Optimizing bearing preload for hyperstatic housing deflection control
Aerospace propulsion has spent two decades chasing the same goal: produce more usable shaft horsepower from every gram of metal that flies. Geared turbofan and open-rotor architectures answered that question by inserting a power reduction gearbox between the turbine and the fan or propeller, and that gearbox quietly became the most thermally and mechanically loaded subsystem in the engine. The bearings that carry the high-speed pinion are no longer the simple rolling-element assemblies of the past; they sit inside casings that are deliberately stiff, deliberately light, and increasingly hyperstatic, which forces a careful conversation about how much preload is actually helpful versus how much is structural noise amplified by the housing itself.
Recent work by the OPTIMIZE project focuses on exactly that conversation. The programme, anchored by a network of European and Australian research partners, treats bearing preload as a tunable variable that interacts with housing distortion, manufacturing variation, lubrication regime, and thermal growth. Engineers working from Melbourne and Adelaide have already contributed measurement data that shows how a hyperstatic casing can multiply a deliberately chosen preload into something far beyond the original specification, and the design of experiments work that follows aims to put a hard envelope around that phenomenon rather than let it drift unchecked.
The role of preload in high-speed reduction gears
Inside a power reduction gearbox, the bearing pack supports a pinion or planet carrier that may spin at fifteen thousand to twenty-five thousand revolutions per minute while transmitting considerable torque. Preload is the residual elastic force applied across the bearing assembly once the mechanical stack has been closed and the rotor is in place, but before any external service load is applied. The intent is to push the rolling elements firmly against both inner and outer raceways so that any subsequent rotor movement is countered by known stiffness rather than by unpredictable internal clearance or by sudden metal-to-metal contact.
A correctly set preload suppresses the rattling that would otherwise occur during low-load flight regimes, holds the gear flanks on a stable line of action, and limits the secondary motion of the shaft that drives bearing wear. If the value is set too low, the rotor wanders and the gear mesh pattern shifts at every throttle burst, accelerating surface fatigue on both the bearing raceways and the gear flanks. If the preload climbs too high, friction rises, lubricant film temperature follows, and the contact ellipse beneath each rolling element can locally overload the raceway in a way that produces spalling within a fraction of the design life.
At the Australian aerospace cluster around Fishermans Bend, and at the Defence Science and Technology Group laboratories at Edinburgh in South Australia, engineers have tracked the impact of preload on bearing endurance for some years. Those studies consistently find that small miscalculations in preload are amplified by any micro-distortion of the casing, which is exactly where the hyperstatic geometry problem enters the picture.
Hyperstatic housings and where the deflection problem starts
A hyperstatic housing is one that has more constraint paths than the minimum needed to fix every point in space. A simple two-bolt cover plate might be statically determinate; a six-bolt casing with closely spaced ribs and stiffening webs is statically indeterminate. The added rigidity is welcome for vibration and acoustic reasons, but it comes with a price: under torque, gyroscopic loads, and thermal growth, the housing cannot freely distort in the direction the load is pushing. The strain is distributed, and often amplified, at the bearing seats.
That distributed strain is what engineering teams call hyperstatic housing deflection. Rather than the casing flexing uniformly, certain ribs and lugs hog inward by measurable amounts, and those inward hogs push inward on the outer race of the bearing, multiplying the preload that was deliberately set on the bench. The bearing is now carrying not only the original preload but the additional squeeze imposed by the housing distortion, and that combined load can rapidly exceed the design envelope at high shaft speeds.
In a thrust class benchmarked against the airworthiness expectations published by Australia's Civil Aviation Safety Authority, the operating window for shaft speed and torque is narrow, so the design team does not have the luxury of letting a bearing pack absorb unexpected radial loads without consequence. The bearing becomes the soft target inside an otherwise stiff structure, and the structure's stiffness itself becomes the bearer of the problem. Any program that wants to fine-tune bearing life has to treat the housing as an active element in the load path, not just a passive container.
How bearing preload counters housing distortion
The mechanism by which bearing preload counters housing distortion is, at its heart, a stiffness matching problem. When the rotor attempts to whirl or to orbit under engine torque, the preload forces the rolling elements to react predictably against the raceways. If the housing footprint under the outer race is too compliant, the bearing recoils inward and the rotor settles into a low-frequency mode that the lubricant film cannot easily damp. A carefully chosen preload elevates the natural frequency of this mode above the operating excitation, locking the rotor into a quieter response.
In reduction gearboxes where the gear mesh is the dominant excitation source, the preload also acts on the pinion position. A positive axial preload removes the small end-float that allows a helical gear to walk along its shaft under load, which keeps the helix lead correct across the whole thermal range. This is one of the reasons design of experiments studies on the OPTIMIZE gear-drivetrain platform focus intensely on preload, since the variable sits at the intersection of bearing behaviour, gear behaviour, and housing behaviour.
Engineers in Brisbane's aviation precinct have built fixtures capable of measuring this housing-borne preload amplification directly. By instrumenting the housing with strain gauges near the bearing seat and using proximity probes over the rotor, they can plot the additional squeeze applied to the bearing as the unit goes from cold idle to full take-off thrust. The plots consistently show that casings designed in a hyperstatic arrangement push an extra load onto the bearing, unless the original preload is deliberately reduced to leave room for the dynamic squeeze. Discipline on the test stand is essential, and the team applies the same care as a forager working to identify edible mushrooms safely, checking each input twice before trusting a result.
Tolerance analysis and design of experiments around preload
Because the additional load depends on small variations in housing wall thickness, rib angle, bolt pretension, and bearing seat roundness, the preload value cannot be selected once and forgotten. The OPTIMIZE project uses a design of experiments framework in which dozens or hundreds of virtual builds of the gearbox are solved with small, statistically chosen perturbations in each variable. The response of interest is the peak contact pressure at the bearing raceway under simulated flight conditions, with secondary responses covering gear mesh efficiency, housing temperature, and lubricant film thickness.
The sweeps include manufacturing variations that are realistic for Australian supply chains. Wall thickness is drawn from a Gaussian distribution centred on the nominal with a standard deviation of a few tens of micrometres. Rib fillet radii vary within the supplier's drawing tolerance, and bearing seat roundness is matched to the metrology reports from local coordinate measuring machine providers. By feeding the responses back through a tolerance analysis, the engineering team can identify the limited set of dimensions that drive most of the housing-induced preload amplification, and the housing drawings can then be steered to where they have the highest leverage on bearing life.
This is where the project's collaboration with its partners matters. A meaningful DOE needs realistic supplier variation, and the partner laboratories help anchor the simulated variation to the actual measured variation in freshly machined casings. Without that loop, the optimised preload would be tuned to a housing that exists only in the simulation, and any divergence between the simulation and the shop floor would reintroduce the original risk the programme set out to remove.
Practical trade-offs and selection of preload method
The choice of preload method shapes everything else. Some designs rely on a precisely sized spacer between the bearing rings; others use a stack of disc springs that flex under load. Aerospace units commonly specify a hydraulic or elastomeric element that absorbs thermal growth while keeping a near-constant face load. Each method interacts differently with hyperstatic housing deflection, because each one reacts not only to shaft thrust but also to the radial squeeze imposed by the casing.
In the OPTIMIZE programme, candidate methods are evaluated against a fixed set of design criteria. The comparison below summarises how four common approaches behave under hyperstatic housing conditions, their sensitivity to manufacturing variation, and the typical cost to inspect and rework on the assembly line. The numbers are normalised for confidentiality, but the relative behaviour is representative of the data captured on the Australian and European test rigs.
| Preload method | Behaviour under hyperstatic housing deflection | Sensitivity to manufacturing variation | Assembly complexity |
|---|---|---|---|
| Precision spacer | Rigid coupling, transfers most housing distortion to the bearing | High | Low |
| Disc spring pack | Moderate compliance, partially absorbs the housing squeeze | Moderate | Moderate |
| Hydraulic preload | Acts as an isolator, decouples the bearing from most housing distortion | Low | High |
| Elastomeric element | Viscous damping, reduces dynamic amplification of housing squeeze | Low to moderate | Moderate to high |
For a housing whose rigidity is part of the value proposition, a hydraulic or elastomeric element sits more naturally, since it accepts the small radial distortion without pushing the full value back into the bearing raceway. The trade-off lives at the interface of lubrication, thermal expansion, and assembly tolerances, and engineers rarely optimise for one alone. Local aerospace manufacturers, including suppliers around the Avalon precinct in Victoria and the TAE Aerospace network in Queensland, generally default toward elastomeric or hydraulic preload for new architectures, partly because Australian maintenance, repair, and overhaul providers can re-pressurise a hydraulic element in the field without replacing the bearing pack. Australian defence and civil programmes also place weight on first-time-right assembly, because the supply chain is small and downtime is expensive. That is why the OPTIMIZE team has pushed to publish the design rules behind the comparison in a form that procurement and airworthiness specialists can reuse when they review a new gearbox specification.
If your engineering team is evaluating a power reduction gearbox for an unmanned aerial vehicle, a regional airliner, or a future civil propulsion demonstrator, the OPTIMIZE project welcomes technical conversations. Engineers can reach the technical leads through the contact channel linked from the programme's web presence, where a short message describing the application envelope usually returns a useful reply within a working week. Partners looking to contribute measurement data, host a bearing fixture, or supply housing castings at the tolerance band required for the DOE sweeps are invited to introduce themselves through the same route so the consortium can add their capability to the published list.