6 Common Weld Defects and Fixes

These are the six weld defects that the consumable makers' troubleshooting guides keep returning to in MIG, flux-core and stick work on steel. They are ranked by how much room AWS D1.1 gives each one, strictest first. Cracks, which the code rejects at any size, come first. Spatter comes last, because none of the D1.1 visual acceptance criteria cited here put a limit on it. The code figures come from the static and cyclic tables as summarized by NDT inspectors. The current edition is D1.1/D1.1M:2025-AMD1, effective January 12, 2026, so check your contract's edition before relying on any number below.

Cracks are the one defect with no dimensional allowance anywhere in AWS D1.1. Atlantis NDT's Level III reviewer calls this the single most stringent criterion in the code. Any size, any direction, any location: the crack is rejected and must be repaired.

The Xiris defects guide splits cracks into two types by when they form, and the cause tells you which one you have. Hot cracks form during solidification and point to high sulfur and phosphorus content. Cold cracks form after cooling and point to hydrogen.

The fixes follow from the type:

• Cold cracks: use low-hydrogen consumables and a clean, dry joint.
• High-carbon steel: preheat. UTI flags this steel as crack-prone because it cools fast and unevenly, and preheating lets the metal expand slowly and evenly.

A cracked weld gets cut out and redone. Grinding the surface until the crack disappears is not a repair.

D1.1 requires complete fusion between weld metal and base metal and between successive passes. It does not permit visible lack of fusion.

The word "visible" is the problem, because this defect usually is not. As Greg Siepert put it in The Fabricator, "lack of fusion doesn't show up well," and a clean cover pass hides it.

The causes are consistent across Lincoln Electric, Hobart and Xiris:

• Insufficient heat input or low current
• Travel speed too fast
• Wrong electrode or work angle
• A dirty joint

Hobart adds one that is easy to watch for in flux-core work: the arc running ahead of the puddle. Lincoln adds one people miss. Heavy spatter left on a root or fill pass can produce incomplete fusion on the next pass.

The repair is to remove the unfused metal by grinding, gouging or machining, then reweld it with a qualified welder. Running another pass over the top does not fix it.

Sonibel Instruments has built a torch-mounted acoustic sensor that listens for porosity and lack of fusion in semiautomatic GMAW and FCAW, according to The Fabricator. The coverage gives no detection rates. Those figures would need to be verified against sectioned or UT-inspected welds before the sensor replaces anything.

Overlap is weld metal that rolls over the toe and sits on the base metal without fusing to it. D1.1 does not permit it, according to OneStopNDT's summary of the visual criteria.

It is a fusion defect at the edge of the bead. It ranks just behind lack of fusion because it is at least visible: a rolled, ropey toe with a sharp re-entrant angle.

The sources here say less about cold lap than about any other defect on this list. UTI ties it to inconsistent travel speed and angle and to poor joint prep, and its prevention advice is to hold both steady pass to pass. Lincoln groups it under improper bead profile.

If the joint is cold and the bead is piling up, look at heat input and travel speed together. Removing overlap is not covered on its own in these sources. Treat it as the toe-of-weld fusion problem it is: remove the unfused metal and reweld.

Undercut is the first defect on this list that the code tolerates in small amounts.

Under the statically loaded criteria:

• Base metal under 1 in. thick: undercut is capped at 1/32 in., with 1/16 in. allowed for an accumulated 2 in. in any 12 in. of weld.
• Base metal 1 in. and thicker: 1/16 in. along any length.

On cyclically loaded connections, undercut transverse to tensile stress drops to 0.01 in. That is a depth you check with an undercut gauge, not by eye.

Lincoln explains why the limits are tight: the groove weakens the weld toe and raises the likelihood of cracking. That makes undercut a feeder for defect number one.

The causes are excess current or voltage and too little dwell at the toes. In GMAW spray transfer, arc pressure can also push molten metal away from the toes.

Hobart's fix list:

• Reduce current and voltage
• Correct the electrode angle (Xiris gives 5 to 15 degrees)
• Slow down
• Weave with a pause at each side

For repair, blend minor undercut by light grinding. Fill severe undercut with a stringer bead.

A January 2025 SENLIS roundup calls porosity the most frequent defect. The code allows a measured amount of it.

Under the static criteria:

• CJP butt welds transverse to stress: no visible piping porosity.
• Other groove and fillet welds: the sum of pore diameters of 1/32 in. and larger may not exceed 3/8 in. per linear inch, or 3/4 in. per 12 in.

Cyclically loaded welds allow one pore per 4 in., at 3/32 in. maximum diameter.

Lincoln names wire chemistry as the leading cause and disrupted gas coverage as the second. The fixes, by cause:

• Wire chemistry: ER70S-6 carries the most deoxidizer, at 0.8 to 1.15% silicon and 1.4 to 1.8% manganese.
• Gas coverage: Xiris recommends 15 to 20 CFH and notes that wind as light as 5 mph can strip the shield.
• Self-shielded flux core: keep stickout at 1¼ in. or less. If you see wormtracking, drop the voltage in half-volt steps.
• Stick: keep low-hydrogen rods in an oven.

Porosity is also over-blamed. A September 2026 Aluminum Workshop column puts its strength penalty at roughly its volume fraction. Even a bad 5% weld loses about 5%. The 6061-T6 specimens in that column reached about half their required yield. The likelier culprit was excess preheat softening the heat-affected zone.

On stainless MIG welds, the glassy black spots are silica islands, not porosity. They are leftover deoxidizer from ER308L and ER316L wire, and more heat input makes more of them. The silica islands article calls them cosmetic but notes they can start corrosion in corrosive service. That is one more reason to manage heat input on stainless.

Spatter is last because it is the only defect here with no acceptance limit in the D1.1 visual criteria these sources cite. It is a cleanup cost, not a rejection.

Lincoln lists three causes:

• Wire feed speed too fast
• Voltage too high
• Stickout too long

Xiris ties most spatter to short-circuit transfer, where the wire repeatedly touches the puddle. Its fix is pulsed GMAW, which keeps the arc stable across a wider range of parameters.

The one structural risk is the multi-pass case already covered under lack of fusion. Clean spatter off between passes and it stays a cosmetic problem.

Start with lack of fusion. Cracks are the strictest line in the code, but lack of fusion is the defect most likely to pass the visual inspection D1.1 requires on every weld, and visual inspection is the step NDT supplements rather than replaces. Porosity and spatter fixes are setup changes you can make on the spot. The first three defects on this list are procedure problems, and they only show up on a weld that gets sectioned or scanned.