What to do when part lag affects precision when machining a three-axis CNC

• 1
  What part lag actually is
• 2
  Why part lag affects precision when machining on 3 axes
• 3
  Multiple faces and the limits of a three-axis setup
• 4
  Where the shift comes from after a re-clamp
• 5
  How to measure part lag before it becomes scrap
• 6
  Re-datum, re-clamp, or re-fixture: choosing the fix
• 7
  Matching the symptom to the cause and the fix

On a three-axis vertical machining center the workpiece sits in one orientation for the whole program. The control works in machine coordinates. It has no idea where the part edges are unless someone tells it. That assumption is the work offset. If the part moves even slightly between setups, the offset is now wrong and the cut lands in the wrong place.

Operators call this lag because the finished feature seems to trail the model. It is not servo lag and it is not backlash. Servo lag is a following error during acceleration and it is usually small and repeatable. Part lag is a rigid-body displacement of the workpiece. The two feel similar on the shop floor but the fix is completely different.

The error budget has a fixed part and a variable part. Spindle runout, thermal growth, and ball screw pitch error are largely fixed once the machine is warm. Clamping induced shift, chip trapped under a locator, and thermal drift of the part itself are variable. Part lag lives in the variable column. That is why it appears suddenly on a job that ran fine yesterday.

• 1
  Fixed errorsGeometry and wear. Checked by ballbar and laser, corrected by compensation.
• 2
  Variable errorsDatum shift, clamp load, chip entrapment, thermal drift. Checked by probing and indicators.
• 3
  Which one you haveIf the error changes when you re-clamp the same part, it is variable.

A three-axis vertical machine works the top face well. A three-axis horizontal machine works the side faces well. Neither can present a new face to the spindle without the operator moving the part. That single fact is the root of most lag complaints. Parts that need work on four or five sides are being asked to do something the machine was not bought to do.

The common workaround is a second op. Rough the part, re-clamp it on a fixture built off the first-op datums, then finish the remaining faces. This works, but it doubles the number of times the part is touched. Every touch is a chance for a chip, a burr, or a clamp load to move the part by a few microns.

Where the geometry is simple and the tolerance is loose, this is fine. Where a bore and a slot must be coaxial within ±0.02 mm and sit on opposite faces, the stacked error from two setups eats the whole tolerance before the cutter touches metal. In that case the honest answer is not a clever re-clamp. It is a fourth or fifth axis, or a different process.

• 1
  One face, one setupBest accuracy. Datum is set once and never disturbed.
• 2
  Two faces, two setupsAdds clamp and chip risk. Budget the tolerance accordingly.
• 3
  Three or more facesConsider a 4-axis or 5-axis center before fighting the fixture.

Clamping force is the first suspect. A vise tightened to 40 N·m on a thin wall will bow the part. The bow springs back when the vise opens, and the finished face is no longer flat. The same vise at 15 N·m holds the part and leaves the wall straight. Soft jaws bored to the part profile spread the load and reduce this further.

Chip entrapment is the second. A 0.05 mm chip under a locator tilts the part. On a 100 mm span that is roughly 0.05 mm of error at the far edge, which is ten times a ±0.005 mm tolerance. Air blast the locators and wipe them before every load. It sounds trivial until you see the scrap.

Thermal drift is the third. A 200 mm aluminium part warmed 5 °C above the reference grows about 0.023 mm. That is enough to lose a tight bore. Rough in the morning, let the part and fixture reach room temperature, then finish. If the shop has no climate control, measure the part at the same temperature it was cut.

• 1
  Clamp loadUse soft jaws and torque wrenches. Thin walls move first.
• 2
  ChipsBlast and wipe locators. One chip ruins a ±0.005 mm job.
• 3
  HeatLet the part soak back to room temperature before the finish cut.

The tool for the job is a lever dial test indicator with 0.002 mm resolution, mounted on the spindle or on a magnetic base. Sweep the finished reference face and read the total indicated runout. Note the values and the direction. A consistent tilt in one direction points at a clamp or a chip. A random scatter points at thermal drift or a loose locator.

If the machine has a touch probe, use it. Probe three points on the reference face and let the control solve the plane. Compare the plane to the value recorded at the first setup. Anything over 0.01 mm is worth stopping for. The probe routine takes under a minute and it is repeatable, which a hand sweep is not.

For parts with a tight positional callout between two faces, add a check on the feature itself, not just the datum. Probe the bore center and the slot center and compare the distance to the drawing. Datum checks tell you the part moved. Feature checks tell you the part moved in the direction that matters.

• 1
  Sweep the datumLever indicator, 0.002 mm resolution, record direction of tilt.
• 2
  Probe the planeThree points, solve the plane, compare to setup one.
• 3
  Check the featureBore to slot distance is what the drawing actually controls.

If the shift is small, under about 0.02 mm, and the remaining stock allows it, the cheapest fix is to re-datum. Sweep the reference face, update the work offset by the measured amount, and take a light spring pass at 0.1 mm depth. The part stays in the vise. Total time is a few minutes.

If the shift is larger, or the part has already been cut to size on one face, pull it and re-clamp. Clean the locators, seat the part against the stops with light hand pressure, then tighten in a cross pattern to the specified torque. Re-probe before cutting. This costs a setup but saves the feature.

If the shift keeps coming back on the same part number, stop patching and change the fixture. Add a positive stop in the direction of the error. Move from a vise to a dedicated plate with dowel-pin location. On a three-axis center, a better fixture is usually cheaper than a fourth axis, and it solves the problem the operator has been fighting all week.

• 1
  Re-datumShift under 0.02 mm and stock available. Fastest option.
• 2
  Re-clampLarger shift or finished face at risk. Clean and torque properly.
• 3
  Re-fixtureRepeat offenders. Add stops, pins, and a dedicated plate.