Using Finite Element Analysis to Predict Gear Web Thickness for Weight Reduction
In a geared aircraft engine, every kilogram removed from the gearbox can contribute to lower fuel consumption, reduced structural loads, and improved power density. Yet gear weight cannot be reduced by simply making the web thinner. The web must transmit torque between the rim, hub, and adjacent components while surviving repeated load cycles, manufacturing variation, thermal effects, and high-speed operation.
Finite Element Analysis (FEA) provides a practical way to estimate the minimum safe gear web thickness before expensive prototypes are produced. By translating gear geometry, material properties, bearing reactions, and operating loads into a detailed numerical model, engineers can identify where stiffness and strength are genuinely required and where excess material can be removed.
The OPTIMIZE Project applies this kind of design thinking to power reduction gearboxes for geared aircraft engines. Its research combines simulation, design of experiments, tolerance analysis, and physical testing to improve efficiency, durability, mass, and reliability under demanding propulsion conditions.
Why Gear Web Thickness Controls More Than Mass
The web is the structural region connecting a gear’s outer rim to its hub. It transfers tangential force from the teeth toward the shaft while resisting bending, torsion, local deformation, and fluctuating loads. A thicker web usually increases stiffness and lowers stress, but it also adds rotating mass and may raise the polar moment of inertia.
Reducing web thickness can therefore improve several performance measures at once. A lighter gear requires less energy to accelerate, imposes lower centrifugal loading on some components, and can support a more compact gearbox architecture. However, reducing thickness too aggressively may cause excessive deflection at the gear rim. That deflection can disturb tooth contact, concentrate load at one side of the face width, and increase transmission error.
The design target is not simply the lowest von Mises stress. A successful design must meet several limits simultaneously:
- Static stress must remain below an allowable value with an appropriate safety factor.
- Cyclic stress must support the required fatigue life.
- Gear rim deflection must preserve acceptable tooth contact.
- Natural frequencies must remain separated from excitation orders.
- Local buckling and manufacturing distortion must be controlled.
- The design must remain robust when dimensions and loads vary.
This is especially important in aerospace gearboxes, where high rotational speed amplifies small imbalances and where a failure can affect the entire propulsion system.
Building a Reliable Finite Element Model
An FEA prediction is only as credible as the model behind it. The first step is to define the geometry with enough detail to represent load paths without making the model unnecessarily expensive. A preliminary design may use a two-dimensional axisymmetric or sector model, while a final assessment usually requires a three-dimensional model containing the hub, web, rim, fillets, and representative tooth geometry.
The web-to-rim and web-to-hub transitions deserve particular attention. Sharp corners produce artificial stress concentrations, while coarse mesh elements can hide real gradients. Fillet radii should match the proposed manufacturing geometry, and the mesh should be refined around these transitions, keyways, bolt interfaces, and any lightening features.
Boundary conditions must represent the way the gear is supported in service. Fully fixing the hub can create unrealistic restraint and artificially high stresses. A better approach may include shaft flexibility, bearing stiffness, contact interfaces, or a remote coupling that distributes the reaction across the actual mounting region. In a hyperstatic gearbox, load sharing can depend on several interacting supports, so simplified constraints should be checked against a more complete system model.
Material data also requires care. Elastic modulus, Poisson’s ratio, density, yield strength, and fatigue properties should correspond to the operating temperature and manufacturing condition. For forged or machined aerospace steels, directional properties, residual stresses, and surface treatments may affect the final result. Centrifugal loading, torque, tooth forces, bearing reactions, and thermal gradients should be applied in load cases that reflect real operating combinations rather than isolated textbook values.
Converting Stress Results Into a Thickness Decision
A useful thickness study begins with a parameter rather than a single geometry. The web thickness can be defined as a design variable, along with fillet radius, web profile, hub diameter, rim thickness, and lightening-hole dimensions. Running the model across a controlled range reveals how each parameter affects stress, stiffness, mass, and modal behavior.
For a first estimate, engineers may calculate a baseline model at several thickness values and record the critical response. The governing condition may be tooth-root loading, web bending, hub transition stress, rim displacement, or a resonant frequency. The minimum acceptable thickness is the smallest value that satisfies every relevant requirement, including uncertainty margins.
Stress interpretation should distinguish between physically meaningful concentration and numerical singularity. A sharp geometric edge can generate a stress peak that is highly mesh-dependent. Engineers should review stress gradients, use averaged or structural stress methods where appropriate, and confirm that the feature exists in the manufactured part. Removing material based on an unexamined peak can produce an unnecessarily heavy design; ignoring a genuine concentration can create a serious fatigue risk.
Large deformation and nonlinear effects may become relevant when the web is made very thin. Contact changes, material plasticity, local buckling, or friction at interfaces can alter the response predicted by a linear static analysis. A staged analysis is often efficient: use linear models to screen concepts, then apply nonlinear, transient, or contact-based analysis to the most promising designs.
Comparing Analysis Approaches
Different FEA methods answer different design questions. A simplified model is valuable during concept development because it can compare many alternatives quickly. A detailed model takes longer but provides stronger evidence for local stress, contact behavior, and fatigue assessment. Neither should be treated as a universal replacement for the other.
| Analysis approach | Best use | Main output | Important limitation |
|---|---|---|---|
| Axisymmetric or sector model | Early web-thickness screening | Global stiffness, mass, basic stress trends | May omit tooth-level and asymmetric effects |
| Linear static 3D FEA | Preliminary sizing under defined loads | Stress, displacement, reaction forces | Does not capture nonlinear contact or large deformation |
| Detailed tooth-and-web model | Local load transfer and fatigue review | Tooth-root stress, rim flexibility, web transition stress | Higher modelling and computational cost |
| Modal analysis | Vibration and resonance screening | Natural frequencies and mode shapes | Requires realistic mass and boundary conditions |
| Nonlinear contact analysis | Interfaces, preload, and changing contact | Load redistribution and local deformation | Sensitive to contact definitions and solver settings |
| Transient or harmonic analysis | Dynamic operating conditions | Time-dependent stress and response | Needs accurate excitation and damping data |
A robust workflow uses these approaches in sequence. A sector model may identify a promising thickness range, a three-dimensional model can refine the geometry, and modal or transient analysis can establish whether the weight reduction introduces a dynamic problem. This staged process limits computational effort while preserving engineering confidence.
Design-of-experiments methods make the process more informative than a simple trial-and-error sweep. Instead of changing one dimension at a time, an experiment plan can evaluate interactions between web thickness, fillet radius, hub diameter, and material properties. Response surfaces or surrogate models then estimate performance across the design space and highlight combinations that offer low mass with sufficient safety margin.
Handling Variation, Fatigue, and Manufacturing Reality
A nominal FEA model represents only one manufactured gear. Actual components vary because of machining tolerances, heat treatment, residual stress, surface finish, profile deviations, and local thickness differences. Operating conditions vary as well. Torque, speed, temperature, lubrication quality, bearing alignment, and gear misalignment can shift the load path.
Tolerance analysis can connect these variations to the FEA results. For example, a small change in web thickness may have little effect when the gear is lightly loaded but become significant when combined with a reduced fillet radius and an unfavorable bearing reaction. Statistical sampling, polynomial response surfaces, or probabilistic simulation can estimate the likelihood of exceeding a stress or deflection limit.
Fatigue assessment must use the correct stress history. Gear webs experience cyclic loading as teeth engage and disengage, while the rotating component may also encounter once-per-revolution forces and transient torque events. Mean stress effects, surface condition, notch sensitivity, and residual stresses should be included where suitable material data is available. A design that passes a static yield check may still fail its intended life requirement if cyclic stress is concentrated near a transition.
Manufacturing constraints should enter the optimization before the final geometry is selected. Very thin webs may be difficult to machine without distortion, difficult to inspect, or vulnerable to residual-stress release. A design with generous fillets, accessible surfaces, and stable machining allowances may weigh slightly more than the theoretical minimum yet provide better repeatability and lower production risk.
The wider research context is illustrated by resources such as this project blog resource, which forms part of the online material associated with the project. When reviewing any external project resource, engineers should distinguish general communication material from validated technical evidence and rely on documented methods, test data, and traceable assumptions for design decisions.
Validating the Model With Physical Testing
Testing is essential because FEA predicts behavior through assumptions about loads, constraints, material response, and geometry. A gearbox component can be instrumented with strain gauges or fiber-optic sensors to compare measured strain with calculated values. Static loading can establish whether the global stiffness is correct, while spin testing can reveal centrifugal, dynamic, and thermal effects.
A useful validation plan begins with a representative component rather than waiting until the complete gearbox is finished. A gear or subassembly can be tested under controlled torque and speed conditions, allowing engineers to compare deflection, strain distribution, and vibration response. If the measured mode frequencies differ from the model, the cause may be incorrect boundary stiffness, missing mass, material data, or an overlooked contact interface.
Correlation should examine trends as well as individual values. If increasing torque produces a proportional strain increase in both test and simulation, the load path may be correctly represented even if a local peak needs refinement. If the discrepancy grows at high speed, centrifugal stiffening, bearing behavior, thermal expansion, or aerodynamic effects may need to be added.
Test results can then update the design model and improve subsequent optimization cycles. This closes the loop between numerical prediction and physical evidence, reducing the chance that weight reduction depends on an idealized geometry that cannot survive service conditions.
A Practical Optimization Workflow
A disciplined process keeps weight reduction connected to structural safety and production capability. The following recommendations help turn FEA into a repeatable engineering method:
- Define thickness, fillet radius, hub dimensions, and rim geometry as explicit design variables.
- Apply realistic torque, tooth forces, speed, temperature, bearing reactions, and constraint stiffness.
- Use mesh convergence and sensitivity studies around the hub, rim, and fillet transitions.
- Evaluate stress, deflection, fatigue life, modal separation, mass, and manufacturing limits together.
- Validate the leading design through strain, spin, vibration, or component-level testing.
Optimization should use a clear acceptance envelope rather than a single target value. For example, the best candidate may be the lightest design that stays below a fatigue damage limit, maintains tooth-contact alignment, avoids a critical frequency range, and remains manufacturable across the specified tolerance band.
The final geometry should be reviewed at system level. Removing mass from one gear may change torsional dynamics, shaft reactions, balance, or bearing loads elsewhere in the gearbox. In a geared aircraft engine, local optimization has to support the performance of the complete power transmission system.
Accurate finite element modelling can make gear web thickness a controlled design variable instead of a conservative guess. Apply the workflow to a parameterized gear model, compare the numerical results with physical evidence, and use the resulting insight to develop lighter, stronger, and more reliable aircraft gearbox components.