High-Accurate Condition monitoring using Digital Gate Control and Convolutional Neural Networks

This paper presents an investigation into the real-time monitoring of Insulated Gate Bipolar Transistor (IGBT) modules using digital gate control (DGC) and Convolutional Neural Networks (CNNs). The system detects junction temperature (Tj) and emitter current (Ie) through gate voltage (Vge) waveform analysis, eliminating the need for external physical sensors. By employing digital gate control, the waveform signals are optimized to enhance the sensitivity and accuracy of the CNN model. Furthermore, the estimated current and temperature are continuously provided for health diagnostics, enabling the early identification of degradation or faults - critical for predictive maintenance. The proposed gate driver control method achieved 100% accuracy in estimating the temperature and current levels of IGBTs, offering comprehensive monitoring under various load conditions and contributing to the development of more intelligent and robust power electronics systems.