Spiral bevel gears transmit motion between intersecting shafts through teeth whose flanks are curved along both the profile and the length direction. The curvature lengthens the contact path, spreads load over several teeth, and reduces noise compared with straight bevel teeth, which is why these gears dominate final drives in automotive axles, helicopter transmissions, and heavy industrial reducers. The same curvature makes the flank difficult to describe: it is not a simple analytic surface but the envelope of a cutting tool swept through a prescribed generating motion, so its geometry is tied to machine-tool settings rather than to a closed-form profile equation.
An accurate three-dimensional model of that flank is the entry point for virtual assembly, tooth contact analysis, finite element stress evaluation, and dimensional inspection. Approximating the tooth with a straight-sided or radially ruled surface introduces deviations of the same order as the elastic deflections being studied, which makes such approximations unusable for contact simulation. The alternative is to compute flank points directly from the meshing condition and rebuild the surface in a CAD environment, so that the solid model inherits the geometry of the generating process rather than a visual resemblance to it.
Contact Conditions Imposed at the Meshing Point
Two gear flanks in mesh must touch at a common point and share a common tangent plane there. Let the two flank surfaces be S1 and S2, each carrying its own moving frame σ1 and σ2 with origins O1 and O2, and let m = O1O2 be the position vector between origins. Writing r1, n1 and r2, n2 for the radius and unit normal vectors of the two surfaces, contact and tangency at the instantaneous point M require
$$
\begin{cases}
\mathbf{r}_2 = \mathbf{m} + \mathbf{r}_1 \\
\mathbf{n}_2 = \mathbf{n}_1
\end{cases}
$$
The first line places the two points at the same location in space; the second forces the normals into alignment, which is the tangency condition. Positional coincidence alone is insufficient for a conjugate pair, because the surfaces would either separate or interpenetrate at the next instant. Continuous mesh therefore adds the velocity condition
$$
\mathbf{v}_{12}\cdot\mathbf{n} = 0
$$
where v12 is the relative velocity of the two surfaces at the contact point and n is the common normal. Physically, the relative velocity must have no component along the normal: sliding along the flank is permitted, penetration is not. Both conditions apply whether the pair is designed for line contact or for the localized point contact produced by a slightly mismatched pair, and together they define the equation of meshing used to generate the flank.
Cutter Geometry and the Machine-Tool Coordinate Chain
Generation is described in a chain of frames that reproduces the kinematics of a mechanical gear cutter: a fixed machine frame Sm, a cradle frame Sg that carries the tool spindle, a cutter-head frame St, auxiliary frames Sn and Sq, and the workpiece frame S1 attached to the gear being cut. Each frame is offset and rotated relative to its parent by the machine settings: cradle angle, radial setting, swivel angle, and the vertical and horizontal adjustments of the cutter head.
The cutting edge of a face-milling cutter lies on a cone. In the cutter frame the cone surface and its unit normal are written as
$$
\mathbf{r}_t =
\begin{bmatrix}
(r_p – u_p \sin\alpha_1)\cos\theta_p \\
(r_p – u_p \sin\alpha_1)\sin\theta_p \\
-u_p \cos\alpha_1 \\
1
\end{bmatrix},
\qquad
\mathbf{n}_t =
\begin{bmatrix}
\cos\alpha_1 \cos\theta_p \\
\cos\alpha_1 \sin\theta_p \\
-\sin\alpha_1
\end{bmatrix}
$$
Here rp is the cutter-head radius at the reference point, up the distance from the cutter tip along the blade, α1 the blade pressure angle, and θp the angular position around the cutter axis. The fourth row of the position vector is the homogeneous coordinate that allows the surface to be carried through the transformation matrices. The sign attached to α1 distinguishes the two flanks of the tooth space: the concave side is generated with the negative sign and the convex side with the positive sign, so a single formulation covers both.
Transforming the Cutter Surface into the Wheel Flank
Composing the frame transformations from the cutter to the workpiece gives the family of cutter surfaces in the wheel frame:
$$
\mathbf{r}_2(u_p,\theta_p,\varphi_p) = \mathbf{M}_{2a}\,\mathbf{M}_{am}\,\mathbf{M}_{mp}\,\mathbf{M}_{pe}\,\mathbf{r}_e
$$
in which φp is the cradle rotation parameter and the four matrices encode the successive frame changes. At this stage the surface depends on three parameters, but only two are independent on a generated flank. The third is removed by the equation of meshing, written in the fixed machine frame to keep the velocity vectors simple:
$$
\mathbf{v}_m \cdot \mathbf{n}_m = 0
$$
Solving this scalar equation expresses up as a function of θp and φp, and substituting back into the transformed surface leaves the two-parameter representation r(θp, φp) of the wheel flank. The same procedure applied to the pinion side, with its own machine settings, produces the mating surface. Because the derivation is analytic up to this point, the flank coordinates carry no discretization error; the numerical work begins only when the surface must be sampled.
Discretizing the Flank into a Point Cloud
A generated flank covers only part of the mathematical surface, so the parameter ranges must first be bounded by the tooth boundaries: the tooth tip, the root, the outer and inner ends of the face width, and the two ends of the active profile. The sampling strategy used in the modeling workflow rotates the surface about the gear axis and projects it onto an axial plane, meshes the resulting planar region, and then maps each grid node back to the three-dimensional surface through the correspondence between the projection and the flank. This produces an ordered grid rather than a scattered cloud, which is what a CAD loft requires.
The procedure was demonstrated on a 43-tooth wheel meshing with an 11-tooth pinion, a combination giving a ratio of about 3.9 with a 90° shaft angle. The design parameters are listed in Table 1.
| Parameter | Pinion | Wheel |
|---|---|---|
| Number of teeth | 11 | 43 |
| Pressure angle (°) | 20 | 20 |
| Module (mm) | 6 | 6 |
| Spiral angle (°) | 35 | 35 |
| Shaft angle (°) | 90 | 90 |
| Face width (mm) | 50 | 50 |
| Hand | Left | Right |
Evaluating the flank equations for this pair over a 5 × 9 grid yields 45 points on the convex flank, written out as three-dimensional coordinates in millimetres. Nine of them are reproduced in Table 2; the spacing between successive rows follows the profile direction, and the spacing between columns follows the lengthwise direction across the 50 mm face width.
| Point | x | y | z |
|---|---|---|---|
| 1 | 80.6964 | −5.8813 | 19.2378 |
| 2 | 86.9161 | −4.0991 | 20.6020 |
| 3 | 93.1028 | −1.5095 | 21.9663 |
| 4 | 99.1978 | 1.9679 | 23.3305 |
| 5 | 105.1224 | 6.4409 | 24.6947 |
| 6 | 110.7673 | 12.0592 | 26.0590 |
| 7 | 115.9710 | 19.0431 | 27.4232 |
| 8 | 120.4719 | 27.7490 | 28.7875 |
| 9 | 123.7743 | 38.8516 | 30.1517 |
The gradual increase of the z coordinate along the series reflects the rise of the flank from the inner to the outer end of the face width, while the much larger change in y captures the sweep of the curved tooth. Because the points are ordered, they can be written to text files in groups and read directly by a solid modeler as spline data.
Lofting the Point Grid into a Solid Tooth
Importing the five point groups into SolidWorks as separate curve sets produces the family of lengthwise and profile splines of the convex flank. Lofting these curves, knitting the resulting patches, and extending them to the tooth boundaries closes the surface. Repeating the calculation with the opposite sign in the cutter-cone equation generates the concave flank, so both sides of the tooth space come from the same mathematical route rather than from a mirror or offset operation.
The tooth space is then obtained by cutting the blank with the two flank surfaces. The blank itself is dimensioned from standard spiral bevel gear blank practice: face width, tip and root cone angles, crown height, and fillet radius are set from the design specification, and the root fillet is completed from the blank geometry rather than from the cutter-cone equations, which describe only the active flank. A circular pattern of the finished tooth space around the axis gives the 43-tooth wheel; the same sequence applied to the pinion settings gives the 11-tooth member.
Accuracy, Assumptions, and Scope of the Model
The accuracy of a model built this way is governed by three factors rather than by the solid modeler. The first is the fidelity of the machine settings and cutter geometry: the flank is only as correct as the cradle angle, radial setting, and blade pressure angle entered into the transformation chain. The second is sampling density. A 5 × 9 grid reproduces the global curvature of a 50 mm face width well, but the loft interpolates between splines, so local deviation grows where curvature changes rapidly, near the tooth ends and the tip. Increasing the number of points along both directions is the direct remedy, at the cost of larger data files and heavier surface knitting. The third is the treatment of the root fillet, which follows blank geometry and therefore does not represent the actual cutter tip radius unless that radius is entered consistently.
What the model represents is the theoretical flank of a face-milled gear with the specified machine settings. It contains no lengthwise or profile crowning beyond what the cutter geometry and generating motion produce, no allowance for heat-treatment distortion, and no elastic deflection under load. These omissions are appropriate for its intended use, which is to supply a geometrically faithful solid for finite element meshing, unloaded tooth contact analysis, interference checking in assembly, and as a nominal reference for coordinate measurement of cut or forged teeth. In those applications the value of deriving the flank from the meshing equation is that the discretization error can be made arbitrarily small and separated from the manufacturing error being investigated, so a measured deviation can be attributed to the process rather than to the model.
The approach also transfers directly to related tooth forms. Once the cutter surface and the generating kinematics are expressed as a parameterized surface plus a meshing equation, the same elimination step and the same projection-based sampling apply to profile-modified spiral bevel gears, to hypoid pairs with offset axes, and to circular-arc tooth profiles, where only the blade description and the frame chain change.