Anomaly Detection for Insulated Gate Bipolar Transistor(IGBT) under Power Cycling using Principal Component Analysis and K-Nearest Neighbor Algorithm

摘要

Insulated Gate Bipolar Transistor (IGBT) is a power electronic transistor used in medium to high power applications such as hybrid cars, railway traction motors, switch mode power supplies, and wind turbines. As more IGBTs find their application into larger and complex systems, the ability to detect and predict failures in IGBTs can provide a key advantage in driving down cost of maintenance and improving system availability and safety. This paper briefly discusses the common failure modes found in IGBTs under power cycling along with the experimental setup. Several electrical parameters are extracted and analyzed for fault using principal component analysis (PCA) and k-nearest neighbor (KNN) classification. The proposed algorithm is successfully shown to detect faults just before the IGBTs enter a final degradation stage toward failure.