| A research study has been conducted to develop an advanced control strategy for a commercial continuous-flow stirred-tank terpolymerization reactor for the production of ethylene-propylene-diene rubber, and to optimize process operation for better product quality control. Based on an original program provided by the sponsor, a nonisothermal dynamic model for the reactor is simulated using an optimum set of kinetic parameters, determined from designed plant experiments and a multivariable direct search optimization technique. The kinetic model is validated using plant data and conventional plant operation guidelines. Results indicate that the dynamic reactor model is in a very good agreement with actual process data. The simulator is implemented on-line using a VAX station, and the plant database and control systems. Sensitivity analysis and identification are conducted in order to determine the extent of interaction among process variables and the effect of operating conditions. The open-loop responses of reactor output variables, due to step changes in feed stream conditions, are quantified in terms of process gain and response time. Reactor input variables affect polymer composition, molecular weight (or Mooney viscosity), polymerization rate and reactor temperature to different extents.; The application of on-line quality control for the continuous polymerization reactor is investigated. Results show that engineering process control concepts could be used, with statistical process control methods, for on-line control of key product quality variables. A multivariable dynamic matrix control technique is considered for a 3 x 3 system to control the compositions and Mooney viscosity of the polymer product. Simulation results prove the excellent capability of the control strategy to control product quality on-line during "desired" changes in product specifications (or setpoints), and process loads or disturbances. By selecting an appropriate set for controlled variable weights and manipulated variable "move suppression factors" for the multivariable system, the control technique is proved to handle process interactions very well, and give very satisfactory responses for changes in quality variable setpoints, production rate, and variations in catalyst activity. |
