Gearbox lubrication CFD (Lattice Boltzmann): methods and results

Paper · Project 42

This paper documents a CFD study of oil–air multiphase flow in a gearbox, using a GPU-accelerated Lattice Boltzmann Method solver to analyse lubricant distribution and torque losses under realistic operating conditions.

Method

The model captures the fluid interaction between two rotating gears at a transmission ratio of 1.54:1. The working fluids are air (ρ = 1.225 kg/m³, µ = 1.789×10⁻⁵ Pa·s) and oil (ρ = 860 kg/m³, µ = 0.01 Pa·s), with the interface tracked to measure oil impregnation and wetted surface area while recording instantaneous torque on each gear.

A mesh sensitivity study was conducted comparing a coarse lattice of 40,500 elements against a fine lattice of 324,000 elements, with element sizes of 0.004 m and 0.002 m respectively.

Results

Wetted surface area converged to 32.5% for the driving gear and 56% for the driven gear, with quasi-periodic torque losses observed on both. The driven (larger) gear carries substantially more oil, as expected from its greater submerged arc.

Discussion

The two mesh resolutions produced significant discrepancies, which is the central methodological finding: the study is not mesh-converged. Reaching convergence requires further lattice refinement together with reduced time steps, and the computational cost gap between the two cases (roughly 2 minutes against 20 minutes) indicates what that would demand.

← See the project