Font Size: a A A

Study On Support Vector Machine Inverse System Method And Its Application

Posted on:2011-10-06Degree:MasterType:Thesis
Country:ChinaCandidate:D Y HeFull Text:PDF
GTID:2178360308955344Subject:Control theory and control engineering
Abstract/Summary:PDF Full Text Request
Support Vector Machine (SVM) has lots of advantages in dealing with machines learning ,including good generalization ability, simple topological structure, et al, which is based on the statistical learning theory. SVM can map linear inseparable input data into a high dimensional linear separable feature space in which regression is done via a nonlinear mapping technique and the SVM's solution just has only one. This dissertation uses SVM to solve the problem of identification and control in nonlinear systems, and studies the disadvantage of SVM inverse method ,and the proposed methods are applied to nonlinear system control.This article firstly introduces SVM and SVM inverse method application in system identification including the development and the research state. On the theory of inverse method, the reversibility and relative vector order of system has been described. Then, the method of choosing the parameters of SVM is analyzed. Simulation of nonlinear SISO system is studied by using SVM which can get good performance, and the method would be extended to MIMO nonlinear system control. Ball mill coal pulverizing process is a nonlinear and multivariable coupling system. Firstly, based on the characteristics and the dynamic mathematical model of ball mill coal pulverizing system, the reversibility of the system is testified. Then, this paper applies SVM to obtain the inverse system of ball mill coal pulverizing process which is combined with the original system. Finally, in order to overcome the modeling errors of SVM inverse system method, a predictive controller is designed to obtain the closed optimized control of the compound system which has been realized approximate decoupling and linearization. Simulation results show that good control performance can be obtained in ball mill coal pulverizing process based on the presented control approach which doesn't rely on the exact process model and parameters, and can adapt the uncertain model.The adaptive inverse control is introduced to expand applications of SVM inverse method which does no resistance to disturbance and only for minimum phase systems. In simulation, the model reference adaptive control strategy based on SVM method gets a good performance which can not only have a favorable dynamic tracking ability to nonlinear non-minimum phase system, but also has resistance to disturbance and model change in the system.
Keywords/Search Tags:support vector machine, reversibility, ball mill coal pulverizing process, predictive control, adaptive inverse, non-minimum phase system
PDF Full Text Request
Related items