Weld Defect Troubleshooter
Most weld defects come down to heat, speed, angle, gas and a clean joint. Holes in the bead are usually dirt or lost shielding gas, a groove at the edge is too much current or speed, and a bead sitting on the plate is too little heat. Pick your process and what you see for the causes in the order to check them, then measure the weld against the AWS D1.1 visual limits.
What is wrong with your weld?
Pick the process, then what you can see. You get what the defect is, how strict the structural code is about it, the likely causes with your process's first, and what to change.
Is it within the code? Check a measurement against AWS D1.1
The visual acceptance limits for statically and cyclically loaded nontubular steel connections, D1.1 Table 8.1, as Atlantis NDT tabulates them. Your contract names the edition and may be stricter.
Fractions or decimals: 1/32, 3/64 or 0.047. Millimeters: type the number with mm, 1.2 mm.
Visual limits only. Radiography and ultrasonic testing have their own criteria, tubular connections theirs, and cracks and lack of fusion are rejected at any size whatever these say.
Reading the answer
The causes and fixes come from the makers' and trainers' own troubleshooting guides: Hobart Brothers on self-shielded flux-cored wire, the Xiris welding defects guide and Universal Technical Institute. Each line names its source. Lines in bold are the ones particular to the process you picked: Hobart's 1 1/4 in stickout limit for flux-core, for instance, or Xiris's 15 to 20 CFH of gas for MIG. They come first because they are the ones a general guide leaves out.
Two of the twelve only show for some processes. Slag inclusions need a flux, so they are for stick and flux-core: as Xiris puts it, solid-wire MIG and TIG produce no slag. Wire feed faults (the bird's nest and the burnback) are for the wire processes; Hobart writes about them for flux-cored wire, which is softer than solid wire and needs knurled V-groove drive rolls.
How strict the code is, defect by defect
On structural steel, AWS D1.1 decides which defects are tolerated. Its visual acceptance criteria are Table 8.1 (Table 6.1 before the 2020 edition), and they split by whether the connection is statically loaded, like a building frame, or cyclically loaded, like a crane runway or a bridge. This is how Atlantis NDT tabulates the limits for nontubular connections:
| Defect | Statically loaded | Cyclically loaded |
|---|---|---|
| Cracks | Rejected at any size and location | Rejected at any size and location |
| Fusion | Complete fusion required | Complete fusion required |
| Undercut, base metal under 1 in | 1/32 in; 1/16 in over a total of 2 in in any 12 in | 0.01 in across the tensile stress; 1/32 in otherwise |
| Undercut, 1 in and thicker | 1/16 in, any length | As above: set by the stress direction |
| Piping porosity, CJP butt weld across the stress | None visible | None visible |
| Piping porosity, fillet and other groove welds | Pores 1/32 in and up: 3/8 in total in any inch, 3/4 in in any 12 in | One in each 4 in, 3/32 in diameter at most |
| Undersized fillet | 1/16 in under for a 3/16 in fillet or smaller, 3/32 in for 1/4 in, 1/8 in for 5/16 in and up, over no more than 10% of the length; none at the ends of girder web-to-flange welds for twice the flange width | |
OneStopNDT adds what visual inspection rejects outright: visible cracks, surface lack of fusion, unfilled craters and overlap, with excessive convexity or concavity not allowed either. Spatter has no number; it is usually acceptable unless the project specification says otherwise. The current edition is D1.1/D1.1M:2025, which Atlantis NDT notes revised the undercut and piping porosity criteria, so check the edition your contract names; a contract may also be stricter than the code. The checker in the tool works these limits for a measurement you type. It is for visual inspection only: ultrasonic testing and radiography have their own criteria, as do tubular connections.
Measuring what you see
Undercut is measured with an undercut gauge, and OneStopNDT lists fillet weld gauges and bridge cam gauges as the other everyday tools; 0.01 in is not a depth you can judge by eye. Clean the weld before you inspect it, since, in OneStopNDT's words, dirt, slag or spatter can be taken for porosity. For the porosity check, add up the diameters of the pores 1/32 in and larger in the worst inch of weld, then in the worst 12 in.
The defects that do not show
Lack of fusion and lack of penetration are often inside the weld, where only ultrasonic testing or radiography finds them, and Xiris counts lack of fusion as a planar defect, more dangerous than porosity. A clean cap says nothing about them. That is why the troubleshooter's repair for them is to remove the unfused metal down to sound metal and weld again, not to run another pass over the top; Xiris's repair for overlap makes the same point, that grinding it smooth is not a repair if the metal underneath is not fused. Grinding a weld flush with a flap disc is finishing, not fixing: see the flap disc grit chart for that, and the flap disc picker on this site for the disc.
For the six defects the code ranks strictest, in more depth, read 6 common weld defects and fixes. Stainless brings its own problems with heat; TIG vs MIG for stainless steel covers them, and grinding stainless without discoloration the finish. If the same defect keeps coming back on a production line, cobot welding vs traditional automation looks at taking the hand out of it.
Sources, read on 19 October 2026: Hobart Brothers, Troubleshooting Common Self-Shielded FCAW Problems (updated 5 December 2025); Xiris, Welding Defects Guide; Universal Technical Institute, Welding Defects Explained; Atlantis NDT, AWS D1.1 weld acceptance criteria (Table 8.1 visual limits); OneStopNDT, welding defects acceptance criteria; American Welding Society, D1.1. General guidance for diagnosing a weld, not a substitute for the welding procedure, the code edition in your contract or a certified welding inspector.