Researches On Performance Monitoring And Fault Diagnosis For Process Industry Based On ICA-PCA Technique

The process monitoring and fault diagnosis is one of the most important problems in the process industry. Through monitoring the state of the production process, detecting the fault, process upsets and other abnormal events promptly, locating and removing the factors causing such event, the safety of production process will be assured and the quality of the product will be improved. With the development of the CIPS (Computer Integrated Process system), a great amount of process data can be sampled and collected. How to fully utilize this deep-level information to improve the performance of the process monitoring has been gradually becoming one of the focuses in the field of process control. Traditional multi-variable statistical monitoring methods, such as PCA (Principal Component Analysis), PLS (Partial Least Square analysis), assume that the variables must subject to the normal-distribution condition in addition to the assumption that the variables are independent and identical distribution, and only the second-order statistical information is imposed. ICA (Independent Component Analysis), which is introduced into the field of process industry as a data analysis method, is a signal decomposing technique based on the higher-order statistical information. This method can utilize the statistical characteristics of the variables more efficiently. The intrinsic characteristics of the process can be described through the decomposing of the monitoring variables under the meanings of the statistical independence.In view of characteristics of the continuous and batch industry process systems, some improvements of tradition PCA have been made at different degree, and some new monitoring algorithms based on ICA are also proposed in this thesis. The works are divided into two parts, the continuous process monitoring and the batch process monitoring. The main contributions of this thesis are as follows:1, A new continuous industry process system monitoring method based on independent component analysis (ICA) is proposed. The maximization of non-Gaussian criterion is adopted to decompose the independent non-Gaussian components from the monitoring variables. It can satisfy the independence with the statistical meaning, not just the decorrelation in PCA. Furthermore, the non-Gaussian sequence technique is used to choose the number of the independent variables and set the relevant confidence limits for thecontinuous monitoring. The monitoring performance of the proposed method and the PCA-based method are compared with application to the Tenessee Eastman process (TE process). The results show that the proposed ICA-based method has advantages over PCA-based method, and less detection delay can be obtained. At the same time the ICA-based method can monitor the faults which the PCA can not.2, A string matching method is proposed for the different fault identification. The latent variables are extracted from the independent component analysis (ICA) in different state of production can be transformed to the character strings. According to the situation of alarm limit violations of independent components, the diagnosis of faults is reduced to a string matching problem. The proposed method is data driven and unsupervised, does not need plentiful of the training data and a process model. The simulation results from the application to the Tennessee Eastman challenge process demonstrate its feasibility and effectiveness.3, An improved step-by-step MPCA, using the process variable trajectories to monitor batch industry process system is presented. A series of MPCA is set up to avoid pre-estimating the unknown part of the process variable trajectory deviation from the current. time to the end. Meanwhile, a forgetting factor makes the control much easier and can be extended to the multi-phase batch process. This algorithm is evaluated on industrial streptomycin fermentation process data and is compared with the traditional MPCA. The step-by step MPCA can detect the abnormal quicker than the traditional MPCA, can also reflect the modification o...