Researches On Continuous Process Performance Monitoring And Fault Diagnosis Based On PCA

Principal component analysis (PCA) has been considered as an important method of fault detection and diagnosis due to its independence of process model and combinition with computer technology. However, the theoretical method and actual application of fault monitoring based on PCA is not perfect and has a lot of problems. The main purpose of this thesis is to make further study on fault detection and diagnosis method based on PCA. The main context consists of the following parts.(1)The outlier in historical data acquired from industry process can decrease ability of process performance monitoring. A new outlier detection combined method based on robust scaling closest distance to center (CDC) and ellipsoidal multivariate trimming (MVT) is proposed after a summary on several robust outlier detection method principle and limitation. The results of an example show it can detect outlier effectively and accurately.(2)Because conclusions are indefinite in the performance monitoring of industry process, a faulty detection approach based on Q statistic separation is proposed. Q statistic is separated into principal component related variable (PVR) statistic and common variable (CVR) statistic, and it can detect adequately the change of process using with T2 statistic. The simulation results show that it can enhance the veracity of process monitor comparing with traditional detection approach.(3)This thesis presents a new determination of principal component in PCA model by using cumulative percent variance (CPV) and multi-correlation coefficients (MCC) because of subjectivity and neglect of fault information using only cumulative percent variance. The application results show that the approach can ensure enough information in principal component subspace.(4)The T2 statistic based on fault reconstruction technology is used to improve the ability of fault diagnosis by defining principal component subspace (PCS) and residual subspace (RS), because of the limitations using variable contributing rate and SPE statistic fault reconstruction technology. This approach is applied to double-effect evaporator process. The simulation results show that it can identify fault effectively.Finally the dissertation is concluded with a summary and discussions of the prospective research on open problems.