(94g) Multi-Objective Surrogate-Based Optimisation for Sustainable Pharmaceutical Manufacturing

The pharmaceutical industry faces growing pressure to enhance sustainability while maintaining efficient and reliable manufacturing processes. Advanced process modeling techniques provide valuable insights, but their high computational complexity often makes direct optimisation impractical. To overcome this challenge, surrogate-based optimisation has emerged as an efficient and practical solution that enables the optimisation of manufacturing flowsheets to meet sustainability goals.

In this work, we introduce a multi-target tree regression approach for surrogate-based optimisation, focusing on sustainability. This mathematical programming method optimises green metrics, such as process mass intensity, while ensuring high yield and purity. By generating Pareto fronts, we visualise trade-offs between competing objectives, aiding decision-making. Our findings demonstrate that mathematical programming surrogate models can effectively approximate complex process behaviors, improve green metrics, and support sustainable pharmaceutical manufacturing. This approach enhances interpretability, computational efficiency, and practical applicability, offering a powerful tool for decision-making in pharmaceutical manufacturing and beyond.

References

Bertsimas, D., Dunn, J., Paschalidis, A. 2017. Regression and classification using optimal decision trees. 2017 IEEE MIT Undergraduate Research Technology Conference (URTC), 1–4. 10.1109/URTC.2017.8284195.

Misener, R., Biegler, L. 2023. Formulating data-driven surrogate models for process optimization. Computers & Chemical Engineering, 179, 108411. 10.1016/j.compchemeng.2023.108411.

Tian, H., Ierapetritou, M. G. 2023. A surrogate-based framework for feasibility-driven optimization of expensive simulations. AIChE Journal, 70 5, e18364. 10.1002/aic.18364.