Multi-objective dynamic optimization of ampicillin batch crystallization: Sensitivity analysis of attainable performance vs. product quality constraint

Antonios Dafnomilis, Samir Diab, Alistair Rodman, Andreas G. Boudouvis, Dimitrios Gerogiorgis

Research output: Contribution to journalArticlepeer-review

Abstract

Ampicillin is a broad-spectrum antibiotic and World Health Organization Essential Medicine whose crystallization is an important unit operation in its production. A published model for the solubility of ampicillin as a function of pH as well as growth and nucleation kinetics allows for dynamic simulation and optimization of its batch crystallization. While experimental approaches to investigating different dynamic pH profiles have been considered in the literature, dynamic mathematical optimization of pH modulations to meet specific production objectives for ampicillin batch crystallization has yet to be implemented; therein lies the novelty of this study. In this study, dynamic simulation and optimization of the batch crystallization of ampicillin are performed to establish optimal pH manipulations for different production objectives. Simulation of already published batch seeded ampicillin crystallization experiments is performed prior to definition and solution of a dynamic optimization problem for maximization of mean crystal sizes and minimization of size distribution width. The effects of seed loading, time domain discretization, and mean crystal size and size distribution width objective function weights are considered and discussed. Pareto fronts showing trade-offs between different objectives and constraints are then investigated.
Original languageEnglish
JournalIndustrial & Engineering Chemistry Research
Volume58
Issue number40
Early online date30 Aug 2019
DOIs
Publication statusE-pub ahead of print - 30 Aug 2019

Fingerprint Dive into the research topics of 'Multi-objective dynamic optimization of ampicillin batch crystallization: Sensitivity analysis of attainable performance vs. product quality constraint'. Together they form a unique fingerprint.

Cite this