Online algorithm for early stage fault detection in IGBT switches

The early detection of incipient faults is desirable in mission-critical applications such as shipboard propulsion drives. This paper presents an online condition-monitoring approach for detecting early stage faults in IGBTs. The proposed algorithm extracts important device features (i.e. on-state resistance, gate charge, etc.) and compares them to healthy values recorded over a range of operating conditions. The algorithm is based on principal-components analysis (PCA). An experimental implementation in an IGBT-based drive is described, and results recorded with two different faults over a range of operating conditions are presented. The scheme integrates well with new FPGA-based gate drives and provides a powerful alternative to rules-based fault detection.