so, for a while now i have been trying to get this simulation running. the idea is simple: I want to crush a hyperelastic cylinder with two plates. all went well until i started playing with friction. in real life simulations you normally hava a liitle gap that stabilises the elastic cylinder and causes it to appear more like a barrel, instead of just a more flattened cylinder. the simulation only works until the firtion coefficient is 0.3, after that it crashes.
here is what i have so far
3 solid parts (with the top and bottom tool both having extruded mesh) with reference points and the very top and very bottom of the tools (created from between two points)
3 solid sections (2 with steel for the crushing tools, and the elastic cylinder is a Mooney-Rivlin (the parameters for M-R I took from a friend that got them out of and experiment))
2 rigid body contraints for the whole parts of the crushing tools
a surface interaction type hard, with the friciton coefficient
2 contact pairs & 2 BCs
the first one for the top tool that does the crushing - it has a U3 displacement of -2mm
the second one is fixed for the bottom tool - where it touches the cylinder
static step with parameters in the photo
i have tried eveything - with additional fixed surfaces, with changing contact parameters, surface values, both static and dynamic steps. i just can’t think of anything else, so all help will be greatly appreciated.
sorry for the photo, but i could have only added one
Do you mean that it stops converging ? Friction is highly nonlinear and really bad for convergence, especially when the coefficient is high (> 0.2).
You should also fix the rotational DOFs of the top reference point.
But more importantly - why don’t you use symmetry and model only a quarter of this system ? Or just define an axisymmetric model (although 2D elements may cause other issues) ? It would also prevent some rigid body motions, which are often the reason for nonconvergence when some parts rely on contact for stabilization.
Hi, I have worked in the past in the rubber industrie, and what we did to overcome the friction uncertainty was make the real test using fine sand paper glued to the plates, so we remove completely the friction in the test and FEA. As we made lot of test with these standard samples (shorter than your picture), two samples for every batch of rubber, static and dynamic properties, the sand paper get “polished” or coated with the wax that some rubbers exude, we change it everyday to be sure of the non slipping condition in test and FEA.
And for the FEA, for rubber use hexa meshing, will be better for convergence and less elements (quicker results). If is only compression test use a quarter or axissimetric modeling as other says.
Using a sandpaper gives the same result as having the rubber glued to the plates, so you can model only the rubber body and play with the BC for getting the result, no need to use contact at all.
We use a very fine sandpaper glued with double face tape. It was independent of the hardness type. You can make this simple test, put sand paper and measure the diameter of the rubber part in loaded state, if the dimeter is the same, then you have reached the non slippling condition (no friction at all).
