This was presented at the NAFEMS World Congress 2025, held in Salzburg, Austria from May 19–22, 2025.

Abstract

We presented our overall automated workflow for power module development at NAFEMS Multiphyics Conference in Singapore, and focused on the EMC aspects of the workflow at the NAFEMS Simulation in Electronics event. This highly technical presentation demonstrates the complete design workflow of a full bridge IGBT module starting with EDA design, electrothermal, thermal, thermomechanical, electromagnetic, optimization and ROM extraction, all of which is performed in a CAD environment. The goal of the optimization is to maximize heat transfer while minimizing pressure drop, inductance, and warpage within the module to enhance its reliability and performance. Particular emphasis is placed on the numerical techniques involved, how the simulation model fidelity can be confirmed and enhanced with transient thermal testing, and a boundary condition independent reduced order model created for use in system and electrothermal simulation solutions. The simulation scope includes Joule heating within the direct bonded copper substrate within the model, the bond wire interconnects and the bus bars. The focus of this study is not the design of the module itself, however, we have included the simulation of the electromagnetic performance, as loop inductance is a parasitic effect that needs to be minimized. Indeed, the design is a trade-off, balancing variations in current density, stray inductance, electrical resistance, thermal resistance, mechanical stress and physical design constraints required for manufacturability. The module is an internal reference design, authored within an EDA environment and transferred into mechanical CAD, preserving its parametric definition and easing design space exploration and performance optimization. The same model can be used for non-linear mechanical stress simulations that use the temperature fields predicted without the need for translation. Wire bond fatigue is a common failure mechanism due to fatigue in the wires from heating and cooling. Design space exploration can involve the simultaneous variation of a large number of input parameters at the same time, which is recommended to ensure that a globally optimum solution can be found. The technique used combines a number of search algorithms to ensure that the scheme does not get stuck at some local optimum. As the number of designs simulated increases, AI is used to accelerate the process where the accuracy is acceptable, with the accuracy increasing as more designs are simulated. The work that will be reported here replicates similar work done in conjunction with a major European automotive OEM, where the module design is authored in mechanical CAD and fully parameterized, so in both cases the starting point for all simulations is a parameterized CAD model.