Resilience by Teaming in Supply Networks
In an increasingly interconnected world, physical, service, and digital supply networks are becoming progressively more complex, dynamic, and interdependent. As this transformation occurs, it gradually reduces supply networks’ capacity to detect all possible vulnerabilities and events leading to disruption. In the past, in order to survive disruptions, supply networks have resorted to protective measures entailing the use of stand-by resources. However, of late, sustainability concerns push for higher efficiency in the use of resources, reducing protective redundancies and making supply networks more susceptible to disruptions. Supply network resilience is an emerging concept related to the inherent ability of a network to tolerate and overcome disruptions, a central capacity to achieve long-term sustainable operations. Nevertheless, current understanding of resilience fundamentals is still at its early stages. Furthermore, there is a dearth of research addressing strategies and protocols for supply network agents to design locally and globally resilient supply network structures, and achieve resilient operations, with minimal use of resources, especially when resources are fault-prone. In this dissertation, a set of resilience fundamentals is derived from research in physical, service, and digital supply networks’ literature. Based on the fundamentals of resilience and inspired by the Fault-Tolerance by Teaming principle of Collaborative Control Theory, the Resilience by Teaming framework (RBT) is developed to address the structure and control dimensions of resilience, both at agent and network level. RBT comprises a suite of agent-level protocols and design rules capable of leveraging the performance of weaker agents through teaming. Local decisions made under RBT protocols not only increase agent-level resilience but also lead to emergent resilient supply networks. Furthermore, as teaming enables agents to lower the level of resources used as stand-by protection vs. traditional approaches, RBT effectively allows a more sustainable achievement of resilience in supply networks. Resilience by Teaming protocols are applied to three case studies: (1) control of unreliable production lines to maximize throughput while minimizing WIP and throughput variability, (2) design and control of distribution networks for small parcel delivery under network congestion and disruptions, and (3) supply network formation and re-configuration for resilience to targeted and random disruptions. Results from case study (1) show that RBT protocols for control of agents’ internal resource networks, such as production lines, increase throughput by 0.45% while simultaneously reducing WIP 24% and throughput variability by 43%, vs. traditional production line control methods. Hence, RBT protocols effectively reduce the level of stand-by resources, in the form of WIP, required to cope with disruptions and are able to produce a higher, and more stable, production line output. Case study (2) results indicate that RBT protocols for design and control of distribution networks achieve equal performance to protocols optimized for delivery timeliness, but require 4.5% lower costs to cope with congestion and disruptions. This advantage persists over a wide range of congestion and disruption scenarios. Therefore, RBT protocols successfully leverage available distribution network flexibility by teaming resources, when needed, to anticipate and overcome congestion and disruptions. Finally, case study (3) analyses the effect of including RBT agents in a supply network on structure and resilience to disruptions, from a global perspective. Results show that, when at least 20% of agents in a supply network use RBT protocols, structure is modified to increase interconnection of a subset of agents which, in turn, reduces QoS loss under targeted and random disruptions vs. networks without RBT agents. Furthermore, the cost of overcoming disruptions is lower when RBT agents are present in the structure. Therefore, RBT protocols are able not only to increase agent-level resilience and sustainability, as shown in case studies (1) and (2), but also lead to the emergence of more resilient and sustainable supply networks.
Nof, Purdue University.
Industrial engineering|Systems science
Off-Campus Purdue Users:
To access this dissertation, please log in to our