we have performed recently a static non linear FE analysis of quite a simply shaped rubber component.
Predicted strains are in the order of 50% for a typical load case, say load case 1.
Deflections for this load case are predicted within 1% accuracy.
Then the task was to predict the failure location should the load be increased.
This was not easy. I have used three criteria, the maximum principal stretch (allegedly Kawabatas criterion), maximum true stress and maximum strain energy.
Following a test it was realized the failure does not happen close to the locations indicated by the mentioned criteria.
It seems to happen instead in an area where none of the three peaks, but on the other hand there is a very significant strain gradient.
I am thinking there must be a mistake in the FE modelling, unless this behaviour has a sense for a Nitrile rubber, i.e. failing where the stress gradient is at a maximum (and the principal stretch is maybe above a threshold?)
Is this possibility familiar to any of you? Or is there simply something wrong with the model / test?
Thank you very much for your help
Interesting study. Here are a few comments:
(1) First, it appears that you are interested in monotonic loading to failure only, is that right?
(2) Was the stress and strain state almost homogeneous? It might be possible that your validation simulation can accurately predict the macroscopic deflection, but not actually predict the local stress and strain distributions in the rubber. What material model did you use? Are you feeling good about the accuracy of that model?
(3) Did you calibrate the material model only to 50% strain, and then attempt to simulat the response to failure (>50% strain). If so, that can lead to problems.
(4) What was the highest values of your three failure criteria in the simulations? What where the corresponding values at the failure location?
(5) There are other failure criteria that might work better. I kind of like the max chain stretch failure condition.
(6) I have never heard of failure caused by a high gradient. Did it happen to be a hydrostatic tension at the actual failure location?
thanks for your comments.
My replies are:
1) Yes, only monotonic loading to failure is of interest.
2) The strain state is far from homogeneous. A 4 constants MR material law was employed and of course the possibility you mention about less accurate local strains is taken very seriously. I do not have much to support my confidence in the model but the deflections of a number of points, which are fairly accurate. Also the work done by the testing machine up to failure is predicted quite well (less than the deflections though, around + 6%).
3)The model is calibrated for strains around 50%, but deflections have been verified even close to failure and the accuracy does not diminish.
4) The highest figure at the failure location is the strain energy density, only 75% of the maximum strain energy which is picked up by the model elsewhere. The principal stretch is even lower (56%), wjhile the true stress is somewhere in between (68%).
I have not found much on the web about the max chain stretch criteria, do you have any source to mention? I have found an article on UHMWPE where the authors have tried also criteria which I thougth where more suited to metals ( Mises stress, Tresca stress, hydrostatic stress, Coulomb stress), I might have a go.
5) The stress state is not exactly hydrostatic but it is true that at the failure location the strain state is more triaxial than elsewhere (in the sense strains are closer to each other). This might represent an inconvenience for maximum strain, for example, but I was hoping this would reflect on the strain energy value.
Thank you again for your input and looking forward for final comments and hints.
Have a nice weekend
It certainly sounds like an interesting dilemma.
I am curious, how high was the hydrostatic stress at the failure location compared to the Mises stress at the failure location, and compared to the Mises stress at failure in uniaxial loading? I am just trying to get a feel for how severe the triaxial stress is at the failure location.
The max chain stretch paper that you found might have been a paper that I wrote a few years ago. At that time I performed a study focusing on the failure of thermoplastics in monotonic loading, and I found that the max chains stretch was more accurate than other traditional stress and strain failure conditions.
thanks for your comments.
The hydrostatic stress at the failure location is of the same order of magnitude than Von Mises. In its turn Von Mises stress at the failure location is roughly 80% to the value in uniaxial tension.
Is there any source available for the maximum stretch criterium?
All the Best
Heres a [URL= https://polymerfem.com/forums/showthread.php?t=34 ]reference[/URL] to the chain stretch failure condition.