A Feature Transfer Learning Method for Open-Circuit Fault Diagnosis of Single-Phase Pulse Rectifier

Owing to the complexity of the operating environment, high-speed trains experience a significant proportion of open-circuit faults in pulse rectifiers, which severely impacts their safe operation. Additionally, the operational data of high-speed trains is characterized by an extremely limited number of fault samples, posing a challenge to the accuracy of data-driven fault diagnosis methods. To address these issues, this article proposes a feature transfer learning-based approach that leverages a large volume of readily available rectifier fault data to assist target rectifiers with scarce fault samples in identifying new feature space representations. This method facilitates fault classification and diagnosis, achieving promising fault diagnosis performance.