GNBoost-Based Ensemble Machine Learning for Predicting Tribological Properties of Liquid-Crystal Lubricants

Area of Science:

• Materials Science
• Tribology
• Machine Learning

Background:

• Accurate prediction of liquid-crystal lubricant tribological performance is crucial for their development.
• Understanding the influence of various parameters on lubricant behavior is essential for optimizing formulations.

Purpose of the Study:

• To propose a classification model for predicting the tribological performance of liquid-crystal lubricants.
• To evaluate the effectiveness of polysorbate-85, polysorbate-80, and graphene oxide as additives in liquid-crystal lubricants.
• To optimize lubricant formulation and performance evaluation through predictive modeling.

Main Methods:

• Fabrication of liquid-crystal lubricants using three additives: polysorbate-85, polysorbate-80, and graphene oxide.
• Testing lubricant coefficients of friction in rotational mode using a universal mechanical tester.
• Development of a data augmentation model to predict the coefficient of friction, with parameters optimized via particle swarm optimization.

Main Results:

• A Gaussian noise extreme gradient boosting (GNBoost) model was successfully developed to predict tribological performance.
• The model demonstrated effectiveness in predicting the coefficient of friction for the tested liquid-crystal lubricants.
• Particle swarm optimization enhanced the predictive accuracy of the developed model.

Conclusions:

• The proposed GNBoost model offers an effective approach for evaluating lubricant performance.
• This study provides a valuable framework for optimizing liquid-crystal lubricant formulations.
• Predictive modeling significantly aids in the timely and accurate assessment of lubricant behavior.