Spiral bevel and hypoid gears carry motion between shafts that intersect or are offset, and their curved teeth allow several teeth to share load while the mesh rolls smoothly. In automotive, aerospace, and mining drives, the difference between a quiet, durable gear set and one that runs hot or fails early is often a contact patch only a few millimetres wide. The manufacturing target is therefore not merely a nominally correct flank: the tooth surface must place a controlled contact ellipse in the middle of the flank and keep it away from the root and the ends.
Conventional development relies on trial cutting. Machine settings are changed, a blank is cut, the contact pattern is checked, and the cycle is repeated. That loop is slow and consumes material. A more controlled route is to treat tooth-surface curvature as the design variable, simulate the cutting process, extract the contact pattern by finite-element analysis, and then correct the curvature through the profile of a grinding wheel. The case discussed here follows that chain for a 43/11 Gleason-type hypoid pair, from virtual cutting to virtual grinding and a rolling check.
Why Tooth-Surface Curvature Sets the Contact Pattern
A gear flank can be described as a parametric surface \(\mathbf r(u,v)\). Its local shape is encoded by the first and second fundamental coefficients, \(E,F,G\) and \(L,M,N\). For a direction on the surface defined by \(\lambda=dv/du\), the normal curvature is
$$k_n(\lambda)=\frac{L+2M\lambda+N\lambda^2}{E+2F\lambda+G\lambda^2}.$$
The two principal curvatures are the extrema of this normal curvature over all directions. With \(H\) as the mean curvature and \(K\) as the Gaussian curvature,
$$H=\frac{EN-2FM+GL}{2(EG-F^2)},\qquad K=\frac{LN-M^2}{EG-F^2},$$
$$k_{1,2}=H\pm\sqrt{H^2-K}.$$
In a gear mesh, the relative principal curvatures of the two flanks, together with the elastic deflection under load, determine the size and orientation of the contact ellipse. If the curvature is too small, the contact line becomes long and the patch spreads across the flank. If the curvature distribution is wrong, the patch moves toward the root or the outer end. Both conditions raise local stress and can increase transmission error, noise, and sensitivity to misalignment.
The correction used here is a micro-geometry change rather than a rigid shift of the tooth. A reference point near the centre of the flank is held fixed, and the surrounding surface nodes are displaced along their local normals by increasing amounts toward the edges. The result is a thin, controlled material-removal field. The grinding wheel profile is the tool-side realization of that field: by changing the wheel section, the grinding pass removes the intended layer and leaves a flank with a different curvature.
Virtual Cutting of the 43/11 Pair
The basic design parameters are summarized in Table 1. The gear has 43 teeth and the pinion 11, with a module of 7.00 mm. The gear is right-hand and the pinion left-hand, the mean pressure angle is 22.5°, and the midpoint helix angles are 31.84° and 50.0°, respectively.
| Parameter | Gear | Pinion |
|---|---|---|
| Number of teeth | 43 | 11 |
| Module | 7.00 mm | 7.00 mm |
| Hand | Right | Left |
| Mean pressure angle | 22.5° | 22.5° |
| Midpoint helix angle | 31.84° | 50.0° |
| Tooth width | 26.92 mm | 34.31 mm |
The gear was generated by the Formate process with a double-sided cutter. In this process the gear blank and cutter have no generating motion during a cut; the tooth space is formed by the cutter profile, and the blank is indexed after each cut. The gear was positioned at an installation angle of 20.4°, with horizontal, vertical, and axial wheel positions of 87.94 mm, 76.59 mm, and 3.63 mm. After each cut the blank was rotated by 8.372°, and 43 cuts produced the complete gear.
The pinion was cut by the cutter-tilt method with a single-sided cutter, so the convex and concave flanks were produced separately. For the convex flank, a 1° cutter rotation was paired with a 4.039° blank rotation, and 49 cutting steps generated the tool-mark lines. For the concave flank, the corresponding roll ratio was 3.4266, and 39 steps completed the surface. The cutter and blank motions were scripted through a Visual Basic interface to CATIA, with Boolean subtraction removing the intersection between tool and blank at each step. The pinion tool marks were then converted into surfaces, joined into a single tooth, and arrayed around the blank.
The resulting model was checked against 45 theoretical points, distributed as 9 points along the tooth height and 5 along the face width. Measured in Pro/E against the CATIA model, the maximum deviation was 0.003 mm on the gear convex flank and 0.004 mm on the concave flank. That level of agreement established the cutting model as a suitable geometric basis for contact analysis.
Finite-Element Contact Before Correction
The assembled pair was simplified before analysis. The shafts were removed, leaving the tooth and root features that participate in mesh. Because the pinion has 11 teeth and only about three pairs share load at a given instant, the gear was trimmed to 12 teeth. This retained a full revolution of pinion mesh while reducing the mesh count and computation time. The pinion flank was set as the master surface and the gear flank as the slave surface. Reference points on the two axes were coupled to the inner rings, and a resisting torque of 500 N·m was applied to the gear to represent service load.
| Item | Setting |
|---|---|
| Material | 20CrMnTi |
| Density | 7860 kg/m3 |
| Elastic modulus | 207,000 MPa |
| Poisson’s ratio | 0.3 |
| Element type | C3D8R |
| Contact friction coefficient | 0.1 |
| Solver | Static, General |
| Applied torque | 500 N·m on the gear |
Contact states were sampled over a meshing cycle at gear rotations of 2°, 4°, 6°, 8°, 10°, 12°, and 14°. Before correction, the contact pattern extended into the root and toward the outer end of the flank, and it formed a long, broad contact line rather than a compact ellipse. In gear terms, the tooth-height curvature was not distributing the load correctly, and the face-width curvature was too low, allowing the contact to stretch along the tooth.
That result defined the correction target: move the patch to the middle of both the tooth height and the face width, and shorten it into an elliptical contact. The existing cutting model was retained; the change would be made in a subsequent grinding pass.
Designing the Grinding Wheel from Node Offsets
Three curvature-correction levels were defined by their maximum normal offsets. The central reference point \(M\) remained fixed, while the surrounding nodes were displaced along the surface normal with a smooth distribution. In the tooth-height direction the offset followed a sine-like profile toward the maximum; in the face-width direction it increased from the large end toward the small end. The third level removed the most material and was selected for virtual grinding and finite-element verification.
| Correction level | Maximum offset along tooth height | Maximum offset along face width | Relative material removal | Use |
|---|---|---|---|---|
| Level 1 (curvature 1) | 0.020 mm | 0.081 mm | Lowest | Design candidate |
| Level 2 (curvature 2) | 0.0215 mm | 0.121 mm | Intermediate | Design candidate |
| Level 3 (curvature 3) | 0.025 mm | 0.161 mm | Highest | Virtual grinding and FE |
The node coordinates and normal vectors were extracted from the gear convex flank. A program then computed the offset of each node for the selected curvature level. Those offsets were applied to the cutter-section coordinates, and the modified section was used to define a bowl-shaped grinding-wheel profile. This is the key process link: the cutting model supplies the base flank, the node offsets define the desired curvature change, and the wheel profile converts that change into a manufacturable grinding pass.
Virtual Grinding and Contact After Correction
The virtual machine was a five-axis FANUC-controlled model of a Gleason Phoenix bevel-gear grinder, built in Vericut. It included the cradle, saddle, workhead, rotary table, and wheel assembly, with three translational axes and two rotary axes representing wheel and workpiece motion. The cutting model was imported, the bowl-shaped wheel was created in the tool library, and a G-code program drove the five-axis form-grinding pass.
After grinding, the gear flank had lost the intended layer of material, the tooth space was wider, and the surface was smoother. The modified gear was reassembled with the pinion and returned to ABAQUS. The resulting contact pattern was compact and elliptical, centred in the tooth-height direction and in the face-width direction. The root and outer-end concentrations of the unmodified model were no longer present.
| Feature | Before correction | After curvature-3 grinding |
|---|---|---|
| Location on flank | Root and outer end | Central in tooth height and face width |
| Shape | Long, broad contact line covering much of the flank | Elliptical patch |
| Surface condition | As-cut flank | Smoother flank; tooth space widened by controlled material removal |
Rolling Check and What the Evidence Supports
A rolling check was used to compare the simulated pattern with physical contact. Red lead was applied to the gear flank, and an electric motor rotated the pair. Where the teeth touched, the red lead was wiped away, exposing the contact region. Before correction, the pattern on the gear was concentrated toward the small end of the flank. After correction, it sat near the middle of the flank and had an elliptical or rectangular shape. The direction of the change agreed with the finite-element result.
The evidence supports a specific conclusion: a grinding-wheel profile derived from normal node offsets can shift a spiral bevel gear contact pattern from root and end contact to a centred ellipse without rebuilding the cutting model. The approach also fits the real process sequence, since cutting is followed by grinding rather than by another trial-cut iteration. It is therefore useful for screening wheel profiles and machine settings before a physical trial.
The validation is mainly geometric and contact-based. The finite-element model uses a simplified pair, a single torque level, and linear elastic material behaviour; the rolling check is qualitative. A full production assessment would also need dynamic transmission-error measurement, wheel wear, thermal distortion, and load sharing over a range of torques. Within those limits, curvature-controlled virtual grinding provides a practical way to connect spiral bevel gear design, finite-element contact analysis, and the grinding process.