Machine-Learning-Assisted Identification and Formulation of High-Pressure Lubricant-Piezoviscous-Response Parameters for Minimum Film Thickness Determination in Elastohydrodynamic Circular Contacts

Abstract

From the earliest theoretical studies on elastohydrodynamic lubrication, it was believed that film build-up is governed by lubricant rheology in the low-pressure contact inlet. Recently, it was discovered that this is only true for the theoretical line contact case, where lubricant out-of-contact lateral flow is absent. In actual contacts, though central film thickness is indeed governed by low-pressure lubricant rheology, minimum film thickness is additionally influenced by the high-pressure response. Thus, a proper prediction of minimum film thickness (either by analytical formulae, or machine-learning frameworks) would require input parameters that define the high-pressure viscous response of the lubricant. The current work identifies and formulates these parameters with the help of machine-learning regression tools. These are fed with minimum film thickness results from finite element simulations of smooth steady-state isothermal Newtonian circular contacts, lubricated with sets of hypothetical fluids having the same pressure-viscosity response at low pressure, but different high-pressure ones. It is found that conventional dimensionless groups are not sufficient to describe minimum film thickness formation, and that an additional pressure-viscosity coefficient—evaluated at half the Hertzian contact pressure—is required.