Works on the Bench, Fails Installed: Decision Tree

Why this matters

A part that tests perfect on the bench and fails the moment it is installed is one of the most frustrating calls in the trade, because the obvious test says the part is good. It is good. The fault is in what changes between the bench and the install: load, heat, vibration, supply, geometry, and the things connected to it. A tech who keeps re-testing the part in isolation will keep proving it works and never fix the call. The skill is to stop testing the part and start testing the difference.

Safety first if it fails hot, trips, or arcs

If the installed failure involves heat, a trip, arcing, or a leak that the bench test never showed, treat the install condition as a live hazard.

• Trips or arcs under install load: de-energize and verify dead before you probe. The energy is real even though the bench was tame.
• Overheats installed: shut down and let it cool. The thing that makes it run hot installed will keep cooking it.
• Leaks or over-pressures installed: relieve and verify zero before you open the connection.

Once it is safe, work the difference, not the part.

Start here: list what is different installed

The part did not change. The conditions did. Write down what the install adds that the bench did not have.

• Real load. The bench had little or none. Installed, it works against full demand, and a marginal part or setting that coasted on the bench now stalls, trips, or overheats.
• Real supply. The bench had a clean, short, known feed. Installed, the voltage may sag, the pressure may droop, the line may be long or shared.
• Heat. A hot cabinet, a sunny attic, a tight enclosure. Heat shifts readings and pushes marginal components over the edge.
• Vibration and movement. Installed equipment shakes. A connection that holds on a still bench backs out, an intermittent contact opens.
• Geometry. Orientation, line length, bends, height, and what it connects to all change once it is mounted. The bench had none of that.

The branch you follow is the difference that best matches the symptom.

If it fails under load but passes at rest

A part that is fine until it works hard is rarely the part. It is the supply or the sizing.

1. Measure the supply under load, not at rest. Voltage, pressure, or flow at the unit's own terminals while it is actually trying to run. A reading that is fine at idle and sags under load is your fault.
2. Check the feed for the install conditions: undersized conductor or line, long run, loose connection that only matters at full current or flow, a shared circuit pulling it down.
3. Confirm sizing and setting for the real load, not the bench load. A correct part set for a smaller demand will trip on the real one.

If the supply holds and the sizing is right and it still fails under load, suspect a connection that only opens under vibration or heat.

If it fails only when hot

Heat is the great revealer of marginal parts and bad joints.

• Find what gets hot first. A connection warming up adds resistance, which adds more heat, which fails the joint. The bench never ran long enough to heat it.
• Check the enclosure and ventilation. A part rated fine in open air can overheat in a sealed, hot install. That is an install problem, not a part defect.
• Let it fail, then read it immediately. A reading taken cold after a hot failure tells you nothing. Catch it at the moment it quits.

If it fails intermittently or with movement

This is the classic bench-passes, install-fails signature, and it is almost always a connection or a marginal contact.

• Wiggle test the connections with the unit running, gently, where safe. A fault that comes and goes as you flex a wire or fitting is a bad joint, not a bad part.
• Look for a connection the install added that the bench did not have: the very terminals you landed, a quick-connect, a splice, a backed-out lug.
• Suspect vibration paths. A part bolted to something that shakes will lose a marginal contact the bench never disturbed.

The honest possibility: it is the install, not the part

Hold this one the whole time. If the part passes every bench test and fails every install test, the most likely answer is that the install is the fault: a supply problem, a heat problem, a connection problem, a setting wrong for the real conditions, or an incompatibility with what it connects to. Swapping the part proves nothing if the install is what broke it, and the new part will fail the same way.

When you finally do find the difference, fix the difference, then re-run the unit installed, under load, and watch it hold. That is the only test that counts here.

Recap: test the difference, not the part

1. Make it safe if it trips, overheats, arcs, or leaks installed.
2. List what the install adds: load, supply, heat, vibration, geometry.
3. Failed under load: measure supply and sizing under load.
4. Failed hot: find what heats first; check enclosure and joints.
5. Failed intermittently or with movement: wiggle-test the connections.
6. Default to "it is the install," because the bench already proved the part.

The rule to bank: the bench tests the part, the install tests everything around it. When they disagree, the install is talking.

References

• Trade-standard practice for load testing and supply verification under operating conditions
• Manufacturer environmental and operating-condition ratings
• See related: Catching the Fault That Hides When You Arrive (decision tree); Forcing an Intermittent to Show Itself
• See related: Why a New Install Fails Early (the usual suspects)