On Friday, 18 March 2022, Yesim will defend her PhD thesis Optimal Decision Making under Uncertainty in Biomanufacturing”. This thesis has been supervised by prof.dr.ir. Ivo Adan and dr. Tugce Martagan. The ceremony will take place at Eindhoven University of Technology, in the Atlas Building, Room 0.710, at 13:30.

Summary

Biomanufacturing methods use living organisms (i.e., viruses and bacteria) to generate active ingredients. This leads to challenges that are different than those in any other industry. In this thesis we address these challenges and develop optimization models to improve biomanufacturing efficiency. In collaboration with MSD Animal Health (MSD AH), we focus on a novel technique: bleed-feed. Bleed-feed is promising in skipping intermediary setups. However, its optimal implementation involves unique trade-offs and challenges, and its potential benefits are not fully understood by the industry yet. We combine the biological dynamics of fermentation and the operational trade-offs of bleed-feed, and present stochastic optimization models. We investigate the optimal bleed-feed decisions in different contexts and generate insights for practitioners.

More specifically, we first develop a finite-horizon, discrete-time Markov decision processes (MDP) model to determine condition-based bleed-feed policies that maximize expected total yield per batch. We analyze the structural characteristics of optimal policies and show that optimal bleed-feed policies have a three-way control-limit structure under mild conditions. We characterize the behavior of the value function as a function of regulatory restrictions. Our analysis reveals that the marginal benefits of an additional bleed-feed are decreasing and converge to a certain value. Second, we use renewal reward theory to determine time-based bleed-feed policies maximizing the expected fermentation throughput. We analyze structural properties of the optimization problem and present a case study from MSD AH. Through several practically relevant scenarios, we assess the potential impact of implementing bleed-feed on current practice. We analyze under which settings implementing bleed-feed is not desirable. Next, we extend our renewal model by including practically relevant constraints. These constraints ensure that bleed-feed is implemented during a shift and consider biomanufacturer’s risk-averse behavior while finding optimal bleed-feed policies. We generate managerial insights through numerical analysis. By relaxing the constraints, we investigate the value of flexibility in bleed-feed decisions. Finally, we present a portfolio of decision support tools to improve biomanufacturing efficiency, consisting of the bleed-feed, yield optimization and rhythm wheel tools, using variety of operations research methods, such as renewal reward theory, Bayesian design of experiments, and simulation-optimization. We elaborate on development of the tools and their implementation at MSD AH. Thanks to these tools, MSD AH had a significant impact with up to 50% increase in the batch yield and an additional revenue of €50 million per year. Additionally, our analyses show that bleed-feed implementation can provide benefits. Real-world implementation of bleed-feed at MSD AH resulted in an 85% improvement in the batch yield per setup using one bleed-feed.