How to Inspect a Weld Without Destroying It

Why this matters

Most of the decisions that determine whether a nondestructive examination will produce a usable answer are made before anyone strikes an arc, and a shop that treats the examination as a service call booked after the fact will pay for a result that cannot be interpreted. Access, surface condition, joint detail and the written acceptance criteria all have to exist first. Buy the examination without them and the report comes back saying an indication was recorded, with no line to judge it against and no way to reach the part of the joint the question was about.

Before any of this: radiography makes the shot area a restricted area under 29 CFR 1910.1096, so nobody enters, welds or works inside the posted boundary while a source is exposed, and the source itself is licensed by the Nuclear Regulatory Commission or by your Agreement State rather than by the shop. Penetrant materials are commonly solvent-based and flammable in aerosol form, so use them away from any ignition source with the ventilation and the chemical-resistant gloves and eye protection named on each product's safety data sheet, and never spray them into a confined space without ventilation meeting 29 CFR 1910.146 for that space. Magnetic particle prods can arc and burn the part and the operator's hand, so use them only on a surface where an arc strike is acceptable and keep the contacts clean and tight. Fluorescent methods use ultraviolet lamps, which need the eye protection the lamp's instructions specify. Grinding a surface to prepare it for examination releases the base metal and any coating as respirable dust, an inhalation route needing capture at the tool or a respirator issued under a written program meeting 29 CFR 1910.134, and gloves do nothing for it.

Step 1: get the acceptance criteria in writing before the joint is detailed

An examination without a stated pass and fail line is a photograph. Ask which document governs, in which edition, and who adopted it. On structural steel that is normally AWS D1.1 in the edition invoked by the project specification or by the building code your authority having jurisdiction has adopted; on pressure work it is normally the applicable ASME code as adopted by your jurisdiction. Both bind through adoption or contract, never on their own.

What you lose by skipping it: the report lands, the indication is real, and nobody in the room can say whether it is rejectable. That argument is settled in the customer's favour every time.

Step 2: choose the method against the flaw you expect and the plane it lies in

Method selection is a geometry problem, and the sibling card on what each nondestructive method can and cannot see owns the capability matrix. The one sentence to carry into this step: every method has a blind orientation, so pick the one whose interrogating direction crosses the plane the expected flaw lies in.

What you lose by skipping it: the classic wasted shot, radiography aimed straight down a joint whose likely flaw is a flat sidewall discontinuity lying nearly parallel to the beam.

Step 3: detail the joint and plan the sequence for the access the method needs

Radiography needs both sides, film or a detector on one and the source on the other. Ultrasonics needs a scanning surface wide enough for the probe to reach the far side of the weld at the refracted angle, clear of stiffeners and clear of the cap unless the cap is dressed. Magnetic particle and penetrant need only the surface, but they need the whole surface.

This is where a backing bar becomes a decision rather than a detail. Backing left in place is an interpretation problem on a radiograph, because the film sees the joint plus the bar, and it is a reflector on an ultrasonic scan.

What you lose by skipping it: an examination that can only cover part of the length, reported honestly as partial coverage, on a joint the contract says is 100 percent examined.

Step 4: set the timing, including any hold

Two timings matter. First, any hold after welding: hydrogen-assisted cracking opens hours to days after the arc is out, so an examination performed while the joint is warm is a true statement about that moment and nothing more. AWS D1.1, in the adopted edition, requires visual inspection of certain quenched and tempered steels to be delayed at least 48 hours after completion, and project specifications commonly extend a hold to the nondestructive examination as well. Second, if the joint gets a post-weld heat treatment, the governing examination is the one after it, because the treatment can both close and open things.

What you lose by skipping it: a clean report dated the same shift as the weld, on a joint that cracks over the weekend.

Step 5: prepare the surface the method actually requires

Each method has a different surface. Penetrant needs a surface clean enough that the penetrant can enter the flaw and the developer can draw it back, which means no paint, no smeared metal and no blasting that peens flaw openings shut. Magnetic particle tolerates more. Ultrasonics needs a smooth, spatter-free scanning surface and a couplant. Radiography is the most tolerant of surface condition and the least tolerant of geometry.

What you lose by skipping it: a false clean. A surface-breaking crack whose mouth was closed by a wire wheel is invisible to penetrant, and the report will say so in the negative.

Step 6: use qualified people and a written procedure

Technician qualification runs through the employer's written practice, which normally adopts ASNT SNT-TC-1A in a stated edition or ANSI/ASNT CP-189, and the practice, not the recommended practice itself, is what binds. Interpretation is usually a Level II function. The written examination procedure states the technique, the calibration or reference standard, the scan plan and the coverage.

What you lose by skipping it: a result nobody can reproduce, which is the same as no result the first time it is challenged.

Step 7: record coverage and the detection limit, not just the verdict

A result reads "no rejectable indications" over a scope: which joints, what percentage of length, from which surfaces, using which technique. Coverage is a number and it belongs on the report. So does the reference standard the equipment was set against.

What you lose by skipping it: the report gets read as a statement about the whole joint when it covered part of it.

Step 8: handle a rejectable indication as a controlled removal, not a grind-and-look

Mark the extent, remove to sound metal, re-examine by the same method with the same coverage after re-welding. The removal itself is a welding operation with its own defect set, and the sibling card on grinding and gouging owns it.

Worked example: six plate splices, one request

A customer asks a shop to "x-ray the welds" on six plate splices in a fabricated frame, 3/4 in material, single-V groove, eight passes each. Walking the steps changed four things before any equipment arrived.

Step 1 result. The contract named AWS D1.1 in a stated edition, statically loaded, and the specification called for radiographic examination of 100 percent of the splices. So the acceptance criteria and the method were both already fixed by contract, which is the easy case.

Step 2 result. Two of the six joints turned out to be corner joints, not butt splices, drawn on a different sheet. Radiography of those two is geometrically poor, and the shop raised it as a request for information rather than shooting them. The engineer changed those two to ultrasonic examination.

Step 3 result. Four of the six had permanent backing bars in the detail. Left in, the radiograph images the bar with the joint. The engineer accepted removal of backing on all four, which added a back-gouge and a back-weld to each of those joints, so 4 of 6 joints gained a pass.

Step 4 result. The material was ordinary hot-rolled structural plate, not quenched and tempered, so no 48-hour delayed inspection requirement applied. The shop held the examination to the shift after welding anyway, because the hold costs nothing when the crew is on another frame.

Step 5 and 6 result. Caps were ground flush on the four radiographed butt joints for interpretation, and the two ultrasonic joints were dressed on the scanning surface only. The examination ran to a written procedure and the interpretation was signed by a Level II under the contractor's written practice.

Step 7 result. Four butt joints radiographed, 100 percent of length. Two corner joints ultrasonically examined, reported at 100 percent of length from one accessible face, which is the honest number and less than a two-face scan would have given. One indication in joint 4, elongated and lying along the fusion line, rejectable under the governing criteria.

Step 8 result. Joint 4 was excavated over the marked length, re-welded, and re-radiographed at 100 percent of the repaired length plus an overlap onto sound metal each side. The re-examination was clean.

What would have happened without steps 2 and 3. Six radiographs, two of them on geometry that produces an uninterpretable image, and four of them imaging a backing bar. The shop would have paid for six shots and been able to act on none of the four backed joints, and the two corner joints would have come back marked uninterpretable rather than accepted or rejected.

How to verify you ran this rather than a nearby version of it

• Every joint has a criterion named with its edition and adopter. In the example that is AWS D1.1 in the edition the contract invoked, not "code."
• Coverage is a printed number, and the honest one. Four butt joints at 100 percent of length; two corner joints at 100 percent of length from one accessible face. The second figure is qualified because a one-face scan is not a two-face scan, and reporting it as plain 100 percent would flatter it.
• Timing is recorded against the rule. No quenched and tempered material, so the 48-hour delayed visual requirement did not apply and is recorded as not applicable rather than silently omitted; the shop's own hold was one shift.
• Pass count after the change is carried through. Backing removal added a back-gouge and back-weld to 4 of 6 joints, so those four are nine-pass joints on the traveller, not the eight the detail started at.
• The repair was re-examined by the method that found it, at the same coverage plus an overlap, not by a different and cheaper method.

References

• AWS D1.1, structural welding code for steel, in the edition invoked by your project specification or the building code your authority having jurisdiction has adopted, for examination requirements, acceptance criteria and delayed visual inspection of certain quenched and tempered steels
• ASNT SNT-TC-1A, or ANSI/ASNT CP-189, in the edition adopted by your employer's written practice, which is the instrument that actually binds technician qualification
• 29 CFR 1910.1096 for ionizing radiation and restricted areas, plus Nuclear Regulatory Commission or Agreement State licensing for the source itself
• 29 CFR 1910.134 for respiratory protection during surface preparation, and 29 CFR 1910.146 where a confined space is involved
• See related: What Each Nondestructive Method Can and Cannot See; Why a Visual Inspection Finds Most of What Matters; Why Grinding and Gouging Are Part of the Weld