Nonlinear elasticity and normal force / soft limiter problem

[bengt]

Hi, dear Elmer team,

In my current model a hyperelastic solid (neo-hooke, ElasticSolver) is getting vertically compressed (stepwise displacement of the top-plane's nodes in -coordinate 3). The vertical displacement of the (curved) bottom surface is restricted by means of the soft limiter, i.e. the nodes of the bottom surface don't move below a certain z value (sort of simulating hard contact). This purely displacement-based model runs perfectly. However, now changing to a "mixed" formulation comprising a normal force (i.e.: pressure) imposed on the top surface, the simulation oscillates from the start and then diverges. If the soft limiter is replaced by the standard dirichlet bc (bottom surface vertically entirely fixed, i.e. displacement 3 = 0), the simulation converges. But utilizing the soft limiter just doesnt work.
Is this a known limitation that I'm just not able to understand, or do you know any trick that could resolve this issue?
I'd really appreciate your suggestions.
Kind regards
bengt

[nguyent]

I am facing the same challenge using a force on a electrostatic cantilever and a soft limiter. Making the timestep very small can help keep ElasticSolver from diverging during the nonlinear iterations, but I'm still not getting my system into equilibrium. At some point it diverges again, at which point I change the timestep again. It's rather time consuming, and I'm not sure it will get me an equilibrium state. If I can show it succeeds, at some point my coworkers and I want to write an adaptive timestepper that is a function of the convergence.

I really hope there's a better way though.