Taguchi-ANN Hybrid Approach for Evaluating Unsupported Overhang Structures in FDM-Printed Polymers

Abstract

Fused deposition modeling (FDM) faces significant challenges in producing unsupported overhangs, particularly regarding dimensional accuracy and surface quality. While support structures can address these issues, they increase material waste and post-processing requirements. This study investigates how process parameters affect the quality of unsupported inclined surfaces in FDM printing, aiming to optimize fabrication without support structures. Test specimens with three surface inclination angles-75 degrees (Surface-1), 60 degrees (Surface-2), and 45 degrees (Surface-3)-were produced using acrylonitrile butadiene styrene (ABS), polylactic acid (PLA), and polyethylene terephthalate glycol (PETG). The study employed an L27 Taguchi orthogonal array to evaluate five key process parameters: material type (MT), infill pattern (IP), wall thickness (WT), infill density (ID), and layer thickness (LT). Dimensional deviations and surface roughness were measured across all angled regions. Dimensional deviations remained below 2.3% across all specimens. ANOVA revealed LT and MT as the most significant factors affecting both dimensional accuracy and surface roughness (p < 0.001), with LT showing the highest contribution to surface roughness variance (F values > 500). Surface roughness improved with decreasing surface angles, while dimensional accuracy showed an inverse trend. Artificial neural network (ANN) models developed for predicting quality metrics achieved R-2 values exceeding 0.90. This study establishes a comprehensive framework integrating Taguchi design, statistical analysis, and machine learning (ML) for optimizing unsupported overhang fabrication in FDM. The findings reveal crucial relationships between process parameters and part quality, demonstrating that carefully controlled parameter selection can achieve acceptable quality without support structures. The developed predictive models offer a reliable tool for parameter optimization in FDM manufacturing of unsupported overhangs.