Dynamic Optimization of an Emulsion Polymerization Process Using an Embedded Monte Carlo Model for Bimodal MWD

摘要

The molecular weight distribution(MWD)of polymers affects many end-use properties and is therefore a major production target.The desired MWD can also be multi-modal,which can be achieved by using chain-transfer agents.As these products are specialty chemicals,they are most often produced in semi-batch operation leading to inherent nonlinear dynamics.By using dynamic optimization,production time can be reduced while producing the targeted quality.In this work,we consider the model-based dynamic optimization of an emulsion polymerization process to achieve a bimodal MWD while reducing the batch time.The degrees of freedom consist of the isothermal reactor temperature,feed rates of monomer,initiator and chain-transfer agent.We use a combined model consisting of a deterministic kinetic model and a stochastic Monte Carlo polymer architecture model.The kinetic model describes macroscopic variables,such as concentrations.Distinct chains are simulated in a polymer particle using the Monte Carlo model,and many particles are computed in parallel.The time-varying reaction rates used in the Monte Carlo model are computed in the kinetic model.By taking all simulated chains and their respective weights,the MWD can be constructed.A Monte Carlo approach is chosen as it allows to simulate properly the transfer to polymer and branching reactions.To solve the dynamic optimization problem,we select a derivative-free surrogate model based optimizer due to the stochastic nature of the Monte Carlo model.For the Monte Carlo model,gradients are not readily available.The results show a reduction of the batch time between 6.2 and 7.5 % compared to the base recipe while the product quality is satisfied.