Data analytics and simulation methodologies for adaptive supply chain risk management in bio-pharmaceutical manufacturing

Abstract

Specically, we propose innovated data analytics approaches to capture the important properties in all sources of real-world data, such as the skewness, multi-modality and dependence in the product demands. The proposed methods can improve the estimation and prediction accuracy. We also provide insights on the underlying generative process for the high dimensional/ non-stationary data streams.

To provide a reliable guidance for real-time decision makings, we build a calibration framework for simulation models. Data-driven stochastic optimization and simulation-based optimization methodologies are developed to further guide decision makings accounting for the overall uncertainty.

To assess the system risk performance and study the random behaviors, we develop simulation analytics and output analysis methodologies, and analyze the sensitivity of the system performance to each source of uncertainty. For complex end-to-end bio-pharmaceutical supply chains, metamodels are constructed as knowledge representation of the system input/output relationship and further employed to predict the system risk behaviors.

The objective of this dissertation is to improve the risk management in bio-pharmaceutical supply chains. The bio-pharmaceutical supply chain is a highly interactive complex system; there exist various uncertainties in supply, production, testing and demands; and the biopharmaceutical evolves rapidly. To address these challenges, we develop a biopharma supply chain risk management platform, which can guide both strategic and operational decision makings.