CNC screw machine accuracy: how it is lost and how to improve it

• 1
  Key takeaways
• 2
  Why a screw machine holds tolerance in the first place
• 3
  Guide bushing and bar stock: the first place accuracy is lost
• 4
  Thermal drift: the error that shows up after lunch
• 5
  Tool wear, offsets, and the numbers you should record
• 6
  Inspection: proving the tolerance instead of hoping for it
• 7
  Which accuracy lever to pull, and what it costs you
• 8
  When a screw machine is the right answer, and when it is not

On a conventional lathe the part hangs out of the chuck. Cutting force bends it, and the deflection grows with the cube of the unsupported length. A screw machine flips that geometry. The bar is fed through a guide bushing and the tool works within a few millimeters of the bushing face, so the unsupported length stays short and the bending moment stays small. That single arrangement is the reason small-diameter parts can be held at tight tolerance at all.

The second part of the mechanism is simultaneity. On a Swiss-type platform the main spindle, sub-spindle, and up to a dozen slides can work at once. Operations that would be three separate setups on a mill or a turret lathe happen in one handling. Every setup removed is a stack of fixture error removed with it.

The third part is thermal and structural. These machines are small, heavily ribbed, and often sit on a polymer concrete base. Lower mass moving at lower speed generates less heat and less vibration than a large turning center. What is left of the error budget is small enough that thermal drift and tool wear become the visible terms, not machine stiffness.

None of this makes the machine automatic. The control interpolates positions; it does not know that the bar just got 0.02 mm smaller or that the spindle has grown 12 μm since the morning warm-up. Accuracy is a property of the whole loop: bar, guide bushing, tool, coolant, thermal state, and measurement.

A useful way to think about it is an error budget. Start from the drawing tolerance, then subtract bar tolerance, bushing clearance, thermal drift over the run, tool wear per part, and gauge uncertainty. If the remainder is negative, no amount of machine skill will save the job. That arithmetic is worth doing before quoting, not after the first rejection.

• 1
  Short overhangCut close to the guide bushing and the part cannot deflect far.
• 2
  Fewer setupsOne handling means one datum, not three.
• 3
  Low thermal massLess heat generated, less drift to compensate.
• 4
  Closed loop mattersMachine, bar, tool, and gauge all contribute error.

The guide bushing is the heart of a sliding-head machine, and it is also the most common source of a drifting diameter. A carbide bushing running on ground and polished bar with a clearance of roughly 0.005–0.010 mm on diameter behaves well. Run the same bushing on cold-drawn bar with a 0.05 mm diameter spread and the part diameter follows the bar, not the program.

Bar quality is therefore a process input, not a purchasing detail. For tight work we look for centerless-ground bar, straightness within about 0.5 mm per meter, and a diameter tolerance band no wider than a quarter of the part tolerance. If the bar arrives with mill scale or handling dents, the bushing will pick those up and the surface finish will tell you before the gauge does.

Bushing condition matters as much as the setting. A scored or bell-mouthed bushing lets the bar whip. The symptom is a diameter that is good on the first 200 parts and then wanders, often with a taper that follows the direction of the cut. Replacing the bushing mid-run is cheaper than scrapping the second half of the order.

Clearance is a trade. Too little and the bar seizes or gauls the bushing in a few hours. Too much and you lose the support that made the machine worth buying. The right number depends on bar finish, coolant, and material, so it should be recorded per job rather than left to whoever sets up the machine that shift.

• 1
  Ground barDiameter band ≤ 25% of the part tolerance.
• 2
  StraightnessAbout 0.5 mm per meter keeps whip manageable.
• 3
  Bushing clearanceRoughly 0.005–0.010 mm on diameter for ground stock.
• 4
  Replace on scoreA tapered diameter at the end of a run points to the bushing.

A screw machine is a small heat engine. Spindle bearings, ballscrews, servo motors, and the cutting zone all add heat, and the casting grows in response. The tool point does not move far in absolute terms, but on a 6 mm part a 10 μm shift is a large fraction of the tolerance. This is why the first parts of a shift and the parts after a long idle period often read differently from the middle of the run.

The practical fix is boring and effective: run a warm-up cycle. Twenty to thirty minutes of spindle rotation and axis motion before the first cutting pass brings the structure to a stable state. Shops that skip the warm-up and compensate with offsets spend the first hour chasing a moving zero.

For long runs, log the diameter of the first part of each hour. If the trend is monotonic, that is thermal. If it steps, that is a tool change or a new bar. Telling the two apart takes one column in a spreadsheet and saves a lot of argument at the inspection bench.

Coolant temperature is the other half. If the coolant swings with ambient air, both the bar and the spindle housing follow it. Keeping the coolant within a couple of degrees of the room, or the room within a couple of degrees of itself, removes a whole class of slow drift.

Materials with high thermal expansion make this worse. Aluminium 6061 expands about 23 μm per meter per °C. A 10 °C rise across a 100 mm part is already 23 μm, which is five times a ±0.005 mm tolerance if the measurement happens at a different temperature than the cut.

• 1
  Warm up first20–30 minutes before the first part is cut.
• 2
  Log hourlyOne diameter reading per hour separates drift from steps.
• 3
  Control coolantHold it near room temperature, not at ambient swings.
• 4
  Mind expansionAluminium moves ~23 μm/m per °C.

On small bar work most of the cutting is done by carbide inserts and small-diameter end mills in a live tool station. Wear land grows steadily, so the effective cutting edge moves. On an external turning tool the part diameter creeps in one direction; on a boring or drilling operation the hole creeps the other way. Both are predictable, which means both are compensable.

The practical approach is a wear offset schedule tied to part count, not to operator feel. Measure the first part, then every 50 or 100 parts depending on material and surface speed, and apply the offset before the dimension walks out of the control band. For stainless and titanium, where tool wear is faster, the interval shortens. For free-machining brass, a full run may need almost no adjustment.

Tool runout is the other tooling term. A live tool with 0.02 mm of runout will cut an oversize slot no matter what the program says. Indicating the tool on the machine, or using a presetter with a repeatable holder interface, removes that error before it reaches the part.

Surface finish is a useful early warning. When Ra starts climbing from the 0.8–1.6 μm band toward 3.2 μm on the same program and material, the edge is wearing. The diameter may still be in tolerance, but the next hundred parts will not be.

Record the offset history per job. A tool that needed 15 μm of offset over 500 parts last month and needs 40 μm this month is telling you something changed: a different lot of material, a different coating, or a coolant problem.

• 1
  Offset by countEvery 50–100 parts, or by measured trend.
• 2
  Indicate live toolsRunout shows up directly in slot and flat dimensions.
• 3
  Watch RaRising finish usually precedes a diameter shift.
• 4
  Keep a historyOffset per 100 parts is a health check for the process.

A part that measures 9.998 mm in the inspection room and 10.004 mm on the shop floor is not two different parts. It is one part measured two ways. Screw machine work is small enough that gauge choice, clamping force, and temperature dominate the last micrometers. Decide the measurement method before the first article, and write it on the inspection plan.

For diameters in the 1–32 mm range, a micrometer with a friction thimble or a constant-force ratchet beats a caliper. Calipers are fine for a quick check but their repeatability is around 0.02 mm, which is four times a ±0.005 mm tolerance. Using one to accept a tight part is a coin toss with paperwork.

Optical and vision measurement avoids contact deflection entirely and is well suited to small turned features, burr-free edges, and fine pitch threads. A CMM with a small stylus works for position and form, but the probing force can deflect a slender part, so fixturing matters.

Statistical process control is what turns inspection into capability. Plot the diameter across the run, calculate Cpk against the drawing limits, and watch the trend rather than the last reading. A process sitting at Cpk 1.67 with a flat trend is stable. A process at Cpk 1.33 with a rising trend is about to fail.

Finally, agree on the reference temperature. Most drawing tolerances assume 20 °C. If the shop runs at 28 °C and the inspection room is at 20 °C, aluminium parts will not agree by roughly 4 μm on a 25 mm diameter, and the argument will be about the machine when it is really about the thermometer.

• 1
  Micrometer over caliperConstant-force measuring for tight diameters.
• 2
  Vision for small featuresNo contact deflection on slender parts.
• 3
  Track Cpk, not one partTrend and capability tell you what happens next.
• 4
  Fix the temperature20 °C reference, or state the actual condition.