Intermittent faults reward technicians who preserve evidence and search for patterns instead of restarting until the problem disappears.
SURVIVAL RULE: Preserve evidence before the machine recovers.
Preserve evidence before reset
Intermittent faults are difficult because the system often returns to normal. Your best chance is to preserve what the machine knew at the moment of failure: alarm code, sequence step, sensor states, drive fault history, network status, operator action, product type, time, temperature, and machine position.
A servo communication fault that clears after power cycling may leave a useful log entry. If the first response is to reboot everything, that evidence is gone.
Field habit: Create a standard intermittent-fault note template and use it before reset when safety and production allow.
Common trap: Do not confuse recovery with repair.
Look for position, temperature, time, and vibration patterns
Intermittent problems become solvable when you discover what changes when they occur. Common variables include machine position, cable flex, warm-up time, ambient temperature, product size, speed, vibration, humidity, and simultaneous loads.
A sensor cable may open only when the carriage reaches the far end. A drive may fault only after the cabinet warms. A pneumatic problem may appear only when several cylinders move together.
Field habit: Ask operators what is always true just before the failure, then verify with observations or trends.
Common trap: Do not accept ‘random’ until you have tested the major changing variables.
Quick check
What should be true at this point in the sequence?
What evidence can prove or eliminate this section of the system?
What changed recently or only under the failing condition?
Compare good cycle versus bad cycle
A strong method is to capture the same signals through a normal cycle and through a failing cycle. The first divergence is often the fault boundary. This can be done with PLC watch tables, trends, drive traces, network logs, or simple timestamped observations.
If a good cycle shows sensor S4 for 220 ms before output Y7, but a bad cycle never shows S4, you have a specific direction: sensor, target, wiring, timing, or input.
Field habit: Choose a small set of relevant signals rather than trending hundreds of tags.
Common trap: Do not drown yourself in data. Capture the sequence around the symptom.
Repair evidence, then monitor recurrence
When you find a likely intermittent cause, make one repair and monitor enough cycles to demonstrate improvement. Because the fault was intermittent, one successful cycle proves little. Record what changed and how long or how many cycles the machine ran afterward.
If a cracked M12 connector was replaced, note the physical damage and then verify the axis through the positions that previously caused dropouts.
Field habit: Define a verification window appropriate to how often the fault occurred.
Common trap: Do not label a fault ‘fixed’ merely because it did not happen again during the next five minutes.
Quick check
What should be true at this point in the sequence?
What evidence can prove or eliminate this section of the system?
What changed recently or only under the failing condition?
Field Exercise
Create a five-line intermittent-fault capture form and use it the next time a fault recovers by itself.