How a validator reviews film and digital radiographic interpretation

Readability comes before interpretation

An interpretation dispute is usually an argument about a shape on an image. The prior question — the one that decides whether the argument is worth having at all — is whether that image was capable of showing the flaw in the first place. A radiograph that is too light, too dark, too unsharp, or shot with the wrong image quality indicator cannot support any interpretation, favourable or unfavourable, and no amount of expertise applied to reading it changes that.

This ordering is convenient for a reviewer, because readability is largely a matter of measurement rather than judgement. Density is read with a calibrated densitometer. IQI visibility is a yes or no. Geometric unsharpness is arithmetic performed on the technique sheet. Coverage is checked against markers imaged from the part. These checks can be repeated by anyone and produce the same answer, which is exactly the property a finding needs when it is going in front of a third party.

Only after the images pass do the calls come under scrutiny. Running the review in that order also prevents a common unfairness: criticising an interpreter for missing something on a film that was never capable of showing it. The finding in that case belongs to the technique and the procedure, and the interpreter is not at fault. Getting the attribution right is most of what makes a review usable.

Density: the gate that decides whether the film can be read

ASME Section V, Article 2 sets transmitted film density limits through the body of the appropriate hole IQI, or adjacent to the designated wire of a wire IQI, and through the area of interest. For single-film viewing the minimum is 1.8 for radiographs made with an X-ray source and 2.0 for those made with a gamma source, and the maximum is 4.0. For composite viewing of multiple film exposures, each film of the set must reach at least 1.3, under the same 4.0 ceiling.

Two failures recur. The first is a film shot at composite densities and then read singly, which happens when the second film of a pair is lost, never printed, or quietly dropped from the package sent to the client. The second is a densitometer with no current verification record behind the readings, which turns the whole density check into an assertion. A reviewer asks for the densitometer verification alongside the film, because a density number without it proves nothing at all.

Article 2 also limits variation, not just level. Density anywhere through the area of interest must not vary by more than minus 15% or plus 30% from the density through the body of the designated hole-type IQI adjacent to the essential hole, or adjacent to the essential wire. Where the requirement is not met, an additional IQI is required for each exceptional area and the radiograph is retaken. Reviewers therefore read the extremes of the area of interest, not the comfortable middle of the film.

IQI selection, placement and the sensitivity actually demonstrated

The IQI is the only thing on a radiograph that proves the technique worked. Article 2 selects the designated hole-type or wire-type IQI from Table T-276 against the nominal single-wall thickness, and permits a thinner or thicker hole-type IQI to be substituted for any listed section thickness provided an equivalent IQI sensitivity is maintained. Getting the thickness basis wrong, including how weld reinforcement is treated, puts the wrong IQI on the film and leaves the demonstrated sensitivity unknown.

Visibility is the check that gets skipped. The requirement is not that the IQI is present. It is that the image of the designated hole IQI and the 2T hole are visible, or the essential wire of a wire-type IQI. A plaque that shows as a rectangle with no resolvable essential hole is a failed technique, not a marginal one. Reviewers examine the IQI under magnification and record precisely what is and is not resolvable, because this is the finding most often disputed and most easily proven either way.

One frequently misquoted point deserves stating plainly: the designated IQI in Table T-276 does not necessarily deliver a 2-2T sensitivity level, and the code does not claim that it does. Reports and procedures reciting 2-2T sensitivity per ASME V as a blanket requirement are repeating a shop convention rather than the code. Correcting that is routine NDT procedure development work, and it removes a recurring source of argument between contractor and client.

Geometric unsharpness and the geometry of the shot

Geometric unsharpness is the penumbra a finite focal spot casts at the edge of a discontinuity: the source size multiplied by the distance from the source side of the object to the film, divided by the source-to-object distance. Small, tight flaws blur first, and they blur out of existence rather than becoming faint. Article 2 caps unsharpness at 0.020 in. for material under 2 in., 0.030 in. from 2 to 3 in., 0.040 in. from 3 to 4 in., and 0.070 in. above 4 in.

The reviewer recomputes it from the technique sheet rather than accepting the value printed there, because the inputs are where the errors sit: an assumed focal spot size that does not match the tube or the source capsule actually used, or a source-to-object distance shortened in the field to save exposure time on a night shift. Neither is visible on the film. Both become visible the moment the arithmetic is redone against the equipment records and the source certificate.

Unsharpness is also the most common technical reason a fine crack is genuinely absent from a radiograph that otherwise looks perfectly acceptable. That is a real limitation of the method rather than a failure by any individual, and a review that says so protects the interpreter while still telling the owner what the radiograph could and could not have shown. The same limitation is one reason ultrasonic methods are specified on thick sections, as covered in our comparison of RT and UT for weld inspection.

Coverage, overlap and the markers that prove it

Coverage on a radiographed weld is proven by location markers, and Article 2 is specific about where they go. Markers appear as radiographic images on the film and are placed on the part, not on the exposure holder or cassette. Their locations are permanently marked on the surface of the part being radiographed where that is permitted, or recorded on a map, in a manner that allows the area of interest on a radiograph to be accurately traceable to its location on the part.

That requirement is what lets a reviewer reconstruct the shot layout from the films alone. Marker sequences that jump, repeat or restart expose gaps between exposures, unrecorded retakes, and films shot in a different order from the one the report describes. For double-wall viewing, at least one location marker is required adjacent to the weld or on the material in the area of interest for each radiograph, which gives the reviewer an anchor on every single image in the package.

Missing markers rarely mean the weld was not shot. They mean nobody can now prove which part of it each film covers, and a retake cannot be located on the joint. That distinction matters enormously when the finding goes to a client: coverage unproven is accurate and defensible, while coverage inadequate is a claim the reviewer usually cannot support and should not make. Precision in the wording is part of the deliverable.

Artefact or indication: telling processing from metallurgy

Film artefacts have signatures. Crescent-shaped pressure marks from rough handling, tree-static from dry-weather stripping, streaks from exhausted or contaminated processing chemistry, screen scratches that repeat in the same position across several films from one cassette, and water spots all behave like defects to a hurried eye, and none of them respects the geometry of the weld. Position relative to the joint is the first discriminator a reviewer applies.

The second discriminator is behaviour. A real discontinuity has a density gradient consistent with a volumetric or planar change in path length through the metal, an orientation that makes sense for the process that made the weld, and continuity across films where exposures overlap. An artefact often crosses the film edge, sits over base metal and weld indifferently, or appears on one emulsion side only — which shows immediately when the original film is examined under oblique light rather than on a viewer.

Digital images add artefacts film never had: dead and stuck detector pixels, imaging plate ghosting from an incompletely erased previous exposure, and structured noise from scatter. These repeat in fixed positions relative to the detector rather than the part, which is exactly how a reviewer identifies them, by comparing images taken with the same detector across different exposures. A defect that never moves when the part does is not in the part.

Does the interpretation match what the image shows?

Only after the qualification gates are settled does the reviewer address the calls themselves, and the question is narrower than it looks. It is not whether the reviewer would have called it the same way. It is whether the report's characterisation is supported by the image, and whether the acceptance criterion was applied as written. Competent interpreters disagree at the margins, and a review treating every marginal difference as an error is not useful to anybody.

Three specific mismatches recur. Characterisation errors, where a rounded indication is recorded as slag or a linear indication as porosity, which changes which acceptance rule applies. Measurement errors, where length or cumulative length across a specified span was estimated rather than measured, or measured on a print at unknown magnification. And criterion errors, where a correct observation is scored against the wrong table, the wrong thickness band, or a superseded edition of the referencing construction code.

Interpreter qualification sits behind all three, and reviewers check it against the employer's written practice rather than a certificate alone — the arrangement described in our note on ASNT SNT-TC-1A. Level II interpretation limits, near-vision and colour-contrast examination currency, and the annual employer evaluation are ordinary records that are ordinarily missing, and their absence is a real finding even when the interpretation itself turns out to be correct.

Computed radiography and digital detector arrays

Digital radiography does not remove the qualification gates; it substitutes different measurements for them. ASME Section V, Article 2 handles the two families separately: Mandatory Appendix VIII covers radiography using phosphor imaging plates, and Mandatory Appendix IX covers digital detector systems, each with requirements for the written procedure, system qualification, image quality indicators, image evaluation and documentation. A report citing digital RT per ASME V without naming which appendix has told the reviewer very little.

Density is replaced by signal-to-noise ratio as the measure of whether the image carries usable information, and spatial resolution takes on a role film density never had. Basic spatial resolution is demonstrated with a duplex wire IQI, whose paired wires are read from a line profile rather than judged by eye. A digital report without signal-to-noise and spatial resolution evidence is the exact equivalent of a film report with no densitometer readings behind it.

Detector condition is the other new axis. Bad pixel maps drift, imaging plates degrade with use and handling, and a plate that was never fully erased carries a ghost of the previous exposure into the next one. The system qualification requirements in the appendices exist precisely because the detector is now part of the technique, and its performance record is part of the evidence a reviewer asks for alongside the images themselves.

Processing that cannot be undone, and the conditions the film was read in

Film viewing conditions used to be the whole argument: illuminator output adequate for the density on the film, ambient light controlled, masking used so stray light does not wash out a dense area. Those still apply to film, and a review notes whether a high-density radiograph was read on an illuminator capable of it. A film at density 3.8 read on a weak viewer is functionally unread, whatever the interpretation report says about it.

Digital viewing moved the problem into software and made part of it permanent. Windowing, contrast and brightness applied at viewing are reversible and harmless. Filters, edge enhancement, noise reduction and re-sampling applied before the image is archived are not: they discard information, and the archived file cannot be returned to what the detector originally recorded. If only the processed image survives, the reviewer can evaluate what is there but cannot recover what was removed from it.

So the retention question for digital RT mirrors the one for phased array. The archive that matters is the original unprocessed image data with its acquisition parameters, not an exported and enhanced viewing copy. Owners who commission independent report validation years after the fact discover this constraint at the worst possible moment, and it is written into examination contracts far less often than it should be.

When the review upholds the original interpretation

Radiographic interpretation is more reproducible than the volume of disputes suggests. Competent interpreters looking at a qualified radiograph agree most of the time, and where they disagree it is usually about a marginal call sitting near an acceptance limit, or about which criterion applied to it. Reviews therefore uphold the original interpretation often, and that outcome should be stated as plainly and as confidently as a defect would be.

The credible finding in many RT disputes is not that the interpreter was wrong, but that the radiograph was never qualified, so nobody's interpretation of it is reliable — a different conclusion with a different remedy, usually a retake rather than a repair. Where a decision, a claim or an acceptance turns on a radiographic record you cannot re-shoot, send the films and the technique records for review and get a defensible answer in either direction.

Request a consultation