Comparative analysis of machine learning techniques for predicting wear and friction properties of MWCNT reinforced PMMA nanocomposites

Abstract

Polymer nanocomposites have received significant scientific and industrial attention due to the synergetic combination of features of a polymeric matrix and organic or inorganic nanofillers. While experiments have been essential for identifying and characterizing new materials, their high costs and limited trials have shifted the focus towards applying machine learning (ML) to predict nanocomposite properties. This study aims to establish a connection with the tribological performance of multi-walled carbon nanotubes (MWCNT) reinforced with polymethyl methacrylate (PMMA) nanocomposites through the comparison of ML techniques. The wear and friction characteristics of MWCNT-reinforced PMMA nanocomposites were predicted based on three input variables: material weight percentage, load weight, and track diameter. The features of nanocomposites were predicted using three different ensemble ML algorithms: random forest (RF), extra tree (ET), and gradient boosting machine (GBM). The dataset was utilized to train the proposed models in Python, followed by hyperparameter tuning to determine the best model for predicting target values. The results demonstrated that the GBM model outperformed the RF and ET models, with an R-squared of 0.99, RMSE of 0.62, and MAE of 0.18. The proposed models’ predictions of wear values were more precise than their friction values. These findings indicate that ML techniques, particularly the GBM model, can effectively predict the tribological properties of MWCNT-reinforced PMMA nanocomposites, potentially reducing the need for extensive experimental trials and contributing to advancements in nanocomposite material science.

Keywords

• Carbon nanotubes
• Gradient boosting machine
• Machine learning
• Nanocomposites
• Predictive models