Spiral bevel gears transmit motion between intersecting axes through curved teeth that engage gradually along the face width. Their flanks are generated by a rotating cutter head and a timed workpiece motion, so the final tooth surface is determined jointly by cutter geometry, blank position, and machine kinematics. On CNC spiral bevel gear generators, modified roll, helical motion, and vertical motion can be superimposed as additional degrees of freedom. In many production settings, however, the setup calculation stops at the basic machine settings and the first-order helical motion coefficient. The higher-order coefficients remain available but unused, largely because their effect on the generated flank is not quantified in a form that machine adjusters can apply to contact-pattern correction.
The relevant question is not whether higher-order helical motion can change the tooth surface, but how each order changes it: whether the effect is a rigid tilt, a diagonal twist, a curvature change, or a combination; whether the concave and convex flanks respond similarly; and which orders are strong enough to be useful for contact-area adjustment without creating unwanted edge contact. A tooth-surface model that incorporates first- to sixth-order helical motion coefficients provides a direct way to examine these effects.
Generating kinematics and the helical coefficient set
In the generating coordinate system, the cutter head is carried by the cradle, and the workpiece rotates in a timed relationship with the cradle. The cutter surface is a cone; a point on that surface has a normal vector, a tangent vector along the cone generatrix, and position vectors referred to the machine center and to the design crossing point. The generated tooth surface is the envelope conjugate to the cutter surface, obtained from the meshing equation, which supplies the cradle angle, cutter phase angle, and workpiece rotation angle for each contact point.
The helical motion is introduced as a polynomial relation between the workpiece slide position and the cradle-angle increment. With \(H_1\) through \(H_6\) denoting the first- to sixth-order helical motion coefficients, the slide position is
$$X_B=X_{B0}+\sum_{k=1}^{6}(-1)^k H_k\frac{(\Delta q)^k}{k!}$$
where \(X_{B0}\) is the reference slide position and \(\Delta q\) is the cradle-angle increment from the reference setting. The associated axial motion rate, expressed as the first derivative with respect to cradle angle, is
$$v_H=\sum_{k=1}^{6}(-1)^k H_k\frac{(\Delta q)^{k-1}}{(k-1)!}.$$
The first-order term produces a linear change in slide position with cradle rotation. The second- and higher-order terms introduce quadratic and higher-power variations, so they do not merely scale the first-order effect; they reshape the contact geometry along the tooth.
After the meshing solution is obtained, the workpiece tooth-surface point and its normal can be written in vector form as
$$\mathbf r_w=(\mathbf p\cdot\mathbf r_c)\mathbf p+\cos\phi(\mathbf p\times\mathbf r_c)\times\mathbf p+\sin\phi(\mathbf p\times\mathbf r_c),$$
$$\mathbf n_w=(\mathbf p\cdot\mathbf n)\mathbf p+\cos\phi(\mathbf p\times\mathbf n)\times\mathbf p+\sin\phi(\mathbf p\times\mathbf n),$$
where \(\mathbf p\) is the workpiece-axis unit vector, \(\mathbf r_c\) is the cutter-surface position vector referred to the design crossing point, \(\mathbf n\) is the cutter-surface normal, and \(\phi\) is the workpiece rotation angle. The same construction applied to a perturbed coefficient set gives the actual tooth surface under that machine setting.
Normal error on a discrete tooth-surface grid
Tooth-surface error is evaluated on a grid laid out on the rotational projection plane of the theoretical flank. A practical grid uses nine columns along the face width and five rows along the tooth height, giving 45 discrete points. Following the AGMA shrinkage convention, the grid is kept away from the root, tip, and end roundings: the two ends are each inset by 10% of the face width, and the root and tip are each inset by 5% of the full tooth height, with a minimum inset of 0.6 mm. The remaining area is divided uniformly into the chosen number of rows and columns.
For a surface point at radius vector \(\mathbf r_w\) and workpiece axis \(\mathbf p\), the axial and radial coordinates in the projection plane are
$$x=-\mathbf r_w\cdot\mathbf p,\qquad y=|\mathbf r_w\times\mathbf p|.$$
Because \(x\) and \(y\) are functions of the cradle angle and cutter phase angle, the corresponding generating parameters for each grid point are obtained by a two-variable iterative solution. The theoretical radius vector and normal then follow from the surface equations. The same procedure is repeated with the perturbed helical motion coefficients to obtain the actual surface point and normal.
The signed normal error at a grid point is obtained by projecting the difference between the actual and theoretical points onto the theoretical normal:
$$\delta_i=(\mathbf r’_{w,i}-\mathbf r_{w,i})\cdot\mathbf n_{w,i}.$$
On a gear measuring center, the probe path is normally arranged around the center of the flank and approaches each nominal point along the theoretical normal, with an S-shaped path toward the toe and heel. The resulting error map gives both the direction and magnitude of the deviation, in micrometres, over the whole flank.
For sensitivity analysis, the error map is approximated by a second-order surface. With the origin at the grid center, \(X\) along the face width, \(Y\) along the tooth height, and \(Z\) as the normal error,
$$Z=a_1X+a_2Y+a_3XY+a_4X^2+a_5Y^2.$$
The coefficients have direct geometrical meanings: \(a_1\) is the spiral-angle error sensitivity and shifts the contact position along the face width; \(a_2\) is the pressure-angle error sensitivity and shifts the contact position along the tooth height; \(a_3\) is the tooth-surface torsion sensitivity and controls the diagonal character of the contact area; \(a_4\) is the lengthwise curvature sensitivity and affects contact length; \(a_5\) is the profile curvature sensitivity and affects contact width. For \(m\) grid points, the error vector is related to the coefficient vector by
$$\{Z\}=[S]\{a\},$$
and the least-squares solution is
$$\{a\}=([S]^{\mathrm T}[S])^{-1}[S]^{\mathrm T}\{Z\}.$$
Thus each helical motion coefficient perturbation can be represented by a set of five fitted coefficients rather than by a single scalar error value.
Example pair and perturbation set
The example is a 17/17-tooth, 90° shaft-angle spiral bevel gear pair with a 40° mean spiral angle, 3 mm module, 17.5° pressure angle, and 12 mm face width. The wheel is right-hand and the pinion is left-hand. The pinion is the analyzed member and is produced by the duplex helical method. Tables 1 and 2 list the basic gear data and the machine and cutter settings used for the pair.
| Parameter | Wheel | Pinion |
|---|---|---|
| Hand of spiral | Right | Left |
| Number of teeth | 17 | 17 |
| Shaft angle (°) | 90 | 90 |
| Mean spiral angle (°) | 40 | 40 |
| Module (mm) | 3 | 3 |
| Pressure angle (°) | 17.5 | 17.5 |
| Face width (mm) | 12 | 12 |
| Parameter | Wheel | Pinion |
|---|---|---|
| Radial cutter position (mm) | 28.7080 | 28.6176 |
| Horizontal wheel position (mm) | 0 | −0.2618 |
| Vertical wheel position (mm) | 0 | −0.7194 |
| Slide base (mm) | 0 | −1.7016 |
| Cutter tilt angle (°) | 2.58 | 7.42 |
| Cutter swivel angle (°) | 219.31 | 84.06 |
| Blank mounting angle (°) | 37.48 | 35.99 |
| Roll ratio | 1.402053 | 1.411378 |
| Center roll position (°) | 81.13 | 86.87 |
| Nominal cutter diameter (mm) | 76.2 | 77.55 |
| Cutter offset (mm) | 1.25 | 1.16 |
| Inner blade angle (°) | 25 | 25 |
| Outer blade angle (°) | 10 | 10 |
| First-order helical motion coefficient \(H_1\) | 0 | 4.9717 |
Only the pinion coefficients were perturbed. The perturbation set was \(H_1+0.2\), \(H_2+2\), \(H_3+20\), \(H_4+200\), \(H_5+2000\), and \(H_6+20000\); the increment rises by an order of magnitude with coefficient order. Each run changed one coefficient while the remaining machine and cutter settings were held fixed. The resulting normal-error maps were fitted with the second-order surface model to obtain the five sensitivity coefficients for each case.
Fitted sensitivity coefficients
Tables 3 and 4 give the fitted coefficients for the pinion concave and convex flanks. The relative magnitudes within each row indicate which geometrical component dominates for that coefficient perturbation. Because the helical coefficients enter the generating motion at different powers and were perturbed by different amounts, the numbers are comparative descriptors of the resulting topology rather than a simple ranking of machine-axis gains.
| Perturbed coefficient | a1 | a2 | a3 | a4 | a5 |
|---|---|---|---|---|---|
| H1 | 0.0001489 | −0.0008759 | −0.0002049 | 0.0001305 | 0.0001092 |
| H2 | −0.0105070 | 0.0113974 | −0.0005773 | 0.0005052 | 0.0000306 |
| H3 | 0.0094212 | −0.0103516 | 0.0001602 | −0.0002695 | 0.0001505 |
| H4 | −0.0078969 | 0.0084628 | −0.0002366 | 0.0002847 | −0.0000734 |
| H5 | 0.0048817 | −0.0052221 | 0.0000960 | −0.0001506 | 0.0000699 |
| H6 | −0.0023761 | 0.0025240 | −0.0000547 | 0.0000779 | −0.0000299 |
| Perturbed coefficient | a1 | a2 | a3 | a4 | a5 |
|---|---|---|---|---|---|
| H1 | −0.0155950 | 0.0103966 | 0.0002281 | 0.0001788 | −0.0001620 |
| H2 | 0.0511271 | −0.0513838 | −0.0028206 | 0.0001809 | 0.0026612 |
| H3 | 0.0135064 | −0.0154027 | −0.0022695 | 0.0007755 | 0.0015837 |
| H4 | 0.0221120 | −0.0223736 | −0.0019513 | 0.0004423 | 0.0015221 |
| H5 | 0.0080402 | −0.0083500 | −0.0010432 | 0.0003231 | 0.0007344 |
| H6 | 0.0047823 | −0.0048448 | −0.0005228 | 0.0001459 | 0.0003797 |
The concave flank shows a consistent pattern across all six orders: the pressure-angle term \(a_2\) has the largest absolute magnitude, while the profile-curvature term \(a_5\) is the smallest. The spiral-angle and lengthwise-curvature terms are intermediate. For the convex flank, the first-order coefficient is exceptional: \(H_1\) has its largest effect through the spiral-angle term \(a_1\), whereas \(H_2\) through \(H_6\) are pressure-angle dominated, with \(a_2\) largest and \(a_4\) generally smallest. The \(H_2\) run is especially strong: on both flanks its \(a_1\) and \(a_2\) coefficients are of comparable magnitude and opposite sign, producing a pronounced combined tilt rather than a pure shift in one direction.
The convex flank is consistently more sensitive than the concave flank. For example, the largest pressure-angle coefficient in Table 4 is about 0.0514, compared with about 0.0114 in Table 3. A helical coefficient change that visibly corrects the convex flank therefore leaves the concave flank with a much smaller, and possibly insufficient, correction. Separate compensation maps for the two flanks are necessary.
Topological signatures on concave and convex flanks
The fitted coefficients describe average tilts and curvatures, while the full error maps show where those components act. The calculated maps indicate a predominantly diagonal modification: the deviation changes from one corner of the flank to the opposite corner, and the ends of the tooth are affected more strongly than the middle. The sign pattern alternates between odd and even orders.
- Concave flank. Changing \(H_1\), \(H_3\), or \(H_5\) produces one diagonal pattern: the actual surface at the toe tip lies below the theoretical surface, while at the heel root it lies above. Changing \(H_2\), \(H_4\), or \(H_6\) reverses this pattern: the toe tip lies above the theoretical surface and the heel root below it.
- Convex flank. The heel tip is displaced above the theoretical surface for every coefficient order considered. For \(H_1\), \(H_3\), and \(H_5\), the toe root lies below the theoretical surface; for \(H_2\), \(H_4\), and \(H_6\), the toe root lies above it.
- Magnitude distribution. The convex flank responds more strongly than the concave flank, and both flanks respond more strongly near the toe and heel than near the middle of the face width.
These patterns explain why the higher-order coefficients are useful for contact-area work but cannot be treated as independent local knobs. A change intended to raise or lower one corner also alters the tilt and curvature across the rest of the flank. The \(a_3\) term, for example, controls the diagonal orientation of the contact area; the \(a_4\) and \(a_5\) terms change its length and width. In a setup correction, a coefficient that is strong in \(a_2\) will mainly move the contact pattern up or down the tooth height, but it will also introduce some face-width shift and curvature change through the coupled terms.
Validation and use in machine setup
Grinding trials on an H350GH CNC spiral bevel gear grinding machine were used to check the calculated trends. The test gear was ground with the selected helical motion coefficient changes and then measured on a gear measuring center. The measured error map for the \(H_2+2\) condition reproduced the predicted diagonal topology and the predicted corner deviations; the remaining differences were of the order expected from machining and measurement error. The validation therefore supports the use of the vector tooth-surface model and the second-order sensitivity fit for setup planning, at least for the range and coefficient increments considered.
For practical machine adjustment, the results lead to a clear division of labor among the helical coefficients:
- Use \(H_1\) when the dominant requirement is a first-order shift of the contact pattern. On the convex flank it acts mainly through the spiral-angle term; on the concave flank its pressure-angle component is stronger.
- Use \(H_2\) when a strong combined pressure-angle and spiral-angle correction is needed. Its coefficients are the largest in the present set, and its action is distinctly diagonal.
- Use \(H_3\) and \(H_4\) for diagonal and curvature corrections near the tooth ends, where the deviation is largest. They produce stronger torsion and profile-curvature components than \(H_5\) and \(H_6\) under the tested increments.
- Use \(H_5\) and \(H_6\) for finer adjustment after the main tilt has been set. In the tested perturbation set their fitted coefficients are smaller, so they are better suited to trimming the pattern than to making a large global correction.
- Check the concave and convex flanks separately. The convex flank is more sensitive, so a correction sized for the convex side may be insufficient on the concave side; a correction sized for the concave side may overcorrect the convex side.
The practical value of the first- to sixth-order helical motion coefficients lies in this ability to shape the tooth surface diagonally and locally, not merely to shift the whole contact pattern. The coefficients add degrees of freedom that basic machine settings do not provide, but their effects overlap with those of cutter parameters and basic settings. A reliable adjustment therefore starts from a measured error map, fits the five topological coefficients, changes the helical terms in the direction indicated by the dominant terms, and verifies the result by remeasurement before the setting is released to production.