Title
Optimising Power Semiconductor Thermal Simulation via Finite Element ModelingPublication Date
2025Other institutions
https://ror.org/00wvqgd19Ingeteam (Spain)
Version
PostprintDocument type
Conference ObjectLanguage
EnglishRights
© 2025 IEEEAccess
Open accessPublisher’s version
https://doi.org/10.1109/ECCE-Europe62795.2025.11238575Published at
Energy Conversion Congress & Expo Europe (ECCE Europe) Birmingham, 1-4 September 2025Publisher
IEEEKeywords
Semiconductor device modelingThermal modeling
FEM
Abstract
This article presents a Finite Element Modeling (FEM) framework for thermal analysis of power semiconductor modules, combining simulation accuracy with enhanced computational efficiency. Power losses ... [+]
This article presents a Finite Element Modeling (FEM) framework for thermal analysis of power semiconductor modules, combining simulation accuracy with enhanced computational efficiency. Power losses are calculated using a universal estimator validated against Infineon's IPOSIM tool and are applied as heat sources in a COMSOL Multiphysics ® 3D thermal model. The geometry and thermal structure of the power module are modeled and validated by comparing the junction-to-case thermal resistance (Rjc) with datasheet values. To reduce simulation time and memory usage, two strategies are introduced: the application of geometric symmetries and the replacement of complex heat sink structures by thermally equivalent simplified geometries. Additionally, LiveLink TM for MATLAB ® enables full automation of iterative simulations and parametric studies. Three functionalities are implemented: (i) dynamic adjustment of time step to accelerate steady-state convergence and enable highresolution analysis, (ii) recalculation of power losses based on updated temperatures, and (iii) internal geometry modifications to simulate ageing effects such as solder voids. The proposed workflow achieves a simulation time reduction of more than four orders of magnitude compared to fixed-step approaches, while preserving accuracy in thermal distribution. This approach is well suited for long mission profile evaluation and provides a foundation for future integration of physics-based reliability models and lifetime prediction. [-]


















