Study On Cross Directional Basis Weight Control Of Paper Machine Based On PCA

Cross-directional basis weight control of paper machine is a key to the quality control of the machine, also along with the increasing of productivity and consumption, the speed of paper machine has been increasing and the web has been widening, which brings new challenges to Cross-directional basis weight control. Currently, due to the modern paper machine is equipped with a new gantry system, basis weight control system and dilution headbox, so that the measurement data of cross-directional basis weight has a higher dimension and the coupling of dilution water actuator is stronger. For cross-directional basis weight control system, the conventional data processing method and the design of controller do not meet the needs of production. In the paper, related to the problems of cross-directional basis weight control, three aspects is forwarded up to solve it.Firstly, as for the measurement data of cross-directional basis weight control system, it is analyzed that the principle and the process of transfer between the cross-directional basis weight control measurement and the control section, also the creating technology of a dynamic OPC client is presented. This OPC client can bind shared variables and deploy variables adaptively, and by through the subscription from gantry OPC server, the measurement of cross-directional basis weight control is acquired successfully.Then based on the PCA data processing principle, the most representative eigenvalues and eigenvectors of measurement data is extracted by obtaining the eigenvalues and eigenvectors of cross-directional basis weight control measurement data, also the transformation matrix is built. Through the establishment of the cross-directional basis weight models, from which the main interference is found using the PCA and autoregressive moving average of the transverse process, as a result those secondary data is discarded and the measurement data dimension is reduced effectively.Finally a predictive control algorithm named dynamic matrix control is designed, which is based on the step response to predict the model and which has the advantage of smaller amount of calculation and the capability of implicit decoupling. Then, combined with the reduced-order model by PCA, the form of objective function is achieved and also the control parameters of dynamic matrix control is selected. It is proved that the controller can track disturbance well and has good dynamic and static performance by simulation in MATLAB.