What a Weld Actually Is and What It Is Not

Why this matters

A weld is a small casting poured in place in the middle of a wrought part. Everything that makes welding useful and everything that makes it fail comes from that one fact. It is the reason a weld is not a fastening you can inspect by looking at it, not a braze, and not "the same metal, made continuous." This card is about what a weld is not, because every one of those four wrong models is in daily use in small shops and each one leads to a specific, predictable field failure.

Before an arc is struck: the arc burns corneas and skin with ultraviolet at distances well past the welder's own station, so screen the arc and give every person with line of sight to it a filter meeting the minimum protective shade for the process and current in OSHA's filter-lens table at 29 CFR 1910.133(a)(5). Welding fume is an inhalation hazard whose constituents come from the base metal, the coating and the consumable rather than from how thick the smoke looks: manganese from ordinary mild-steel consumables carries a ceiling limit in 29 CFR 1910.1000 Table Z-1, stainless generates hexavalent chromium under 29 CFR 1910.1026 in general industry and 29 CFR 1926.1126 in construction, and galvanized coating generates zinc oxide fume. All three need local exhaust at the arc or a respirator issued under a written program meeting 29 CFR 1910.134, and a glove and a face shield do nothing about any of them. Fire prevention, ignition control and the fire watch belong to the hot work permit and the hot work cards in this library, under 29 CFR 1910.252(a) or 29 CFR 1926.352 in construction.

The definition that does the work

Fusion welding melts the parent metal on both sides of the joint, lets the two liquid pools become one, and lets that combined pool freeze. Filler metal, where used, joins the pool rather than gluing the faces. The joint that results is continuous metal because it was, for a moment, one body of liquid.

That is a physical event with three consequences that a mechanical joint never has: the parent metal's composition ends up inside the deposit, the deposit solidifies with a cast structure, and the metal around the deposit gets a heat treatment nobody specified. The last of those three is the heat-affected zone and its property changes are owned by the sibling card on what a welded joint changes about the parent metal. The first two are this card's subject.

Not a fastening

A bolt or a rivet holds two separate parts in contact by clamping them. The parts stay two parts, they can be taken apart, and every element of the joint is a manufactured item that arrived with a certificate.

A weld makes the parts one part. Nothing about that is reversible, and nothing in it arrived with a certificate covering the joint as installed. The practical difference on a job: a bolted connection is inspected by counting bolts and checking tightness, both external and both simply true or false. A welded connection cannot be inspected that way, because the thing carrying load is inside.

Not a braze and not a solder

This library covers brazing and soldering in their own cards. The line between them and welding is not temperature, it is whether the parent metal melts.

• Soldering and brazing: filler melts and wets the parent surfaces, which stay solid. The joint's strength comes from the filler and from the geometry of the lap that holds it. The parent metal is heated but never becomes liquid.
• Welding: parent metal melts. The joint is parent metal plus filler, mixed.

That distinction is why a brazed joint needs a lap and a controlled gap while a butt weld does not, and why brazing can join dissimilar metals that welding cannot. It also means the failure vocabulary does not transfer. A braze fails by dewetting, voids in the capillary gap, or the filler's own strength. A weld fails by things that only exist in a casting or a heat treatment: solidification cracking, trapped gas, unmelted sidewall, a hardened band.

Not "the same metal, continuous"

The weld deposit is not the filler you bought. It is the filler diluted by however much parent metal melted into the pool. Dilution is the percentage of the deposit that came from the parent, and it varies by process and joint far more than most shops expect: a covered-electrode fillet is commonly in the low tens of percent, a spray-transfer wire deposit runs higher, and a submerged-arc single pass can be around half. Your procedure and your process supplier own the number for your setup; the point that travels is that it is never zero.

So the filler certificate describes the wire or the electrode, not the joint. If the parent is what the drawing says, the difference is usually small and the filler was chosen with dilution already in mind. If the parent is something else, the deposit inherits whatever it brought - carbon, sulfur, copper, an unknown alloy from a reclaimed plate - in proportion to the dilution.

And the deposit is cast. It solidifies from the fusion boundary inward with columnar grains growing toward the centre of the pool, it segregates low-melting-point impurities toward the last liquid to freeze, and it shrinks against restraint as it cools. Those are casting behaviours, and wrought plate has none of them. That is exactly why a plate can be perfectly sound and the weld across it can be the weak line.

Not the bead the customer looks at

The visible bead is one region of four and the only one anyone photographs. A weld you can see the whole face of is a weld whose fusion boundary, root and interpass surfaces you cannot see at all. Surface inspection is worth doing; it is just not a statement about the interior. Which method reaches which defect is covered in the card on lack of fusion.

Worked example: a lifting lug on a reclaimed plate

A shop is asked to weld a lifting lug onto a plate salvaged from a decommissioned frame. The drawing for the new work calls for a plain low-carbon structural plate. The salvaged plate is unmarked.

The setup, with the illustrative values labelled as such. Say the filler deposits 0.08 percent carbon undiluted, which is representative for a common low-carbon structural electrode. Say the salvaged plate turns out on a mill certificate chased down later to be a medium-carbon grade at 0.35 percent carbon. The joint is a two-pass fillet with a covered electrode, and we will use 30 percent dilution as an illustrative figure in the range that process and joint commonly produce, not as a spec value.

Correction, printed: the certificate figure is undiluted, and the deposit is not. The 0.08 percent on the filler's test certificate is the composition of an all-weld-metal test pad made under the conditions of its AWS A5 filler metal specification, with essentially no parent metal in it. That number already contains "no dilution" as a condition. Re-basing it to this joint is not an addition, it is a weighted average of two sources:

• Filler share: 70 percent of 0.08 = 0.056 percent carbon.
• Parent share: 30 percent of 0.35 = 0.105 percent carbon.
• Deposit carbon: 0.056 + 0.105 = 0.161 percent.

The deposit carries about twice the carbon of the electrode that made it, and 0.105 of the 0.161, which is 65 percent of the carbon in the weld metal, came from a plate nobody selected.

What that changes. Carbon is the dominant term in hardenability, so a deposit at 0.161 percent hardens on cooling in a way a deposit at 0.08 percent does not. The parent metal beside it, at 0.35 percent, is more hardenable still. Whether that matters depends on how fast the joint cools, which the heat-affected-zone card governs, and on what preheat is required, which the preheat card governs by chemistry, thickness and hydrogen level together. Neither number is decidable from the filler certificate alone, and that is the whole finding.

Check against the sibling rules, with the figures printed. The preheat card states that a preheat value needs chemistry, combined thickness and consumable hydrogen level together; this example has chemistry (0.161 percent deposit, 0.35 percent parent) and neither of the other two, so it names no preheat temperature and routes instead. The heat-input card states that arc energy compares two of your own procedures rather than describing one absolutely; this example quotes no heat input, so it makes no claim that rule governs.

The failure mode this produces in the field. Not a bad-looking bead. The lug was welded, the bead was even and well-tied, and it passed a visual check the same afternoon. It cracked at the toe some days later, which is the delayed pattern the cracking card reads as hydrogen-assisted, and the shop's first response was to blame the electrode. The electrode was fine. The shop's model of the joint was wrong: it believed it had a joint made of filler metal, when it had a joint made of a mixture it never characterised on a parent metal it never identified.

What would flip the recommendation. If the plate had carried legible marking matching the drawing, dilution would still be roughly 30 percent, but it would be diluting with a low-carbon plate and the arithmetic lands near 0.08 rather than 0.161. Same process, same technique, same dilution, no finding. Dilution is only a problem when you do not know what you are diluting with, which is why verifying the material before you cut is worth more here than any change to the welding.

How to verify you got this right

• Can you name the parent grade from a document? A marking, a mill certificate, a line list or a drawing the piece is traceable to. "It came off the rack" and "it is the same size" are not identifications. Where the grade cannot be established and the joint carries load, that is an engineer's question, not a welder's.
• Does your filler choice account for dilution rather than assume none? If the only basis is that the filler's published strength exceeds the parent's, ask what condition that figure was published under: a filler's stated properties come from a test in a defined position and heat treatment condition set by its AWS A5 specification, and an as-welded joint in another position is not that test.
• Have you distinguished the three questions? What the filler is, what the deposit is after dilution, and what the parent metal beside the joint became. Confusing any two is the model error this card exists to break.
• Grinding is its own exposure, not a continuation of the weld. Grinding stainless or a chromate primer releases hexavalent chromium, galvanizing releases zinc oxide, and old paint can release lead under 29 CFR 1910.1025 or 29 CFR 1926.62 in construction. Each is an inhalation route needing capture at the tool or a respirator under a 1910.134 program, and none is addressed by eye protection.

References

• 29 CFR 1910 Subpart Q (welding, cutting and brazing) and 29 CFR 1926 Subpart J (construction), with 29 CFR 1910.252(a) and 29 CFR 1926.352 for hot work fire prevention
• 29 CFR 1910.133(a)(5) filter-lens table for minimum protective shade; 29 CFR 1910.134 respiratory protection; 29 CFR 1910.1000 Table Z-1 for manganese and zinc oxide fume; 29 CFR 1910.1026 and 29 CFR 1926.1126 for hexavalent chromium
• ANSI/AWS Z49.1, Safety in Welding, Cutting, and Allied Processes, in the edition your employer's written safety program adopts; it binds through that program rather than on its own
• AWS A5 series filler metal specifications, which define the test condition, position and heat treatment behind every published filler property
• See related: What a Welded Joint Changes About the Parent Metal; What Brazing Does That Soldering Cannot; What a Soldered Joint Actually Is; What Lack of Fusion Is and Why It Hides