Research on On-Board Acoustic Fault Diagnosis Method of Train Bearing Based on Near-Field Frequency-Domain Smoothing Minimum Variance Distortionless Response

Aiming at the problem of enhancing the multi-target bearing signals of the train in the complex acoustic environment during on-board fault diagnosis, this paper proposes a near-field frequency-domain smoothing MVDR (Minimum Variance Distortionless Response) algorithm based on a circular microphone array. First, the near-field MUSIC (Multiple Signal Classification) sound source localization algorithm is used to estimate the position of the sound source. Then, the sound source localization result and the data are imported into the near-field frequency-domain smoothing MVDR algorithm. The frequency-domain smoothing is utilized to assist in constructing the covariance matrix of the array, improving the estimation accuracy of the covariance matrix and realizing the enhancement of each sound source signal. Finally, the fault information of the train bearing is determined by analyzing the envelope spectrum characteristics. Through experiments, it is verified that the proposed method has an 11.45% improvement in SNR (signal-to-noise ratio) compared with the traditional near-field MVDR algorithm.