A Quantitative Structure Tribo-Ability Relationship Model for Ester Lubricant Base Oils

Abstract

Friction tests with point–point contact were carried out using a microtribometer to investigate the tribological characteristics of steel/steel rubbing pair immersed in 57 kinds of esters as lubricant base oils. A set of 57 esters and their wear data were included in the back-propagation neural network (BPNN)-quantitative structure tribo-ability relationship (QSTR) model with two-dimensional (2D) and three-dimensional (3D) QSTR descriptors. The predictive performance of the BPNN-QSTR model is acceptable. The findings of the BPNN-QSTR model show that the extent of polar groups cannot be too large in the molecule to achieve good antiwear performance; and the polar groups with a high degree of relative concentrated charge are favorable for antiwear. A low degree of molecular hydrophobicity of lubricant base oil is beneficial for antiwear behavior. Large molecular dipole moment is disadvantageous for antiwear properties. It is necessary to maintain one large molecular surface in one plane, to have a long and short chain length to be present within the same molecule, and to keep small difference between the long and short chain length to enhance the antiwear performance. Finally, lubricant base oil candidate molecules will have beneficial antiwear properties that they should contain more N groups with three single bonds and more C groups with one double bond and two single bonds; the presence of O atoms with any bonds or CH groups with three single bonds leads to a decrease in the wear resistance performance.