A fault diagnosis method for single-phase rectifiers based on open set transfer learning

By constructing a domain-adaptive transfer learning network based on open set transfer learning, and combining feature center matching with the Openmax method, the problem of identifying unknown faults in single-phase rectifiers in open set scenarios is solved, achieving high-precision and fast fault diagnosis.

CN121009289BActive Publication Date: 2026-06-30SOUTHWEST JIAOTONG UNIV
View PDF 2 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
SOUTHWEST JIAOTONG UNIV
Filing Date
2025-07-04
Publication Date
2026-06-30

AI Technical Summary

Technical Problem

Existing technologies struggle to accurately identify unknown faults in open-set scenarios during single-phase rectifier fault diagnosis, and traditional methods have limited generalization capabilities, failing to meet the demands for rapid response and high accuracy.

Method used

An open-set transfer learning approach is adopted, which constructs a domain-adaptive transfer learning network, combines feature center matching with the Openmax method, and utilizes the Weiber distribution and bimodal feature storage to achieve the identification and classification of unknown faults.

Benefits of technology

It effectively identifies unknown faults, improves the accuracy and stability of fault diagnosis, and enhances the system's rapid response capability and adaptability to complex operating conditions.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN121009289B_ABST
    Figure CN121009289B_ABST
Patent Text Reader

Abstract

This application discloses a fault diagnosis method for single-phase rectifiers based on open-set transfer learning, comprising: collecting experimental and simulation data of the single-phase rectifier under normal and fault conditions; constructing a domain-adaptive transfer learning network and training the network using source domain data; inputting target domain data into the trained network and extracting feature vectors and classification score vectors from the target domain data; calculating the distance to the known fault category centers based on the feature vectors and estimating the probability that the distance belongs to the tail of the known fault distribution based on the Weiber distribution and the Openmax method; if the tail probability exceeds a preset threshold, it is determined to be an unknown fault and triggers the circuit breaker protection mechanism; otherwise, the known fault type with the highest probability in the classification score vector is output. This invention introduces an open-set transfer learning model, combined with feature center matching and Openmax method classification correction, to achieve effective detection and isolation of unknown faults and improve the transfer robustness of the model's target domain.
Need to check novelty before this filing date? Find Prior Art