Troubleshooting guide

CNC Maintenance and Transformation: Symptoms, Causes, Fixes

This page is for maintenance engineers and plant managers who keep older machining centers running. Work through the fault table, the diagnostic steps, and the repair-or-retrofit thresholds, then decide with numbers instead of opinions.

Get Instant Quote Talk to an Engineer

±0.005 mm capability127 CNC machinesISO 9001 / IATF 1694915 years in Dongguan

CNC maintenance and transformation guide for older machining centers

On this page

Fault table: symptom, cause, countermeasure

Read the left column first. Three columns must line up before you touch the machine.

Symptom	Likely cause	Countermeasure
Taper size drifts over a shift	Spindle thermal growth	Warm up 20-30 min, re-zero after warm-up
Round holes come out oval	Backlash in X or Y	Measure backlash, adjust or replace ballscrew nut
Surface chatter at high rpm	Bearing preload loss	Check runout, replace spindle bearings as a set
Steps in a contour cut	Servo gain mismatch	Re-tune velocity loop, check encoder coupling
Part face not square	Machine not leveled	Re-level bed, check foundation pads and anchor bolts
Position error grows with travel	Pitch error in ballscrew	Laser-compensate pitch error, re-check every 6 months
Alarm on rapid moves only	Drive overload or binding	Check way lube, gib clearance, then drive current limit

Fix the mechanics, then decide

Measure backlash, level, and squareness first. If they pass, a CNC maintenance and transformation project is worth quoting. If they fail, no control upgrade will save the part.

Get Instant Quote Talk to an Engineer

Fault mapping

Start with the symptom, not the part

Most CNC maintenance and transformation decisions stall because the conversation starts with the part and not the machine. Two operators report the same complaint: a bore drifts 0.02 mm over a 200-part run. One machine has spindle thermal growth, the other has a worn ballscrew nut. The fix is different, and the cost is different by an order of magnitude.

Write the symptom in a way that can be measured. "Bad finish" is not a symptom. "Ra rises from 1.6 to 3.2 μm after 40 minutes of continuous cutting" is. Once the number and the trigger are on paper, the cause list usually shrinks to two or three items, and you can test them in the order of cheapest first.

Check the trigger before the machine. Does the fault appear only after a cold start, only at high spindle speed, only on one axis, or only after a long run? A fault tied to warm-up points to thermal growth. A fault tied to direction reversal points to backlash. A fault tied to feed rate points to servo tuning or mechanical binding.

Keep a one-page log per machine: date, symptom, axis, measured value, action taken, result. After six months that log tells you whether the machine is drifting or the process is drifting. It also gives the retrofit quote something real to work from.

1
Cold start faultWarm-up and thermal growth, not the control
2
One axis onlyMechanical wear or a single drive, not the program
3
Grows with run timeThermal or lubrication, not geometry
4
Appears after reversalBacklash or lost motion in the drive train

Wear limits

Backlash, level, and when geometry is the cause

Backlash is the first number to measure because it is cheap to check and it explains a large share of position faults. Mount a dial indicator on the table against a ground block, jog the axis in one direction, zero the indicator, then jog the same distance back. The difference is the total lost motion, including the nut, the thrust bearings, and the coupling. On a typical 750 mm travel vertical mill, keep total backlash under 0.010 mm on X and Y for finishing work. Beyond 0.020 mm, contours will show witness marks at direction changes and bores will go oval.

Level comes second. A machine that has settled on a soft pad will twist the bed, and the twist shows up as a squareness error that no amount of compensation removes. Check level with a precision level at four corners and re-check after 24 hours. Twist of more than 0.02 mm per 1,000 mm across the bed is enough to spoil squareness on a 400 mm part. Re-grout or re-shim the pads before you spend money on the control.

Squareness and parallelism come third, and they are the numbers that decide repair versus retrofit. Measure them with a granite square and an indicator, or with a laser interferometer if one is available. If squareness error stays under 0.010 mm per 300 mm after leveling and backlash correction, the machine still has life for general work. If the error is in the rails or the bed itself, no control upgrade will fix it.

Pitch error is the one geometry item that can be corrected electronically. A laser measurement maps the screw error, and the control applies a compensation table. This works well up to about 0.03 mm of accumulated error over the travel. Beyond that, the screw is usually worn unevenly and replacement is the honest answer.

1
Backlash limitUnder 0.010 mm for finishing, investigate past 0.020 mm
2
Level limitTwist under 0.02 mm per 1,000 mm across the bed
3
Squareness limitUnder 0.010 mm per 300 mm after leveling
4
Pitch compensationPractical up to about 0.03 mm accumulated error

Electronics

Power, drives, and the retrofit decision

Electrical faults on older machines rarely come alone. A drive that trips on rapid moves often sits behind a contactor with pitted faces, and the same cabinet may have a cooling fan that stopped two years ago. Before replacing a drive, check the incoming voltage under load, the DC bus ripple, and the cabinet temperature. A cabinet running above 45 °C will shorten drive life no matter how good the drive is.

Encoder faults deserve a specific check. A dirty or loose encoder coupling produces position errors that look random and appear on one axis only. Clean the disc with dry air, check the coupling runout, and re-seat the connector. This costs an hour and saves a drive replacement more often than most people expect.

This is where the retrofit question becomes concrete. If the mechanics pass their checks and the failure is in the control, the drives, or the feedback system, a retrofit can be economical. If the mechanics are worn, the retrofit just moves the problem to a newer screen.

Set the threshold in money, not in feeling. Get a written quote for the retrofit, a written quote for a comparable used machine, and a written quote for a new machine at the capacity you actually need. Compare three-year cost including downtime, spares, and operator retraining. Retrofit wins when the mechanical checks pass and the control is the only weak link.

1
Check under loadMeasure voltage and DC bus ripple while cutting, not idle
2
Cabinet temperatureKeep below 45 °C for drive life
3
Encoder firstClean disc and check coupling before replacing a drive
4
Retrofit thresholdMechanics pass, electronics fail: retrofit is worth quoting

Transformation

What transformation actually changes

Transformation of a manual or older CNC machine usually means three things: new drives and motors, a new control, and new feedback. It does not mean the machine becomes a 5-axis center. If the part needs five simultaneous axes, the geometry and the kinematics are not there, and no control swap creates them.

The realistic gains from a control and drive retrofit are shorter setup, better contour control, and modern network and probing support. Typical position accuracy after a well-executed retrofit lands near the original machine specification, not better. Plan for the machine to hold what it held when new, not what a new machine holds today.

Spindle and way condition set the ceiling. A spindle with worn bearings will still chatter after the retrofit. Ways with scoring will still show stick-slip at low feed. Deal with those first, or the retrofit budget partly pays for problems you did not fix.

For parts that need tight tolerances, ±0.005 mm at our shop comes from machines built for it, with 16 simultaneous 5-axis centers in the fleet. When a retrofit cannot reach the drawing, moving that part to the right machine is cheaper than forcing the old one.

1
Realistic gainSetup time, contour control, probing and network
2
Accuracy ceilingBack to original spec, not to new-machine spec
3
Fix mechanics firstSpindle and ways decide the final result
4
Wrong machineIf ±0.005 mm is required, plan a machine change

Procedure

Step by step: from complaint to decision

1
1. Write the symptom with a numberRecord axis, feed, spindle speed, and the measured value. Note whether it appears cold or after warm-up. Keep the same measuring method between checks.
2
2. Warm up and re-measureRun the spindle 20-30 min at 2,000-4,000 rpm, then repeat the measurement. If the error drops, thermal growth is part of the story.
3
3. Measure backlash on every axisUse a dial indicator and a ground block. Jog one direction, zero, jog back. Record total lost motion. Flag anything above 0.020 mm.
4
4. Check level and foundationUse a precision level at four corners, wait 24 hours, re-check. Twist above 0.02 mm per 1,000 mm means re-shim or re-grout before anything else.
5
5. Check squareness and parallelismGranite square plus indicator, or laser if available. Under 0.010 mm per 300 mm is acceptable for general work. Worse points to rails or bed.
6
6. Inspect the electrical cabinetMeasure incoming voltage under load and DC bus ripple. Check fan operation and cabinet temperature. Keep below 45 °C. Clean encoder discs and check couplings.
7
7. Price the three optionsGet written quotes for retrofit, used replacement, and new machine at the required capacity. Compare three-year cost with downtime and training included.
8
8. Fix mechanics before electronicsIf backlash, level, or spindle runout fail, correct those first. A new control on worn mechanics gives a clean screen and the same bad part.

FAQs

Common questions

How often should backlash be measured?

Check every six months on production machines, and after any crash or spindle change. If a machine runs two shifts, move it to quarterly. The measurement takes about 20 minutes per axis and gives you a trend line, which is more useful than a single number.

Replace the ballscrew nut, not just the compensation value, once total lost motion passes 0.020 mm on a finishing machine. Software compensation hides the error in one direction and leaves it visible in the other.

Can a retrofit improve accuracy beyond the original specification?

No. New drives and feedback can recover original accuracy and reduce lost motion, but the rails, bed, and spindle set the physical ceiling. If the bed is twisted or the ways are scored, the retrofit result will follow the mechanics.

Plan for the machine to hold what it held when new. If the drawing needs ±0.005 mm, that is a machine-class question, not a retrofit question.

When is a retrofit a waste of money?

When squareness or parallelism is out of tolerance after leveling and backlash correction, the geometry is gone. When the spindle housing or the ways need re-machining, the cost usually passes the value of the machine.

A retrofit also loses when the real bottleneck is not the machine. If parts wait for tooling, inspection, or programming, a new control will not move the schedule.

What spare parts should be kept for older machines?

Keep one spare drive of each model in use, a set of encoder cables, a contactor set, and the cooling fans. These are the parts that stop production and they are often discontinued.

Store drives in a dry cabinet and cycle the power once a year. Capacitors age faster in storage than in service if they never see voltage.

Does maintenance affect the certification of the parts we buy?

Process control does. Our shop holds ISO 9001:2015, IATF 16949:2016, ISO 13485:2016, and ISO 27001:2022, and every shipment is inspected before it leaves. Machine maintenance records are part of that chain.

If you need documented inspection reports with the parts, ask for them on the order. Reports are issued on request, not by default.

How do we decide between repairing the machine and moving the part?

Compare three numbers: the tolerance the drawing needs, the tolerance the machine holds after maintenance, and the lead time for the alternative. If the machine cannot hold the tolerance, repair will not change that.

For tight work we run 16 simultaneous 5-axis centers with a ±0.005 mm capability and ship in 3-5 days. Moving one hard part is often faster than rebuilding a machine around it.