Acoustic Detection Algorithm of Train Bearings Based on Phased Derivative of CFD and the Fallow-GSC

Trackside acoustic detection system (TADS) is one of the high-speed train online detection technologies, which has the advantages of early fault detection, low cost and non-contact measurement. In order to extract bearing fault characteristics from a strong noise environment effectively, an acoustic fault detection algorithm for train bearings is proposed based on the motion parameter estimation and noise cancellation with microphone array in this paper. This method uses the phase center finite difference (CFD) based on phase derivative to estimate to motion parameters of the sound source, and constructs the Fallow-generalized sidelobe canceller (GSC) through the motion parameters to perform beamforming (BF) to achieve noise elimination. Finally, a resampled times series is constructed to perform Doppler correction on the output signal after BF. The Fallow-GSC consists of variable fractional delay (VDF) digital filters, a fixed beamformer (FBF), a blocking matrix and an adaptive noise canceller (ANC) based on the recursive least squares (RLS) method. This method solves the problem of mutual cancellation between noise signal and the upper branch expected speech in the process of adaptive filtering of moving sound source faced by the traditional GSC algorithm, avoiding the distortion of expected speech. The effectiveness of the method is verified by practical experiments. Compared to the traditional GSC algorithm, it has a higher diagnostic accuracy and improves the signal-to-noise ratio.