Spiral bevel gears transmit high torque between intersecting shafts with smooth, low-noise motion because their teeth are curved and gradually tapered. They are common in automotive drivelines and aerospace transmissions. Their flanks are generated surfaces, and the conventional design route makes the pinion a point-conjugate mate of the gear: contact is deliberately localized, and flank modifications control the contact ellipse. That route is rooted in the motion capability of mechanical generators. It limits the instantaneous contact area, so load is concentrated on a small patch of the working flank.
An alternative is full working-flank line contact. If the pinion flank approximates the true conjugate of the gear flank, the pair can touch along a line at each instant, and that line can sweep the full working flank over the meshing cycle. The idea is geometrically attractive but not automatic. The meshing limit line and the undercut limit line must lie outside the working flank, and the pinion addendum must remain thick enough. These conditions are governed by basic blank parameters—spiral angle, working tooth height coefficient, and modification coefficient—so the design problem is one of constrained multi-objective optimization, not merely local flank correction.
Theoretical Line Contact and Machine-Setting Compensation
The line-contact target starts from the conjugate relation between the two flanks. Let the gear flank be $\mathbf{r}_2(u,\theta)$, where $u$ and $\theta$ are surface parameters. The conjugate pinion flank $\mathbf{r}_1$ satisfies
$$
\mathbf{n}\cdot\mathbf{v}_{12}=0,\qquad
\mathbf{r}_2=\mathbf{r}_2(u,\theta),\qquad
\mathbf{r}_1=\mathbf{M}_{12}\mathbf{r}_2,
$$
where $\mathbf{n}$ is the gear-flank normal, $\mathbf{v}_{12}$ is the relative velocity at the mesh point, and $\mathbf{M}_{12}$ transforms coordinates from the gear to the pinion. The first equation is the meshing condition; the second and third generate the theoretical pinion flank that is line-conjugate to the gear flank.
Conventional machining does not produce this theoretical flank directly. It produces a pinion flank $\mathbf{r}_1^{*}$ whose machine settings are calculated from a cutting card, and that flank is point-conjugate to the gear. To approach line contact, the pinion machine settings are corrected so that $\mathbf{r}_1^{*}$ approaches $\mathbf{r}_1$. The correction is computed on a grid over the pinion flank. At grid point $i$, the normal mismatch between the theoretical and machined flanks is
$$
h_i=\left[\mathbf{R}_i-\mathbf{R}_{i1}(\varepsilon_i,\mathbf{d})\right]\cdot\mathbf{n}_{i1}(\varepsilon_i,\mathbf{d}),
$$
where $\mathbf{R}_i$ is the position vector of the grid point on the theoretical flank, $\mathbf{R}_{i1}$ and $\mathbf{n}_{i1}$ are the position and normal vectors on the machined flank, $\varepsilon_i$ is the grid-point coordinate, and $\mathbf{d}$ is the vector of machine settings. The settings are then updated by a least-squares problem,
$$
\min_{\mathbf{d}} f(\mathbf{d})=\frac{1}{2}\mathbf{h}(\mathbf{d})^{\mathsf T}\mathbf{h}(\mathbf{d}),
$$
which is solved by a trust-region Levenberg–Marquardt iteration. The result is a pinion flank that approaches the conjugate flank within a prescribed tolerance. Line contact is therefore a manufacturing target, not only a theoretical construct: the machine settings are adjusted until the normal mismatch over the flank is small enough for contact to spread along a line instead of collapsing to a point.
Basic Parameters as Design Variables
Three basic parameters dominate the global contact topology and are treated as optimization variables:
$$
\mathbf{x}=[\beta,h_k,\xi]^{\mathsf T}.
$$
Here $\beta$ is the spiral angle, $h_k$ is the working tooth height coefficient, and $\xi$ is the modification coefficient. The spiral angle changes the obliquity and length of the contact lines and influences the contact ratio; in this design context, it is kept in a range that gives a contact ratio of about 1.25–1.75. The working tooth height coefficient moves the addendum and root boundaries, and therefore changes where the meshing limit line and the undercut limit line fall relative to the working flank. The modification coefficient changes tooth thickness and addendum thickness, trading contact strength against the risk of a sharp tooth tip.
These variables are coupled. A larger spiral angle or working tooth height can lengthen the contact lines and improve load sharing, but it can also increase sliding and change the lubrication condition. A larger modification coefficient can strengthen the tooth, but it can thin the addendum. The design therefore has competing objectives and active geometric constraints, which is why table-based selection or trial-and-error tuning becomes inefficient.
Constraints: Keeping Limit Lines off the Working Flank
At every instant, the meshing condition $\mathbf{n}\cdot\mathbf{v}_{12}=0$ defines contact. Differentiating this condition with respect to the motion parameter gives the meshing limit condition. In parameter form, the meshing limit line satisfies
$$
\mathbf{r}_2=\mathbf{r}_2(u,\theta,\varphi),\quad
\mathbf{n}\cdot\mathbf{v}_{12}=0,\quad
\Phi_\varphi(u,\theta,\varphi)=0,
$$
where $\varphi$ is the motion parameter and $\Phi_\varphi$ is the derivative of the meshing function with respect to $\varphi$. The meshing limit line separates the gear flank into a contact region and a non-contact region. If it crosses the working flank, part of the designed flank cannot be used for contact.
The undercut limit line comes from the second-order contact condition. It satisfies
$$
\mathbf{r}_2=\mathbf{r}_2(u,\theta,\varphi),\quad
\mathbf{n}\cdot\mathbf{v}_{12}=0,\quad
\psi(u,\theta,\varphi)=0,
$$
where $\psi$ is the second-order condition for the generated surface. If this line lies inside the working flank, the pinion can be undercut or solid interference can occur during meshing. For full-flank contact, the meshing limit line must lie below the root curve of the gear in the axial section, and the undercut limit line must lie above the tip curve.
These geometric requirements are converted into equality constraints by comparing four equally spaced axial points from the toe to the heel. Let $R_{fj}$ be the axial projection of the meshing limit line at point $j$, and $r_{fj}$ the projection of the root curve. Let $R_{aj}$ be the projection of the undercut limit line, and $r_{aj}$ the projection of the tip curve. The constraints are
$$
h_1(\mathbf{x})=\sum_{j=1}^{4}\max\left(R_{fj}-r_{fj},0\right),\qquad
h_2(\mathbf{x})=\sum_{j=1}^{4}\max\left(r_{aj}-R_{aj},0\right).
$$
If all four comparison points lie outside the working flank, the corresponding sum is zero. If any point lies inside, the sum becomes positive and penalizes the design. This gives a simple numerical test for a geometric condition that would otherwise be awkward to impose directly.
A sharp addendum is prevented by a tooth-thickness constraint on the pinion. The single-tooth addendum thickness $s_{a1}$ must satisfy
$$
g(\mathbf{x})=s_{a1}-0.35\frac{m_n}{\cos\beta}\ge 0,
$$
where $m_n$ is the pinion normal module. This inequality keeps the tip from becoming too thin while still allowing the modification coefficient to improve tooth strength.
Objectives: Lubrication and Load Capacity
The first objective represents lubrication. In elastohydrodynamic line contact, the minimum film thickness follows the Hamrock–Dowson form. With lubricant properties and load terms fixed, the film thickness scales as
$$
A_{\min}\propto \alpha^{0.54}(\eta_0 u_0)^{0.7}R^{0.43}E^{-0.03}Q^{-0.13},
$$
where $\alpha$ is the pressure–viscosity coefficient, $\eta_0$ is the dynamic viscosity, $u_0$ is the entrainment velocity, $R$ is the equivalent radius of curvature at the mesh point, $E$ is the equivalent elastic modulus, and $Q$ is the contact load per unit width. The terms that can be changed through flank geometry at fixed operating conditions are $u_0$ and $R$, so they are combined into a composite lubrication parameter:
$$
v_\rho=u_0^{0.7}R^{0.43}.
$$
The entrainment velocity is computed from the surface velocities $\mathbf{v}_1$ and $\mathbf{v}_2$ and the unit vector $\mathbf{g}_{12}$ along the relative velocity:
$$
u_0=\left|\frac{\mathbf{v}_1\cdot\mathbf{g}_{12}+\mathbf{v}_2\cdot\mathbf{g}_{12}}{2}\right|,\qquad
\mathbf{g}_{12}=\frac{\mathbf{v}_{12}}{|\mathbf{v}_{12}|}.
$$
The equivalent radius $R$ is obtained from the local flank geometry and the contact-line normal. A higher $v_\rho$ means a thicker lubricating film at the same load and lubricant condition. Because a single poorly lubricated point can dominate scuffing risk, the objective uses the smallest value over all flank grid points:
$$
\min f_1(\mathbf{x})=\frac{1}{\min_i v_{\rho i}},\qquad i=1,\ldots,n.
$$
The second objective represents load capacity through the total length of the instantaneous contact lines. If $l_t$ is the sum of all contact-line lengths on one tooth at meshing time $t$, a longer total line distributes the normal load over more material. To avoid a design that performs well only at one instant, the model maximizes the smallest total length over the meshing cycle:
$$
\min f_2(\mathbf{x})=\frac{1}{\min_t l_t},\qquad t=0,\ldots,T.
$$
Here $T$ is the single-tooth meshing duration. The reciprocal form converts the two maximization goals into minimization objectives.
Multi-Objective Formulation and Pareto Solution
The constrained problem is
$$
\min \mathbf{F}(\mathbf{x})=(f_1,f_2)
$$
subject to $h_1=0$, $h_2=0$, and $g\ge 0$. The constraints are incorporated by an exterior penalty function:
$$
f_1=\frac{1}{\min_i v_{\rho i}}+\sigma_{st},\qquad
f_2=\frac{1}{\min_t l_t}+\sigma_{st},
$$
$$
\sigma_{st}=\sigma_1\sum_{b=1}^{2}h_b^2(\mathbf{x})+\sigma_2\left[\max\{0,-g(\mathbf{x})\}\right]^2,
$$
with penalty factors $\sigma_1=\sigma_2=1000$. The variables are bounded as
$$
35^\circ\le\beta\le40^\circ,\qquad
1.5\le h_k\le1.8,\qquad
0.2\le\xi\le0.5.
$$
The two objectives conflict. A larger spiral angle or working tooth height can lengthen the contact lines, but the changed sliding and curvature can reduce the minimum lubrication parameter. There is therefore no single best parameter set. A Pareto front is required, followed by a design decision that selects a compromise between lubrication and load sharing.
The model is a composite of meshing equations and discrete flank integrals, so gradient information is expensive to form. A derivative-free Pareto optimizer is appropriate. A bare-bones multi-objective particle swarm optimization algorithm is used to generate an evenly distributed set of non-dominated solutions. The search is stopped after 200 generations; the Pareto front changes little after about 100 generations. A compromise solution is then selected from the front rather than an extreme solution that favors only one objective.
Case Study: A 29/11 Low-to-Medium-Speed Pair
The procedure is illustrated with a spiral bevel gear pair having 29 gear teeth and 11 pinion teeth. The gear is right-hand and the pinion is left-hand; the shaft angle is 90°. The gear convex flank and the pinion concave flank are treated as the working flanks. The gear is machined by the formate method, the pinion by the modified roll method, and the pinion machine settings are compensated so that its flank approaches the gear-conjugate surface.
| Parameter | Pinion | Gear |
|---|---|---|
| Number of teeth | 11 | 29 |
| Hand | Left | Right |
| Shaft angle | 90° | 90° |
| Module | 5 mm | 5 mm |
| Pitch cone angle | 20.77° | 69.23° |
| Outer cone distance | 77.54 mm | 77.54 mm |
| Inner cone distance | 54.54 mm | 54.54 mm |
| Face width | 23 mm | 23 mm |
| Parameter | Value |
|---|---|
| Cutter radius | 76.2 mm |
| Inner blade profile angle | 22° |
| Outer blade profile angle | −18° |
Before optimization, the basic parameters were taken from conventional design tables: $\beta=35.00^\circ$, $h_k=1.65$, and $\xi=0.39$. The optimization returned a compromise set $\beta=35.07^\circ$, $h_k=1.70$, and $\xi=0.48$. The spiral angle changed only slightly, while the working tooth height and modification coefficient increased more noticeably.
| Parameter | Before optimization | After optimization | Change |
|---|---|---|---|
| Spiral angle $\beta$ | 35.00° | 35.07° | +0.07° |
| Working tooth height coefficient $h_k$ | 1.65 | 1.70 | +0.05 |
| Modification coefficient $\xi$ | 0.39 | 0.48 | +0.09 |
| Minimum composite lubrication parameter $v_{\rho,\min}$ | 6.1542 | 6.4248 | +4.4% |
| Minimum total contact-line length $l_{t,\min}$ | 20.3257 mm | 21.3472 mm | +5.0% |
Results: From Interference to Full-Flank Contact
The geometric simulation shows the decisive difference. Before optimization, the undercut limit line crossed the gear working flank. When the gear convex flank enveloped the pinion concave flank, the heel tip of the gear entered the pinion body, producing undercut or solid interference. After optimization, the undercut limit line lay outside the working flank, and both flanks remained outside each other during meshing. The meshing limit line stayed outside the working flank in both cases within the chosen parameter bounds, so the active geometric defect in this example was the undercut limit line rather than the meshing limit line.
The contact-performance metrics improved consistently with the geometric result. The minimum composite lubrication parameter increased from 6.1542 to 6.4248, a gain of 4.4%. The minimum total instantaneous contact-line length increased from 20.3257 mm to 21.3472 mm, a gain of 5.0%. The first result indicates a thicker lubricating film at the weakest flank point; the second indicates a longer contact line at the least favorable meshing instant, which improves load distribution.
A finite-element contact simulation agreed with the geometric prediction. Before optimization, stress concentrated at the heel tip and the mating root because the undercut limit line fell inside the working flank. After optimization, contact spread across the working flank in a line-contact pattern without interference. The simulation therefore supports the central design claim: when the two limit lines are excluded from the working flank and the addendum remains thick enough, the compensated pinion flank can approach full-flank line contact rather than localized point contact.
Engineering Interpretation and Limits
The method changes the role of basic parameters. Instead of being chosen from tables and then corrected locally, they become the variables that set the feasible contact topology. Local flank modification can shape the loaded contact pattern, but it cannot reliably rescue a design whose undercut limit line crosses the working area. Basic-parameter optimization and local flank correction are therefore complementary: the former sets the global geometric envelope, while the latter controls the detailed loaded contact pattern.
The optimization is also practical because it uses a small number of variables and a derivative-free Pareto solver. It can be run before detailed machine-setting calculations, and it returns a target parameter set rather than a single arbitrary choice. The Pareto front makes the trade-off between lubrication and contact-line length visible, so the final selection can reflect whether scuffing risk or load sharing is the more critical constraint.
The result is a design target, not a finished gear. The line-contact condition depends on machine-setting compensation reducing the normal mismatch to a small tolerance. Residual mismatch, cutter wear, heat-treatment distortion, and mounting misalignment can shift the contact and reintroduce edge loading. The present case is a low-to-medium-speed 29/11 pair, and the two objectives are proxies for lubrication and load sharing rather than full tribological or structural predictions. Loaded tooth contact analysis, efficiency measurement, and fatigue testing remain necessary for qualification. Because the design intentionally uses the whole working flank, edge relief and alignment control become more important, not less.
Within those limits, the case demonstrates a clear design logic: compensate the pinion machine settings toward the conjugate flank, exclude the meshing and undercut limit lines from the working area, prevent addendum sharpening, and optimize the basic parameters against lubrication and contact-line length. The optimized pair showed full working-flank line contact without interference and measurable gains in both performance proxies.