How to Diagnose a Drifting Temperature Reading
A temperature point that slides a few degrees over days or weeks, with no matching process change, is one of the most common instrument complaints and one of the easiest to chase in the wrong order. The symptom is slow bias, not spikes; erratic jumping is a different fault family with different causes. This page orders the realistic causes of slow drift by likelihood and gives the test that confirms or clears each one.
Diagnose a Drifting Temperature Reading in one line: To diagnose a drifting temperature reading, first separate instrument drift from real process change by comparing against an independent measurement, then test causes in order: sensor element degradation, moisture lowering insulation resistance in the sheath or head, reference-junction or transmitter electronics drift, and installation effects such as lost tip contact or degraded insulation. The trend shape helps: element and moisture problems drift one way, while wiring-related error often tracks daily ambient swings.
First Checks
Before touching hardware, pull the trend and characterize the drift. A smooth one-directional ramp over weeks suggests element degradation or progressing moisture ingress; a bias that breathes with day-night cycles points at wiring, junction boxes, or CJC; a step change dates to an event worth finding in the maintenance log. General patterns of aging are covered in instrument drift over time, but the trend shape is your cheapest diagnostic.
Next, get one independent reading: a reference probe in an adjacent test well, a comparable redundant sensor, or even a surface measurement corrected for the installation. If the independent reading agrees with the suspect point, the process really changed and the instrument is innocent. This single check saves more wasted sensor swaps than any other step.
Test the Sensor Element First
The element is the most likely culprit. Platinum RTDs drift from contamination and mechanical strain, especially after high-temperature service or vibration; thermocouples decalibrate as the alloy composition changes near the hot junction. Pull the sensor and check it against the standard table at a known temperature - the ice-point procedure in bench-testing an RTD gives a definitive answer for a Pt100 in minutes on the bench.
The fix for a drifted element is replacement, not adjustment; trimming the transmitter to mask a drifting element bakes a moving error into the loop. For thermocouples, replace with the same type and junction configuration, and treat repeated fast decalibration as a signal that the sheath or the service temperature is wrong for the sensor.
Check for Moisture and Insulation Breakdown
Moisture in the sensor head or inside the sheath creates a shunt path that lowers the measured resistance of an RTD, so the reading drifts low and often gets worse in wet weather or after washdowns. On a thermocouple, moisture and degraded insulation can form a partial virtual junction that pulls the reading toward the temperature where the fault sits. Test insulation resistance between the leads and the sheath with the element disconnected; a healthy sensor reads very high, and a value that has sagged into the low megohm range explains a drifting point.
The fix is drying and sealing if the moisture is in the head - replace the gasket, reseat the cable gland, face the conduit drain properly - or sensor replacement if the ingress is inside the sheath, which does not recover.
Isolate the Transmitter and the Cold Junction
If the element and insulation check clean, substitute a simulator for the sensor and watch the loop for a day if you can: a decade box for an RTD or a millivolt source for a thermocouple. If the loop drifts with a rock-solid input, the transmitter electronics or the reference junction are drifting, and the split between input and output error is found with the procedure in calibrating a temperature transmitter. On thermocouple loops, a failing terminal-temperature sensor shifts every reading equally, the signature described in cold-junction compensation.
Finally, consider the installation: a spring-loaded element that has lost tip contact in its well, insulation stripped from the pipe near the well, or a changed flow regime all shift the measured temperature away from the true process value without any component being broken.
When to Escalate
If the point feeds a safety function, a custody calculation, or a product-quality decision, stop diagnosing informally once you confirm genuine drift and route the loop through the site's formal calibration program, with as-found values recorded before anything is adjusted. Chronic drift across many points in one area is worth escalating as a pattern - shared heat, shared moisture, or a shared marshalling cabinet - rather than as a series of individual sensor swaps.
Frequently Asked Questions
How do I tell instrument drift from a real process change?
Get one independent measurement of the same process point: a reference probe in a test well, a redundant sensor, or a corrected surface reading. If it agrees with the suspect instrument, the process changed. If it disagrees, characterize the drift shape on the trend: one-directional ramps implicate the element or moisture, daily-cycling bias implicates wiring and reference-junction effects.
Why does an RTD drift low when moisture gets in?
Water creates a parallel conductive path between the leads or from a lead to the sheath, and a parallel resistance always lowers the total the input measures. Since lower resistance means lower indicated temperature on an RTD, moisture classically shows up as a point that slowly reads colder, often worsening after rain or washdown, and an insulation-resistance test with the sensor disconnected confirms it.
Automation services
Need help turning this into a working system?
Merobix integrates SCADA, programs Allen-Bradley and Siemens PLCs, and designs and fabricates industrial control panels.
Meeting requests are reviewed before confirmation.