Another requirement for these multiphysical effect safetyfactors is to keep possible a verification test in theultimate conditions; it leads to avoid safety factors on"calculation beings" physically inseparable by testconditions as with the present thermal stress safety factorof AIR2004-E and other regulations.

5.4 Safety factors evolution with innovations

5.4.1 The particular case of fly by wire aircraft

Knowing that the flight control system, with a "care freepiloting functions, can protect against limit loadovershoots, a debate may arise as to the pertinence of achange to the safety factor (currently 1.5) ; suchdiscussions come up against great difficulties :

• The current safety factor covers aspects other than theoccurrence of load conditions that are severer thanthe limits ; they involve, amongst others :

✓ potential flaws in the load computation models (forcefields applied to the airframe) in function of loadingconditions (flight mechanics state vector ).

✓ every unknown differences between the airframes inservice and the one that was qualified (non-detectedmanufacturing or material defects, various non-detected corrosion-, fatigue- or impact-induceddamage types, etc…).

• For all of these factors, there are non sufficientlyconclusive probability models available that give theload or structure strength overshoot statisticaldistributions ; we do not know how to quantify thesefactors separately within the global safety factor.

• The global safety factor of 1.5 can be justifiedquantitatively only by the acquirements of

experience, based on observation over half a centuryof a globally satisfactory structural strength ofaircraft in service ; but this safety factor cannot bedecorrelated from the rest of the environment of theused construction techniques, analysis methods andverification process. Any partial change that occurredin the technical environment requires a demonstrationto establish that there is no regression in Safety(cf. qualification rules for composite materials),although this would not mean that any likely gain inone point can be exchanged against a reduction of themargin in another point.

A further element for debate bears on the advantagesthat might be drawn from a potential safety factorreduction:

• For new projects, the potential gain in terms ofstructure mass is likely to be slim, the safety factor-to-mass exchange ratio will remain far belowproportionality (fatigue sizing of metallic parts,design to technological minimal for large areas, areaswith design-sizing aeroelasticity constraints, …).

• The discussion is somewhat more open, for existingand proven by flight service airframes, whenconsidering any specific or circumstance-relatedmaneuver performance characteristics improvement.

5.4.2 Towards probabilistic approaches

At long range a complete reconstruction of structuralanalysis process would be required , to get out of theabove mentioned piling of safety margins, resulting fromignorance of the part, within present global safety factor,assigned to any innovation of design solution or ofanalysis method .

This long range research could be founded on a fullprobabilistic approach, considering all items of airframequalification : loads, types of design ,calculation and testprocess , manufacturing process, flight service use,fatigue & corrosion and any other aging effects, controlplan , …, and human error possibilities everywhere insidethe process .

It is a subject in itself, which could be proposed to furtherRTO discussions .

References :

1. C. PETIAU, M. DE LA VIGNE AnalyseAéroélastique et Identification des Charges en VolAGARD conference proceedings No 373 -"operational load data" - Sienne 1984 .

2. C. PETIAU Evolution de la philosophie descharges de dimensionnement des avions militaires.AGARD report No 815 - "Loads and requirementsfor military aircraft" - Florence 1996