Font Size: a A A

Study On Support Vector Machine And Its Application In Control

Posted on:2004-02-04Degree:DoctorType:Dissertation
Country:ChinaCandidate:Z H SunFull Text:PDF
GTID:1118360122471281Subject:Control Science and Engineering
Abstract/Summary:PDF Full Text Request
Statistics is to inference the law of nature according to observation data. Statistical learning theory is a newly developed theory for studying the statistical estimation and prediction problem based on small number of samples. It studies the nature of machine learning, so more and more people are interested in it.Support vector machine (SVM) is a new method for pattern recognition based on the statistical learning theory. It is an implementation of structure risk minimization principle in the statistical learning theory. By mapping input data into a high dimensional characteristic space in which an optimal separating hyperplane is built, SVM presents a lot of advantages for resolving the small samples, nonlinear and high dimensional pattern recognition, as well as other machine-learning problems such as function fitting. This thesis studies SVM and its applications in control. It consists of two parts, one is the study on SVM in which new SVMs are proposed, and the other is the application in optimal control, soft sensor construction and fault diagnosis.In detail, the major contributions of this thesis are as following:1. Research contributions and major problems in statistical learning theory study are reviewed. In order to explain the implementation problem of statistical learning theory, basic concepts and theory of the SVM ?including the developments and research situation of SVM itself, SVM algorithms and applications ~ are summarized, and problems in the research for each aspect are put forward.2. Fuzzy SVMs for regression estimation are developed in Chapter 2. They are used to resolve the problem that ordinary SVM could not deal with contaminated samples well. By applying generalized SVM and least square SVM of classification to regression estimation problem, fuzzy generalized weighted SVM and fuzzy multiplayer least square generalized SVM are proposed.3. Fuzzy C-clustering SVM and least square SVM for multiclass regressionestimations is presented in Chapter 3, This method can estimate multiple regressions while clustering the samples. Multi-output SVM and multi-output least square SVM are discussed for multiclass regression estimations.4. Chapter 4 introduces the use of weighted least square generalized SVM for optimal control systems. Weighted least square SVM for robust regression estimation is generalized and then used to solve N-step optimal control problem based on the SVM. The weights are determined by fuzzy method.5. To meet the needs of various soft sensors in microbiological fermentation process, new kind of soft sensors based on SVM are proposed in Chapter 5. SVM based soft sensor and least square SVM based soft sensor for off-line and on-line estimation are discussed respectively.6. Chapter 6 studies how SVMs are applied to fault diagnosis. A new method of fault diagnosis for microbiological fermentation process is presented, two relevant vector machines act as observer and classifier respectively. The observer is applied to estimate release rate of carbon dioxide to get residual sequence. The classifier is applied to classify the residual sequence. In order to reduce the loss arisen by polluted mycelia, it is very important to diagnose abnormal states in time. We adopt another fault diagnosis method for the checking of polluted mycelia in the microbiological fermentation process. The method combines nonlinear principal components analysis technology with support vector machines to construct multi-layer support vector machines. The multi-layer support vector machines are able to extract main monitoring variables from many process variables, also obtain decision function with excellent generalization performance from limited samples of fault.Finally, a brief review of this thesis is given. Some future research directions are highlighted.
Keywords/Search Tags:Statistical learning theory, support vector machine (SVM), least square SVM, Fuzzy C clustering, Fuzzy, Microbiological fermentation, Soft sensor, Fault diagnosis
PDF Full Text Request
Related items