How to Repair CNC Machines: A Shop-Floor Procedure

• 1
  Key takeaways
• 2
  Diagnose first: how to repair CNC machines starts with a written symptom
• 3
  Sort the fault into mechanical, electrical or control
• 4
  Isolate the faulty element before ordering parts
• 5
  When to stop and call the machine builder
• 6
  Stop the same fault coming back
• 7
  Step by step: how to repair CNC machines in a working shop
• 8
  Symptom, likely cause and first action

A stopped machine is a data source. Before the covers come off, write down what the operator saw: the alarm code, the axis or spindle involved, the program block being executed, spindle speed and load, coolant state, and whether the fault repeats at the same point in the cycle. A fault that repeats at one block is almost always program, tool or fixture related. A fault that appears at random times points to electrical or thermal causes.

Pull the alarm history from the control. Older Fanuc, Siemens and Heidenhain controls keep a log you can page through; capture the last twenty entries, not just the last one. The first alarm in a cascade is the useful one. Everything after it is a consequence.

Then ask what changed. A new fixture, a different material batch, a recent pallet change, a power event over the weekend, a coolant concentration that has drifted. Repair work is often just finding the recent change.

Do not start with the control parameters. Parameter corruption is real but it is far less common than a loose connection, a worn coupling or a dull tool. Keep parameters as the last suspect, not the first.

• 1
  Photograph the alarm screenScreens get cleared when someone cycles power. A photo survives.
• 2
  Note ambient temperatureThermal drift faults often show up in the first two hours of a shift.
• 3
  Check the maintenance logA fault that follows a recent repair is usually related to that repair.

Mechanical faults show up as noise, vibration, heat or geometry error. A spindle that runs warm and starts leaving a chatter pattern is usually bearings. A table that moves in steps instead of a smooth sweep points to a worn ball screw, a failing thrust bearing or a loose coupling. Gearbox backlash shows as a dimensional error that changes direction with the axis travel direction.

Electrical faults are intermittent by nature. Aging electrolytic capacitors on drive boards bulge or leak and cause faults that appear after the machine has been running for an hour. A contactor that chatters, a relay with pitted contacts, a cable that has been flexing in a drag chain for years. This is where most intermittent stop-start faults live, especially on machines past ten years old.

Control faults are the least common and the easiest to misdiagnose. They show as a single axis losing position, an encoder fault, or a drive that will not enable. Servo drive faults often trace back to a bad encoder cable rather than the drive itself. Read the drive LED pattern before you pull the unit.

A useful test: command a slow move in one axis with the spindle off and your hand on the motor housing and the ball screw nut. You will feel a bad bearing, a rough ball screw and a loose coupling within a few strokes.

• 1
  MechanicalNoise, heat, vibration, geometry error, backlash that changes with direction.
• 2
  ElectricalIntermittent faults, faults after warm-up, contactor chatter, cable damage.
• 3
  ControlPosition loss, encoder alarms, drive enable faults. Check the cable first.

Isolation means proving which single element is at fault, not which subsystem. On a slipping axis, disconnect the motor from the ball screw and run the motor alone. If the motor holds position and the fault disappears, the problem is downstream in the mechanical train. If the fault stays, it is the motor, encoder or drive.

For spindle faults, run the spindle at increasing speed in steps, for example 1,000, 4,000, 8,000 and 12,000 rpm, and watch vibration and current draw at each step. A fault that only appears above a threshold speed is usually balance or bearing related. A fault at every speed from the first revolution is usually drive or wiring related.

For hydraulic and pneumatic systems, check pressure at the gauge and then at the actuator. A pressure drop between the two means a leak or a blocked line. A pressure that holds with the actuator disconnected but drops when connected means the cylinder seals are passing.

Only after isolation does the parts list make sense. Ordering a servo drive before you have proven the encoder cable is where repair budgets quietly disappear.

• 1
  One variable at a timeDisconnect, test, reconnect. Do not test two subsystems together.
• 2
  Use the machine's own diagnosticsMost controls show following error and motor current in real time.
• 3
  Write the result downWhat you ruled out is as valuable as what you found.

Some repairs belong in the shop and some do not. Spindle cartridge replacement on a high-speed spindle needs a controlled environment, a press and a balancing step. If you fit bearings without balancing, you will be back inside the spindle within weeks. The same applies to direct-drive rotary tables and linear motor axes.

Control software and parameter work is another boundary. Firmware updates, drive parameter restoration and PLC logic changes should come from the machine tool builder or a qualified integrator, because a wrong parameter can drive an axis into a hard stop at full rapid. Rare, but expensive when it happens.

Geometry work is the third boundary. If the machine needs way grinding, column realignment or a full volumetric calibration, the job needs laser interferometry and a reference square, plus the time to bring the machine back into thermal equilibrium. That is a scheduled project, not a shift repair.

A useful rule: if the repair needs a clean room, a balancing step, a firmware change or a laser, schedule it with the builder. Everything upstream of that is fair game for a competent maintenance team.

• 1
  In scopeCables, couplings, belts, sensors, lubrication, coolant, tooling, fixtures, air and hydraulic lines.
• 2
  Needs the builderSpindle cartridges, direct-drive tables, firmware, PLC logic, volumetric calibration.
• 3
  Before restarting after any repairRecheck lubrication, coolant concentration and safety interlocks.

A repair without a change to the maintenance plan is a scheduled repeat. Once the machine is running, go back to the record you wrote at the start and decide which early signal you could have caught. A spindle temperature trend, a backlash measurement, a coolant concentration reading, a pressure log.

Backlash is the clearest example. Measure it every quarter on each linear axis and write the number down. When the value starts to climb, you have weeks of warning before the surface finish degrades. Replacing a coupling on a planned stop costs far less than scrapping a batch of parts.

Coolant is the quiet one. Concentration outside roughly 6 to 10 percent for common water-miscible fluids leads to corrosion, poor finish and tool wear. Check it with a refractometer weekly and top up with premix, not neat water.

Finally, keep a machine-specific fault file. The next time the same alarm appears, the diagnosis takes twenty minutes instead of two days.

• 1
  Quarterly backlash logOne number per axis, compared against the last reading.
• 2
  Weekly coolant checkConcentration, pH and tramp oil. Adjust with premix.
• 3
  One fault file per machineAlarm, cause, fix, date. This is the cheapest tool in the shop.