Experimental and machine learning studies on flexural properties of 3D-printed PLA/wood composites

Abstract

This study investigated the impact of critical input parameters in fused filament fabrication (FFF) on the flexural strength of PLA/wood bio composite specimens. This study considers the layer height (LH), printing temperature (PT), and printing speed (PS) as input parameters at five levels. Taguchi design of experiments (L25 orthogonal array) was employed to minimize experimental runs, followed by analysis of variance (ANOVA) to identify significant factors. The results demonstrated that layer height (78.91% contribution) is the most critical parameter, with 0.1 mm identified as the optimal amount for achieving the maximum flexural response, while printing temperature (9.39%) had a moderate effect and printing speed (2.69%) showed negligible influence. Machine learning methodologies such as Support Vector Machines (SVM), Decision Trees (DT), Random Forests (RF), Gradient Boosting Regression (GBR), and hybrid methods including SVM + DT, RF + DT, and GBR + DT were utilized to predict the flexural strength of the composite filament. In contrast to previous studies relying only on single models, this work shows that utilizing hybrid frameworks (specifically GBR + DT) achieves even better generalization performance, with a test accuracy of 94.30% and considerably lower prediction error (MSE = 10.8109 and RMSE = 3.288) than other models. The usage of Taguchi-ANOVA experimental design with advanced hybrid ML methods achieves a new contribution to using the approaches in additive manufacturing materials property modeling for suggesting comprehensive experimental investigation while improving predictive accuracy and robustness.