Intelligent Diagnosis of Open Circuit of IGBT in Electric Drive Inverter Based on Transformer

Abstract

The insulated gate bipolar transistor (IGBT) is the core switching device in electric drive inverters. Its open-circuit faults pose a significant threat to system reliability due to their stealthy nature and potential to induce secondary failures. This paper proposes an intelligent IGBT open-circuit fault diagnosis method based on the Transformer architecture, which directly uses normalized three-phase output voltage and current sequences without manual feature engineering. A field-oriented control drive system model is built in MATLAB/Simulink, and a fault injection matrix is employed to systematically simulate 22 operating states, including single-switch and double-switch open-circuit faults. A sliding window mechanism is applied to construct a labeled dataset. A Transformer encoder tailored for multidimensional time-series signals is designed to capture long-range dependencies and fault correlations across electrical signals. After systematic hyperparameter tuning, the model achieves 100% classification accuracy on an independent test set. The confusion matrix and t-SNE visualizations confirm that the proposed method autonomously learns highly discriminative fault features, resulting in clear separation of all 22 state clusters. This study provides an end-to-end technical pathway free of prior knowledge for online intelligent maintenance of inverters in electric drive systems.