2025 AIChE Annual Meeting

Supply Chain Resilience through Monitoring, Real-Time Optimization and Continuous Learning

Authors

Miriam Sarkis - Presenter, Texas A&M Energy Institute
Shivam Vedant, Texas A&M University
Eleftherios Iakovou, Texas A&M University
Efstratios Pistikopoulos, Texas A&M Energy Institute, Texas A&M University
Supply chain resilience relates to the capacity of a manufacturing network to withstand and recover from disruptive events [1,2,3]. On the one hand, resilience is enhanced through proactive measures at the design and planning stage, to make networks resilient against known and, increasingly, unknown unknowns and to ensure business continuity. On the other hand, following a disruptive event, reactive measures are needed for response and the timely recovery of network performance. Post-disruption, the mitigation of lengthy recovery times relies on the ability of the system to learn and integrate new acquired knowledge regarding uncertainty.

To this end, this work presents an end-to-end supply chain network optimization and control tool for the monitoring and operation of manufacturing systems, leveraging on a rolling horizon-moving horizon estimation approach. Initially, a scenario-based optimization problem is solved based on an approximation of the available uncertainty information. Here, the set of sampled scenarios represents the knowledge of uncertain parameters based on fitted historical data of known unknowns. During simulation, optimal planning decisions are gradually implemented and fixed and system states are computed. Upon realization of an unforeseen uncertainty, the ability of the manufacturing network to withstand the disruption through dynamic adjustments of decision variables and corresponding recovery times are quantified. Post-recovery, the tool integrates a methodology for real-time uncertainty based on the new gathered uncertainty information and recorded network trends. The methodology is designed to trigger resampling through detection of significant deviations of real-time realized uncertainty data from the assumed pre-disruption uncertainty information. Overall, the ability of the network to learn and improve its performance compared to initial states is quantified.

The proposed methodological tool enables systematic assessments of supply chain designs alongside design-specific recourse actions and constraints. Hence, design strategies are compared with respect to key performance indicators of: (i) expected cost pre-disruption; (ii) time to recover (TTR) post-disruption; and (iii) learning capacity, which is captured as the TTR reduction percentage following disruption.

[1] Chrisandina, N. J.; Vedant, S.; Iakovou, E.; Pistikopoulos, E. N.; El-Halwagi, M. M. Metrics and methods for resilience-aware design of process systems: advances and challenges. Current Opinion in Chemical Engineering 2024, 43, 100984.

[2] Bechtsis, D., Tsolakis N., Iakovou E., Vlachos D., 2022. “Data-Driven Secure, Resilient and Sustainable Supply Chains: Gaps, Opportunities, and a New Generalised Data Sharing and Data Monetisation Framework”; International Journal of Production Research; Vol. 60, No. 14, pp. 4397-4417; DOI: https://doi.org/10.1080/00207543.2021.1957506 .

[3] How to build more Secure, Resilient, Next-Gen U.S. Supply Chains, Eleftherios Iakovou, Chelsea C. White III (Georgia Tech), Brookings Institute TechStream, Dec 3, 2020; https://www.brookings.edu/techstream/how-to-build-more-secure-resilient-next-gen-u-s-supply-chains/ .