Research On Bearing Fault Identification Based On Multi-sensor Information Fusion

Rolling bearing is one of the important components in the structure of mechanical equipment parts.The state of rolling bearing itself will directly affect the whole mechanical system,so people focus on the condition monitoring of rolling bearings.Early condition monitoring of rolling bearings,timely detection of faults and diagnosis can fundamentally change the post-event and regular rolling bearing maintenance to rolling bearing maintenance as appropriate,so as to reduce maintenance costs and avoid accidents.In the traditional rolling bearing fault diagnosis technology,the most widely used method is the vibration analysis method,but the signal source of this method is single,and the characteristic reflection of the tested object can only rely on a certain side.Moreover,a single detection sensor is easy to cause error in fault diagnosis results due to external information interference or its own fault.Therefore,a method of fault diagnosis based on the fusion of rolling bearing vibration signal and sound signal is proposed.In this paper,the multi-sensor information fusion technology is applied to calculate the vibration and sound signals of rolling bearing through RBF neural network and D-S evidence theory.On this basis,considering the shortcomings of multi-sensor information fusion method,such as complex eigenvalues and long calculation time,a RBF neural network based on Fisher score is proposed to screen the eigenvalues,which can not only meet the high recognition rate,but also reduce the recognition time.In this paper,an acquisition program is designed based on LabVIEW.The QPZZ-Ⅱrotating machinery vibration analysis and fault diagnosis test platform is selected as the experimental platform.The vibration and sound data of N205 rolling bearing in inner ring fault,outer ring fault,rolling element fault and normal conditions are collected through the test-bed.After verifying the reliability of the data and preprocessing,the single vibration signal,single sound signal,vibration and sound signal are input to Fisher＿RBF network for calculation.The comparison results of the three fault recognition rates fully verify the reliability and necessity of multi-sensor information fusion technology in the application of rolling bearing.Then the vibration and sound signals are input into Fisher＿RBF＿D-S model for calculation.The comparison results prove the necessity of the fusion of feature level and decision level in multisensor information fusion technology,and provide a new method for rolling bearing fault diagnosis.