FDM 3D printing surface quality optimisation: multi-objective parameter control based on hybrid machine learning and response surface methodology

Fused Deposition Modelling (FDM) offers notable advantages in material utilisation and manufacturing flexibility, but surface quality issues such as warping, stringing, and roughness limit its application in high-precision fields. This study developed and validated an integrated framework for the high-accuracy prediction of multi-region surface morphology and the intelligent coordinated optimisation of process parameters for Polylactic Acid (PLA) specimens fabricated by FDM, thereby significantly enhancing surface quality control. Among them, we employed a Design of Experiments (DOE) approach to systematically investigate key process parameters. Using non-contact measurement, we acquired 3D point cloud data to calculate areal roughness parameters (Sa, Sq). Machine learning algorithms, including random forest (95.83% accuracy), evaluated parameter importance, while response surface methodology modeled interactions. A multi-objective optimization framework integrating desirability function and entropy weight method identified the optimal parameter combination: extrusion temperature 180.0°C, infill density 52.50%, extrusion rate 95.00%, Z-axis height -1.565 mm, and heated bed temperature 62.5°C. This integrated methodology combining data-driven modeling with multi-objective optimization provides both theoretical and practical advancements for surface quality improvement in high-precision FDM applications.