Experimental investigation and data-driven prediction of gearbox churning loss in high-rotational-speed flow regimes

Churning loss in gearboxes, arising from hydrodynamic resistance due to interactions between rotating gears and lubricating fluid, significantly reduces power transmission efficiency and becomes increasingly difficult to predict under high rotational speed flow regimes characterized by complex viscous, inertial, and turbulent effects. As conventional empirical models exhibit clear limitations in capturing such nonlinear behavior, this study conducted churning loss experiments on 54 spur and helical gears over a wide range of rotational speeds and lubrication conditions and developed a random forest regression model using experimental measurements. The model was trained with key variables including gear geometry, oil immersion level, and rotational speed, and its performance was evaluated against the ISO 14179–1 and Changenet models. The proposed model shows excellent agreement with experimental results across the entire operating range, achieving a coefficient of determination of 0.994, while maintaining consistent accuracy in high rotational speed and turbulent flow conditions where empirical models exhibit rapidly increasing prediction errors.