Multi-objective Modeling of Additively Manufactured Bio-Composite Based on Machine Learning Regression

Combining eco-friendly materials like wood fibers and biodegradable polymers, bio-composites offer exciting potential for sustainable innovation. However, printing these bio-composites using Fused Filament Fabrication (FFF), presents challenges, and optimizing the printing process is crucial for high-quality, strong, and efficiently produced objects. This study explores the use of Support Vector Machine (SVM) regression for modeling and optimizing the FFF process with Wood/PLA composites as filament. By employing a full factorial plan and constructing 64 test cubes, the study unravels the complex relationships between printing parameters and critical outcomes like relative density and productivity. To capture non-linear dependencies, a cubic kernel function was used within the SVM model. Cross-validation with a value of 5 ensured generalizability and prevented overfitting to specific data subsets. The outcomes underscore the robust predictive capability of the model, substantiated by elevated R-squared values and precise predictions (indicated by low metric values) observed across both validation and test datasets. Furthermore, the Newton-Raphson method identified the optimal printing parameters that maximize both relative density and productivity (calculated as a portion of volume built in a second). The optimal setting was found with a nozzle temperature of 205 °C, a printing speed of approximately 87 mm/s, and a layer height of approximately 1.5 mm.