7. Tips on using GVNO

It is preferable to use modeling GVNO with the traditional Newton method in the case of damage calculations that do not present strong instabilities (snap-backs). Indeed, Newton’s method is faster but it does not allow to overcome snap-backs without the use of an additional algorithm, such as piloting, which has not yet been developed here. If not, it is best to use the « secant » method, triggering a PAS_MINI_ELAS large enough to ensure that the condition is checked at each time step. This method makes it possible to overcome instabilities and therefore to always converge on a solution. However, the number of iterations required for convergence can be quite large (e.g. 800 iterations). It is therefore necessary to allow a large maximum number of iterations in the command file (1200 approximately). The speed of convergence depends on the law used and also on the temporal discretization. Too long steps of time can lead to a convergence that is far too difficult. This method is very effective for quadratic behavior laws in damage and displacement since we then work with a robust alternating minimization algorithm.