Reduced-Order Modelling for Coupled Thermal-Mechanical Analysis and Reliability Assessment of Power Electronic Modules with Nonlinear Material Behaviours

This paper presents a compact and time-efficient reduced-order modelling method for thermal-mechanical analysis and material nonlinearity study of power electronics modules (PEMs). Generally, the thermal-mechanical analyses in research are conducted utilising a sequential coupling method, where the thermal model is solved first, and the temperature distributions are used as a load in the structural (mechanical) system. The direct-coupling method has been employed in the present thermomechanical analysis to evaluate the thermal and structural governing equations together, i.e., to solve for its thermal and directional deformation distributions, which are degrees of freedom (DOFs) of the coupled system. A new approach, utilising the Krylov-subspace- and Arnoldi-based model order reduction (MOR) technique, Newmark method and Newton-Raphson al-gorithm within the reduced-order modelling framework, has been developed and demonstrated here for analysing material nonlinearity of PEMs. The reduced-order model (ROM) solutions agree exceptionally well with the full-order (FOM) solutions, showing excellent consistency in inelastic strains and plastic work results for materials with time-independent and time-dependent nonlinearities. The MOR method offers a remarkably compact model with a ROM order of just 20x20 for reduced-order computation compared to a FOM order of about 400,000x400,000 and up to 83% reduction in computational time.