CNC System Control of CNC Grinders: Requirements and Development

• 1
  Why the CNC system control of CNC grinders differs from a milling control
• 2
  The core requirements the control has to meet
• 3
  Wheel condition, dressing, and why the control must track both
• 4
  In-process gauging and the closed grinding loop
• 5
  Where grinding control reaches its limits
• 6
  When to specify grinding instead of milling or turning
• 7
  Grinding control vs milling control, by requirement

A milling control moves a tool through space and lets the cutter remove material in discrete bites. A grinding control holds a wheel in contact with a workpiece that is already close to size, then removes the last few hundredths of a millimeter. Contact is continuous. The wheel touches the part the entire time the axis is feeding.

That single difference drives everything else. Continuous contact means the servo loop, the spindle, the coolant, and the machine structure are all coupled through the grinding zone. A tiny deflection shows up as a dimension error, not as a slightly different chip load. The control has to manage forces, not just positions.

Grinding also runs at surface speeds that milling never reaches. A 400 mm vitrified wheel at 35 m/s spins at roughly 1,670 rpm at the rim, and the grit interacts with the workpiece on a microsecond scale. Feed rates are low, often 0.1–5 mm/min in the finishing pass, so the control spends most of its time regulating position at very fine resolution.

The practical result: a general-purpose milling control can be adapted to a grinder, but it will be fighting its own architecture. Grinding-specific controls add the loops, the compensation tables, and the fast I/O that this process needs.

• 1
  Continuous contactThe wheel never leaves the cut during a grinding pass, so force and position are coupled.
• 2
  Fine feed resolutionFinish passes often run at 0.1–5 mm/min, where a 1 μm axis error matters.
• 3
  Wheel changes shapeThe cutting edge is consumed during the pass, so the control must compensate continuously.

Loop stiffness comes first. A grinding axis needs a position loop bandwidth high enough to reject the periodic force variation that comes from wheel runout and grit breakout. If the bandwidth is too low, the axis lags, and the lag shows up as taper or as a wavy surface. Mechanical stiffness and control bandwidth have to be matched; a stiff machine with a slow loop performs worse than a softer machine with a fast one.

Resolution is the second requirement. Most grinding controls interpolate at 0.1 μm or finer and use linear scales rather than rotary encoders on the feed screw. Scales remove backlash and screw pitch error from the loop. On a Ø50 mm bore, a 1 μm scale error is a 2 μm diameter error, which already eats a large share of a ±0.005 mm tolerance.

Thermal management is the third. Grinding puts 70–90% of its energy into the workpiece as heat. Even with flood coolant, the wheelhead, the workhead, and the bed grow at different rates. The control has to compensate for that growth, either by modeling it or by measuring it. Machines that ignore thermal growth drift out of tolerance within the first hour of a run.

The fourth requirement is fast, deterministic I/O. Dressing, gauging, gap elimination, and wheel-head retraction all happen on a millisecond timescale. Standard PLC scan times are too slow. Grinding controls use dedicated high-speed inputs for these signals.

• 1
  Position loop bandwidthMust be high enough to reject force variation from wheel runout.
• 2
  0.1 μm interpolationLinear scales on the feed axis, not rotary encoders.
• 3
  Thermal compensationModeled or measured growth of wheelhead, workhead, and bed.
• 4
  High-speed I/ODressing and gauging signals cannot wait for a standard PLC scan.

A grinding wheel is a cutting tool that wears continuously. As the grit dulls, the wheel pushes instead of cuts. Force rises, heat rises, and the surface finish degrades. The control cannot see this directly, so it relies on a model: spindle power, normal force, or acoustic emission, monitored against a baseline.

Dressing restores the wheel geometry. A single-point diamond traverses across the wheel face at a defined depth, typically 5–20 μm per pass, and the control records the total material removed. That number matters, because every dress changes the effective wheel diameter, and the wheel diameter sets the surface speed.

Surface speed is held constant by raising spindle rpm as the wheel shrinks. A wheel that starts at 400 mm and wears down to 380 mm needs about 5% more rpm to keep the same rim speed. Without that compensation, the finish drifts across a long production run.

Dressing also opens or closes the wheel face. A coarse dress with a fast traverse leaves sharp, open grit that cuts cool and leaves a rougher finish. A slow, fine dress closes the face and produces a finer finish but generates more heat. The control stores these dress recipes as parameters, so the operator selects a finish target rather than a diamond feed rate.

• 1
  Dress depthTypically 5–20 μm per pass with a single-point diamond.
• 2
  Constant surface speedSpindle rpm rises as the wheel diameter shrinks.
• 3
  Dress recipeTraverse speed and depth are stored, not set by hand each time.

The most reliable way to hit ±0.005 mm on a production bore or shaft is to measure the part while it is still in the machine. In-process gauging uses a caliper head that rides on the workpiece, or a probe that touches the surface between passes. The control reads the size and adjusts the remaining stock removal.

A typical cycle runs in three stages. Rough grinding removes most of the stock at a higher infeed rate. A gauging pass measures the actual size and lets the control correct for wheel wear and thermal growth. Finish grinding removes the last few micrometers at a low infeed rate, then the wheel retracts before spark-out ends.

The hard part is the signal path. A gauge that updates every 50 ms is fine for roughing but too slow for the final micrometer. Grinding controls use faster analog or digital inputs, often 1 ms or better, so the retraction command lands before the wheel overshoots. The gap between the gauge reading and the axis stopping must be modeled, not guessed.

Closed-loop gauging adds cost and maintenance. It also adds a failure mode: if the gauge drifts, the control grinds to the wrong size with full confidence. Calibration with a master ring or a setting plug before each run is not optional.

• 1
  Three-stage cycleRough, gauge, finish, with a correction between stages.
• 2
  Fast inputs1 ms or better for the retraction signal on the finish pass.
• 3
  Master calibrationSetting plug or master ring before every production run.

Grinding is a finishing process, not a stock-removal process. If a part needs 2 mm removed from a hardened bore, grinding it directly is slow and expensive. The usual route is to mill or turn the part soft, heat treat it, then grind the last 0.1–0.3 mm. The control is built for that last fraction.

Very deep features are a second limit. A grinding wheel is a disc, and the wheel diameter sets the maximum depth it can reach without hitting the workpiece. Internal grinding spindles are small and flexible, so bore depth-to-diameter ratios above roughly 3:1 start to deflect. The control can compensate, but only up to a point.

Third, grinding needs a part that can be held without distortion. Thin-wall tubes and unsupported plates spring under the wheel force. The control can reduce infeed, but a part that deflects 20 μm under load will not hold a 5 μm tolerance no matter how good the loop is. Fixturing and part geometry set the real limit.

Finally, grinding is a single-point-operation process. It grinds one surface at a time, or a few with a formed wheel. If a part has 12 features with a 0.4 μm finish requirement, grinding all of them is usually slower than milling most and grinding only the critical two.

• 1
  0.1–0.3 mm stockGrind only the material left after heat treatment.
• 2
  3:1 depth-to-diameterInternal grinding spindles deflect beyond this ratio.
• 3
  Thin wallsParts that spring under load cannot hold a 5 μm tolerance.

Pick grinding when the material is hard. Anything above 45 HRC is difficult to mill with acceptable tool life, and the surface finish from a hardened mill is usually Ra 0.8–1.6 μm at best. A grinder holds Ra 0.2–0.8 μm on the same material with a stable wheel.

Pick grinding when the tolerance is tight and the quantity justifies the setup. A ±0.005 mm bore on a hardened steel part is a grinding job. The same bore on a soft aluminum part is a boring job, because the material is easy to cut and the finish requirement can be met with a fine boring pass.

Pick grinding when the surface has a functional requirement. Seal faces, bearing journals, hydraulic spools, and valve seats need a finish and a geometry that milling cannot deliver. The control's ability to hold constant surface speed across the wheel life is what makes that finish repeatable.

Do not pick grinding for prototypes of soft parts. A mill with a fine finishing pass gets close enough for fit checks, and the setup is faster. Save the grinder for the production run, after the geometry is fixed and the material is in its final state.

• 1
  Hardness above 45 HRCGrinding is usually the only practical finishing route.
• 2
  Seal and bearing surfacesFunctional finish requirements that milling cannot repeat.
• 3
  Soft prototypesUse a mill; save grinding for the hardened production part.