Every other inspection on this site looks at a joint and tries to leave it intact. The bend test does the opposite. It takes a coupon welded to the same procedure, bends it until the metal yields, and then reads the fracture. If the joint survives the bend without opening, the weld did what it was supposed to do; if it opens, the way it opens is the diagnosis.
That is the point of destroying a good weld: a joint can look correct, measure correct and still be attached to the parent metal only at its edges. A bend puts the root and the fusion boundary into tension, which is precisely where that kind of failure hides.
Bending a welded coupon puts one face in tension and the other in compression. Which face you put in tension is a choice, and it is the whole design of the test. Bend the root in tension and you test penetration and root fusion. Bend the face in tension and you test the cap and the fusion line at the toes. Bend the side and you test the through-thickness stack, which is where layered defects show up.
So a bend test result means nothing until you know which face was in tension. A shop that reports a pass without saying which bend it ran has reported a number, not a result.
A nick break takes a different route to the same question. A notch is cut into the weld, the coupon is broken across it, and the fracture surface is examined. Instead of bending the joint, it tears the weld metal apart to expose the throat.
What you are looking for is the character of that surface. A sound weld tears through the base metal or through the weld metal with a dull, fibrous appearance, because the material yielded before it separated. A defective weld shows flat, bright, crystalline facets, because the crack ran along a boundary that was never fused. Once you have seen the difference, it is unmistakable.
A good fracture is boring. It is rough, it is dull, and it does not follow a straight line. It wanders, because ductile tearing follows whichever path requires the most work.
It also shows a clear transition. If the coupon failed in the parent metal rather than in the weld, that is usually the strongest possible result, because it proves the joint is stronger than the material around it. When a fracture runs through the weld metal with the same dull appearance, that is also acceptable, provided the whole throat is continuous.
A bad fracture is diagnostic in a way that a photograph of a bead never is.
Lack of fusion appears as a region of the surface that is flat and shiny, sitting alongside the torn area like a separate plate. It is smooth because nothing joined there, and the crack simply travelled along the gap. Porosity appears as rounded voids, and their position matters: voids along the root line point to shielding or fit-up, while scattered voids across the section point to a wet electrode or contaminated filler. Slag or oxide trapped between passes appears as dark, glassy lines that the fracture steps around.
None of these are visible from outside. The joint passed a visual inspection the day it was made.
The coupon is not cut from a customer's part. It is welded from the same material, with the same filler, on the same machine, by the same operator, in the same position, using the same settings, immediately before or alongside the production run.
That is what makes the result meaningful. A coupon welded on a different day, in flat position, by the supervisor, proves that the supervisor can weld a coupon. It says nothing about the part in the crate.
The bend is a comparative test, not an absolute one. It does not tell you the joint will survive a service life; it tells you the process that made it was capable at the moment the coupon was made.
That distinction matters when a result is marginal. A fracture that is mostly ductile with one small bright patch is not a pass with a cosmetic blemish. It is a signal that the fusion boundary was marginal, and the useful response is to find out why before the next run rather than to record a pass and move on.
Neither test says anything about fatigue, corrosion or the effect of the heat the weld put into the part. A coupon that bends cleanly can still be part of an assembly that distorts, or a joint that sensitises and later rusts at the seam.
Destructive testing answers one narrow question very well: is the metal joined. Everything else on this site exists because that question is necessary and not sufficient.
If a supplier can describe which bend they run, which face is in tension, where the coupon comes from and what a marginal result triggers, they have a process. If the answer is that the parts look fine, they have an opinion.
The test costs a coupon and a few minutes. It is the cheapest evidence in the building, and it is the only evidence that comes from inside the joint rather than from its surface.
