What question did this study set out to answer?

The aim is to optimize the mechanical properties of bio-based composites fabricated by fused deposition modeling.

June 3, 2026Journal of Applied Polymer Science0 citations

Multi‐Parameter Optimization, Machine Learning‐Based Prediction, and Surface Characterization of Bio‐Based FDM Composites

Key Points

The aim is to optimize the mechanical properties of bio-based composites fabricated by fused deposition modeling.
Utilized a Taguchi L27 orthogonal array for parameter optimization.
Performed tensile (ASTM D638) and compression (ASTM D695) tests to evaluate mechanical properties.
Developed predictive models using decision tree regression and Gaussian process regression for data analysis.
PLA samples exhibited the highest tensile strength of 48.66 MPa and compressive strength of 45.69 MPa.
Material type was identified as the most significant factor affecting tensile strength, while infill density had the strongest effect on compressive strength.
Gaussian process regression model achieved R² values above 0.995 and MAPE values below 14%, demonstrating high predictive accuracy.

Abstract

ABSTRACT In this study, mechanical and morphological behavior of the bio‐based composites produced by fused deposition modeling are experimentally and data driven investigated. Three distinct material configurations were fabricated and examined with the help of a Taguchi L27 orthogonal array which included pure PLA, PLA/Wood composite and multilayer sandwich structures. Five important process parameters were optimized viz. material type (MT), infill density (ID), printing speed, infill pattern and nozzle temperature. Tensile and compression tests were performed according to ASTM D638 and ASTM D695 standards. The PLA samples showed the highest tensile strength (48.66 MPa). The compressive strength was also highest for the PLA samples (45.69 MPa). ANOVA results showed that MT was the most influential factor on tensile strength and ID was the strongest effect on compressive strength. Scanning electron microscopy observations showed differences in interlayer adhesion, porosity, and fracture behavior of the material systems. Furthermore, the Taguchi based prediction models, decision tree regression and Gaussian process regression (GPR) were also developed. Among these models, GPR showed the best predictive performance with R 2 values above 0.995 and MAPE values below 14%, exhibiting good agreement with the experimental results.

Ask AI

Helpful

Bookmark

View Full Paper