Access Type

Open Access Dissertation

Date of Award

January 2021

Degree Type


Degree Name



Industrial and Manufacturing Engineering

First Advisor

Ratna Babu Chinnam

Second Advisor

Saravanan Venkatachalam


In today's environment with high global and complex supply chains for engineered products, the ability to assess and manage the resilience of supply chains is not a luxury but a fundamental prerequisite for business continuity and success. This is particularly true for firms with deep-tier supply chains, such as the automotive original equipment manufacturers (OEMs) and their suppliers. Automotive supply networks are particularly facing growing challenges due to their complexity, globalization, economic volatility, rapidly changing technologies, regulations, and environmental/political shocks. These risks and challenges can disrupt and halt operations in any section of the supply network. Given that supply chains have become quite lean in the 21st century with relatively little slack, the COVID-19 pandemic has fully exposed these vulnerabilities. According to Allianz's Business Risk Report from 2014, half of all supply chain disruptions stemming from tier-2 and tier-3 suppliers. However, the industry's supply network assessment practice is primarily limited to immediate (i.e., "tier-1") suppliers with no real consideration for the deep-tiers. The added complication due to poor supplier relations is that there is no visibility to the upstream deeper-tiers of the supply network, which could lead to severe vulnerabilities and impose massive disruption costs.

Our research goal is to enhance the resilience of deep-tier automotive supply networks through improved resilience assessment and management mechanisms. In this collaborative study with a global automotive OEM (Ford Motor Company), we seek to develop methods to assess and manage the resilience of deep-tier supply networks. This research considers the multi-dimensional nature of resilience management, focusing on metrics around cost efficiency, effective inventory management, demand fulfillment, capacity management, and delivery performance. We develop and evaluate our proposed resilience assessment and management framework with a real case study supply network for an automotive climate control system. The supply network contains 20 firms (nodes) located in various global regions and 21 connections (edges) between firms. The network includes three tiers of suppliers with different transportation modes, making the network a rich illustrative example for proposed resilience assessment and management methods and analysis. All inventory and shipping policies with related parameters have been defined and set for each supplier and their connections.

The proposed resilience assessment framework relies on discrete-event simulation for effectiveness; computational efficiency is maintained by relying on modern open-source packages for modeling, optimization, and analysis. The framework starts by generating a digital supply network model that includes the focal firm and its suppliers and deeper-tiers based on the available visibility. Disruption scenarios, including disruption sources, frequency, and severity, are then efficiently generated using private and public regional risk sources. For illustrative purposes, we primarily relied on public secondary data sources. The secondary regional risk indices that we relied upon aggregate political, economic, legal, operational, and security risks for the given region. Finally, the digital supply network is simulated with an adequate number of replications for reliable assessment. In this research, discrete-event simulation is implemented using NetworkX and SimPy Python packages. We employ the network analysis techniques combined with discrete-event simulation informed by secondary data sources for improving the assessment framework. Our resilience assessment results confirm that visibility into the deeper-tiers of the supply network (through primary or secondary data sources) leads to a more accurate network resilience assessment. Finally, we offer a global sensitivity analysis procedure to determine the supply network players, parameters, and policies that most influence the network performance.

We also propose an effective resilience management framework that efficiently leverages simulation-based optimization. For illustrative purposes, we considered the mitigation strategies typical in the automotive industry, such as dual sourcing, reserve capacities (at primary or secondary suppliers), and contracts with backup suppliers besides carrying safety stock. Sourcing and transportation mode decisions can be easily incorporated into the framework. The method seeks to minimize the cost of risk mitigation strategies while attaining the target resilience. The framework is flexible and can entertain other objectives and constraints. Given that simulation-based optimization methods can be computationally expensive, we employ surrogate models that relate supply network resilience performance to network design parameters within our mathematical programming formulation. Without loss of generality, the surrogate models are based on linear regression models that define the relationship between the focal firm and tier-1 suppliers' resilience levels and network design decision variables. The imperfections of the regression models are accounted for in the formulation through constraints with slack (function of the RMSE of the regression model). We demonstrate that optimal resilience management would stem from jointly allocating safety buffers (e.g., capacity, inventory levels) across the network and not by independently applying a simplistic/static set of rules for all nodes/arcs. Our validation experiments with a real-world case study informed by secondary data from public data sources confirm the effectiveness and efficiency of the proposed supply network resilience management method.