PLS-Based Cycle-to-Cycle and within-Cycle Control Strategy for a Simulated Moving Bed Process
| Autor: | Kai-Ting Hsieh, 謝凱珽 |
|---|---|
| Rok vydání: | 2007 |
| Druh dokumentu: | 學位論文 ; thesis |
| Popis: | 95 Separation of enantionmers is common in industries, with chromatography widely used. To overcome discontinuity of batch operation, simulated moving bed (SMB) is developed. The SMB model is multi-variate and contains difficult to obtain physical parameters. Design and control based on physical model is a rigorous process and its accuracy is affected by the presence of disturbance, whereas empirical model requires continuous correction and no physical characteristic can be observed. Hence, a strategy which employed the empirical modeling with physical model acting as a constraint is proposed. Two method of control is proposed. In the first method, partial least squares (PLS) is applied. It decouples the variables and transformed the MIMO process into a series of uni-variate SISO control problem. The model is updated by double exponentially weighted moving average (dEWMA). This is carried out in two ways, at the end of cycle and within each switching. The controller is designed for set point tracking as well as disturbance rejection. The second method employs wave propagation velocities as the input variable. The actual process is represented implicitly because wave propagation is a derivation of physical model. In control, the direction of improvement can be determined by visualization and acts as a constraint to improve control performance. The performance is evaluated by implementation on a virtual eight column SMB unit for set point tracking and disturbance rejection. |
| Databáze: | Networked Digital Library of Theses & Dissertations |
| Externí odkaz: |
|