CNC Machining Unstable: Why Part Size Drifts Between Passes

• 1
  What CNC Machining Unstable Actually Means
• 2
  Loose Coupling, Backlash, and Guideway Drag
• 3
  Thermal Growth: the Slowest Failure Mode
• 4
  Servo Tuning and Position Error
• 5
  Tool Deflection and Fixture Effects
• 6
  When the Process Is the Wrong Tool
• 7
  A Step by Step Check for CNC Machining Unstable
• 8
  Size Pattern to Probable Cause

A stable process holds its size. Cut a bore at 12.000 mm on part one and part fifty should land within a few thousandths of that. When CNC machining unstable, the number walks. It drifts up over a run, or swings back and forth between batches with no change to the program.

Drift of a few microns is normal and comes from tool wear and thermal growth. Drift of 0.05 mm or more, or a size that changes direction, is a fault. The two behave differently and need different fixes, so measure the pattern before touching the machine.

The first question on any unstable process: does the size move in one direction, or does it jump? One direction points at wear, heat, or a loose clamp. Jumping points at backlash, lost motion, or a signal fault in the servo loop.

• 1
  One-directional driftTool wear, spindle growth, thermal expansion
• 2
  Two-directional swingBacklash, lost motion, loose coupling
• 3
  Random scatterFixture slip, chip recutting, servo tuning

The classic fault on a lathe or mill is a loose coupling between the servo motor shaft and the lead screw. The motor moves, the screw lags, and the tool reaches its position a fraction late. Because the axis is still catching up when the cut starts, the size shifts in one direction.

Test it without pulling the machine apart. Mark the coupling and the screw with a paint line, then command a fast traverse and a reversal. The two marks separate if there is play. Any visible offset means the coupling needs re-tightening or replacement.

Guideway drag is the same idea, but slower. Binding on the linear rail, or an over-tightened gib on a box way, adds resistance that the servo fights through. The result is irregular size change in the low-micron range, plus heat in the slide. Re-adjust the gib and restore lubrication.

• 1
  Coupling playMarks separate on reversal
• 2
  BacklashPosition deviation peaks at direction change
• 3
  Guideway dragSlide runs warm, current draw climbs

Metal grows when it gets hot. A 300 mm aluminum part rises about 0.007 mm per °C. A 20 °C swing over a morning shift moves the size by 0.14 mm, which is far more than most shops will accept on a tight callout.

The spindle is the usual source. Run it at 12,000 rpm for 30 minutes and the nose can grow 0.02-0.04 mm. Start cutting immediately and the first ten parts run small. Add a warm-up cycle and the drift flattens out.

Coolant temperature matters just as much. Plumb the chiller to the machine and hold it near 20 °C. On machines without a chiller, measure the coolant at the start and the end of a run. A 5 °C rise is enough to show up in the final inspection report.

• 1
  Aluminum≈0.007 mm growth per 300 mm per °C
• 2
  Steel≈0.003 mm growth per 300 mm per °C
• 3
  Warm-up20-30 min at cutting rpm before first part

A servo loop that is not tuned correctly will leave position error during acceleration. On a move that combines X and Z, one axis trails the other and the tool path shifts. The part size then depends on feed rate, which is the giveaway.

Read the position deviation on the drive. On a Fanuc control that is DGN 800-804. Command a fast traverse and watch the peak value. A healthy axis recovers to near zero within the settling time. A large peak that decays slowly means the gain is too low, or the mechanical load is too high.

A worn or poorly adjusted rolling bearing shows up the same way. The bearing adds drag, the loop cannot keep up, and the size wanders by a few microns in no fixed pattern. Replace the bearing and re-tune, not just one or the other.

• 1
  Gain too lowSlow settle, size depends on feed
• 2
  Gain too highAudible hum, overshoot marks on finish
• 3
  Bearing dragSlow settle with no tuning improvement

Tool deflection is a size error in its own right. A 6 mm end mill hanging 40 mm out of the holder bends under cutting load. Push past about 0.3 mm radial depth of cut and the wall bows. The part measures oversize after the tool leaves, because the spring-back is gone.

Shorten the gauge length or step down to a smaller radial engagement. A 0.2 mm step-over at 2,000 mm/min often holds size better than 0.5 mm at 800 mm/min on thin walls.

Fixtures bend parts too. A three-point clamp on a thin ring will ovalize it while held, then the bore springs back round after release. Check by measuring while clamped and again after. If the two numbers differ, reduce clamp torque or add support under the cut.

• 1
  L/D over 5:1Expect measurable deflection
• 2
  Thin wallsClamp low, support opposite the cut
• 3
  Measure releasedCompare clamped vs free size

Some size instability is not a fault. It is the process telling you it cannot hold the callout at that geometry. A 500 mm long aluminum extrusion with a 0.02 mm flatness callout will move with the weather. No amount of tuning fixes that.

Thin, long, or asymmetric parts are the usual offenders. If the wall is under 1 mm, or the length-to-thickness ratio is past 20:1, expect to fight the material. Better to change the design, add a stress-relief step, or split the operation.

Send the drawing before the order and we will flag the features that are hard to hold. Free DFM analysis comes back within 12 hours, with a note on which dimensions we can hold at ±0.005 mm and which need a different approach.

• 1
  Stress reliefNormalize before the finishing pass
• 2
  Rough and restLeave 0.3 mm, cool, then finish
• 3
  Design changeAdd a rib instead of chasing tolerance