Forming Simulation Fails when unloading model


I am trying to perform a simulation of a sheet metal ring being formed and I want to calculate the springback in the ring.

The forming part of this problem runs just fine. As you can see from Run 6.

The error occurs when I start to remove the load on the sheet metal.

As you can see this error occurs at t=1.003, the first time step where I am removing the load. I have tried using half the maximum displaced value ( -.001875) and I still get this error at the point I start removing the load. At this point the material is just starting to bend so I don’t think this is a deformed element or mesh issue.

I have also tried changing the the time so the maximum displacement happens at t=.5 and the error still happens at the first time step that starts to unload the model.

I tried changing the loading from a table to cosine function and I still get the same error. If did however, shift the point at where the error occurred. You can see this in Run 9 and the error occurred at t-.5.

I did find an old post on the Code Aster forum about this issue (from 2012).

Indeed when unloading an elasto-plastic structure you need to tweak some numerical parameters :

  • since unloading occurs with elastic stiffness, using the TANGENTE matrix is usually not advised (unless as you pointed out you do very small time steps and use line search)
  • the recommended way of treating elastic unloading is by setting PREDICTION=‘ELASTIQUE’ in order to use the elastic stiffness matrix for the first Newton iteration of each time step, this will help the algorithm get back on track and then the TANGENTE matrix will be OK for the remaining iterations (particularly is plastic hardening occurs during the unloading)

I did try changing this parameter but it has no effect on the simulation.

I would appreciate any help anyone can provide. My project lnk is here.

Thank you,

Hi Christopher!

First of all, thanks for using the forum, I hope everything is going alright!

By looking at the solver log, I can see that the Elastic matrix is actually being used at the first Newton iteration:

So the setup does have an effect. I think that it is just not enough to properly compute the spring back.

This error message is very tricky in the sense that it lends you to believe that the problem is in the plasticity integration routine. What I have found is that sometimes the source of the error is a badly constrained body, and somehow this condition causes the error in the material routine.

This model reassures my position, because I can see that the sheet has no vertical constraint, and the spring-back action can make it move too fast.

All that being said, I think the solution is to add an elastic support to the sheet. I would select orthotropic properties and set the horizontal springs to zero. For the vertical direction, use the smaller value that gets rid of the error (might need a couple iterations to find the optimum).

Please let me know if you can overcome the issue!

Hi @ggiraldof,

I spent all weekend trying to get this to work. This morning I finally gave up and posted my problem on the forum. The very last run I started before I made my post actually completed successfully!

The only things is, Run 10 that finished successfully should have been the same as Run 6. Now I need to go look to see what I changed.

Thanks for taking a look at this. You always have great insight into nonlinear problems and I enjoy learning from your experience.


1 Like

Thanks a lot for the kind words!

I will check your successful run and see what can I also learn from it.