It is generally known that non-associated plasticity models (models in which the yield function is different from the plastic potential function, for example using a dilatancy angle less than the friction angle in the Mohr-Coulomb model) do not provide a unique solution. The use of such models in the finite element method can lead to numerical problems in the form of alternating mechanisms, irregular load-displacement curves and mesh alignment. So far, these problems have barely occurred. However, with the increasing accuracy in terms of large number of high-order elements and stricter error control, calculations tend to suffer from such problems nowadays more than in the past. Plaxis is working together with the research community towards a solution for this fundamental problem. Meanwhile, some suggestions are provided here to overcome such problems in existing versions only if such problems occur:

• Use 6-node elements instead of 15-node elements
• Use mesh refinements only in areas where really needed
• Use a tolerated error of 0.02 instead of the default of 0.01