Fault Diagnosis Approach Research Of The Three-Tank Process System

Three-tank process system with nonlinear, large delay and coupling is typicalabstract model of the controlled objects in industrial process, and then the faultdiagnosis approach on three-tank process system has been widely concerned about.Firstly, in this thesis the background and significance on fault diagnosis researchis introduced, the current research status of fault diagnosis on the process controlsystem is stated and the fault diagnosis approaches domestic and overseas aresummarized on three-tank process system.Secondly, the main structure and the working principle of three-tank system arestudied, and the mechanism model is structured, according to its mathematical model,Simulink model of system is built for fault diagnosis research.Thirdly, the signal de-noising pre-processing technique is studied on three-tanksystem, the principle of wavelet transform, wavelet packet and1ifting wavelet is statedin signal de-noising, summarizing the advantage of1ifting wavelet. Combined withprobabilistic neural network (PNN), the fault diagnosis approach based on1iftingwavelet and PNN for isolating the faults of the three-tank system is presented. The datathrough1ifting wavelet pre-processing is as the input characteristic vector of PNN, andthen the diagnosis result is obtained through PNN diagnosis analysis. The advantage ofthe approach presented is making pattern recognition of PNN more reliable through theanalysis and reconstruction of the signal to eliminate noise interference. The MATLABexperiment results show that the fault diagnosis approach based on1ifting wavelet andPNN is correct and effective.Finally, the fault diagnosis research platform is designed through forcecontrol andMATLAB, including the system interface, database, C/S and OPC technology.