AI-Driven Tribological Prediction of Ultrasound-exfoliated F-MWCNTs, MoS₂, and Graphite Self-lubricating Epoxy Composite

Abstract

This thesis presents a comprehensive study on the design, characterization, and predictive modeling of self-lubricating epoxy composites reinforced with exfoliated graphite, ultrasonicated MoS₂, and selectively functionalized multi-walled carbon nanotubes (MWCNTs). It integrates experimental investigations with advanced artificial intelligence (AI) methodologies to enhance the mechanical and tribological performance of polymer composites for advanced industrial applications.

Experimentally, low filler loadings (0.1–0.5 wt%) of graphite and MoS₂ significantly improved composite performance without compromising structural integrity. The 0.3 wt% graphite composite achieved a 91% reduction in specific wear rate and a 44% decrease in friction coefficient, while MoS₂ composites demonstrated up to 86% wear reduction and 44% COF improvement. Among functionalized MWCNTs, COOH-MWCNTs exhibited superior self-lubrication through friction-induced graphitization, a novel mechanistic insight, whereas NH₂- and silane-modified MWCNTs enhanced tensile and ductile behavior with varied wear responses. Structural and tribological mechanisms were confirmed via SEM, TEM, XRD, and FTIR analyses.

A key innovation of this research lies in the integration of AI with traditional tribology. A hybrid modeling framework combining Kragelsky’s friction law and Archard’s wear law with a feedforward Artificial Neural Network (ANN) captured complex nonlinear relationships between experimental inputs such as contact pressure, sliding velocity, material hardness, and filler composition and tribological outputs, including wear rate and COF. The ANN demonstrated high predictive accuracy for graphite composites (R² > 0.98) while highlighting environmental effects, such as oxidative degradation in MoS₂, on tribological performance. This approach enables efficient 3D surface-based optimization, reducing experimental workload.

The study further expands computational capabilities through a multimodal AI architecture incorporating ANN, Recurrent Neural Networks (RNN), and Convolutional Neural Networks (CNN). RNNs effectively modeled time-dependent friction evolution across repeated sliding cycles, capturing mechanisms such as abrasion, adhesion, fatigue, and delamination. CNNs successfully classified wear mechanisms from SEM micrographs, achieving high accuracy, providing a powerful tool for surface diagnostics.

This work presents an interdisciplinary investigation integrating materials science, tribology, polymer chemistry, and artificial intelligence, combining experimental characterization with advanced predictive modeling. The study provides new insights into wear mechanisms in nano-reinforced epoxy composites and establishes a comprehensive framework for the predictive design, optimization, and monitoring of self-lubricating polymer systems.