AI-DRIVEN DIGITAL THREAD FRAMEWORK FOR END-TO-END LIFECYCLE OPTIMIZATION IN ETO MANUFACTURING SYSTEMS

Abstract

Engineer-to-Order (ETO) manufacturing systems represent the most complex segment of production paradigms, characterized by high customization, prolonged design cycles, and fragmented information silos. Traditional Product Lifecycle Management (PLM) and Manufacturing Execution Systems (MES) struggle to maintain semantic consistency across bidding, design, procurement, production, and field service. This paper proposes a novel AI-driven Digital Thread framework (AI-DThread) that leverages Large Language Models (LLMs), Graph Neural Networks (GNNs), and reinforcement learning to achieve end-to-end lifecycle optimization. Unlike conventional model-based definition (MBD) approaches, the proposed framework creates a dynamic, bidirectional knowledge graph that connects unstructured customer requirements (RFQs) to low-level machine toolpaths. Through a case study of a heavy-duty ETO machinery manufacturer, we demonstrate a 31% reduction in engineering change orders (ECOs), 24% compression of order-to-delivery lead time, and 18% decrease in warranty claims. The paper concludes with an implementation roadmap and discussion of organizational enablers for AI maturity in ETO contexts.