How Much Radiography Is Enough? Weld Porosity, Inspection Coverage, and Confidence in Strategic API 650 Potable-Water Storage Tanks for Emergency Preparedness

Abstract Background Potable-water tanks built to API 650 are accepted on a package of examinations, but only a sample of the shell welds is radiographed (RT), and the adequacy of that sample is rarely examined with field data. Methods Case study of the inspection records of two nominally identical, field-erected A36 steel tanks (five shell courses, 10–20 mm plate). Rejections were analysed by shell course, RT batch, joint location and listed welder with exact confidence intervals and exact tests, and two sampling scenarios estimated the chance of detecting defective seams. Results Of 86 traceable RT spots listed (122 planned for both tanks), 67 were radiographed and 6 (9.0%; 95% CI 3.4–18.5%) were rejected for porosity. All six were in one tank and one RT batch, in the two 10 mm upper courses, on four joints credited to four different welders; the same joints in the other tank were accepted. In the 10 mm fourth course, rejection was 5/12 versus 0/12 (risk difference 41.7 percentage points, 95% CI 8.7–68.0; Fisher exact p = 0.037). After repair all rejected spots were accepted; all other examinations reported no indications, and both tanks passed the hydrostatic test at the 8 m design liquid level. A 25% random sample would include a single defective seam in 25% of cases, and would hit a cluster of 3 to 6 defective seams in 59–84% of cases, after which extending RT to the whole batch would find all of them. Conclusions Porosity clustered by batch rather than by welder, thickness or joint type. Batch-triggered extension of RT, recording of consumable and ambient conditions, and documentation of any reduction in RT scope are proposed for consideration. The findings come from two tanks and support analytical rather than statistical generalisation.