Research On Ensemble Fault Diagnosis Method For CSTH Process

With the development of modern industrial technology,the fault type of industrial process presents the characteristics of diversification and complication.Once the sources and types of equipment failures are not detected and diagnosed correctly,the whole industrial process will not be able to run normally,so as to affect the production efficiency and economic benefits of industrial processes.In view of this,a timely and effective fault monitoring and diagnosis for industrial process is particularly important.A ensemble fault diagnosis method is proposed to improve the performance of fault diagnosis,which is based on the variational mode decomposition(VMD)-improved fuzzy c-means(IFCM)-instant learning(JITL)-recursive least squares support vector machine(RLSSVM).Data collected from the industrial field often have a lot of noise due to external factors or the device itself.Therefore,it is necessary to deal with the collected data.In this thesis,the VMD method is used to deal with the data.It has the advantages of strong robustness and small sampling effect compared with the traditional algorithm.VMD method decomposes the data into independent components,and obtains the denoised data through independent components.In modal identification,in order to identify the failure mode belongs to which operating mode,the IFCM method is used to distinguish the modal.In order to set up the model timely and adapt to the changes of working conditions,this thesis use a JITL algorithm to solve the problem of updating model timely.Based on the above analysis,a fault diagnosis method based on VMD-IFCM-JITL-RLSSVM is proposed.The method was applied to the experiment of continuous stirred tank heater(CSTH),and the diagnosis results were compared with the traditional method.The theoretical and experimental results show that the fault diagnosis method based on VMD-IFCM-JITL-RLSSVM is better than traditional method in fault diagnosis,and it has a remarkable improvement in the speed and accuracy of fault diagnosis,and its performance is superior to traditional fault diagnosis algorithm.