Tuesday, October 18, 2011: 4:30 PM
102 D (Minneapolis Convention Center)
The proliferation of biologic based therapeutics and the complex nature of their manufacturing makes accurate detailed methods for generating production schedules valuable. The large capital costs and expenses in these plants, the convergent flow of buffer and media preparation to support mainline manufacturing, the need to satisfy regulations and medical demands, yet revise plans and designs in the face of changing knowledge, have motivated our work. Mathematical programming has proven to succeed when we layer it with a well-developed framework for model specification, algorithm customization, scenario analysis, and graphical results. We demonstrate an approach that can apply widely to biologics and allow for further custom detail. We discuss biologic processes in general and illustrate the methodology on perfusion based processes, which present some unique challenges to mathematical programming based approaches. The inputs take a natural, concise format to describe the plant and process steps. The model captures how perfusion reactors have varying yield over time with uncertainty. It also schedules supporting activities to supply growth media and buffers, clean in place, and satisfies restrictions on storage vessels and duration. The formulation ordinarily covers time scales from the daily variation of yield to many months for capacity analysis case studies. Scheduling has become rapid enough for large-scale scenarios to mine for data, helping to answer design questions and support routine rescheduling.