Optimization-based digital design of a commercial pharmaceutical crystallization process for size and shape control

This work demonstrates a systematic development of a digital twin of a commercial active pharmaceutical ingredient (API) - Compound A - crystallization processes to be used for in-silico design of experiments (DoE) and process optimization. Eleven highly corelated kinetic parameters were estimated by exploiting an optimization approach with an evolutionary algorithm, to fit to concentration profile, mean size and the shape distribution of the product. Two more experiments were then performed to validate the model showing that the developed model could predict the product size and shape properties. The digital twin of the process was used to perform in silico design of experiments to understand the attainable operating space of the system and provide a framework for rapid quality-by-design (QbD). Subsequently the model was used in a dynamic optimization framework for the digital design of the API crystallization process to achieve the critical quality attributes (CQAs) without extensive experimentation.