Rail health monitoring using acoustic emission technique based on NMF and RVM

In order to detect the health status of high-speed railway, this paper proposes a detection method based on non-negative matrix factorization (NMF) and relevance vector machine (RVM) by acoustic emission (AE) signals. AE signals are obtained by tensile testing machine and AE data acquisition system. According to the stress-time curve, AE signals with safe state and unsafe state are obtained. Based on the frequency spectrum analysis of AE signals, the ratio of each frequency component relative to maximum frequency component is used as a feature vector to distinguish safe and unsafe states. Vectors with compressed and optimized features are obtained based on NMF, and these vectors are used to train and test the classifier by RVM. The classification accuracy of 10-folds cross validation on the whole dataset is up to 96%. The results illustrate that the proposed method can detect the safe status of rail effectively.