The Fault That Tracks With Usage: Decision Tree
Why this matters
The equipment is fine when you test it and fails when the customer runs it hard. The fault tracks with use: it shows up after a long run, under heavy load, at high demand, or once things heat up - and it hides the moment you arrive and run a quick check cold. This is the fault that gets misdiagnosed as "no problem found" and sent home, only to fail again that afternoon. A fault that scales with usage is telling you exactly where to look: at whatever gets stressed when the work gets heavy. This tree turns "it only does it when I use it" into a diagnosis.
Start here: does the fault under load create a hazard
A fault that appears under heavy use can be more dangerous than a steady one, because it strikes when the equipment is working hardest and the customer least expects it. If the usage-driven fault involves a safety system - something that overheats under load, an electrical fault that appears at high current, a pressure issue that shows at peak demand - treat the loaded condition as the dangerous one and control it before testing. Do not run the equipment to full load to reproduce a fault if reaching that load could injure someone or damage the equipment. De-energize, limit, or stage the test safely, and if you cannot reproduce the fault without creating real risk, advise against running the equipment hard until it is repaired. The fact that it only fails under load does not make it minor; it makes it ambush-prone.
For everything below, assume reproducing the load is safe.
First move: characterize the usage relationship
"Tracks with usage" covers several different patterns. Get specific about which one.
- Time-under-load. Fails after running for a while, fine on a short run. Points at heat building up or a component fatiguing as it works.
- Magnitude of load. Fails under heavy demand, fine under light. Points at a part that is marginal and only exposed near its limit.
- Cumulative cycles. Fails after many starts and stops, fine when left running. Points at something stressed by the transition, not the run.
- Frequency of use. Worse the more often it runs, recovers when rested. Points at inadequate cooling, recovery, or capacity between uses.
Write it as a sentence: "Fine for the first ten minutes, faults once it's hot, recovers after it cools." That sentence aims the whole diagnosis.
Second move: map the usage pattern to the stress
Each usage pattern stresses something specific. Match the pattern to the mechanism.
| Usage pattern | What gets stressed | Common fault mechanism |
|---|---|---|
| Long runs | Heat accumulates | Thermal expansion opens a joint, a part drifts hot, a cutout trips |
| Heavy load | Capacity and current peak | Marginal component fails near its limit, voltage sags, undersized part overheats |
| Many cycles | Start/stop transitions | A connection fatigues, an inrush stresses a weak part, wear at the moving point |
| High frequency | Recovery time shrinks | Inadequate cooling or capacity between runs, accumulation that never clears |
The pattern names the stress, and the stress names where to look. A heat-after-running fault sends you to thermal expansion and rising resistance; a heavy-load fault sends you to whatever is sized closest to its limit.
Third move: reproduce the load and measure while it fails
The cardinal rule of usage-tracking faults: do not diagnose them cold. Recreate the usage and take your readings while the fault is actually present.
- Run it the way the customer does. Long enough, hard enough, or often enough to bring the fault on. A quick cold check is exactly the test that misses this fault.
- Measure during the failure, not before or after. A reading taken while the equipment is hot, loaded, and faulting is worth ten taken when it is behaving. Catch it in the act (see related: the recurring seasonal fault).
- Watch what changes as you load it. A reading that drifts as the system heats or loads up points straight at the stressed component. The thing that moves with the load is usually the thing failing under it.
If you cannot reproduce it on this visit, do not call it fixed or fine - leave instrumentation, get the customer's exact trigger, or schedule a return under real conditions.
Fourth move: confirm the stressed part, then verify under load
Once the usage test points at a component, confirm it and then prove the repair against the real condition.
- Test the suspect under the stress, not at rest. A part that passes cold and fails hot must be judged hot. A bench test at room temperature can clear a part that dies under load.
- Fix the cause, not just the part if the part is a casualty. If a component fails under load because something upstream is undersized or restricted, replacing it just buys time. Address why the stress exceeded what the part could take.
- Verify by recreating the failing condition after the repair. Run it hot, run it heavy, run it long - whatever brought the fault on. A fix that survives the original usage is real; a fix that only passes a cold recheck is unverified (see related: the fix worked for the wrong reason).
Recap
- If the loaded condition is a hazard, control it before testing and never force a dangerous load just to reproduce the fault.
- Characterize the usage relationship - time, magnitude, cycles, or frequency.
- Map the pattern to the stress, and let the stress point you at the mechanism.
- Reproduce the load and measure while the fault is happening, not cold.
- Judge the suspect part under the stress, fix the real cause, and verify by recreating the failing condition.
The judgment to bank: a fault that only shows under use will only be found under use. Stop testing it cold and confirming it is "fine." Load it the way the customer does, measure it in the act, and the fault stops hiding.
References
- Trade-standard practice for diagnosing load-related and intermittent faults
- Manufacturer documentation for component ratings and operating-condition limits
- OSHA guidance on safe testing under load and verifying safety-device function
- See related: The Recurring Seasonal Fault (decision tree); The Fix Worked for the Wrong Reason (decision tree)