1. Introduction: Why the Duval Triangle Matters
Every power transformer contains thousands of liters of insulating oil. When this oil is subjected to thermal or electrical stress, it decomposes — releasing a characteristic cocktail of dissolved gases. Dissolved Gas Analysis (DGA) is the single most powerful diagnostic tool we have for assessing transformer health, and the Duval Triangle is arguably its most elegant interpretation method.
Developed by Michel Duval at Hydro-Québec’s research institute (IREQ) in the 1970s and refined continuously over five decades, the Duval Triangle transforms raw gas concentration data — numbers that are meaningless to the untrained eye — into a precise fault classification. A single point plotted inside a triangle tells you whether your transformer is experiencing partial discharge, arcing, or thermal runaway, and at what temperature.
This guide takes you on a complete journey — from understanding why gases form, through plotting your first triangle, to mastering the full diagnostic workflow with all seven Duval tools and the latest AI-assisted interpretation techniques.
2. Part 1 — Foundations: The Gas Physics of Transformer Faults
2.1 The Seven Key Fault Gases
Transformer mineral oil is a complex mixture of hydrocarbon molecules. When these molecules break down under thermal or electrical stress, they produce seven diagnostically significant gases:
| Gas | Formula | Primary Fault Association | Formation Temperature |
|---|---|---|---|
| Hydrogen | H₂ | Partial discharge, corona — the universal early warning gas | > 100 °C |
| Methane | CH₄ | Low-temperature thermal faults, stray gassing | > 150 °C |
| Ethane | C₂H₆ | Low-to-medium temperature thermal faults | > 200 °C |
| Ethylene | C₂H₄ | High-temperature oil overheating (> 500 °C) | > 300 °C |
| Acetylene | C₂H₂ | High-energy arcing & discharges — the arc signature gas | > 700 °C |
| Carbon Monoxide | CO | Cellulose (paper) thermal degradation | > 105 °C |
| Carbon Dioxide | CO₂ | Cellulose aging & general oxidation | Broad range |
Figure 1: Gas formation temperature spectrum. The key insight: C₂H₂ (acetylene) forms only above 700 °C — making it the definitive marker for arcing faults.
2.2 Why Gas Concentrations Alone Are Not Enough
Absolute gas concentrations (measured in ppm or µL/L) are useful for trend monitoring, but they cannot directly identify the type of fault. A transformer with 50 ppm of acetylene might be experiencing a dangerous high-energy arc, or it might simply be a large transformer with a large oil volume where the concentration is diluted. This is where gas ratios become essential.
The fundamental insight behind all ratio-based DGA methods is: the relative proportions of gases — not their absolute values — reveal the physics of the fault. Different fault types produce different gas “fingerprints”:
- Partial Discharge (PD): Almost pure H₂, with traces of CH₄. Very little else.
- Thermal faults (T1/T2/T3): Dominated by CH₄ and C₂H₄, with C₂H₄ proportion increasing with temperature. Little or no C₂H₂.
- Electrical arcing (D1/D2): High C₂H₂ (the arc “smoking gun”), with moderate CH₄ and C₂H₄.
3. Part 2 — Duval Triangle 1: The Core Diagnostic Method
3.1 The Three Diagnostic Gases
Duval Triangle 1 uses only three gases — deliberately excluding H₂:
| Gas | Role in Diagnosis |
|---|---|
| Methane (CH₄) | Indicates low-temperature thermal decomposition of oil. Dominant in T1 faults. |
| Ethylene (C₂H₄) | Indicates high-temperature oil overheating. Proportion increases with fault temperature. |
| Acetylene (C₂H₂) | The arc indicator. Any significant presence (> few ppm) is a red flag for electrical discharge. |
3.2 The Seven Fault Zones
After normalizing the three gas concentrations to percentages that sum to 100%, the resulting coordinate falls into one of seven zones:
| Zone | Fault Type | Typical %CH₄ | Typical %C₂H₄ | Typical %C₂H₂ | Physical Meaning |
|---|---|---|---|---|---|
| PD | Partial Discharge (corona) | ≥ 98% | — | — | Gas bubbles in voids. Low energy, but degrades insulation over time. |
| T1 | Thermal fault < 300 °C | Up to 98% | ≤ 20% | ≤ 4% | Mild overheating — overloaded windings, blocked oil ducts. |
| T2 | Thermal fault 300–700 °C | — | 20–50% | ≤ 4% | Moderate hot spot — loose connections, circulating currents. |
| T3 | Thermal fault > 700 °C | — | ≥ 50% | ≤ 15% | Severe overheating — carbonized paper, metal hot spot, tank wall heating. |
| D1 | Low-energy discharge | — | ≤ 23% | ≤ 13% | Spark discharges, tracking in paper, partial discharge in oil wedges. |
| D2 | High-energy discharge (arcing) | — | 23–40% | 13–29% | High-current arcing — winding flashover, tap changer failure, lead break. |
| DT | Mixed thermal + electrical | — | 40–50% | 4–29% | Combination fault — or a fault in transition between categories. |
Figure 2: Duval Triangle 1 — schematic representation of the seven diagnostic zones. Each vertex represents 100% of one gas; the opposite side represents 0% of that gas. A point’s position is determined by the normalized percentages of CH₄, C₂H₄, and C₂H₂. Zone boundaries per IEC 60599:2022 Annex B. (Schematic illustration — exact boundaries are defined by the standard’s coordinate table.)
3.3 Step-by-Step: How to Use Duval Triangle 1
📝 Worked Example: Real Transformer Data
Scenario: A 120 MVA, 220 kV power transformer has been in service for 8 years. Routine DGA sampling returns the following values from the laboratory:
| Gas | H₂ | CH₄ | C₂H₆ | C₂H₄ | C₂H₂ | CO | CO₂ |
|---|---|---|---|---|---|---|---|
| ppm | 220 | 80 | 40 | 90 | 25 | 350 | 4200 |
Step 1: Extract the three Triangle 1 gases
CH₄ = 80 ppm C₂H₄ = 90 ppm C₂H₂ = 25 ppm
Note: H₂, C₂H₆, CO, and CO₂ are not used in Triangle 1.
Step 2: Calculate the sum
Sum = 80 + 90 + 25 = 195 ppm
Step 3: Normalize to percentages
%CH₄ = 100 × 80 ÷ 195 = 41.0%
%C₂H₄ = 100 × 90 ÷ 195 = 46.2%
%C₂H₂ = 100 × 25 ÷ 195 = 12.8%
Quick check: 41.0 + 46.2 + 12.8 = 100.0 ✓
Step 4: Plot and classify
| Parameter | Value | Which Zone? |
|---|---|---|
| %C₂H₂ | 12.8% | < 13% → below D1 threshold; but > 4% → above thermal thresholds |
| %C₂H₄ | 46.2% | Between 40% and 50% → DT zone boundary |
| %CH₄ | 41.0% | Not at PD level (< 98%) |
Diagnosis: Boundary between D2 (high-energy arcing) and DT (mixed thermal-electrical).
4. Part 2 — Beyond Triangle 1: The Complete Duval Family
4.1 The Diagnostic Hierarchy
Duval Triangle 1 is your starting point — but it is only the first tool in a seven-tool diagnostic suite. The complete family resolves fault sub-types that Triangle 1 can only suggest:
Figure 3: Complete Duval diagnostic workflow — from DGA sample to final diagnosis. Triangle 1 is the primary classifier; secondary tools refine the diagnosis.
4.2 Duval Triangle 4 — Resolving Low-Energy Faults
When to use: Only when Triangle 1 returns PD, T1, or T2. Do NOT use for D1/D2/DT results.
Purpose: Distinguishes genuine partial discharge from stray gassing of oil, and low-temperature overheating from paper carbonization.
Gases: H₂, CH₄, C₂H₄, C₂H₆ (note: different gas set from Triangle 1)
| Zone | Interpretation |
|---|---|
| PD | Genuine partial discharge — investigate insulation condition |
| S | Stray gassing of oil — common in new transformers; typically benign but requires monitoring |
| O | Overheating of oil — elevated temperature without paper involvement |
| C | Carbonization of paper — serious; indicates cellulose degradation at hot spot |
4.3 Duval Triangle 5 — Pinpointing High-Temperature Faults
When to use: Only when Triangle 1 returns T2 or T3.
Purpose: Answers the critical question: “Is the hot spot in the oil only, or is paper being carbonized?”
Gases: H₂, CH₄, C₂H₄, C₂H₆.
| Zone | Interpretation | Urgency |
|---|---|---|
| T3-H | Hot spot in oil only — no paper involvement | Medium — plan outage |
| C | Paper carbonization — cellulose is being destroyed | High — immediate action; paper is irreplaceable |
4.4 Duval Pentagon 1 — The Five-Gas Cross-Check
Triangle 1’s main limitation is that it only uses three gases and ignores H₂. Pentagon 1 addresses this by incorporating all five hydrocarbon gases (H₂, C₂H₆, CH₄, C₂H₄, C₂H₂) into a weighted centroid calculation.
When to use: When Triangle 1 result is ambiguous (near boundaries), when H₂ is disproportionately high, or as a routine cross-verification.
The calculation is more complex than Triangle 1 — each gas contributes a vector in 2D space based on its percentage, and the Bourke area-weighted centroid determines the fault zone. This centroid approach makes Pentagon 1 more robust against single-gas measurement errors.
4.5 The LEDT — Catching Faults Before They Become Critical
The Low Energy Degradation Triangle (LEDT), introduced in 2022, is the newest addition to the Duval toolkit. It targets the earliest stages of fault development — before gas concentrations reach the threshold where Triangle 1 becomes reliable.
Duval Triangle 1
✓ Classifies well-developed faults
✓ 50+ years of field validation
✓ Standardized in IEC/IEEE
✘ Cannot assess fault severity
✘ Requires sufficient gas levels
✘ Blind to early-stage faults
LEDT (Low Energy Degradation Triangle)
✓ Detects incipient faults early
✓ Gives operators time to plan outages
✓ Effective for ester-filled transformers
✘ Newer method — less field data
✘ May over-alert (false positives)
✘ Not yet in major standards
5. Part 4 — Advanced Topics: AI, Nanofluids, and Multi-Method Fusion
5.1 The Severity Problem — Solved by AI (2025 Research)
The single biggest criticism of the Duval Triangle has always been: it classifies the fault type, but not the fault severity. A T1 classification could mean a minor hot spot requiring monitoring, or it could mean incipient failure. How severe is it?
A landmark 2025 IEEE paper addressed this gap using artificial intelligence combined with synthetic data generation (SDG). Analyzing over 1.1 million data instances across D1, D2, T1, T2, and T3 fault categories, the researchers discovered clear severity thresholds within the Triangle 1 zones:
| Severity Indicator | Threshold | Interpretation |
|---|---|---|
| %CH₄ > 43% | Indicates severe faults | Points to T3 or D2 with high confidence |
| %C₂H₄ < 29% | Suggests low-to-medium thermal | T1 or T2 — not yet critical |
| C₂H₂ absolute > 35 ppm | Rapidly increasing | Active arcing — urgent action required |
| CO₂/CO ratio < 3 | Paper degradation | Solid insulation damage — high severity |
5.2 Adaptation for Nanofluids and Ester Oils
As the industry transitions toward more environmentally friendly dielectric fluids (natural esters, synthetic esters, silicone oils), a critical question has emerged: do the Duval Triangle zone boundaries still apply?
A December 2024 study published on IEEE Xplore tested this directly and found that:
- Nanofluids produce significantly more total dissolved combustible gas than mineral oil under identical fault conditions
- Using unmodified Duval Triangle 1 on nanofluids caused systematic fault misclassifications — PD faults were misclassified as T1, and T1 faults appeared as T2
- The researchers proposed adjusted zone boundaries specifically calibrated for nanofluid applications
5.3 Multi-Method Fusion: No Single Method Is Perfect
A comprehensive 2025 comparative study evaluated four DGA interpretation techniques — IEC 60599 Ratio Method, Duval Triangle, Duval Pentagon, and the 90th Percentile Gas Population method — against 2,538 real transformer oil samples. The key finding:
5.4 Software Tools for Automated DGA Interpretation
Manual plotting of Duval Triangles is tedious and error-prone. Modern DGA interpretation is increasingly automated:
| Tool | Platform | Capabilities |
|---|---|---|
| DGA_Diagnostic (MATLAB) | MATLAB / GNU Octave | Triangles 1, 4, 5; trend analysis; IEC/IEEE compliance reports |
| TriboTech DGA Calculator | Web-based | Triangle 1 + Pentagon 1; sampling drift analysis; Monte Carlo error estimation |
| Reinhausen DGA Expert | Commercial software | Full Duval suite + proprietary algorithms; integrated with monitoring hardware |
| Open-source (GitHub) | Python / JavaScript | Multiple implementations available; customizable for research |
6. Part 5 — Practical Workflow: From Oil Sample to Diagnosis
6.1 The Complete Diagnostic Protocol
🛠 Recommended DGA Diagnostic Workflow
1 Verify data quality. Check that the lab report includes all seven gases (H₂, CH₄, C₂H₆, C₂H₄, C₂H₂, CO, CO₂) plus moisture. Confirm sampling date, oil temperature at sampling, and transformer load at time of sampling.
2 Check absolute gas levels. Compare each gas against IEEE C57.104 Table 1 condition thresholds (normal / caution / warning / alarm). If ALL gases are in the normal range, stop — no fault diagnosis needed. Continue only if at least one gas exceeds caution level.
3 Run Duval Triangle 1. Calculate %CH₄, %C₂H₄, %C₂H₂. Plot the point. Record the primary fault classification (PD / T1 / T2 / T3 / D1 / D2 / DT).
4 Refine with secondary tools:
If PD, T1, or T2 → run Triangle 4 (check for stray gassing vs. genuine fault)
If T2 or T3 → run Triangle 5 (check for paper involvement)
If D1, D2, DT, or boundary case → run Pentagon 1 (cross-verify with all 5 gases)
5 Check paper condition. Calculate CO₂/CO ratio. If < 3, suspect paper degradation. If CO is rising rapidly, investigate cellulose hot spots.
6 Run cross-verification. Apply at least two additional independent methods: IEC 60599 ratios, Rogers ratios, Key Gas Method, or Doernenburg ratios. Look for convergence — methods that agree increase confidence.
7 Integrate trend data. Plot gas concentrations over time. A gas at 15 ppm that has been stable for 5 years is very different from a gas that rose from 2 to 15 ppm in 6 months. Duval Triangle should be re-run at each sampling interval to track fault evolution.
8 Formulate the diagnosis. Combine all results into a single assessment: fault type (from Triangle 1), severity (from trend + AI analysis), paper involvement (from CO₂/CO + Triangle 5), confidence level (from method agreement), and recommended action.
6.2 Three Real-World Case Studies
Case 1: Clear T3 — Severe Overheating
| Gas | H₂ | CH₄ | C₂H₆ | C₂H₄ | C₂H₂ | CO | CO₂ |
|---|---|---|---|---|---|---|---|
| ppm | 60 | 180 | 110 | 950 | 4 | 85 | 680 |
Triangle 1: %CH₄ = 15.9%, %C₂H₄ = 83.8%, %C₂H₂ = 0.4% → T3 (severe thermal, >700 °C)
Triangle 5: T3-H (hot spot in oil, no paper carbonization yet)
CO₂/CO: 8.0 → paper not involved. Consistent with Triangle 5.
Action: Plan outage within 3 months. Inspect for loose connections, circulating currents, or blocked oil flow. Transformer can continue operating under increased monitoring.
Field Confirmation: Loose bolted connection on LV bushing found during inspection. Localized hot spot had carbonized the oil but not the paper insulation.
Case 2: Clear D2 — High-Energy Arcing
| Gas | H₂ | CH₄ | C₂H₆ | C₂H₄ | C₂H₂ | CO | CO₂ |
|---|---|---|---|---|---|---|---|
| ppm | 420 | 120 | 15 | 65 | 85 | 240 | 2900 |
Triangle 1: %CH₄ = 44.4%, %C₂H₄ = 24.1%, %C₂H₂ = 31.5% → D2 (high-energy arcing)
Pentagon 1: D2 — agrees with Triangle 1
CO₂/CO: 12.1 → paper may not be directly involved, but CO absolute value is elevated
Key finding: C₂H₂ at 85 ppm and 31.5% is unambiguous — this is an active arc. The H₂ spike (420 ppm) combined with high C₂H₂ is the classic arcing fingerprint.
Action: Trip the transformer immediately. Do not re-energize until the arc source is located and repaired.
Field Confirmation: Winding flashover between HV winding and grounded core at a damaged insulation point. Transformer required rewind.
Case 3: Ambiguous — Multi-Method Resolution
| Gas | H₂ | CH₄ | C₂H₆ | C₂H₄ | C₂H₂ | CO | CO₂ |
|---|---|---|---|---|---|---|---|
| ppm | 180 | 55 | 35 | 48 | 8 | 540 | 5100 |
Triangle 1: %CH₄ = 49.5%, %C₂H₄ = 43.2%, %C₂H₂ = 7.2% → DT (mixed boundary zone)
Pentagon 1: T2 — thermal fault only; does not confirm electrical component
IEC ratios: C₂H₂/C₂H₄ = 0.17 (not in arcing range); C₂H₄/C₂H₆ = 1.37 (T1–T2 range)
Triangle 5: C (paper carbonization) — this is the critical finding
Resolution: Despite the ambiguous Triangle 1 result, convergence of Pentagon 1 (T2), IEC ratios (thermal), and Triangle 5 (C) points clearly to a thermal fault with paper involvement. The DT indication from Triangle 1 is a false positive — likely caused by C₂H₂ from a previous minor discharge event that has since resolved.
Action: Plan outage. Investigate paper hot spot. The elevated CO (540 ppm) and low-ish CO₂/CO ratio (9.4, trending down) support the paper degradation hypothesis.
7. Part 6 — Common Pitfalls and How to Avoid Them
| Pitfall | Why It’s Dangerous | How to Avoid |
|---|---|---|
| Using Triangle 1 on non-mineral oils | Zone boundaries calibrated for mineral oil; esters produce different gas profiles. Systematic misclassification. | Note oil type. Use ester-calibrated methods or cross-reference with Key Gas Method. LEDT preferred for early ester fault detection. |
| Over-relying on a single method | All methods have blind spots. IEC ratios return “unresolved” ~40% of the time. Triangle 1 ignores H₂. | Always use at least 3 independent methods. Multi-method agreement = confidence. |
| Ignoring absolute gas concentrations | A transformer with 2 ppm C₂H₂ may plot as D1 due to normalization, but the fault is negligible. | IEC 60599:2022 §6.1: only interpret Triangle 1 when at least one of the three gases exceeds 10 ppm (or is rapidly rising). |
| Failing to trend over time | A single DGA snapshot is a moment in time. A “normal” result today could follow a rapid increase. | Plot Triangle 1 at every DGA interval. Track zone transitions — PD → T1 → T2 → D1 signals escalating fault. |
| Neglecting CO and CO₂ | Triangle 1 only uses three hydrocarbon gases. Paper degradation (the most serious fault) goes undetected. | Always check CO₂/CO ratio. Always note CO trend. When Triangle 5 suggests “C”, take it seriously. |
| Assuming the Duval Triangle is infallible | Even with 50+ years of validation, the method has known limitations — especially with mixed faults and new oil types. | Treat the Duval Triangle as a powerful but imperfect tool. Combine with physical inspection, electrical testing, and operational history. |
| Misinterpreting stray gassing as PD | New transformers can produce H₂ and CH₄ without any fault — stray gassing of the oil itself. | Use Triangle 4 to differentiate stray gassing (S zone) from genuine PD. Monitor for 3–6 months before concluding it’s a fault. |
References & Further Reading
- IEC 60599:2022 — Mineral oil-filled electrical equipment in service — Guidance on the interpretation of dissolved and free gases analysis. Annex B: Duval Triangle 1 method.
- IEEE C57.104-2019 — IEEE Guide for the Interpretation of Gases Generated in Mineral Oil-Immersed Transformers. Annex D: Duval Triangle and Pentagon methods.
- CIGRE TB 771:2019 — DGA Interpretation Systems — Complete zone boundary definitions for Duval Triangles 1, 4, 5 and Pentagons 1, 2.
- Aliyari, Samimi & Sanaye-Pasand (2025). Comparative Accuracy of IEC Ratio Approaches and Duval Triangle/Pentagon in Transformer Fault Diagnosis from DGA Data. ICEMD 2025. — Comprehensive evaluation against 2,538 samples.
- IEEE (2025). Enhancing Severity Level Detection of Transformer Discharge and Thermal Faults Using Artificial Intelligence in Duval Triangle 1. — AI-enhanced severity classification with 1.1M+ data instances.
- IEEE (Dec 2024/Jan 2025). Adaptation of the Duval Triangle for Enhanced Dissolved Gas Analysis and Diagnosis of Dielectric Nanofluids. — Nanofluid calibration and modified zone boundaries.
- Ngubane, M.M. (2024). Condition Assessment Using Dissolved Gas Analysis on Transmission Power Transformers. University of Cape Town. — LEDT evaluation for incipient fault detection.
- Reinhausen (2025). Duval Method FAQ — DGA Interpretation for Power Transformers. — Practical implementation guidance and error analysis.
- TriboTech ApS (2025). Inside Duval Triangle 1 and Pentagon 1 — Coordinates, Zones, and Live Diagnostics. — Detailed coordinate calculations and worked examples.
- Duval, M. & Buchacz, J. (2022). Refined Duval Triangle 1 with D1-P/D2-P and D1-H/D2-H Sub-Zones. — Latest zone refinement for distinguishing paper vs. oil arcing.
- IEEE (2025). Review of Dissolved Gas Analysis (DGA) for Transformer Fault Detection from Standards to AI. — Comprehensive historical review from 1940s to present.
- ScienceDirect (2025). Experimental Investigation of Transformer Fault Diagnosis Using Integrated DGA and Physical Inspection. — Field-verified validation of Duval and other methods.