Dynamic Optimization Of Molecular Weight Distribution For Batch Free Radical Polymerization

As the processing performance and the using performance of polymer is greatly determined by its molecular weight distribution (MWD), how to produce the polymer product with specified MWD is of great significance. Dynamic optimization on free radical polymerization will be studied for the polymer design. The molecular weight distribution will be calculated and optimized.Based on the improved Stickler-Panke-Hamielec model, the free radical polymerization of PMMA has been studied. The MWD calculation can be implemented by model decomposition and sequential variable decoupling method. Based on the MWD calculation, dynamic optimization of reaction conditions for specified molecular weight distribution can be implemented by optimizing the reaction temperature. Through parameterization of the temperature profiles, the dynamic optimization is converted into a nonlinear programming. Then the sequential quadratic programming algorithm, SNOPT, is applied for solving the optimization.In this thesis, the dynamic optimization of MWD has been successfully achieved with the following results:1. Constant temperature control profile optimization.The type of profile with constant temperature is first studied. Under this condition, the optimized variables are temperature level and time. Given a specified MWD curve, a dynamic optimization is constructed to search for the optimal profile to produce polymer meeting the specified MWD requirement. This specification is implemented by adding constraints with relative error at chain length less than 5%. Both the steady-state and the non-steady-state reaction products are considered. The simulation results shown that the specified MWD can be successfully met via the dynamic optimization.2. Varying temperature control profile optimization.As the constant temperature profile can only produce limited polymers with certain-shape of MWD, varying temperature profile is next studied. In this thesis, the multistage step-change profiles are used. Different optimization formulations have been constructed for the tasks the steady-state and the non-steady-state reaction products by utilizing the different features. Both cases can be successfully solved via the dynamic optimization.