Classification of impact signals from insulated rail joints using spectral analysis

Abstract

Insulated Rail Joints (IRJs) are a railway track component that generates impact noise and requires close maintenance. Instrumented Revenue Vehicles (IRVs) developed by the Institute of Railway Technology at Monash University measure the interaction between the vehicle and track. Impact signals were measured and post-processed from vibration sensors located on the side-frame at IRJ locations. Wavelet analysis was used to interrogate the non-stationary impact signals. Wavelet energy was used as the wavelet feature extraction techniques in the frequency domain. The wavelet energy impact signatures were clustered using multi-signal discrete wavelet transform clustering. These clusters classified the empty and loaded conditions of the wagon from the vibration response. Frequency identifications were created from the clustering and the severity of the impacts in the frequency domain could be determined from the cluster numbers.