Reduced Order Modeling And Control Of Cross Directional Basis Weight On Paper Machines

In the papermaking processes, basis weight is a fundamental property of paper requiring precise control for quality of production. Because of user's improvement of the quality requirement of the paper and control on cost of the manufacturer correctly, cross-directional control of paper is being paid close attention to more and more, it is also one of the most complicated control targets in the papermaking processes. Because of changes in raw materials and the complicated pulping process prior to the paper machines, variations will undoubtedly exist in the formed paper sheet. The controlling problem on paper machines has become more difficult, with the increase in sheet width and machine speed.K-L (Karhunen-Loeve) transform is widely used in system modeling and signal processing, it based on the same class characteristic choice, and reduce characteristic space dimension, the method employs a sequence of order reduction techniques based on the Karhunen-Loeve expansion of the disturbances affecting the properties of interest .Once the dominant disturbances are identified, the space of the manipulated variables is also projected down to the space of the dominant disturbances, effectively creating a reduced order input-output description of the original physical system, then ,the optimal control action in the lower order subspace is found by solving the formulation, and the corresponding solution of the optimization problem is projected back to the original space to be implemented on the real plant.The contents of thesis are outlined as follows:(1) The overview of the cross-directional control technology of the survey paper machine, the newest progress of developing and present research, and has analysed the question existing among them .(2) Application of K-L theory in system modeling and signalprocessing are introduced.(3) The method employs a sequence of order reduction techniques based on the Karhunen-Loeve transformation of the disturbances affecting the properties of interest , effectively creating a reduced order input-output description of the original physical system.(4) The model predictive control strategy is employed to control the system, the size of such optimization and control problems associated with large dimensions can be significantly reduced, such make the good result greatly.Finally , the conclusion are drawn , and the challenges and problems that are expected to meet in the future are discussed.