Gear root fillets and bending stress in aircraft gearboxes
Geared aircraft engines rely on reduction gearboxes to transmit substantial power at high rotational speed, elevated temperature and demanding vibration levels. Understanding the effect of gear root fillet radius on bending stress concentration factor is central to choosing a tooth form that can survive repeated mesh loading without adding avoidable mass. A sensible radius supports fatigue life, but it also changes local stiffness and interacts with the rest of the gear geometry.
At mesh, tangential force creates a steep stress gradient where the tooth flank blends into the root. A sharp transition concentrates that gradient and provides a likely crack-initiation site; a generous blend can spread the load over more material. The best radius is therefore not an isolated cosmetic choice. It must preserve adequate root thickness, tip clearance and contact ratio while meeting the fatigue duty cycle of a geared aircraft engine.
Engineering teams can map this trade-off with design-of-experiments methods, finite-element analysis, tolerance studies and physical tests. Those methods are especially valuable when several tooth parameters move together during manufacture. Rather than treating a nominal geometry as a fixed shape, the program can expose which combinations control peak stress, fatigue scatter and power density.
The factor is also easy to misread. A theoretical elastic concentration factor predicts a local stress peak; it does not, by itself, predict crack life. Fatigue notch sensitivity, material strength, surface condition, residual stress, contact loads and operating history determine how much of that peak becomes damaging cyclic stress. The analysis must separate geometry effects from service and manufacturing effects.
Why the root fillet governs tooth bending
The fillet is the curved transition between the working tooth flank and the root or dedendum region. It carries a large share of the bending load because material there must redirect internal force towards the tooth centre. A small-radius blend produces a rapid change in stress flow, while a larger blend generally lowers the elastic peak and gives a crack-initiation site more surrounding material. These trends should be confirmed for each cutter-generated tooth form rather than assumed from handbook results.
Changing that blend also changes the tooth’s local compliance. Under a measured tangential load, a larger fillet may deflect differently, redistribute load across a planetary set and alter how neighbouring teeth share the mesh cycle. Engineers should check tip-root clearance and contact ratio, root thickness, pressure-angle effects and any profile modification. Interference with a mating gear tip can outweigh the fatigue benefit suggested by a single finite-element result.
This combination of performance, geometry and manufacturability shapes the project objectives behind OPTIMIZE. A useful study connects analytical trends to test conditions, states which variables were controlled and records the limits of each result. It should also show whether a lower stress peak came from the intended radius or from unintended changes in contact position, tooth spacing or load share.
How to interpret the stress concentration factor
For a linear-elastic bending case, the theoretical stress concentration factor is (K_t = \sigma_{\text{local,max}}/\sigma_{\text{nominal}}). It compares the highest local stress with the nominal stress associated with the applied loading and geometry. This definition makes (K_t) valuable for comparing fillets, but it does not imply that a 20% reduction in (K_t) will produce a 20% increase in gearbox life. Local yielding, tooth contact, dynamics and fatigue can alter that simple relationship.
The engineering estimate that enters many fatigue calculations is (K_f = 1 + q(K_t - 1)), where (q) is the fatigue notch sensitivity. It is closer to the effective stress multiplier for cyclic loading because it accounts for how readily the material responds to the notch. Even (K_f) remains a simplified term; material scatter, surface finish, case treatment, residual stress and load spectrum are assessed separately. Dynamic mesh factors should not be hidden inside a geometry-only concentration factor.
Radius sensitivity should be reported alongside the variables that can move it: pressure angle, tooth thickness, dedendum, lead modification, root geometry and cutter path. Results should include both the absolute radius and a normalised ratio such as radius divided by module. A graph of (K_t) against radius is more informative when it includes mesh-convergence evidence and the exact load application used. If stress at adjacent model boundaries differs, the comparison may reflect the model rather than the gear.
Building a representative engineering model
A finite-element model for tooth-root bending needs enough detail at the fillet to resolve curvature, contact and stress gradients. An 8-node or 20-node solid-element mesh should undergo a convergence study, with special attention to element aspect ratio and contact-patch size. For planetary stages, shaft, carrier, bearing and housing flexibility may determine load sharing; if that is neglected, the predicted benefit of a larger radius may be unreliable. A hyperstatic arrangement can hide local overload when several load paths are assumed to share perfectly.
Design-of-experiments can then separate radius from interacting variables. A useful matrix might vary fillet radius, tooth thickness, profile shift, alignment error and surface finish, while keeping speed, torque and temperature at clearly defined levels. It can include quadratic terms where curvature is expected and repeated runs where manufacturing scatter matters. Responses should include nominal (K_t), fatigue (K_f), root strain, contact-pressure distribution, mass and manufacturing time, rather than one peak-stress number alone.
Tolerance analysis turns nominal CAD into a representative hardware range. Gear milling or grinding leaves waviness and roughness; heat treatment, case hardening, lapping and handling can alter the root condition. Monte Carlo studies, worst-case bounds or measured process distributions can show whether a favourable radius loses its margin after manufacture. A robust conclusion identifies the combination of geometry and process variables that controls fatigue risk, not merely the largest radius that fits the available cutter or grinding process.
Verifying the model under real operating conditions
Physical testing should verify the chain from simulation to component life. Instrumented gear rigs can apply representative torque, speed and temperature, while strain gauges, rosettes or optical methods measure root response at several tooth positions. Tooth-level fatigue testing is useful for screening, but a gearbox test is needed to capture carrier loading, bearing compliance and load sharing among planets. Destructive specimens should be separated from high-speed spin tests so their safety cases and instrumentation plans remain distinct.
A staged campaign reduces ambiguity. First, compare radii under identical material, finish, load and alignment. Next, test the selected geometry across tolerance limits and an engine-representative duty cycle. Record crack locations, slip, bulk oil temperature, local temperature where available, and the progression of tooth-root damage. Repeatability matters: a design change that works only for one unusually clean or well-aligned article has not established a robust margin. Test reports should preserve raw measurements and state uncertainty beside derived factors.
Oil flow can also couple the static fillet result to the assembled gearbox. The project’s oil flow distribution study considers temperature gradients across a planetary gear set, which can affect lubricant viscosity, bearing clearance, distortion and thermal growth. These changes may alter mesh position and measured root strain without changing the nominal fillet. Measure oil flow and temperature at locations justified by the thermal model; do not treat a uniformly heated test stand or an assumed bulk-oil value as proof that every tooth sees the same environment.
Australian constraints and practical recommendations
Australian implementation adds supply-chain and operating realities that belong in the test plan. Melbourne supports a large engineering, precision-manufacturing and maintenance base, and Brisbane is a major hub for aerospace maintenance, repair and overhaul. Programs may also draw on specialist capability and freight links in Perth before supporting mining and energy users in the Pilbara. In remote operations, inspection access and spare-part lead times can be limited, so a validated service life and conservative replacement interval carry substantial value.
The local market often combines capable domestic machining and MRO work with imported gear cutters, grinding equipment, high-grade test instrumentation and specialist materials. Programs coordinate CAD reviews, courier movements and test bookings across interstate schedules, making configuration control and traceability essential. Record the revision, serial number, heat-treatment batch and inspection history for every test article. A backup measurement method should be agreed before a component enters a long campaign, since an imported transducer failure can otherwise create a costly gap.
For airworthiness work in Australia, the compliance plan must identify the applicable requirements under the Civil Aviation Regulations 1988 and relevant Civil Aviation Safety Authority published airworthiness standards. The responsible approved design organisation should map gearbox interfaces, material and process controls, and test evidence to the certification basis; a low fatigue factor does not replace that mapping. Where the gearbox is part of an engine installation, configuration changes also need review against the approved engine and aircraft definitions. Local regulatory evidence should therefore be planned alongside the technical matrix.
- Define each candidate radius at the same tooth thickness, root diameter and profile convention, then report (K_t) and (K_f) separately with the calculation method and mesh assumptions.
- Run mesh-converged design-of-experiments models that include pressure angle, profile shift, load sharing and speed, rather than changing the fillet while holding unrealistic geometry fixed.
- Represent material condition, roughness, residual stress, heat-treatment batch and dimensional tolerances using measured supplier data or justified bounds before ranking the candidates.
- Separate elastic screening, tooth fatigue, high-speed spin and assembled-gearbox tests, with independent safety reviews and acceptance criteria established before each campaign.
- Validate the selected design across torque, temperature, oil-flow and alignment conditions representative of Australian service, including remote-site maintenance and long lead-time constraints.
- Preserve traceability from CAD revision and test-article serial number through raw strain data, uncertainty, fatigue result, compliance mapping and final design approval.
Together, these actions prevent a promising finite-element plot from becoming an unsupported hardware claim. They also allow procurement, MRO and test teams to see which tolerances must be held locally, which can be verified by incoming inspection and which require supplier surveillance. The result is a traceable decision that can withstand design review, customer scrutiny and eventual airworthiness assessment.
Turn the evidence into a qualified design choice
The immediate action is to nominate candidate radii, freeze the common tooth geometry and issue a single test matrix for modelling and hardware review. Require every proposed benefit to name its evidence: geometry analysis for (K_t), cyclic response for fatigue life, tolerance data for robustness and component testing for assembled performance. This chain makes technical trade-offs legible to design, manufacturing, procurement and airworthiness teams.
At the next design gate, review the measured margin against mass, durability, manufacturing cost and power-density targets, then release the qualified configuration to prototype production. Record unresolved uncertainty as a controlled follow-up rather than disguising it as a nominal improvement. The strongest outcome is not the largest fillet radius; it is a repeatable root form whose stress reduction survives realistic tolerances, operating temperatures and documented Australian engineering controls.