How to run experiement verification

[oso]

I am doing some FEA analyses and also have a chance to run tests to verify my models.

I have two questions:

1. What do you measure when you run verification test? Displacement or strain?

2. How to verify the permanent deformation start to happen? Some localized yield stress may not be able to observe permanent deformation. How can I justify that my stress level in model will cause permenent deformation?

Thanks for your help!

 

[GregLocock]

We don't correlate the static results, because it is often VERY difficult to do in any meaningful way (there are a few exceptions), particularly in the 'interesting' parts of the structure. It is much easier experimentally to correlate the modal results.

However, given you limited choices, I'd try and correlate strains in an area that can be accurately strain gauged, ie not necessarily where the stresses are highest.

I strongly suggest that you talk to whoever is doing the physical testing, you are about to run into the new brave exciting world of real testing.



Cheers

Greg Locock

 

[oso]

Greg:

Thanks for your advice and glad to know you are also in Automotive industry:->.

Actually, my first correlation test is not too bad on displacement. It's inside 10% range. By the way, I am working on brake pedal asm. The specification requires certain deflection under certain load.

I am still more wondering on how to predict when the permanent deformation will happen. Foe example, my FEA model shown there is localized yield area on 1200N. Then large yield area on 1500N. If I observed that permanent deformation happens under 2500N, is my model still correct?

I understand that material yield stress may vary. But I am still very interested to know the experience of other people on permanent deformation prediction (both FEA and experiment). I also planning to do some non-linear analysis to predict permanent deformation and correlate my model to experiment results. Is that the correct way to proceed?

Thanks!

Oso

 

[GregLocock]

The reason I don't think displacement correlation is especially useful is that static stiffness is a product of the material properties, which you have probably got right, and the BULK structure of the model, which is insensitive to things like fillets and other stress raisers.

I think in some respects you are asking an unanswerable question. Real structures yield locally the first time they are exercised, but it doesn't matter, that's just a question of assembly misalignment getting worked out. I suggest you talk to whoever set the spec about this, that is , force to get to first yield is not inherently a useful characteristic to report.

Witthout giving anything away if you don't want to - is this brake pedal plastic? Australian? AI? BA? If the answer is yes to all four questions then I can probably give you some contacts to talk to.

Cheers

Greg Locock

 

[GSC]

Hi,
Although some time has passed since you posted the message, maybe I can still add something useful. I was working a lot on large deformation problems on soft materials. As a rule of the thumb, for validation you work with values which you can directly measure and preferably also available for direct comparison to reduce the resultant errors.
Q1: In respect of the above I would measure global displacements and local strains however I don't quite understand the load vs displacement criteria if permanent deformation exist.
Q2: I agree with Greg about the unanswerable thing. It is hard to capture the yield on the way you may want. An explicit model could give you an idea what may happen at impact if you manage to set realistic conditions for the time. Otherwise you have to trust your implicit model and more importantly the material properties and safety factors you work with.
Cheers
GSC