A Practical Method for the Reduction of Linear Thermo-Mechanical Dynamic Equations

Abstract

Linear thermo-mechanical equations are widely used for the dynamic modeling of electric motors/generators or gas turbines. In order to use them in the context of a digital twin, real-time capable versions of the models must be achieved. In principle, model order reduction techniques for coupled thermo-elastic physics are known. However, commercial tools and even open-source tools allow only limited access to the necessary information in terms of coupling terms. This paper aims at providing an algorithm for the reduction of thermo-elastic equations in the framework of given software tools. After sampling and running some simple test cases in the offline stage, the reduced coupling term can be obtained and directly applied in the online stage to solve the reduced thermo-mechanical equations without intrusion into the commercial FEM software. Moreover, the numerical residual due to sampling and grouping techniques is also discussed in the paper. Basically, an algorithm similar to operator inference [15] is applied for extracting the coupling term. The complete workflow for extracting the coupling matrix is demonstrated on the open-source software Code_Aster.