Fatigue Studies?

I have a request from a customer that I don't think is possible & having a hard time understanding

Can anyone help me with this

They want to know the difference between (2) designs, A - Steel Frame & B - Aluminum Frame

A Design calculated accuracy is: 1.8/1.5 Cp/Cpk; and after 200K cycles the estimated Cp/Cpk is 1.3/1.1…

B Design calculated accuracy is: 1.4/1.2 Cp/Cpk; and after 200K cycles the estimated Cp/Cpk is 1.0/0.9…

And so on…

Please help

SolidworksSimulation
Answers 8
SF
SF 2013-08-08 Best Answer

cp/cpk are we talking about this: http://www.ask.com/question/cp-cpk-formula

JP
JP 2013-08-09

Yes , Jared, I don't think there is anyway in SWX simulation to forcast the Cp/Cpk using fatigue studies

But I wanted to get some feedback prior to telling my customer this isn't possible

SF
SF 2013-08-09

here's how i look at it. you have a part that is constructed. you test say 10 of them, they should fall in that statistical configuration.

then you cycle them 200k and test 10 of them, they should fall in that statistical configuration.

in simulation, if you test a part 10 times, the result will be the same because the construction of each part and the materials of each part are perfect. so statistics doesn't come into play.

also with fatigue, it is not possible to create a "fatigued" component with a fatigue test. you'd need to use nonlinear. but even then, you're performing a perfect test. so statistically, the component is perfect.

not sure if that makes sense. the short answer is no. but wnated to make sure you understand why.

when it comes down to it

static analysis tells you the displacements, stresses and strains

fatigue tells you the damage and number of cycles based on the max stress compared to the part's SN curve

nonlinear could tell you about how a "fatigued" component behaves but 200k cycles would result in a long analysis.

JP
JP 2013-08-09

That makes sense, I have tried to perform a non-linear fstigue study using a creep material in the past & didn't have much success

Any suggstions on how to perform  "nonlinear could tell you about how a "fatigued" component behaves but 200k cycles would result in a long analysis"

SF
SF 2013-08-09

its not a creep problem its a plasticity problem. so run the analysis with a plasticity model. and setup the BC to be a sine wave for 200k cycles. then at the end of that, add your normal loading to see how it behaves.

but remember, this is just pure material fatigue. often there are construction issues and material differences that are the cause of failures that you can't account for (without some work) in simulation.

JP
JP 2013-08-09

"with a plasticity mode-" change to non linear?

"BC" Bend curve?

Shouldn't the load be on & off?

What would be involved with "there are construction issues and material differences that are the cause of failures that you can't account for (without some work) in simulation"

SF
SF 2013-08-09

you need nonlinear because you need a "time effect" in your boundary condition. it has to change over time. if you plug in a sine curve, the load will be on and off.

plasticity is one of the material models. it allows for plastic deformation.

construction issues and material issues, this one is going to require your knowledge of the product. consider that maybe some bolts don't get tightened down completely and that leads to a load path change. you could setup your "perfect analysis" and then another one with less bolt preload and see how they differ. or if you're casting this or injection molding, maybe part of the material is weaker. you could setup another one with a chunk of your part as a different material. so basically YOU have to predict what is going to cause the statistical differences.

feel free to contact me at jared@hawkridgesys.com if you want us to scope out this analysis and then work with you to teach you at the same time as we get it completed.

JP
JP 2013-08-13

Thnaks Jared