Most defects that later fail a pressure check were visible on the bench, in ordinary light, within a minute of the torch coming off. They travel further because nobody defined what looking means, not because the inspector was careless. Visual inspection is the cheapest check a shop owns and the one most often reduced to a glance.
A trained eye does something different from a quick scan. It moves in a fixed order, at a fixed distance, with a light held at a fixed angle, and compares what is there against a written expectation rather than a general impression. The order does more work than the talent.
A bead that looks perfect under a ceiling fixture can look like a ploughed field under a low raking light. A smooth convex surface reflects light away from the eye, while a depression or a crack edge catches it and throws a shadow back. Light sitting directly above the work kills that shadow.
Shops settle on a small hand lamp held at a shallow angle, close to the weld, with the eye offset from the light so the viewer looks along the shadow rather than into the glare. A pocket torch does the same job inside a tight engine-bay part. Consistency matters more than the lamp itself.
A finished bead carries three pieces of information at once: how the puddle was controlled, how the travel was paced, and whether the arc was held at a sensible distance. A uniform crown, even ripple spacing and toes that blend without a crease say the welder kept the same relationship with the work throughout the pass.
Ripple spacing is the most readable signal. Even spacing means even travel. Spacing that widens toward the end of a pass usually means the part was heating and the welder moved faster to stop the puddle spreading. Spacing that tightens suddenly means a hesitation, and on thin wall a hesitation is where burn-through appears.
The toe is where a fatigue crack starts on a part that vibrates. A toe that rolls smoothly into the parent metal spreads stress; a sharp transition concentrates it. On an exhaust component that sees cycling load, that difference decides service life rather than appearance.
Then there are the lines that should not exist: a faint dark trace parallel to the bead, a sudden change in ripple pattern, a small arc mark away from the joint. The arc mark is usually a repositioned tack, and on titanium it is also an oxidation site. The parallel line is often a lap where two passes never merged.
On stainless steel and titanium the colour of the finished weld is a rough log of how well the hot metal was protected. A clean straw or faint blue tint across bead and heat-affected zone indicates the shield held. A dull grey or chalky white surface on titanium means the metal reacted with the atmosphere while it was still hot enough to absorb it.
Colour has to be read with the material in mind. Stainless tolerates more of it than titanium before the surface oxide becomes a problem, and the same tint means different things on a fillet and on a butt joint. What matters is that the shop defines an acceptable band, photographs a reference sample and keeps it at the bench.
Some of the worst findings at final inspection were built in before the first tack. A joint with an uneven gap closes on one side and opens on the other, and the welder fills it. The result looks acceptable, the part has pulled out of alignment, and no amount of dressing brings it back.
This is why the trained eye looks at the assembly as well as the weld. Sighting along a tube run, checking that a flange sits square to a straight edge, running a thumb along a seam to feel a step: crude methods that find real problems. A step at a seam is a mismatch that existed before welding and got locked in by shrinkage.
A visual inspection cannot see lack of fusion at the root, cannot measure penetration and cannot prove a joint is sound. It can only tell you that nothing is visible from the surfaces you can reach. That distinction has to be stated plainly in the record, because a visual pass is often written up as a full release.
Where the root is inaccessible, the visual becomes a check on the process rather than the result. You confirm preparation, purge, fit-up and colour, and accept that the root is inferred from those. That is legitimate when the procedure is controlled and repeated. It is not legitimate on a one-off part built entirely by eye.
An inspection that is not recorded is an opinion. The record needs enough structure that a second person can repeat the check and reach the same conclusion: the part, the joint, the lighting, the reference sample used, and the features examined.
Short and specific beats long and vague. A note saying the bead is acceptable, colour within reference, toe blend good at the inlet and slightly undercut at the outlet flange, tells the next person where to look. A note saying the weld is fine tells them nothing and moves the risk to whoever signs the dispatch note.
A photograph is a useful attachment and a poor substitute. It flattens the surface, loses the shadow that revealed the defect, and depends on a camera the shop may not control. Keep the photograph, but write the note as though the photograph did not exist.
The value of visual inspection is not that it finds every defect. It is that it makes the state of the work visible to everyone, including the person who welded it. When a welder knows the part will be examined in a fixed order, under a raking light, against a reference sample, the standard becomes a physical object on the bench.
That is where a shop's public description of its work has to hold up. The claims on our company profile are only meaningful if the checks behind them are routine, from titanium welding parts through to fabricated titanium components, all leaving through the same bench.
Inspectors who work this way stop treating the bench as a gate at the end of the line and start treating it as feedback. When the same finding appears twice in a week, the answer is a change in the fixture, the tack sequence or the joint preparation, not a stricter inspection.
For buyers, the practical question is not whether a supplier inspects, since everyone says they do. It is whether the inspection has a fixed method, a reference kept at the bench and a written trace. You can see the range of work that passes through that routine on our other parts pages, and the custom scope in our customisation section. Trade enquiries run through the dealership page.
