Krylov Subspace Methods for Model Order Reduction in Computational Electromagnetics

This paper presents a model order reduction method via Krylov subspace projection, for applications in the field of computational electromagnetics (CEM). The approach results to be suitable both for SISO and MIMO systems, and is based on the numerically robust Arnoldi procedure. We have studied the model order reduction as the number of inputs and outputs changes, to better understand the behavior of the reduction technique. Relevant CEM examples related to the reduction of finite element method models are presented to validate this methodology, both in the 2D and in the 3D case.