Prediction of Tribological Properties of PC-PBT/GNP-MWCNT Nanocomposites Using Machine Learning Models

Abstract

This study investigates the effect of the incorporation of multi-walled carbon nanotubes (MWCNTs) and graphene nanoplatelets (GNPs) in polycarbonate-poly(butylene terephthalate) (PC-PBT) blends on the mechanical and tribological blend properties. PC-PBT/GNP-MWCNT nanocomposites were synthesized via melt-compounding with various filler loadings (0.5, 1, 3, 5, and 7 wt.%). SEM analyses revealed adequate dispersion and strong interaction between the nano-fillers and the polymer matrix. Mechanical testing demonstrated up to ~16%, ~38%, and ~9% improvement in tensile modulus, bending modulus, and impact strength, respectively, with the optimum nano-filler fraction of 0.5 wt. %. Tribological assessments, conducted using a pin-on-disc apparatus, showed marked reductions in specific wear rates (SWRs) reaching ~87% at the optimal filler loading of 0.5 wt.%. The mechanical behavior of the nanocomposites was found to depend primarily on dispersion state, whereas tribological properties were found to be dictated by a transfer film formation mechanism facilitated by filler addition. The experimental results were corroborated by a Random Forest machine learning model yielding the highest accuracy with R² = 0.94 for tensile modulus estimations and R² = 0.82 for SWR estimations under a 10 N load.