A Multi-Objective Predictor Approach for IGBT Trigger Pulse Intermittent Fault Diagnosis

The reliability studies of Insulated Gate Bipolar Transistor (IGBT) are significant in guaranteeing the stability of system operation. This paper concentrates on the abnormality between the IGBT module and the controller, investigating the diagnosis method for its trigger pulse intermittent fault (IF). The proposed method focuses on the single-phase pulse rectifier, constructing a data-driven multi-objective predictor to simultaneously monitor four-way IGBT pulses. The Nonlinear Autoregressive with Adaptive Exogenous Inputs (adaptive NARX) model establishes the time-varying architecture for the IGBT pulse signal, while the Random Vector Functional Link (RVFL) determines its input-output mapping relationship. Additionally, periodic fault assessment criteria were developed to provide a clear demonstration of pulse randomness fault severity. Through experimental and robustness tests, the method is evidenced to be effective in detecting the occurrence of IF and achieving the security maintenance of the system for early faults.