Transformer health indices are used in fleet management to rank assets by condition and prioritize inspection, testing, and replacement planning. A health index reduces the multidimensional condition of a transformer to a single number, which makes it possible to sort a fleet of dozens or hundreds of assets and direct limited maintenance resources toward the highest-priority units.
The appeal is obvious. The risk is that a single-number index compresses a lot of information and can be misleading if the compression is done poorly or if the inputs are incomplete. This article explains what transformer health indices measure, how the major input categories contribute, where each input's limitations matter, and how continuous EM signature monitoring fits as an additional input layer that addresses a specific gap in conventional index construction.
What a Health Index Is Actually Expressing
A transformer health index is typically expressed as either a percentage score (100 = new/pristine condition, lower = more degraded) or as a probability estimate (0 to 1 representing failure probability over a defined time horizon). These are mathematically equivalent representations of the same underlying concept, but they communicate differently to different audiences.
The probability framing is more precise about what the index represents and what it does not. A health score of 65 on a 0-100 scale does not mean the transformer is 65 percent of the way through its operational life. It means that, given the current input data, the transformer's estimated failure probability over the next 30 or 90 days is X, where X is calibrated against the failure rates observed in the reference population used to build the index model.
This calibration against a reference population is where the first significant limitation appears. The reference population for any commercially deployed health index was assembled from some set of transformers, with some distribution of ages, designs, loading histories, and operating environments. If your fleet differs substantially from that reference population, the absolute probability estimates may be biased. A 35-year-old transformer in a coastal environment with high humidity and salt exposure that is scored against a reference population of inland industrial transformers may have its failure probability systematically underestimated.
DGA as a Health Index Input
Dissolved gas analysis is the most information-dense single input in a transformer health index. The key gas method and Duval triangle interpretation provide fault type classification (thermal, electrical, arcing), and the gas concentration levels and rate-of-change can be mapped to severity ratings that feed directly into a composite score.
Within a composite health index, DGA inputs typically carry high weight because they are directly diagnostic of internal condition. The principal dissolved gases, hydrogen, methane, ethane, ethylene, acetylene, carbon monoxide, and carbon dioxide, each have specific physical interpretations. Acetylene, for example, is specifically produced by high-energy arc discharge, and its presence above trace levels is a direct indicator of a serious fault. Carbon monoxide and dioxide indicate cellulosic insulation degradation. The mapping from gas concentrations to a health score is not arbitrary; it reflects the established physical interpretation of what each gas indicates.
The inputs to DGA-based health scoring have two limitations that are worth naming. First, sampling frequency: a quarterly sample provides a snapshot, not a trend. The health score component derived from DGA is as current as the most recent sample. Between samples, the DGA-derived health component is static regardless of what is happening inside the transformer. Second, background variation: some transformers have elevated baseline hydrogen from normal hot-spot activity or from oil chemistry that produces hydrogen through non-fault mechanisms. Interpreting hydrogen elevation against a flat absolute threshold rather than against that unit's own historical baseline will produce either false positives (for transformers with naturally high background) or reduced sensitivity (if the threshold is raised to accommodate high-background units).
Power Factor and Bushing Condition
Power factor testing measures the dielectric loss in the transformer's insulation system, including the main tank insulation, bushings, and oil. Bushing power factor is particularly important because bushing failures are a significant contributor to in-service failure statistics, and bushing condition can deteriorate faster than main insulation condition under some aging and contamination scenarios.
In a composite health index, bushing power factor is typically weighted to reflect that a bushing in poor dielectric condition represents an elevated failure risk independent of the transformer's overall insulation condition. The IEEE C57.152 standard for field testing of liquid-immersed power transformers provides reference ranges for bushing power factor that inform the mapping from measured values to health score contribution.
The limitation here is testing frequency. Power factor testing is typically performed during planned outages, which may occur every 5 to 8 years for a well-maintained fleet asset. Between tests, the power factor component of the health index is stale. A bushing that develops moisture ingress or contamination between testing cycles does not produce any signal in the health index until the next test. This creates the same between-test coverage gap as periodic DGA sampling, but potentially over a longer interval.
Load and Thermal History
The IEEE C57.91 thermal aging model, built around the Montsinger relationship between temperature and insulation life consumption, provides a principled framework for translating load and temperature history into a cumulative thermal aging factor. This aging factor expresses how much of the transformer's insulation life has been consumed relative to rated life at design operating conditions.
In a composite health index, the thermal aging component reflects the cumulative impact of the transformer's operating history: how many hours at what fraction of rated load, how many times the transformer was subjected to emergency overload conditions, what the ambient temperature profile looked like. A transformer that has operated at modest load in a mild climate will have a lower thermal aging contribution to its health index than a same-age unit that has repeatedly been run at 120 percent load during peak demand periods.
The thermal aging component is relatively data-rich compared to the periodic-sample inputs, because SCADA systems typically record load and temperature at 15-minute or hourly intervals over the full operating history. The main uncertainty is in the hot-spot temperature calculation, which requires the thermal model parameters for the specific transformer design, and these parameters vary across designs in ways that are not always well-documented for older equipment.
Age and Inspection Records
Age and physical inspection findings typically contribute to the health index as modifying factors on the other inputs. A transformer that is 35 years old, at the end of its expected design life range, has its scores from DGA and power factor interpreted differently than the same results from a 15-year-old unit. The age factor reflects that residual insulation life is lower for an older unit even if its test results are currently within normal limits.
Physical inspection findings, including oil leaks, corrosion, mechanical damage to the tank and bushings, and visible evidence of previous overheating or arcing, are typically incorporated as binary or categorical modifiers. A transformer with a visually intact external condition scores differently from one with noted oil weeping from the radiator seals or discoloration of the bushing porcelain.
Where Conventional Indices Have a Structural Gap
The four input categories above, DGA, power factor, load and thermal history, and age plus inspection, give a composite picture of transformer condition that is substantially better than any single indicator alone. But there is a structural gap in how these inputs update.
DGA updates on a schedule. Power factor updates on a longer schedule. Load history is continuous. Age and inspection are updated episodically. The result is that a health index based only on these inputs is partially continuous and partially stale at any given moment. For a transformer where the most recent DGA was taken 3 months ago and the last power factor test was 4 years ago, the health index is reflecting conditions as they were at those measurement times, not as they are today.
A developing winding fault that began after the last DGA sample will not appear in the health index until either a DGA sample is taken that shows elevated gases, or the fault progresses far enough to produce a measurable thermal or electrical effect visible in other signals. The interval between onset and detection in the conventional health index can easily be the full duration of the DGA sampling cycle, during which the true condition is worse than the index reflects.
EM Signature as a Continuously Updating Health Index Component
The EM signature deviation signal that Magnefy tracks provides a continuously updating component that specifically addresses the winding fault detection gap. Unlike DGA, which updates with each sample, or power factor, which updates with each test, the EM signature component updates on the same cadence as the measurement, which is continuous or near-continuous in a deployed sensor configuration.
The EM-derived health component reflects the current state of the winding current distribution relative to the transformer's own established baseline. When the winding current distribution is consistent with baseline, the EM component contributes a healthy baseline score. When a deviation appears that exceeds the statistical threshold for the baseline variability of that specific transformer, the EM component scores lower, and the composite health index drops to reflect the current anomaly.
In a composite index that incorporates EM signature alongside DGA, power factor, and load history, the combined index has higher temporal resolution than one based on periodic samples alone. For the winding fault family, the EM component provides advance signal before the DGA signal rises to detectable levels, extending the effective detection window for that fault family.
We are not saying that EM monitoring makes DGA or power factor testing obsolete for health index construction. DGA provides fault type classification that EM signature alone cannot match for the thermal and high-energy fault families. Power factor testing provides bushing condition assessment that EM signature does not address. The composite index incorporating all available inputs is more informative than any single source, and the EM component adds value specifically in the winding fault coverage and temporal resolution dimensions that conventional inputs leave partially covered.
The practical question for an engineer building or evaluating a transformer health index program is: what fault families and what detection timelines are covered by the current input set, and where are the gaps? For most conventional health indices, the winding fault family with late-DGA-detectable onset is the most significant coverage gap in terms of failure consequences and detection window. Continuous EM signature monitoring addresses that gap directly and is the cleanest addition to a conventional index program for operators who want to close it.