Numerical Challenges in Modeling Gearbox Lubrication

Numerically, modeling gearbox lubrication involves significant challenges. Lubricant splashing is a highly turbulent and transient process, necessitating a transient CFD analysis with a sufficiently long simulated time to achieve a developed flow. This typically requires four to eight seconds of physical time, depending on the geometry’s size and complexity, and demands substantial computational power. The new XFlow 2023 solver supports both NVIDIA and AMD GPU cards, allowing for significantly reduced runtimes, from weeks of computing time to days, making this process accessible. XFlow employs a particle-based Lattice Boltzmann solver, offering multiphase and moving parts modeling capabilities, irrespective of the system complexity, gear types and lubrication methods (Ref. 1). Heat transfer, being a slower process, is typically resolved through steady-state analyses. The Abaqus implicit solver is widely recognized in both industry and academia for its effectiveness. Lubrication involves heat extraction from solid parts through convection, a challenging aspect to quantify in transient analyses with rotating parts. The equation for convection, also known as the heat transfer convection equation, is typically expressed by Newton’s law of cooling. This law describes the heat transfer between a surface and a moving fluid.

The convective heat transfer coefficient (h) is evaluated during the transient CFD analysis at every time increment, on all surfaces. The values of h are highly uneven across the circumference of the gears due to their rotation and the turbulent nature of the splashing. To use these values in a steady-state heat transfer analysis, the h values at each finite element node must be averaged over their respective displacements. Rotating parts are represented by hundreds of thousands of nodes, making the data processing and transfer from a transient CFD analysis to a steady-state heat transfer analysis a major challenge. In the company’s project, we faced these numerical challenges head-on. The process required meticulous discretization of gear profiles and the modeling of roller and ball bearings, involved using advanced post-processing tools and visualization techniques, along with automating data processing. Utilizing advanced solvers and leveraging our GPU card power enabled us to overcome these challenges, providing reliable and actionable insights into the lubrication issues and potential solutions.