Compare Of Nonlinear Process Fault Detection And Diagnosis

Industrial processes are complex nonlinear processes. The stability and security ofindustrial processes play very important role. In order to ensure the industrial process running,monitoring the industrial production process automation is requested. Fault detection anddiagnosis techniques (Fault Detection and Diagnosis, FDD) widely applied to industrialautomated production systems.In this thesis, wastewater treatment process which is more complex, non-linear,non-Gaussian among industrial process and penicillin production process which is batchprocess are treated as two research objects.This thesis uses multi-model fault detection methods to the wastewater treatmentprocess. Multi-model fault detection uses a combination of several different methods. In thisthesis, the wastewater treatment process uses multi-model fault detection methods whichcombine a variety of clustering methods with principal component analysis to compare. Atlast an improved multi-model method which combines subtract fuzzy C-means clusteringsubtraction and principal component analysis with lifting wavelet transform denoising obtainbetter results in wastewater treatment process fault detection.In the course of penicillin process fault diagnosis use a batch fault diagnosis method. Inthis paper, SOM method is used to divide penicillin process phases. BP neural network israrely used in penicillin process fault diagnosis. It also uses a genetic algorithm to optimizeBP neural network. Multi batch SOM-GABP method is put out as a improved penicillin batchprocess fault diagnosis method.Finally, this thesis will propose a method using Labview platform used in industry.