Hybrid AI Improves 3D Printing Quality Prediction

This research investigates several advanced machine learning strategies for improving predictions in additive manufacturing, specifically focusing on Fused Filament Fabrication (FFF). The goal is to accurately predict critical properties like bond quality and porosity of 3D-printed parts. The core innovation lies in incorporating physics knowledge into data-driven deep learning models. Three main strategies are explored: embedding physics constraints directly into the neural network's loss function, using physics model outputs as additional inputs, and pre-training models with physics simulations before fine-tuning with experimental data. By combining these methods, the study demonstrates that hybrid machine learning models can enforce physically consistent relationships between bond quality and tensile strength. This approach yields accurate predictions even when experimental data is scarce, making porosity predictions more physically meaningful and reliable.