A new DFT-based dynamic detection framework for polygonal wear state of railway wheel

Abstract

Polygonal wear of railway wheels will accelerate the failure of critical structure. It is of great significance to monitor their state in service. However, affected by the frequently changed operating speed and track irregularity, the operation process of the vehicle is typically non-stationary. The traditional analyzing methods are not ideal for dealing with the non-stationary processes, causing the accurate detection of wheel polygon wear state very hard. This paper proposes a dynamic detection framework based on iterative modified discrete Fourier transform to solve the above problem. Firstly, extract the relatively stationary short-time signals from the original signal by setting appropriate stationarity test conditions. Secondly, obtain the wheel polygon's vibration frequency and period by taking the iterative calculation and analysis. Thirdly, truncate the extracted short-time signal again to get a new short- time signal representing an integer multiple of the wheel polygon's vibration period. Finally, combined with the wheel polygon's dynamic characteristics, take the frequency domain analysis to the new short-time signal and realize the accurate estimation of the wheel polygon wear state. The simulation and experiment results show that the detection framework can effectively eliminate the influence of non-stationary factors and better solve the inherent defects of traditional analysis methods.