How to Interpret DGA Results: The Duval Triangle, Gas Ratios, and the Key-Gas Method

Dissolved gas analysis tells you that a fault is developing. Interpretation tells you what kind of fault, and how urgent it is. Fault gases dissolve into transformer oil in patterns that reflect the physics that produced them — a hot connection, a partial discharge, an arc — and decades of engineering practice have distilled those patterns into a handful of interpretation methods. This guide explains the three that matter most, and, just as importantly, what they can and cannot tell you on their own.

A note before the methods: interpretation is the second question. The first is whether the gassing is abnormal at all, which is a trend question, not a single-sample one — see DGA trend analysis. With that established, interpretation classifies the fault.

1. The key-gas method

The key-gas method is the most intuitive: the gas that dominates points to the fault type. Hydrogen with methane suggests low-energy electrical activity; ethylene indicates higher-temperature thermal faults; acetylene is the signature of arcing and high-energy discharge; carbon monoxide and carbon dioxide point toward cellulose (paper) involvement. IEEE C57.104 frames how these gases are generated and screened. The key-gas view is excellent for building intuition, but it is coarse — real faults often generate a mix of gases, which is where ratios and the triangle earn their place.

2. Gas ratios (IEC 60599 and Rogers)

Ratio methods compare pairs of gases rather than absolute amounts, which makes them more robust to how much oil is in the tank or how long gases have accumulated. IEC 60599 defines the basic ratios — C2H2/C2H4, CH4/H2, and C2H4/C2H6 — and the closely related Rogers ratio set uses the same pairs to place a fault into categories such as thermal faults of different severity, partial discharge, and arcing. Ratios are powerful precisely because they normalize away the “how much” and focus on the “what kind.” Their limitation is that some gas combinations fall outside the defined ranges, returning no diagnosis — a known gap that the graphical methods handle more gracefully.

3. The Duval triangle

The Duval triangle is the method most engineers reach for, because it always returns a zone. Duval Triangle 1 plots the relative percentages of three gases — methane (CH4), ethylene (C2H4), and acetylene (C2H2) — as a single point in a ternary diagram, and the point’s location classifies the fault into one of seven zones: partial discharge (PD); thermal faults of rising temperature (T1, T2, T3); low- and high-energy discharge (D1, D2); and a mixed thermal-and-electrical region (DT). Because every point lands somewhere, the triangle avoids the “no result” gap of the ratio methods. The Duval triangle method is grounded in the same IEC 60599 fault physics as the ratios.

One honesty check carries across all three methods, and it is worth stating plainly: the Duval triangle classifies a fault, it does not detect one. You first determine that the gassing is abnormal — a rising trend, an elevated generation rate — and only then does the zone tell you what type of fault is most consistent with the gas mix.

Try it on your own numbers

GridAPM’s interactive DGA Analyzer runs all three at once — the Duval Triangle 1 classification, the basic IEC 60599 gas ratios, and IEEE C57.104-2019 screening — entirely client-side, so your gas values never leave your browser. It is a reference aid for engineers, not a diagnosis: it shows where a sample sits and what the standard methods say, and leaves the judgement where it belongs.

Why no single method is the answer

Modern interpretation guidance, including CIGRE TB 771, treats these methods as complementary, cross-checking one against another rather than trusting any one in isolation. And even a confident zone is not yet a decision. The action depends on:

• Trend velocity and generation rate — is the fault active and accelerating, or a stable historical artifact?
• Measurement quality — laboratory versus online monitor, calibration, and sampling integrity.
• Operating context — recent loading, cooling, oil processing, or maintenance that could explain the gas.
• Consequence — the asset’s criticality, spare availability, and outage constraints.

This is why interpretation sits inside a broader condition-assessment practice (CIGRE TB 761) and connects to the full picture of why transformers fail and the diagnostics that catch them. Where the asset warrants it, continuous online DGA monitoring shortens the time from a gas change to a reviewed decision.

How GridAPM uses interpretation

GridAPM does not replace the interpretation methods or the engineer who applies them. It assembles the DGA history, computes the standard classifications, flags where the trend and the zone agree or disagree, surfaces missing context and measurement-quality caveats, and drafts a source-linked case — so a qualified reviewer can decide whether to monitor, retest, inspect, plan an outage, or intervene, with every conclusion traceable to the gas data behind it.

Interpretation turns gases into a hypothesis. Trend, context, and review turn a hypothesis into a defensible decision. Try the DGA Analyzer to see the three methods on a sample, or request a GridAPM pilot to put interpretation to work inside a human-reviewed workflow on your own fleet.