Spiral bevel gears are used where high contact ratio, load capacity, and smooth motion must be combined in a compact right-angle drive, as in marine propulsion, precision machinery, and aerospace transmissions. Their meshing quality is commonly assessed through tooth contact analysis (TCA), which predicts two coupled outputs: the contact path on the tooth flank and the transmission error (TE) curve. In an actual assembly, however, bearing and tooth-flank manufacturing deviations, thermal growth, and housing tolerances move the gear pair away from its theoretical mounting position. The resulting misalignment shifts the contact pattern and changes both the symmetry and the fluctuation amplitude of the TE. An error TCA (ETCA) model is therefore needed not only to evaluate sensitivity but also to determine mounting corrections during assembly.
Mounting adjustment as an inverse assembly problem
For a spiral bevel gear pair, the theoretical assembly is defined by the shaft angle α, the gear offset E, the pinion mounting distance P, and the gear mounting distance G. In the usual setting, G is zero and E and P are chosen so that the initial contact occurs near the gear tooth midpoint and the instantaneous ratio at that point equals the nominal tooth ratio. Real assembly introduces offsets Δα, ΔE, ΔP, and ΔG, so the gears run at α + Δα, E + ΔE, P + ΔP, and G + ΔG.
Once the housing is machined, the shaft angle and offset are effectively fixed; Δα and ΔE are not practical adjusters. The pinion and gear mounting distances, however, can be changed with shims. This creates an inverse problem: given a desired contact-path position and a desired TE shape, what values of ΔP and ΔG should be introduced? In this context, “misalignment” is not only an inadvertent error. It is also a controlled adjustment variable used to move the mesh toward a better operating condition.
Tooth surface generation and TCA formulation
The flanks are generated by a flat-top generating wheel on a five-axis gear machine. The straight cutting edge of the cutter head sweeps a generating surface, and the tooth flank is its envelope under cradle rotation. For the gear, a double-sided cutter produces both flanks; for the pinion, separate inner and outer cutters are used, with a modified roll motion. In the cutter frame, the generating surface of the gear can be written as
$$
\mathbf r_g(\mu_g,\theta_g)=
\begin{bmatrix}
(R_g\pm \mu_g\sin\alpha_g)\cos\theta_g\\
(R_g\pm \mu_g\sin\alpha_g)\sin\theta_g\\
-\mu_g\cos\alpha_g
\end{bmatrix}
$$
where μg and θg are surface parameters, Rg is the cutter radius, and αg is the cutter profile angle. Its unit normal follows from
$$
\mathbf n_g(\theta_g)=\frac{\mathbf N_g}{\|\mathbf N_g\|},
\qquad
\mathbf N_g=\frac{\partial \mathbf r_g}{\partial \mu_g}\times
\frac{\partial \mathbf r_g}{\partial \theta_g}
$$
which gives the compact form
$$
\mathbf n_g=
\begin{bmatrix}
\cos\alpha_g\cos\theta_g\\
\cos\alpha_g\sin\theta_g\\
\pm\sin\alpha_g
\end{bmatrix}.
$$
Transforming the cutter surface into the workpiece frame by the homogeneous matrix M2t2 and enforcing the meshing condition Ng · Vg = 0, where Vg is the relative velocity at the cutting point, yields the gear tooth surface and normal:
$$
\mathbf r_2=\mathbf r_2(\theta_g,\phi_{c2}),
\qquad
\mathbf n_2=L_{2t2}\,\mathbf n_g(\theta_g).
$$
The pinion is treated in the same way with its own cutter geometry and modified roll parameters. The resulting surface is r1(θp, φc1), with normal n1 transformed by the corresponding 3 × 3 submatrix of the machine transformation.
The theoretical mounting settings are obtained from the initial contact condition. The gear tooth midpoint M is selected first. Its axial coordinate L and radial distance R from the gear axis satisfy
$$
x_2(\theta_g,\phi_{c2})=L,
\qquad
y_2^2(\theta_g,\phi_{c2})+z_2^2(\theta_g,\phi_{c2})=R^2,
$$
which determines the gear surface parameters at the midpoint. The two tooth surfaces are then represented in a fixed assembly frame Sh. Continuous contact requires equality of position vectors and unit normals:
$$
\mathbf r_1^{(h)}(\theta_p,\phi_{c1},\phi_1)
=
\mathbf r_2^{(h)}(\theta_g,\phi_{c2},\phi_2),
$$
$$
\mathbf n_1^{(h)}(\theta_p,\phi_{c1},\phi_1)
=
\mathbf n_2^{(h)}(\theta_g,\phi_{c2},\phi_2).
$$
At the initial point, the instantaneous ratio must equal the nominal ratio z2/z1. With p1 and p2 as unit vectors along the pinion and gear axes, and l1 and l2 as vectors from the contact point to the axes, this condition is
$$
\frac{\omega_1}{\omega_2}
=
\frac{z_2}{z_1}
=
\frac{\left|\mathbf n_2\cdot(\mathbf p_2\times\mathbf l_2)\right|}
{\left|\mathbf n_1\cdot(\mathbf p_1\times\mathbf l_1)\right|}.
$$
The position and normal equalities provide five independent scalar equations, and the ratio condition provides one more. Together they form six equations in six unknowns: E, P, θp, φc1, φ1, and φ2. Solving this system gives the theoretical mounting adjustments and the initial pinion surface parameters.
With misalignment included, the actual assembly uses α + Δα, E + ΔE, P + ΔP, and G + ΔG. The position and normal equalities are re-established in Sh using the corresponding transformation matrices. Taking the pinion rotation φ1 as the input, five scalar equations are solved for θp, φc1, θg, φc2, and φ2. The contact path is the locus of the contact points on the two flanks. The transmission error is
$$
\Delta\phi_2(\phi_1)
=
\left[\phi_2(\phi_1)-\phi_{20}\right]
–
\frac{z_1}{z_2}
\left[\phi_1-\phi_{10}\right],
$$
where φ10 and φ20 are reference angular positions. This ETCA formulation preserves the actual mounting offsets in the coordinate transformations rather than treating them as a post-hoc correction to the contact pattern.
Feature parameters for contact path and transmission error
Optimization requires a numerical description of the two meshing outputs. For the contact path, let (xi, yi) be the contact points and (xm, ym) the tooth midpoint. The contact-path midpoint is
$$
\bar x=\frac{1}{n}\sum_{i=1}^{n}x_i,
\qquad
\bar y=\frac{1}{n}\sum_{i=1}^{n}y_i.
$$
Its position relative to the tooth midpoint is expressed by
$$
\lambda_x=\frac{\bar x-x_m}{b},
\qquad
\lambda_y=\frac{\bar y-y_m}{h_m},
$$
where b is the face width and hm is the midpoint tooth height. These coefficients define the desired shift: a negative λx moves the unloaded contact path toward the small end, and a negative λy moves it along the tooth height. For light load, a contact path slightly toward the small end is preferred because under load the contact area expands back toward the middle of the flank. This reduces the risk of edge contact and stress concentration.
For the TE curve, two features are used. The first is the peak amplitude ETmax at the intersection of adjacent tooth-mesh curves, which represents the fluctuation amplitude during mesh transfer. The second is the droop deviation
$$
\Delta ET=\left|ET_a-ET_b\right|,
$$
where ETa and ETb are the TE values at the two ends of a single tooth-mesh cycle. A small ΔET indicates a nearly symmetric curve, while a large value indicates that one end of the cycle is more heavily loaded and is more likely to produce edge contact. An ideal geometric TE curve is therefore shallow, downward-curved, and as symmetric as possible.
Constrained optimization of mounting offsets
Using ΔP and ΔG as design variables, four residuals can be defined:
$$
(f_{\mathrm{auto}})_1=\bar x-x_m-\Delta_1,
\qquad
(f_{\mathrm{auto}})_2=\bar y-y_m-\Delta_2,
$$
$$
(f_{\mathrm{auto}})_3=ET_{\max}-\Delta_3,
\qquad
(f_{\mathrm{auto}})_4=\Delta ET-\Delta_4,
$$
where Δ1 = λxb and Δ2 = λyhm are the desired contact-path shifts, and Δ3 and Δ4 are the desired TE amplitude and symmetry targets. Because the tooth-surface equations are strongly nonlinear, solving four objectives simultaneously is not reliable. Instead, one residual is minimized and the other three are converted into constraints:
$$
\min (f_{\mathrm{auto}})_4(x)\big|_{x=(\Delta P,\Delta G)}
$$
$$
\text{s.t.}\qquad
\bar x-x_m\le \Delta_1,\qquad
\bar y-y_m\le \Delta_2,\qquad
ET_{\max}\le \Delta_3.
$$
The solution maps the prescribed contact-path and TE features to a pair of mounting offsets (ΔP*, ΔG*). This is an assembly-level correction: the cutting parameters and tooth geometry remain unchanged, while the mesh position is shifted by the mounting adjustment.
Example: 40/79-tooth SGM spiral bevel gear pair
The procedure was applied to an SGM spiral bevel gear pair with a 40-tooth left-hand pinion and a 79-tooth right-hand gear. The module was 3 mm, the face width 30 mm, the pressure angle 20°, and the shaft angle 90°. The gear was generated by the double-sided method, and the pinion by the modified-roll single-side method with a first-order roll coefficient of −0.05. The cutter radii were 226.6202 mm for the pinion and 227.0462 mm for the gear.
| Parameter | Pinion | Gear |
|---|---|---|
| Number of teeth | 40 | 79 |
| Module | 3 mm | 3 mm |
| Face width | 30 mm | 30 mm |
| Pressure angle | 20° | 20° |
| Pitch cone angle | 26.5651° | 63.4349° |
| Face cone angle | 28.2787° | 64.3804° |
| Root cone angle | 25.6196° | 61.7213° |
| Whole depth | 5.664 mm | 5.664 mm |
| Hand | Left | Right |
The theoretical mounting settings obtained from the initial-contact solution were P = −0.5751 mm and E = 0.4979 mm, with G = 0. At zero misalignment, the TCA contact-path midpoint was (15.59, 2.54) mm, while the tooth midpoint was (15.00, 2.52) mm. The unloaded contact path therefore sat near the large end and toward the addendum. The geometric TE curve had ETmax = 12.0025″ and ΔET = 5.3211″, indicating a pronounced droop and poor symmetry.
With ΔP and ΔG as variables, target position coefficients λx = −0.01 and λy = −0.01, and a prescribed TE amplitude limit, the nonlinear program returned ΔP = 0.0722 mm and ΔG = −0.0795 mm. The optimized contact-path midpoint moved to (13.83, 2.47) mm, a shift of 1.76 mm along the face-width coordinate toward the small end and 0.07 mm along the tooth-height coordinate. Under load, the contact area is expected to expand from this small-end position back toward the middle of the flank, which improves load sharing and reduces the risk of edge contact.
The TE curve also changed. The peak amplitude decreased from 12.0025″ to 11.3376″, about 5.5%, and the droop deviation fell from 5.3211″ to 0.009″. The optimized curve is therefore nearly symmetric. This matters because a symmetric geometric TE curve gives a more uniform transition between adjacent tooth pairs under load, reducing the excitation that can arise from an asymmetric curve and lowering the likelihood of edge contact.
| Quantity | Baseline | Optimized |
|---|---|---|
| Contact-path midpoint x̄ (mm) | 15.59 | 13.83 |
| Contact-path midpoint ȳ (mm) | 2.54 | 2.47 |
| Tooth midpoint (xm, ym) (mm) | (15.00, 2.52) | (15.00, 2.52) |
| ETmax (reported, ″) | 12.0025 | 11.3376 |
| ΔET (reported, ″) | 5.3211 | 0.009 |
| Mounting offsets | 0 | ΔP = 0.0722 mm, ΔG = −0.0795 mm |
Practical interpretation and limits
The example shows that a sub-0.1 mm change in mounting distance can have a large effect on TE symmetry. The contact path is deliberately moved toward the small end at light load, not centered on the tooth midpoint. This is a load-aware choice: as the teeth deflect under torque, the contact area grows toward the middle of the flank, where it can share load over a larger region. The simultaneous reduction of ΔET to nearly zero means that the two ends of the mesh cycle carry more equal TE, which is favorable for smooth transition between tooth pairs and for reducing vibration excitation.
The method is an inverse assembly tool for face-milled spiral bevel gears generated by a flat-top cutter. It does not alter the cutting parameters; it finds the mounting offsets that best reproduce a prescribed contact-path position and TE shape. Because the optimization is based on geometric TCA, it predicts the unloaded contact pattern and geometric TE. Loaded contact analysis or bench testing remains the appropriate verification step when contact stress, bending stress, or actual dynamic response must be assessed. For other manufacturing processes, the same optimization structure can be used, but the tooth-surface generation model and machine transformations must be rebuilt for that process.