Unsupervised Dual Convolutional Autoencoder Models for Efficient Group Anomaly Detection of HST Bogies via Domain Adversarial Learning

The bogie system is known as the “legs” of high-speed train (HST), various failures will inevitably occur under large disturbances, high speeds, and heavy loads. Abnormal detection (AD) is an important means to detect the health status of its key components. Nevertheless, the cross-correlation of fa...

Full description

Saved in:
Bibliographic Details
Published in:Journal of physics. Conference series Vol. 2999; no. 1; pp. 12029 - 12035
Main Authors: Chang, Yuanhong, Zhong, Shuncong, Pan, Tongyang, Xie, Jingsong
Format: Journal Article
Language:English
Published: Bristol IOP Publishing 01.04.2025
Subjects:
Anomalies
Cross correlation
High speed rail
Learning
Undercarriages
ISSN:1742-6588, 1742-6596
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:The bogie system is known as the “legs” of high-speed train (HST), various failures will inevitably occur under large disturbances, high speeds, and heavy loads. Abnormal detection (AD) is an important means to detect the health status of its key components. Nevertheless, the cross-correlation of failures causes the confusion of health data and fault data under highly-coupled components, which leads to the issue of false detection and missing detection. Hence, this paper proposes a dual convolutional autoencoder (d-CAE) network combined with unsupervised domain-adversarial learning for group anomaly detection of bogies. Firstly, the d-CAE adopts temporal window aggregation to construct initial inputs. Afterwards, the domain-adversarial learning strategy is utilized to make the d-CAE realize multi-level encoding and reconstruction of multi-channel time-series. Finally, a parameterized dynamic AD index is designed to accurately establish the health sample guided abnormal decision boundary. The experimental results indicate that the d-CAE is competitive in the aspects of detection accuracy and robustness compared with the state-of-the-art methods.
Bibliography:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:1742-6588
1742-6596
DOI:10.1088/1742-6596/2999/1/012029