Spiral bevel gears carry high torque with smooth motion, but their curved tooth surfaces are sensitive to micron-scale geometry. A small change in flank topography can shift the contact pattern, alter the load distribution, and change the motion error that excites the driveline. Transmission error is therefore not only a manufacturing defect to be minimized; it is also a design variable. In the unloaded state it is the geometric transmission error (GTE) prescribed by the tooth-surface design. Under torque it becomes the loaded transmission error (LTE), which includes elastic deflection and contact deformation. The practical question is which GTE a loaded mesh needs so that contact remains away from the tooth edges and motion excitation stays low.
A finite-element comparison of a 27-tooth wheel and a 16-tooth pinion illustrates the trade-off. Two tooth-surface designs were generated with GTE levels of 15″ and 30″, while the contact trace angle and contact-ellipse size were held constant. At 200 N·m, the larger-GTE design reduced edge-contact severity and lowered the peak-to-peak LTE amplitude from 52.82″ to 36.72″. It also reduced the RMS displacement and velocity at the pinion shaft bore. The acceleration result was much weaker. These results frame the central point: GTE is useful when it compensates for load-induced contact migration, but its benefit is specific to the load case.
Transmission Error and Edge Contact
Transmission error is the difference between the actual output rotation and the rotation that a perfectly conjugate gear pair would produce. The geometric component is fixed by the tooth-surface design and is calculated by tooth contact analysis (TCA) without load. When torque is applied, the mesh deflects; the wheel lags or leads, and the resulting LTE curve contains the elastic compliance of the teeth and the surrounding structure. The mean LTE reflects the average deflection, while the peak-to-peak amplitude reflects the cyclic motion excitation that drives vibration. A design that lowers the amplitude without creating a stress concentration is preferable, even if its unloaded GTE is larger.
Edge contact occurs when the instantaneous contact ellipse crosses a tooth boundary. The contact stress rises sharply, the effective mesh stiffness changes, and the LTE curve can develop a pronounced fluctuation. In a loaded spiral bevel gear mesh, the contact pattern often migrates toward the toe or heel because the teeth deflect under torque. A prescribed GTE can pre-position the contact so that deflection moves it back toward the center of the flank. This is why GTE should be evaluated together with the load case rather than as an isolated unloaded error.
Tooth-Surface Model and Load Case
The analyzed pair had 16 pinion teeth and 27 wheel teeth, a module of 4.25 mm, a face width of 17 mm, a pressure angle of 20°, a shaft angle of 90°, and a spiral angle of 35° (Table 1).
| Parameter | Value |
|---|---|
| Number of teeth (pinion/wheel) | 16/27 |
| Module | 4.25 mm |
| Face width | 17 mm |
| Pressure angle | 20° |
| Shaft angle | 90° |
| Spiral angle | 35° |
Instead of building a CAD solid and meshing it, the model placed nodes directly on the flank from the tooth-surface equation. The computational grid had 25 nodes along the face width and 9 nodes along the working tooth height. The root transition was generated by a 0.6 mm cutter-tip fillet and was meshed more densely with 6 additional nodes along the height, because the curvature and stress gradient are highest there. The convex and concave flanks were angularly positioned so that a reference point at mid-face on the pitch cone had zero y-coordinate; the flanks were then rotated by the tooth-thickness angle
$$\phi=\frac{\pi}{z}-\frac{j_n}{z m_t}$$
where \(z\) is the number of teeth, \(j_n\) is the normal backlash, and \(m_t\) is the outer transverse module. This preserved the intended tooth thickness and backlash. TCA supplied the relative rotation between pinion and wheel, and the nodes and elements were written directly into an Abaqus input file. The resulting assembly avoided two geometry conversions and retained micron-scale flank topography.
The final mesh used 8-node brick elements (C3D8). A five-tooth sector was selected on each member according to the contact ratio, which reduced the model size while keeping all teeth that can share load. The material was steel with density \(7.9\times10^{-9}\ \mathrm{t/mm^3}\), elastic modulus \(2.1\times10^{5}\ \mathrm{MPa}\), and Poisson’s ratio 0.3. The contact pair included the flanks, root fillets, and tip transitions, with the pinion flank as the master surface and a friction coefficient of 0.06. The pinion was fixed, and the wheel carried 200 N·m about its axis with the rotational degree of freedom free. Two implicit static steps were used: 1 s for stable contact establishment and 5 s for loading.
Loaded Contact Response
Both tooth-surface designs had a contact trace angle of 70° and an instantaneous contact-ellipse major-axis radius equal to 0.3 times the face width. This held the contact-pattern geometry constant so that the comparison isolated the effect of GTE. Under 200 N·m, contact at mesh entry and exit reached the tooth boundary and produced edge contact, with a sharp rise in contact stress. The 15″ pair showed the more severe edge contact; the 30″ pair still experienced edge contact, but to a milder degree. The stress distribution on the working convex flank of the wheel at 0° pinion rotation showed the same ranking. The contact-stress history over one mesh cycle, corresponding to 22.5° of pinion rotation, was formed by extracting the maximum stress at each instant.
| Quantity | GTE 15″ | GTE 30″ |
|---|---|---|
| Mean LTE | 201.66″ | 188.53″ |
| Peak-to-peak LTE | 52.82″ | 36.72″ |
The peak-to-peak LTE of the 15″ pair was 52.82″, 44% larger than the 36.72″ of the 30″ pair; from the other direction, the 30″ design reduced the amplitude by about 30% relative to the 15″ design. The mean LTE was also higher for the 15″ pair (201.66″ versus 188.53″). Since the mean LTE is largely set by elastic deflection, this ranking is consistent with the more severe edge loading and the larger displacement response discussed below. The key result is that the design with the larger unloaded GTE had the smaller loaded fluctuation. Here, GTE acted as a compensation term rather than merely as an error.
Vibration Response
The contact-force history from the loaded contact analysis was used as dynamic excitation. The force was decomposed into axial, radial, and tangential components and applied along the contact trace. A modal superposition analysis then gave the response at a point on the pinion shaft bore. The coordinate system used the pinion axis as X and the wheel axis as Z, with Y orthogonal to both. Root-mean-square (RMS) values are listed in Table 3; the total value is the resultant of the three directional components.
| Quantity | Component | GTE 15″ | GTE 30″ |
|---|---|---|---|
| Displacement (μm) | X | 12.740 | 10.968 |
| Y | 9.512 | 7.853 | |
| Z | 5.402 | 5.051 | |
| Total | 16.792 | 14.404 | |
| Velocity (mm/s) | X | 19.574 | 18.905 |
| Y | 23.721 | 21.421 | |
| Z | 2.023 | 2.209 | |
| Total | 30.820 | 28.655 | |
| Acceleration (m/s²) | X | 70.881 | 72.846 |
| Y | 71.454 | 68.623 | |
| Z | 17.332 | 16.337 | |
| Total | 102.128 | 101.403 |
Displacement and velocity followed the LTE trend. The total RMS displacement fell from 16.792 μm to 14.404 μm, a 14.2% reduction, and the total RMS velocity fell from 30.820 mm/s to 28.655 mm/s, a 7.0% reduction. The Z velocity component was slightly higher for the 30″ design, but the total velocity was lower. The acceleration result was not uniform: total RMS acceleration fell only from 102.128 m/s² to 101.403 m/s², a 0.7% reduction, and the X component was slightly higher for the 30″ design (72.846 versus 70.881 m/s²), while the Y and Z components were lower. The displacement and velocity data therefore support the conclusion that the larger-GTE design was smoother; the acceleration data show only a marginal, component-dependent benefit.
The total displacement ranking agrees with the mean LTE ranking, and the total velocity ranking agrees with the LTE amplitude ranking. This consistency is useful because it links tooth-surface design to a measurable motion response: a lower peak-to-peak LTE corresponded to lower RMS velocity at the shaft bore. The acceleration result is a reminder that a single scalar can hide directional effects and that the vibration benefit should not be overstated.
Design Implications and Scope
The practical lesson is that GTE should be selected together with the load case. A small GTE can produce an acceptable unloaded contact pattern that migrates to the edges under torque; a larger GTE can pre-position the contact so that deflection moves it toward the center of the flank. At 200 N·m, increasing GTE from 15″ to 30″ reduced edge-contact severity, lowered the LTE peak-to-peak amplitude from 52.82″ to 36.72″, and reduced the RMS displacement and velocity. The acceleration benefit was marginal.
This is not a general rule that larger GTE is always better. GTE also affects light-load behavior, where elastic deflection is small; an excessive unloaded motion error can itself become the dominant excitation. The optimum depends on the torque spectrum, misalignment, contact ratio, tooth-flank topology, and manufacturing scatter. The comparison here covers one gear pair, one load, and two GTE levels. It establishes a mechanism and a quantitative trend within that window, not a universal optimum.
For design, the useful workflow is to treat GTE as a controlled variable in TCA, then verify edge contact, contact stress, LTE amplitude, and modal response under the actual load spectrum. Peak-to-peak LTE is a compact metric because it connects tooth-surface design to motion excitation; the vibration RMS values show whether that excitation reaches the shaft. A design should be accepted only when the stress and vibration metrics improve together, since a lower LTE amplitude achieved through a stress concentration at the edge would not be a real improvement.
Within the tested window, the 30″ GTE design was the better compromise at 200 N·m: it accepted a larger unloaded motion error in exchange for less edge contact and lower loaded motion fluctuation. That is the sense in which geometric transmission error can serve as a design lever for spiral bevel gears.