Three-dimensional Thermal Modeling Method of Power Device Chip Based on Intelligent Algorithm

Thermal management of power devices has always been the subject of research. A three-dimensional thermal modeling method is proposed in this paper. Firstly, based on the principles of finite element analysis (FEA), the dataset of heat sources and temperatures is established. Secondly, time-dependent long short term memory (LSTM) neural network is used to achieve prediction of transient temperature distribution. Finally, the three-dimensional thermal distribution of the chip is predicted by using FEA. The proposal of this method can provide a beneficial reference value for the stress of chips and solder layers, because there is a temperature gradient in three-dimensional temperature.