Machine learning-based optimization of mechanical and morphological performance of polylactic acid nanocomposites with lignin nanoparticles

Abstract

This study investigates machine learning-based optimization of the mechanical properties of environmentally friendly biopolymer nanocomposites produced by incorporating lignin nanoparticles (NLPs) and maleic anhydride (MA) into a polylactic acid (PLA) matrix. Lignin was extracted from black pine using a deep eutectic solvent method and melt-compounded with PLA via twin-screw extrusion, followed by injection molding. Mechanical performance was evaluated using tensile and three-point bending tests, while fracture morphology was examined by scanning electron microscopy (SEM). Random Forest (RF) and Extreme Gradient Boosting (XGBoost) models were applied to predict and optimize mechanical properties using lignin and MA contents as input variables, with an 80/20 training–testing data split. Experimental results showed that neat PLA exhibited the highest tensile strength (61 MPa) and modulus (5.2 GPa). The addition of low lignin contents with MA slightly reduced tensile properties but significantly enhanced flexural strength (≈58 MPa) and modulus (≈3.9 GPa). SEM observations revealed uniform nanoparticle dispersion and crack-free fracture surfaces at low lignin loadings, whereas higher lignin contents resulted in agglomeration and brittle behavior. Both machine learning models demonstrated high predictive accuracy, with RF outperforming XGBoost. The results confirm that MA improves interfacial adhesion and that data-driven approaches effectively support optimization of biopolymer nanocomposite compositions.