18.1 Evaluation and Reporting

The ASNT NDT Level II visual general exam lists Evaluation and Reporting as official VT topic 8. Those items are not asking whether you can see a weld bead. They are asking whether you can measure, locate, and describe what you see, tell relevant geometry from a discontinuity, know when visual testing has reached its limit, and leave an auditable record that a Level III or a customer can reconstruct without standing next to you.

Visual testing (VT) reports conditions that are open to the viewed surface — or that change the surface contour enough to be seen under the procedure's lighting and viewing rules. An indication is what you observe. A discontinuity is an interruption in the typical structure or contour of the material or weld. A defect is a discontinuity that the named acceptance standard rejects. Evaluation is the work of moving from the first word to the third without guessing depth you did not measure and without calling every shadow a crack.

Measure, locate, and describe

A Level II who writes "undercut noted" has not evaluated anything. The method's product is a description that another qualified person can take to the same weld and find the same condition. Three verbs sit on every serious VT report.

Depth if measurable is the phrase that separates honest VT from fiction. A fillet-weld gauge, a V-WAC or similar undercut gauge, a high-low / mismatch gauge, and a reinforcement / cam gauge measure contour. They do not measure how far a tight crack runs into the throat. If the acceptance table rejects on undercut depth and the gauge fits the groove, you measure depth and apply the table. If the acceptance table rejects on crack depth and all you have is a surface line, you do not assign a millimetre depth from appearance.

Orientation is not decoration. A longitudinal crack-like line in weld metal is a different evaluation problem from a transverse line that crosses the toe into the HAZ. Branching or a sharp change of direction is part of the description because it is how a crack distinguishes itself from a straight tool mark or a weld ripple crest.

Location without a datum is not a location. "Undercut on the downstream toe, 40 mm starting 15 mm from the upstream edge of stamp W12, clockwise from the 12 o'clock flow mark" is a location. "Undercut on that weld we did Tuesday" is a conversation.

Relevant, nonrelevant, and false — VT versions

The same three classes used in magnetic particle and penetrant evaluation apply here. Collapsing them into "real crack" versus "junk" misses items.

Relevant is not rejectable. Undercut is a relevant geometric condition even when the measured depth is inside the table. You still measure it. You do not ignore it because "codes allow some undercut." You also do not reject every relevant indication on sight. The table decides defect versus acceptable discontinuity after you have described the condition in the table's units.

Weld ripple versus crack

Weld ripple (the regular, repeating solidification lines of a stringer or weave) is typical weld contour. It follows the travel and oscillation. Crests are blunt, spacing is roughly even, and the lines stay inside the weld face. That pattern is nonrelevant geometry unless a specification demands a blended or ground finish and the ripple violates that finish requirement.

A crack that a Level II is expected to catch visually is a sharp, often irregular line. It may start in a crater, at a toe, or in a stop-start. It does not keep the ripple spacing. The teaching contrast the outline wants is this: a line that leaves the ripple, crosses the toe, and continues into the HAZ is not more contour. It is a relevant crack-like indication until proven otherwise. Do not grind it "to see if it goes away" unless the procedure authorizes metal removal as part of evaluation, and never grind away the only evidence before it is recorded.

A crater crack sits in the termination puddle and is often star-shaped or short and jagged. It is relevant. A neatly filled crater with only the same ripple as the rest of the bead is contour.

Mill scale versus pitting

Mill scale is the oxide layer from hot rolling or high-temperature exposure. Intact scale can look dark, flaky, or mapped. Loose flakes, color change alone, and scale that the specification allows to remain on an as-rolled surface are not automatically pits. Pitting is metal loss: a cavity with a measurable mouth, a remnant wall, or a cluster of depressions in the base metal or weld. You distinguish them by contour and metal loss, not by how brown the surface is. A rust bloom that wipes off and leaves the original surface is stain (often false or nonrelevant discoloration). A cavity that remains after cleaning is pitting and is relevant. Depth of a pit is recorded only if a gauge or optical method the procedure allows can measure it; otherwise you record mouth size, location, and that depth was not visually measurable.

Related pairs use the same logic. A grinding mark is a broad, directional scratch with a visible lay; a crack is sharp and does not follow the wheel. Spatter is attached droplets; porosity is a cavity in the metal. Heat tint is color; underfill is missing weld metal below the plate surface. Learn the pair, then measure the one that is a discontinuity.

When to stop and request another method

VT is a surface method. It is excellent at contour and at open, visible discontinuities. It is the wrong last word when the acceptance question is something the eye cannot answer. Stop the visual accept/reject — not the job, unless safety requires it — and request the complementary method the procedure or the ITP already names, or elevate to the Level III if the document is silent.

Stop and call another method when:

• A linear indication might be a tight crack and the procedure names MT or PT for confirmation of surface breaking.
• The table rejects on depth (crack depth, remaining wall, or an undercut the gauge cannot enter) and VT cannot measure that depth.
• The concern is subsurface — incomplete penetration not open to the viewed face, lamellar tearing, or a buried lack of fusion — and the contract already requires UT, RT, or another volumetric method at that joint.
• Surface condition blocks VT (coating, heavy scale that this hold point does not allow you to remove, or a geometry the eye cannot resolve) and remote optics still cannot meet the procedure's resolution.
• Remote VT in a tube or vessel shows a feature smaller than the system can reliably measure, and the procedure requires a measurable record.

What you do not do is invent a PT or MT exam because you are curious, skip a required VT hold point because "UT will catch it later," or accept a crack-like line because "it is only visual." Complementary methods answer the question VT cannot. They do not erase the duty to describe what VT already saw.

Report forms, sketches, and photographs

The report is the product. A traveler check-box is not a substitute unless the written practice and the procedure explicitly treat that box as the complete record for a no-indication exam. When relevant indications exist, or when the procedure requires a narrative even for accept, the record typically carries:

• Item identity: job, drawing, weld or part number, heat or serial if used
• Procedure identity and revision, plus the named acceptance standard
• Technique: direct or remote, magnification if used, lighting arrangement if the procedure records it
• Measuring aids actually used (fillet gauge, undercut gauge, scale)
• Date, inspector identification, and certification level the written practice requires on the form
• For each relevant indication: type, length, depth or height if measured, orientation, location from a datum, surface, photo or sketch reference, and disposition (accept, reject, hold, or "evaluate by [method]")

A sketch needs a datum, a weld cross-section or plan view that matches the joint, and dimensions. A photograph needs identification in the frame or in the file name the procedure specifies, a scale or a known feature, and lighting that does not hide the contour you claim to have measured. A beautiful photo of an unlocated weld is not a report. A report that says "see photo" with no measurements is not an evaluation.

If you cleaned, ground under authorization, or changed lighting to dismiss a false indication, record that action. Tomorrow's auditor should not have to guess why the first photo showed a black line and the second did not.

Hold points

A hold point is a planned stop in fabrication or in-service work where VT (or another examination) must be completed and signed before the next operation hides the surface. Typical visual hold points include fit-up (alignment, root opening, cleanliness), root pass, after back-gouge, final visual of the completed weld, after postweld heat treatment when the procedure wants a re-look at contour, and in-service or shutdown VT of the same geometry.

You sign a hold point only after the examination the procedure describes is finished and the record exists. You do not sign ahead so welding can continue. You do not sign a hold point you could not access. You do not waive a hold point because the weld "looks fine from the platform." If the surface will be covered — insulation, coating, the next weld pass, a fitting — the hold point is there because later VT cannot reconstruct what you were supposed to see.

A reject or a hold at a hold point stops that operation until disposition. The Level II's job is to document the condition and the stop. The Level III, engineering, or the customer, as the written practice names them, owns a repair or use-as-is decision that changes acceptance.

Realistic exam scenarios

A stem shows a regular weave on the face and a sharp line leaving the toe into the HAZ: classify the weave as contour and the line as relevant; measure and locate the line. A stem gives length and a photo but asks for depth of a hairline crack: the correct move is that depth is not measurable by VT. A stem offers a rust-colored plate with flakes that wipe off versus a cavity that remains after cleaning: stain versus pitting. A stem asks what belongs on the report when the procedure requires a record: type, measurable size, orientation, datum location, disposition, and the named sketch or photo — not a lone accept check. A stem places you at a root-pass hold point you cannot see: you do not sign; you get access or you elevate the missed hold point.