Surrogate-Assisted Adaptive Experimentation for Fused Filament Fabrication Process Optimization

Abstract

Fused Filament Fabrication (FFF) is an advanced manufacturing process that requires precise control of multiple parameters, including nozzle temperature, print speed, and layer height. Due to the complexity of this high-dimensional process design space, experimental evaluations are often constrained. A key challenge in FFF is understanding how these parameters influence print quality and identifying optimal process conditions efficiently. This study addresses this challenge by developing a physics-based thermal model for FFF, implemented using a graphics processing unit-accelerated finite element method. The model is calibrated and validated against experimental thermal data for printing polylactic acid (PLA). It is then used to investigate the effects of nozzle temperature, print speed, bed temperature, and layer thickness on print quality by developing a cooling rate metric. A series of simulations is conducted within the process window using the physics-based model, and the resulting data are analyzed with SHapley Additive exPlanations to understand the influence of process parameters on print quality. The results indicate that layer height is the most critical factor affecting the quality of tensile samples. To enhance process optimization, a surrogate model is trained and optimized using data generated from the physics-based model, enabling the identification of an optimal processing window for PLA. By combining physics-based and data-driven modeling, this approach accelerates thermal prediction in the FFF process, facilitating the study of high-dimensional design spaces and the optimization of material-specific printing parameters. The proposed methodology provides a scalable framework for improving the efficiency and quality of extrusion-based additive manufacturing processes, demonstrating its potential for broader applications in process optimization.