Fast Algorithm of Flat Sliding Detection in Flat Wheel Detecting System

In this paper, a fast algorithm for flat sliding detecting on wheels of passenger train is developed. Firstly, the input signal is denoised by using filter bank to enhance its SNR, and the first and second derivative of denoised signal is calculated. Secondly, when the local maxima and minima are sought based on the two derivatives, the singularity and regularity of the signal is revealed and the sub-signals including only one maximum are separated from the denoised signal. Thirdly, the sub-signals of singularities are checked by the decision rule, thereby the flat sliding detection (FSD) of train wheels is implemented. The algorithm is also effective for different sampling-rate input signals. As an application, we have emulated the algorithm with the measured data from the flat wheel detecting system (FWDS) and done experiments on railway. The results show that: the correct rate of FSD exceeds 99% and the measurement precision of the flat sliding depth can reach 0.2mm