Fast genetic algorithm approaches to solving discrete-time mixed integer linear programming problems of capacity planning and scheduling of biopharmaceutical manufacture

Karolis Jankauskas,Lazaros G Papageorgiou,Suzanne S Farid

doi:10.1016/j.compchemeng.2018.09.019

Karolis Jankauskas, Lazaros G Papageorgiou + Show 1 more

Open Access

https://doi.org/10.1016/j.compchemeng.2018.09.019

Copy DOI

Abstract

The previous research work in the literature for capacity planning and scheduling of biopharmaceutical manufacture focused mostly on the use of mixed integer linear programming (MILP). This paper presents fast genetic algorithm (GA) approaches for solving discrete-time MILP problems of capacity planning and scheduling in the biopharmaceutical industry. The proposed approach is validated on two case studies from the literature and compared with MILP models. In case study 1, a medium-term capacity planning problem of a single-site, multi-suite, multi-product biopharmaceutical manufacture is presented. The GA is shown to achieve the global optimum on average 3.6 times faster than a MILP model. In case study 2, a larger long-term planning problem of multi-site, multi-product bio-manufacture is solved. Using the rolling horizon strategy, the GA is demonstrated to achieve near-optimal solutions (1% away from the global optimum) as fast as a MILP model.

Highlights

Biopharmaceutical drug development requires substantial investments of capital, human resources, and technological expertise
This paper presents a fast genetic algorithm (GA)-based approach to both mediumand long-term capacity planning and scheduling of singleand multi-site biopharmaceutical manufacture using discrete-time models
The GAs discussed in the previous sections for case study 1 and case study 2 are used to solve the respective scheduling problems, and the results are compared with the recreated mixed integer linear programming (MILP) models

Summary

Introduction

Biopharmaceutical drug development requires substantial investments of capital, human resources, and technological expertise. The cost of development of a single drug entering human trials between 1989 and 2002 was estimated to be in excess of $800M (DiMasi et al, 2003). The likelihood of a new biopharmaceutical drug product gaining approval for marketing and the rate of approval for new products has been getting lower over the years. According to Kaitin and DiMasi (2010), only one in six new drugs that entered clinical trials in the United States during 1993–98 and the 1999–2004 sub-periods were successfully approved for marketing. Given the high costs and the uncertainty of the biopharmaceutical development process, building new capacity for products which may or may not reach the market is not the most desirable option.

Methods

Results

Conclusion