| About this Abstract |
| Meeting |
2025 TMS Annual Meeting & Exhibition
|
| Symposium
|
Additive Manufacturing Modeling, Simulation and Machine Learning
|
| Presentation Title |
LLM Agents for 3D Printing Error Detection and Correction |
| Author(s) |
Yayati Jadhav, Peter Pak, Amir Barati Farimani |
| On-Site Speaker (Planned) |
Yayati Jadhav |
| Abstract Scope |
Additive manufacturing, particularly fused deposition modeling (FDM), has revolutionized production by creating customized, affordable products with minimal material waste. Despite its advantages, FDM often encounters errors that require expert intervention to detect and mitigate defects, impacting print quality. Existing automated and machine learning models for error detection lack generalizability across diverse 3D printer setups and require extensive labeled datasets, limiting scalability. This study introduces a novel framework leveraging the embedded information and emergent reasoning capabilities of pre-trained large language models (LLMs) to detect errors and seamlessly integrate with 3D printers, ensuring high-quality prints. The LLM evaluates print quality based on images of each layer, identifies failure modes, queries relevant print parameters, and executes a solution plan. This process ensures error-free subsequent layers by addressing defects in previous layers. Our results demonstrate that LLM-based agents can autonomously detect, analyze, and resolve failures without human intervention. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Additive Manufacturing, Machine Learning, |