Catching the Fault That Hides When You Arrive: Decision Tree

Why this matters

The unit works perfectly while you stand in front of it. The customer swears it failed an hour ago. This is the call that wastes the most time and damages the most trust, because if you leave saying "could not find anything," you will be back, for free, and the customer will doubt you. An intermittent fault is not a no-fault. It is a fault you have not recreated yet. The skill is a method for cornering a problem that hides on arrival, so you fix the actual fault instead of declaring it gone.

Safety first if the hidden fault is a hazard

If the reported symptom involves a real hazard, treat it as present even when you cannot see it.

• Reported arcing, burning, shock, or a trip: the absence of the symptom now does not mean it is safe. De-energize and inspect for evidence (scorching, heat marks, a tripped device) before you energize and probe.
• Reported gas or fuel smell: do not dismiss it because you do not smell it now. Check for a leak with proper method before you create an ignition source.
• Reported overheating or over-pressure: look for the evidence of it (discoloration, a relieved device) even if it reads normal at the moment.

An intermittent hazard is still a hazard. Rule it out with evidence, not with "seems fine now."

Start here: get the story before you touch anything

The customer watched the fault happen and you did not. Their account is your best data.

• When does it happen? Time of day, after how long running, hot or cold, wet or dry. A pattern in the "when" points straight at a cause family.
• What is it doing when it fails? Starting, stopping, under heavy load, idling, after a cycle. The condition at failure is where you will recreate it.
• What is different when it works versus fails? Weather, usage, what else is on. The difference is the lever.
• How often, and is it getting worse? A fault that is becoming more frequent is easier to catch and more urgent.

Write down the conditions at failure. You are going to try to recreate them.

Look for the evidence the fault left behind

An intermittent fault that is hiding now usually left a mark when it last happened. Hunt the mark.

• Heat evidence: discoloration, a melted or distorted surface, a smell, a warm component. Heat marks persist after the unit cools.
• A tripped or latched device: a protection that fired and is holding, a code stored in memory, a manual reset that was pressed.
• Corrosion or staining at a connection: a halo at a terminal or a stain under a fitting marks an intermittent leak or a warming joint. See related: Reading Rust and Corrosion Patterns.
• Witness marks of movement: a connection that has been working loose, a part that has shifted, rub marks.

The evidence often names the fault even when the unit will not reproduce it on demand.

Recreate the failure conditions on purpose

Do not wait passively for the fault. Reproduce the conditions the customer described.

1. Run it the way it fails. If it fails hot, run it long enough to get hot. If it fails under load, load it. If it fails on the third cycle, run it through three. Idle observation rarely catches an intermittent.
2. Stress the suspected family while it runs, where safe. Flex a wire, tap a component, vary the load, change the condition that differs between working and failing.
3. Watch a live reading, not just the unit. A meter or gauge left on the suspect parameter while you provoke the fault catches a momentary drop or spike the unit itself swallows. See related: Forcing an Intermittent to Show Itself.

If you reproduce it, you are no longer chasing a ghost; treat it as a now-present fault and diagnose normally. If you cannot reproduce it in the time you have, move to the next branch rather than guessing.

If you still cannot reproduce it

Sometimes the fault will not show in one visit. Do not swap a part on a hunch and call it fixed; that is how you end up replacing several good parts. Instead:

• Leave a witness. A data logger, a recording meter, a marked connection, or even an indicator that latches a fault, so the unit reports what happened while you were gone. See related: The Witness Mark or Data Logger Method.
• Fix the most-likely root cause you have evidence for, and say so honestly. If the evidence points at a warming joint, remake the joint. Name it as the probable cause, not a guarantee.
• Set the follow-up. Tell the customer what you did, what to watch for, and when to call. A documented probable-cause repair beats a confident wrong swap.

Recap: a hidden fault is a fault you have not recreated

1. Treat a reported hazard as present until evidence clears it.
2. Get the story: when, what condition, what is different, how often.
3. Hunt the evidence the fault left: heat, a tripped device, corrosion, movement.
4. Recreate the failure conditions on purpose, watching a live reading.
5. If it still hides, leave a witness, fix the evidenced probable cause, and set a follow-up.

The rule to bank: never write "no fault found." Write "could not reproduce today, here is what I left to catch it." One is an excuse, the other is a plan.

References

• Trade-standard practice for intermittent-fault diagnosis and data logging
• Manufacturer documentation on stored fault codes and latching protections
• See related: Forcing an Intermittent to Show Itself; The Witness Mark or Data Logger Method; The Customer Can't Reproduce It (decision tree)
• See related: Reading Rust and Corrosion Patterns