Temperature correction for insulation resistance measurement — why 500 MΩ at 30°C means something different than at 20°C

The guide on insulation resistance measurement covers the basic measurement itself. This article covers a factor that is at least as important as the measurement itself when comparing insulation resistance measurements over time — in trend analysis of a motor or transformer across multiple years: the temperature of the measured winding at the moment of measurement.

Why insulation resistance is temperature-dependent

The insulation resistance of a winding is determined not only by the physical condition of the insulation (contamination, moisture, ageing), but also strongly by the temperature of that insulation at the moment of measurement. Insulating material behaves electrically somewhat like a semiconductor: as temperature rises, the conductivity of the material increases (and hence insulation resistance decreases), largely independent of the actual condition of the insulation.

The rule of thumb: halving per 10°C

A commonly used rule of thumb is that the insulation resistance of an average winding roughly halves for every 10°C rise in temperature, and conversely roughly doubles for every 10°C drop. Two measurements on the same winding — for example 500 MΩ at 20°C on one day, and 250 MΩ at 30°C on another day — can therefore be identical in terms of the actual insulation condition, despite an apparent halving of the measured value.

Note: this factor-of-2-per-10°C is a practical rule of thumb for general purposes, not an exact physical constant. IEEE 43 (recommended practice for testing insulation resistance of rotating AC machinery) publishes more detailed temperature correction factors per insulation system, which are preferable to the simple rule of thumb for an accurate trend analysis.

The correction: reducing to a reference temperature

To meaningfully compare two measurements taken at different winding temperatures, each measured value is reduced to a fixed reference temperature — in practice usually 20°C — using a temperature correction factor that corresponds to the temperature difference between the measured temperature and the reference temperature. Only after this correction are the two corrected values mutually comparable and usable for a trend analysis across multiple inspections.

Why this is more than a theoretical subtlety

Without temperature correction, a trend analysis can lead to two kinds of erroneous conclusions:

An apparent decline in the insulation is flagged, while the winding on the most recent measurement day was simply warmer than at the previous measurement (for example measured shortly after operation, instead of in a cold state).
An actual decline in the insulation is masked, because the most recent measurement happened to be taken at a lower winding temperature than the previous one, causing the uncorrected value to incorrectly appear equal or higher.

Both situations undermine the purpose of a periodic insulation resistance measurement as a trending tool for predictive maintenance.

Practical relevance

When setting up or assessing a trending insulation resistance measurement of a motor, generator or transformer, it is important to record not only the insulation resistance at each measurement, but also the winding temperature at that moment (preferably via a built-in temperature sensor or, in its absence, a representative estimate), and to correct each measured value to the same reference temperature before comparison. Without this step, a trend graph of uncorrected values across multiple years says little.

Common mistakes

Comparing insulation resistance values from different measurements without temperature correction in a trend analysis, while the winding temperature may differ significantly between measurements.
Not recording the winding temperature at the moment of measurement, making a later correction impossible.
Applying the simple "halving per 10°C" rule of thumb to an insulation system for which more detailed IEEE 43 factors are available, resulting in a less accurate correction.
Assessing a single, uncorrected measurement against an absolute threshold without considering the temperature at which that threshold was defined — even an absolute rejection limit belongs to a reference temperature.