Machine learning algorithms to optimize the properties of bio-based poly(butylene succinate-co- butylene adipate) nanocomposites withcarbon nanotubes

A simple, cost-effective and scalable solution to improve the mechanical properties of Poly[(butylene succinate)-co-adipate], PBSA, is reported by using functionalized single-walled carbon nanotubes (SWCNTs). Different SWCNT loadings have been incorporated in the PBSA matrix via simple solution casting, and the ultrasonication conditions have been optimized to attain a homogenous SWCNT dispersion. The nanocomposites have been characterized in detail by scanning electron microscopy (SEM), Infrared spectroscopy, thermogravimetric analysis (TGA), differential scanning calorimetry (DSC), tensile and impact strength tests. Unprecedented increments in stiffness were found at low nanotube loadings. Further, four machine learning (ML) algorithms, Polynomial Regression (PR), Support Vector Machines for Regression (SVR), Gradient Boosting (GB) and Artificial Neural Network (ANN), were applied to predict their mechanical properties. The algorithm´s performance was assessed using analytical parameters such as the coefficient of determination (R²), the mean square error (MSE) and the mean absolute error (MAE). The developed models exhibited strong performance, achieving R² values ranging from 0.69 to 0.99 across the evaluated properties. The results corroborate that even when the same prediction model is used, its performance varies depending on the physical property to be predicted. Thus, SVR, GB, PR, and ANN were found to be the most effective for estimating the Young’s modulus, tensile strength, elongation at break and impact strength, respectively. This research holds great potential for advancing the field of modelling the mechanical properties of polymeric nanocomposites and their practical applications in various industries such as food, pharmaceutical and biomedicine.