This is an issue when you use second order elements in combination with node to surface discretization. The matter is explained in the pdf file I have included.
This is an issue when you use second order elements in combination with node to surface discretization. The matter is explained in the pdf file I have included.
Contact problems in elasticity are non-trivial to approximate computationally. Algebraically, after the finite element discretisation, they involve solving a constrained optimisation problem. Therefore, relative errors can arise from either the finite element approximation or the numerical optimisation procedure. I suggest that, as you increase the order of the finite elements, try to decrease the mesh-size (of course, this will increase the computational cost of your simulation). If the problem persists, then you are probably in a case where FEM solvers require further improvement/refinement. However, various reliable numerical algorithms are available.
Here is some nice work that you might find relevant in general
Article The Contact Dynamics method: A nonsmooth story