Rubber experimental testing for combined load
Hello dear friends,
Im trying to predict with FEA the stiffness of a rubber spring.
The spring will get a axial pretension of x mm and while is compressed a lateral displacement of y mm.
My question is, for modelling do I need other experimental testing beside the uniaxial experimental test?
Usually this test was enough.
I have a similar problem. I am modeling gasket compression between two thin metal plates (gasketed plate heat exchanger). Until now I have used uni-axial tensile test and uni-axial compression test.
In a recent project we had some problems with gasket sealing and we could not reach the required pressures due to leakages. We then used a different and a bit stiffer material and the problem got solved. So my aim is to find out why the first material and the numerical model did not achieve the expected (and calculated) results. Among many variables I chose to take a look into material models and the experimental data need to fit the hyperelastic model.
From the literature I have read (a lot of advice is given by Axel Products Inc., check: [url] http://axelproducts.com/ [/url]) it is often suggested that for complex strain states being analyzed it is desirable to have experimental data from multiple states of strain. Therefore you need additional experimental data to model the behaviour correctly.
I had a discussion with my former professor about the tests needed for my numerical model. He suggested that additional shear and biaxial tension test would not increase my model accuracy nor give me any advantage. However it would be better if a compression experiment would be made in such a way that it resembles the actual geometry in state of strain. I could then measure the stress - strain relation similar as it appears in real case. By real case I mean a gasket between two thin metal plates, compressed together.
For now I would somewhat disagree for the case that a shear test is not needed. When you perform a uni-axial compression test you get frictional force at the contact of the pressing plates and the test piece. Therefore shear stresses are generated. This frictional force has an effect to the stress-strain response of the material. The same could be assumed for a gasket pressed between two heat plates in a heat exchanger assembly - shear stresses are also generate since the physical problem is practically the same. From this point of view I think shear stress experimental data is necessary since shear stresses will appear. I have not yet studied the numerical models and tested how additional data affects the numerical results so I am just hypothesizing at this moment.
Thats my two cents. If anyone with more experience can elaborate further on this topic it would be useful.
I would like to throw my thoughts in. Your problem is quite complex. When compressed above a certain threshold, rubber in compression can possibly be well tricky to model. Pushed to the limit, the idea that volumetric and distortional modes can be decoupled can be shown faulty. And you are of course right that the deformation close to the gasked edge will be very complex. I dare do not agree with your former professor, but for reasons different than yours (at the end of the day, I think it is often overlooked that a tensile test imposes shear: as a matter of fact, there are only shear strains in an incompressible rubber..). What I think would be necessary for your problem is the ability to characterise shear under severe hydrostatic pressure. But before this challenging task I have one question: you state that the numerical model did not achieve the expected (and calculated) results. You were modelling leakage, so what were your expected results ? Unless you were actually modelling a liquid in contact with the gasket, what were the results you were expecting to be able to verify leaking? Maybe you were monitoring contact pressures between the gasket and the steel plate and comparing to your fluid pressure? Or what else? This can be important. Maybe the solution is already there and only needs interpreting. And when you said you switched to a stiffer material, how did the compressibilities compare? Do you also use lubricated and confined compression tests or just compress a (laterally free) cubic specimen with a plate?
All the best
Thanks for the response.
1. Could you please elaborate on the statement there are only shear strains in an incompressible rubber
2. No I was not modeling leakage but just the compression of the gasket and looking at the contact pressure. Then, as you stated, compared the contact pressure against the fluid pressure. The model I used (and also my former coworker) until now does not incorporate the fluid pressure. We are also monitoring the metal plate bending after the end of the compression too see how much the plates are bent (plastic deformation).
3. Regarding the material: The new material we used has on average from 25% to 58% higher compression stresses at the same deformation rate. The compression samples were measured without any additives on the contact surfaces, with grease and with talc powder. When measuring without any additives and with talc powder at some point the rubber sample slipped and caused a sharp decrease of stress and increase of strain on the diagram (relatively low values). We are using a standardized ISO 7743 compression test.
I admit that the model is not covering some aspects of the physical problem but I am also limited with my resources (2 core licence). However the method I am using currently works if the whole assembly of metal plates and gaskets is stable. By stable I mean that during the assembly the metal plates dont shift in any direction (or at least shift minimally). But that is an entirely different problem.
I think I also need to make a sensitivity analysis of the numerical model to see how the friction coefficient between metal and rubber affects the results, gasket position (assembly is not always perfect), different parts of the real model etc. Seems like I will be writting a thesis. 😀
Anyway, I would be happy to hear your thoughts.
in reference to your points:
1. any deformation on elastic materials can be decomposed in volumetric (volume changing) and distorsional (shape changing, shear) deformation. By saying that rubber is incompressible, it is certainly not meant that it is rigid in compression: certainly steel is stiffer in compression than rubber. What is meant by that is that the volumetric stiffness is so much bigger than the distorsional one, that practically the first can often be neglected. In this sense, rubber is a liquid. Of course one can use the term shear strain when referring to how the angles between material lines change: in this sense a tensile deformation seems shear-less. But this definition depends on your reference system: that is why for example the pure shear test piece, essentially a very wide tensile tests, get its name from. The topic is addressed in much more rigour in many excellent books.
2. I understand better now. It is difficult to help. You are certainly right in that sensitivities analysis might help. The problem with those very high compression is, as I said, that a lot of assumptions might get too inaccurate: for example, hydrostatic and shearing stress might not decouple like commonly assumed. Or maybe, the very small viscoelastic relaxation in compression becomes not negligible. The stiffer a rubber is, the bigger impact little errors might have. I understand the test you performed: I would still perform the compressive test in fully confined, constrained set-up, not so sure if it covered by any Standard but would clearly tell you, by comparison with the test you performed, how well compression is captured. You can test and model both the set-ups, lubricated and laterally constrained, and compare. The bonded compression test (I think, mentioned in the same ISO standard you quote) might be also very useful: it will not give you a material property, but you might FE model it and see how your comparison works: in the bonded compression test the barrelling induces plenty of shear strains and might be mimiciking what happens in practice. As an additional validation, maybe a more complex test such as measuring the deflection and force exterted on the upper metallic plate on your gasket might or might not be feasible.
This is what I can say without touchin the gasket...I think very few could say that gasket-modelling is easy, so best of luck
Thanks for the additional comments. Much appreciated.
I guess Ill start researching into the matter now.