a “real Crash test” (with high impact speed, self-contact, large plastic deformations, rupture) is currently not possible on the platform. This would require a fully explicit solver.
The nonlinear solvers we are currently using are fully implicit solvers (with some explicit time schemes). Compared to an explicit solver they have a very high accuracy also for large time steps, but since they have to fulfill the balance equations in an implicit way (every equation term is evaluated at the current time step and iteration) they are computationally very expensive for fast, high speed impacts.
But we are currently working towards enabling an explicit solver on SimScale.
The geometric nonlinearity setting is used to decide if large rotations or large strains can appear in the simulation. If they are expected this property should be set to nonlinear (in the end this triggers the use of a rotation-invariant deformation tensor). In our model we would not expect large deformations, but large rotations could appear.