Transformer, Liquid-Filled
ASTM D3612 measures the gases dissolved in electrical insulating oil by gas chromatography. Three procedures extract the gas: Method A by vacuum extraction, Method B by a stripper column with the oil injected directly, and Method C by headspace sampling. The gases are hydrogen, oxygen, nitrogen, carbon monoxide, carbon dioxide, methane, ethane, ethylene and acetylene, with propane and propylene possible. Each is reported in microliters per liter, parts per million by volume, corrected to 0 C and 760 torr. Interpreting the result is outside the method; that belongs to IEEE C57.104.
Whenever internal arcing, partial discharge, overheating, or cellulose degradation occurs inside a liquid-filled tank, the energy released breaks down the liquid and insulation into smaller molecules, releasing characteristic gases that dissolve into the remaining liquid. Different fault types produce different gas signatures - for example, arcing favors acetylene and hydrogen, while sustained overheating favors methane and ethane - so a DGA result can reveal a developing fault long before it manifests as a failure. A rising trend over time is an earlier warning than a single snapshot, which is why NETA maintenance protocols call for periodic DGA sampling alongside the rest of the insulating-liquid screen and dissolved-gas analysis to detect slow changes before they become catastrophic.
Acceptance: baseline dissolved-gas analysis is taken before a new or reconditioned transformer is first energized, to establish normal reference levels for that unit so future samples can be compared against its own history rather than against fleet-wide tables. Maintenance: the same test is repeated on a periodic interval, typically annually or bi-annually depending on equipment age and condition class, and is always captured immediately after an event that could generate internal gas such as a through-fault, bushing work, or an unexpected pressure rise, so the evolution of fault gases can be tracked over days, weeks, or months and trended alongside winding temperature estimates and dissolved-gas interpretation methods.
The sample is taken in a syringe or a rigid metal cylinder, kept from light and air, and tested as soon as practical. Method A extracts the gas under vacuum and corrects each gas for incomplete extraction; Method B strips the gas from an injected oil sample in a column and corrects by a measured efficiency; Method C equilibrates 15 mL of oil in a sealed vial at 70 C and samples the headspace. A gas bubble in the syringe is handled per method, and Method C reports a sizable bubble. Detection limits differ: hydrogen 5 uL/L by Method A and 20 by Method B, carbon oxides 25 and 2, atmospheric gases 50 and 500. Silicone is not recommended for Method B on an instrument that also runs mineral oil.
The concentration of each gas in uL/L at 0 C and 760 torr; for a gas the laboratory reports as below its detection limit, the detection limit is entered as the value and the below-detection-limit box is ticked. Also the method letter, the sample identification, the oil temperature at sampling, the presence of a sizable bubble for Method C, and the total dissolved gas, percent, which sums every gas including oxygen, nitrogen and carbon dioxide and is not the same as total dissolved combustible gas.
ASTM D3612 supplies the measurement, not the judgment. The seven fault gases are rated against IEEE C57.104-2019 status levels, chosen by the oxygen to nitrogen ratio and transformer age; a gas with the below-detection-limit box ticked counts as zero there, and a blank gas is left out. Total dissolved combustible gas is defined by neither document: D3612 does not define it, and IEEE C57.104-2019 removed the total-combustible-gas interpretation of its previous edition, naming the quantity only in its gas-space annex. The precision data cover mineral oil only and differ by gas and method, with reproducibility between laboratories of roughly 40 to 90 percent of the value for the fault gases, so a single result is best read as part of a trend. Method B hydrogen is the least reliable figure at low levels.
The equipment classes and standard sections that call for this test, and the UltraDb forms that carry it.
UltraDb references ASTM, ANSI, IEEE, NETA, NFPA, NEMA, FAA, TIA, and IEC standards descriptively, to identify the published methods its forms are built from. UltraDb and Verlecta are not affiliated with, endorsed by, certified by, or licensed by NETA, ASTM, ANSI, IEEE, NFPA, NEMA, FAA, TIA, IEC, or Megger. Standard names and section numbers are the property of their respective organizations. Always test to the edition your contract specifies. Where a standard is sold through the ANSI webstore, Verlecta may earn a commission on a purchase made through our link, at no extra cost to you. A commission never decides which standards we cite or how a form is built.