Research On WaysideAcoustic Monitoring And Diagnosis Of Train Bearing

Wayside train bearingacousticcondition monitoring and diagnosis system maintains many advantages that the on-board system could not compare. But the practical application of wayside system is restricted because of some characteristics of the acquired acoustic signal, such as strong noise, steep distortion and multi-sources. We select NJ(P)3226X1 bearing, an active type of domestictrain bearings, as the research object and mainly focus on solving the first two problems based on time resampling doppler correction method and morphological filter algorithm.Both the simulation and experimental signals have verified the effectiveness of the proposed algorithms, providing some ideas and methods to realize the wayside acousticcondition monitoring and diagnosis.First, weformulate the main structure of NJ(P)3226X1 bearing and analyze the vibration mechanism and main fault types of rolling bearing. The calculation equations of fault characteristic frequency are also provided. We employed a static collection and dynamic transmission strategy to approximate the real acquisition environment of wayside train bearingacousticcondition monitoring and diagnosis system, where the acoustic bearing fault signals were firstly collected at the static experiment platform and then transmitted by the microphone on a moving vehicle which travelswith a constant speed.With the implementation of single source single-microphone and single source multi-microphone dynamic experiments, the wayside acoustic signal corrupted by the Doppler Effect are acquired, which provides experimental data for the following research.Afterwards, the Doppler Effect of the acquired wayside acoustic train bearing signal is explored, and the time domain resampling technique for doppler correctionis elaborated, whose key procedure is the acquisition of time resampling sequence. Based on the feature that the instantaneous frequency of doppler signal is symmetric to the time center, an acquisition method is proposed by rotating and matching the time-frequency ridge. In addition, by taking advantage of the far field uniform linear array and the space sparsity of sound source, the space information of the sound source is obtained by matching pursuit and thus the resampling sequence can be calculated according to the model of wayside acoustic diagnosis. Both the simulation and experimental analysis has indicated that the two methods can effectively correct dopplerdistortion.Finally, based on the research of the vibration mode of bearing fault signal and morphological filter, an improved morphological filtering algorithm is proposed to extract the fault feature from acoustic signal with strong noise. On the basis of the similarity between sinusoidal signal and fault transients, the structure element is constructed based on their correlation coefficients, which successfully distinguishes the noise and fault information. What’s more, the improved method can extract the fault feature from both simulation and experimental signals.