What a Welded Joint Changes About the Parent Metal

Why this matters

A weld is not something you add to a member. It is a length of the member removed and replaced with three materials you did not order: cast weld metal, a fusion boundary, and a band of parent metal that never melted but came back different. The band is the heat-affected zone, and it is where most weld failures actually live. Nobody selects it, nobody can see it, and every routine acceptance check a small shop performs looks straight past it. The rest of this article is about what got changed in there and which of those changes your inspection is blind to.

Before striking an arc: welding stainless steel puts hexavalent chromium into the breathing zone, which is regulated under 29 CFR 1910.1026 in general industry and 29 CFR 1926.1126 in construction and controlled with local exhaust ventilation at the arc plus respiratory protection supplied under a written program meeting 29 CFR 1910.134. That is an inhalation route and a glove, a face shield and a leather sleeve do nothing about it. Welding or grinding through a galvanized coating adds zinc oxide fume, and mild steel electrode fume carries manganese, both by the same route and needing the same class of control. Arc radiation burns unprotected eyes and skin at a distance, so screen the arc and use the filter shade for the process and current per 29 CFR 1910.133 and 1910.252(b)(2). Hot work needs its fire prevention and a watch under 29 CFR 1910.252(a) in general industry or 1926.352 on a construction site, with NFPA 51B in the edition your authority having jurisdiction has adopted or your insurer requires.

The three zones, and which one you chose

• Weld metal. Melted and resolidified. Its composition is the filler diluted by whatever parent metal melted into it, and its structure is cast, not wrought. You chose the filler, so you chose most of this.
• The fusion boundary. A narrow transition where the two mixed incompletely. Lack of fusion, if it exists, is here, and it is invisible from outside.
• The heat-affected zone. Parent metal that never reached melting but got hot enough to change. Grain size, hardness, temper, and corrosion behaviour all change here, in a band whose width is set by how much heat went in and how fast it left. You did not choose any of it. It is a consequence of the procedure.
• Unaffected parent metal. What the drawing was written about.

What actually changes in that band

Grain size grows next to the fusion line. Coarse grain has lower toughness than the fine grain the mill gave you. On a member that only ever sees static load at room temperature this may not matter; on anything that sees impact, low temperature, or cyclic load, it is the least tough material in the assembly and it sits right beside the highest residual stress.

Hardness in carbon and low-alloy steel is set by cooling rate. Cool a hardenable steel fast from the HAZ peak temperature and you get a hard, brittle structure that is susceptible to hydrogen-assisted cracking when hydrogen from moisture, oil, rust, or a damp electrode is present. This is the cracking that shows up hours or days after the arc is out, which is why AWS D1.1, in the edition invoked by your project specification or by the building code your authority having jurisdiction enforces, requires visual inspection of certain quenched and tempered steels to be delayed at least 48 hours after the weld is finished. Preheat and interpass temperature exist to slow that cooling.

Chromium carbides precipitate in austenitic stainless. Held roughly in the 800 to 1500 degrees F band, carbon in the steel combines with chromium at the grain boundaries. The boundary keeps its carbon and loses its chromium, which is the element the corrosion resistance came from, so a depleted line forms beside every grain boundary. In a corrosive service that line is attacked and the metal comes apart along the grain boundaries in a band parallel to the weld, a short distance out from it, while the weld itself looks perfect. Low-carbon grades and stabilized grades exist specifically to buy time in that band.

Cold work is annealed out. If your member got its strength from being drawn, rolled, or formed cold, the heat takes that strength back in the HAZ and does not give it back. This applies to hard-drawn tube, cold-formed sections, and work-hardened non-ferrous material, and it applies to brazing temperatures as well as welding ones.

Coatings burn off exactly where you least want them gone. Galvanizing, paint and plating are destroyed across the HAZ, which is also where residual tension is highest, so the least protected metal and the most stressed metal are the same metal.

Residual stress is locked in. Weld metal shrinks as it solidifies and the surrounding restraint will not let it, so the joint region ends up in tension whether the structure is loaded or not. Service stress adds to it rather than starting from zero.

The same knob, two opposite directions

Heat input is the one variable a welder controls minute to minute, and it does opposite things depending on what is being welded, which is why a habit carried between materials is dangerous.

Arc energy for a single pass is volts times amps times 60, divided by travel speed in inches per minute, giving joules per inch. That is energy at the arc, not energy in the plate; the fraction that actually enters the work depends on the process, and a buried arc transfers a much higher share than an open arc throwing spatter. Use it as a ratio between two of your own procedures, not as an absolute.

• On hardenable carbon steel, more heat input helps and then hurts. Slower cooling means a softer HAZ and less hydrogen cracking risk. Push it further and the grain coarsens and toughness drops.
• On austenitic stainless, more heat input only hurts. More energy means longer in the sensitizing band, which means more carbide precipitation. There is no softening benefit to trade against it.

If someone tells you they slow down to get a better-looking weld, ask which metal. On one it is a reasonable trade. On the other it is the whole failure mechanism.

Worked example: a repair that passed everything and failed anyway

A stainless line carrying a chloride-bearing condensate cracked at a support. Before the first cut into any line: isolate, drain, vent and confirm zero at the piece you are cutting rather than at an upstream valve, which is the stored-energy duty at 29 CFR 1910.147, and a pipe or container that has held a flammable or its residue is not cut or welded until it has been cleaned or inerted per 29 CFR 1910.252(a)(3). With that done, the shop cut out a short section and welded in a replacement piece with two circumferential butt welds. Six months later both welds leaked, and the leaks were not in the welds.

What the shop actually did. The drawing specified a low-carbon grade. The rack held the standard-carbon grade of the same family and the pieces were dimensionally identical, so that is what went in. The procedure on file called for two passes at 12 inches per minute. The welder ran three passes at 6 inches per minute to get a cap he was happy with.

Put the arc energy side by side. Take 26 volts and 200 amps for both, since only travel speed and pass count changed.

• Procedure as written: 26 x 200 x 60 / 12 = 26,000 joules per inch per pass, two passes, 52,000 joules per inch total.
• Procedure as run: 26 x 200 x 60 / 6 = 52,000 joules per inch per pass, three passes, 156,000 joules per inch total.

Three times the energy into the joint. Both changes, the material substitution and the slower travel, push the same direction on the same mechanism: more carbon available to precipitate, and more time spent in the range where it precipitates.

Where it leaked. Not at the cap, not at the root. A line of pinholes on each side of each weld, roughly parallel to it and set back from the toe by a small, consistent distance. That distance is the signature: it is the band of parent metal whose peak temperature landed inside the sensitizing range and which then cooled slowly enough to stay there. Closer to the weld the metal went above the range and came back through it quickly; further out it never got hot enough. The damage is where the thermal history was worst, which is neither at the arc nor away from it.

What the acceptance actually covered. This is the part worth copying into your own process, because every item on it passed.

Every check looked at the weld. The failure was in the parent metal the weld had changed, and no check was pointed at it. The two steps that would have caught this were both paperwork rather than inspection: verifying the material against the drawing before cutting, and following the heat input the procedure specified.

What would change the answer. If the same line carried clean, non-corrosive water, that sensitized band would still be there and would very likely never do anything. Sensitization is a metallurgical condition, not damage; it becomes damage only when a corrosive service finds it. That is why the same shortcut can be taken for years on one system and produce a failure the first time it is taken on another. It is also why "we have always done it this way" is not evidence on this particular question.

What would flip the correction entirely. Move this repair to a hardenable carbon steel and the slow travel speed is no longer the fault. On that material the risk runs the other way, toward a fast-cooled hard HAZ, and the corrective actions are preheat, interpass temperature control, dry consumables, and a delay before final inspection rather than a faster travel speed.

What to do about it in a small shop

You will not buy a metallurgical lab. Three controls carry most of the risk, and all three are administrative.

1. Verify material before you cut, not after you weld. Confirm the grade against the drawing or the line list, and treat "same family, same size, on the rack" as a substitution request rather than a match. Where a substitution is genuinely needed, route it to the engineer of record or the specifying authority, because the drawing owns that answer.
2. Write the heat input into the procedure and make it a number the welder can hit. Travel speed and pass count are the accessible form of it. A procedure that says only "two passes" leaves the variable that matters unspecified.
3. Qualify to the right document and say which one. Pressure piping and pressure vessels run under ASME Boiler and Pressure Vessel Code Section IX, invoked through the piping code and through your jurisdiction's adoption of it; structural steel welding runs under AWS D1.1 in the edition the specification or adopted building code names. Both bind through adoption or contract, not on their own, and a welder qualified under one is not automatically qualified under the other.

When you grind a weld or a coated surface, treat it as its own exposure rather than an extension of welding. Grinding stainless or a chromate-bearing primer releases hexavalent chromium, grinding galvanizing releases zinc, and grinding a concrete or masonry surface releases respirable crystalline silica under 29 CFR 1926.1153 or 1910.1053. All three are inhalation routes needing capture at the tool or respiratory protection under 1910.134, not eye protection alone.

References

• 29 CFR 1910.1026, hexavalent chromium, and 29 CFR 1910.134, respiratory protection, for stainless welding and grinding fume; 29 CFR 1910.133 and 1910.252(b)(2) for eye and arc protection
• 29 CFR 1910.252(a) (general industry) and 29 CFR 1926.352 (construction) for hot work fire prevention; NFPA 51B in the edition adopted by your authority having jurisdiction or required by your insurer
• AWS D1.1, structural welding code for steel, in the edition invoked by your project specification or the adopted building code, including its delayed visual inspection requirement for certain quenched and tempered steels
• ASME Boiler and Pressure Vessel Code Section IX, welding and brazing qualification, as invoked by the applicable piping or vessel code and adopted by your jurisdiction
• See related: Why a Joint Fails at the Interface and Not in the Metal; Thermal Expansion Mismatch Inside a Joint