Research And Application Of The Fault Diagnosis Of Rolling Bearing Based On The Sound Signal

Rolling bearing is widely used in lots of machinery and it is one of the most easily damaged parts in rotating machinery.Bearing faults have direct influence on the normal operation of the whole machine and even endanger the personal safety.So analysis of fault monitoring and fault diagnosis is very significant.The energy distribution of the sound signals will change if the rolling bearing emerges fault,then the fault feature information can be extracted by analyzing the sound signal.The sound fault diagnosis technology has the characteristics of non-contact measurement,high speed,simple equipment and convenient signal acquisition.However,the sound signal is very low signal to SNR,it is difficult to extract the feature information.Therefore,diagnosis of rolling bearing and the compound fault model using sound signal as the failure signal source,the main research on the sound signal of rolling bearing is as follows:(1)Aiming at the case that characteristics of the sound signal and the traditional BSS only obtain single-channel signal in actual bearing fault diagnosis,a signal-channel BSS algorithm is presented.This method is based on wavelet packet and FastICA algorithm,it can solve the main defects of BSS.(2)We estimate the source number of the virtual multichannel sound signal using the SVD method,and rebuilt the multichannel signal according to the number of source signal.Then,FastICA algorithm is applied to separate the sound signal that we get at the second step.The example analysis shows that this method used in the fault diagnosis for bearings is effective.(3)Based on the Shannon entropy module,Shannon entropy is combined with different sound signal analyzing algorithms in time domain,frequency domain and time frequency domain separately to construct the three Shannon entropy feature indexes,such as singular spectrum entropy,power spectrum entropy and wavelet packet energy spectrum entropy.Which could realize the quantitative description of the energy distribution character of the sound signal in different transform space.(4)Fault recognition based on Shannon entropy feature extractions and SVM.Making singular spectrum entropy,power spectrum entropy and wavelet packet energy spectrum entropy as the feature vector.And then put SVM to judge the fault type and fault degree.The method has high recognition rate.(5)Establish a set of fault diagnosis system based on sound signal,and analyzed the experimental data.The analysis shows that these methods used in early fault diagnosis based on sound signal are effective.