The test’s value comes from the physics behind it. An internal fault produces heat and electrical stress that decompose the oil and the paper insulation into characteristic gases, and those gases dissolve in the oil, from which a sample can recover them. The composition of the gas mixture carries information about the temperature and the mechanism of the fault that no electrical measurement can provide, because the fault may be small enough not to affect the electrical quantities at all.

What DGA detects that other tests cannot

DGA detects active degradation. Ratio, resistance, insulation resistance, dielectric loss and frequency response are all condition measurements: they describe the state of the transformer at the moment of the test. A developing fault that has not yet changed the electrical characteristics leaves no trace in any of them.

The gases produced by that fault do leave a trace. A localised hot spot, a partial discharge site inside the insulation, a poor connection carrying current that heats it, or a core problem generating eddy currents all produce gas at temperatures determined by the mechanism, and the mixture that accumulates in the oil is a record of what has been happening since the last sample.

That record is also cumulative in a useful way. Because the oil circulates and mixes, gas produced at one location is distributed throughout the tank, so a sample drawn from a single point can detect a fault anywhere in the oil volume. The exception is a fault in a region where the oil does not circulate freely, which is one reason a gas result is read alongside the equipment’s design and its cooling arrangement.

Key gases and what each points to

Hydrogen is produced by partial discharge activity and by some low-temperature thermal processes. Its presence in isolation, at a stable level, is common in transformers and does not by itself indicate a fault. A rising hydrogen concentration with nothing else changing is a reason to look at the discharge measurement rather than to schedule an inspection.

Acetylene is produced by arcing at high temperature and is the gas most closely associated with a serious condition. Because it requires very high energy to form, its presence indicates that an arc has occurred, and a rising acetylene level indicates that arcing is continuing. It is the single result that most often justifies immediate action, and it does so regardless of the ratio code.

Ethylene forms at the higher end of the thermal range and indicates a hot spot. Methane forms at lower temperatures and is associated with general overheating. Ethane accompanies both. Carbon monoxide and carbon dioxide are produced by the decomposition of the cellulose insulation rather than the oil, so they indicate that the paper has been thermally stressed, which is a different concern from a hot spot in the oil.

Ratio methods: use and limitations

Ratio methods take the concentrations of the characteristic gases and compute ratios between them, then map the resulting values onto a code that corresponds to a fault type. Their purpose is to compress a multi-dimensional result into something a person can compare against a reference table without recomputing the underlying physics.

That compression is what limits them. Several different physical conditions can produce the same code, because the ratio boundaries are drawn at values chosen for practical utility rather than for physical uniqueness. The boundaries themselves are the second limitation: a sample whose ratio sits near a boundary can move from one code to another on the basis of measurement uncertainty that is small relative to the concentrations being compared.

The third limitation is that ratio methods were developed for oil-immiscible designs and for particular gas extraction methods. Where the oil type, the equipment design or the analytical method differs materially from the original basis, the code retains its usefulness as a summary and loses its usefulness as a diagnosis. The interpretation framework in IEC 60599 is written with those dependencies in mind, which is why it is presented as guidance rather than as a decision table.

Rate of change versus absolute values

The absolute concentration of a gas and its rate of change answer different questions, and the rate usually decides. A concentration that is above a typical reference value but has been flat for a decade describes a condition that has stabilised, which may be a slowly ageing insulation system or a benign generation mechanism inherent to the design.

A concentration within the typical band that has doubled over a six-month interval describes an active process. The mechanism may be small, but it is operating, and its trajectory is what determines whether the transformer needs attention now or at the next planned outage.

The practical consequence is that the sampling interval should be set by the rate rather than by the concentration. A unit whose values are flat does not need frequent sampling to be managed; a unit whose values are moving needs frequent sampling precisely because the data will change, and the change is the evidence.

Overdiagnosis: faults that are normal ageing

Several benign mechanisms produce gas in service. Normal thermal ageing of the paper produces carbon monoxide and carbon dioxide continuously, and the ratio between them changes with the ageing process. Oil oxidation produces its own products, some of which appear in the gas analysis. Certain materials inside the transformer, including some paints and surface treatments, generate small quantities of hydrogen under normal operating conditions.

A recent oil addition or an oil treatment changes the gas content of the oil, and the first sample after treatment can look unlike the samples that preceded it. A transformer that has been overloaded, or that has experienced a heavy through-fault, will show a thermal and possibly arcing signature that reflects the event rather than a persistent fault.

The defence against overdiagnosis is the unit’s own history. A transformer with ten years of stable values and a single elevated sample is a different situation from one with a steady upward trajectory. Where the history exists, the comparison is immediate. Where it does not, the first sample establishes a baseline, and the second sample is what gives the baseline meaning.

Correlating DGA with electrical tests

Gas analysis indicates that something is happening and, through the composition, what kind of something. It does not indicate where. Locating the mechanism requires the electrical tests, and the gas pattern is what tells the test team where to look.

A pattern dominated by thermal gases directs attention to connections, tap changer contacts and circulating currents, which are examined through resistance measurement, contact resistance and frequency response. A pattern dominated by hydrogen directs attention to discharge activity, which is examined through partial discharge measurement. A pattern that includes carbon oxides directs attention to the paper insulation, which is examined through furan analysis and dielectric response testing.

Reading the two together also resolves cases where one test is ambiguous. A marginal gas pattern with a clean frequency response and stable resistance suggests a benign mechanism. The same gas pattern with a shift in the frequency response band associated with winding geometry is a much stronger indication of a mechanical or thermal problem inside the winding.

When to increase sampling frequency

Sampling frequency should increase when a direction has been established but the interpretation is not yet conclusive. That is a defined condition rather than a matter of discomfort: a concentration that has moved beyond its previous range, a rate of change that exceeds what the history would predict, or a gas ratio that has moved across a reference boundary all count.

The increased frequency should come with a defined criterion for returning to normal. A rule that reverts to the standard interval once a concentration has been stable across a defined number of consecutive samples prevents an escalated programme from becoming permanent.

The interval is also worth shortening after an event, because the period immediately following a fault is when the information content of the analysis is highest. A sample taken shortly after an event and a second taken at a defined interval later will distinguish an event that has finished from one that is continuing, and that distinction determines whether the unit can be returned to service on its normal basis.

Reporting language that guides action

A gas report has two audiences: the engineer who has to decide what to do and the record that will be read years later. Both are served by stating the measured values, the previous values with their dates, the rate of change, the sampling conditions and the interpretation with its basis.

What should not appear is a conclusion that the data cannot support. A report that says a transformer has an arcing fault, on the basis of a ratio code, invites an outage that may not be justified. A report that says the acetylene concentration has risen from one defined value to another over a defined interval, that the rise is consistent with an arcing mechanism, and that the unit should be re-sampled at a stated interval and examined with a stated test, is both defensible and actionable.

The record should also state the analytical method and the laboratory, because a change in either introduces a difference that will otherwise be attributed to the transformer. The interpretation guidance is published as IEC 60599 and the supervision guidance for the oil is published as IEC 60422, with the general standards framework maintained by ASTM. Research and utility practice on gas interpretation is coordinated through CIGRE study committees and published in the transformer research programme of EPRI, with applied case material documented by Transformer Consultants. The laboratory and on-site instrument range is grouped on the oil testing hub.