Fillet Welds That Pass Visual but Fail UT
A shop is fabricating 220 ft of 3/8-in. fillet welds on girder stiffeners with FCAW and 0.045-in. E71T-1, flat and horizontal. The in-house CWI visually inspected 100% and accepted all of it. The owner then commissioned third-party ultrasonic testing on 14 randomly selected joints; 6 showed lack-of-fusion indications at the root. The welder's qualification record shows SMAW in 6G only - he has been running FCAW for four months because "the machine was free." The WPS calls for 24 V, 210 A, and a maximum 3/8-in. root gap; the machine dial reads 22 V, and the fitter reports gaps up to 5/8 in. are being pulled shut. The customer's inspector arrives in two days.
1. Before you act on the UT results, what do you verify - and what makes this systemic rather than random?
Read the worked answer
Two separate questions.
Verify the NDT first. Fillet-weld UT is technique-sensitive. Confirm the procedure is validated for 3/8-in. fillets, check calibration and scan pattern, and before condemning 220 ft, re-scan two or three of the six joints with a second method or a second Level II. Condemning on one technique is how shops pay for other people's errors.
Then test for systemic cause. Three findings already point that way:
- The welder's performance qualification is SMAW 6G; qualification is process-specific, so he is not qualified for the process he is running.
- Machine at 22 V against a 24 V WPS.
- Gaps up to 5/8 in. against a 3/8-in. maximum force fast travel and poor root fusion.
Any one of those makes the 14-joint sample unrepresentative of the 220 ft. The sample size stops mattering; scope becomes the whole run.
2. What do you tell the customer, and what do you do with the 220 ft?
Read the worked answer
Disclose before the inspector arrives, with the data: sample size, indication type, and the three systemic findings. Inspectors forgive defects; they do not forgive surprises.
Proposal:
- 100% UT of the affected 220 ft using a validated procedure.
- Repair rejectable indications under an approved repair WPS.
- Re-inspect every repair with the same method that found it.
- Requalify the welder on FCAW before he returns to that joint.
Do not offer to "grind out the six." That treats symptoms and leaves 80% of the footage untested. State the rework cost and the schedule impact in the same conversation, and note that a repaired girder is cheaper than a rejected shipment or a field failure - voluntary disclosure is also what preserves your approval to weld for that owner.
3. What single change do you make first on Monday?
Read the worked answer
Pull the welder off FCAW production. Compliance comes first because unqualified welding invalidates every downstream record regardless of how good the arc looks - you cannot defend the weld even if it passes.
Then, in order:
- Machine audit and parameter lock (22 V vs. the 24 V WPS is a WPS deviation, not a preference).
- Fix fit-up at the source - the fitter owns the 5/8-in. gaps, and no welder can fuse a root that is being levered shut.
Sequence matters. Requalifying the welder and then letting him run out-of-parameter on bad fit-up puts you back in this room in a month, with a better-qualified welder making the same defective weld.