Internal and external cylindrical grinding: how the process controls part precision

• 1
  Key takeaways
• 2
  What internal and external cylindrical grinding actually removes
• 3
  External cylindrical grinding: between centers or chucked
• 4
  Internal cylindrical grinding and why bores are harder
• 5
  Coolant, thermal drift, and in-process gauging
• 6
  When grinding is the wrong process
• 7
  How a grinding cycle is set up
• 8
  OD grinding vs ID grinding vs hard turning

Grinding is not a high-volume removal process. On a cylindrical grinder, the wheel takes 0.05–0.5 mm of stock in a normal finishing pass, and most of that disappears in the first few seconds. The rest of the cycle is spark-out: the wheel keeps passing over the same band with the infeed stopped, letting the elastic deflection of the wheel, workpiece, and centers relax. That is why the last 3–8 seconds of a cycle often decide the final diameter.

The abrasive grains act as thousands of tiny cutting edges. Each grain shears a chip a few micrometres thick, and the force per grain is small. The sum of those forces, however, is large enough to push the workpiece away from the wheel. The machine then measures a diameter that is larger than the one it is cutting. A grinder that removes 0.02 mm in one pass while deflecting 0.01 mm is really only cutting half of what the dial says.

This is the engineering reason cylindrical grinding is separated from turning. Turning leaves a helical feed mark and a tear-prone surface layer. Grinding replaces that layer with a shallow, compressive-dominated cut that produces a consistent Ra and a rounder cross-section. Roundness of 1–3 μm is routine on a well-set OD grinder; hitting the same number on a lathe requires a very rigid setup and a careful tool.

The same mechanism applies inside a bore, but the geometry changes the numbers. An internal spindle with a Ø20 mm wheel running at 30,000 rpm has a much smaller contact arc than a Ø400 mm external wheel. Smaller arc means lower force per pass, but also far less stiffness, so the wheel deflects and the bore tends to come out tapered or bell-mouthed.

• 1
  Stock removal0.05–0.5 mm total for finishing; up to 3 mm if the part is soft and the setup is rigid.
• 2
  Spark-out3–8 seconds of zero infeed; skimping here leaves size drift.
• 3
  Typical result±0.005 mm diameter, roundness 1–3 μm, Ra 0.2–0.8 μm.

Between-centers workholding is the default for shafts. The part rotates on two centers, the wheel feeds in radially, and the table traverses along the axis. Stiffness is high because the part is supported at both ends, so this setup holds concentricity across long spans. A Ø50 mm steel shaft 500 mm long is a comfortable job; the same shaft at 1,500 mm needs a steady rest or the middle will spring away from the wheel.

Chucking one end is simpler and works when the part has no center holes, but it cantilevers the workpiece. Deflection grows with the cube of the overhang, so a 100 mm overhang is roughly eight times softer than a 50 mm one. When you see a grinder operator re-dress the wheel halfway through a short batch, it is usually because the wheel loaded up, not because the part moved.

Wheel choice follows the same logic as any abrasive operation. Aluminum oxide (white or pink) suits steels up to about 60 HRC. Silicon carbide suits cast iron and non-ferrous metals. Cubic boron nitride (CBN) holds form on hardened steel and production runs, and it cuts cooler at high wheel speed, but it costs several times more and needs a rigid machine to pay back. Grit 46–60 is a common starting point for general OD work; grit 80–120 when the finish matters more than the removal rate.

Wheel speed for conventional abrasives sits around 30–35 m/s. Pushing to 45–60 m/s with CBN lowers the chip thickness per grain, which reduces burn risk on hardened parts. The trade-off is that the wheel must be balanced to a finer grade, and the coolant nozzle has to keep up.

• 1
  Between centersBest for shafts. High stiffness, good concentricity along the length.
• 2
  ChuckedUse when there are no center holes. Keep overhang short.
• 3
  Wheel speed30–35 m/s conventional; 45–60 m/s CBN on rigid spindles.

Inside a bore, the wheel has to fit. A Ø60 mm bore with a 10 mm wall allows a wheel of roughly Ø40 mm, and that wheel needs a spindle small enough to reach in without touching the far wall. Small spindle means small bearings, and small bearings mean the whole tool is far less stiff than an external wheelhead.

The contact arc is the second problem. Because the wheel is small relative to the bore, it wraps around a large angle of the hole. More grains are in contact at once, so the force per unit width goes up even though the wheel is tiny. The wheel also wears faster, and a worn internal wheel loses its form quickly, which shows up as taper. Operators compensate by dressing more often, sometimes every few parts.

Coolant is harder to deliver. An external nozzle floods the contact zone from above. Inside a bore, the wheel throws coolant out of the hole, and getting fluid to the contact point needs a high-pressure jet aimed through the spindle or a well-placed external nozzle. Poor coolant in an ID grind almost always ends in burn on the bore wall, which is invisible until the part is etched or cut open.

Typical ID results are looser than OD results. A good internal setup holds ±0.01 mm diameter and Ra 0.4–0.8 μm on a hardened steel bore. Holding ±0.005 mm inside a deep bore is possible but needs a rigid quill, a dressed wheel, and a stable thermal state. Let the machine warm up and measure the first part before trusting the offset.

• 1
  Wheel sizeRoughly two-thirds of the bore diameter, limited by quill clearance.
• 2
  DressingMore frequent than OD; a dull ID wheel tapers the bore.
• 3
  Realistic hold±0.01 mm diameter, Ra 0.4–0.8 μm on hardened steel.

Grinding converts a large share of its energy into heat at the contact zone. Most of that heat leaves with the chip, but a fraction goes into the workpiece surface. When the surface passes roughly 600 °C for steel, the structure changes and the surface can re-harden and crack. This is grinding burn, and it does not always show as a blue tint. Light burn is only visible after a nitric acid etch.

Coolant type matters less than delivery. Water-based emulsion at 4–8% concentration is standard. Straight oil cools and lubricates better and is common on CBN and on hard-to-grind alloys, but it needs fire protection and a mist extraction system. Flow rate should be high enough to clear the contact zone, not just wet the part.

Thermal drift is the other quiet error source. A grinder that has been running for an hour is physically different from one that just started. Spindle growth and coolant temperature move the wheel-to-part relationship by several micrometres. Shops that hold tight tolerances let the machine idle through a warm-up cycle and then re-zero the offset before the first production part.

In-process gauging closes the loop. A caliper gauge rides on the part and signals the control when the target size is reached, so the machine stops on the diameter rather than on a position. This is how a grinder holds ±0.005 mm across a batch without touching every part. Final inspection still happens before shipment, but the gauge is what keeps the batch centered.

• 1
  BurnPasses 600 °C on steel and can crack the surface. Etch test reveals it.
• 2
  CoolantEmulsion 4–8%; straight oil for CBN and tough alloys.
• 3
  Warm-upIdle the machine, then re-zero the offset before production.

Grinding is slow and it is a finishing operation, so it earns its place only when the specification demands it. If the print calls for ±0.05 mm and Ra 1.6 μm on a soft aluminum part, a good CNC lathe will hit it in a fraction of the cycle time. Adding a grinding step there raises cost and lead time without improving the part's function.

Thin-wall parts are a poor fit. A Ø80 mm tube with a 2 mm wall will deflect under grinding force, and the wheel will cut a lobed cross-section. The same geometry on a lathe with light finishing passes is more predictable. If grinding is unavoidable, the part needs internal support or a very light infeed with extended spark-out.

Very long, slender shafts also fight the process. Deflection grows with the third power of length, so a shaft that grinds cleanly at 300 mm may need multiple steady rests at 1,200 mm. Each rest adds setup time and a potential witness mark. Sometimes the better answer is to grind in sections or to accept a slightly looser roundness spec.

Interrupted surfaces are another boundary. A keyway or cross-hole crossing the ground diameter produces impact loading on the wheel, which chips abrasive grains and shows up as a random size variation. Operators handle it with a softer wheel grade and slower infeed, but the process never becomes as stable as grinding a continuous surface.

• 1
  Soft, loose-tolerance partsTurn or mill them. Grinding adds cost with no gain.
• 2
  Thin wallsDeflect under wheel force; lobing appears.
• 3
  Interrupted cutsChip the wheel and scatter the size. Expect slower infeed.