Intelligent detection of rail corrugation using ACMP-based energy entropy and LSSVM

In this paper, an intelligent method to diagnose rail corrugation based on signal decomposition and entropy theory is proposed. The axle box acceleration signals are first decomposed into several components with different frequency bands by ACMP, EEMD and MODWT. By comparison, ACMP is able to successfully extract rail corrugation component from original signal without mode mixing. Energy entropy is then introduced here to quantify the degree of the rate of energy concentration. The analysis results show that the energy will change when rail corrugation occurs and the entropy will become small. It has been also proved that the entropy difference of rail corrugation and normal signal based on ACMP is the most significant. In addition, to intelligently diagnose rail corrugation, we combine energy entropy with energy index and the first mode energy, regarded as the input feature vector of LSSVM, to distinguish rail corrugation from mass data sets. It is obvious that the accuracy of ACMP-based technique is the highest.