| With the introduction of computer data acquisition and control to the process industries, it became possible to obtain huge volumes of data. Such tools as multivariate statistical projection has been shown to be more effective than normal algorithms, which often build ill models, in compressing this large volume of noisy correlated data into a subspace of much lower dimention than the original data set. Because most of what is eliminated is the collinearity of the original variables and the noise, the bulk of the information contained in the original data set is obtained. The resulting low dimentional representation of the data set has been shown to be of great utility for process analysis and monitoring, as well as fault diagnosis. These types of models can also be used directly in contol system design, which seldom resumed in literatures. Most control systems are based on model. With the development and innovation and inosculation of control schemes, solutions to multiple-input multiple-output(MIMO) system have captured eyeballs of researchers. At the same time, how to make advantage of the characters of multivariate statistical projection technique to simplify the structure of control system and cotroller design, as well as optimize the control performance, is becoming a new point with much attention.Among the regression algorithms in multivariate statistical projection which model the relationship between two blocks of data, partial least saquares(PLS) model two blocks while simultaneously compressing them, and outer relationship is projected to inner latent variables spabe with orthogonal component model. Some advantages of using this approach as part of control system design include automatic decoupling and efficient loop paring, as well as natural handling of nonsquare systems and poorly conditioned systems. In this paper, a methodology is proposed for process development and control based on optimization in the subspace defined by the latent variable models built from data blocks. The methodology is applied to several industrial process and some simulation are made to shown its efficiency in systemcontrolling. Contents in this paper are as following:1. The paper summarizes the definition, developing tendency and problems existing in Control Theories and design practices, as well as provided a comprehensive review of development of Multivariate Statistical Projection (MSP) method, especially Partial Least Squares (PLS)technique, and its application in process control field.2. The paper briefly introduced the theoretical foundation of MSP method, which include Principle Component Analysis(PCA), Principle Component Regression(PCR), and Partial Least Squares(PLS). In addition, the paper provided a brief introduction of PID controller.3. Methodological principles for control based on optimization in the subspaces defined by the latent variable models built from PLS are expatiated. Control structure is designed based on the methodology and compared with ordinary control structure. Merits as well as deficiencies of the method iss analyzed. Some simulations are applied to several processes to testify the performance of the method and the results turned out to be affirmative.4. Dynamic PLS modeling for process control is developed to adapt to the dynamic characters of control systems and improve the utility of PLS which usually arrived from steady datas. Dynamic PLS is applied to two systems, known as Shell Standard Control Problem and some kind of distillation.5. A method is proposed to monitor the controlled system using PC A model and multivariate statistical control charts. Points out of control and disturbances then can be detected with a direct and clear way. The method is applied to two controlled system, and some analysis is made according to the control charts. Finally, the paper concluded the research findings.and pointed out some futureresearch areas... |
PDF Full Text Request