© nCode 2005 Fatigue Design of Vibrating Components Slide 1
Accelerated DurabilityAccelerated DurabilityTestingTesting
Dr. Andrew Halfpenny&
Frederic Kihm
© nCode 2005 Fatigue Design of Vibrating Components Slide 2
Introduction toIntroduction toAccelerated TestingAccelerated Testing
Mission ProfilingAnd
Test Synthesis
Fatigue Design of Vibrating Components Slide 3
What Do We Want From A Durability Test?What Do We Want From A Durability Test?
• Durability test that’s suitablefor the item in question:– a component,– sub-assembly– or a whole vehicle
• Test must replicate the same failuremechanisms as seen in the real world
• Test should be representative of the realloading environment
• Test should be accelerated where possible to reduce project timescales and costs
• Test specification can be used in FE based virtual test or realphysical test
Fatigue Design of Vibrating Components Slide 4
What Steps Are Involved?What Steps Are Involved?
1. Duty or Mission Profiling– Find out what’s expected of the vehicle / component– How long should it last?– Determine the ordinary loads that it’s likely to see every day– Determine the extraordinary loads it might see and be expected
to survive
2. Test Synthesis– Synthesise a test that exhibits the same damage as the Mission
Profile
Fatigue Design of Vibrating Components Slide 5
How Do I Obtain Real Load Data?How Do I Obtain Real Load Data?
In-service Measurement - usingwheel force transducers, load cells,strain measurements, etc. frominstrumented prototypes & customervehicles inrealconditions
Proving Ground / Test Flights /Engine Test Cells –using measurements taken over knownevents with well establishedcorrelation to the real world
Pros.• Good source of data• Much less data to analyse than
aboveCons.• Requires a prototype• Tends to be biased towards extreme
events
Pros.• Good source of data• Much less data to analyse than
aboveCons.• Requires a prototype• Tends to be biased towards extreme
events
Pros.• The best source of data possibleCons.• Requires a prototype• Long record length required• Ideally should include many
statistically representative samples
Pros.• The best source of data possibleCons.• Requires a prototype• Long record length required• Ideally should include many
statistically representative samples
Fatigue Design of Vibrating Components Slide 6
How Do I Obtain Real Load Data?How Do I Obtain Real Load Data?
Analytical modeling - usingMulti-Body Simulationsoftware.
Pros.• Doesn’t require a prototype, fully
analytical simulation• Carried out early in design cycle• Ideal for the new ‘Breed’ product
where no legacy data is availableCons.• Variable quality of data depends a lot
on experience• Needs a prototype test to support it
Pros.• Doesn’t require a prototype, fully
analytical simulation• Carried out early in design cycle• Ideal for the new ‘Breed’ product
where no legacy data is availableCons.• Variable quality of data depends a lot
on experience• Needs a prototype test to support it
Engineering Judgment – Estimate loadsusing experience of what worked in thepast
Pros.• Ideal where long standing experience availableCons.• Many tests have poorly recorded heritage• Often have unknown safety factor• Often intolerant to changes in material or
changes in intended use
Pros.• Ideal where long standing experience availableCons.• Many tests have poorly recorded heritage• Often have unknown safety factor• Often intolerant to changes in material or
changes in intended use
Fatigue Design of Vibrating Components Slide 7
What is Mission Profiling?What is Mission Profiling?
ProvingGround
• A Mission Profile is a set of loading events thatconstitute the expected loads seen by a vehicleover its life.
• Events might be time series records of discreteevents like curb strikes or potholes or extra-ordinary emergency cases (deterministic)
• or PSD records of continuous vibratory loadingsuch as standard road surface, engine inducedvibration, etc. (stochastic)
Mission Profile
x 200 rpts
x 100 rpts
x 100 hrs
+
+
etc…
Fatigue Design of Vibrating Components Slide 8
What is Mission Profiling?What is Mission Profiling?
• The damage on most Automotive structuralcomponents is dominated by deterministic timeevents
• Aerospace components and automotive bracketsare usually dominated by continuous stochasticprocesses
Take off
Cruise
Air combat
Intercept
Cruise
Descent
Land
-10
0
10
20
30
55.8 56 56.2 56.4Time (s)
Acc
eler
atio
n (g
)
-0.2
0
0.2
0.4
100 200 300Time (sec)
Acc
eler
atio
n (g
)
Deterministic
Stochastic
Fatigue Design of Vibrating Components Slide 9
Test Synthesis – Type of TestTest Synthesis – Type of Test
Vibration Controlled
• Test rig driven by an accelerationor displacement input
• Component is fixed only to test rigand is vibrated against its owninertia, ∴ frequency and amplitudemost important
• Force in component:F =a.m
m = massa = acceleration
Fatigue Design of Vibrating Components Slide 10
Test Synthesis – Type of TestTest Synthesis – Type of Test
Rig Controlled(Cyclic or “Closed Load Path”)
• Test rig driven by a displacementinput
• Component is fixed to test rig andan external rigid restraint,amplitude important, frequencynot important
• Typical of many structuralcomponent tests
• Force in component:F = k.x or F = c.v
k = stiffness c = dampingx = displacement v = velocity
Fatigue Design of Vibrating Components Slide 11
Test Synthesis – Type of TestTest Synthesis – Type of Test
Remote Parameter Control (RPC)(Simulation)
• Enter a number of measuredchannels
• Let rig controller determine inputlevels that recreate measured data
• Run test and monitor responses tomake sure they match originalmeasured channels
• Tests are usually inertia reactedbased on whole vehicle or subassemblies, ∴ frequency andamplitude important
Fatigue Design of Vibrating Components Slide 12
Test Synthesis – Route MapTest Synthesis – Route Map
DeterministicStochastic
Quasi-staticDynamic
UniaxialMultiaxial
Test Synthesis Frequency DomainTime DomainPeak-Valley Domain
Dynamic Quasi-static
UniaxialUniaxial
Load ScalingLoad Scaling
Fatigue Design of Vibrating Components Slide 13
Load ScalingLoad Scaling
• Scaling up the load will reduce the test durationexponentially.
• Target life is influenced by endurance limit andonset of local plasticity as well as dynamicresponse of component
• Scaling should be used with extreme care toavoid local yielding and changing the load paths
• Not suitable for most inertia reacted tests
• Destroys Amplitude• Maintains Sequence• Maintains Phase between multiple
channels• Maintains Frequency Content
1 10 100 1 .103 1 .104 1 .105 1 .106 1 .107 1 .108 1 .109 1 .1010100
1 .103
1 .104
Number of Cycles to Failure
Stre
ss R
ange 2UTS
1000 NC1
Scaled Range
Original Range
Real Duration
Test Duration
1/ b
Where b is the Basquin Exponent (gradient of SN curve)This is only approximate!
Fatigue Design of Vibrating Components Slide 14
Test Synthesis – Route MapTest Synthesis – Route Map
DeterministicStochastic
Quasi-staticDynamic
UniaxialMultiaxial
• Peak Valley Extraction• Block load sequence• Statistical Exceedence• Equivalent Sinus
Test Synthesis Frequency DomainTime DomainPeak-Valley Domain
Dynamic Quasi-static
UniaxialUniaxial
Load ScalingLoad Scaling
Fatigue Design of Vibrating Components Slide 15
Peak Valley Extraction (PVX)Peak Valley Extraction (PVX)
• Typical 90% reduction in signallength
• ‘Gate’ small cycles on range,rainflow or fatigue contribution
• Take care with slew rates, etc.
• Maintains Amplitude• Maintains Sequence• Destroys Phase between
multiple channels• Destroys Frequency content
360 Points
36 Points
Fatigue Design of Vibrating Components Slide 16
Block Load Sequence (with mean)Block Load Sequence (with mean)
• Very simple test specification• No sequence information ∴
cannot distinguish overloadeffects in EN analysis norcrack closure effects in LEFM,etc.
• Maintains Amplitude• Destroys Sequence• Destroys Phase between
multiple channels• Destroys Frequency content
Data available in lists of:• Max, Min, Cycle count• Range, Mean, Cycle count• Range, R, Cycle count• Etc…
Fatigue Design of Vibrating Components Slide 17
Block Load Sequence (no mean)Block Load Sequence (no mean)
• If mean stress effects are notimportant then you can outputin simple load blocks.I.e. 306 cycles at 200MPafollowed by 56 cycles at11MPa, etc…
• Maintains Amplitude• Destroys Sequence• Destroys Phase between
multiple channels• Destroys Frequency content
Fatigue Design of Vibrating Components Slide 18
Statistical ExceedenceStatistical Exceedence
• Very simple test specification• G-Exceedence test data is
widely used in the Aerospaceindustry
• Maintains Amplitude• Destroys Sequence• Destroys Phase between
multiple channels• Destroys Frequency content
Fatigue Design of Vibrating Components Slide 19
Damage Equivalent Constant AmplitudesignalDamage Equivalent Constant Amplitudesignal
• Much easier signal to reproduce ontest rigs
• Damage Equivalent and not CyclesEquivalent like Block Load Sequence
• Makes the comparison of the damagepotential of different signals easy
• Destroys Amplitude• Destroys Sequence• Destroys Phase between
multiple channels• Destroys Frequency content
Both sets of signalscreate same damage! … 100 000 or more
cycles
Fatigue Design of Vibrating Components Slide 20
Test Synthesis – Route MapTest Synthesis – Route Map
DeterministicStochastic
Quasi-staticDynamic
UniaxialMultiaxial
• Peak Valley Extraction• Block load sequence• Statistical Exceedence• Equivalent Sinus
Test Synthesis Frequency DomainTime DomainPeak-Valley Domain
• Is it proportional i.e.dominant plane?
• Multiaxial PeakValley Extraction
Dynamic Quasi-static
• Increase Load Frequency
UniaxialUniaxial
Load ScalingLoad Scaling
Fatigue Design of Vibrating Components Slide 21
Resultant / Critical Plane AnalysisResultant / Critical Plane Analysis
• Proportional multi-axial, or caseswith a dominant fatigue plane
• Establish critical plane• Eliminate non-damaging channels• Determine a single drive channel
with fixed proportions betweeninputs or align component on theuniaxial test rig at a given angle Resultant Load Plane
Fatigue Design of Vibrating Components Slide 22
Multi-axial PVXMulti-axial PVX
• Maintains phase relationshipbetween multiple channels bykeeping points that correspondwith a peak or valley in adifferent channel
• Ordinary peak valley wouldapply all peaks / valleyssimultaneously thereforechanging the load paths
• ‘Gate’ small cycles
• Maintains Amplitude*• Maintains Sequence• Maintains Phase between
multiple channels• Destroys Frequency content
Fatigue Design of Vibrating Components Slide 23
Increase Loading FrequencyIncrease Loading Frequency
• Doubling the frequency will halfthe test time
• Limit acceleration to max 1/3 firstmode natural frequency
• Not suitable for inertia reactedtests
• Maintains Amplitude• Maintains Sequence• Maintains Phase between multiple
channels• Destroys Frequency content
1/3 * natural frq
Fatigue Design of Vibrating Components Slide 24
Test Synthesis – Route MapTest Synthesis – Route Map
DeterministicStochastic
Quasi-staticDynamic
UniaxialMultiaxial
• Peak Valley Extraction• Block load sequence• Statistical Exceedence• Equivalent Sinus
Test Synthesis Frequency DomainTime DomainPeak-Valley Domain
• Is it proportional i.e.dominant plane?
• Multiaxial PeakValley Extraction
• Buffered FatigueAnalysis:Damage Editing
Dynamic Quasi-static
• Increase Load Frequency
UniaxialUniaxial
Load ScalingLoad Scaling
Fatigue Design of Vibrating Components Slide 25
Buffered Fatigue AnalysisBuffered Fatigue Analysis
Rainflow cycle count and calculate fatiguedamage for each cycle proportioning damage tothe start and end times of the cycle
Divide the time signal into buffers and Removenon or low damaging buffers of the original drivesignals and splice remaining data together using awindowing envelope to maintain frequencycontent and prevent high slew rates and impactringing
Take a time signal at the critical location(s)
Fatigue Design of Vibrating Components Slide 26
Buffered Fatigue AnalysisBuffered Fatigue Analysis
• Maintains all key attributes• Typical acceleration 50-80%
depending on amount of damageto be retained and number offailure locations assessed
• Can be used with uniaxial ormultiaxial fatigue solvers
• Can be used with a potentialdamage solver
• Maintains Amplitude• Maintains Sequence• Maintains Phase between
multiple channels• Maintains Frequency content
• You should always compare thedamage before and afterreduction to make sure it’s stillequivalent
Fatigue Design of Vibrating Components Slide 27
Test Synthesis – Route MapTest Synthesis – Route Map
DeterministicStochastic
Quasi-staticDynamic
UniaxialMultiaxial
• Peak Valley Extraction• Block load sequence• Statistical Exceedence• Equivalent Sine
Test Synthesis Frequency DomainTime DomainPeak-Valley Domain
• Is it proportional i.e.dominant plane?
• Multiaxial PeakValley Extraction
• Buffered FatigueAnalysis:Damage Editing
Optional Load ScalingOptional Load Scaling
Dynamic Quasi-static
Later
• Increase frequency of Time Series
UniaxialUniaxial
Fatigue Design of Vibrating Components Slide 28
The Final Test SpecificationThe Final Test Specification
• If Deterministic factors are dominant the test can bedefined wholly in the time domain
• If Stochastic factors are dominant the test can bedefined wholly in the frequency domain
• Where components have both Stochastic andDeterministic events (or Extraordinary events) thenTime and Frequency tests may be required
Test Profile
x 2 hrs
x 4 rpts
x 2 hrs
+
+
etc…
x 4 rpts
+
Fatigue Design of Vibrating Components Slide 29
SummarySummary
• Loading environment can be splitinto Deterministic and Stochastic
• Deterministic loads should berepresented in time domain byblock loads or edited time signals(Chassis, Steering, Suspension &Drivetrain)
• Stochastic loads should berepresented in frequency domainby PSDs (Engine Components,Brackets, Electronics & Ancillary)
• Load inputs can be Uniaxial orMultiaxial
• Components can behaveStatically or Dynamically
• Tests can be Vibration Controlled,Rig controlled or RPC.
• Loading can be Inertial, Viscose orDisplacement induced.