A High-Efficiency IGBT Health Status Assessment Method Based on Data Driven

As a critical device of power electronics, the health status and residual life of the insulated gate bipolar transistor (IGBT), as well as its high-efficiency assessment should be concerned. High-efficiency assessment methods can provide dangerous warning for transportation equipment to avoid catastrophic accidents. This article focuses on the IGBT solder layer failure. We use the finite element simulation combined with the small current junction temperature measurement method to obtain the relationship between the internal structure degradation of the solder layer and the sensitive parameters of the external port under different load conditions. According to the characteristics of port parameters, a data processing method called multiple weight function (WF) network (MWFN) is developed, and the parameter monitoring platform is built. Finally, an IGBT health assessment method based on MWFN is formed. The main advantage of this method is that the best WF is introduced into the neural network to solve the time-consuming problem of updating the high-dimensional weight matrix. In the case study, the active power cycling test was carried out. We obtained the solder layer delamination images by using the 3-D tomography system and the corresponding port parameters by monitoring platform. From the perspective of evaluation results, the proposed method not only performs as well as other mainstream data-driven methods in terms of evaluation accuracy but also has significant advantages in terms of running time.