Spiral bevel gears transmit motion and torque between intersecting shafts. Their curved teeth engage gradually, giving high load capacity, smooth transmission and a large contact ratio; these properties explain their use in locomotive drives, petroleum equipment and aerospace transmissions. The same flank geometry makes inspection demanding. A spiral bevel tooth is a generated helicoid rather than a surface of revolution, and the acceptance quantities—single-pitch deviation, cumulative pitch deviation, runout and profile deviation—are defined at the midface pitch circle in GB/T 11365-2019. Inspection therefore has to establish both the three-dimensional position of the flank and the correct midface reference.
Industrial practice follows four main routes: single-flank rolling to obtain functional transmission error; individual geometric-error measurement on a gear measuring center; on-machine measurement on a multi-axis machine tool; and non-contact optical measurement. CNC gear measuring centers—including systems from Harbin Measuring & Cutting Tool Plant, Chengdu Tool Research Institute and Harbin Jingda Measuring Instruments—automate bevel-gear measurement, while coordinate measuring machines and five-axis on-machine methods have also been demonstrated. Contact probing remains constrained by the stylus: a finite tip radius may not reach the root of a strongly curved flank, the measuring path must be planned from the gear parameters, and probe-radius compensation and tip wear enter the uncertainty budget.
Optical standoff sensing removes the stylus from the loop. A laser displacement sensor measures the distance to the flank along a known beam; combining that distance with the machine pose gives a surface point without a probe-center offset. The difficulty shifts to optical access: the beam must intersect the flank at a usable angle, and the reflected spot must remain within the receiver window. A two-axis sensor mount can satisfy this condition for the large spiral angles common in spiral bevel gears.
Instrument configuration and scan sequence
The measurement arrangement has three orthogonal linear slides, a rotary worktable and a laser displacement sensor carried on a two-axis bracket. The sensor can be tilted so that the incident beam is normal to the pitch cone and the plane containing the emitted and received light is aligned with the local midface tangent plane. This second adjustment is not cosmetic: at a large spiral angle, a beam that is normal in the pitch-cone sense can still produce a reflected spot outside the receiver if the sensor plane cuts across the helix.
For a scan, the gear is mounted on a high-precision arbor on the rotary table. The sensor is positioned at the midface pitch radius, the bracket angles are set, and the slide coordinates are recorded. Linear gratings on the slides and a circular grating on the table provide the position feedback. The table then rotates at constant speed through one revolution while the system records the sensor standoff and the rotary-table angle at 100 ms intervals. Each surface point follows from the machine pose and the measured standoff along the known beam direction. No stylus radius compensation is required, and the scan does not need a tooth-by-tooth path generated from the gear data; the rotary sweep provides the flank intersections used for the midface deviations.
Generating model of the theoretical flank
Gleason face-milling is a generating process. A cutter disk is mounted eccentrically on a cradle. The cradle rotates, and the cutter disk rotates about its own axis, so the blade edges trace one tooth of an imaginary crown gear. The workpiece rotates in a timed ratio with the cradle, and the envelope of the blade positions cuts the spiral flank. The machine settings—cradle angle q, axial cutter position S, work offset T, bed position P and mounting angle δ—enter the theoretical surface directly.
Let rd be the nominal cutter radius, w the point width, rp0 the tip radius, α the blade pressure angle, θ the cutter rotation angle and u the radial parameter along the blade. The cutter surface in its own frame is
$$
\mathbf{r}_p(\theta,u)=
\begin{bmatrix}
(r_{p0}-u\sin\alpha)\cos\theta\\
(r_{p0}-u\sin\alpha)\sin\theta\\
-u\cos\alpha
\end{bmatrix}.
$$
For the inner flank, rp0 = rd − w/2 and α = α1; for the outer flank, rp0 = rd + w/2 and α = −α2. After transformation to the fixed machine frame, the cutter surface and normal are known functions of q, θ and u. The generating condition is
$$
\mathbf{n}\cdot\mathbf{v}_{12}=0,
\qquad
\mathbf{v}_{12}=\mathbf{v}_{2}-\mathbf{v}_{4},
$$
where v12 is the relative velocity between the imaginary crown gear and the workpiece at the contact point. Solving this condition gives u(q,θ) and hence the imaginary-gear surface rg. Transforming to the workpiece frame gives the theoretical spiral bevel flank and its normal:
$$
\mathbf{r}_4=\mathbf{M}_{4f2}\mathbf{M}_{f21}\mathbf{r}_g,
\qquad
\mathbf{n}_4=\mathbf{L}_{4f2}\mathbf{L}_{f21}\mathbf{n}_g.
$$
Here M denotes a homogeneous transform and L its 3 × 3 rotation submatrix.
At the midface, the bevel tooth profile can be evaluated on an equivalent cylindrical gear. With the face-width coefficient φR = b/Re,
$$
m_m=m_e\left(1-\frac{\phi_R}{2}\right),
\qquad
z_v=\frac{z}{\cos\delta},
\qquad
r_b=\frac{m_m z_v\cos\alpha}{2}
=\frac{m_e(1-\phi_R/2)z\cos\alpha}{2\cos\delta}.
$$
Here me is the large-end module, z the tooth number, δ the pitch cone angle, α the pressure angle and rb the equivalent base-circle radius. The involute in the midface coordinate system is
$$
x_p=r_b\left(\sin u_k-u_k\cos u_k\right),
\qquad
y_p=r_b\left(\cos u_k+u_k\sin u_k\right),
\qquad
u_k=\alpha_k+\theta_k,
$$
where αk is the local pressure angle and θk the involute roll angle. For the standard 20° pressure angle used in the test, the pitch point is at αk = 20°.
Extracting pitch and profile deviations
The sampled points near the pitch circle are fitted with a cubic polynomial,
$$
y=a_0+a_1x+a_2x^2+a_3x^3,
$$
by minimizing
$$
S=\sum_{i=1}^{n}\left[y_i-\left(a_0+a_1x_i+a_2x_i^2+a_3x_i^3\right)\right]^2
$$
over n = 8 points selected close to the pitch line. The resulting curve is intersected with the pitch circle to locate the actual pitch point Ci. Because the fit is local, it smooths point noise without forcing the entire flank to a low-order shape.
For tooth i, let Di(xi,yi) be the theoretical involute–pitch circle intersection. The normal to the theoretical involute at Di meets the fitted actual profile at Ei(x,y). The relative profile deviation is the signed normal distance
$$
\Delta f_{ci}=\pm\sqrt{(x_i-x)^2+(y_i-y)^2},
$$
with a positive sign when Ei lies above the theoretical flank and a negative sign when it lies below. The aggregate fc is the sum of the per-tooth values over the inspected teeth.
Corresponding flanks of adjacent teeth meet the pitch circle at Ci and Ci+2; the two-step index accounts for the two flanks of each tooth. With Dm = mez(1 − φR/2) as the midface pitch diameter, the actual pitch arc is
$$
s_1=\widehat{C_iC_{i+2}}
=D_m\arcsin\!\left[
\frac{\sqrt{(x_{i+2}-x_i)^2+(y_{i+2}-y_i)^2}}{D_m}
\right].
$$
The theoretical pitch is
$$
p=\pi m_m=\pi m_e\left(1-\frac{\phi_R}{2}\right),
$$
so the single-pitch deviation is Δfpt = s1 − p, where s1 is the measured one-pitch arc. For K pitches, FpK = sK − Kp; for a full revolution, Fp = sN − Np, with N the number of pitches. Once the flank–pitch intersections are known, individual tooth thickness, backlash and space width can be calculated from the same data set without a second scan.
Experimental results
The test gear set was declared IT8 and had a 1:1 ratio. Its main parameters are listed in Table 1.
| Parameter | Value |
|---|---|
| Accuracy grade | IT8 |
| Gear ratio | 1:1 |
| Large-end module me | 3 mm |
| Number of teeth z1 = z2 | 30 |
| Cone distance Re1 = Re2 | 63.64 mm |
| Face width b1 = b2 | 20.00 mm |
| Pitch cone angle δ1 = δ2 | 45° |
| Spiral angle β1 = β2 | 35° |
| Pressure angle α | 20° |
After processing one rotary sweep at the midface, the reported deviations were those in Table 2.
| Quantity | Symbol | Value |
|---|---|---|
| Mean single-pitch deviation | fpt,mean | 0.000 829 mm |
| Extreme single-pitch deviations | — | −0.027 88 mm and 0.000 97 mm |
| Total cumulative pitch deviation | Fp | −0.055 57 mm |
| Relative profile deviation | fc | 0.010 93 mm |
According to GB/T 11365-2019, the single-pitch deviation, total cumulative pitch deviation and relative profile deviation all lay in the grade-8 intervals. The test gear had been declared IT8, so the measured result was consistent with its specified accuracy. The difference between the mean single-pitch value and the extreme values also shows why acceptance uses both single-pitch and cumulative quantities rather than an average alone.
Practical scope and limitations
The main advantage of the laser method is mechanical simplicity at the scan stage. A rotary sweep and a fixed sensor pose replace a planned stylus path; there is no tip-radius compensation and no wear at the contact. The two-axis mount allows the beam to follow the pitch cone and the local helix, so the same arrangement can be adapted to different cone angles and spiral angles. One data set yields pitch, profile, tooth thickness, backlash and space width at the midface.
The method does not, by itself, produce a full topographic map of the entire flank. The deviations discussed here are defined on the midface pitch circle, and the theoretical reference comes from the generating model. Full flank-form analysis, root stress assessment or contact-pattern prediction would require additional scans at other cone distances and a validated model of the actual cutter and machine settings. Laser standoff also depends on surface reflectivity and local slope; polished or coated flanks may need signal optimization. A traceable uncertainty budget would be required before the method could serve as a calibration-grade instrument.
Within that scope, the approach is useful for production inspection: it reduces fixture and path-planning time, keeps the sensor out of contact with the flank, and converts the measured intersections into the midface quantities used by GB/T 11365-2019. The same logic—known machine pose, optical standoff and a generating model for the reference surface—can be extended to other complex helical surfaces when optical access and surface response are adequate.