Comparison of machine learning models for the potential prediction of tensile strength of Poly(lactic acid)

This thesis presents an approach to analyze the 3D printing processes through the application of machine learning (ML) models. We aim to analyse models that predict tensile strength which is a mechanical property of materials, by analyzing the relationship between printing parameters (bed temperature, layer thickness, extrusion temperature, infill density, printing speed) and tensile strength for the printed parts. The primary focus is on determining tensile strength, a critical factor in the functionality and durability of 3D printed objects. Using three different datasets compiled from both published academic research and open-sourced experimental data, we employ four ML regression models -multilinear, support vector, decision tree, and XGBoost- to identify effective predictors of material properties based on printing settings. The thesis's findings highlight the potential for machine learning to significantly increase the precision and efficiency of additive manufacturing. By providing parameters capable of predicting the properties of prints, this approach can save time, reduce material waste, and enable more customized applications of 3D printing in various industries. We also discuss the development of a user-friendly interface to make this tool available to a broader audience, from professionals in the biomedical and aerospace industries to hobbyists and educators. The potential for commercialization through an initial model will also be explored, providing a way to integrate these predictive capabilities directly into 3D printing hardware and software ecosystems. This work advances additive manufacturing and opens new paths for applying machine learning in materials science.