How to detect the machining accuracy of horizontal towers
A shop-floor procedure for boring mills and horizontal machining centers. You get the check order, the instruments to use, the numbers to write down, and the errors that make a good machine look bad.
In this article
- 1
- 2
- 3
- 4
- 5
- 6
- 7
- 8
What decides the result
1. What machining accuracy of horizontal towers actually covers
Horizontal towers is the shop-floor name for horizontal boring and milling machines and horizontal machining centers. Their accuracy is not one number. It is a stack of errors: bed level, column squareness to the table, spindle axis alignment, slide straightness, and the repeatability of the control. When a part comes out tapered or out of square, only one of those layers is usually at fault.
The checks below follow ISO 230-1 practice, where geometry is measured on the unloaded machine and then confirmed with a test cut. Geometry tells you what the machine can do in theory. The test cut tells you what it does with a spindle running hot, a tool clamped in a holder, and material pushing back.
You need five instruments: a 0.02 mm/m precision level, a granite square or cylinder square, a dial test indicator reading 0.001 mm, a magnetic base, and a 100 mm test bar. A laser interferometer or ballbar helps for positioning work, but the hand tools catch most problems first.
Do the work with the machine warmed up. A cold spindle can grow 20–30 μm as it reaches running temperature, which is larger than the tolerance you are trying to verify. Run a 30-minute warm-up cycle, then start measuring.
2. Level, foundation and thermal baseline before any dial work
Leveling is the first check because every other reading sits on the bed. Put the precision level on the table in the middle of the travel, then at four corners. On a typical boring mill you want the bubble inside 0.02 mm/m in both directions. A twist of 0.04 mm/m across a 2,000 mm bed lifts one corner about 40 μm.
Check the foundation bolts and the pads before you adjust anything. On older machines the floor has often settled more than the leveling screws can correct. If a pad is crushed or a bolt is loose, torque it to the maker's figure and let the bed settle for a day before re-reading.
Thermal state matters as much as geometry. Log the ambient temperature and the spindle temperature at the start and end of the check. If the shop swings 8 °C between morning and afternoon, repeat the checks at the same hour so the numbers can be compared month to month.
Write the readings into a log with date, ambient temperature and operator name. A single sheet is useless. Twelve sheets show a drift trend, and a drift trend is what tells you to schedule realignment instead of chasing a one-off error.
3. Squareness and parallelism of column, table and spindle
Squareness is measured with a granite square or a cylinder square on the table, and a dial indicator held in the spindle. Sweep the indicator along the vertical face of the square while moving the column or the spindle head up and down. The total indicator reading across 300 mm should stay within 0.01 mm for a machine held to ±0.005 mm.
Repeat the sweep in the horizontal plane against the other face of the square. This catches yaw, where the column leans left or right relative to the table travel. Yaw is common after a crash and shows up as a part that is parallel on one side and tapered on the other.
Table parallelism to the X and Z axes is next. Mount the indicator on the spindle, touch the table surface at the front, middle and back of travel, and record the variation. On a 4,000 mm machine a spread of 0.02 mm is normal; 0.05 mm means the bed needs leveling again.
One trap: the square itself. A worn or dropped granite square can be off by more than the machine. Keep a master square for reference only, store it in its case, and check it against a cylindrical square once a year.
4. Spindle runout, taper contact and axis alignment
Spindle runout is split into two readings. Inner runout is measured on a test bar seated in the taper, close to the gauge line. Outer runout is measured on the spindle nose diameter. For a machine rated at ±0.005 mm, keep inner runout under 0.005 mm and outer runout under 0.010 mm.
Clean the taper before you insert the test bar. A single chip or a film of oil changes the reading by several microns. Wipe with a lint-free cloth, seat the bar with a firm push, and rotate the spindle by hand for the first pass. If the indicator jumps once per revolution, the problem is in the bar or the taper, not the bearings.
Taper contact tells you whether the holder sits properly. Blue the test bar or a known-good holder and check that contact covers at least 80% of the taper length, spread evenly. Poor contact at the small end pushes the tool off center and shows up as a hole that is oversized in one direction.
Spindle axis alignment to the Z travel matters on horizontal machines because the spindle moves in and out. Sweep a test bar in the spindle against a square on the table while moving Z. Any rise or fall across the stroke is a spindle-to-Z alignment error, and it will bore a hole that drifts off position as the quill extends.
5. Positioning repeatability, backlash and the test cut
Positioning accuracy is checked by commanding the same point ten times and recording where the machine lands each time. Approach from the same direction for a repeatability figure, then approach from both directions to expose backlash. On a machine with ±0.005 mm capability, unidirectional repeatability should be under 0.005 mm and backlash under 0.010 mm.
Use a dial indicator on the table with the spindle moving in small increments, or a laser interferometer for a full stroke map. Record the reading at 10% intervals along X, Y and Z. Compensation tables can be rebuilt from that data, but only after the mechanical backlash is fixed. Software compensation hides a loose thrust bearing for a while, then stops working.
The test cut is the final word. Machine a 100 mm × 100 mm pocket and a 100 mm bore in aluminium 6061-T6 or mild steel, using a sharp tool and a finishing pass of 0.2 mm radial depth. Measure the bore with a bore gauge and the pocket with a micrometer. Roundness within 0.01 mm and squareness within 0.015 mm per 100 mm is a healthy result.
Compare the test cut with the geometry readings. If geometry is clean but the cut is out, look at the tool holder, the clamping, or the coolant. If geometry is out and the cut matches it, the machine needs realignment. That comparison is the whole point of doing both checks in the same session.
Step by step: run the check in order
- 1Warm up and log conditionsRun the spindle for 30 minutes, then record ambient and spindle temperature. Do not start cold.
- 2Level the bedPlace a 0.02 mm/m level at the table center and four corners. Target under 0.02 mm/m in both axes; re-torque foundation bolts if needed.
- 3Sweep squarenessUse a granite square and a 0.001 mm indicator. Keep total reading under 0.01 mm per 300 mm vertically and horizontally.
- 4Check the taper and runoutClean the taper, seat a 100 mm test bar, and read inner runout under 0.005 mm and outer runout under 0.010 mm.
- 5Check Z-axis alignmentSweep the test bar against a square while moving Z through its stroke. Any drift means the spindle axis is off.
- 6Test repeatability and backlashCommand one point ten times from each direction. Keep repeatability under 0.005 mm and backlash under 0.010 mm.
- 7Cut the test partFace a 100 mm pocket and bore a 100 mm hole in 6061-T6 with a 0.2 mm finishing pass. Measure with a bore gauge.
- 8Record and compareLog every number with the date. Compare the cut result against the geometry readings before deciding what to fix.
Check, instrument and target values
Targets assume a machine rated at ±0.005 mm.
| Check | Instrument | Target | Common error |
|---|---|---|---|
| Bed level | Precision level 0.02 mm/m | Under 0.02 mm/m both axes | Floor settled under a pad |
| Column squareness | Granite square + indicator | Under 0.01 mm per 300 mm | Dirty square face |
| Table parallelism | Indicator on spindle | Under 0.02 mm over 4,000 mm | Re-leveled without settling |
| Inner spindle runout | Test bar + indicator | Under 0.005 mm | Chip in the taper |
| Outer spindle runout | Indicator on nose | Under 0.010 mm | Worn nose diameter |
| Repeatability | Indicator or laser | Under 0.005 mm | Approach from one side only |
| Backlash | Indicator, both directions | Under 0.010 mm | Compensation masking wear |
| Test cut squareness | Micrometer, bore gauge | Under 0.015 mm per 100 mm | Dull tool, loose clamp |
Do the geometry pass before the test cut
If you only have time for one check, run squareness and level. A machine that is out of square will fail the test cut every time, no matter how good the tooling is.
Questions engineers ask before checking
How often should a horizontal tower be checked?
For a machine running two shifts, run the full geometry check every six months and the test cut every quarter. After any crash, after a move, or after a foundation repair, run the whole sequence regardless of the calendar.
Keep the log. If repeatability drifts 0.003 mm in two months, that is a trend worth acting on even though the machine is still inside specification.
Can we check accuracy without a laser interferometer?
Yes. A precision level, a granite square, a test bar and a 0.001 mm dial indicator cover level, squareness, runout and repeatability. Most shops never need more for routine checks.
A laser or ballbar becomes useful when you need a full stroke positioning map or want to rebuild a compensation table. It also shortens the time on a large 4,000 mm machine.
Why does the test cut fail when the geometry readings look fine?
Look at the tooling and the setup first. A holder with poor taper contact, a dull insert, or a clamp that lifts the part will all push the cut out of tolerance while the machine itself is straight.
Check coolant and chip evacuation as well. Heat going into the part during a roughing pass will move the wall position on a long bore.
Does temperature control matter that much?
Yes. Steel grows about 11 μm per meter per °C. On a 2,000 mm bed an 8 °C swing moves the geometry more than the tolerance you are verifying.
Measure at the same time of day, with the same warm-up routine, and note the temperature on every log sheet.
What if backlash is inside tolerance but parts still come out tapered?
Taper usually comes from squareness or from deflection, not backlash. Re-check the column against the table travel in the horizontal plane, and look at the tool overhang.
A boring bar sticking out 6× its diameter will deflect under load even on a perfect machine. Shorten the overhang or take a lighter finishing pass.
When should a machine be realigned instead of compensated?
Fix the mechanics first. If level, squareness or backlash is out of tolerance, realign or repair before touching the compensation table.
Software compensation can only correct repeatable, measurable errors. It cannot correct a loose bearing or a twisted bed, and it will hide the problem until a heavy cut exposes it.
Send us your drawing and tolerance callout
We machine horizontal-tower parts to ±0.005 mm with 100% inspection before shipment, and we return a quotation with free DFM analysis within 12 hours.
12-hour quote100% inspectionNDA on request
More machining notes
We publish setup notes, tooling trials and inspection data from the factory floor.
