Detection of Rail Wheel Flats using Wavelet Approaches

Abstract

Sudden application of heavy traction/braking torque induces flats in rail wagon wheelsets. Flat wheels generate high frequency impact forces sufficient to cause severe damage to the rail head surface. Early detection of flats would facilitate re-profiling the wheels before they cause serious damage to critical rail components, such as switches, crossings, and insulated rail joints. Although there are many sensors available in the market to detect/record wagon accelerations, there are no powerful signal processing tools readily available to detect the presence of wheel flats from the recorded acceleration signatures. This study presents two wavelet approaches to overcome the difficulties in the on-board monitoring and detection systems of rail wheel flats using vibration signals. Signal average techniques, wavelet local energy average concept, and wavelet decomposition are employed in this study. A Matlab-Simulink based dynamic simulation system capable of modeling the wheel flats and track irregularities is also developed for predicting the wheelset/bogie frame acceleration time series. An analysis of the numerical simulation results demonstrates that the methods proposed in this study are effective for the on-board monitoring of wheel flats of sizes smaller than the condemning limits.