Fault Only Under a Named Combined Condition: Reproduce Decision Tree
Why this matters
The hardest faults are the ones that only appear when several conditions stack at once: hot day and high load and a particular cycle, or peak demand and a specific fixture and low supply pressure. On a normal service call, none of those conditions is present, the unit tests fine, and the technician leaves having "found nothing." The customer is then certain the problem is being ignored. A combined-condition fault is not an absence of a fault; it is a fault gated behind a specific combination. The work is to extract the customer's description of when it happens, name each contributing condition, and deliberately reproduce the combination instead of testing under convenient steady-state. This applies across HVAC, plumbing, electrical, and appliance faults.
Symptom presentation
The complaint is conditional: "only when it is really hot," "only when the dishwasher and the AC are both running," "only first thing in the morning," "only when it has been running a while." The unit performs normally during the visit because the gating conditions are not all present. The temptation is to record "could not duplicate" and move on, which guarantees a callback when the conditions return.
The customer's narrative of "when" is the most valuable diagnostic asset you have. It is the recipe for reproduction.
Quick checks
Turn the complaint into a named condition set:
- Extract every "when": time of day, weather, what else was running, how long it had been operating, what the customer was doing. Write each as a separate condition.
- Separate triggers from coincidences: a condition that is always present at failure is a candidate; one that is sometimes present is weaker.
- Identify the physics each condition stresses: heat raises pressures and resistances; combined load drops voltage or pressure; long run time builds thermal soak; cold start changes clearances and viscosities.
- Decide what is reproducible on site and what requires monitoring over time because you cannot manufacture the weather or the customer's full demand profile.
Isolation tree
Can I name every gating condition from the complaint?
- No: interview further or leave instrumentation. You cannot reproduce a combination you have not fully named.
- Yes: proceed to staged reproduction.
Can I reproduce the conditions on site, one at a time, then stacked?
- Apply each named condition individually first (load, heat, run-time, combined draw). If the fault appears under a single condition, you have simplified the problem. If not, stack them in the order the customer reported.
Does the fault appear when the full combination is present?
- Yes: you have reproduced it. Now hold the combination and isolate which element pushes the system past its limit.
- No: a condition is still missing or mischaracterized. Recheck the interview, or instrument for the real event.
Is the gating condition a single weak element exposed by stress, or a true interaction?
- A single marginal part that only fails under stress (a connection that opens when hot, a valve that leaks under peak pressure) is the common case. A true multi-factor interaction is rarer but real. Distinguish by whether removing any one condition stops the fault.
Can I not reproduce it on site at all?
- Move to monitoring: data loggers, max/min capture, thermal records, or scheduled return during the named condition. Do not declare "no fault found"; declare "fault gated behind conditions not reproducible on this visit" and set up to catch it.
When deliberately reproducing a fault, you are intentionally stressing a system toward its failure point. Do this only within safe operating limits, with isolation and protection in place. Forcing combined high-load, high-heat, or high-pressure conditions to chase a fault can cause real damage or injury if pushed past rated limits.
Confirming diagnosis
Confirmation means the fault appears on demand when you assemble the combination and disappears when you remove one element:
- Reproduce the failure under the named stack.
- Remove one condition at a time and identify which removal stops it; that is the limiting factor.
- Tie the limiting factor to a measurable, marginal component or supply condition.
A fix is confirmed only when the system survives the reproduced combination, not just the convenient idle test that hid the fault originally.
Cross-trade examples
The combined-condition trap recurs across trades:
- HVAC: a system that only loses cooling on the hottest afternoons, after running for hours, at peak load. Heat raises head pressure, long run-time builds thermal soak, and high load pushes a marginal capacitor or contactor past its limit. None of those is present on a mild-morning visit.
- Plumbing: a fixture that only runs weak when the irrigation and the laundry both draw at once. The combined demand drops supply pressure below what the fixture needs; each draw alone is fine.
- Electrical: a connection that only opens when it is hot and loaded, after the circuit has carried current long enough to warm a loose terminal into intermittent contact. A cold continuity check reads perfect.
- Appliance: a unit that only faults on a long cycle once a heating element and a motor are both demanding from a marginal supply.
In each, the convenient idle test hides the fault, and only the assembled combination exposes it.
Next steps
Document the exact condition stack that reproduces the fault and the measurement at the limiting element, so the repair can be verified under the same combination and a future tech is not fooled by a clean idle test. Where the conditions cannot be reproduced on site, leave instrumentation and document the monitoring plan rather than closing as no-fault. The discipline that distinguishes a real diagnosis here is refusing to accept steady-state "fine" as proof against a conditional complaint.
References
- ISO 13379-1, Condition monitoring and diagnostics of machines, on intermittent and load-dependent fault interpretation.
- ISO 17359, Condition monitoring and diagnostics of machines, general guidelines, on monitoring when faults are not reproducible on demand.
- ISO 14224 on recording operating context and conditions at the time of failure.
- ACCA technical guidance on diagnosing weather- and load-dependent HVAC faults under representative conditions.