>
>
2026-10-19
Positioning accuracy loss in a vertical machining center is rarely a sudden electronics failure. In most shops it develops slowly and shows up first as a drifting dimension at the end of a long shift, a bore that needs an offset change after lunch, or a part that measures well on the fixture and out of tolerance after the table has been running for six hours. The dominant causes, in practical order, are thermal growth of the ballscrew and spindle, wear or preload loss in the guideways, backlash in the thrust bearing or coupling, a foundation that is no longer level, and servo tuning that no longer matches the mechanical condition of the machine.
The important distinction is between three numbers that buyers often treat as one. Positioning accuracy describes how close the axis arrives to the commanded target at any point in travel. Repeat positioning accuracy describes how consistently it returns to the same point. Backlash describes the lost motion when the axis reverses direction. A machine can repeat to within a few microns and still be out of position by 30 microns across a long travel, and that pattern points to a different root cause and a different repair.
Published specifications usually quote both values over full travel, and the ratio between them is informative. On the GUSS VMC-1165HL, for example, positioning accuracy is stated as ±0.005 mm with repeat positioning accuracy of ±0.004 mm over 1100/650/600 mm of travel. On the larger VMC-1680HL with 1600×800×700 mm of travel, the published figures are ±0.01 mm and ±0.008 mm.
That difference is not a quality gap; it is physics. Longer travel means longer ballscrews, more screw length exposed to spindle and nut heat, and more accumulated pitch error. Understanding this prevents a buyer from expecting a 1.6 m X-axis machine to hold the same absolute number as an 800 mm machine, and prevents an operator from blaming the control for a tolerance band that the machine was never specified to hold.
When the screw rotates under preload and carries cutting thrust, friction in the nut and in the support bearings generates heat. The screw expands and the effective lead changes. A screw that warms by 5 °C over a 1 m length moves its effective reference point by roughly 60 microns, which is already an order of magnitude larger than the ±0.005 mm class of a mid-size VMC. The symptom is characteristic: the part is correct at the start of the shift and progressively oversize or undersize in the same direction as the screw heats, then recovers after a stop.
Linear roller guideways lose preload as the recirculating rollers and raceways wear, and the table begins to lift or yaw under an offset load. On a machine used for heavy face milling at the edge of the table, wear is faster at the ends of travel than in the centre, so the error appears as a taper across the part rather than a constant offset. Rectangular (box) guideways behave differently: they resist load well but depend on gib adjustment and lubrication quality, and a starved way surface will show stick-slip marks before it shows a number on a test report.
Lost motion of 5–15 microns on reversal, measurable with a dial indicator and a programmed step move, usually comes from a worn angular-contact bearing pair, a coupling that has slipped on the shaft, or a locknut that has relaxed. This is the cheapest failure to fix and the most often misattributed to the control.
A machine that has settled on a floor slab, or was leveled once and never re-checked, twists the bed. Squareness between X and Y changes, and circular interpolation produces an oval instead of a circle. This is why a ballbar circularity test is worth running before any parameter is touched.
Spindle growth moves the tool downward as the spindle reaches working temperature. In a long finishing cycle this appears as a gradual depth error, and in a batch it appears as the first part being high and later parts being low. Machines with a spindle oil cooler and a programmed warm-up cycle control this far better than machines that start cold and go straight into a finishing pass.
Pitch-error compensation and backlash compensation are stored in the control and refer to the machine as it was when it was calibrated. After a screw replacement, a coupling change or significant wear, those tables are wrong. A machine that is mechanically healthy can still miss position if its compensation data is stale.
The table below is the working sequence we recommend to service engineers, ordered from the cheapest check to the most expensive. It is a diagnostic ladder, not a list of guaranteed causes.
| Observed symptom | Most likely cause | Diagnostic method | Corrective action |
|---|---|---|---|
| Dimension drifts in one direction as the shift progresses | Ballscrew or spindle thermal growth | Laser interferometer before and after a 2–4 h warm-up; part measurement at intervals | Programmed warm-up cycle, spindle cooler service, thermal compensation parameters |
| Constant offset, same part every cycle | Pitch-error compensation no longer valid, or tool offset error | Laser measurement at 100 mm intervals; compare with stored compensation table | Re-calibrate and rewrite the compensation table |
| Error only on reversal, 5–20 µm | Backlash in thrust bearing, coupling or nut | Dial indicator with programmed ±0.01 mm step moves | Re-preload or replace bearing pair, re-tighten coupling, reset backlash compensation |
| Taper across the part in X or Y | Guideway wear or preload loss, unevenly distributed | Straightness measurement, load test at table ends | Re-preload or replace sliders, re-scrape box ways, re-check gibs |
| Oval circles, out-of-square features | Machine geometry, leveling, foundation | Ballbar circular test per ISO 230-4; squareness with granite square | Re-level, re-check anchor torque, re-cut squareness compensation |
| Depth error grows through a long finishing pass | Spindle Z-axis thermal growth | Test cut with incremental depth measurement | Warm-up routine, spindle cooling, Z-axis thermal compensation |
Four instruments cover almost every case. A laser interferometer following ISO 230-2 gives absolute positioning and repeatability over the full travel and is the only method that produces a defensible number for acceptance. A ballbar following ISO 230-4 gives circularity, squareness and servo mismatch in a single test and is fast enough to use quarterly. A dial indicator and granite square cover backlash, squareness and parallelism in minutes. A test cut — a stepped block or a circular pocket in a stable material — is what actually matters to the customer, because it integrates geometry, thermal state, tool deflection and control behaviour into one result.
One practical note: measure the machine at the temperature it will actually run at. A cold machine measured at 8 a.m. and a warm machine measured at 4 p.m. can differ by more than the specification band, and neither reading is wrong.
Two design choices also reduce the rate of loss before the machine ever ships. The first is a rigid, thermally symmetric structure with a heavily ribbed cast base and column, which slows and evens out distortion. The second is a consistent transmission specification: the HL series uses PMI roller guideways from 35 mm to 55 mm depending on axis and model, with preloaded double-nut ballscrews, and on the heavy-duty HD series the X/Y axes move on rectangular rail with a larger slider count so that cutting force is spread over more contact length. Those choices do not eliminate wear, but they make the machine's accuracy decay predictable and measurable instead of erratic.
Buyers should scale expectations with travel rather than compare across classes. In the GUSS range, the compact VMC-855HL with 800×550×550 mm travel is specified at ±0.008 mm positioning and ±0.005 mm repeat; the mid-size VMC-1165HL at ±0.005 / ±0.004 mm; the long-travel VMC-1680HL at ±0.01 / ±0.008 mm; and the large VMC-1890HL at ±0.012 / ±0.008 mm over 1800×900×800 mm.
The pattern is consistent: as travel grows, absolute positioning tolerance relaxes while repeat accuracy stays tight, because repeat accuracy depends on the servo, the screw preload and the guideway preload rather than on accumulated length. For most production work the repeat figure is the more commercially relevant number, since offsets can correct a constant bias but cannot correct scatter.
A useful rule is that compensation is appropriate when the error is smooth and repeatable, and repair is required when the error is erratic, directional after reversal, or accompanied by surface finish deterioration. Smooth, monotonic pitch error across the travel is a compensation task. A backlash step at every reversal is a mechanical task. Vibration marks in the finish combined with a positioning error is a bearing or preload task. Treating a mechanical fault with a compensation table only hides it until the next tight-tolerance job arrives.
If you are specifying a new vertical machining center and want the accuracy clause written in a way you can actually enforce at acceptance, our engineering team can supply the model-specific geometry and accuracy data for the VMC range so that the purchase order, the acceptance test and the laser report all reference the same numbers.
