Research On Multi Fault Diagnosis Method With Coupling Characteristics

In the modern industrial process,with the development of automation and computer technology,the scale and complexity of industrial systems are gradually increasing,so that it is also more difficult to establish an accurate model.Complex working conditions make each fault occur in multiple causes,once a fault occurs,it can easily lead to a chain reaction,and then there is a multi fault concurrent situation.Due to the coupling characteristics between faults,multi fault diagnosis is more difficult than single fault diagnosis.Therefore,how to diagnose multiple faults with coupling characteristics,enhance the reliability of the system,ensure the reliable operation of complex systems is really important.In this thesis,a multi fault diagnosis method with coupling characteristics is mainly studied.The main research work is summarized as follows:Firstly,Industrial data is detected by kernel principal component analysis,we used industrial data to extract principal component,and then fault diagnosis model was built by using the principal component.T2 statistics and Q statistics was used to set the control limits.The fault detection of air compressor is used as an example to comparison of kernel principal component analysis and principal component analysis result.The simulation showed that KPCA method could have a better detection of the air compressor fault than PCA method.In order to overcome the complex effect of KPCA method during fault diagnosis,Designated Component Analysis was used.In this part the multi fault diagnosis method based on the designated component is studied.Through the analysis,it is found that when the fault exists coupling characteristic,the fault diagnosis based on step DCA has not only one diagnosis result.This paper used DSmT to consummate fault diagnosis theory,this method can show the relationship between multi faults,and the general framework of identification and ultrapower set concept was built to represent multi faults,multi fault diagnosis model was constructed based on free DSmT and mixed DSmT.Set the mix rule based on the coupled degree of fault symptom.And multiple diagnostic results was fused by using DSmT model.The result shows that DSmT model can effectively fuse the diagnosis results of step DCA methods,and get more accurate fault diagnosis results.Finally through the research on DC A residual subspace,DC A space projection framework is used.The unknown fault diagnosis algorithm is build based on residual space of DCA,statistical methods was used to estimate unknown fault.The validity of this method has been proved by the example.