Multi-scale Convolutional Neural Network for Fault-locating of High-speed Train Bogie

As the high-speed train (HST) is developing towards safer and faster, the stricter safe operation mechanism of HST has become a research hotspot. The performance degradation and fault state of the key parts of the HST bogie directly lead to the change of the vibration form of the car body and the bo...

Full description

Saved in:
Bibliographic Details
Published in2021 CAA Symposium on Fault Detection, Supervision, and Safety for Technical Processes (SAFEPROCESS) pp. 1 - 6
Main Authors Wu, Qin, Na, Zhang, Longguan, Li, Kaifu, Liu, Suimei, Huang, Deqing
Format Conference Proceeding
LanguageEnglish
Published IEEE 17.12.2021
Subjects
Convolution
Convolutional neural network
Convolutional neural networks
Data models
Fault diagnosis
Feature extraction
High-speed train bogie
Representation learning
Safety
Vibrations
Online AccessGet full text

Cover

More Information
Summary:As the high-speed train (HST) is developing towards safer and faster, the stricter safe operation mechanism of HST has become a research hotspot. The performance degradation and fault state of the key parts of the HST bogie directly lead to the change of the vibration form of the car body and the bogie, which also seriously threatens the safety of the train operation. In this paper, the multi-scale convolutional neural network (MS-CNN) is proposed to realize multi-scale feature fusion in convolutional neural network, which can realize the intelligent fusion of multi-channel information, intelligent extraction of fault features and intelligent fault-locating of HST bogie The experimental tests of the proposed method are conducted using the data set acquired by SIMPACK via the HST model CRH380A. Overall, the proposed MS-CNN achieves an average accuracy of 91.4%, demonstrating the remarkable performance of fault-locating of HST bogie.
DOI:10.1109/SAFEPROCESS52771.2021.9693709