Inferential MIMO predictive control of the particle size distribution in emulsion polymerization

A new inferential 2-step multiple input/multiple output (MIMO) model predictive control (MPC) of the particle size distribution (PSD) in emulsion polymerization processes is proposed. The bulk-like model describing the PSD is used with the material balances of initiator, radicals, monomer and surfactant. The inferential 2-step control strategy uses two measurements available online (without delay): the concentration of surfactant in the aqueous phase by conductimetry, and the concentration of monomer by calorimetry. In a first step, the optimal trajectory of surfactant concentration leading to the target PSD is calculated offline. In a second step, a multivariable model predictive control manipulates online the monomer and surfactant flow rates in order to track the precalculated surfactant concentration trajectory and to maximize the monomer concentration in the polymer particles in a constrained setpoint tracking. Two control strategies are compared (nonlinear MPC and linearized MPC) with and without modelling errors.