How a CNC Forming Wheel Gear Grinding Machine Is Designed

• 1
  Key takeaways
• 2
  How a CNC forming wheel gear grinding machine removes a tooth gap
• 3
  Why the dressing loop decides the final gear quality
• 4
  Heat, burn, and thermal drift on the machine
• 5
  Gear geometry that suits form grinding, and geometry that does not
• 6
  From ground gear to finished assembly: what we control
• 7
  How to measure and qualify a form-ground gear
• 8
  Form grinding against generating grinding: where each one fits

A CNC forming wheel gear grinding machine uses a wheel whose axial cross-section is the mirror image of the tooth space. The wheel plunges into the gap and the flank is produced by the wheel surface itself, not by a rolling motion between tool and workpiece. That single geometric fact drives everything else on the machine: the dressing unit, the axis layout, and the way the controller compensates for wheel wear.

Because the profile is copied rather than generated, the machine needs fewer coordinated axes. Typically the table indexes the tooth, the wheel head feeds radially, and a dressing spindle sits beside the work zone. The controller synchronizes index and feed but does not have to hold a precise roll ratio. On a generating machine, a small error in the roll chain shows up as a profile slope error. Here, the same size of error has to come from the wheel or the dresser.

The trade-off is direct. Copying the profile means the wheel is only as good as its last dressing. It also means the wheel contact length can get long on coarse-pitch gears, which raises heat and spindle load. Generating spreads contact over a shorter arc and rolls it along the flank. Form grinding loads the whole flank at once. That is why the process is comfortable on module 2 to 8 gears and gets harder as the module climbs.

Wheel speed matters more than many operators expect. Vitrified bonded wheels in the 35 to 60 m/s range are common for profile grinding, and the specific removal rate has to stay low enough that the coolant reaches the contact zone. Push the feed per pass and the flank tempers instead of cutting cleanly.

• 1
  Profile is copiedThe wheel shape is the tooth shape, mirrored.
• 2
  Fewer coordinated axesNo roll chain means no roll-chain error.
• 3
  Contact length grows with moduleLonger contact means more heat and more spindle load.

On a CNC forming wheel gear grinding machine, the dressing unit is not a maintenance accessory. It is the accuracy source. A diamond roll or a formed diamond tool is plunged into the wheel to reproduce the tooth space profile, and whatever error lives on that tool is transferred to every flank ground until the next dress. If the dresser wears 5 μm on one flank, that 5 μm appears as profile deviation on the gear.

Dress depth per pass is usually kept small, often 0.01 to 0.03 mm, with several passes to reach the finished profile. Two or three spark-out passes at zero infeed clean up the wheel surface and reduce the effective runout. Skipping spark-out is a common shortcut. It shows up later as a rougher flank and a profile that drifts across the batch, because the wheel face never settles into a stable condition.

Dress interval is set by material removal, not by clock time. A shop running hardened 20MnCr5 at 60 HRC will dress more often than one running soft 4140. The useful rule is to log the number of gear flanks ground per dress against the measured profile deviation. When deviation crosses about 70 percent of the tolerance band, shorten the interval. That single log usually does more for consistency than any machine parameter change.

The dresser itself needs a reference. A rotary diamond roll is trued or replaced on a schedule, and the truing result is checked with the same metrology used for the gear. Without that, wheel profile error and dresser error are indistinguishable, and operators end up chasing the wrong variable for weeks.

• 1
  Dress depth0.01–0.03 mm per pass, plus spark-out passes.
• 2
  Set interval by flank countDress when profile deviation reaches about 70 percent of tolerance.
• 3
  Reference the dresserTruing data keeps wheel error and tool error separate.

Grinding a full flank in one contact patch puts a lot of energy into a small volume. The visible result is a bright flank, but the meaningful result is the subsurface. Tempering and re-hardening show up as a light or dark band after nital etch, and they start well before the gear measures out of tolerance. This is the main reason a form grinding process can pass a size check and still fail a metallurgical one.

Coolant delivery has to match the contact length. Through-spindle coolant at 40 to 80 bar is common on these machines because it reaches under the contact arc where a flood nozzle cannot. Nozzle position is worth checking every shift. A nozzle that has drifted 5 mm off the contact line will not show up in any alarm, but it will show up as burn on one flank.

Thermal drift in the machine structure is the second heat problem. Spindle growth and table growth move the wheel relative to the workpiece over a run of several hundred gears. A warm-up cycle of 20 to 30 minutes, plus a reference part ground and measured at the start of the run, gives you a baseline. Without it, the first parts and the last parts of a shift can sit at opposite ends of the tolerance band.

In-process or post-process gauging closes the loop. Measuring profile and lead on a sample every 20 to 30 parts lets the controller or the operator correct dress compensation before the drift becomes a scrap event.

• 1
  Burn before size error
• 2
  Through-spindle coolant
• 3
  Warm-up and reference part

Form grinding is comfortable when the tooth space is wide enough for a stable wheel and the gear has a shoulder or flange that would block a generating wheel from reaching the full face width. Interference is a common reason to switch. If the generating wheel cannot exit the gap without touching the shoulder, a form wheel that fits inside the space is the practical answer.

Coarse pitch is where the process starts to fight back. As module grows, the contact length along the flank grows with it, and the wheel has to remove more material per pass. The usual response is to reduce feed per pass and accept a longer cycle. On very coarse gears, the cycle time advantage that made form grinding attractive disappears, and generating grinding or hard turning plus a finishing pass becomes worth re-examining.

Thin, long pinions are the other weak spot. A form wheel pushes radially on a slender part, and the deflection can show up as lead error or as a taper across the face. If the length-to-diameter ratio is high, a generating process that spreads the load or a support fixture is the better route. On short, stiff gears the radial push is a non-issue.

Hardened gears above roughly 58 HRC are where the process proves itself, because the alternative is often hard turning with a worse surface finish. Below about 45 HRC, the economics are less clear and a milled and ground combination may be cheaper.

• 1
  Good fitWide gears, shoulder interference, module 2–8, hardened flanks.
• 2
  Poor fitVery coarse module, thin long pinions, soft gears with easy alternatives.

A ground gear usually does not ship on its own. It goes into a housing, a shaft, or a gearbox where bore position, face runout, and mounting features decide whether the mesh is quiet. GreatLight machines those mating features on the same part or on the mating component, which keeps the gear-to-bore relationship inside one tolerance stack instead of two.

We run 127 high-precision CNC machines across three wholly-owned plants, including 16 simultaneous 5-axis machining centers, 12 four-axis mills, 27 three-axis machines, and 16 mill-turn centers. Maximum processing size reaches 4,000 mm, with travels of 4,000 × 400 × 150 mm on the large frame and 750 × 1,150 × 550 mm on the medium frame. That range covers most gearbox housings, shafts, and covers that pair with a ground gear.

Machined mating surfaces are held to ±0.005 mm (±0.0002 in) where the drawing calls for it, with finishes from Ra 0.2–0.8 μm on fine work to Ra 1.6–3.2 μm as-machined. Materials run from 6061 and 7075 aluminium through 303, 316, 17-4PH stainless, 4140 and 4340 steel, and up to TC4 titanium and Inconel when the application needs it.

Every lot is inspected before it ships: raw material check, in-process monitoring, final inspection, with reports on request. Qualification rate sits at 99.99 percent. That matters when a ground gear has to meet a housing bore that was machined in a different process, because the assembly stack is where the quiet running is won or lost.

• 1
  Machine capacity127 CNC machines, 16 five-axis centers, up to 4,000 mm.
• 2
  Tolerance and finish±0.005 mm, finishes from Ra 0.2–0.8 μm.
• 3
  Inspection100 percent before shipment, reports on request.

A form-ground gear is qualified on profile, lead, pitch, and runout, not on a single over-pin dimension. Profile and lead traces show how the flank deviates from the design curve, and on a form-ground part those deviations usually come from the dresser. If the profile trace tilts in the same direction on every tooth, look at the dressing unit before you touch the machine geometry.

Pitch and index error point somewhere else. On a form grinding machine the index is set by the table drive, so uniform pitch error across all teeth suggests an index calibration issue, while random tooth-to-tooth variation suggests clamping or thermal movement. Separating those two patterns is the fastest way to cut troubleshooting time.

Surface finish on the flank should be checked along the profile direction and across the lead. A finish that is acceptable along the profile but rough across the lead usually means the axial feed or the wheel conditioning is off, not the dress. Ra 0.8–1.6 μm is a realistic target for a production form-ground flank, and Ra 0.2–0.8 μm is achievable with a finer dress and spark-out passes.

Keep a first-article record that includes the etch result alongside the dimensional data. When a batch later shows burn or a profile shift, that record tells you whether the process ever had margin or whether it was running at the edge from the start.

• 1
  Profile and leadTilt on every tooth points at the dresser.
• 2
  Pitch and indexUniform error points at the table drive.
• 3
  First-article recordPair the etch result with the dimensional data.