| The feedback control of a discrete representation of emulsion polymerization particle size distribution (PSD) is investigated. The relative particle growth of a bidisperse polystyrene seed is controlled via manipulation of the monomer feed in a custom-built reactor control facility.; The facility is designed and constructed to overcome obstacles which impede the implementation of advanced control methodologies to emulsion polymerizations. It is capable of isothermal conditions, semi-batch operation, creating a suitable mixing and reaction environment, and measuring critical latex characteristics on-line. The on-line sensors include a density meter for near-continuous conversion information and an on-line adaptation of capillary hydrodynamic fractionation (CHDF) for particle size. A distributed-intelligence computer system manages, in real-time, the automation, communication, data acquisition, and control of the facility.; Two dynamic growth models, validated with on-line data from the facility, are developed. A monodisperse model is used to investigate the interesting dynamics which occur in the semi-batch reactor including pseudo-steady states, autoacceleration of the rate, and limiting conversions. A bidisperse model is used to study the suitability of the growth mechanism to influence relative growth via a local stability analysis and a time-varying output controllability analysis. The validation of this model proves that competitive growth experiments provide richer data than monodisperse experiments to distinguish between different radical absorption mechanisms.; A combined state estimator and controller, made up of a parameter adaptive steady state Kalman filter and a nonlinear shrinking horizon model predictive controller, is designed and implemented in the facility to control diameter ratio of the bidisperse system. The estimator estimates the system states, measured variables, controlled variable, and two model parameters from frequent density measurements and infrequent particle size measurements. The controller incorporates these estimated states and the adapted parameters to predict the diameter ratio at the end of batch using the nonlinear bidisperse model. Monomer feed rate moves are calculated from an iterative optimization that minimizes deviations from a final diameter ratio set point. The ability of the framework to adequately realize final set points and reject disturbances for several semi-batch emulsion polymerizations is demonstrated. |
