Parameterized model order reduction for fast transient electromagnetic simulations

Abstract

Transient Electromagnetic problems constitute an area of significant investigative effort. The principal computational issue in these problems is the solution of large system of differential algebraic equations (DAE's) obtained after Finite Element (FE) discretization. Model Order Reduction (MOR) Techniques provide a mechanism to generate reduced order models from the detailed description of the original FE network. This is achieved by using moment matching techniques, where the reduced order model matches the moments of the original system to approximate the response with a low order transfer function. However, these numerical techniques all conserve the original system moments only with respect to frequency. While this provides a significant CPU cost advantage when performing a single frequency sweep, a new reduced order model is required each time a parameter is varied in the structure under study. This necessitates the use of parametric MOR strategies so as to expedite optimization and design space exploration cycles. In this work, we present a methodology for transient electromagnetic field simulations through parameterized model order reduction (pMOR). The proposed methodology is illustrated for a generic system with promising results and a significant saving in computational effort.