A gear fault diagnosis method based on auditory cortical frequency-domain response in real workshop noise environment

Acoustic-based diagnosis (ABD) has become a research hotspot in gear fault diagnosis due to its noncontact advantage. Interference from ambient noise is one of the critical problems that urgently need to be solved in the practical application of ABD methods. The human auditory system has a strong ability to resist noise, with the auditory cortex playing a core role. Therefore, a gear fault diagnosis method based on auditory cortex frequency-domain response under real workshop noise conditions is proposed in this paper. This method simulates the auditory cortex processing mechanism, calculates the auditory cortical frequency-domain responses of gear sound signals containing workshop noise, and thereby obtains the noise-resistant processed data. Subsequently, the auditory cortical frequency-domain responses are clustered and subjected to dimensionality reduction through cross-correlation analysis, and appropriate features are extracted from the dimensionally reduced data to form a feature vector set. Finally, the classification part is implemented by the support vector machine. The classification results indicate that an average accuracy of 99.26% under different signal-to-noise ratio conditions can be achieved by this method, which shows better diagnostic outcomes compared with existing methods. This method can provide a new idea for gear fault diagnosis in noisy environments.