| An anticipatory plant operation algorithm, the Adaptive Dynamic Optimization (ADO), is established for minimizing production cost and maximizing product quality and controllability for the semibatch emulsion polymerization process. The ADO is developed from application of concepts from modern systems and control engineering and optimization theory. Open-loop optimal recipe and reactant feed-rate policies for the semibatch emulsion polymerization of the model system methylmethacrylate are generated from each polymerization run, using a predictive, mechanistic 'tendency' model (Georgakis, 1986) and data filtering, fitting and forecasting methods. The behavior of each run is used in the automatic fitting of the model to maintain fidelity to measured plant variables as the ADO is applied in a serial manner, from run to run, over a polymerization campaign.; Optimal policies allow specified polymerization rate and average molecular weight distribution profiles to be followed, minimizing kettle time and interbatch product variability, subject to a maximum polymerization/heat generation rate safety constraint. These policies are responsive to uncontrolled, unmonitored, unmodelled, but relatively slow variations in reactant quality due to the presence of inhibitor, as well as other factors contributing to model/plant mismatch. Polymerization progress for each run is observed solely by online densitometry, and the ADO procedure is successfully validated on the lab-bench scale for a short series of polymerizations.; Additionally, simulations of closed-loop regulatory control the semibatch emulsion polymerization system using very rapidly decomposing ('hot') redox initiator (Jansen, 1987) are executed. The polymerization rate is shown to be theoretically (if not practically) controllable around a given constant value by adding hot initiator to the reaction mass based on either open-loop optimized policies, as above, or by scheduled gain feedback of the polymerization rate. |
