| About this Abstract |
| Meeting |
2025 TMS Annual Meeting & Exhibition
|
| Symposium
|
Additive Manufacturing Modeling, Simulation and Machine Learning
|
| Presentation Title |
Large Language Models for Distilling Knowledge in Additive Manufacturing
|
| Author(s) |
Achuth Chandrasekhar, Jonathan Chan, Francis Ogoke, Olabode Ajenifujah, Amir Barati Farimani |
| On-Site Speaker (Planned) |
Achuth Chandrasekhar |
| Abstract Scope |
Additive manufacturing (AM), or 3D printing, is transforming industries by allowing intricate designs, minimizing waste, and speeding up prototyping. However, keeping up with the rapidly growing knowledge base in AM is challenging. General large language models (LLMs) like GPT-4 often fall short in delivering detailed, specific answers needed by materials science researchers. To address this, we present "AMGPT," a specialized LLM text generator for metal AM. AMGPT leverages a pre-trained Llama2-7B model in a Retrieval-Augmented Generation (RAG) setup, integrating information from approximately 50 AM papers and textbooks. Using Mathpix, we convert these documents into TeX format for seamless integration into the RAG pipeline managed by LlamaIndex. Expert evaluations indicate that AMGPT improves response times and maintains coherence, demonstrating the effectiveness of specialized models in providing precise, relevant information to advance metal AM research. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Additive Manufacturing, Machine Learning, Modeling and Simulation |