The Development of CNC Plane Grinder: How Flat Surfaces Get Made
A plane grinder removes material with the rim or the face of a rotating abrasive wheel, and the table carries the work past it. This page explains how that arrangement changed over time, which machine suits which part, and where the limits still sit. Read it if you specify flatness, parallelism or surface finish on a print.

In this article
- 1
- 2
- 3
- 4
- 5
- 6
- 7
What the development of CNC plane grinder design actually controls
A plane grinder is a reciprocating machine. The work sits on a magnetic chuck or a fixture on the table, the table strokes left and right, and the wheel head drops a few micrometres per pass. The wheel does not cut like an end mill. It scratches. Each abrasive grain takes a chip measured in micrometres, and the sum of thousands of those scratches is the surface you measure with a profilometer.
That is why the machine controls three things at once. Table position sets where the grind lands. Wheel head infeed sets how deep. Wheel condition sets whether the grain cuts or rubs. Get the third one wrong and the wheel glazes, the temperature climbs, and the part moves.
The early development of CNC plane grinder platforms replaced handwheel infeed with a servo axis on the cross slide. The table still reversed hydraulically on most machines, but the downfeed became a programmed number. Operators stopped counting graduations and started calling a subroutine.
Flatness on a ground surface is rarely a machine accuracy problem alone. It is the sum of wheel wear, coolant delivery, chuck condition and thermal drift in the bed. A machine that holds ±0.005 mm on a 100 mm part can still bow a 600 mm plate if the coolant misses the contact zone.
- 1Infeed axisServo downfeed replaces the handwheel; increments of 1 μm are repeatable.
- 2Table driveHydraulic reversal is giving way to linear motors on newer platforms.
- 3Wheel conditionDressing interval decides whether you cut or burn.
Horizontal spindle, vertical spindle and the rotary table variant
Most flat grinding happens on a horizontal-spindle reciprocating machine with the wheel periphery doing the work. It is the default for plates, blocks, dies and shims. The wheel is narrow, the contact arc is short, and heat has somewhere to go. Depth per pass usually sits between 0.005 mm and 0.03 mm for finishing, with a spark-out pass at zero infeed.
A vertical-spindle machine uses the face of a cup wheel instead. The contact area is much larger, so it removes stock faster and leaves a cross-hatch pattern rather than straight scratches. Use it when flatness matters more than a directional finish, or when you need to clean up a wide face in one setup.
The rotary-table variant feeds parts under the wheel on a continuous circle. It suits small parts in volume: bearing races, spacers, valve plates. Load and unload happen while the wheel keeps cutting, so cycle time drops. The trade-off is that the table diameter caps the part size.
A double-disc grinder is the other branch. Two wheels face each other and grind both sides of a part in one pass. Parallelism comes from the gap between the wheels, not from a chuck, which is why it holds tight thickness control on washers and compressor vanes.
Linear motors, hydrostatics and thermal behaviour
The table drive changed the most over the last two decades. A hydraulic cylinder reverses fast but its speed varies with oil temperature. A ballscrew and servo give a cleaner motion profile. A linear motor removes the screw entirely, so there is no pitch error and no wind-up. Reversal happens in milliseconds and the stroke stays consistent from the first part to the thousandth.
Linear motors bring their own problem. They put heat directly into the table and the guideways. A machine that grinds 4,000 mm long parts will drift unless the bed is cooled. Most builders run chilled coolant through the bed casting, and some add a thermally controlled oil circuit through the ways.
Hydrostatic guideways address stiffness and damping rather than speed. An oil film a few micrometres thick supports the table with almost no friction and absorbs vibration that would otherwise print onto the surface. The cost is a pump, a filtration loop and a warm-up routine before the first cut.
For most shops the practical question is not which drive is theoretically better. It is whether the machine holds size at 7 a.m. and at 4 p.m. on the same day. Ask for a test that runs the same part across a shift and charts the thickness spread.
In-process gauging and closed-loop size control
Older grinders relied on the operator to mike the part and adjust the offset. Modern controls close that loop. An in-process gauge sits against the work or the wheel head, measures as the grind proceeds, and signals the control when the target size is inside a set band. The control then shifts to a fine feed and stops on size.
This matters most on thin parts, where the material springs back after the wheel leaves. A gauge that reads during the cut and again after a dwell can compensate for that deflection. Without it, the operator ends up chasing a moving target and scrapping the last few tenths.
Gap control and crash detection ride on the same servo data. The control watches the load on the spindle motor and the following error on the infeed axis. If the wheel touches a hard spot or a clamped part shifts, the axis backs off before the wheel breaks.
None of this replaces a first-article inspection. A gauge reads one point. Flatness across a 300 mm plate still needs a surface plate, a height gauge or a CMM, and that reading is what tells you whether the machine itself is drifting.
Coolant delivery, filtration and burn control
Grinding puts a lot of energy into a very small contact zone. If the coolant does not reach the arc of contact, the surface burns, temper colours appear, and residual stress changes the part after it cools. High-pressure coolant through the wheel or through nozzles aimed at the contact point is standard on newer machines. Flow rates around 40–80 L/min are common for wider wheels.
Filtration matters as much as pressure. Swarf and abrasive fines recirculate if the filter is coarse, and those particles scratch the finish you just paid for. Paper band, magnetic separator and centrifugal units are the usual choices, often stacked in series.
Oil-based coolant used to be common on plane grinders because it lubricates and resists burning. Most shops moved to water-based fluids for cost and disposal reasons, but oil still appears where the material is difficult: titanium, Inconel, some hardened tool steels.
Coolant temperature should be controlled, not just filtered. A chiller holding the tank within ±1 °C keeps the part and the machine from growing through the shift, which is often the difference between holding ±0.005 mm and holding ±0.015 mm.
Loading, dressing and unattended running
The current push is unattended time. A robot or gantry loader moves blanks onto the chuck and finished parts off, so the wheel keeps cutting through the shift. That only works if the process is stable enough to run without an operator watching every pass.
Automatic wheel dressing is what makes it stable. The control triggers a dress cycle after a set number of parts or a set volume of material removed, and the compensation offset updates in the same motion. Skip that step and the wheel loads up, the spindle load rises, and the size drifts.
CMM feedback closes the outer loop. Measure a sample part, feed the deviation back to the control, and the offset corrects before the next batch. Some shops do this manually once a shift. Others wire it in and let the machine adjust itself.
The limit is part handling, not grinding. Small flat parts with a clean clamping face are easy to automate. Thin plates that need shimming, or parts with interrupted surfaces, still need a person to load them properly. That is where most automation projects stall.
Which plane grinding setup fits the part
Match the machine type to the geometry, the tolerance and the volume.
| Machine type | Best for | Typical flatness | Watch out for |
|---|---|---|---|
| Horizontal spindle reciprocating | Plates, blocks, dies, shims | 0.005 mm per 300 mm | Wheel wear across a long stroke |
| Vertical spindle rotary | Wide faces, flatness first | 0.01 mm per 300 mm | Cross-hatch finish on some prints |
| Rotary table feed | Small parts in volume | 0.005 mm on thin parts | Table diameter caps part size |
| Double disc | Both faces in one pass | Parallelism within 0.005 mm | Needs matched wheels and gap control |
| Creep feed | Deep forms, hard alloys | Profile-dependent | Dresser cost and cycle time |
The honest trade-off
If you need the tightest flatness on long plates, buy a horizontal-spindle machine with a linear-motor table, chilled coolant and in-process gauging. If you are grinding thousands of small flat parts, a rotary-table or double-disc setup will outproduce it at a lower cost per part. Pick by part family, not by spec sheet.
Questions engineers ask about plane grinding
How much stock should I leave for grinding after milling?
Leave 0.2–0.4 mm per face for a normal heat-treated steel part that will be ground after hardening. That gives the wheel enough material to clean up distortion from the heat treat without a long cycle.
If the part is already close to size and only needs a finish pass, 0.05–0.1 mm is workable. Anything under 0.03 mm risks leaving hard spots or a decarburised skin in place.
Can a plane grinder hold ±0.005 mm on a 600 mm long plate?
Yes, but the machine is only part of the answer. The bed has to be thermally stable, the coolant has to reach the contact zone, and the chuck has to be flat and clean. A magnetic chuck with a burr under it will bend the plate more than the machine error.
Ask the shop how they verify flatness on a part that size. A surface plate and height gauge reading across a grid is normal. A single mike reading is not enough.
What surface finish is realistic on a ground flat face?
Ra 0.8–1.6 μm is routine for a finishing pass on steel with a dressed aluminium oxide wheel. Pushing to Ra 0.2–0.8 μm means a finer wheel, a lighter depth of cut and a spark-out pass, and it costs cycle time.
On aluminium, the wheel loads up quickly and the finish degrades. Use a coarser, open-structure wheel and more coolant, and expect to dress more often.
When should I not grind a flat face?
If the flatness callout is looser than about 0.05 mm and the surface finish is not critical, milling or face turning will get you there faster and cheaper. Grinding earns its cost when flatness, parallelism or finish actually drive the function of the part.
Thin, flexible parts are the other case. If the part cannot be held flat on a chuck without deforming, grinding will not fix it. Change the geometry or the fixturing first.
Does the shop need a hardened part to grind it?
No. Grinding works on soft steel too, and it is sometimes used simply to hit a flatness or finish that milling cannot reach. But the economics usually favour grinding only when the part is hardened or the tolerance is tight.
For soft materials, an abrasive process also introduces its own risks: burn, embedded abrasive and residual stress. That is one reason we usually mill soft parts and grind only what needs it.
How do I know the grinder is drifting during a long run?
Run a control part at the start and the end of the shift and measure both. If the thickness spread across the run is wider than the tolerance band, the machine is moving. Common causes are coolant temperature, wheel wear and chuck condition.
A shop that tracks this will have the data. Ask for it before you place a production order on a flatness-critical part.
Send us your flatness callout
Upload the drawing and we will come back with a process route, a tolerance review and a quote within 12 hours.
12-hour quote100% inspectionNDA on request