The Fault a Specific Condition Triggers Decision Tree

Why this matters

The fault that only shows up sometimes is the one that makes a tech look bad. It works fine while you are there, you find nothing, you leave, and it acts up again the moment you are gone. The trap is treating a condition-triggered fault like a random one. It is not random. It has a trigger, and the trigger is a specific condition: a temperature, a load, a time in the cycle, a vibration, a humidity level. Find the condition that brings it out and you have found the fault, because reproducing it on demand is most of the battle. This tree maps the fault to its trigger so you can catch it on purpose instead of hoping.

Safety first

A condition-triggered fault can be a condition-triggered hazard. If the fault that appears under heat, load, or a specific cycle point involves arcing, overheating, gas, or pressure, do not keep forcing the trigger on a live system. De-energize and verify dead, relieve pressure, or shut off the fuel, then investigate the triggered condition safely. Repeatedly driving a system into a fault that arcs or overheats to "catch it in the act" can turn an intermittent into a failure while you stand there. Make it safe, then reproduce under control.

Start here: what condition brings the fault out

The whole method is identifying the trigger. Ask the customer and watch the pattern: when, exactly, does it fail. Sort the trigger into a family.

• Temperature - fails only when hot, or only when cold.
• Load - fails only under heavy demand, only at peak, or only at startup inrush.
• Time or cycle - fails after running a while, or at a specific point in the operating cycle.
• Movement - fails with vibration, or when something is bumped (see the related fault-that-moved method).
• Moisture or weather - fails only when humid, wet, or after rain.

Nail the trigger and you have aimed the whole diagnosis. A fault that "happens randomly" almost always has a trigger the customer has not connected to it yet.

If heat triggers it

Branch A: fails when hot. Something changes as the system warms: a connection expands and opens, a marginal component drifts out of spec, a thermal protector approaches its limit, a fit binds as metal grows. Run the system until it reaches the failing temperature, then test at temperature, not cold. The fault is invisible cold and obvious hot.

Branch B: fails when cold. Contraction opens a marginal joint, lubricants thicken, seals shrink, condensation freezes, a fit binds the other direction. Test cold-started, before things warm and the fault hides.

If load triggers it

Branch C: fails under heavy load. A connection that carries fine at light current sags or heats under heavy draw, a supply that holds at idle drops under demand, a marginal component that passes unloaded fails loaded. Read under real load, not at rest. Many faults live only under draw.

Branch D: fails at startup. The inrush at start is the heaviest moment for many systems. A weak start component, a marginal supply, or a high-resistance connection shows only during that surge. Watch the start, not the run.

If time or cycle triggers it

Branch E: fails after running a while. This is usually heat (see heat branch) or a slow drift, a clogging path, a filling reservoir, a thermal buildup. Let it run to the failure point; do not test it in the first minute when it is still fine.

Branch F: fails at a specific cycle point. The fault appears at one stage every cycle: a particular valve position, a certain stroke, a specific mode change. That stage is your suspect. Watch the cycle and pin the exact moment it faults.

If movement triggers it

Branch G: vibration or bump. A loose, cracked, or barely-contacting part that lets go when shaken. Reproduce by vibrating or tapping the suspect area (safely), and read the related fault-that-moved method for the contact-fault tree.

If moisture triggers it

Branch H: humidity or wet. Moisture bridges a contact, swells a part, corrodes a connection, or overwhelms a drain. The fault tracks the weather. Note the conditions and, where you can, reproduce with controlled moisture at the suspect point.

If you cannot identify a trigger

Branch I: no trigger found. If careful questioning and observation find no condition that correlates, you may be dealing with a truly random intermittent (a deeply marginal part or a stack-up; see the related tolerance method) or a trigger you have not caught yet. Have the customer log exactly when it happens, and leave instrumentation if you can. A logged pattern usually reveals the trigger you missed live.

Confirming the diagnosis

Confirm by reproducing the fault on demand. Once you know the trigger, force the condition and watch the fault appear, then test under that exact condition to find the component that fails there.

• The fault appears reliably when you apply the trigger - you own it. Test at that condition and find the failing part.
• The fault will not reproduce even under the suspected trigger - the trigger is wrong or incomplete. Re-question the pattern; the real trigger is something else.

A condition-triggered fault is proven fixed only when you apply the same trigger after the repair and it no longer faults. Reproduce, repair, then reproduce again to verify.

Next steps

Once the trigger reveals the failing component, repair it and re-apply the trigger to confirm the fix holds (see the related verification method). Document the trigger plainly: "fails only after 20 minutes run, traced to connection opening when hot, remade joint, ran to temperature and confirmed." That record turns the next "it does it randomly" call into a known pattern. When you genuinely cannot reproduce it on site, say so and set up a log rather than guessing at a part.

References

• ISO 13379-1 (Condition monitoring and diagnostics of machines; intermittent and condition-dependent faults).
• NFPA 70B (Standard for Electrical Equipment Maintenance; thermal and load-dependent fault diagnosis).
• ASHRAE Handbook (HVAC Applications; temperature- and humidity-triggered fault behavior).
• See related: The Fault That Moved After You Touched It Decision Tree; Proving It's Fixed Before the Customer Tests It.