How to Set Up an Extended Run Test Without Camping On Site

Why this matters

An eight-hour run test is the right diagnosis and the wrong use of a tech. Sitting with a machine for a full shift consumes a day of capacity to collect maybe fifteen minutes of useful data, and no small shop can absorb that on a job that has already produced two no-fault-found visits. The alternative is not to skip the test. It is to run it with the tech somewhere else - which is a different problem than passively monitoring a fault that happens on its own schedule, because here you are deliberately commanding equipment to operate outside its normal control while nobody qualified is standing next to it. That deserves a real setup, not a shrug and a data logger.

Lead with the decision that comes before any of it: is this safe to leave running

Do not design the instrumentation first. Decide whether unattended extended operation is acceptable at all, and be willing to answer no.

Leave it attended, or do not run it, when any of these are true:

• Any safety in the chain is questionable, defeated, or has been reset more than once. A protective device that has already tripped repeatedly on this equipment is the thing you are testing. You do not leave that alone.
• Combustion is involved and the venting has not been verified this visit. Flue integrity, combustion air, and spillage checks are prerequisites, not follow-ups.
• The failure mode you suspect could produce an uncontained release: water, steam, fuel, pressurized fluid, or a fire path. The whole point of the test is to reach failure, so plan the failure as though it will happen, because you are trying to make it happen.
• Nobody responsible is on the premises for the full run. An empty building plus equipment you told to run past its normal pattern is a risk transfer you should not be making silently.

If you clear all four, you still install the shutdown path described in step 3 before you leave. If you cannot clear them, the honest options are an attended run, a shorter instrumented run that stops short of failure, or handing the run to the customer under their normal operating pattern with a log sheet.

Step 1: Get explicit, written permission for what you are about to do

The customer's authorization for a diagnostic visit is not authorization to run their equipment continuously for a shift. Put four things in writing before you set it up, in plain language:

• What will run, for how long, and outside what normal control (for example, "the unit will run continuously rather than cycling with the thermostat, for up to 8 hours").
• That the test is designed to reach a failure, and that the failure may be the same one they are already experiencing.
• Who on their side is present, what they watch for, and the single number they call.
• What it costs them in hours: your setup time, your return time, and the run itself, expressed as tech hours rather than a vague "we will be monitoring."

This is not paperwork for its own sake. The most common way an extended run test goes wrong commercially is a customer who believed you were "checking on it" and later learns you deliberately ran their equipment to failure. Say it out loud first and the same event is a diagnosis instead of a complaint.

Step 2: Instrument for the shape, not the moment

A passive fault monitor answers "did it happen and when." An extended run test answers "what was the system doing on the way there," so you need continuous logging, not just a min/max capture.

Log at minimum three channels:

1. The suspect parameter - the temperature, pressure, current, or flow you believe drifts toward the failure.
2. A reference channel - ambient at the equipment, or supply voltage, or inlet condition. Without it, a rise is unreadable, because you cannot tell the equipment's contribution from the room's.
3. A run-state channel - something that proves the equipment was actually running the whole time. A current clamp on the main load is the usual answer. Without it, a fault at hour five is ambiguous: did it run five hours, or did it stop at hour two and sit?

Sample every 30 to 60 seconds for a run of several hours. That is fine enough to show the curve and to place the failure within a minute, and coarse enough that memory is not a concern on any modern logger. Point a camera at the display or indicator panel if the equipment reports faults visually and does not log them.

Step 3: Install a way to stop it that does not require you

Every unattended run needs an abort path that works when you are 40 minutes away.

• Set a hard limit on the logging device if it supports an alarm output, at a value below the damage threshold and above normal operation.
• Brief one on-site person on exactly what to do: which switch or breaker to operate, in what order, and that they should do it without calling you first if they see, hear, or smell the listed conditions. Write those conditions on a card and tape it to the equipment.
• Leave the normal controls capable of stopping the unit. If your forced-run method involves jumpering a control, the safeties must still be able to shut it down, and you must be able to describe how in one sentence.
• Label everything you touched. A tag on every jumper, clamp, and probe saying what it is, that a test is in progress, your name, and the time it must be removed. Untagged test leads found the next morning by a different trade is how someone gets hurt.

Step 4: Split your attendance to the two moments that matter

You do not need to be there for the middle of the run. You need to be there at the start and near the expected failure.

• Attend the first 30 to 45 minutes. This is where the fast masses settle and where a setup error announces itself. Confirm all three channels are logging real numbers, not zeros or open-circuit values, before you drive away.
• Return roughly 45 minutes before the expected time-to-fault. The failure-moment readings are the payoff and they are perishable, often gone in under two minutes of cooling. Arriving after it tripped costs you the visit.

If the time-to-fault is uncertain within a couple of hours, arrange for the on-site contact to call the moment it happens and use the logged data as your fallback. A logged curve plus a timestamped fault is worth most of an attended failure; an unlogged run with a phone call is worth very little.

A worked example, carried through

A refrigeration-side system in an occupied commercial space quits about 6 hours into the day, roughly four days out of five. All values below are illustrative.

Attended cost of a full test: 8 run hours, which as pure attendance is a full tech day. The split-attendance version costs 0.75 hours on setup and start-up verification, a drive back, 1.25 hours covering the failure window, and 0.5 hours on teardown and data pull. That is 2.5 attended hours against 8, so roughly 30 percent of the attended cost, and the tech ran two other calls in the gap.

The setup: a two-channel logger on suspect temperature and ambient at the equipment, a clamp-on current logger on the main load for run state, all sampling at 45 seconds. Written authorization for a continuous run up to 8 hours, an abort card taped to the panel listing three conditions, and the shop's supervisor briefed on the disconnect.

Day one result: the fault hit at 5 hours 40 minutes, and the tech arrived at the 5 hour 15 minute mark with 25 minutes to spare. The logged curve showed the suspect temperature climbing steadily through hour 4, flattening, then climbing again from hour 5 - the late second rise that is the signature of something changing partway through the run. The current trace showed the load dropping by about a fifth at the same moment the second rise started, which located the change: a cooling airflow component had dropped out, not the thing that eventually tripped.

Without the run-state channel, that fifth-of-load drop would have been invisible and the trip would have looked like the primary fault. The reference channel earned its place too: ambient rose 9 degrees F over the run, so the raw suspect temperature overstated the equipment's own contribution by that much, and only the delta showed the true shape.

How to verify you set this up right

Before you leave the site, pull five minutes of data off each channel and look at it. Not the display, the actual recorded data. A channel reading a flat unchanging value, a channel reading zero, or a timestamp in the wrong time zone are all common and all fatal, and every one of them is a five-minute fix while you are standing there and a wasted day once you have driven off.

Then answer out loud: if this fails in the worst way I can imagine at hour four, what happens? If the answer includes water on a floor, an unvented combustion product, or a fire, go back to the safety gate and change the plan.

What changes the answer

• Residential and occupied spaces often cannot tolerate a forced continuous run for comfort reasons. Hand the run to the customer under their normal pattern and instrument it; you lose control of the duty cycle but keep the data, and the run-state channel becomes essential rather than optional.
• If the time-to-fault is longer than a work day, the split-attendance model breaks down and the test becomes multi-day passive monitoring, which is a different method with different tooling.
• If the equipment is under warranty or a service agreement with defined operating conditions, confirm that a forced continuous run is not an excluded condition before you set it up.
• If the customer cannot provide a responsible person on site for the run, drop back to an attended test or a shorter one. The unattended model depends on that person existing, and no amount of instrumentation substitutes for them.

References

• OSHA general industry guidance on machine guarding, lockout, and test-equipment use
• NFPA 70E and NFPA 70B guidance on securing test equipment on energized systems
• Manufacturer documentation on continuous-duty limits and forced-operation procedures
• See related: How to Test a System That Only Fails After Hours of Running; The Extended Run Test SOP; How to Log Temperature Drift Across a Full Duty Cycle