Date of Award


Publication Type


Degree Name



Industrial and Manufacturing Systems Engineering


Recovery Indicators;Supply Chain Digital Twin;Supply Chain Recovery;Supply Chain Resilience


Waguih ElMaraghy


Jessica Olivares-Aguila



Creative Commons License

Creative Commons Attribution 4.0 International License
This work is licensed under a Creative Commons Attribution 4.0 International License.


ABSTRACT The increasing risks and disruptions faced by global supply chains have emphasized the need for enhanced supply chain robustness. Organizations must develop procedures to recognize, assess, and manage disruptive influences to remain competitive and resilient. This study focuses on establishing a technology framework, specifically a digital twin, to anticipate floods, a common natural disaster and disruptor, and predict recovery indicators using discrete events simulation and machine learning algorithms. The objective is to provide supply chain management with a decision-support tool for making informed decisions. This thesis presents a three-phase digital twin framework for supply chains. In the first phase, machine learning algorithms, logistic regression, and long short-term memory (LSTM) are trained using precipitation data from Kerala, India, to predict floods. LSTM performs slightly better than logistic regression, achieving 73% recall, 75% accuracy, and 84% area under curvereceiver operating characteristics (AUC-ROC) score in flood prediction. The second phase utilizes simulation tools to create what-if scenarios for disruptions in the supply chain. A discrete event simulation of a real-world supply chain network, driven by process flow logic, simulates operational disruptions. FlexSim is employed to mimic service-level failures, and the performance of the supply chain model is evaluated based on the average service level at distribution centers. In the third phase, a multi-layer perceptron neural network (MLPNN) is trained using simulated case scenario data features to forecast supply chain recovery after disruptions. The MLPNN considers multiple disruptive inputs and monitors mean squared errors (MSE) during training and validation. The steady test MSE reduction over the epochs represented the recovery indicator. Scenario 4 had an optimal test MSE of 2.06 at the 87th epoch, while scenario 5 had that same MSE value at the 100th epoch for distribution center 1. The validation MSE values of 1.11 occurred at epochs 78 and 100, respectively, for scenarios 6 and 7. This recovery indicator helps anticipate the time to restore service levels to pre-disruption conditions. Overall, this study develops a conceptual digital twin that can forecast the risk of severe supply chain interruptions and predict recovery indicators. By utilizing disruption prediction and recovery indicators, the output of the digital twin can assist supply chain planners in making decisions and strategies to ensure supply chain resilience. The proposed supply chain digital twin architecture serves as a valuable tool for planners to support decision-making processes. Keywords: Supply chain digital twin, supply chain resilience, supply chain recovery, recovery indicators

Available for download on Thursday, September 26, 2024