Statistical and machine-learning models to predict the flexural properties of wood-based composites fabricated via material extrusion technique
Vishal Mishra, Nikhil Bharat, Dhinakaran Veeman, Sushant Negi, Vijay Kumar
Abstract
Sustainable 3D printing using material extrusion (MEX) is a transformative technique for fabricating high-performance, eco-friendly products. This study investigated the flexural properties of polylactic acid (PLA) and wood dust composites fabricated via fused filament fabrication (FFF). A Taguchi method was adopted for parameter optimisation, focusing on factors like printing temperature, layer height, print speed, raster angle, and infill density. Experimental and statistical analyses, including signal-to-noise (S/N) ratio evaluation and analysis of variance (ANOVA), were performed to identify the optimal printing conditions. Furthermore, machine learning models, specifically the Levenberg-Marquardt neural network (LM-NN) and scaled conjugate gradient neural network (SCG-NN), were employed to predict the flexural properties. Results revealed that raster angle and print speed significantly influence flexural modulus and load, with a 90° raster angle yielding the highest modulus of 3121 MPa. The combination of experimental validation, statistical optimisation, and machine learning prediction demonstrates a robust framework for enhancing the mechanical performance of 3D-printed PLA/wood dust composites. This study paves the way for the development of sustainable composite materials suitable for advanced engineering applications.