Improved scheduling model for a multistage biopharmaceutical process in a multiproduct facility

This study critically evaluates deterministic and stochastic/evolutionary strategies for long-term scheduling of multiproduct facilities in batch and continuous biopharmaceutical manufacturing, as reported in existing literature using discrete-time and continuous-time models. It identifies key limitations in literature model/algorithm, including (i) real-time storage violation, (ii) real-time violation of shelf-life due to its calculation over a single event, (iii) minimum campaign length calculated over single events leading to suboptimal use of resources, and (iv) early product deliveries. To overcome these limitations, a deterministic mathematical model is proposed extending the existing unit-specific event-based model(s) with additional features including (i) modified material balances, (ii) new minimum campaign length constraints captured over multiple events, (iii) new shelf-life constraints captured over multiple events, (iv) new setup constraints based on product sequencing, (v) updated tightening constraints, (vi) updated sales and penalty constraints for late deliveries, (vii) updated bounds, and (viii) updated objective function. These improved features lead to better modeling in terms of handling intermediate and final product storage without real-time violations of storage and shelf-life, handling of initial setup in production tasks, and timely delivery of final products. The proposed improved model yields up to 36 % increase in profit when applied on four literature examples.