Fused deposition modeling component quality enhancement by experimental investigation and ANN prediction

Abstract Fused Deposition Modeling (FDM) is a widely adopted additive manufacturing process in which surface roughness (SR) critically influences the performance and appearance of printed components. This study presents an Artificial Neural Network (ANN) model to predict and optimize SR based on three key process parameters: extrusion temperature, layer height, and printing speed. A full factorial design of 64 experiments was conducted using PLA material, and the resulting data were used to train a 3-10-1 ANN architecture in MATLAB R2021a. The model was trained using the Levenberg–Marquardt algorithm with a 70:15:15 split for training, validation, and testing. The ANN achieved a high correlation coefficient (R² >0.99) and a maximum prediction error of 1.23%, confirming its robustness and accuracy. Analysis showed that layer height had the greatest impact on SR due to the staircase effect, while extrusion temperature and printing speed had secondary effects. Optimal surface roughness was achieved at an extrusion temperature of 210 °C, layer height of 0.1 mm, and printing speed of 40 mm/s, consistently improving surface quality across horizontal, vertical, and inclined orientations. These settings minimize surface roughness by balancing thermal adhesion, layer resolution, and deposition accuracy across varied geometries. The developed ANN model offers a reliable, data-driven tool for process optimization in FDM, enabling improved print quality and reduced trial-and-error in industrial applications. Graphical abstract