Domain-Adversarial Transfer Learning-Driven Adaptive Diagnosis Method for Multi-Faults in Subway Train Air Conditioning Systems

Read full text Summarise with VIRAT
Authors: Tongle Zhang, Yang Chuang, Chen Huanxin, Zhang Xue
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Efficient fault diagnosis of metro train air-conditioning systems is essential for reducing energy consumption and ensuring passenger comfort. This paper proposes an unsupervised transfer learning method based on a domain-adversarial neural network (DANN) to address the challenges involving diverse feature distributions and complex faults. Data obtained from three single faults (condenser fouling, ventilation fouling, and refrigerant leakage) and their concurrent combinations are collected from a multifunctional test vehicle under various operating conditions and compressor frequencies. The DANN achieved accuracy values ranging from 97.30%-98.90% for single faults and 77.80%-86.70% for concurrent faults. Uniform manifold approximation and projection (UMAP) and SHapley Additive exPlanations (SHAP) analyses revealed the underlying reasons for the less accurate concurrent fault diagnosis. Compared with the two baseline transfer learning models, DANN exhibited markedly smaller performance fluctuations, maintaining high accuracy, even under large feature distribution shifts and overlapping concurrent fault features.

Fields of Study & Subjects

Machine Fault Diagnosis Techniques Fault Detection and Control Systems Refrigeration and Air Conditioning Technologies

Publication details

Journal / source
DOAJ (DOAJ: Directory of Open Access Journals)
Publication date
1 October 2026
Language
ZH