Computation algorithms for efficient coupled electromagnetic–thermal device simulation

Abstract

Coupled electromagnetic thermal field problems using nonidentical finite-element meshes for the magnetic and thermal discretisation, as encountered for instance in the simulation of electromagnetic energy transducers such as motors and transformers, require the application of nonlinear iterative solution algorithms. The paper gives an overview of the commonly used weakly coupled block iterative Picard methods and relaxation techniques. Strongly coupled Newton methods, both with explicit and implicit Jacobian matrix computations are discussed. Local as well as global convergence issues are treated. In this respect, the use of an alternative continuation technique, the pseudotransient coupled algorithm, using transient calculations in the frequency domain by means of an envelope approach, is discussed. The performance of the algorithms is compared using representative benchmark problems with both moderate and strong interaction. This leads to indications and a choice table on how to select appropriate algorithms for these coupled problems.