Doppler distortion elimination using short-time sparse singular value decomposition strategy for wayside acoustic source fault diagnosis

Abstract

With the rapid development of transportation, wayside condition monitoring and fault diagnosis of key acoustic sources has attracted considerable attentions because of low cost and high efficiency. However, serious Doppler distortion exists in the wayside acquired signals when the monitored moving source passes by the system at high velocity, which makes it difficult for condition monitoring and fault diagnosis. This paper presents a novel method involving short-time sparse singular value decomposition to eliminate Doppler distortion in the wayside acquired signals based on a microphone array. The procedure of the proposed short-time sparse singular value decomposition is performed as follows. First, the Doppler distorted array signals are decomposed into a series of array segments by a sliding window with a proper window length. Afterwards, the time-varying direction of arrival of the corresponding array segments is acquired by individual sparse singular value decomposition. Then the fitting time-varying directions of arrival and time-domain interpolation resampling are employed to correct the Doppler distorted signals for recovering the objective characteristic frequency. Simulation analysis has validated that under heavy background noise situation, better localization accuracy and effectiveness of the proposed short-time sparse singular value decomposition could be achieved in comparison with other strategies like short-time multiple signal classification. Besides, the real data cases have verified the effectiveness of the proposed method, which shows great potential applications in wayside condition monitoring and fault diagnosis system for moving vehicles, trains, planes, etc.