CFD optimises gearbox oil flow paths to reduce drag
Using computational fluid dynamics (CFD) to optimise the oil flow path in the gearbox for reduced drag treats lubricant as an active part of the power transmission, not a passive supply. Oil can be flung from gear faces, churned in a sump, compressed through bearing passages or carried away as spray. The resulting shear and pumping losses consume shaft power, while air entrainment can compromise the very films meant to protect bearings and tooth contacts.
The engineering task is therefore a balance, not a simple race to remove oil. The gearbox must retain enough lubricant for cooling and durability, deliver it where contact occurs, and keep the fluid moving without needless circulation. The OPTIMIZE Project brings design-of-experiments methods, multiphase simulation, tolerance analysis and physical testing together to meet that balance in geared aircraft engines.
Why uncontrolled oil flow wastes power
At high operating speed, centrifugal force pushes oil towards the outer radius of the casing. Gears pick up lubricant at their lower edges and release it as droplets or sheets, creating churning and hydrodynamic drag as the droplets are sheared and accelerated. Oil recirculating between the sump and upper galleries also has to overcome pressure restrictions, while narrow bearing clearances generate viscous shear. None of these losses is obvious from a simple fill-level calculation.
The consequence reaches beyond gearbox temperature. Every unit of torque lost to oil drag is unavailable for thrust, electrical generation or useful work elsewhere in the engine. This makes even a modest reduction valuable to overall engine efficiency and power density. Lower drag can also reduce the oil agitation and heat that must be removed, potentially allowing a smaller sump, cooler or pump. Yet reducing flow too far would starve bearings and the gear mesh, so the useful target is the lowest drag that still provides a robust film, adequate cooling and safe pressure at every operating condition.
CFD helps expose where that balance is being missed. A resolved flow model can show whether oil is short-circuiting from the pump to the tank, whether a gear is pumping a cavity full of aerated fluid, or whether a return passage is creating a high-speed jet. It also turns drag from an empirical allowance into measurable quantities such as shear torque, pressure loss, oil-air ratio, bearing-feed rate and temperature rise. That evidence lets engineers compare competing housing and lubrication changes before committing to expensive tooling.
Building a model that earns trust
A useful model begins with the complete lubrication circuit. The computational domain should include the sump, pump or pressure source, filter, inlet and return galleries, restrictors, bearings, gear spaces, seals and tank boundary. Flow paths that are hidden in cast housings need plausible geometric representation, while the rotating boundaries must carry the correct speed and direction. Otherwise, the model may look tidy but fail to reproduce the pressure pulses and trapped volumes that create real drag.
Oil and air must be treated according to how they occur in the machine. A volume-of-fluid approach can capture the moving free surface in a sump; an Eulerian or phase-field model may suit dispersed mist. Gear rotation may be represented with moving mesh, overset cells or a rotating reference frame, depending on whether tooth-passage detail is needed. Temperature-dependent viscosity, non-Newtonian behaviour at low temperature, cavitation near sharp restrictions and conjugate heat transfer can all change the result. The physics should match the decision, rather than adding every available model by default.
Mesh quality is especially important around gear-entry regions, bearing races, sharp corners and oil-entry holes. Engineers should perform mesh refinement and time-step sensitivity checks, then confirm that the predicted drag torque and flow split are stable. Reporting only velocity contours is insufficient: useful outputs include integrated surface shear, bearing flow allocation, pressure drop across each restriction, trapped-air fraction, film thickness and bulk oil temperature. Independent energy checks can reveal whether the model is creating or removing more momentum than the imposed pump and shaft conditions allow.
Contact conditions also need to be realistic. Tooth deflection and mesh-stiffness variation alter local loading, film thickness and lubricant entrainment. The project’s method for gear mesh stiffness variation provides a useful bridge between tooth-contact analysis and the flow model. This matters because an optimistic load case can make a high-flow design appear adequate while a tougher production tolerance exposes starvation. The same principle applies to pump capacity, surface roughness and bearing clearances: uncertainty belongs in the model, not in a final paragraph of caveats.
Converting flow behaviour into better hardware
A design-of-experiments programme can turn many possible changes into a manageable search. Variables might include gallery diameter, restrictor size, feed-hole position, sump baffle height, vent location, seal geometry, return-port orientation and the clearances around rotating components. CFD can then screen which combinations materially affect drag and lubrication, after which a response model can identify interactions that a one-change-at-a-time approach would miss.
The objective should be multi-functional. Minimum drag is valuable only if bearing-film thickness, tooth-contact temperature and oil temperature remain inside limits. The optimisation can also seek secondary gains such as lower pressure-drop demand, better cooling uniformity and less aeration. This prevents a narrow result—often a very low pump flow—from winning merely because it produces the smallest headline drag figure. A practical scorecard exposes the trade-offs and keeps structural, thermal and manufacturability requirements in the same decision.
Passage shaping deserves as much attention as passage sizing. The project’s exploration of cascading reels and slots shows how deliberately shaped internal features can be assessed alongside conventional galleries and drilled passages. CFD can compare how features meter flow, create low-shear regions or move air away from bearing inlets without weakening the housing. The promising arrangement must still be checked against pressure loss, temperature, strength and cleanability, because a feature that performs well in a clean bench model may trap more oil or debris in service.
Tolerance analysis then converts the best nominal geometry into a robust design. Clearances, hole sizes, surface finishes and tooth forms will vary through manufacturing and wear, so several variants should be simulated around the target. In a hyperstatic bearing arrangement, load sharing and leakage paths can make a small change in one clearance alter flow at several other locations. The optimised gearbox should therefore work across the tolerance band, not collapse whenever a nominally correct part lands just outside it.
Proving the gain on a physical rig
Physical testing is where a promising flow model becomes credible engineering evidence. An instrumented gearbox rig can reproduce selected speeds, loads, oil temperatures and inlet pressures while measuring shaft torque, bearing temperatures, differential pressures and oil consumption. Flow indicators, targeted temperature sensors and calibrated flow probes can reveal whether the CFD predictions appear in hardware. Where visual access is useful, a transparent test section or a compatible fluid may aid observation, but its different properties must be considered when comparing results.
The test matrix should include the operating envelope and the worst credible tolerance combinations. Designers should separate changes caused by speed, load, oil temperature, pump variation and leakage, then compare each trend with the simulation. Any mismatch becomes a modelling task: inspect the boundary condition, update the oil properties, refine the mesh or improve the physical rig. A close final value matters less than agreement in behaviour across conditions, because the latter shows that engineers understand the mechanism.
That evidence has particular value for Australian programmes. Melbourne, Brisbane and Adelaide each host important parts of the country’s aerospace ecosystem, from commercial design and propulsion work to defence manufacturing and maintenance, but the centres are separated by thousands of kilometres. Test articles, specialist instrumentation and certified components often move long distances, and many aircraft engines or gearboxes are imported before local overhaul, where CASA and approved maintenance-data requirements shape the work. International collaboration can draw on the European aerospace research network, helping teams compare test methods, uncertainty and evidence standards across borders.
Turning efficiency into an Australian case
The business case should convert shaft-power recovery into operational value. Less oil drag can reduce fuel burn and emissions per flight hour, decrease cooling-system demand and preserve more of the engine’s available output. For Australian airlines, regional operators and defence fleets, those benefits must be weighed against certification, modification and maintenance costs. A modest saving repeated across an engine fleet can justify a change, while a bespoke low-volume gearbox must also demonstrate a credible route to manufacture and sustain.
A staged implementation keeps the risk controlled. First, establish a validated baseline and rank the loss mechanisms by size and uncertainty. Next, optimise the most influential geometry with DOE, then manufacture representative features and repeat the critical rig tests. Finally, feed manufacturing variation and service conditions back into the model before committing to production. In Australian design reviews, a “fair-dinkum” result is one that survives that loop: it predicts the rig, tolerates the parts bin and remains inspectable by local maintenance teams.
The immediate action is to use the OPTIMIZE Project’s videos, objectives and methodology as the basis for a focused CFD-to-rig case. Bring forward a representative gearbox, its duty cycle and known pain points; identify the oil paths that consume the most power; and test a small set of manufacturable changes against both drag and film requirements. That sequence gives Australian researchers, suppliers and operators a common path from simulation to verified hardware, with efficiency gains that can survive certification scrutiny and real operating conditions.