Study On Compound Fault Diagnosis Method And Application Based On Multi-symptom Characteristic Analysis

With the increasing complexity of industrial systems and the diversification of system components,the probability of compound faults increases greatly.However,the diversity of fault modes and the complexity of fault forms has brought difficulties and challenges to compound fault diagnosis.Furthermore,the fault diagnosis method for a single fault is difficult to accurately identify all fault modes of the compound faults.Therefore,this study aims at the problem of compound fault diagnosis,starting from the system symptom characterization,taking the rolling bearing as the research object,adapting the symptom analysis as the research method,and the compound fault diagnosis is studied accordingly.The main work of the thesis is as follows:(1)This thesis constructed adaptive noise reduction methods for vibration signal under the condition of unknown noise signal distribution.Firstly,the improved genetic algorithm is used to optimize the parameters of the pulse-coupled neural network with dynamic adaptive characteristics,and then the noise-reducing is processed for signal with noise.On this basis,considering the band aliasing of noise and vibration signals,the vibration signal is first processed by wavelet transform to obtain the estimated noise signal.A signal mixing matrix is formed from the initial estimated noise signal and the original vibration signal.Then the mixing matrix is separated by blind separation algorithm to obtain a signal after noise reduction.Experimental simulation results prove that the adaptive noise reduction method proposed in this research can effectively suppress noise points and improve the quality of vibration signals when the noise characteristics are unknown.(2)This thesis starts from the fault symptoms and constructs the fault symptom set from different angles,so as to characterize the compound fault.Furthermore,considering the redundant phenomenon between the features of the symptom set,the symptom sets is optimized by using the improved Relief F algorithm,so as to obtain the optimal symptom set to characterize the compound fault,and improve the efficiency and accuracy of fault identification.The simulation results prove that in the case of extracting more comprehensive fault symptom sets through multiple domains,the original symptom set optimized through the improved Relief F algorithm,which can eliminate redundancy while retaining the features that are strongly related to the fault category,can eliminate he influence of redundant features on the classifier,thereby improving the efficiency of diagnosis.(3)On the basis of the above research,this thesis uses RBF neural network to establish the mapping relationship between fault category and fault feature,of which the similar structure is similar with the mapping relationship between fault category and fault feature,so as to complete the identification of compound fault.The simulation results show that the fault identification model constructed by RBF can accurately identify the compound fault types.In this thesis,the rolling bearing is taken as the research object,and the vibration signals of the rolling bearing are analyzed according to the steps of signal noise reduction,symptom set extraction and optimization,and fault mode recognition.The effectiveness of the compound fault diagnosis method proposed in this thesis is verified by the public data set and the measured data set respectively.Combining the neural network method with the traditional signal processing method,the compound fault is diagnosed by the method of symptom analysis,and the recognition accuracy of the compound fault is improved.