Diagnosis of IGBT Module Bond Wire Independent of Temperature Based on BP Neural Network

Yanyong Yang,Xiaofeng Ding,Zhenyu Shan,Pinjia Zhang

DOI: https://doi.org/10.1109/ecce53617.2023.10362371

2023-01-01

Abstract:The fault diagnosis of the insulated-gate bipolar transistor (IGBT) module bond wire is crucial to monitor the IGBT module's state and avoid the converter's catastrophic fault. However, most of the existing methods of IGBT module bond wire diagnosis are affected by temperature. In this paper, a technique based on Back-Propagation (BP) neural network for IGBT module bond wire diagnosis is proposed. In this method, two parameters affected by both IGBT temperature and bond wire state, namely IGBT on-state voltage and IGBT gate voltage overshoot, are combined by BP neural network. The influence of temperature on bond wire failure diagnosis is eliminated. The IGBT on-state voltage and IGBT gate voltage overshoot are monitored and extracted by a double pulse test. Lots of experimental data are collected. A BP neural network is trained for IGBT module bond wire diagnosis. Experimental results verify the feasibility of the proposed method. The proposed IGBT module bond wire diagnosis technique has the advantages of not being affected by temperature and accurate monitoring results.