19. Summary of results

The results obtained on all the models of this test case are satisfactory, both in 2D and in 3D. On points \(A\), \(B\), \(C\) and \(D\), we get a relative error of less than 1% compared to the results of GRECO. On the other hand, for point \(E\), the relative error, of the order of 2.5%, remains acceptable.

From a calculation time point of view, it can be seen that the penalization method is quite fast. However, its use is not obvious since the convergence and the quality of the results obtained are conditioned by the penalty coefficient.

From a qualitative point of view:

• in 2D, there is a positive effect on the results of quadratic meshing compared to linear meshing (see models A and B),

• in 3D, we see that calculations with quadratic meshes do not improve the results. This is explained by the fact that the number of cells has been reduced by 50% compared to the linear mesh to have an equivalent number of nodes. In all cases, only the use of TETRA10 or HEXA27 meshes is recommended in discrete formulation, the HEXA20 meshes needing to be linearized (they therefore lose their interest).

With regard to the L modeling, inclined plate, good convergence and satisfactory results are noted.

The method of ADAPTATION =” CYCLAGE “or ADAPTATION =” TOUT” makes it possible to better converge cases of grazing pressure.

Finally, for the continuous method (M, N, P, P, P, Q, Q, R, R, V, W and X models), results are obtained in 3D with relative errors that are slightly larger than those obtained by the penalized methods. These errors come from the blocking in the DZ direction, which is not complete in continuous formulation in models that do not use the SANS_GROUP_NO_FR feature for excluding knots from friction.