Optimal operation of semi-batch polymerization reactors

  • Optimaler Betrieb von Semi-Batch Polymerisationsreaktoren

Faust, Johannes Michael Magnus; Mitsos, Alexander (Thesis advisor); Asua, José Maria (Thesis advisor)

Aachen : RWTH Aachen University (2022)
Book, Dissertation / PhD Thesis

In: Aachener Verfahrenstechnik series - AVT.SVT - Process Systems Engineering 26
Page(s)/Article-Nr.: 1 Online-Ressource : Illustrationen, Diagramme

Dissertation, Rheinisch-Westfälische Technische Hochschule Aachen, 2022


We address the optimal operation of emulsion polymerization reactors with different approaches for distinct polymer properties. Polymers are versatile products with wide ranges of properties. A distinct feature of polymers is that the reaction step is crucial for quality as modifications after the reaction are usually not feasible. Therefore, research on optimal operation of polymerization reactors is an active research area. While in-silico studies allow for the assumption of state-feedback, online measurements are required to assess the current state of the reactor in reality. Raman spectroscopy allows for online determination of individual components. In order to extract this information from the spectrum, we create a model using the method of indirect hard modeling. The model allows to determine concentrations of monomers and polymers online. The predictive capabilities of the model assessed in terms of root mean square error of prediction are good .One property of emulsion polymers is the particle morphology evolving from the successive polymerization of different monomers. The particle morphology can be described by population balances of polymer clusters. Due to the predominant production of these polymers in semi-batch reactors, recipes are of great interest. We compute optimal recipes by employing dynamic optimization. For the improved production of existing products, time-optimal recipes are sought. With this model-based approach, we can also determine if new product qualities can be produced using the given setup. Here, the objective is to minimize the error between the desired cluster distributions and the obtained distributions. In a next step, feedback control schemes for particle morphology are evaluated in-silico. We evaluate a nonlinear model predictive controller, which is tracking an offline computed optimal recipe. This controller uses the polymerization kinetics model only. A more complex scheme is the economic nonlinear model predictive controller aiming at the minimization of the batch time while achieving the desired particle morphology within tolerances. Different disturbances are studied for both controllers. A different property studied for optimal operation is the molecular weight distribution. Based on a hybrid model, containing a deterministic reaction kinetics model coupled with a Monte Carlo simulation for individual polymer chains, we compute optimal recipes. We extend this approach to bimodal molecular weight distributions using chain transfer agents. Chain transfer agents increase the reaction rate for chain transfer and therefore reduce the molecular weight of formed polymers. For all case studies considered, we follow a surrogate-based optimization route and determine batch time reductions. Overall, we contribute to the transition from fixed time-based recipe operation to optimal state-based operation for semi-batch polymerization reactors.