What Changes Inside a System During a Long Run Cycle

Why this matters

Techs are trained to take a reading and compare it to a spec. What almost nobody is taught is that the reading has a shape over time, and that most of the shape happens after the twenty minutes you were on site. A system at hour six is not the same machine you measured at minute ten: metal has moved, oil has thinned, resistances have climbed, a filter has loaded, and a component is sitting at a temperature it never reaches during a short call. Knowing what actually drifts, and how fast, is the difference between "it tested fine" and knowing exactly which reading you had no business trusting.

Nothing settles at the same time, and that is the whole idea

Every mass in a system approaches its final operating temperature on its own schedule, governed by how much heat it holds versus how fast it can shed it. That approach is exponential, not linear, and the shorthand for it is the time constant: the interval it takes to close about 63 percent of the gap between where a part started and where it will end up. After two of those intervals you are at roughly 86 percent of the final rise, after three about 95 percent, after four about 98 percent. Treat "settled" as three to four time constants and you will not be far off.

The practical consequence is that a system has fast masses and slow masses running at once:

So at the 20 minute mark, the small parts are essentially at their final temperature and the enclosure has barely started moving. Any fault that lives in the enclosure temperature, or in the difference between a hot component and a slowly warming surround, is invisible to you at that moment. That single table explains most of what people call an intermittent.

Electrical properties that drift with the run

• Conductor and winding resistance rises with temperature. For copper, resistance climbs roughly 0.39 percent per degree C of rise, which is a large enough effect that the industry uses it backwards as a measurement method (see the article on measuring a component at operating temperature). A winding that reads clean cold reads meaningfully higher hot, and that is normal, not a fault. What is a fault is a connection whose resistance climbs far more than the conductor around it.
• Contact resistance and heating feed each other. A slightly loose or lightly corroded connection dissipates heat in proportion to resistance times current squared, that heat oxidizes the joint further, and the resistance climbs again. This runaway is slow, which is exactly why it shows up as an hour-five problem rather than a minute-five problem.
• Capacitors lose capacitance and gain internal resistance as they heat. A component that measures inside tolerance cold can drift outside it at operating temperature, and the equipment it supports fails only once the drift crosses the working threshold.
• Insulation resistance falls as temperature rises. Marginal insulation that passes a cold test can leak enough hot to trip a protective device.

Mechanical and fluid properties that drift

• Clearances close. Bearings, bushings, shafts, and guides are machined to tolerance at a reference temperature. As parts heat at different rates, running clearance shrinks. Something that turned freely at minute ten can drag at hour four, and the drag makes more heat.
• Lubricant viscosity drops. Warm oil or grease flows more freely, which is usually good, but past a point the film gets thin enough to stop separating surfaces. A bearing can be quiet cold and noisy hot for that reason alone.
• Fluid density falls and vapor pressure rises. Hot liquid is easier to boil locally, which is why pumps that are quiet on a cold start start rattling deep into a long run. Nothing mechanical changed; the fluid did.
• Gas volumes expand and closed sections pressurize. Any section that is isolated while its contents heat will climb in pressure with no leak, no fault, and no external cause.

Accumulation: the changes that do not reset overnight

Everything above resets when the system cools. This group does not, and separating the two is the most useful cut in this article.

• Filters, strainers, and screens load progressively. Restriction rises across the run, and it rises again the next run from where it left off.
• Condensate and drainage volumes accumulate. A pan or trap that handles a 45 minute cycle can overflow on a 6 hour one, and the failure is a function of total volume, not temperature.
• Debris migrates. Lint, dust, and shed material move with airflow and settle where the flow slows, which is often the cooling path for the thing that later overheats.
• Consumables deplete. Fuel, chemical, refrigerant charge in a leaking system, and fluid levels all fall with run time.

The diagnostic test between these two families is the warm restart. If a partial cool-down and restart makes the fault come back much faster, you are in the thermal family, because you restarted partway up the curve. If a partial cool-down and restart makes no difference to time-to-fault, you are in the accumulation family, because cooling does not un-load a filter.

Control behavior changes too

Long runs change what the controls are doing, not just what the hardware is experiencing. Staged equipment brings on a second or third stage only after the first cannot keep up, so a component that only energizes on a late stage may not run at all during a short call. Anti-short-cycle timers, defrost or regeneration routines, and self-test intervals fire on schedules measured in hours. A control that adapts to conditions is by definition a different control at hour six than at minute ten. When a fault appears at a repeatable run hour rather than a repeatable temperature, look at what the control does at that hour before you look at hardware.

A worked example, carried through

A pumped circulation system runs continuously through a shift and develops a noise and a nuisance trip roughly five hours in. All values are illustrative.

Start with the shape. Fluid temperature at the pump reads 12 degrees F above its cold starting value at 30 minutes, 26 above at 2 hours, 33 above at 4 hours, and 35 above at 5 hours. The curve is flattening, so the fluid reached near-equilibrium at about the 4 hour mark. If the fault were driven purely by fluid temperature it should have appeared at 4 hours and then stopped getting worse. It did not; it appeared at 5 hours and got worse after.

Next check the slower mass. The equipment room air is 4 degrees F above its morning value at 2 hours and 15 above at 5 hours, still climbing. That is the slow mass, and its timing lines up with the fault. So the driver is the surround, not the fluid.

Now separate thermal from accumulation. Shut down for 30 minutes, restart. The fault returns in 1.5 hours instead of 5.0. Time-to-fault dropped to 30 percent of the cold-start value from a partial cool-down, which puts this firmly in the thermal family. If it had returned at 5 hours again, the strainer would have been the first suspect.

Finally, name the mechanism. Fluid near equilibrium, room still climbing, and a noise that arrives with the room rather than the fluid points at a component whose cooling depends on room air rather than on the process fluid. The fix targets the room's heat rejection, not the pump.

Note what a tech who took one reading at 30 minutes would have concluded: fluid 12 degrees above cold, room 2 above, everything quiet, no fault found. Every number was accurate. Every number was also taken before either mass had gone anywhere.

How to verify you are reading drift correctly

Take the same measurement at least three times across the run and write down the interval between readings each time. One reading is a value; two are a direction; three are a shape, and the shape is what carries the diagnosis. Always log ambient at the equipment alongside every reading, because a 20 degree F rise means something completely different in a 60 degree F room than in a 95 degree F one, and the delta is the number that travels between visits.

The failure mode to watch for in yourself is anchoring on the first reading. A tech who records enclosure temperature once at hour four and once at hour six, and sees a 3 degree F difference, will call it stable - but if the first four hours produced 40 degrees of rise, a curve still moving 3 degrees per two hours at hour six has not settled, it is crawling toward a number nobody has measured yet.

What changes the answer

• A system that cycles rather than running continuously never reaches the equilibrium described here, and the relevant variable becomes duty cycle percentage. Two units running 8 hours can be at wildly different temperatures if one ran 30 percent of the time and the other 90 percent.
• Outdoor or unconditioned equipment has ambient moving underneath the run curve, so the raw temperature is nearly useless and only the rise above ambient is readable.
• A large thermal mass in the process itself (a big tank, a slab, a full vessel) can dominate everything above and stretch the settling time to a full day, which means a same-day extended run test may not reach the failure condition at all.
• Very short time constants everywhere, as in small light-duty equipment, mean everything settles inside the first 45 minutes and a genuine hours-later fault is far more likely to be accumulation or a control routine than heat.

References

• Manufacturer documentation on continuous-duty ratings and allowable temperature rise
• NFPA 70B, recommended practice for electrical equipment maintenance
• Trade-standard practice for temperature-rise testing and thermal equilibrium
• See related: How to Test a System That Only Fails After Hours of Running; How to Log Temperature Drift Across a Full Duty Cycle; Heat-Soak Faults Versus Cold-Start Faults