Intelligence Diagnosis System Research Of Rolling Bearing Composite Faults

Rolling bearing is one of the most widely used and easily damaged mechanical parts.Its operating state will directly affect the whole system, and may lead to equipmentdamage, even cause catastrophic accidents. Therefore, developing rolling bearingsmonitoring and fault diagnosis has great practical significance. This paper choose rollingbearing fault diagnosis model based on support vector machine as the research object, themain work is as follows:Firstly, it makes detailed analysis of rolling bearing mechanical structure, the faultmechanism and the fault forms. Using the vibration signal analysis method, it makesconcrete analysis of several typical faults time and frequency domain characteristics.According to the insufficiency of fault data samples in engineering practice, it putsforward using the support vector machine classifier to solving the rolling bearing faultdiagnosis problem.Secondly, it proposes a multiple classifiers model based on features extraction,according to the single method being not able to extract the complete features, thencausing the incorrect diagnosis. This model combines with three methods of waveletpacket transform, the ensemble empirical mode decomposition and the time-frequencyinformation entropy analysis. It can obtain complete information from different angle, andthen it is input into the multiple classifiers group, which is composed of three supportvector machine classifiers. And then it use information fusion method of self-adjustingweighted decision templates to make the final decision. Through the simulation results itshows that the model can be applied to the bearing composite faults diagnosis.Finally, in order to analysis the effects of the parameters optimization on theclassification results and further simplify the multiple classifiers group’s structure, itcombines the time and frequency domain characteristics to extract rolling bearing faultinformation, and it also uses multiple classifiers group based on support vector machine tomake the final diagnosis. Meanwhile it uses an improved artificial bee colony algorithm,whose searching ability is stronger, to search for the support vector machine bestclassification parameters, in order to improve the classifier’s performance. Through the simulation analysis of the different bearing signals it proves the superiority of thediagnosis model.