Research on Multi-Objective Optimization Design of Double-Sided Cooling SiC Power Module Based on Intelligent Algorithm

Double-sided cooling power module has low parasitic parameters and excellent heat dissipation performance, which is one of the development directions of power modules. However, its electrical, thermal, and mechanical properties are contradictory, so achieving the complete optimization of all performance is difficult. As a new packaging structure type, the double-sided cooling power module lacks a systematic multi-objective optimization design method. Therefore, this paper proposes a multi-objective optimization design method based on an intelligent algorithm. Firstly, the structure of the double-sided cooling SiC power module was designed, and the material of each key component within the module was determined. The electrical, thermal, and mechanical properties of the designed power module were analyzed by parasitic parameter simulation and finite element electro-thermal coupling simulation. The parasitic inductance was calculated using ANSYS Q3D, and the junction-to-ambient thermal resistance was calculated using COMSOL. The mechanical properties, including die attach stress and chip stress, were calculated using COMSOL considering the temperature gradient in the heat conduction process. After that, the influence of the size parameters on the performance indexes of the power module was studied. There are three key points: (1) h2 has a large effect on the parasitic inductance through eddy current effect; (2) a negative relationship between h1 and the thermal resistance exists because a thicker copper layer can spread more heat horizontally; (3) the chip stress optimization and the die attach stress is contradicted. Secondly, a multi-objective optimization method based on an intelligent algorithm was proposed using the method of parametric modeling and simulation. Based on the randomly generated size parameters, hundreds of module samples were simulated, and the results were used as training samples for the artificial neural network. The functional relationship between size parameters and performance indexes can be obtained, which can speed up the calculation of the performance index and ensure calculation accuracy. After obtaining the objective function, a genetic algorithm is used to solve the multi-objective optimization. A three-objective optimization of die attach stress, parasitic inductance, and thermal resistance was conducted. The Pareto front is a curve because thermal resistance and die-attach stress have the same optimization direction. Then a three-objective optimization of chip stress, parasitic inductance, and thermal resistance was conducted. The Pareto front is a curved surface because these performance indexes have different optimization directions. To achieve customized optimization, a stricter size constraint was set, and a four-objective optimization was conducted including die-attach stress, chip stress, parasitic inductance, and junction-to-ambient thermal resistance. A solution that can improve all performance exists. Finally, based on four-objective optimization, the power modules before and after optimization were manufactured, and their performance indexes were tested and compared. The parasitic inductance was measured using an LCR meter, and junction-to-ambient thermal resistance was measured using a thermal characteristic test platform. The results show that the multi-objective optimization method can significantly improve the overall performance of the double-sided cooling SiC power module, the loop parasitic inductance is reduced from 7.475 nH to 6.489 nH, and the thermal resistance is reduced from 0.373 K/W to 0.355 K/W. © 2023 Chinese Machine Press. All rights reserved.