Semiconductor Supply Chain Synthetic Data Generation and Network Analysis under Disruption of System Order 125 views

Semiconductors are critical inputs for a variety of consumer goods and technology systems, but their supply chains are prone to disruptions including natural hazards, technological advances, counterfeiting, and geopolitical tensions causing cascading impacts. Modeling, simulation, and resilience analysis efforts aim to reduce the impact of disruptions; however, these methods are often limited by the shortcomings of available supply chain data. This thesis introduces a semiconductor supply chain synthetic data generation framework using large language models and presents a sample integration of the synthetic data with a network modeling application to help address data scarcity and demonstrate an application for the synthetic datasets. The data generation framework incorporates two-stage verification and knowledge graphs to improve synthetic data quality. Four prompting strategies and six models are compared on their ability to generate synthetic data including their verification success rates, error correction efficacy, and resource usage providing best practices for generating synthetic data. The synthetic datasets are integrated into a network-based system order analysis of the supply chain entities. The datasets are converted into single-layer and multilayer network representations, and seven centrality metrics are used to determine a baseline system order. Scenarios, including natural hazard, geopolitical, and technological disruptions, are applied to the network to determine the system order sensitivity to disruption and the most disruptive scenarios. The contribution of this thesis is thus the novel synthetic data generation and verification framework for semiconductor supply chains as well as extending network-based system order analysis through study of novel network and scenario constructions and centrality metrics. The methodology and results provide insights into semiconductor supply chains and analysis approaches for stakeholders across industry, government, and academia.

MS (Master of Science)

Supply Chain Management; Large Language Models; Centrality Measures; System Order Analysis; Risk Analysis; Resilience; Knowledge Graphs

English

2026-04-23