(Dr. Yogendra Joshi, advisor)
"Thermal-Based Multi-Objective Optimal Design of Liquid Cooled Power Electronics"
A computationally efficient modeling framework for predicting and optimizing the thermal performance of liquid cooled power electronic modules has been developed to be used as a rapid scouting tool for preliminary layout designs. The framework is based on an integration of a compact multi-dimensional, transient thermal model with a genetic algorithm based multi-objective optimizer. By combining a simplified thermal impedance network model with the impulse response of multiple devices on a common substrate, an overall transient thermal assessment of a power electronic module can be performed. This approach enabled the compact model to rapidly determine the thermal conditions at any specified time for a power module structure consisting of multiple substrate layers and power devices. This methodology has been employed for analyzing power electronic modules that do not operate long enough to reach steady-state conditions. Thereby, designing for specified transient conditions, over conservative estimation of design features associated with steady-state design can be avoided. Examples of Pareto solution sets of power electronic module layouts based on optimizing maximum module temperature; overall coolant pressure drop, and power device placement were developed. The layouts were validated with full field numerical simulations, and compared to steady state results to determine the advantages of avoiding conservative design features. The rapid computational times make this methodology useful in analyzing a vast sample space of possible design alternatives. These advantages allow a designer to quickly identify potential optimal designs and perform detailed numerical simulations on them, rather than spend time searching for these optimal layouts through extensive computations.