Research And Development Of Multi-Channel Fault Diagnosis System Based On Acoustic Signal

The degree of automation in modern society is getting higher and higher with the progress of science and technology.All walks of life have different kinds of equipments to improve production efficiency.Now the equipment is becoming more and more perfect,but the structure is becoming more and more complex.According to the basic theory of reliability:the more complex products,the greater the probability of failure.Once a failure comes to some pivotal equipment,it will not only cause huge economic losses,but also may threaten the safety of the staff and even cause catastrophic accidents.Therefore,it has urgent market demand to monitor and diagnosis the running state of the equipment.When the device fails,its acoustic signal will change in both frequency domain and time domain.The acoustic signal can be measured without affecting the running state of the equipments.So takeing the acoustic signal as the research object to monitor and diagnosis the running state of the equipment has great practical signifiicance on ensureing the safe operation and avoiding major economic losses.This paper takes the research and development of the fault diagnosis system based on acoustic signal as the subject,and extracts the wavelet packet decomposition energy features,the MFCC features,and the hybrid features to train the SVM classifier by using these feature vectors.This paper uses the Case Western Reserve University's bearing dataset to test these algorithms,and the expirement results show that the diagnostic system based on these above features has a high recognition rate.At the same time,considering that there are more normal data and few failure data,the anomaly detection method of Gauss model is applied to fault diagnosis to monitor the equipment's abnormal state.Then this paper tests this method on the same dataset,the results show that it can distinguish normal and abnormal data with higher recognition rate.and the gauss fitting result also has a satisfactory recognition rate on abnormal data classification.This paper carries out a detailed analysis and design of this fault diagnosis system according to the actual demands,and uses suitable data acquisition equipment and sensor equipment to obtain multichannel acoustic signal data.Through Matlab and C/C++ hybrid programming technology,this paper achieves calling the Matlab processing algorithm functions in the MFC application.Then,this paper uses database technology to save the system's various parameters,and realizes a multi-channel fault diagnosis system,which has an adjustable sampling rate,drived by acoustic signal data,and can be retrained and improved under the operator's control.Next I use the fan datas,bearing datas,and the centrifugal pump datas to test the diagnosis system.The system can set different sampling frequencies on different datas,and can use different feature extraction method and classification method to train the classifier under the operator's control and use the classifier to classify the on-line data.Finally,this paper achieves the expected goal.