A new rail crack detection method using LSTM network for actual application based on AE technology

Abstract In order to use acoustic emission (AE) technology in the actual application of rail crack detection, an important problem to be solved is how to overcome the noise interference of wheel-rail contact movement. In this paper, a new method is proposed to eliminate noise interference and detect rail crack signal based on AE technology, which has a two-level structure with Long Short-Term Memory (LSTM) network. At the first level, an improved noise model from multiple kinds of noise signals is built by the LSTM network. This model is used to eliminate the known noise signals. At the second level, the model of crack signal is built to remove the unknown noise interference from the denoised signal of the first level. Based on the proposed two-level structure, the crack signals can be detected. All the AE signals are acquired from the real noise environment of railway. Meanwhile, the detection ability of the proposed method is analyzed and verified. The results demonstrate that the proposed method is effective to detect crack signals in actual application. It can provide a useful guidance for AE detection of rail cracks.

[1]  Hongyan Ma,et al.  Defect detection and location in switch rails by acoustic emission and Lamb wave analysis: A feasibility study , 2016 .

[2]  Jürgen Schmidhuber,et al.  Parallel Multi-Dimensional LSTM, With Application to Fast Biomedical Volumetric Image Segmentation , 2015, NIPS.

[3]  Jürgen Schmidhuber,et al.  Long Short-Term Memory , 1997, Neural Computation.

[4]  Mathieu Feuilloy,et al.  Acoustic emission pattern recognition approach based on Hilbert–Huang transform for structural health monitoring in polymer-composite materials , 2013 .

[5]  Yi Shen,et al.  Detection method based on Kalman filter for high speed rail defect AE signal on wheel-rail rolling rig , 2018 .

[6]  Kristoffer Bruzelius,et al.  An initial investigation on the potential applicability of Acoustic Emission to rail track fault detection , 2004 .

[7]  Yi Shen,et al.  Acoustic emission detection of rail defect based on wavelet transform and Shannon entropy , 2015 .

[8]  Hengshan Hu,et al.  An improved AE detection method of rail defect based on multi-level ANC with VSS-LMS , 2018 .

[9]  Yi Shen,et al.  An analysis of the simulated acoustic emission sources with different propagation distances, types and depths for rail defect detection , 2014 .

[10]  Alan Hase,et al.  Correlation between features of acoustic emission signals and mechanical wear mechanisms , 2012 .

[11]  K. Ono.  Application of acoustic emission for structure diagnosis , 2011 .

[12]  Erik Marchi,et al.  Non-linear prediction with LSTM recurrent neural networks for acoustic novelty detection , 2015, 2015 International Joint Conference on Neural Networks (IJCNN).

[13]  Guoqiang Peter Zhang,et al.  Time series forecasting using a hybrid ARIMA and neural network model , 2003, Neurocomputing.

[14]  D. Bianchi,et al.  Wavelet packet transform for detection of single events in acoustic emission signals , 2015 .

[15]  Robert Lewis Reuben,et al.  Rail–wheel interaction monitoring using Acoustic Emission: A laboratory study of normal rolling signals with natural rail defects , 2010 .

[16]  Yoshua Bengio,et al.  Learning long-term dependencies with gradient descent is difficult , 1994, IEEE Trans. Neural Networks.

[17]  Miao He,et al.  A new method to classify railway vehicle axle fatigue crack AE signal , 2018 .