Condition monitoring of SiC power assembly by using time-series analysis of acoustic emission during power cycling tests

Condition monitoring using acoustic emission (AE) technology is a widely used method for detecting failures and analyzing working conditions in various industrial facilities. However, there has been limited research conducted on utilizing AE for monitoring the condition of power devices. In this study, we implemented AE-based online monitoring during a power cycling test of a SiC power assembly and successfully captured a significant amount of AE signals. Through the analysis of the power assembly failures, it was determined that the AE signals originated from stress waves caused by solder layer cracking. We also compared the AE monitoring to traditional condition monitoring methods that rely on thermal signals such as junction temperature and thermal resistance. The obtained AE signals demonstrated superior sensitivity in identifying initial cracking. Furthermore, we investigated the variations in AE peak frequency centroid and AE energy as the power cycling test progressed, which provided multiple dimensions for understanding the failure process of the power assembly.