Intermittent fault finding fails when it becomes a vigil. The fault appears when you are not there, disappears when you arrive, and returns after you leave. Progress comes from two directions only: making the fault happen on demand, or leaving an instrument watching so the moment is captured whether or not anyone is present.
Gather the pattern before touching anything
Intermittents almost always correlate with something, and the correlation is the most valuable clue available. Ask when it happens: at start-up, after running for a while, on cold mornings, during rain, when a particular machine runs, when someone walks past, at a specific time of day. Ask what changes around it. Look at whatever logs exist, including equipment error logs, building management records and maintenance history. A fault that only occurs when the heating starts points toward a supply disturbance or a thermal effect; one that only occurs when a door is closed points toward a cable being pinched. This costs nothing and frequently narrows the search more than an afternoon of probing. Write the pattern down, because it is also the thing you will use to verify a repair.
Provocation turns waiting into testing
Once you have a suspected mechanism, try to reproduce it deliberately. Thermal faults respond to controlled heating or cooling: warm a suspect area gently with a hot air source at a safe setting, or cool a component with freezer spray, and watch for the fault to appear or clear. Mechanical faults respond to controlled movement: flex a cable along its length, press connectors, tap components with an insulated tool, and vibrate assemblies. Load-related faults respond to running the equipment at the condition that triggers them. Do this methodically, one variable at a time, so that when the fault appears you know what caused it. Provocation on energized equipment carries the usual hazards and requires appropriate training, rated instruments and personal protective equipment; where it can be done de-energized, do it that way.
Logging catches what provocation cannot
When the fault refuses to be provoked, leave something recording. A data logger on the supply, a scope armed in single mode with a trigger set for the abnormal condition, a meter with a minimum and maximum capture function, or the equipment’s own diagnostic logging all work. The essential step is defining the trigger condition precisely enough that it catches the fault and not everything else. A scope set to trigger on a voltage falling below a threshold will capture the moment a rail collapses and show what preceded it, which is often the whole answer. Synchronize clocks between every instrument and the equipment itself, because correlating two records is the point and a few seconds of offset can make two related events look unrelated.
Instrument the most likely places, not the most convenient
Because you usually get one capture, where you place the instrument matters enormously. Choose points that will distinguish between the leading hypotheses rather than points that merely confirm the fault exists. If the question is whether a supply sags or a signal glitches first, record both on the same timebase so their order is unambiguous. If the question is whether the disturbance comes from upstream or from a local load, record voltage and current together, since voltage falling while current rises points to a local load. Two channels chosen to discriminate are worth more than four chosen for convenience. Note the connection points precisely, because a later capture must be taken the same way to be comparable.
Avoiding the false conclusion
The greatest hazard in intermittent work is declaring victory after a change that coincided with a quiet period. Reseating a connector, cleaning a contact or replacing a plausible part often precedes a few good days regardless of whether it fixed anything. Guard against this by defining in advance what success looks like: the fault must stay absent through the conditions that previously provoked it, for a period longer than the longest previous gap between occurrences. Where possible, confirm by reproducing the fault before the repair and failing to reproduce it after. And if several changes were made at once, the record should say so rather than attributing the fix to whichever one seems most satisfying.
What to check before you act
- Write down when the fault occurs and what else is happening at the time.
- Provoke with heat, cold, flexing and tapping, changing one variable at a time.
- Leave an instrument armed with a trigger that defines the fault condition.
- Define in advance how long the fault must stay away to count as fixed.
Common questions
How long should I log before giving up?
Longer than the longest previous interval between occurrences, and long enough to include the conditions the pattern points to. Ending a log early is the most common wasted effort.
Is tapping components a legitimate technique?
Yes, with an insulated tool and appropriate precautions. It finds cracked joints and failing connectors that no static measurement will reveal.
The practical takeaway
Stop waiting and start causing. Collect the pattern, provoke with one variable at a time, and where the fault will not cooperate leave an instrument triggered on the condition itself with clocks synchronized. Then hold yourself to a defined standard of proof before calling it repaired.
References and further reading
- NFPA 70E: standard for electrical safety in the workplace
- CDC/NIOSH: electrical hazards and safe work practices
- CPSC: home electrical safety guidance
Numerical scenarios are illustrative unless identified otherwise. Follow the exact product instructions; component ratings and local installation requirements can differ.
