65 crack indications in 1,900 mm of weld, and two NDT teams had already signed it off as acceptable.

The job was a tubular with a 60 mm wall. Its longitudinal seam went through the fabricator's UT and passed, then through the contractor's UT and passed again. Two independent sign-offs, both clean.

What the client's cross check found

On the client's cross check, the inspector added one scan the earlier teams had not run: probe on top of the weld cap, beam pointed along the weld. The screen lit up. There were 65 indications between 29 and 32 mm deep, sitting right through the middle of the weld cross section. They were chevron cracks.

Why the first two scans saw nothing

Chevron cracking is hydrogen-induced cold cracking in the weld metal itself, most often reported in submerged arc welds on C-Mn and low alloy steels. It does not look like lack of fusion or a slag line, and that is the whole problem.

• Transverse, not longitudinal. The cracks run roughly across the weld rather than along it, and sit at about 45° to the plate surface. Cut the weld lengthwise and they stack up in a herringbone pattern, which is where the name comes from.
• Mid-wall, tight and flat. At ultrasonic frequencies the crack faces behave almost like mirrors. Sound that hits them edge-on or at a glancing angle is reflected away from the probe.
• A standard weld scan points the wrong way. The usual UT procedure puts the probe on the parent metal on either side of the cap and aims the beam across the weld. That is ideal for sidewall lack of fusion. For a transverse crack, the beam runs almost parallel to the crack face, and very little energy comes back.

So the teams scanned from both sides, saw nothing worth recording, and accepted the weld. Nobody was careless. The scan they ran simply cannot see a flaw in this orientation.

The scan that can find chevron cracks is the one that gets dropped most easily. It needs the cap dressed, it takes longer, and because the cracks can lean either way, it has to be run in both directions along the weld. The literature is blunt about the alternative: angling the probe from beside an unground cap is prone to missing these flaws.

The defect was always there. The only variable was whether someone scanned in the direction that could see it.

Why it matters after hand-over

A planar crack at mid-wall in a main pressure or structural seam does not stay harmless just because it is quiet today. Cyclic load is exactly what grows a flaw like this, and a crack found in service costs far more than grinding a cap in the shop. Codes do not size-assess cracks the way they treat porosity; a confirmed crack is rejected.

On the fabrication side, chevron cracking is a hydrogen problem first, and the usual controls apply: dry flux and consumables, correct preheat and interpass temperature, and hydrogen bake-out where the WPS calls for it on thick sections.

I have watched this on enough jobs to say it plainly. On thick submerged arc seams, a UT procedure that treats the cap scan as optional is only testing the flaws it expects to find.