Why Porosity Appears and What It Tells You
Why this matters
A pore is a record. It says the pool dissolved more gas than the freezing solid could hold, and that the gas was still trying to get out when the metal closed over it. Its shape says how fast it was escaping, its position says where it came from, and its distribution says whether the source was constant or intermittent. It is also the most misread defect on a weld, because the two instincts it triggers - more gas, more heat - are corrections for two of its causes and actively wrong for the rest. The sharper point, and the one this card exists to make, is that porosity is the gas you can see leaving. The gas that stayed in is a different problem with a different timetable.
Before touching a contaminated surface with an arc: never degrease with a chlorinated solvent anywhere the vapour can reach an arc. Ultraviolet from the arc decomposes chlorinated hydrocarbon vapour into phosgene and other irritant gases, an inhalation route with delayed onset, and ANSI/AWS Z49.1 in the edition your employer's written program adopts requires vapour degreasing to be separated from welding by distance and ventilation rather than managed at the arc. Welding through paint, primer, galvanizing or an unknown coating puts that coating's decomposition products in the breathing zone: zinc oxide from galvanizing, lead from old paint under 29 CFR 1910.1025 or 29 CFR 1926.62 in construction, hexavalent chromium from chromate primers and stainless under 29 CFR 1910.1026 or 29 CFR 1926.1126, and cadmium from plated hardware under 29 CFR 1910.1027. Each needs local exhaust at the arc or a respirator under a written program meeting 29 CFR 1910.134, and none is addressed by gloves or a face shield. Grinding out a rejected weld is its own exposure by the same route, with capture at the tool. Where the inspection method is radiography, the shot area is a controlled area under 29 CFR 1910.1096 and nobody welds inside it while a source is out.
The mechanism in one paragraph
Liquid steel dissolves hydrogen, nitrogen and, to a lesser degree, carbon monoxide formed in the pool. Solid steel dissolves far less of any of them. When the pool freezes the solubility collapses and the dissolved gas comes out of solution as bubbles. A bubble that reaches the surface before the metal closes leaves, and you never know it existed. One that does not is a pore. That race between bubble escape and the solidification front is why the same contamination gives fine scattered pores in one weld and long wormholes in another.
The field key
| What you see | What it usually means | What to change |
|---|---|---|
| Fine pores scattered evenly through the deposit | Continuous low-level contamination or marginal shielding across the whole run | Find the constant source: flow band, liner, base metal condition |
| Pores clustered at arc starts, clean thereafter | Cold, damp metal at the start, or no pre-flow before the arc | Pre-flow, start technique, surface condition at the start point |
| Pores in a line along the root or fusion line | Contamination in the joint faces, a gap drawing air in, or an unpurged root | Joint cleaning, fit-up, root-side purge |
| Elongated pores angled toward the surface, tapering | Gas evolving heavily during freeze with a partial escape path, classically a coating burning under the pool | Remove the coating back from the joint before welding |
| A single cavity at the crater | Shielding lost at arc stop, or a crater filled too fast | Post-flow, crater fill technique |
| Surface-breaking pinholes on an otherwise sound bead | Gas escaping right at the end of freezing | Same causes, later in the race; treat as coverage first |
Read the distribution before the shape. Distribution tells you whether the source was constant, intermittent or positional, and that is what narrows the list. Shape tells you how fast it was coming out, which is useful confirmation and a poor starting point.
Where the gas comes from
Sources fall into three groups and they are worth keeping separate, because the corrections do not transfer.
Moisture, in every form it takes. Condensation on cold metal, damp flux on a covered electrode, humid air drawn into a leaking liner, water in a compressed-air line used to blow off a joint, rain on a plate an hour earlier. Moisture is the dominant hydrogen source in ordinary shop work and the one people most often rule out because the surface looked dry.
Hydrocarbons and coatings. Oil, grease, cutting fluid, paint, primer, plating, galvanizing, marker ink. These decompose under the arc and deliver hydrogen and other species straight into the pool, and they deliver a fume hazard by the same event.
Shielding failures. Low flow, turbulent flow, a bridged nozzle, a draft, a leak. Coverage mechanics belong to the shielding-gas card; what matters here is that a coverage failure feeds the pool nitrogen and oxygen from the atmosphere rather than hydrogen, and usually produces the evenly scattered pattern rather than the clustered one.
The gas you can see and the gas you cannot
Hydrogen that comes out of solution and escapes into a bubble is porosity: visible, findable, rejectable, and mostly a section-loss problem in the amounts a shop normally sees. Hydrogen that stays dissolved diffuses into the hardened band beside the weld and, given a susceptible structure and tensile stress, cracks the joint hours or days later. That is the cracking card's subject and its control is the preheat card's.
The consequence is worth saying plainly: a clean radiograph is not evidence of low hydrogen. A weld can be free of porosity because it froze slowly enough for the gas to escape, or because the hydrogen went into the metal instead of into bubbles. The film cannot tell you which. Controlling hydrogen is a consumable-handling and preheat question and it is not verified by an inspection that looks for pores.
A run of brackets that failed only in the morning
A fabrication shop ran a batch of 24 identical brackets, single-pass fillets, gas metal arc on plain carbon plate, one welder, one day. The customer's specification called for radiography on the full batch. Seven were rejected for scattered porosity. All seven were welded before mid-morning, and nothing after that time was rejected.
What the distribution ruled out immediately. Seven of the first seven and none of the last seventeen is a time pattern, not a rate. A gas mixture problem, a spool contamination problem or a technique problem would each spread failures across the day at some rate. None of them would stop.
Shielding was checked and did not move. Flow read 28 cubic feet per hour with the trigger pulled at 07:40 and 28 again at 13:00, both mid-band for the nozzle in use, and the bottle was not changed. Doors stayed shut, so the roughly 5 mile per hour air movement figure the shielding-gas card gives as the derate point was never approached, and no draft is claimed here.
What was true was in the staging, not the welding. The plates had been stacked outside overnight and brought in at 07:00. Shop air that morning was 68 degrees F at 60 percent relative humidity, which puts the dew point at roughly 54 degrees F. The plates came in near 41 degrees F, about 13 degrees F below that dew point, so the moment they entered the shop the room's moisture condensed onto them. The surfaces looked dry to a hand because the film was thin and the plate was cold, and a cold plate does not feel wet the way a cold glass does. By roughly 09:30 the stock had warmed above the dew point and stopped collecting water, and the porosity stopped with it.
Correction, printed, and what it is not. The control adopted was to stage plate inside for at least 12 hours before welding and, where that is not possible, to warm the joint area and a few inches either side of it until a contact thermometer reads above the shop dew point, then weld. That is a moisture-removal step and it is not preheat. Preheat is a cooling-rate and hydrogen-diffusion control whose value comes from steel chemistry, combined thickness and consumable hydrogen level together, and the preheat card owns that determination. Warming a plate from 41 to above 54 degrees F satisfies the dew point condition and satisfies no preheat requirement whatever. Confusing the two leaves a shop believing it has addressed hydrogen when it has only addressed condensation.
Confirmation. The next batch of 22 was staged indoors overnight and welded with the same wire, bottle, welder and settings. None were rejected. Against 7 of 24 rejected in the first batch, which is about 29 percent, and 0 of 22 in the second, the change is attributable to the one variable that moved.
Check against the sibling rules, with the figures printed. The shielding-gas card states that gas coverage is judged against a flow band with the trigger pulled; this case prints 28 cubic feet per hour at both readings, inside the 20 to 40 band that card states, so shielding is excluded on a figure rather than on an impression. That card also gives roughly 5 miles per hour as the derate point for air movement; this case records doors shut and no measured movement, so draft is not invoked. The preheat card states that a preheat value requires chemistry, combined thickness and hydrogen level together; this case names none of the three and correspondingly claims no preheat, only a dew point threshold of 54 degrees F against a plate at 41 degrees F.
What would have changed the conclusion. If the seven rejects had been spread through the day at roughly the same rate, the dew point finding would be irrelevant and the investigation would move to the constant sources: liner, flow band, base metal condition. If the rejects had been clustered at arc starts within every part rather than clustered in the morning, the answer would be start technique and pre-flow rather than staging. And if the shop had been welding a hardenable grade rather than plain carbon plate, solving the condensation would have removed the pores and left the more serious question open, because the hydrogen that did not form pores went into the metal.
How to verify you found the source rather than a correlation
- Write down the distribution before you write down a cause. Time-clustered, position-clustered, start-clustered or uniform. A cause that does not explain the distribution is not the cause, however plausible.
- Re-read the shielding figures with the trigger pulled and compare them against a stated band, not against memory. A number that did not change cannot explain a defect that appeared.
- Change one variable and re-run a full batch. A single good part proves nothing when the baseline reject rate was under a third.
- Check plate surface temperature against the shop dew point with a contact thermometer whenever stock is staged outdoors or moved between temperatures, and record both numbers. It is a two-reading check that takes under a minute and most shops have never made it once.
- Do not close a porosity finding on a hardenable steel without a separate hydrogen answer. The pores are the visible part of the problem, and the invisible part is on a delay.
References
- 29 CFR 1910.1000 Table Z-1 for welding fume constituents, 29 CFR 1910.134 for respiratory protection, 29 CFR 1910.1025 and 29 CFR 1926.62 for lead in old coatings, 29 CFR 1910.1026 and 29 CFR 1926.1126 for hexavalent chromium, 29 CFR 1910.1027 for cadmium
- 29 CFR 1910.1096 for ionizing radiation areas where radiographic inspection is performed on site
- ANSI/AWS Z49.1, Safety in Welding, Cutting, and Allied Processes, in the edition your employer's written safety program adopts, including its separation requirement for chlorinated solvent degreasing
- AWS D1.1, in the edition invoked by your project specification or the adopted building code, for porosity acceptance criteria applicable to the connection type
- See related: What Shielding Gas Is Doing and What Happens When It Stops; What Cracking Tells You About When It Happened; Why Preheat Exists and What It Is Controlling