Robustness of supply chain networks against underload cascading failures

TL;DR

This paper investigates supply chain network robustness against cascading failures caused by load decreases and fluctuations, highlighting the effectiveness of recovery strategies like surplus inventory and backup suppliers.

Contribution

It introduces an underload cascading failure model considering inventory and cost constraints, and compares robustness under different load scenarios and distributions.

Findings

Recovery strategies improve system robustness.

Supply chains are more fragile to demand shocks than load fluctuations.

System shows discontinuous phase transition under load decrease without recovery.

Abstract

In today's global economy, supply chain (SC) entities have become increasingly interconnected with demand and supply relationships due to the need for strategic outsourcing. Such interdependence among firms not only increases efficiency but also creates more vulnerabilities in the system. Natural and human-made disasters such as floods and transport accidents may halt operations and lead to economic losses. Due to the interdependence among firms, the adverse effects of any disruption can be amplified and spread throughout the systems. This paper aims at studying the robustness of SC networks against cascading failures. Considering the upper and lower bound load constraints, i.e., inventory and cost, we examine the fraction of failed entities under load decrease and load fluctuation scenarios. The simulation results obtained from synthetic networks and a European supply chain network [1] both confirm that the recovery strategies of surplus inventory and backup suppliers often adopted in actual SCs can enhance the system robustness, compared with the system without the recovery process. In addition, the system is relatively robust against load fluctuations but is more fragile to demand shocks. For the underload-driven model without the recovery process, we found an occurrence of a discontinuous phase transition. Differently from other systems studied under overload cascading failures, this system is more robust for power-law distributions than uniform distributions of the lower bound parameter for the studied scenarios.