Intermittent Fault: How to Catch It (Decision Tree)

Why this matters

An intermittent fault (a problem that comes and goes) is the hardest thing in any trade to fix, because the system usually works fine while you are standing in front of it. The temptation is to swap the most-suspected part and leave. That is a gamble, and intermittents punish gamblers with callbacks. The skill is to make the fault show itself, or to capture proof of it so you can act on evidence instead of hope.

Start here: pin down "when"

Before any test, interview hard. An intermittent always correlates with something. Ask:

• When does it act up? Time of day, after running a while, right at startup, only in heat or cold, only when wet.
• What is the system doing at that moment? Heavy load, switching on, vibrating, a particular cycle.
• Has anything changed recently (a repair, a part, a setting, weather, new equipment nearby)?

Write the answers down. The pattern is your map. An intermittent with no pattern is rare; usually the customer just has not noticed it yet.

If the fault correlates with temperature

Heat-related intermittents point at connections and components that change with temperature. If it fails only after warming up: suspect a marginal connection that expands and opens, a component drifting out of spec as it heats, or a heat-soaked control. If it fails only when cold: suspect moisture, a stiff seal, or a part that needs warmth to conduct. Use controlled heat or cold (a heat source or freeze spray) on one suspect at a time and watch for the fault to appear or clear.

If the fault correlates with movement or vibration

This is the loose-connection signature. Gently wiggle, flex, and tap suspect wiring, connectors, and boards one at a time while the system runs. If a tap makes the fault appear or disappear, you have found the zone. Then look for the specific bad joint: a cracked solder point, a backed-out terminal, a corroded crimp, a chafed wire. Movement-triggered faults are almost never the expensive component; they are the cheap connection feeding it.

If the fault correlates with load or run time

If it fails under heavy load or after running a while: suspect a marginal supply (voltage sagging under demand), an undersized or failing component that holds up until stressed, or thermal cutout. Measure the relevant value (voltage, current, pressure, temperature) during the heavy condition, not at rest. A reading that is fine at idle and out of range under load is your answer.

If no pattern emerges: log it

When you cannot force the fault and cannot find a correlation, stop guessing and capture data over time:

• Leave a data-logging meter or recorder on the suspect signal.
• Mark the suspect connections so you can tell on the return visit whether anything moved or arced.
• Have the customer note the exact time and conditions each time it acts up.

You return, read the log, and now you have the moment of failure recorded. This is slower but it converts an unwinnable guess into a normal diagnosis.

Confirm before you leave

Once you believe you found it, try to make it fail again by reproducing the trigger (heat it, shake it, load it). A fix you cannot break on purpose is a fix you can trust. A part you swapped without ever seeing the fault is not a fix, it is a hope.

References

• Trade-standard intermittent-fault practice: wiggle test, thermal cycling, data logging.
• Multimeter manufacturer documentation on min/max recording and event capture.
• See related: Symptom Won't Reproduce on Arrival, Measure Don't Guess.