Chip-Level Electrothermal Stress Calculation Method of High-Power IGBT Modules in System-Level Simulation

Abstract

Although with good robustness, insulated gate bipolar transistor (IGBT) modules suffer from catastrophic failures due to excessive electrothermal stress in field use. In view of the coupled and multitimescale electrothermal dynamics of IGBT modules, the conventional simulation methods are difficult to calculate the stresses accurately and efficiently. Therefore, this article proposes an artificial neural network (ANN) based method to calculate the chip-level stress of IGBT modules in system-level circuit simulation. The first contribution in this article is that the research focus goes from the entire device to chips within the device. Correspondingly, the uneven dynamic current and thermal coupling effects among chips are fully considered based on the established uneven power loss and three-dimensional (3-D) thermal models. The second one is to solve the contradiction between simulation efficiency and accuracy in multitimescale operation analysis through the joint application of IGBT physical model and ANN. The third one is to propose the complete chip-level stresses calculation and automated modeling processes, based on multisoftware cosimulation. An inverter experimental prototype was established to verify the accuracy of the proposed calculation method. The comparison results show that the method can accurately describe the transient junction temperature, overcurrent and overvoltage stresses, thus providing an effective support for the safety design of converters.