In a helicopter main gearbox, the input stage can run above 20,000 r/min. At that speed, the spiral bevel pinion does more than transmit torque: it drags air and oil through the housing, and the resulting windage loss is independent of load. The loss appears as a parasitic torque, adds heat, and increases the cooling burden. Because it grows strongly with speed, windage becomes a first-order term in high-speed aerospace transmissions rather than a small correction.
Windage is produced by pressure and viscous shear on moving surfaces. A spiral bevel gear is not a simple disc. It has a large end face, a small end face, two conical rims, curved tooth flanks, and tooth cavities that exchange fluid with the surrounding housing. Empirical correlations calibrated on single discs or single gears are useful, but their range is limited and they do not resolve which surface dominates. A surface-decomposed model can answer that question and provide a basis for geometry changes.
Rotating-Disc Baseline: Boundary Layers and Regime Transition
The starting point is a disc of radius \(R\) spinning at angular speed \(\omega\) in otherwise quiescent fluid. The local flow regime is set by the rotational Reynolds number
$$Re_\omega=\frac{\rho\omega R^2}{\mu}$$
where \(\rho\) is density and \(\mu\) is dynamic viscosity. Near the axis the flow tends to remain laminar. As \(R\) or \(\omega\) increases, the outer region can become unstable, then transitional, and finally turbulent. Reported critical ranges place the onset of instability at about \(1.78\times10^5\) to \(2.12\times10^5\), the transition to fully turbulent flow at roughly \(2.7\times10^5\) to \(2.99\times10^5\), and fully turbulent conditions near \(7\times10^5\). A representative transition value is often taken near \(2\times10^5\), but the actual value depends on local flow conditions and surface roughness.
A rotating disc drives fluid by shear. Fluid adjacent to the surface approaches the local tangential velocity \(\omega R\), while fluid farther away slows. Centrifugal action then pushes fluid radially outward inside the boundary layer, and mass conservation forces fluid to enter axially from the surrounding region. For a laminar boundary layer, the thickness is approximately
$$\delta\approx 5.5\sqrt{\frac{\nu}{\omega}}$$
where \(\nu=\mu/\rho\) is kinematic viscosity. The thickness is therefore independent of disc radius; it is controlled by speed and viscosity. The entrained mass flow can be written in dimensionless form as
$$\dot m_0=2.779\rho R^2\sqrt{\omega\nu},\qquad C_w=\frac{\dot m_0}{\mu R}=2.779Re_\omega^{1/2}$$
For one face of the disc, the laminar and turbulent windage powers are expressed as
$$P_{w,\mathrm{face}}^{(L)}=0.484R^4\rho\nu^{1/2}\omega^{5/2}$$
$$P_{w,\mathrm{face}}^{(T)}=0.0366R^5\rho\omega^3Re_\omega^{-1/5}$$
Summing the two faces gives the two-sided laminar disc loss \(0.967R^4\rho\nu^{1/2}\omega^{5/2}\). In the turbulent case the weaker \(Re_\omega^{-1/5}\) dependence reflects the fact that the torque coefficient falls slowly as Reynolds number rises. For turbulent field closure, the SST \(k\)-\(\omega\) model is a practical choice because it blends near-wall and free-stream behavior while resolving strong near-wall gradients.
Conical Rims and the Step from Disc to Cone
Disc faces are not the only wetted surfaces. The cylindrical rim of a disc also produces viscous drag. For a rim of width \(b\), the rim power is
$$P_{w,\mathrm{rim}}=4\pi\mu bR^2\omega^2$$
A cone can be treated as a stack of thin cylindrical slices. If the cone has end radii \(r_1\) and \(r_2\) and axial width \(b\), integrating the local shear over the sloping rim gives
$$P_{w,\mathrm{cone\ rim}}=\pi\mu\omega^2b(r_1+r_2)^2$$
These rim terms are purely viscous in the present treatment. They are usually much smaller than the tooth-generated pressure loss in a high-speed gear, but they are retained because they complete the surface budget and become relatively more visible when the teeth are shrouded or when the gear runs in a low-density medium.
Surface Decomposition of a Spiral Bevel Gear
The spiral bevel gear is split into five contributions: the large end face, the small end face, the large-end conical rim, the small-end conical rim, and the tooth cavities. The tooth flanks themselves are treated through the cavity pressure field rather than as independent smooth plates. Shaft drag is neglected. Each surface is first evaluated as if it rotated independently, and the contributions are summed. This is a deliberate simplification: it ignores interference between adjacent surfaces, but it preserves the geometric scaling that controls the ranking of losses.
Let \(R_1\) and \(R_2\) be the large- and small-end face radii. For local Reynolds numbers \(Re_{\omega,1}=\rho\omega R_1^2/\mu\) and \(Re_{\omega,2}=\rho\omega R_2^2/\mu\), the end-face losses are
$$P_{w1}^{(L)}=0.484R_1^4\rho\nu^{1/2}\omega^{5/2},\qquad P_{w1}^{(T)}=0.0366R_1^5\rho\omega^3Re_{\omega,1}^{-1/5}$$
$$P_{w2}^{(L)}=0.484R_2^4\rho\nu^{1/2}\omega^{5/2},\qquad P_{w2}^{(T)}=0.0366R_2^5\rho\omega^3Re_{\omega,2}^{-1/5}$$
The conical rims use the back-cone and front-cone widths \(b_1\) and \(b_2\), together with the large- and small-end tip radii \(R_{a1}\) and \(R_{a2}\):
$$P_{w1}=\pi\mu\omega^2b_1(R_1+R_{a1})^2,\qquad P_{w2}=\pi\mu\omega^2b_2(R_2+R_{a2})^2$$
The tooth cavity is the most important and least smooth part of the flow. Fluid is drawn axially toward the small end, turned outward by centrifugal action, and then split between the surrounding housing, the small-end rim, and the tooth spaces. Fluid that enters the tooth spaces can leave partly at the large end and partly across the tooth flanks. The model treats the cavity fluid as a lumped rotating volume and drops axial velocity gradients and body forces in the cavity momentum balance. With the large-end outlet taken as the zero-pressure reference, the radial pressure distribution is
$$p(R)=\frac{\mu^2}{8\rho}\left(\frac{1}{R^2}-\frac{1}{R_{a1}^2}\right)+\frac{1}{2}\rho\omega^2(R^2-R_{a1}^2)$$
The average cavity pressure follows from integrating over the cavity radius:
$$\bar p=\frac{\mu^2}{16\rho}\left(\frac{1}{R_{a2}^2}-\frac{1}{R_{a1}^2}\right)+\frac{1}{4}\rho\omega^2(R_{a2}^2-R_{a1}^2)$$
The pressure difference is then applied over the effective cavity area and a representative cavity radius to obtain a per-tooth loss \(P_m\). For a gear with \(z\) teeth, the total windage for a purely laminar or purely turbulent estimate is
$$P_w^{(L)}=zP_m+P_{w1}^{(L)}+P_{w2}^{(L)}+P_{w1}+P_{w2}$$
$$P_w^{(T)}=zP_m+P_{w1}^{(T)}+P_{w2}^{(T)}+P_{w1}+P_{w2}$$
In a real gearbox the flow may be mixed: laminar near the axis, transitional over part of the face, and turbulent farther out. The model selects the local regime through \(Re_\omega\) and sums the corresponding face terms. When oil is present, the same equations are retained with mixture properties. For a second-phase volume fraction \(f_m\),
$$\rho_{\mathrm{mix}}=\rho_1(1-f_m)+\rho_2f_m,\qquad \mu_{\mathrm{mix}}=\mu_1(1-f_m)+\mu_2f_m$$
The mixture density and viscosity replace \(\rho\) and \(\mu\) in the face, rim, and cavity terms. This is a homogeneous-flow approximation; it does not resolve separate air and oil films, but it captures the first-order effect of oil on the effective inertia and shear stress.
Example Calculation for a 27/81 Tooth Pair
The model was applied to an aerospace spiral bevel gear pair with 27 pinion teeth and 81 gear teeth. The basic geometry is listed in Table 1, and the structural parameters used to define the end faces and rim widths are listed in Table 2.
| Parameter | Pinion | Gear |
|---|---|---|
| Number of teeth | 27 | 81 |
| Large-end module (mm) | 3.85 | 3.85 |
| Pressure angle (°) | 20 | 20 |
| Spiral angle (°) | 35 | 35 |
| Face width (mm) | 38 | 38 |
| Outer cone distance (mm) | 164.31 | 164.31 |
| Shaft angle (°) | 69.77 | |
| Pitch cone angle (°) | 16.91 | 52.86 |
| Hand | Left | Right |
| Parameter | Pinion | Gear |
|---|---|---|
| Hub radius (mm) | 20 | 35 |
| Small-end web radius (mm) | 27 | 106.6 |
| Large-end web radius (mm) | 35 | 148.4 |
| Front-cone rim thickness (mm) | 3.6 | 6 |
| Back-cone rim thickness (mm) | 10 | 20 |
The surrounding fluid was initialized at \(1.0\times10^5\) Pa and 298 K. Air density was 1.205 kg/m³ and air dynamic viscosity was \(1.81\times10^{-5}\) Pa·s. Oil density was 998 kg/m³ and oil dynamic viscosity was \(6.04\times10^{-2}\) Pa·s. The specific heats were 717 J kg⁻¹ K⁻¹ at constant volume and 1005 J kg⁻¹ K⁻¹ at constant pressure. The dry case had zero oil volume fraction. After oil injection and flow stabilization, the reported air–oil volume-fraction ratio was about 1:33, and the corresponding mixture density and viscosity were substituted into the model.
The calculated loss distribution is strongly skewed toward the teeth. For a single gear, pressure drag on the tooth flanks accounted for about 92% of total windage. The large and small end faces together contributed about 6%. Conical rims, tooth tips, and cavity/root vortex losses filled the remainder. Over the speed range considered, total windage for the single gear remained on the order of a few watts to tens of watts, while the cavity/root vortex term reached only a few watts at its maximum.
This ranking is consistent with independent work on spur, helical, and spiral bevel gears. Those results also place the tooth region above 90% of the total, with the large end face generally ahead of the small end face in spiral bevel geometry, and they identify pressure difference rather than viscous shear as the main contributor. The example therefore supports the model’s use as a ranking tool: it reproduces the dominant surface hierarchy without requiring a full three-dimensional simulation of every case.
Interpretation, Design Implications, and Model Limits
The central result is that wetted area alone is not a sufficient design variable. The tooth-space pressure field dominates because the teeth act as a row of centrifugal impellers: fluid is drawn in near the small end, accelerated outward, and discharged through restricted gaps. Smooth disc faces and conical rims contribute, but they are secondary at high speed. This is why a shroud that reduces the external disc flow may still leave a large tooth-space loss if the cavity pressure field is not managed. Effective mitigation has to address the axial inlet near the small end, the large-end outlet, and the oil distribution together.
The model also provides a direct way to compare design changes. Increasing the large-end outlet area, reducing the small-end inlet restriction, or changing the cavity depth alters the pressure term before it alters the wetted area. The face and rim terms remain useful for cases where the teeth are fully shrouded or where a smooth rotating component dominates the loss. The two-phase correction allows the same framework to be used for dry air and for oil-laden air, although the homogeneous mixture assumption cannot resolve stratified oil films or local oil pooling.
The limitations are important. The flow is treated as steady, axisymmetric, and incompressible. Surface contributions are summed independently, so interference between the tooth flanks, end faces, and housing is not resolved. The cavity is represented as a lumped fluid volume, shape drag and wake separation are neglected, and shaft drag is omitted. The model evaluates a single gear rather than a meshing pair; it does not capture the air displacement caused by the mating tooth or the mutual interaction of two rotating gears. Transition Reynolds numbers are sensitive to local conditions and surface roughness. The example verification is a consistency check against measured and simulated rankings, not a full validation over all speeds, oil fractions, and housing clearances.
Within those boundaries, the model is useful at the preliminary design stage. It identifies the tooth cavity as the controlling region, gives a physically based scaling for the end faces and rims, and shows why windage mitigation in high-speed spiral bevel gears must be based on pressure management rather than surface area alone. Final thermal and efficiency predictions still require CFD or instrumented spin testing, especially when the gearbox operates with unsteady oil flow or tight shroud clearances.