CNC repair tips and tricks from the shop floor
Most CNC breakdowns start as a small signal: a spindle load spike, a finish that drifts from Ra 0.8 to Ra 1.6 μm, or a single alarm that clears on reset. This guide is for maintenance techs, machinists, and process engineers who want to catch those signals early. Read it and you will know which faults you can fix at the machine, which ones need a specialist, and how to document both.
In this article
- 1
- 2
- 3
- 4
- 5
- 6
- 7
Key takeaways
Start with alarm triage and machine history
Every repair starts with evidence, not with a wrench. Before you touch anything, pull the alarm history and the maintenance log. On most controls you can export the last 200 to 500 alarms with timestamps. Read them in order. The alarm that stopped the machine is rarely the first one that mattered. A servo overload at 14:32 often follows a lubrication pressure warning at 14:06.
Write down three things: what the operator was doing, what the machine was doing, and what changed. Was it a first run after a long idle? A new program? A new fixture? A cold morning in an unheated bay? Thermal growth on a 4,000 mm machine can shift geometry by more than the tolerance band itself, so ambient conditions are part of the evidence, not background noise.
Check the obvious consumables next. Way lube level, air pressure, coolant concentration, and spindle chiller temperature. A surprising share of "axis faults" trace back to a clogged lubrication line or air pressure below the machine specification. These checks take five minutes and cost nothing. Do them before you open an electrical cabinet.
Finally, reproduce the fault if it is safe to do so. Run the same program block in single step with feed override at 10% to 25%. If the fault appears only above a certain feed rate or only in one quadrant of a circular move, you already know more than the alarm code tells you.
- 1Export the alarm logKeep at least 200 entries with timestamps before clearing anything.
- 2Note ambient temperatureLog bay temperature and how long the machine has been running.
- 3Check consumables firstWay lube, air pressure, coolant, chiller setpoint.
- 4Reproduce at low feedSingle step at 10–25% override to isolate the trigger.
Diagnose spindle and axis symptoms before you dismantle
Spindle problems usually announce themselves in three ways: noise, vibration, or thermal drift. Put a vibration meter on the spindle housing and read velocity in mm/s. A healthy spindle at 8,000 rpm typically sits under 1.5 mm/s. Between 1.5 and 2.8 mm/s, plan a bearing inspection at the next scheduled stop. Above 2.8 mm/s, stop running production parts.
Rule out tooling before bearings. A holder with 0.01 mm runout or a tool that is 20% heavier on one flute will produce vibration that looks like a spindle fault. Swap in a known-good holder, rebalance if the tooling is adjustable, and re-measure. This costs one hour. A spindle rebuild costs weeks.
On the axes, separate mechanical backlash from servo tuning error. Command a short move of 0.05 mm and read the actual position with a dial indicator or a laser interferometer if you have one. Repeat ten times in the same direction and ten times reversing. Repeatability under 0.005 mm with backlash under 0.02 mm is generally serviceable for finishing work. Larger backlash shows up as poor circularity in interpolation, and no amount of parameter tuning fixes worn thrust bearings.
Heat is the other axis variable. Run a warm-up cycle of 20 to 30 minutes before measuring anything. Cold measurements on a machine that has been idle overnight will mislead you on both backlash and squareness.
- 1Spindle vibrationWatch for velocity above 2.8 mm/s at operating speed.
- 2Toolholder runoutAnything over 0.01 mm is worth replacing before blaming bearings.
- 3Axis backlashUnder 0.02 mm is usually acceptable; check repeatability too.
- 4Warm-up first20–30 minutes of cycle time before taking measurements.
Fix surface finish and dimensional drift at the source
Finish drift is a slow fault, which makes it easy to ignore and expensive to live with. If a part that held Ra 0.8–1.6 μm last month now reads Ra 1.6–3.2 μm, do not change the program. Change one variable at a time and measure. Start with the tool: edge wear, built-up edge on aluminium, and a chipped corner all show up as finish problems before they show up as dimensional problems.
Then look at coolant delivery. On deep pockets and long-reach tools, poor chip evacuation causes recutting and chatter. Verify nozzle aim, flow rate, and concentration. A refractometer reading between 6% and 10% is a normal band for many aluminium jobs, but check your coolant supplier's range for the specific product.
Dimensional drift in the same direction across many parts usually points to thermal growth or tool wear, not to a bad servo. Measure the first part, the part in the middle of the run, and the last part. If the trend is monotonic, it is thermal or wear. If it is random, it is mechanical or fixturing.
For tight work at ±0.005 mm, keep spindle and axis temperatures stable and let the machine soak. Many shops schedule their tightest jobs in the middle of the day, after the machine has been running for hours. That is not a superstition; it is thermal stability.
- 1Change one variableTool, then coolant, then parameters. Never all three at once.
- 2Measure first, middle, lastA monotonic trend means thermal or wear, not servo.
- 3Watch built-up edgeCommon on aluminium and 300-series stainless at low speed.
- 4Soak before tight workLet the machine stabilize before running ±0.005 mm jobs.
Preventive checks that stop most CNC repair calls
Most emergency calls we see start as a skipped preventive task. A daily check of way lube level, air pressure, and coolant flow takes ten minutes. A weekly check of filter condition, chiller temperature, and spindle taper cleanliness takes thirty. Shops that keep this rhythm spend far less time chasing faults than shops that run until something breaks.
Back up parameters and programs on a schedule, not after a crash. Controls lose absolute position references, and recovery without a backup can take a full shift. Keep an offline copy of parameters, macro variables, tool offsets, and ladder logic. Store it somewhere other than the machine's own hard drive.
Track tool life and replace on counts, not on feel. A tool that has run 20% past its normal life will produce scrap before it produces an obvious failure. Log tool changes and correlate them with finish and dimension data. That data tells you when to shorten the interval.
Document every repair, even the small ones. What failed, what you checked, what you replaced, and what you ruled out. The next technician saves hours, and patterns show up that no single event reveals. A simple log beats a perfect memory.
When a fault crosses into geometry, alignment, spindle rebuilding, or thermal compensation, that is the point to call a specialist. Five-axis platforms and large gantry machines carry alignment work that general maintenance crews are not equipped to verify.
- 1DailyWay lube, air pressure, coolant flow, chip evacuation.
- 2WeeklyFilters, chiller temperature, taper cleanliness, way covers.
- 3MonthlyBacklash checks, level check, parameter backup verification.
- 4Every repairWrite down what you checked and what you ruled out.
Step-by-step CNC repair tips for a fault on the floor
Work in this order. Skipping a step usually sends you back to it later.
- 1Lock out and record the stateStop the program, note the block number, and lock out the machine before opening any panel. Photograph the screen and the alarm log. Do not clear alarms until you have exported them.
- 2Read the alarm log in orderLook at the first 10 to 20 alarms before the stop. Note any warning that repeats. A lubrication or air pressure warning that appears twice in an hour points to the root cause more often than the final servo alarm.
- 3Check consumables and utilitiesVerify way lube level and pump operation, air pressure at the machine inlet, coolant flow and concentration, and spindle chiller setpoint. Restore anything out of range and retest before going further.
- 4Isolate mechanical from electricalCommand a 0.05 mm move and read actual position with an indicator. Check backlash and repeatability. If the servo holds position and the mechanical reading is off, the problem is mechanical. If the reading drifts with no command, look at the drive and feedback.
- 5Measure spindle conditionCheck toolholder runout first, then spindle vibration in mm/s. Under 1.5 mm/s is healthy at typical operating speeds. Above 2.8 mm/s, stop production and plan a bearing inspection.
- 6Test the suspect axis under loadRun a warm-up cycle of 20 to 30 minutes, then cut a test part with a circular interpolation move. Measure circularity. Roundness error in one quadrant usually points to backlash or a loose gib, not to servo gains.
- 7Change one thing at a timeReplace the tool, re-test. Then adjust coolant, re-test. Then look at parameters. Changing multiple variables at once makes the result unreadable and often hides the real fault.
- 8Escalate when geometry is involvedAlignment, squareness, spindle rebuilds, and thermal compensation need specialist tooling and a controlled environment. Document your findings and hand off with the data, not just a description.
Symptom, likely cause, and who should fix it
Use this to decide whether to keep working at the machine or call a specialist.
| Symptom | Likely cause | In-house fix | Call a specialist when |
|---|---|---|---|
| Alarm clears on reset | Intermittent sensor or loose connector | Yes, check wiring and connectors | The same alarm returns within one shift |
| Finish drifts to Ra 1.6–3.2 μm | Tool wear or built-up edge | Yes, replace tool and re-test | Finish stays rough with a new tool |
| Circularity error in one quadrant | Backlash over 0.02 mm or loose gib | Sometimes, adjust gib and re-measure | Backlash remains after adjustment |
| Spindle vibration over 2.8 mm/s | Worn bearings or unbalanced tooling | Only tooling swap | Vibration persists with good tooling |
| Dimensional drift across a run | Thermal growth | Yes, add soak time and monitor | Drift exceeds the tolerance band |
| Poor chip evacuation in deep pockets | Coolant aim or pressure | Yes, adjust nozzles and flow | Design needs a different strategy |
| Repeated position loss on startup | Absolute encoder or battery | No, needs diagnostic tools | Always, before running production |
Fix what you can measure, escalate what you cannot
Use the alarm log, backlash readings, and spindle vibration numbers to decide. If the numbers point to mechanical wear or thermal growth beyond your tolerance band, hand it to a specialist with the data attached.
CNC repair questions engineers ask
How often should we check axis backlash?
For production machines running most days, check backlash monthly and after any crash. Use a dial indicator with 0.001 mm resolution on a 0.05 mm command move, ten times each direction.
Under 0.02 mm is usually workable for finishing. Above that, circularity and interpolation accuracy suffer, and the fix is mechanical, not a parameter change.
When is spindle vibration too high?
Read velocity in mm/s at operating speed. Under 1.5 mm/s is healthy for most spindles at typical production speeds. Between 1.5 and 2.8 mm/s, plan an inspection at the next scheduled stop.
Above 2.8 mm/s, stop running production parts. Check toolholder runout and balance before assuming bearings. A holder with more than 0.01 mm runout can produce vibration that looks like a spindle fault.
Can we fix thermal drift ourselves?
Partially. You can add warm-up cycles of 20 to 30 minutes, keep the bay temperature stable, and schedule tight-tolerance work after the machine has soaked. These steps reduce drift a lot.
You cannot fix compensation tables or alignment without specialist tooling. If drift still exceeds the tolerance band after stabilizing the environment, that work belongs with a specialist.
What should be in a machine repair log?
Date, machine, operator, alarm codes with timestamps, what you checked, what you changed, and the result. Include the ambient temperature and whether the machine was cold or warm.
The log is most useful when it records what you ruled out. That is what saves the next technician time and prevents repeat work.
How do we reduce unplanned downtime overall?
Daily checks of way lube, air pressure, and coolant take ten minutes. Weekly checks of filters, chiller, and taper cleanliness take thirty. Parameter and program backups should be on a schedule.
Tool changes should follow counts, not feel. Log tool life against finish and dimension data, then shorten intervals where the data shows wear before failure.
When should we send a part out instead of repairing in-house?
When the fault involves geometry, alignment, spindle rebuilding, or thermal compensation, the verification tooling and environment matter as much as the repair. General maintenance crews usually cannot verify squareness or spindle taper geometry to the needed level.
For everything else, in-house triage is faster and cheaper. Document the evidence, escalate with data, and keep the machine running where you can.
Need parts machined while your machine is down?
Send your drawings and get a quotation with free DFM analysis within 12 hours. Production can start within 24 hours, with 100% inspection before shipment.
12-hour quote100% inspectionNo minimum order quantity
More from the shop floor
We publish setup notes, tooling trials and inspection data from the factory floor.
