Theoretical Control and Experimental Verification of a Calender Roller System

• Main Authors: Phung Van Hiep, PhungVanHiep
• Other Authors: Chung-Feng Jeffrey Kuo
• Format: Others
• Language: en_US
• Online Access: http://ndltd.ncl.edu.tw/handle/40101935674372543744

碩士 === 國立臺灣科技大學 === 自動化及控制研究所 === 97 === Control of amount of deflection which is produced by the pressing and pressed rollers in calender system is big challenge in industrial fiber. Normally, trial and error is used to determine each dimension of both rollers. Hence, objective of this thesis is dealing with the control of the pressed roller to get high quality of the calendering products. Both theoretical control and experimental verification are performed. Firstly, the pressed roller is assumed as an Euler-Bernoulli beam and the forces which are produced by pressing roller are considered as distributed forces. The mathematical model of the system is derived as a distributed parameter system using Hamilton’s principle, Lagrange’s equations and assumed-modes method. System model is a two-input one-output system. Since the distributed forces are constants, system can be simplified to single-input single-output system by changing variables. In order to investigate effect of non-collocation sensor and actuator, positions of sensor and actuator are also considered as arbitrarily. Then, an advanced control technique, Model Predictive Control (MPC), is used in this thesis. Because it has high robustness so it can deal with model uncertainties and other disturbances. A state estimator is designed in order to avoid using a lot of sensors. A new control structure is build up by combining the state estimator with traditional MPC controller. It is called Model predictive control with a state estimator (MPC-SE) controller. Computer simulations using MATLAB software are carried out. Proposed control structure is experimented on PC-based control system with RT-ADC4/PCI card and real-time control software, VisSim. Positions of sensor and actuator are chosen different. This not only investigates non-allocation problem but also solves physical problem. It is impossible to locate two devices in only one point. Experimental results confirm the feasibility of the proposed controller as well.