LLM-Empowered Monitoring and Process Optimization of Additive Manufacturing

Authors

• Amogh Modgekar Desai, University of Texas at ArlingtonFollow

Document Type

Article

Abstract

Although widely adopted, additive manufacturing (AM) processes such as wire arc additive manufacturing (WAAM) are prone to defects caused by process uncertainties, yet most existing monitoring solutions operate as centralized, black-box systems that lack interpretability and real-time decision support. This work develops the mitigation and interpretation components of an edge-enabled intelligent framework for WAAM. Building on a teammate’s edge-deployed YOLOv8 model for real-time weld defect detection, I designed and implemented an API-driven system that integrates Retrieval-Augmented Generation (RAG) with a large language model (LLM) to provide evidence-based guidance for process optimization. A domain-specific knowledge base of research literature is constructed to retrieve defect-relevant publications, which are supplied to the LLM along with detected defect types and welding parameters. The LLM generates literature-backed mitigation recommendations and supports interactive, human-in-the loop dialogue, enabling secure, interpretable, and real-time decision support for defect reduction in AM systems.

Disciplines

Artificial Intelligence and Robotics | Databases and Information Systems | Industrial Engineering | Software Engineering | Theory and Algorithms

Publication Date

Spring 5-25-2026

License

This work is licensed under a Creative Commons Attribution 4.0 International License.

Recommended Citation

Modgekar Desai, Amogh, "LLM-Empowered Monitoring and Process Optimization of Additive Manufacturing" (2026). 2026 Spring Honors Capstones Projects. 9.
https://mavmatrix.uta.edu/honors_spring2026/9