Explainable AI for Structure–Property Analysis of FDM-Printed Biodegradable Polymer Nanocomposites

Abstract

Biodegradable polymer nanocomposites have gained significant attention for sustainable engineering applications, particularly when processed through fused deposition modeling (FDM) to create complex, customizable structures. Despite their potential, understanding how processing conditions and nanoscale reinforcements collectively influence the final properties remains a persistent challenge. The primary difficulty arises from the nonlinear, multivariate nature of structure–property interactions in FDM-printed biodegradable systems, which conventional modeling approaches often fail to capture or interpret. This study aims to develop an explainable artificial intelligence (XAI) framework capable of predicting and interpreting the mechanical and thermal behavior of biodegradable polymer nanocomposites fabricated via FDM. Biodegradable polymer matrices reinforced with 0–5 wt% nanoscale fillers were printed under controlled variations of nozzle temperature, layer height, infill density, and raster orientation. Machine learning models—including random forest and gradient boosting regressors—were trained on experimentally obtained structural, morphological, and thermal descriptors, while SHAP-based explainability tools were used to identify dominant contributors to property variation. The proposed framework achieved high predictive accuracy for tensile strength (R² = 0.93, RMSE = 3.1 MPa) and elastic modulus (R² = 0.91, RMSE = 45 MPa), and reliably predicted thermal stability (R² = 0.89 for T₅%). Explainability analysis revealed that infill density, nanofiller dispersion quality, and crystallinity index contributed up to 78% of the variance in mechanical response, whereas extrusion temperature and filler–matrix interfacial compatibility dominated thermal behavior. These findings provide mechanistic insights into the structure–property relationships governing FDM-printed biodegradable nanocomposites and demonstrate the potential of XAI to guide systematic material design and process optimization.