The development of a nonlinear dynamic model of the seeded semicontinuous emulsion copolymerization of vinyl acetate/n-butyl acrylate for copolymer composition control is presented. The model accounts for the monomer partitioning between the two phases in the system by employing the concept of a "quasi-static" thermodynamic equilibrium with expressions based on the Flory-Huggins theory. An excellent agreement with experimental data is reported.;Because the available process measurements do not yield the entire state vector directly, nonlinear state estimation was used. Process disturbances and measurement errors were modeled as stochastic processes and a hybrid extended Kalman filter is employed for state estimation. The filter was based on the local linearization of the process model around the sub-optimal filter estimates which are needed for the effective composition control of the produced copolymer. Different Kalman filter designs were developed and implemented in real time to assess the estimator performance. A noise adaptation scheme which estimates the process and measurement noise levels on-line was found to increase dramatically the robustness of the filter.;A combined deterministic nonlinear feedforward/feedback control policy was employed to achieve the desired copolymer composition of the latex particles. A significant improvement over current practices was observed. The design of a distributed-intelligence computer system for real-time implementation of the state estimation and control strategies is also discussed.
展开▼
