Optimising gearbox mount placement to minimise housing strain
Geared aircraft engines deliver remarkable power density from compact drivetrains, yet their housings rarely experience steady loading. Centrifugal forces, torque reversals and thermal growth push casing walls through complex strain cycles that resist simple calculation. A mounting arrangement that looks reasonable in static analysis can introduce bending that elevates stress well beyond design margins.
The OPTIMIZE Project has spent years examining how small architectural changes affect gearbox efficiency, durability and weight. Within that work, mount location has proven a rich area for systematic study. Researchers in Melbourne and Adelaide, collaborating with European partners, have tested whether disciplined experimental design can tame the variability seen in service.
Australia's aerospace sector has a direct stake in housing longevity. Maintenance hubs near Sydney and engineering capability at Fishermans Bend support operators such as Qantas and Virgin Australia, where every gram saved and every fatigue cycle avoided extends time on wing. Mounting optimisation sits at that intersection of structural and economic benefit.
Design-of-experiments offers a structured path through this complexity. Instead of varying one mount at a time, the technique sweeps multiple variables together, mapping interactions rather than isolating them. The output is guidance grounded in measured response surfaces, evidence that engineering teams can apply with confidence.
The strain problem in geared aero-engine housings
Gearbox casings face a cocktail of static and dynamic loads that shift with flight phase. Takeoff brings peak torque and uneven heating. Cruise settles into thermal equilibrium that still produces relative movement at bolted joints. Touchdown delivers millisecond inertia spikes. Each regime excites different strain paths through the housing walls.
Mounting points define the boundary conditions for those scenarios. A support that is too stiff in one direction forces the casing to bend around it during thermal expansion. One that is too compliant allows resonant whirl under torque pulses. In hyperstatic arrangements with more than three constraints, the geometry of the mount pattern shapes the strain field more strongly than local wall thickness often does.
The Australian Defence Science and Technology Group has documented similar sensitivities at its Fishermans Bend site, observing that moving a mount by fifteen millimetres changed principal stress direction in an adjacent rib by over twenty degrees. Such findings explain why mounting layout deserves the same rigour as gear tooth geometry.
Why design-of-experiments is a good fit
A full factorial sweep of mount coordinates would demand hundreds of finite element runs. Design-of-experiments collapses that workload by selecting a smart subset that still captures main effects and interactions. Fractional factorial or Latin hypercube designs let engineers see how six or seven geometric variables behave together with far fewer samples.
Mounting problems also suit the hierarchical nature of DoE. A screening pass eliminates negligible variables, such as the exact fore-aft position of a small accessory pad. A follow-up response surface study then zooms into survivors with higher resolution. That staged approach mirrors how gearbox design proceeds, from rough architecture to detailed definition.
Australian research groups have embraced this workflow because it pairs well with the National Computational Infrastructure in Canberra and the Pawsey Supercomputing Centre in Perth. Running hundreds of non-linear FEA cases overnight is now routine, allowing experimental matrices richer than was practical a decade ago. The result is guidance that holds across the operating envelope rather than at a single design point.
Constructing the experimental matrix
The first step is choosing which geometric parameters vary in service. Bolt pattern pitch, mount pad height, longitudinal position of the forward support and lateral offset of the rear support are typical candidates. Each factor receives a range that brackets current product variation with margin for tolerance extremes.
Levels are then selected to populate the design space. Two-level screening designs need only extremes, but response surface work adds centre points and axial runs to capture curvature. Engineers at the University of Sydney have shown that five levels per factor give enough resolution to fit a quadratic model without overwhelming the schedule. The choice depends on whether the team needs a screen or a tunable surrogate.
Boundary conditions matter as much as the factors themselves. The finite element model must apply identical torque, gyroscopic and thermal loads at every point in the matrix, so that only the mounting layout changes. That discipline turns a collection of FEA runs into a genuine experiment rather than a series of unrelated analyses, keeping factor effects distinct from mesh or load path assumptions.
Measurement and simulation under representative loads
Strain gauges remain the ground truth for any mounting study. Rosettes near junction radii and around bolt holes capture the multi-axial stress state that single-axis gauges miss. Acquisition rates must resolve torque pulses, which on some turboprop applications exceed a kilohertz. In OPTIMIZE rig trials conducted in Melbourne and Brisbane, the team logged millions of cycles at representative load spectra.
Simulation feeds the DoE loop rather than replacing it. Each candidate layout undergoes a coupled thermal-structural analysis that mirrors rig conditions, filling regions where physical testing is impractical. Agreement between measured and predicted strain at validation points gives confidence to extend the surrogate into corners the rig never visited.
Correlating the two sources also exposes modelling blind spots. If simulation consistently under-predicts strain at a pad, it usually points to a contact stiffness or bolt preload assumption needing tightening. Several Australian research students have built thesis projects around these correlations, deepening understanding of how local details drive global response in tightly packaged drivetrains.
Tolerance, lubrication and hyperstatic interactions
Mounting layout does not exist in isolation. The same forces that strain the housing walk the gears through mesh, and contact pattern reacts to small changes in relative position. A study published alongside the oil debris monitoring system work shows how bearing alignment shifts under thermal soak can accelerate pinion surface fatigue. Mounting optimisation is therefore a prerequisite for meaningful gear pitting analysis.
Tolerance stack-up adds another layer. Manufacturing variation in bolt hole position, casing flatness and pad thickness consumes part of the margin designers assume is available. A DoE that includes those tolerances as factors reveals whether the recommended layout is robust or fragile, potentially trading small nominal performance for a wider process window.
Hyperstatic mounting arrangements deserve particular care. With four or more constraints, internal loads distribute according to relative stiffness, so a small change in one support can redirect load through another in ways that are hard to anticipate. Statistical experimental design is one of the few practical tools that can map those redistributions across the full operating range.
Extracting actionable mounting guidelines
Once matrix runs complete, analysis turns raw strain data into design rules. Pareto charts identify which factors carry the most variance, while interaction plots show where two variables together produce more or less strain than their individual effects suggest. Engineers can then rank candidate layouts by predicted strain and balance performance, weight and manufacturability.
A useful output is a contour map of strain against the two most influential mount coordinates. Such maps can sit in design manuals, letting future variants be checked without repeating the DoE. The team has shared versions with a Queensland-based component supplier, so proposed changes can be screened before tooling is modified.
The guidelines also feed back into the gearbox specification. If moving the rear mount fifteen millimetres aft cuts peak strain by twelve per cent, that becomes a requirement for the structural layout team. Conversely, if a candidate location shows high sensitivity to bolt preload, the recommendation might be tighter assembly control rather than abandoning the location. Either way, decisions are evidence-based.
Validating findings on a physical rig
Even the best surrogate must face the bench. A dedicated gearbox rig instrumented with strain gauges and torque transducers loads representative housings under controlled spectra. The optimum layout identified by the DoE is installed, and its strain response is compared with a baseline. Successful validation shows the predicted reduction actually materialises in metal rather than just in the solver.
These campaigns are resource-intensive, so they are reserved for the most promising candidates. Australian facilities, including advanced laboratories at RMIT and the DSTG site in Melbourne, have hosted several such trials recently. Collaboration with European partners ensures the rigs cover operating points relevant to both regional and long-range platforms.
Field follow-up closes the loop. Operators monitor in-service strain histories of a small fleet fitted with the new layout, looking for drift from the predicted patterns. If service data tracks the model, the guidelines move into the standard manual. If surprises emerge, the DoE matrix can be revisited. That iterative rhythm keeps the method honest.
Variables commonly examined in a mounting DoE
- Longitudinal position of the forward mount relative to the main gear mesh
- Lateral offset of the aft mount to control bending moment distribution
- Stiffness ratio between forward and rear support points
- Bolt pre-load range covering torque retention during thermal cycling
- Angular orientation of mount pads under differential expansion
Practical outcomes from a well-designed mounting study
- Reduced peak von Mises strain at junction radii and rib intersections
- Lower sensitivity to manufacturing variation in bolt spacing and pad flatness
- Improved alignment retention over representative thermal cycles
- Decreased risk of fretting wear at mount interfaces during vibration
- Clear quantitative guidance for service teams during overhaul
If your team is wrestling with housing strain in a geared engine programme, the methodology above offers a path from ad-hoc placement to defensible layout. You can explore the wider research context and access additional resources through the OPTIMIZE Project website, where videos, technical reports and contact details are kept current. Bringing structured experimental reasoning to mounting decisions pays dividends in durability, weight and smoother certification conversations with CASA, and it sets a template other gearbox subsystems can follow.