| This thesis presents a systematic approach towards setting up real-time control schemes to provide improved operation for the solution polymerization of methyl methacrylate (MMA) in continuous stirred-tank reactors. An automated pilot-scale experimental reactor system with facilities for on-line measurement of flow rate, temperature, density, etc. has been set up for control studies. Control of conversion (productivity) is the target of the first phase of this long-term project, with the control of both productivity and quality (molecular weight) being the target in the immediate future.; The control objective of this thesis is to maintain a desirable conversion level using initiator flow rate as the manipulated variable for disturbance rejection. During operation, known or unknown types and amounts of reactive impurities are injected into the reactor as disturbances. All monomer, initiator and solvent flow rates are regulated by computer-controlled variable speed dosing pumps in order to maintain constant monomer/solvent ratio and residence time. Data acquisition is facilitated by an Opto-22 digital and analog input/output system which operates as a slave device to the process computer to which various measurements from the reactor are interfaced.; A very accurate on-line densitometer, which relies upon changes in the frequency of oscillation of a fixed volume tube through which liquid is flowing, is used to monitor the density change and therefore conversion in the reactor system. In the later stages of this thesis, an on-line rotational viscometer has also been installed to monitor the viscosity of the reaction product. The on-line viscometer, in combination with the on-line densitometer, will be used for the future on-line multivariable control studies.; The reactor was first run under open-loop conditions using pseudo-random-binary-sequence (PRBS) perturbation and step response analysis techniques to identify a process transfer function model and to verify a mechanistic polymerization model developed for this process. Consequently, control algorithms were developed and implemented on the reactor system considering variations of the reactive impurities introduced in the feed streams as process uncertainties.; Conventional Proportional-Integral-Derivative (PID) control algorithms (e.g., PID, Smith Predictor, Dahlin's control) and stochastic control strategies (e.g, unconstrained minimum variance control (MVC), constrained MVC, and one-step optimal control) were developed and evaluated first at the simulation level to identify promising control runs. Then, closed-loop experimental runs followed in a series of six two- to three-day continuous runs to verify the simulation results. The experimental results have also been cross-checked by off-line gravimetry and gel-permeation chromatographic analysis.; In all, this study not only verifies theoretical control designs with closed-loop experimental runs, but also triggers many interesting control issues for practical polymer reactor control. There is no such study in the open literature at this point, at least to the author's knowledge. |
