Using DOE To Improve Bearing Raceway Performance In Aircraft Gearboxes
In a geared aircraft engine, a bearing raceway is a small geometric feature with a large influence on system performance. Its curvature controls how rolling elements contact the inner and outer rings, how loads spread through the contact patch, and how much sliding occurs near the edges. Those effects influence fatigue life, friction, heat generation and, ultimately, gearbox reliability.
Design of experiments (DOE) provides a practical way to study these relationships without testing every possible combination of radius, load, speed, lubricant and manufacturing tolerance. Instead of changing one factor at a time, engineers can identify the interactions that matter and build a response model from a carefully selected set of simulations or physical tests.
This approach is especially valuable for geared propulsion systems, where high shaft speeds, compact packaging and strict power-density targets leave little room for trial and error. A raceway profile that appears efficient in a nominal model may behave differently when ring misalignment, housing distortion, surface finish or thermal growth is included.
For an Australian aerospace programme, the method also supports disciplined design decisions across geographically distributed teams. A supplier in Melbourne, a test facility near Brisbane and an operator supporting long routes from Perth may all need confidence that a bearing design remains robust under different operating and maintenance conditions.
Bearing Raceway Curvature In Geared Propulsion
Raceway curvature describes the cross-sectional shape of the path followed by a rolling element. It is often expressed through an osculation ratio, which compares the raceway radius with the rolling element radius. A tighter or looser curvature changes the size and shape of the contact ellipse, the pressure distribution and the amount of conformity between the two surfaces.
High conformity can distribute load over a broader region, potentially reducing peak Hertzian stress. However, excessive conformity may increase sliding, lubricant shear and sensitivity to local geometry errors. Lower conformity can reduce some sliding effects but may produce higher contact stress and edge loading if the bearing is misaligned or deflects under load.
In an aircraft gearbox, these trade-offs are coupled with speed and preload. A bearing may experience radial and axial forces from helical gears, transient torque during power changes and structural deflection from the casing. The preferred curvature is therefore a system-level decision rather than a value selected from a catalogue in isolation.
A useful analysis begins by defining the actual duty cycle. Cruise, climb, take-off, descent and transient conditions should be represented with realistic load and speed combinations. For Australian applications, long sectors between capital cities can create extended periods of steady operation, while regional aircraft may experience more frequent power changes and landing cycles.
Why Curvature Changes Contact Behaviour
The primary fatigue mechanism in a rolling bearing is usually subsurface stress accumulation beneath the raceway. Repeated stress cycles can initiate a crack, leading to pitting, spalling or progressive loss of material. Raceway curvature affects the maximum contact pressure and the depth and orientation of the stressed volume, so it has a direct connection with rolling contact fatigue life.
Curvature also affects the contact ellipse. A broad ellipse can lower peak pressure but may increase the total area exposed to sliding and lubricant traction. A narrow ellipse may concentrate load and make the design more sensitive to roughness, contamination and local waviness. These changes can alter the relationship between calculated fatigue life and measured friction torque.
Friction is not caused by a single mechanism. Rolling resistance, sliding at the contact edges, cage interaction, seal drag and lubricant churning may all contribute. When testing a curvature variation, engineers should separate bearing-level friction from the losses of the complete gearbox. Otherwise, a promising raceway design may be incorrectly rejected because of unrelated gear mesh or seal behaviour.
Misalignment deserves particular attention. Ring tilt changes the load distribution between rolling elements and can move the maximum stress toward a raceway shoulder. Housing and mount stiffness can influence this response, which is why the project’s work on housing stiffness effects is relevant to a curvature study.
Building A Useful DOE
A DOE for raceway curvature should include controllable design variables, operating conditions and nuisance factors. Curvature or osculation may be the central factor, but it should be assessed alongside radial load, axial load, rotational speed, lubricant viscosity, preload, surface roughness and alignment error. Manufacturing tolerances can be introduced as noise factors or as explicit variables in a robust design experiment.
The experiment may use a response surface design, a definitive screening design or a fractional factorial plan, depending on the number of factors and the maturity of the design. Early screening identifies influential variables; a second-stage response surface model can then locate a practical region that balances fatigue life, friction, temperature and weight.
Inputs That Strengthen The Analysis
- Raceway curvature or osculation ratio, including manufacturing limits
- Radial and axial load combinations across the engine duty cycle
- Shaft speed, lubricant temperature, viscosity and supply rate
- Misalignment, ring tilt, waviness, roughness and preload
- Measured or simulated contact stress, friction torque and temperature
A computational DOE can use bearing contact codes, multibody dynamics, finite element models and thermal networks. The model should retain enough physical detail to capture contact pressure and sliding while remaining efficient enough for dozens or hundreds of design points. Surrogate models can then estimate performance between sampled cases.
Questions To Resolve Before Testing
- Which response defines an acceptable design: fatigue life, power loss or both?
- Are friction measurements repeatable enough to distinguish curvature effects?
- Which factors represent normal variation, and which represent worst-case conditions?
- How will lubricant degradation and temperature changes be reproduced?
- Can the selected test conditions be linked to aircraft operating time?
Physical tests remain important because surface finish, lubricant film behaviour and cage dynamics can be difficult to predict. A staged campaign can first validate the simulation at a few boundary points, then use the model to explore the design space. This reduces laboratory time while preserving evidence for engineering decisions.
Measuring Fatigue Life And Friction
Fatigue life should be evaluated using more than a single calculated L10 value. That statistical rating is useful for comparison, but a geared aircraft engine requires attention to reliability, duty-cycle damage, lubrication quality and transient loads. A bearing that performs well under a constant load may have a different life prediction when variable loads are accumulated using a spectrum-based method.
Test rigs should capture torque, speed, bearing temperature, lubricant flow, vibration and acoustic emissions where possible. Friction can be reported as torque, power loss or friction coefficient, but the definition must be consistent. At high speed, even a small torque difference can become significant power loss, while temperature rise may reveal changes in lubricant shear that are not obvious from torque alone.
For fatigue testing, failure criteria should be established before the experiment starts. Options include a defined vibration increase, spall detection, a temperature excursion or physical inspection after a specified number of cycles. Censored results, where a bearing completes the test without failure, should be handled statistically rather than discarded.
Uncertainty analysis is equally important. Sensor accuracy, lubricant conditioning, mounting repeatability and specimen variation can obscure a small curvature effect. Replicated centre points and selected repeat tests help estimate pure error, while randomising the test order reduces the influence of rig drift or progressive lubricant contamination.
Connecting Results To Gearbox Reality
A bearing is part of a load path that includes gears, shafts, splines, housings and mounts. Changes in raceway curvature can alter stiffness and deflection, which may affect gear alignment and mesh behaviour. The resulting vibration may then feed back into the bearing through additional dynamic loading.
This is where a wider design-of-experiments framework is valuable. A separate study of spline fit selection illustrates how a local interface can be examined through interacting geometric, load and manufacturing factors. The same principle applies to bearing curvature: local contact performance must be considered alongside adjacent interfaces and tolerance stack-ups.
Australian operating conditions can add useful boundary cases to the model. Aircraft operating from northern Queensland may encounter high ambient temperature and humidity, while dusty regional environments can increase contamination concerns during maintenance. Long-distance operations between Sydney, Adelaide and Perth may place greater emphasis on stable thermal performance over extended cruise periods.
Manufacturing capability should also be represented realistically. Raceway grinding, heat treatment and inspection processes impose limits on achievable curvature and surface quality. If the DOE identifies a narrow optimum that cannot be held economically by an Australian or international supplier, the result is less useful than a slightly less efficient design with a wider tolerance window.
Interpreting The Design Space
DOE results should be presented through main-effects plots, interaction plots, response surfaces and sensitivity rankings. An interaction between curvature and axial load may show that a design is beneficial during cruise but poor during take-off. An interaction between curvature and lubricant viscosity may indicate that the preferred geometry changes as temperature rises.
The best choice is rarely the point with the absolute maximum fatigue life or minimum friction. A multi-objective assessment can identify a Pareto region in which improvements in one response would cause an unacceptable penalty in another. Designers can then include weight, manufacturability, thermal margin and inspection capability before selecting a configuration.
| Raceway curvature choice | Typical contact effect | Likely fatigue and friction response | Design consideration |
|---|---|---|---|
| Lower conformity | Smaller contact area and higher local pressure | Potentially higher fatigue stress, with lower sliding in some conditions | Check overload and misalignment sensitivity |
| Moderate conformity | Balanced contact ellipse and load sharing | Often a practical compromise between life and power loss | Suitable starting region for optimisation |
| Higher conformity | Broader contact area and lower peak pressure | May improve stress distribution but increase sliding or lubricant traction | Check temperature, churning and high-speed losses |
| Tight manufacturing spread | More consistent geometry | Reduces response variation across production bearings | May increase inspection or process cost |
| Wide manufacturing spread | Greater variation in contact behaviour | Can create scattered life and friction results | Requires tolerance analysis and robust design limits |
The analysis should distinguish statistical significance from engineering significance. A factor can be mathematically influential but have a negligible effect on gearbox power loss, while a small friction shift may matter greatly at a high shaft speed. Reporting confidence intervals and practical thresholds keeps the design review focused on operational consequences.
Turning Analysis Into Better Gearbox Decisions
The final output of the study should be a traceable design rule: an acceptable curvature range linked to loads, speeds, lubricant conditions and alignment limits. It should also identify which tolerances require tighter control and which can remain at standard manufacturing values. This information can flow into drawings, inspection plans, supplier specifications and maintenance documentation.
Validation should include at least one representative bearing configuration in a rig that reproduces the relevant load path and thermal environment. Where full-aircraft testing is impractical, correlated subsystem tests and validated simulations can provide evidence for scale-up. The objective is a defensible link between raceway geometry, rolling contact fatigue life, frictional power loss and gearbox performance.
For organisations developing geared aircraft engines in Australia, this method supports collaboration between universities, component suppliers, test houses and operators. It creates a common technical language for discussing tolerance, durability and efficiency, whether the work is carried out in a Melbourne laboratory, a Brisbane test facility or a maintenance environment supporting aircraft across the national network.
A well-designed DOE turns curvature from an isolated geometric parameter into an evidence-based optimisation variable. Applying that method through the OPTIMIZE project can help engineering teams select bearings that deliver lower losses, longer fatigue life and stronger tolerance to real manufacturing and operating conditions.