How to Log Temperature Drift Across a Full Duty Cycle

Why this matters

A single temperature reading is the weakest number in the trade. It tells you where something is, not where it is going, and on a fault that develops over hours the direction is the entire diagnosis. A logged curve turns "it ran hot" into "it rose 40 degrees F in the first hour, flattened, then started climbing again at hour four when the second stage came on," which names the moment something changed and usually names the component. It also settles arguments: a curve on a screen is evidence the customer can see, and it is the difference between recommending a repair and selling one.

Lead with the safety of the probe placement

Probes go on and near equipment that is running, hot, and often energized. Before you place anything:

• Route every lead away from moving parts, hot surfaces, and the path a panel door swings through. A probe wire pinched in a door or drawn into a fan is a failed test at best and a short at worst.
• Use leads and probes rated for the surface temperature and the circuit. A general-purpose probe pressed against a surface far above its rating gives you a wrong number and a melted lead.
• Do not open an energized enclosure to place a probe if a non-contact reading through an existing opening will do. Where contact placement inside a live enclosure is genuinely necessary, de-energize, verify dead with a known-good tester on a known-live source before and after, place the probe, and re-energize with the enclosure secured.
• Anchor probes mechanically, not with tape alone on a hot surface. Tape adhesive fails at temperature, and a probe that falls off at hour three produces a curve that looks exactly like a fault.

Step 1: Decide what the curve has to prove before you place anything

There are only a few questions a temperature log can answer, and picking one determines the whole setup.

• Is this component reaching a limit? One probe on that component, plus ambient.
• Is the system rejecting the heat it makes? Probes on inlet and outlet of the heat-rejecting path, plus ambient. The difference between them is the answer, not either one alone.
• Which of two suspects is heating first? One probe on each, matched sensor types, plus ambient.
• Did something change partway through the run? One probe on the most responsive point, at a fine sample interval, plus a run-state channel.

Placing four probes because you have four is how people end up with four curves and no conclusion.

Step 2: Always log ambient, and log it at the equipment

Every temperature you record is meaningless without a reference, because a rise of 40 degrees F in a 65 degree F room is a different machine than the same rise in a 95 degree F attic. Log ambient on its own channel, physically near the equipment but out of its discharge, its radiant field, and any sunlight.

The number you actually diagnose with is the delta - measured temperature minus ambient at the same timestamp. Delta is what stays comparable between a spring visit and an August one, and it is what you compare against your own notes from the last unit of the same type.

Step 3: Match the sensor type to the surface, and know what it is really telling you

• Surface contact probes read the surface, and they read it low if contact is poor. Clamp or magnet-mount them, and use thermal contact material on rough surfaces where the manufacturer's guidance allows it. A loose contact probe on a hot pipe can read tens of degrees below the true surface.
• Air probes in a duct or an enclosure read the air, which lags the metal around it. If you are chasing a component limit, air is the wrong channel.
• Non-contact infrared reads what the surface emits, and shiny or bare metal emits poorly, which is why an infrared reading on polished metal reads low, sometimes drastically. Read a dull, painted, or taped patch instead. See the sibling article on reading a thermal image for the full method.
• Clamp-on pipe sensors read close to fluid temperature on a small line and poorly on a large insulated one. Insulate over the sensor after placing it so ambient air does not pull the reading down.

Mixing sensor types across channels in one test is fine as long as you know which is which. What is not fine is comparing an air reading to a surface reading and calling the difference a delta.

Step 4: Set the sample interval to the fastest thing you care about

The rule of thumb: sample at least five times faster than the shortest event you want to see. If a protective device trips and resets in about 2 minutes, a 5 minute interval will step over the whole event and leave a curve that looks smooth and clean. If you only care about a multi-hour drift, 60 seconds is plenty and 10 seconds just fills memory.

For a duty-cycle log covering several hours, 30 to 60 second sampling is a good default. Drop to 5 to 10 seconds only when you are specifically hunting a fast transient, and expect a much shorter maximum run length when you do.

Set the logger clock against your phone before you start. Every channel and every note has to line up on one timeline, and a logger that is 40 minutes off turns a clean correlation into a mystery.

Step 5: Take a written baseline in the first 5 minutes

Before the curve goes anywhere, write down every temperature, the ambient, the time, and the state of the equipment. This costs two minutes and it is what makes the log readable a week later when you have run six other calls. It is also your protection against a channel that was reading garbage from the start.

Step 6: Read the shape, not just the peak

Once you have the data, the diagnosis is in the shape. Four shapes cover almost everything.

The third shape is the one that pays. It gives you a timestamp for a change of state, and a timestamp is something you can correlate against the run-state channel, the control's own behavior, and the customer's description.

A worked example, carried through

A cabinet-mounted assembly trips a thermal protective device most afternoons. All values below are illustrative.

Setup: one surface probe on the suspect component's housing, one air probe in the cabinet, one air probe outside the cabinet for ambient, and a current clamp for run state. Sampling at 45 seconds. Baseline written at 07:02: component surface 74 degrees F, cabinet air 73, ambient 72, unit off.

Hour 1: component surface delta above ambient rises to 52 degrees F, cabinet air delta to 9. That ratio is expected; the component is the heat source and the cabinet lags it.

Hours 2 to 4: component delta flattens at about 61 degrees F and holds within 2 degrees. Cabinet delta keeps climbing slowly, reaching 21 degrees F at hour 4. The component reached equilibrium relative to the cabinet, but the cabinet has not reached equilibrium relative to the room. This is the fast-mass, slow-mass split, and it means the component's absolute temperature is still rising even though its delta looks stable.

Hour 4:40: the component delta jumps from 61 to 74 degrees F over about 12 minutes and keeps going. The cabinet delta rises with it. The current trace shows total load dropping by roughly 15 percent at 4:38, two minutes before the temperature moved.

That two-minute lead is the diagnosis. A drop in current followed by a rise in temperature is a cooling component dropping out - something that was consuming power to move air or fluid stopped, so the load fell and the heat stopped leaving. If the temperature had risen first and the current had fallen after, the story would be the reverse: an overheating component reducing its own output or a protective device beginning to act.

Trip occurred at 5:05, so 27 minutes after the current drop. Verify the arithmetic against the shape: component delta went from 61 to 74 degrees F in 12 minutes, a rate of about 1.1 degrees F per minute, and it had roughly another 25 degrees F to travel to reach the limit, which at that rate is about 23 minutes. 12 plus 23 is 35 minutes from the current drop, against an observed 27. Close enough that the mechanism holds, and the small gap is explained by the rate accelerating as the cabinet heated too.

A tech reading only the peak would have recorded "component reached its limit, replace the component." The curve says the component was fine and the thing that cools it quit.

How to verify you got this right

Check three things before you trust a log. First, does the baseline match a hand reading you took with a separate instrument? A channel that is offset by 15 degrees F from the start will be offset all day. Second, does every channel show plausible movement, or is one flat? A flat channel is usually a fallen probe or a dead sensor, not a stable component. Third, does the run-state channel confirm the equipment ran continuously through the period you are interpreting? A gap in operation explains a dip in temperature that would otherwise look like a fault clearing itself.

The characteristic failure mode here is the probe that came loose at hour two. The curve shows a temperature that stops rising and slowly falls toward ambient, which reads exactly like the system suddenly started rejecting heat properly. If a curve heads toward ambient while the equipment is still running hard, suspect your probe before you suspect a miracle.

What changes the answer

• On outdoor equipment, ambient moves several tens of degrees across a day and the delta is the only usable number. Absolute temperatures on an outdoor log are close to meaningless.
• On cycling rather than continuous equipment, log run state and compute duty cycle percentage. Two units at the same peak temperature are in very different trouble if one ran 30 percent and the other 85 percent of the period.
• Where the process fluid itself carries a large thermal mass, the curve may not settle within a work day, so a same-day log will show a rise that never flattens and that is normal rather than a fault.
• If you cannot place a probe at the point that actually matters, say so in your notes rather than substituting a nearby point and treating it as equivalent. A reading from the wrong place, logged beautifully, is still the wrong reading.

References

• Manufacturer documentation on allowable temperature rise and sensor placement
• Trade-standard practice for temperature-rise testing and data logging
• OSHA guidance on test-equipment use around energized and moving equipment
• See related: Reading a Thermal Image: Generic Method; What Changes Inside a System During a Long Run Cycle; How to Set Up an Extended Run Test Without Camping On Site