Research On Wheel-rail Noise Separation Based On Partial Correlation Technique

In recent years,with the rapid development of the domestic high-speed railway business,the problem of high-speed railway noise has become more and more prominent,and has attracted widespread attention.Wheel-rail noise occupies a major position in high-speed railway noise.Therefore,to reduce high-speed railway noise,we must start with the suppression of high-speed railway wheel and rail noise.Due to the complexity of the wheel-rail noise generation mechanism and the coupling between the wheel and rail,it is difficult to analyze the noise generated by the wheel and rail separately,and the beamforming sound source recognition technology based on the microphone array can accurately identify the noise of the train wheel and rail area.Therefore,this paper takes a different approach.Starting from the microphone array sound source recognition technology,the 143-channel spiral array is used to collect the noise data of the passing train on the Great West Line,and then the deconvolution beamforming algorithm is used to identify the noise in the train wheel track area At the same time,by arranging vibration sensors on the rails to obtain the vibration data of the rails when the train passes by,the noise generated by the rail vibrations is obtained;finally,the identified wheel and rail noise and the collected The rail vibration data realizes the separation of wheel and rail noise,which provides a theoretical basis for the subsequent targeted reduction of wheel and rail noise.The main research contents and conclusions of this article are as follows:(1)The mathematical model of the microphone array receiving signal is established,the basic principles of the traditional delay-sum beamforming algorithm and cross-spectrum beamforming algorithm are analyzed,and the specific solution based on the two deconvolution algorithm theories of FFT-NNLS and FISTA step.(2)The simulation results of three beamforming algorithm programs to identify the narrowband sound source,broadband sound source,single sound source and dual sound source,the comparative analysis of the recognition performance of the three algorithms,and through the static sound source recognition test test Verified.Then the beamforming algorithm is extended to the direction of the recognition of the moving sound source.Because the microphone will generate the Doppler effect when receiving the sound pressure signal of the moving sound source,the interpolation method isintroduced to eliminate the Doppler effect,and verified by simulation analysis.(3)The basic principle of partial coherence analysis method and maximum signal-to-noise ratio blind separation algorithm in partial correlation technology is studied and analyzed.The separation effect of these two separation algorithms is simulated and verified.(4)Based on the noise data and rail vibration data of high-speed trains at different speed levels,the wheel-rail noise is identified,the frequency characteristics of the rail vibration are analyzed,and the noise radiated by the wheels is separated using the partial dry function and the maximum signal to noise ratio separation algorithm Provide effective guidance for the future project to reduce the wheel and rail noise to do targeted vibration and noise reduction measures for the wheels and rails.