Research Of Explosion-proof Motor Fault Monitoring Based On Audio Recognition

The explosion-proof motor fault monitoring can response its running state. It’s convenient for maintenance personnel to control the motor. The voice as an important means of communication can be used to monitor the fault. As we know, the fault monitoring based on audio recognition has advantages of simplification and smartness. Even laypeople can also utilize this method to find out and maintain the fault in advance. Nowadays more and more researchers carried out the study of audio recognition technology. This paper conducted in-depth research of explosion-proof motor fault monitoring based on audio recognition.In this background, we studied the fault detection system based on audio recognition first. Then we preprocessed the signal by means of pre-emphasis, framing, add window and endpoint detection technology. In terms of characteristic parameters selection, we extracted linear prediction cepstrum coefficient(LPCC) and Mel frequency cepstrum coefficient(MFCC). After the preprocessing and feature extraction of audio signal, we found out the characteristic parameters.They can show the signal nature. Based on hidden markov model, we set up the acoustic model and study the main problems involved in the HMM. Finally study several noise reduction technologies.During the study of endpoint detection technology based on double threshold, we found that threshold setting only depends on the experience of researchers or repeated experiments. But in complex environment, the performance of this method is unstable. When the noise changed little, accuracy is relatively high. Otherwise the accuracy gets very low in noisy environment. Aiming at this problem, the method has been improved. We used the adaptive threshold instead of fixed threshold. And the length of window was also modified. When making characteristic parameters extraction, the commonly used parameter is MFCC. But this parameter can only describe static characteristics of audio signal. Because the auditory system is more sensitive to dynamic characteristics, we put forward an improved MFCC parameter. This improved parameter can describe dynamic characteristics of sound. We got it by secondary extraction of MFCC parameter. And extraction method is weighting and difference.Based on technical research and improvement, we developed a mature explosion-proof motor fault monitoring system. This system has good recognition performance.