Enterprise-wide optimization under uncertainties and disruption: developing a framework for integrated, resilient, and sustainable supply chain modelling and optimization
Date
2024
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
University of Delaware
Abstract
The fast-changing pace of globalization conditions and the unprecedented flow of information have transformed industries and their supply chains into complex, interconnected networks. This transformation introduces numerous challenges and opportunities for improvement in enterprise operations. While these advancements facilitate faster decision-making and improved efficiency, current enterprise systems are susceptible to issues such as the bullwhip effect, where small uncertainties propagate across the network, leading to amplified distortions in supply and demand signals; significant hurdles in recovering from disruptions; and difficulties in balancing objectives, particularly those related to sustainability. Consequently, the need for innovative frameworks to manage and mitigate the inherent complexities and risks within modern supply chains has become paramount. ☐ This thesis addresses these challenges through a structured approach divided into three parts: (i) developing a framework for data-driven integrated supply chain optimization, which focuses on leveraging advanced analytics to enhance decision-making and operational efficiency; (ii) designing resilient supply chains that ensure flexibility and robustness, enabling adaptation and recovery from uncertainties and disruptions; and (iii) developing a framework for sustainable supply chain design and applying this to plastic waste transformation. ☐ The benefits of an integrated supply chain model are numerous, including cost reduction, lean management, and reduced redundancy. However, implementing such models also presents challenges, particularly in managing the complexity related to multiple scales, often leading to the curse of dimensionality. These challenges result in a compromise between computational efficiency and quality of decision. Recognizing the importance of integrated models in supply chain optimization, the first part of this thesis addresses these challenges through a data-driven integration framework. Some constraints are replaced with a surrogate model developed from enterprise data and machine learning. This framework was applied to enterprise problems of integrating tactical planning with operational planning and scheduling. Dataset generated from a high-fidelity scheduling model was used to build this surrogate, capturing the feasibility space. To validate, we compared solutions from the surrogate model with those from a monolithic model which integrates planning and scheduling constraints. Case studies showed that the surrogate model provided non-inferior solutions with computational advantages, motivating a move to higher-level integration. This solution reduces computational time without compromising solution quality. ☐ The second part of this thesis addresses the critical issues of uncertainty and disruption within the supply chain, challenges exacerbated by increased competition and market volatility. Flexibility in planning decisions is crucial, making modular manufacturing a promising strategy to address these uncertainties. We developed a comprehensive framework for a modular supply chain, offering advantages such as economies of scale, reduced time to market, and improved management of demand and supply variability. To tackle disruptions, we integrated both proactive and reactive measures within the supply chain model, incorporating strategic planning, risk assessment, real-time adjustments, and contingency planning. Additionally, we proposed a stochastic modeling approach to handle uncertainties and disruptions more effectively, considering various probabilistic scenarios to enable dynamic adaptation. This stochastic model outperforms its deterministic counterpart by enhancing overall supply chain objectives like cost efficiency, reliability, and responsiveness, proving especially effective in situations where traditional models fall short. ☐ The third part of this thesis focuses on designing a sustainable supply chain model. There is increasing awareness of the impact that economic activity can have on environmental health. As economic activity scales, achieving sustainability becomes even more challenging. To address these issues, we developed an enterprise method to ensure that our supply chain operations are sustainable. This method integrates sustainable practices into every stage of the supply chain, from sourcing raw materials to delivering the final product, with decisions evaluated in a multi-objective fashion. This methodology was applied to a plastic waste supply chain problem, resulting in a sustainable network design and operations. By incorporating environmental considerations, the model demonstrates that it is possible to create a supply chain that supports economic growth while minimizing environmental impact and maximizing profit.
Description
Keywords
Machine learning, Mixed integer programming, Resilience, Supply chain optimization, Uncertainty disruption