It Works Then Stops Then Works Decision Tree

Why this matters

The intermittent fault is the one that humbles experienced techs. It works fine while you are there, fails the moment you leave, and runs perfectly again by the time you come back. You cannot fix what you cannot catch in the act. The instinct is to replace a likely part and hope, which is a coin flip that usually loses and earns a callback. The key insight is that intermittents are not random - they are triggered. Something brings the fault on and takes it away: heat, vibration, moisture, load, time, temperature. Find the trigger and the intermittent becomes a repeatable fault you can actually diagnose. This tree walks the search for the trigger.

Start here: stop trying to fix it, start trying to trigger it

Do not replace anything yet. Your first job is to make the fault happen on command. An intermittent you can reproduce is a normal diagnosis; an intermittent you cannot is a guess. Gather the pattern from the customer and your own observation, then chase the trigger.

Ask and record:

• When does it fail? After running a while, on cold mornings, during certain weather, at a time of day, under heavy load.
• What brings it back? Cooling off, time, a power cycle, drying out, nothing obvious.
• Is there a physical event nearby? A bump, a door slam, another appliance kicking on, a vibration.

Step 1: Is it heat-triggered?

The most common intermittent trigger by far. Components and connections behave differently hot than cold. A connection that is fine when you arrive opens as it warms; a part that tests good cold drifts out of spec hot; a thermal protector trips after a run and resets after a cool-down.

• If the fault appears only after running a while and clears after cooling, it is heat-triggered. Let the unit run to failure, then test while it is hot - that is the only state where the fault is present. Testing cold tells you nothing here.
• If heat is not the trigger, move on.

Step 2: Is it vibration- or movement-triggered?

A fault that comes and goes with a bump, a slam, a nearby motor starting, or just the equipment's own running vibration points at something mechanical and loose. A cracked solder joint, a loose terminal, a marginal connector, a wire chafed nearly through, a hairline break.

• If tapping, flexing, wiggling a harness, or a nearby vibration source brings the fault on or off, you have a loose or broken connection. Trace and secure it. The fault location is wherever your tap reproduces it.
• If movement does nothing, move on.

Step 3: Is it moisture- or weather-triggered?

Faults tied to rain, humidity, time of day, or season are usually moisture or temperature acting on a connection or component. Water bridges where it should not, a seal weeps under certain conditions, condensation forms, a connection corrodes and conducts only when dry.

• If the fault tracks weather, humidity, or wet conditions, hunt for water intrusion, condensation, or corroded connections. Recreate the wet condition if you safely can.
• If weather is not the trigger, move on.

Step 4: Is it load- or demand-triggered?

Some intermittents appear only under heavy load or peak demand and vanish at idle. A supply that sags only when everything is running, a connection that carries light load fine but heats and opens under heavy current, a component that fails only at full output.

• If the fault shows only under load, recreate the load and test there. A unit tested at idle will pass while hiding a load-only fault.
• If load is not the trigger, move on.

Step 5: Is it time- or cycle-triggered?

A fault that appears after a set runtime, a number of cycles, or at a particular point in the operating sequence points at timing, sequencing, or a control that misbehaves only at a certain step.

• If the fault is tied to a stage of operation or a runtime, watch the unit through the full sequence and catch where it drops. The failure point in the cycle is your clue.
• If no time or cycle pattern emerges, move to monitoring.

Step 6: If you cannot trigger it, monitor and capture

When no trigger reveals itself in a reasonable visit, do not guess. Set up to capture the fault when it happens: leave monitoring in place where practical, instruct the customer on exactly what to note and photograph next time, and schedule a return for the conditions when they actually see it. Documenting "could not reproduce, here is the capture plan" is a real, honest outcome - far better than a hopeful part swap.

The pattern across trades

Heat, vibration, moisture, load, timing, then monitor. An electrical connection that opens hot, a control that fails after a run, a pump that loses prime only on hot days, an appliance that quits under heavy load: same family, same hunt for the trigger.

References

• Trade-standard intermittent-fault diagnostic practice across the trades.
• Reliability engineering guidance on trigger-based failure analysis.
• See related: Confirming the Complaint Before You Start; Documenting What You've Ruled Out; The One-Variable-at-a-Time Rule.