Doctor of Philosophy (PhD)


Chemical Engineering

Document Type



It is hard to point a daily activity in which polymeric materials or plastics are not involved. The synthesis of polymers occurs by reacting small molecules together to form, under certain conditions, long molecules. In polymer synthesis, it is mandatory to assure uniformity between batches, high-quality of end-products, efficiency, minimum environmental impact, and safety. It remains as a major challenge the establishment of operational conditions capable of achieving all objectives together. In this dissertation, different model-centric strategies are combined, assessed, and tested for two polymerization systems.

The first system is the synthesis of polyacrylamide in aqueous solution using potassium persulfate as initiator in a semi-batch reactor. In this system, the proposed framework integrates nonlinear modelling, dynamic optimization, advanced control, and nonlinear state estimation. The objectives include the achievement of desired polymer characteristics through feedback control and a complete motoring during the reaction. The estimated properties are close to experimental values, and there is a visible noise reduction. A 42% improvement of set point accomplishment in average is observed when comparing feedback control combined with a hybrid discrete-time extended Kalman filter (h-DEKF) and feedback control only. The 4-state geometric observer (GO) with passive structure, another state estimation strategy, shows the best performance. Besides achieving smooth signal processing, the observer improves 52% the estimation of the final molecular weight distribution when compared with the h-DEKF.

The second system corresponds to the copolymerization of ethylene with 1,9-decadiene using a metallocene catalyst in a semi-batch reactor. The evaluated operating conditions consider different diene concentrations and reaction temperatures. Initially, the nonlinear model is validated followed by a global sensitivity analysis, which permits the selection of the important parameters. Afterwards, the most important kinetic parameters are estimated online using an extended Kalman filter (EKF), a variation of the GO that uses a preconditioner, and a data-driven strategy referred as the retrospective cost model refinement (RCMR) algorithm. The first two strategies improve the measured signal, but fail to predict other properties. The RCMR algorithm demonstrates an adequate estimation of the unknown parameters, and the estimates converge close to theoretical values without requiring prior knowledge.

Committee Chair

Romagnoli, José A.

Available for download on Sunday, September 08, 2019