Deim-based pgd for multi-parametric nonlinear model reduction

Abstract

A new technique for efficiently solving parametric nonlinear reduced order models in the Proper Generalized Decomposition (PGD) framework is presented here. This technique is based on the Discrete Empirical Interpolation Method (DEIM)[1], and thus the nonlinear term is interpolated using the reduced basis instead of being fully evaluated. The DEIM has already been demonstrated to provide satisfactory results in terms of computational complexity decrease when combined with the Proper Orthogonal Decomposition (POD). However, in the POD case the reduced basis is a posteriori known as it comes from several pre-computed snapshots. On the contrary, the PGD is an a priori model reduction method. This makes the DEIM-PGD coupling rather delicate, because different choices are possible as it is analyzed in this work.