Measurement, Decision, Action: Which One Is Wrong

Why this matters

A control complaint arrives as a description of an outcome: it runs too hot, it never shuts off, it will not hold. That description does not tell you which third of the loop failed, and the three thirds cost wildly different amounts to chase. Two comparisons, both of which you can make in the first twenty minutes, split every one of those complaints into measurement, action, or decision. Techs who skip them end up in the controller's parameter list, which is the one place the answer almost never is.

Before you compare anything

The comparisons below need the equipment running and an instrument at a physical point on the system. That combination is where people get hurt on this kind of call.

• Reading a controller display and an actuator's travel indicator requires opening nothing. Do that first and record it.
• Where a measurement must be taken inside an energized control enclosure, 29 CFR 1910.333(a)(1) permits energized work only where de-energizing introduces additional or increased hazards or is infeasible due to equipment design or operational limitations, and troubleshooting a live control circuit is a recognized case; use a meter and leads rated CAT III at or above the circuit voltage and work within the boundaries and protective equipment NFPA 70E-2021 assigns to the task.
• Where the check can be made dead, open the disconnecting means, apply your own lock and tag under 29 CFR 1910.333(b)(2) in general industry or 29 CFR 1926.417 in construction, and prove dead per NFPA 70E-2021, 120.5 - meter on a known live source, then the conductors, then the meter again.
• A sensor in a well on a hot or pressurized line comes out of the well; the well does not come out of the line. If the well itself must come out, isolate, drain to a safe point, and confirm zero on a gauge before a thread moves, and address the scald route directly - let the line cool or use insulated gloves and eye protection, because a hot pressurized line flashes when it opens.
• Do not disturb pipe or vessel insulation to reach a sensor on older equipment. Thermal system insulation of unknown vintage is an inhalation hazard, treat it as asbestos-containing until it has been sampled, and if it must be disturbed that is respiratory protection selected under a program meeting 29 CFR 1910.134, not a dust mask.
• A spring-return actuator stores energy. Never put a finger, a screwdriver or a linkage-freeing pry into a live actuator's travel path. To work on the linkage, isolate the actuator, release or restrain the spring, and lock or tag under 29 CFR 1910.147.

The two comparisons

Comparison one, on the measurement stage: what the controller says it is reading, against an independent instrument at the same physical point. Same point matters more than same accuracy. A reference reading taken six feet away in different air, different water, or a different part of the vessel is not a comparison, it is a second unrelated number.

Comparison two, on the action stage: what the controller says it is commanding, against what the final device is actually doing. Read the commanded percentage or state off the controller, then read the actuator's own position indicator, valve stem travel, damper blade angle, or contactor state. Do not accept a feedback signal as the answer to this one; a position feedback signal is part of the actuator's own story and can be wrong in the same direction as the fault.

Between them, these two partition the loop. If the measurement agrees and the action matches the command, and the outcome is still wrong, only then are you looking at the decision stage - and that is the rarest of the three by a wide margin.

Why decision comes last

The decision stage is the only one of the three you cannot observe directly. You infer it from the other two agreeing. That ordering is not politeness toward the controller, it is a consequence of what is measurable: a sensor and an actuator both have physical states you can put an instrument on, while the controller's internal comparison is a computation you can only test by watching what it does with inputs you trust.

There is a second reason. A wrong decision is usually a wrong setting, not broken logic - a setpoint someone changed, a schedule fired in the wrong time zone after a power event, an authority or reset schedule that shifts the target on outdoor condition and is doing exactly that while everyone thinks the target is fixed. Those are configuration findings, and you find them by reading the configuration, not by testing. Genuinely broken control logic is real but uncommon, and everything that looks like it in the first hour is normally one of the other two stages.

Case one: the comparisons resolve to action

A hot water supply loop, setpoint 180 F, complaint is that it overshoots at night and the customer hears the relief lifting.

The controller displays 196 F. An independent instrument in the same well reads 195 F. That is 1 F apart, inside the combined tolerance of two field instruments, so the measurement stage is telling the truth: it really is 196 F.

The controller reports its output at 0%. The valve's travel indicator sits at about 15% open. The command and the action disagree, so the fault is in the action stage.

Fifteen percent of stroke with no call is enough to overrun a light night load, and it explains why the complaint is a night complaint: during the day, load is large enough that a small leak-through never accumulates. On this valve the cause turned out to be a set screw that had backed off on the linkage, so the actuator reached its own zero while the stem still sat short of the seat. A worn seat or a stroke limit adjusted at commissioning and never rechecked produce the identical symptom, and all three are found by the same comparison.

What would flip this. If the stem had read a true zero and the temperature still climbed with the loop satisfied, the leak is somewhere the loop does not command at all - a bypass valve left cracked open from a previous service call, a check valve holding backward, a second heat source on the same vessel. That is a different finding, and it is still not a controller fault.

The failure mode. Reading the actuator's position feedback signal off the controller instead of the stem. Feedback signal and command agreed perfectly on this job. The linkage had slipped downstream of where the feedback was taken, so the controller and the actuator both believed the valve was closed, and both were wrong.

Case two: the same complaint resolves to measurement

Same equipment type, same setpoint of 180 F, same complaint of running hot.

The controller displays 168 F, so it is calling for heat, output at 45%. The independent instrument in the same well reads 182 F. Fourteen degrees apart is far outside anything two field instruments explain between them, so the measurement stage is lying and everything downstream of it is behaving correctly on bad information.

Predict before you keep testing. If the controller under-reports by 14 F, it will hold the real temperature at about 180 plus 14, or 194 F. Twenty minutes later the reference read 193 F while the controller read 179 F, a 14 F spread again and within a degree of the prediction. That confirms an offset that is roughly constant across this narrow span rather than a slope error.

Be careful how far you carry that. A constant error in resistance at a corroded splice or a poor termination maps to a roughly constant error in temperature only over a narrow span, because the element's sensitivity changes with temperature. Do not extrapolate this 14 F down to ambient and conclude the sensor reads 14 F high in a cold building.

The direction of the error is informative only once you know the element. On a negative-coefficient thermistor, added series resistance reads as a colder temperature, which fits what this controller showed. On a platinum resistance element, whose resistance rises with temperature, the same added resistance reads hotter. Confirm the input type from the controller's configuration before reasoning backward from direction, or you will chase the wrong end of the circuit.

The other common producers of this exact spread, all cheaper than a sensor: a probe not bottomed in the well, a well with no heat transfer compound and a large air gap, a sensor mounted where it reads a mixture rather than the stream it is meant to control. Check the mechanical fit before you condemn the element, and check the terminations before you check the fit, because a termination is the fastest of the three to prove.

When you cannot make one of the comparisons

Some loops do not give you a place to stand. A humidity, flow or pressure measurement may have no second instrument available at the same point. A valve may be buried with no travel indicator.

Substitute a response test, not a guess. Command a large change from hand mode and confirm the measured value moves in the right direction, by a plausible amount, in a plausible time. That does not prove accuracy, and you must say so on the ticket, but it does prove the circle is intact, which separates a stuck action from an offset measurement. Leave every limit and interlock in service while you do it, and stop the test at the first protective device that begins to approach its trip point rather than pushing through.

Verifying the call before you order a part

Three checks that are not restatements of the comparisons, and each has caught a wrong call in the field.

Move the disagreement. If you blamed the sensor, put your reference in a second location that the sensor should also track and confirm the spread follows the sensor rather than the location. A spread that changes when you move the reference was a placement problem all along.

Command through the full range, not just at the current point. An actuator that matches its command at 45% and diverges above 80% has a stroke or an air-supply limit, which is a different repair than a slipped linkage and gets missed entirely by a single-point check.

Re-read the configuration after the repair, before you leave. A previous tech chasing this same complaint may have shifted a setpoint, an offset trim, or a reset schedule to compensate for the fault you just removed. Fix the fault and leave the compensation in place and the customer calls back with the opposite complaint inside a week.

References

• 29 CFR 1910.333(a)(1) and (b)(2), OSHA general industry work practices for energized and de-energized electrical work
• 29 CFR 1926.417, OSHA construction lockout and tagging of circuits
• 29 CFR 1910.147, OSHA control of hazardous energy, for actuator spring tension and mechanical isolation
• 29 CFR 1910.134, OSHA respiratory protection program requirements
• See related: How Sensors Fail; What a Control Loop Actually Is