CNC System Troubleshooting: Two Rules That End Repeat Faults
Most Fanuc faults come back because the first repair was aimed at the symptom, not the loop that produced it. This page covers two rules we apply on our own machines: follow the alarm chain to its first trigger, then verify the servo loop end to end. Read it and you can judge in a few minutes whether a fault belongs to the control, the drive, or the mechanics.
Symptom, likely cause, and what to do
Work down the rows in order. If two rows match, fix the first one and re-run the part before touching anything else.
| Symptom | Likely cause | What to do |
|---|---|---|
| Alarm clears after a restart, returns hours later | Intermittent contact in a peripheral switch or cable | Log alarm number and time, then meter the switch and connector |
| Alarm number changes between events | Power or ground fault upstream of the drive | Check 24 V supply ripple and cabinet grounding before parts |
| Single axis drifts while others hold position | Loose coupling, worn belt, or encoder gap | Clamp a dial indicator on the axis and measure lost motion |
| Loud squeak or squeal at high feed | Servo gain too high or mechanical binding | Lower velocity loop gain in steps and re-run the same block |
| Overheat alarm after 30-60 minutes | Cooling fan stalled or filter clogged | Replace the fan, clean the filter, then re-check the load meter |
| Alarm only on one program or tool | Wrong offset, worn tool, or bad macro call | Run the block in single step and compare commanded vs actual |
| Spindle load spikes with no alarm | Tool wear or chip packing in the cut | Inspect the edge and chip evacuation before touching parameters |
Follow the alarm chain to its first trigger
A Fanuc control reports the alarm it sees, not the alarm that started the sequence. When a drive trips, the CNC often shows a second or third alarm on the same screen, and the operator writes down the one at the top. That is usually the last link. Our first rule is simple: read the whole alarm chain and find the earliest trigger before you remove a single cover.
On a 0i or 30i control, press the SYSTEM key and open the alarm screen, then page through the history rather than only the active list. Compare the timestamps. If the drive alarm appears 200 ms before the CNC alarm, the CNC alarm is a reaction, not a cause. On many lathes we have traced a repeated servo alarm back to a motor switch on a peripheral unit that dropped out under load, not to the servo amplifier that kept flagging the fault.
Write the sequence on paper: time, alarm number, axis, and what the machine was doing. This takes two minutes and stops the most expensive habit in a shop, which is swapping a board to see if the alarm goes away. A spare drive proves nothing if the trigger sits in a cable, a relay, or a door interlock.
Once you have the first trigger, ask what condition produced it. Overcurrent, overheat, and position error all mean different things. The alarm text tells you the threshold that was crossed. It does not tell you why.
- 1Read the history, not the top lineThe active alarm is often the last one in the chain.
- 2Sort by timestampThe earliest event in the chain is the one to investigate.
- 3Record axis and program blockA fault tied to one block points at the process, not the hardware.
Verify the servo loop before adjusting parameters
The second rule is to close the loop on paper before you touch a parameter. A Fanuc servo loop runs from the command through the amplifier to the motor, then back through the encoder or scale. A fault can live anywhere in that ring. Parameter changes only address one part of it, and they often hide a mechanical problem for a few weeks.
Start at the mechanics. Push and pull the axis by hand with the servo off and feel for play, then clamp a dial indicator on the table or turret and measure lost motion at the reversal point. Bearing preload, coupling screws, belt tension, and gib clearance all show up here. If lost motion is above the value in your acceptance sheet, no gain setting will fix it.
Then check the feedback path. Clean the scale or encoder head, check the gap against the manufacturer value, and inspect the cable shield bond at both ends. A shield that is grounded at one end only is a common source of faults that appear when a nearby welder or VFD runs.
Only after mechanics and feedback are clean should you look at gain and filter settings. Change one value at a time, run the same program block, and record the result. A parameter change with no recorded baseline is guesswork, and it makes the next fault harder to read.
- 1Measure before you tuneLost motion and backlash are mechanical facts, not tuning problems.
- 2Check both ends of the shieldA floating shield invites noise faults that come and go.
- 3One change at a timeRecord the value, the block, and the result.
When these rules do not apply
These two rules assume the machine was running and then stopped. If the machine has never run correctly after a retrofit, an install, or a parameter restore, the problem is usually configuration, not a fault. In that case compare the parameter backup against the last known good set before you chase alarms.
They also assume you have a backup. If there is no parameter and ladder backup, stop and make one first. A troubleshooting session that ends with a corrupted SRAM is far more expensive than the original fault. Fanuc controls let you save parameters to a CF card or over Ethernet, and it takes minutes.
Finally, some faults belong to the process rather than the machine. Chatter, poor finish, and short tool life on a stable machine often come from speeds, feeds, and workholding. If the alarm list is empty and the part is still wrong, the control is not the place to look.
Step by step: two rules in practice
Run these in order. Do not skip a step because the alarm looks familiar.
- 1Capture the alarm chainOpen the alarm history on the SYSTEM screen and photograph every page. Note the timestamp, alarm number, and axis for each entry.
- 2Identify the first triggerSort by time. The earliest entry is the candidate cause. If two alarms share a timestamp, treat the drive-level one as the trigger.
- 3Match the symptom to the tableUse the symptom table above to narrow the cause to control, drive, feedback, or mechanics before opening any cabinet.
- 4Check the mechanical side with the servo offMeasure lost motion with a dial indicator at reversal. Compare against your acceptance value, typically 0.01-0.02 mm on a positioning axis.
- 5Inspect the feedback pathClean the encoder or scale head, verify the gap to the manufacturer spec, and confirm the shield is bonded at both ends.
- 6Check power and groundingVerify the 24 V supply under load and inspect cabinet ground bonds. Loose bonds cause faults that move between axes.
- 7Change one parameter at a timeRecord the old value, the new value, and the program block you used. Revert if the fault returns within one shift.
- 8Re-run the same part and log the resultRun the identical program and tool. If the fault is gone, write the fix into the machine log so the next operator does not repeat the search.
Questions engineers ask next
The alarm clears after a restart. Does that mean the fault is gone?
No. A fault that clears on restart is usually intermittent, and the restart only resets the state that tripped it. The underlying contact, cable, or thermal condition is still there.
Log the alarm number and the time of day for three or four events. Patterns in time or in the program block point at heat, vibration, or a specific move.
How do I know whether the fault is in the control or the drive?
Look at which unit raises the alarm first. If the drive LED shows a fault before the CNC screen updates, the drive or its load is the source. If the CNC raises the alarm with no drive indication, look at the command path and interlocks.
The alarm history timestamps usually settle this in under a minute.
Is it safe to swap a servo amplifier to test?
Only if you have confirmed the mechanical and feedback sides first. Swapping a good amplifier into a machine with a shorted motor cable can damage the spare as well.
Meter the cable and the motor windings before you move any hardware.
How often should we back up parameters and the ladder?
After every change and at least once per quarter. Keep two copies on separate media, and label them with the machine serial number and date.
A backup made before troubleshooting is worth more than any spare part in the cabinet.
Can we run production while a fault is intermittent?
Only with a tighter inspection step on the affected feature. An intermittent axis fault can produce a part that measures correctly at the start of the run and drifts later.
If the feature is safety-related or a mating surface, stop the run until the cause is found.
When should we call the machine builder instead of working it out ourselves?
Call when the fault involves the ladder, the safety circuit, or a parameter set you cannot compare against a known good backup.
Bring the alarm history and your measurements to the call. It shortens the conversation and usually the downtime.
Need parts cut while the machine is down?
Send us the drawing and we will quote within 12 hours, with free DFM feedback on tolerances and features.
12-hour quote100% inspectionNo minimum order
More from our shop floor
We publish setup notes, tooling trials and inspection data from the factory floor.
