TY - GEN
T1 - An improved method of rail health monitoring based on CNN and multiple acoustic emission events
AU - Zhang, Xin
AU - Wang, Kangwei
AU - Wang, Yan
AU - Shen, Yi
AU - Hu, Hengshan
N1 - Publisher Copyright:
© 2017 IEEE.
PY - 2017/7/5
Y1 - 2017/7/5
N2 - Rail health monitoring plays an important role in the railway system, and how to accurately obtain the rail state is very significant for the railway safety. This paper proposes an improved method of rail health monitoring based on convolutional neural network (CNN) and probability analysis of multiple acoustic emission (AE) events. By tensile testing machine, AE signals with safe and unsafe states are obtained. The CNN method of deep learning (DL) is employed to classify the defects, and the results of CNN are also compared with that of other methods. From the output of CNN, the probability values of each sample belonging to a class can be obtained, and then the improved classification method based on multiple events is investigated. The detection errors caused by one-time classification are eliminated, and the classification accuracy are improved. The results illustrate that the proposed method can effectively recognize the rail state for rail health monitoring.
AB - Rail health monitoring plays an important role in the railway system, and how to accurately obtain the rail state is very significant for the railway safety. This paper proposes an improved method of rail health monitoring based on convolutional neural network (CNN) and probability analysis of multiple acoustic emission (AE) events. By tensile testing machine, AE signals with safe and unsafe states are obtained. The CNN method of deep learning (DL) is employed to classify the defects, and the results of CNN are also compared with that of other methods. From the output of CNN, the probability values of each sample belonging to a class can be obtained, and then the improved classification method based on multiple events is investigated. The detection errors caused by one-time classification are eliminated, and the classification accuracy are improved. The results illustrate that the proposed method can effectively recognize the rail state for rail health monitoring.
KW - Acoustic emission
KW - Deep learning
KW - Probability analysis
KW - Rail health monitoring
UR - https://www.scopus.com/pages/publications/85026754055
U2 - 10.1109/I2MTC.2017.7969693
DO - 10.1109/I2MTC.2017.7969693
M3 - 会议稿件
AN - SCOPUS:85026754055
T3 - I2MTC 2017 - 2017 IEEE International Instrumentation and Measurement Technology Conference, Proceedings
BT - I2MTC 2017 - 2017 IEEE International Instrumentation and Measurement Technology Conference, Proceedings
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 2017 IEEE International Instrumentation and Measurement Technology Conference, I2MTC 2017
Y2 - 22 May 2017 through 25 May 2017
ER -