Use this URL to cite or link to this record in EThOS:
Title: Biopharmaceutical scheduling using a flexible genetic algorithm approach
Author: Jankauskas, Karolis
ISNI:       0000 0004 7964 9768
Awarding Body: UCL (University College London)
Current Institution: University College London (University of London)
Date of Award: 2019
Availability of Full Text:
Access from EThOS:
Full text unavailable from EThOS. Please try the link below.
Access from Institution:
The goal of biopharmaceutical capacity planning and scheduling is to identify an optimal production schedule (solution) that would satisfy multiple financial and operational objectives. It is a complex combinatorial optimisation problem characterised by features such as multi-product portfolios, variable process durations and yields, long product lead and approval times, and uncertain market forecasts. The bulk of research in the area of biopharmaceutical capacity planning and scheduling has focused on Mixed Integer Linear Programming (MILP) formulations. Heuristic optimisation methods such as Genetic Algorithms (GAs) have received very little attention even though they are reportedly more flexible, easier to implement and, in certain cases, have the potential of outperforming mathematical programming models. Therefore, this thesis addresses this knowledge gap by describing the development of a flexible GA-based Decision Support Tool (DST) for single- and multi-objective biopharmaceutical capacity planning and scheduling under deterministic and uncertain product demand. This thesis makes four broad contributions. Firstly, a GA is designed for solving biopharmaceutical capacity planning and scheduling problems using a discrete-time representation. The effectiveness of the algorithm is demonstrated on two industrial case studies and compared with discrete-time MILP models from the literature. A rolling time horizon strategy is applied to improve solution quality and the performance of the GA. A Particle Swam Optimisation (PSO) algorithm is utilised as a metaoptimiser to automatically tune the parameters of the GA. Secondly, a novel variablelength chromosome structure and an entirely new continuous-time scheduling heuristic are developed for more realistic and efficient medium-term scheduling of biopharmaceutical manufacture. The variable-length chromosome enables the GA to generate production schedules from a single gene. The novel variable-length GA with an embedded continuous-time scheduling heuristic is shown to outperform related discrete- and continuous-time MILP models on two literature-based examples. Thirdly, a multi-objective component is added to the variable-length GA and the continuous-time scheduling heuristic is extended with additional constraints and features, including rolling product sequence-dependent changeovers and lengthy product quality control and assurance (QC/QA) checks. A real-life industrial case study is used to demonstrate the functionality and benefits of the multi-objective optimisation. The multi-objective variable-length GA is used to optimise both the total production throughput and monthly product inventory levels of a multi-product biopharmaceutical facility. Finally, the multi-objective variable-length GA is combined with a Graphics Processing Unit (GPU)-accelerated Monte Carlo simulation for biopharmaceutical capacity planning and scheduling under uncertain product demand. The merits of the approach are highlighted by comparing the production schedules generated when the uncertainty in demand is ignored and when it is accounted for by characterising it with a probability distribution. In the final sections of this thesis an example of a commercial application of this work is presented.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available