An Improved Method for Parametric Model Order Reduction by Matrix Interpolation

Abstract

Parametric model order reduction (PMOR) can effectively reduce analysis time and memory cost for finite element models. However, the calculation precision of the previous methods is not usually high enough. In this paper, an improved method is presented to approximate parametric models with higher accuracy. First of all, Krylov subspace projection-based method generally adopted in model order reduction (MOR) is introduced. Based on that, the method to adjust the local reduced-order models (ROMs) is described briefly. Then an approach to interpolate the compatible ROMs on the tangent space is presented in detail and the calculation flow chart for parametric model order reduction (PMOR) of finite element models is provided. In the end, the analysis results for the moving coil of electrical-dynamic shaker demonstrate that the generated parameterized ROM can approximate the direct ROM with higher fidelity than the original method.