Spiral bevel gears are widely used in aerospace and automotive drivetrains because they combine high load capacity with smooth motion, a high contact ratio, and a contact area that can be controlled through flank design. Their meshing performance depends on the conjugate motion of two curved flanks under a specific set of machine settings and mounting conditions. Tooth contact analysis (TCA) converts that geometry into two practical outputs: the contact pattern, which locates the touching region on the flank, and the transmission error (TE), which measures the departure from a constant angular ratio.
TCA was introduced for hypoid gears by Baxter in 1961 through an unloaded contact model that was also used to examine mounting-error effects. Since then, the method has been extended to loaded contact, error-sensitive contact analysis, ease-off descriptions, finite-element contact models, and dynamic loaded contact analysis. The numerical difficulty is not merely the number of equations. A contact point is defined by two surface parameter pairs and two rigid-body rotation parameters; the flanks are generated implicitly through cutter envelopes; normals, tooth-blank boundaries, and adjacent tooth contacts are coupled. Solving all conditions in a single system tends to be sensitive to initial values and computationally expensive.
A more robust route is to separate the problem into four linked stages: generate the numerical flanks from the cutter envelope, solve an initial contact point with a well-conditioned five-unknown system, sweep the meshing path and truncate it using transmission-error curves, and extract the contact pattern from the local normal gap. The sections below follow that sequence and then examine a reducer gear pair against computations from Gleason CAGE.
Flank Geometry from the Cutter Envelope
For a straight-sided cutter, the cutting surface is a cone. Its position vector and unit normal can be written as
\[
\mathbf F_C(\theta,s)=
\begin{bmatrix}
(R_{\mathrm{tip}}+s\sin\alpha)\cos\theta\\
(R_{\mathrm{tip}}+s\sin\alpha)\sin\theta\\
-s\cos\alpha
\end{bmatrix},
\qquad
\mathbf N_C(\theta)=
\begin{bmatrix}
-\cos\alpha\cos\theta\\
-\cos\alpha\sin\theta\\
-\sin\alpha
\end{bmatrix},
\]
where \(s\) is the height parameter along the cutter flank, \(\theta\) is the rotary parameter of the cone, \(\alpha\) is the pressure angle, and \(R_{\mathrm{tip}}\) is the cutter tip radius. The normal is independent of \(s\), a property that simplifies the envelope condition.
The workpiece transform contains the machine settings: mounting angle, radial cutter position, vertical offset, sliding base, horizontal position, cutter tilt, cutter swivel, and the angular cutter position. Modified roll and helical motion are represented through the workpiece rotation and sliding-base correction,
\[
\phi_P=R_a\left(\phi_C-C\phi_C^2-D\phi_C^3\right),
\qquad
X_{BH}=X_B+H_e\phi_C,
\]
where \(\phi_C\) is the cradle rotation, \(\phi_P\) is the workpiece rotation, \(R_a\) is the roll ratio, \(C\) and \(D\) are modified-roll coefficients, \(X_B\) is the basic sliding base, and \(H_e\) is the helical-motion coefficient. For a forming process, the roll ratio and cradle rotation are set to zero, so the cutter and workpiece have no relative generating motion.
The cutter is moved through the machine coordinate system by a rigid-body transform \(\mathbf M(\phi_C)\), giving the timed position \(\mathbf F(\theta,s,\phi_C)=\mathbf M(\phi_C)\mathbf F_C(\theta,s)\). The envelope condition is
\[
(\mathbf F_\theta\times\mathbf F_s)\cdot\mathbf F_{\phi_C}=0.
\]
Because the normal is independent of \(s\), this condition is linear in \(s\). Solving it gives \(s=s(\theta,\phi_C)\), which reduces the three-parameter cutter motion to a two-parameter numerical flank \(\mathbf F(\theta,\phi_C)\). The pinion and gear use the same construction; their differences enter through the machine settings and modified-roll terms.
Initial Contact with Five Unknowns
At an initial contact point, two flanks contribute four surface parameters, and the two rigid-body rotations contribute two more. The contact conditions provide three coordinate equalities and two independent normal-direction equalities, leaving five equations for six unknowns. Adding an instantaneous-ratio equation can close the system, but it also increases nonlinearity and sensitivity to the initial guess.
A geometric closure avoids that extra equation. The tooth-blank section midpoints of the pinion and gear are first computed and rotated to opposite sides of the \(Z\) axis. The two flanks are then rotated by the same angle magnitude, \(\varpi_P=\varpi_G=\varpi\), so they approach or recede together. This removes one unknown and leaves the five-unknown system
\[
\begin{aligned}
\mathbf F_P(\theta_P,\phi_{C,P},\varpi)&=\mathbf F_G(\theta_G,\phi_{C,G},\varpi),\\
n_{x,P}&=n_{x,G},\qquad n_{y,P}=n_{y,G}.
\end{aligned}
\]
The two normal equations are sufficient because the unit normals already have fixed magnitude; the third component follows up to sign, and the flank orientation fixes that sign. With five unknowns and five equations, the aligned configuration provides a physically meaningful starting point for Newton-type iteration.
Meshing Path and Tooth-Blank Limits
Once the initial point is known, the pinion is rotated in increments and the contact system is solved again at each step. For a pinion with \(P_N\) teeth and a requested number of path points \(N_n\), the angular step is
\[
P_{\mathrm{step}}=\frac{2\pi}{P_NN_n},
\qquad
\varpi_{P,j}=\varpi_{P,0}+jP_{\mathrm{step}}.
\]
Each solve returns the contact point and the corresponding gear rotation. The previous point is a natural initial guess for the next one, which improves convergence along the path.
The theoretical envelope extends beyond the actual tooth blank. A valid contact point must therefore lie inside both blank sections. In axial-radial coordinates \(L=z\) and \(R=\sqrt{x^2+y^2}\), each section is bounded by four lines: addendum, dedendum, heel, and toe. The admissible region satisfies
\[
R\le K_aL+B_a,\qquad
R\ge K_fL+B_f,\qquad
R\le K_eL+B_e,\qquad
R\ge K_iL+B_i.
\]
For the pinion, the contact point is first transformed from the gear frame into the pinion frame using the shaft-angle rotation. If a point falls outside either blank, the sweep is interrupted. This produces a blank-limited contact path and prevents the calculation from continuing over a theoretical flank that does not exist on the finished tooth.
Transmission-Error Truncation and Multi-Tooth Contact
Blank limits alone are not sufficient for the active contact interval, because the transmission-error curves of adjacent tooth pairs overlap. The TE is defined from the difference between the actual gear rotation and the rotation implied by the ratio, referenced to the first point of the sweep:
\[
T’_{e,j}=\left|\varpi_{G,j}-\varpi_{G,n}\right|
-\left|i\left(\varpi_{P,j}-\varpi_{P,n}\right)\right|,
\qquad
T_{e,j}=T’_{e,j}-\max_j\left(T’_{e,j}\right).
\]
Here \(i\) is the transmission ratio. The second expression shifts the curve so that its maximum is zero; the first reference point need not itself lie on the ratio line. Adjacent tooth pairs have the same TE shape shifted by one pitch:
\[
T_{e,-1}=\left(\varpi_{P,j}-\frac{2\pi}{P_N},\,T_{e,j}\right),
\qquad
T_{e,+1}=\left(\varpi_{P,j}+\frac{2\pi}{P_N},\,T_{e,j}\right).
\]
The intersection of adjacent single-tooth TE curves marks the handover between tooth pairs and therefore the boundary of the active contact interval. The curves are treated as ordered piecewise linear segments; an intersection is detected when the endpoints of one segment lie on opposite sides of the other segment’s line. Because the angular step is small, the lower endpoint at the intersection is used as the truncation point. If the curves do not intersect within the blank, the result indicates edge contact, and the contact pattern would expand further under load.
This truncation avoids solving a ten-equation system for two flanks in simultaneous contact. It also ties the contact-path boundary directly to the same TE curve that is used for the transmission-performance evaluation.
Contact Pattern from the Local Normal Gap
The contact pattern is the envelope of instantaneous contact ellipses along the meshing path. At each contact point, the local tangent plane is constructed from the unit normal \(\mathbf n\). With an arbitrary unit vector \(\mathbf e\) not parallel to \(\mathbf n\), an orthonormal basis is
\[
\mathbf a=\frac{\mathbf e\times\mathbf n}{\|\mathbf e\times\mathbf n\|},
\qquad
\mathbf b=\mathbf n\times\mathbf a.
\]
Rays are generated in the tangent plane,
\[
\mathbf R(\theta,k)=\mathbf P+k\left(\mathbf a\cos\theta+\mathbf b\sin\theta\right),
\]
and the corresponding points on the pinion and gear flanks are located for each ray direction. The normal gap at a candidate point is
\[
d_j=\left|\left(\mathbf P_{P,j}-\mathbf P_{G,j}\right)\cdot\mathbf n_{P,j}\right|.
\]
The instantaneous contact boundary is the set of ray endpoints for which \(d_j\) reaches a prescribed separation. In the numerical example, that separation was \(d=0.003\) mm. The pair of boundary points with the largest separation defines the major axis of the contact ellipse. Because the instantaneous ellipse is narrow in a spiral bevel gear contact, the minor axis is generally small and the pattern is represented mainly by its major-axis trace.
This gap search avoids explicit first- and second-order derivatives of an implicit envelope flank, which are difficult to obtain when the meshing equation is not available in closed form. It also produces the contact pattern directly in three-dimensional coordinates rather than only as a two-dimensional projection.
Numerical Example
The procedure was applied to a reducer spiral bevel gear pair. The pinion has 31 teeth and the gear has 38 teeth; both have a face width of 32 mm and an outer cone distance of 208.81 mm. The pitch cone angles are \(23.1667^\circ\) and \(28.8333^\circ\), giving a non-orthogonal arrangement. The machine settings, cutter dimensions, and blank dimensions are listed in Tables 1–3.
| Machine setting | Pinion | Gear |
|---|---|---|
| Mounting angle (°) | 34.34 | 27.73 |
| Roll ratio | 2.35222 | 2.073214 |
| Radial cutter position (mm) | 152.0721 | 163.33767 |
| Vertical wheel position (mm) | 1.95331 | 0 |
| Sliding base (mm) | −37.138 | 2.69674 |
| Horizontal wheel position (mm) | −15.909 | 0 |
| Angular cutter position (°) | 61.19 | 49.81 |
| Spiral correction coefficient | 38.95265 | 0 |
| Cutter tilt angle (°) | 17.46 | 1.84 |
| Cutter swivel angle (°) | 106.62 | 194.81 |
| \(2C\) | 0 | 0 |
| \(6D\) | 0 | 0 |
| Cutter parameter | Pinion | Gear |
|---|---|---|
| Nominal cutter radius (mm) | 301.8666 | 304.8 |
| Point width (mm) | 2.5479 | 1.905 |
| Pressure angle (°) | 17.6 outer / 22.6 inner | 20 outer / 25 inner |
| Modification radius (mm) | 150 outer / 150 inner | 0 outer / 0 inner |
| Blank parameter | Pinion | Gear |
|---|---|---|
| Number of teeth | 31 | 38 |
| Pitch cone angle (°) | 23.1667 | 28.8333 |
| Face cone angle (°) | 24.2667 | 29.6 |
| Root cone angle (°) | 22.4 | 27.7333 |
| Addendum (mm) | 5.09 | 3.44 |
| Dedendum (mm) | 4.65 | 6.30 |
| Face width (mm) | 32 | 32 |
| Outer cone distance (mm) | 208.81 | 208.81 |
After the full-flank sweep, the TE curve was used to truncate the active contact interval. The computed TE variation was 108.6 μrad. The contact pattern was then extracted with a normal-gap threshold of 0.003 mm.
Gleason CAGE was run with its default mid-face contact-pattern setting. CAGE produced a TE variation of 110.3 μrad, a difference of 1.7 μrad, or about 1.5%, from the coupled calculation. The CAGE contact pattern was located 5.43 mm from the large end, 0.23 mm from the addendum, 1.86 mm from the dedendum, and 6.17 mm from the small end. Its position and trend agreed with the calculated pattern, supporting the consistency of the flank geometry, contact-path truncation, and contact-pattern extraction.
Interpretation and Scope
The staged formulation reduces numerical coupling in three places. The equal-rotation closure removes one unknown from the initial contact problem without imposing an artificial instantaneous ratio. TE-curve intersection replaces a simultaneous two-flank contact system with a one-dimensional truncation condition. The normal-gap search replaces explicit principal-curvature derivatives with ray-based distance evaluation on the numerical flanks.
The result remains an unloaded TCA. The 0.003 mm separation is a geometric criterion for rendering the no-load contact pattern, not a prediction of elastic deflection, contact pressure, or tooth-root stress. Loaded contact analysis requires a separate compliance or finite-element model and a load distribution over the same flank geometry. The present comparison is also a single numerical case against commercial software; it establishes consistency for that reducer pair, but it does not by itself quantify accuracy over the full range of machine settings, tooth counts, or mounting errors.
Within those limits, the coupled scheme is useful during design iteration because the contact pattern and TE are derived from the same numerical flank and the same contact path. Changes in cutter parameters, modified roll, or blank geometry propagate consistently to both outputs, making the method suitable for evaluating flank modifications before a loaded contact or experimental study is undertaken.