Why tooth-body temperature governs nutation drive reliability
Nutation drives transmit motion through a bevel gear whose axis is tilted by a small angle relative to its mate and which wobbles rather than spins about its own centreline. Because the reduction ratio is set by the difference in tooth number between the two conical members — only two teeth in the 30/32 pair examined here — a single stage delivers a large ratio, high load capacity and a short axial envelope, which is what makes the layout attractive for aerospace actuators and robot joints.
The same compactness that makes the arrangement attractive also throttles heat rejection. A single-stage reducer of this type packs the gear pair, bearings and support structure into a small housing with restricted oil circulation, while the tooth contact itself carries combined rolling and sliding. Frictional heat generated in the contact raises the bulk, or tooth-body, temperature above the lubricant inlet value. Bulk temperature is the baseline on which the short-lived flash temperature at each contact instant is superimposed; it therefore feeds directly into scuffing and seizure risk, thermal distortion of the tooth profile, contact fatigue life, vibration and noise. Once the flank temperature is high enough for the lubricant film to break down, adhesive wear escalates quickly and the drive can jam.
Published gear thermal work has largely concentrated on external operating conditions — rotational speed, load, oil supply and nozzle placement — for cylindrical gears, spiral bevel gears in helicopter transmissions, worm gears, face gears and rolling bearings. For the double circular-arc spiral bevel gear used in a nutation drive, the influence of the gear’s own design parameters has received much less attention, even though module, helix angle and nutation angle are frozen at the design stage and cannot be corrected later. A steady-state finite element model of a single tooth carrying a moving frictional heat source makes it possible to rank the sensitivity of bulk temperature to four variables: module, helix angle, nutation angle and lubricant inlet temperature.
Gear pair geometry and mesh kinematics
The analysis uses a double circular-arc spiral bevel gear pair whose basic parameters are listed in Table 1. Both members share the same helix angle of 25°, module of 2.5 mm and pitch cone distance of 60 mm, while their pitch cone angles of 51.04° and 123.96° sum to 175°, i.e. the shaft angle is 180° minus the 5° nutation angle. The 30-tooth outer gear is the component examined in the thermal model.
| Parameter | Outer bevel gear | Inner bevel gear |
|---|---|---|
| Helix angle β (°) | 25 | 25 |
| Nutation angle ε (°) | 5 | 5 |
| Number of teeth | 30 | 32 |
| Pitch cone angle (°) | 51.04 | 123.96 |
| Pitch cone distance (mm) | 60 | 60 |
| Module m (mm) | 2.5 | 2.5 |
Tooth contact analysis locates the heat source
Heat is generated only where the two flanks slide against each other, so the meshing zone must be identified before any thermal load can be applied. According to the gear meshing principle, a valid contact point requires the position vectors and the unit normal vectors of the outer and inner members to coincide when expressed in a common fixed frame $S_m(X_m, Y_m, Z_m)$:
$$\mathbf{r}_{m1}(\alpha_1,\theta_1,\phi_1)=\mathbf{r}_{m2}(\alpha_2,\theta_2,\phi_2),\qquad \mathbf{n}_{m1}(\alpha_1,\theta_1,\phi_1)=\mathbf{n}_{m2}(\alpha_2,\theta_2,\phi_2)$$
Here $\mathbf{r}_{m1}$, $\mathbf{r}_{m2}$ are the position vectors and $\mathbf{n}_{m1}$, $\mathbf{n}_{m2}$ the normal vectors of the outer and inner gear flanks in $S_m$; $\alpha_1$, $\alpha_2$ are the circular-arc angles of the two flanks; $\theta_1$, $\theta_2$ are the tooth-trace angles; and $\phi_1$, $\phi_2$ are the rotation angles of the two flank surfaces. The system contains six unknowns, five of which can be solved once $\phi_1$ is prescribed. Sweeping the solution from the heel to the toe of the outer gear yields 15 discrete contact points, numbered 1 to 15, which together define the meshing interval used as the heat-source path.
Frictional heat flux entering the flanks
Sliding between the flanks is the dominant heat source. Contributions from rolling friction and from elastic–plastic deformation of the metal are small by comparison and are neglected, so the frictional heat at a contact point is written as
$$Q = f P_{\max} v_{12}$$
where $Q$ is the frictional heat, $f$ the sliding friction coefficient taken as 0.05, $P_{\max}$ the maximum contact stress and $v_{12}$ the relative sliding velocity between the two teeth. Assuming that the meshing gears absorb the whole of the generated heat, the mean heat flux entering each member at a corresponding pair of contact points is
$$q_1=\frac{\beta’ Q \omega_1}{2\pi v_1 b},\qquad q_2=\frac{(1-\beta’)Q \omega_2}{2\pi v_2 b}$$
with $q_1$, $q_2$ the mean input heat flux of the two gears, $\omega_1$, $\omega_2$ their angular velocities, $\beta’$ the heat-flux partition coefficient (0.5, i.e. an equal split between the two flanks), $b$ the contact half-width, $v_1$, $v_2$ the tangential velocities at the contact point, $R$ the composite radius of curvature and $E$ the composite elastic modulus of the pair:
$$b=\sqrt{\frac{8P_{\max}R}{\pi E}}$$
Because relative sliding velocity, maximum contact stress and hence heat flux all vary from point to point along the path of contact, the thermal load applied to the finite element mesh is not uniform. The flux differs between the convex and concave flanks as well, since the curvature and the sliding geometry are not symmetric between them. This spatial variation, rather than the absolute magnitude of the flux, is what shapes the resulting temperature field.
Governing equation and thermal boundary conditions
At the speeds of interest, one revolution of the gear takes far less time than the temperature field needs to redistribute, so every tooth on the wheel sees essentially the same thermal history. The model can therefore be reduced to a single tooth. For a body without internal heat sources, the steady-state conduction equation is
$$\lambda\left(\frac{\partial^2 T}{\partial x^2}+\frac{\partial^2 T}{\partial y^2}+\frac{\partial^2 T}{\partial z^2}\right)=0$$
where $\lambda$ is the thermal conductivity of the gear material, $T$ the bulk temperature field and $x$, $y$, $z$ Cartesian coordinates. The field is closed by five boundary conditions applied to the surfaces identified in the single-tooth model:
Meshing working surface (zone N), which receives the frictional flux $q$:
$$-\lambda\frac{\partial T}{\partial n}=\alpha_t (T_f-T_\tau)-q$$
Tooth tip, tooth root and non-meshing working surfaces (zone t):
$$-\lambda\frac{\partial T}{\partial n}=\alpha’_t (T_f-T_\tau)$$
Gear end faces (zone S):
$$-\lambda\frac{\partial T}{\partial n}=\alpha_s (T_f-T_\tau)$$
Tooth bottom (zone d), treated as adiabatic:
$$\frac{\partial T}{\partial n}=0$$
And the two dividing-tooth sections (zones p, q), which are coupled by continuity of temperature and of normal gradient:
$$T\big|_p=T\big|_q,\qquad \frac{\partial T}{\partial n}\Big|_p=\frac{\partial T}{\partial n}\Big|_q$$
In these expressions $\alpha_t$ is the convective coefficient of the meshing flank, $\alpha’_t$ that of the non-meshing flank, $\alpha_s$ that of the end faces, $T_f$ the ambient (lubricant) temperature, $T_\tau$ the local body temperature, $n$ the outward normal of the heat-exchange surface and $q$ the frictional heat flux entering the meshing flank. The values used in the calculation are collected in Table 2; the remaining surfaces have negligible convective exchange and are treated as adiabatic. The gear is made of 45 carbon steel, the ambient temperature is 100 °C, and the moving heat source is applied through the 15 contact points by means of an APDL routine that sweeps the flux along the path of contact.
| Boundary | Symbol | Coefficient W/(m²·K) |
|---|---|---|
| Meshing flank | αt | 2 198.4 |
| Non-meshing flank, tip and root | α′t | 1 681.3 |
| End faces | αs | 5 043.8 |
Predicted bulk temperature field
The converged field on the outer bevel gear is markedly non-uniform. The hottest region lies in the central part of the meshing contact zone, where the frictional flux is concentrated and where the surrounding material offers the longest conductive path to a cooled surface. The tooth tip runs cooler because it presents a large area to the oil, and the gear end faces form distinct low-temperature zones: they are washed by lubricant with a convective coefficient of 5 043.8 W/(m²·K), more than twice that of the meshing flank, while receiving very little conducted heat.
Heat flows outward from the contact centre in all directions, and along the face width the temperature follows a parabolic profile. Traced along the path of contact from heel to toe, the nodal temperatures on both the convex and the concave flank rise to a peak near the middle of the engagement interval and fall away towards both ends, mirroring the distribution of the frictional flux that drives them. The same qualitative pattern appears on the convex and concave flanks, although the absolute levels differ because the two surfaces carry different curvature and sliding conditions.
Module: coarser teeth run cooler
Holding every other quantity at its baseline, the module was varied over 2, 2.25, 2.5, 3 and 3.5 mm, with the boundary conditions recomputed for each geometry. The bulk temperature falls at every point of both flanks as the module increases. Taking the convex flank and the 2 mm case as the reference, the maximum bulk temperature is 2.1% lower at 2.25 mm, 3.5% lower at 2.5 mm, 7.8% lower at 3 mm and 10.5% lower at 3.5 mm.
Two mechanisms act in the same direction. A larger module increases the local radii of curvature at the contact, so the load is spread over a wider Hertzian area and the peak contact stress drops; since the frictional flux is proportional to $P_{\max}$, the heat input falls with it. At the same time, the thicker tooth exposes more surface to the lubricant, so the convective coefficient per unit of heat input rises. The combination produces the monotonic cooling trend observed across the sweep.
Helix angle: axial thrust raises the contact load
Sweeping the helix angle over 18°, 20°, 25°, 30° and 35° produces the opposite trend. Relative to 18°, the maximum bulk temperature on the convex flank rises by 4.6% at 20°, 7.2% at 25°, 11.2% at 30° and 22.2% at 35°. The penalty is clearly non-linear: the increment between 30° and 35° is roughly twice that between 25° and 30°.
The cause is the axial thrust that a helical tooth trace generates. As the helix angle grows, so does the axial force component, which raises the contact pressure between the flanks and therefore the frictional heat flux. The rise steepens at large helix angles because the thrust term grows faster than the contact area over which it is distributed, so each additional degree of helix angle costs progressively more in temperature.
Nutation angle: two competing mechanisms
Nutation angle behaves differently from the other two geometric variables because it does not act through a single mechanism. Sweeping ε over 3°, 4°, 5°, 6° and 7° at a fixed 5° baseline produces a non-monotonic response. Relative to 3°, the maximum bulk temperature on the convex flank falls 10.3% at 4° and 23.6% at 5°, then recovers to 12.9% below the 3° value at 6° and only 0.74% below it at 7°.
On one hand, enlarging the nutation angle increases the entrainment velocity at the inlet of the contact, so more lubricant is drawn into the gap between the flanks and more heat is carried away. On the other hand, it increases the curvature at the contact point, which bends the flanks more sharply, raises the contact stress and lowers the convective coefficient — all of which push the temperature up. Which mechanism prevails depends on the operating point. Up to about 5°, the entrainment effect dominates and the tooth runs cooler as ε grows; beyond 5°, the contact-stress and convection penalties take over and the temperature climbs again, so the curve passes through a minimum in the region of the design value. The existence of this minimum means that the nutation angle should not be treated as a free parameter to be increased for greater entrainment without checking the thermal response.
| Parameter | Values examined | Reference value | Change in maximum bulk temperature |
|---|---|---|---|
| Module m (mm) | 2 / 2.25 / 2.5 / 3 / 3.5 | 2 | −2.1% / −3.5% / −7.8% / −10.5% |
| Helix angle β (°) | 18 / 20 / 25 / 30 / 35 | 18 | +4.6% / +7.2% / +11.2% / +22.2% |
| Nutation angle ε (°) | 3 / 4 / 5 / 6 / 7 | 3 | −10.3% / −23.6% / −12.9% / −0.74% |
| Lubricant temperature (°C) | increments of 10 | initial value | +10 °C per +10 °C (linear) |
Lubricant temperature: a one-to-one offset
Lubricant temperature is the one variable that shifts the whole field without reshaping it. Raising the oil temperature by 10 °C raises the bulk temperature of the gear by 10 °C, and the relationship holds in both directions. The reason is that the oil temperature enters every convective boundary condition as the sink temperature $T_f$, while the heat flux and the convective coefficients are fixed by the geometry and the kinematics; for a linear conduction problem the field is therefore linear in $T_f$, with a slope of unity.
This makes oil inlet temperature the most predictable thermal lever available to the designer, and equally the most direct route to overheating if the oil circuit is undersized or the cooler is degraded. Because the offset is one-to-one, a modest rise in supply temperature is transmitted undiminished to the hottest point of the mesh, exactly where scuffing resistance is already lowest.
What the sensitivities mean for design
Ranked by magnitude over the ranges examined, the helix angle is the most aggressive geometric lever — a 35° helix angle raises the peak bulk temperature by more than a fifth relative to 18° — followed by the nutation angle, whose effect is non-monotonic and reaches roughly a quarter of the reference value at its optimum, and then by the module, which offers a steady 2–10% reduction over the 2–3.5 mm range. Lubricant temperature acts as a pure offset with a 1:1 gain and no saturation.
The practical consequences differ accordingly. Increasing the module buys thermal margin through lower contact stress and better convection, but it must be traded against the tooth-number difference that defines the reduction ratio and against the space available in the housing. Increasing the helix angle improves engagement smoothness and load sharing but is thermally expensive, and the penalty accelerates beyond about 30°. Selecting the nutation angle requires the coupled entrainment and contact-stress response to be evaluated rather than assumed, since a value on either side of the minimum at about 5° can raise the peak temperature by a few per cent to more than 20%. Controlling the oil inlet temperature remains the most reliable single measure.
The results apply to a steady-state, single-tooth model of a 45 steel gear pair with the geometry of Table 1, in which rolling friction and elastic–plastic deformation are neglected and the frictional heat is split equally between the two flanks. Under those assumptions the reported percentages describe trends for this class of double circular-arc spiral bevel gear rather than absolute temperatures for any particular reducer, and they provide a basis for setting design parameters and thermal protection thresholds before the more detailed transient and experimental work that a production design requires.