Application context and scope
Spiral bevel gears transmit rotation between intersecting shafts. Their curved teeth provide a large contact ratio, smooth motion, and low noise, which explains their use in automotive axle drives and other compact right-angle transmissions. Five-axis CNC gear milling is the flexible production route: cutter tilt and root-angle adjustment allow flank correction and a wide design space. The cost of such a machine is high, however, and not every spiral bevel gear requires that freedom.
For non-offset spiral bevel gears designed by local synthesis with pre-controlled meshing and cut by a generating process without cutter tilt, the root-angle swing can be omitted. The generating motion can then be carried by two cutter translations, one workpiece-slide translation, and workpiece rotation. This four-axis arrangement reduces machine complexity and cost. Its scope is narrower than that of five-axis machining: it applies to non-offset gears and to cutting setups in which cutter tilt is not required. Generating modes such as SGM, SFM, and SFT use different motion mappings, so the conversion must be derived for the mode in use.
Machine architecture and generating cycle
The machine is built around a cutter box, a workpiece box, a column, three linear slides, and a common bed. The cutter spindle is driven by a variable-frequency motor through a single-stage gear and rotates independently; it is not one of the contouring axes. The workpiece spindle is driven by a servo motor through a worm gear. Its B axis provides the generating rotation and, after each slot, the indexing motion. The X and Y slides move the cutter center, the Z slide moves the workpiece box, and the B axis rotates the blank. The relative position and motion of cutter and blank must follow the generating function throughout the roll.
The operating cycle is:
- Clamp the workpiece and start coolant and chip removal.
- Home the machine.
- Move Z to a safe position.
- Move X and Y to the centering reference.
- Move X and Y to the generating start point while B rotates to its start angle.
- Feed Z to the tooth depth.
- Interpolate X, Y, and B through the generating roll.
- Move X and Y to the generating end point and retract Z.
- Index B, return X and Y to the start point, and repeat until all tooth slots are cut.
The Z coordinate is part of the kinematic solution, although in the validation setup it remains constant over the tabulated generating interval. The cutter speed is set independently and affects cutting speed rather than the generating geometry.
Kinematic equivalence between the CNC machine and the cradle generator
The tooth surface is defined by a cradle-type generating model; the task is to reproduce that motion on the CNC machine. Two transformation chains are compared. In the CNC chain, \(\mathrm{S}_m\) is fixed to the machine, \(\mathrm{S}_t\) to the cutter, and \(\mathrm{S}_w\) to the blank. The cutter rotation is \(C_A\), the blank rotation is \(C_B\), the blank mounting angle is \(C_\gamma\), the cutter-center coordinates are \(C_X\), \(C_Y\), and \(C_Z\), and the axial wheel-position correction is \(\Delta L\). For a point \(\mathbf r_b\) on the blade, the workpiece coordinates are
\mathbf r_w=\mathbf M_C\mathbf r_b,\qquad
\mathbf M_C=\mathbf M_{wa}\mathbf M_{an}\mathbf M_{nm}\mathbf M_{mt}\mathbf M_{tb}.
$$
In the cradle-type generating model, \(\mathrm{S}_p\) is the cradle, \(\phi_g\) is its rotation, \(S_r\) is the radial cutter position, \(q\) is the angular cutter position, \(\delta_M\) is the blank mounting angle, \(E_m\) is the vertical offset, \(X_b\) is the blank offset, \(X_g\) is the horizontal offset, and \(R_b\) is the roll ratio. The same point maps as
\mathbf r_w=\mathbf M_G\mathbf r_b,\qquad
\mathbf M_G=\mathbf M_{wa}\mathbf M_{an}\mathbf M_{nm}\mathbf M_{mp}\mathbf M_{pt}\mathbf M_{tb}.
$$
Each \(\mathbf M\) is a homogeneous transformation between adjacent frames. To reproduce the intended flank, the two representations must coincide throughout the generating motion:
\mathbf M_C=\mathbf M_G.
$$
The rotational part aligns the cutter and blank orientations; the translational part determines the linear axes. The mounting and cutter-phase terms are matched by
C_\gamma=\delta_M,\qquad C_A=\phi_c,\qquad \Delta L=X_g,
$$
where \(\phi_c\) is the cutter phase angle in the generating model. The translational entries of \(\mathbf M_G\) are written \(g_{14}\), \(g_{24}\), and \(g_{34}\). Equating the translational parts gives the axis positions:
\begin{aligned}
C_X &= X_g\cos\delta_M+g_{14}\cos\delta_M-g_{34}\sin\delta_M,\\
C_Y &= g_{24},\\
C_Z &= -X_g\sin\delta_M-g_{14}\sin\delta_M-g_{34}\cos\delta_M,\\
C_B &= \phi_w.
\end{aligned}
$$
The terms \(g_{14}\), \(g_{24}\), and \(g_{34}\) contain the radial and angular cutter positions, the vertical and blank offsets, and the axial correction. The first three equations therefore determine the linear axes directly from the generating parameters, while the fourth sets the workpiece rotation. The B axis is slaved to the cradle rotation through the roll ratio \(R_b\); the tabulated values follow the NC axis definition. No iterative numerical solution is needed for the nominal motion.
Validation case and machining data
The validation case is the right-hand large wheel of a 28/17-tooth spiral bevel gear pair. Table 1 lists the blank data and Table 2 lists the cutting setup. The concave and convex flanks use different cutter radii and blade profile angles; the remaining setup parameters are common to both flanks. The validation setup used zero vertical, horizontal, and blank offsets, while the equations retain these terms for general setups.
| Parameter | Large wheel | Pinion |
|---|---|---|
| Number of teeth | 28 | 17 |
| Module (mm) | 10.3572 | 10.3572 |
| Pressure angle (°) | 20 | 20 |
| Mean spiral angle (°) | 35 | 35 |
| Face width (mm) | 50 | 50 |
| Large-end pitch diameter (mm) | 290 | 176.07 |
| Whole tooth depth (mm) | 17.9487 | 17.9487 |
| Pitch cone angle (°) | 58.7363 | 31.2637 |
| Face cone angle (°) | 62.0664 | 34.5938 |
| Root cone angle (°) | 55.4062 | 27.9336 |
| Parameter | Concave flank | Convex flank |
|---|---|---|
| Cutter radius (mm) | 143.902 | 137.902 |
| Blade profile angle (°) | −18 | +22 |
| Blank mounting angle (°) | 55.4062 | 55.4062 |
| Radial cutter position (mm) | 131.7678 | 131.7678 |
| Angular cutter position (°) | 61.1561 | 61.1561 |
| Horizontal wheel offset (mm) | 0 | 0 |
| Vertical wheel offset (mm) | 0 | 0 |
| Blank offset (mm) | 0 | 0 |
| Roll ratio | 1.167906 | 1.167906 |
Solved motion coordinates
Table 3 gives representative solutions from a 0.15° cradle increment sequence. The B axis follows the cradle rotation through the roll ratio, while X and Y vary continuously through the roll. For this setup, \(C_Z\) remains at −71.325 mm over the listed interval.
| φg (°) | CX (mm) | CY (mm) | CZ (mm) | CB (°) |
|---|---|---|---|---|
| −30.00 | 111.7235 | 79.6475 | −71.325 | 35.0262 |
| −29.85 | 111.5146 | 79.9397 | −71.325 | 34.8510 |
| 30.75 | −14.9016 | 136.3563 | −71.325 | −35.9018 |
| 30.90 | −15.2586 | 136.3563 | −71.325 | −36.0769 |
Virtual cutting and physical trial
A VERICUT model was built from the machine architecture, the cutter geometry, and the blank data. The axis coordinates from the solution were loaded into the virtual NC system, and the generating cycle was simulated. The simulated flank was compared with a theoretical solid generated from the tooth-surface equations. The maximum deviation was less than 0.02 mm. The virtual run also served as a collision and program check before a blank was committed.
The same large-wheel parameters were then used for a physical cutting trial. The measured flank deviated from the theoretical surface because of cutter, fixture, and machine errors, but the deviation remained within the accepted range for the trial. The agreement between the virtual and physical results supports the kinematic model and the axis-position solution for this gear and setup.
What the four-axis result establishes
For the tested 28/17-tooth pair, the four-axis generating method offers a practical route to lower machine cost and simpler structure for non-offset spiral bevel gears that do not require cutter tilt. The kinematic equivalence converts the cradle-type generating motion into explicit X, Y, Z, and B coordinates. The virtual comparison shows a small kinematic deviation, and the physical trial shows an acceptable flank result within the scope of the setup.
The result is not a universal replacement for five-axis machining. Offset hypoid gears, cutter-tilt designs, and other generating modes require their own motion mappings. Within its scope, the four-axis approach also depends on accurate machine calibration, tool setting, fixture location, and synchronization of the B axis with the X and Y slides. Those production conditions determine whether the nominal kinematic accuracy is preserved on the shop floor.