The weld tests had 3" backing, the root was clearly fused, the line was 1/16" apprx' outside where the root should be.
it appeared to be excess material that bled under the b/up strip away from the root.
I had previously discussed with the owner (a level III) of the test facility whether to remove the backing prior to submission of the plates. He told me they could be removed, but did not need to be due to their size.(3") The b/up would not interfere with the welds location.
Here is a similar post from 10-8-14:
(Our track marks were clearly not undercut though.)
Non-relevant indications are common in radiographic examination of groove welds with backing. If the backing is not fit up very tight (<0.015" clearance) to the base metal, an undercut can occur at each side of the root pass, which looks like lack of penetration or lack of fusion on the radiographic film. This indication is common enough that it has its own name, "railroad tracks." It sounds like this condition was verified, when the contractor "cut several of the coupons." I would like to see photos of the coupon cross section and copies of the rejected film, if they are available.
Although AWS D1.1, Table 4.11 specifies bend tests to qualify welders, and although Table 4.11, note b allows radiography in lieu of bend tests, it is inappropriate to nullify a rejected radiograph with a bend test. AWS D1.1, Paragraph 6.17.3 indicates that the correct way to evaluate a possible non-relevant indication due to surface conditions would be to blend the surface and reinspect, because the surface condition "may cause objectionable weld discontinuities to be obscured in the radiograph." In this case, the contractor should remove the backing ring and blend per AWS D1.1, Paragraph 6.17.3.2, without violating minimum wall, and perform a new radiographic examination. If the indication is truly non-relevant, the new radiography will not show the previous indication.