Defect guides
What your bead is trying to tell you
Five things a critical eye picks up first — and what each one says about your technique.
Almost every surface defect is a symptom of one of four things: too much heat, too little heat, the arc in the wrong place, or the torch moving at the wrong speed. Learning to read which is which is most of what separates a welder who improves from one who just does more reps.
These guides describe what each defect looks like, what typically causes it, and the single change most likely to fix it on your next pass. They are written for people practising — trade-school students, apprentices, and anyone grinding toward a test.
- Undercut — a groove melted into the base metal along the edge of your bead, left unfilled. Usually heat, arc length or travel speed.
- Porosity — holes left by trapped gas. Almost always contamination or lost shielding, and the only defect here that reliably tells you something about the inside of the weld.
- A ropey bead — high, narrow and lumpy, with coarse ripples. The classic too-cold-and-too-slow combination.
- Running too hot — wide, flat, thin-edged, with burn-through and undercut following it around.
- Running too cold — the most dangerous of the five, because a cold weld can look perfectly tidy and still have almost no fusion.
Before you use any of this
Everything on these pages is about the surface. A photograph — and your own eyes, at the bench — can only tell you what the outside of the weld looks like. Lack of fusion, incomplete penetration, internal porosity, slag inclusions and root defects are invisible from the top, and those are the ones that fail under load.
A bead can look excellent and be structurally worthless. If the answer matters — anything load-bearing, anything over a person — you need a Certified Welding Inspector and RT or UT, not a guide on a website and not an app.
Why there are no acceptance numbers here
You'll notice these guides describe defects without quoting exact tolerances — no "undercut shall not exceed X". That's deliberate. Acceptance criteria depend on the code you're working to, the thickness, the joint, whether the stress is transverse or longitudinal, and which edition of the standard your inspector is holding. A single number lifted out of that context and put on a web page is how people end up confidently wrong.
If you need the real figures, get them from the code itself and from the person who will be inspecting your work. What these pages are for is the other half of the problem: knowing why the defect appeared, so there's less of it next time.