Real-time optimization of semi-batch reactors

Semi-batch processes are frequently used in the industry for the production of high value products such as pharmaceuticals, fine chemicals, and fragrance chemicals. These processes are inherently dynamic in nature and very often several batches are run in the same equipment. This may cause batch-to-batch variation in operating conditions resulting in problems in safety, product quality and productivity. Furthermore, the operating strategy may not be the optimal strategy in presence of parametric uncertainty and disturbances. One approach to address these issues is to develop accurate mathematical models and use off-line optimization techniques to find the optimal operating policies for each batch operation. However, accurate mathematical models are very time consuming to develop. In this thesis, a real-time optimization framework is developed to account for uncertainty. First, an approximate model is utilized to compute optimal solution. Then, periodic corrections are made to the optimal solution during the batch operation to account for uncertainty. Two different approaches are proposed to make these periodic corrections: (1) a measurement-based approach and (2) a simulator-based approach. The proposed optimization algorithm is applied to four different classes of single reactions that can take place in a semi-batch reactor. Finally, the results of the proposed optimization methodology are illustrated via several simulation examples.