Enhancement and Machine Learning-Based Prediction of Tribological Properties of PC/PBT/GNPs Nanocomposites

Abstract

Ternary polycarbonate-poly(butylene terephthalate)/graphene nanoplatelets (PC-PBT/GNP) nanocomposites were fabricated by melt-compounding. The nanofiller dispersion, microstructural changes, and mechanical and tribological properties of the produced samples were investigated. The friction and wear performance of the produced samples were evaluated with a pin-on-disc test rig under 5 and 10 N loads against an AISI 52100 steel ball to evaluate the effect of GNP filler fraction on the friction and wear performance of PC-PBT blends subject to polymer-metal contact in automotive and aviation industries. The impact strength, tensile modulus, and flexural modulus of the neat PC-PBT blend were improved by 78, 46, and 38%, respectively, with the optimum nanofiller fraction of 5 wt %. In parallel to the improved mechanical properties, similar to 86 and similar to 90% reduction in specific wear rates were achieved under 5 and 10 N loads, respectively, compared to the neat sample, which is attributable to multiple factors such as increased stiffness contact surface, intrinsic lubricating characteristics of GNPs, a more tribo-layer-oriented wear regime at higher filler fractions, and increased crystallinity via the reduced extent of transesterification. The Least-Squares Boosting (LSBoost) machine learning model provided the highest prediction accuracy with R 2 = 0.9922 via incorporation of contact pressure calculation results into the model as dependent variables.