Reduced Order Modelling in Non-Linear Structural Mechanics

Finite Element Methods are well established in structural mechanics, however, in many engineering applications fast numerical evaluations of parametric solutions are required, e.g. for optimisation, sensitivity analysis or uncertainty quantification. Model Order Reduction (MOR) are currently under investigation for drastically reduction of the computational effort in comparison to FEM simulations. However, to this point no clear guidelines on the treatment of non-linearity have been developed. In this project kernel based methods will be investigated with regard to their performance on tackling non-linear structural dynamics problems, in particular for structural failure to loss of material resistance. A goal oriented comparison of different branches of kernel-based methods, i.e. kernel POD, support vector regression and Kriging with special emphasis on damage and plasticity is performed.