Leveraging Large Language Models for Process Parameter Optimization in 3D-Printed ABS Polymer Specimens

Abstract The optimization of process parameters in Fused Deposition Modeling (FDM) remains a critical challenge in predicting the mechanical strength of 3D-printed parts. Conventional methods, such as the Design of Experiments (DOE), are often costly and time-consuming, necessitating the exploration of Artificial Intelligence (AI)-based predictive models. This paper investigates the potential of Large Language Models (LLMs), including Microsoft Phi-2, Qwen2.5-Math-1.5B, DeepSeek-R1-Distill-Qwen-1.5B, and StableLM-3B-4e1t to optimize FDM process parameters for maximizing the mechanical strength of 3D-printed Acrylonitrile Butadiene Styrene (ABS) polymer specimens. We aim to enhance the prediction and optimization of critical mechanical property by applying few-shot inference with these advanced LLMs using tensile test data with varying print parameters. Model performance is analyzed using Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and R-squared (R²), highlighting each model’s effectiveness in identifying optimal settings for enhanced mechanical performance. Our results demonstrate the potential of LLMs in data-driven parameter optimization, providing a novel perspective on applying large-scale language models to predictive tasks in additive manufacturing. This approach opens new avenues for leveraging state-of-the-art AI to refine material properties in 3D printing processes.