Method of Splicing Constant Speed Signals with Short Signals During Wheelset Acceleration and Deceleration
The characteristic value of train wheelsets is closely related to the speed,and the quantitative analysis of wheelset faults requires signals with a certain length collected at a certain speed.However,such a signal is difficult to be obtained by the system working in the variable speed condition.Therefore,a signal reconstruction method that did not require any velocimeters was proposed to realize the above quantitative analysis.This method used short signal segments to splice the constant speed signals under the acceleration and deceleration conditions.Firstly,a short-time Fourier transform(STFT)peak search method was adopted to obtain the speed curve and extract multiple signal segments passing through a specified speed.Secondly,the signal segments were converted into an angular domain signal,and the phase difference between signals was determined by cross-power spectra density(CPSD).Then,according to the phase difference and the length of the front signal,the splicing position of signals was obtained,and interpolation and fusion techniques were used to ensure the continuity of spliced signals.Finally,the spliced angular domain signal was resampled into the time domain signal.A rotor test bench with a similar wheelset structure has verified that during acceleration and deceleration of a cracked shaft,the frequency spectrums of the spliced signals of each speed obtained by this method are highly close to the frequency spectrums of the constant speed signals obtained by the experiment.The relative errors of the 1x and 3x values between the spliced signal of 15.0 Hz and the constant speed signal are only 0.9%and 1.0%,respectively.After adding noise to the signals to be spliced and shortening them to 1.5 cycles,the splicing phase error does not exceed 10°.
国家科技期刊平台
NSTL
万方数据
