Prediction of Dimensional Accuracy and Surface Quality in Additively Manufactured Biomedical Implants Using ANN

Abstract

This study investigates the prediction of fused deposition modeling (FDM) process parameters for manufacturing biomedical implants with high dimensional accuracy and surface quality. Biomedical implants were fabricated in circular, triangular, and pentagonal geometries to accommodate different anatomical requirements, using three materials selected for their biomedical applicability and mechanical properties. These materials are polylactic acid (PLA), polyethylene terephthalate glycol, and thermoplastic polyurethane (TPU). This research utilizes the Taguchi L27 orthogonal array methodology to analyze the influence of five critical printing parameters: material type, layer thickness (200-300-400 mu m), infill density (30%-60%-90%), infill pattern (zigzag, cubic, and triangle), and wall thickness (1-2-3 mm). The analysis of variance demonstrated that material type and layer thickness are the most significant factors, contributing 49.25% and 17.97%, respectively, to dimensional accuracy in circular geometries. Surface roughness measurements showed that layer thickness (30.95%) and material type (31.28%) are dominant factors affecting surface quality. The optimum parameters for dimensional accuracy were determined as PLA material, zigzag infill pattern, 2 mm wall thickness, 30% infill density and 200 mu m layer thickness, while the highest surface quality was achieved with PLA material, triangle infill pattern, 3 mm wall thickness, 90% infill density and 200 mu m layer thickness. An artificial neural network model was developed to predict dimensional accuracy and surface quality, achieving high correlation coefficients (R2 > 0.96) between predicted and experimental results across all geometric configurations. These findings offer valuable guidelines for predicting and optimizing parameters in FDM-based biomedical implant manufacturing, advancing precision medicine by enhancing additive manufacturing processes and implant performance.