Application Research Of LS-SVM Inverse System Method In Headbox Decoupling Control

Headbox is one of the essential parts of the high-speed paper machine, and plays a keyhub role in connecting "for slurry flow" and "sheet forming". With the constant improvementof the paper machine speed and market demand for paper quality, it is necessary to researchand invent the air-cushion headbox and control strategy which adapt to the situation of ourcountry, high performance and low cost. This thesis analyzes the internal mechanism of theair-cushion headbox and the control systems which have been existed on the basis ofunderstanding domestic and foreign research contemporary situation of headbox. So LS-SVM(Least Squares Support Vector Machine) inverse system method of headbox decouplingcontrol method is proposed. The main contents are as follows:(1) Build inverse model of multivariable system. Mathematical model must be knownprecisely in the Traditional inverse system method, and it is difficult to calculate analyticalsolutions of the inverse model, thus to identify inverse model of the original by combiningnonlinear modeling features of LS-SVM, and put it with the original system in series formpseudo linear system. The effectiveness and feasibility of the method were verifiedrespectively by the simulation of single-variable and multivariate instance. On the one hand,multivariable inverse method can achieve pseudo-linear system. On the other hand, a complexsystem can be decoupled into single independent pseudo linear subsystems. Therefore a newdecoupling method is supplied for a class of uncertain nonlinear strong coupling system.(2) Design controller for multivariable system. Decoupling ability of multivariableinverse system method is limited. So in this thesis pseudo linear composite system will be ascontrolled object. PID and DMC (Dynamic Matrix Control) two control methods werediscussed for the closed-loop control. The simulation results show that although PIDcontroller can achieve good control effect, and have certain anti-interference ability androbustness, but it is difficult to adjust parameters, and overshoot amount is relatively large.For better dynamic and static performance and robustness, further analyzed and designeddynamic matrix controllers.(3) Applied research of LS-SVM inverse system decoupling control method. Inversemodel of headbox is obtained using LS-SVM identification by analysis internal mechanism ofthe headbox. The original headbox system is decoupled into two independent single-inputsingle-output pseudo linear subsystems by the principle of inverse system method. In order toovercome the influence of modeling error and external disturbance, the headbox pseudo linearsystem was controlled by using PID and DMC method as closed-loop controller. Thesimulation results show that the control method has a good robustness, simple structure andeasy for engineering implementation. From the research of theory analysis and simulation, headbox system is effectivelydecoupled using LS-SVM inverse system method. Thus the pump flow rate is uniform, thephenomenon of cross flow and concentration changes is reduced. At the same time, itprovides a new way for a class of uncertain nonlinear strong coupling system decouplingcontrol.