Choosing Between Radiography and Ultrasonics on a Weld

Start With the Defect You Expect, Not the Kit in the Van

Every argument about RT versus UT collapses once you name the flaw. A root pass with a cold lap, a bevel face that never fused, a hydrogen crack running down a heat-affected zone — those are planar. They have almost no volume to remove density from a radiographic image, and they sit at angles that put a flat reflecting face square into an angled sound beam. Porosity, slag inclusions and shrinkage cavities are the opposite: real volume, low aspect ratio, no preferred orientation. Decide which family the process is likely to produce and the method chooses itself.

Process history tells you which family to expect. Submerged arc on heavy plate with a tight bevel generates lack of sidewall fusion. Manual SMAW out of position with damp consumables generates porosity and slag. Pipe welded with a hot-pass problem generates lack of penetration at the root. A fabricator running the same procedure for two years already knows their failure mode from the repair log. Pull that log before writing the NDE plan rather than after the first rejection, and the method selection stops being a preference argument.

This is the step most NDE specifications skip, and it is the one that decides whether a programme pays for itself. Writing the method choice into the procedure with the reasoning attached is exactly what a written practice and NDT procedure development engagement produces. Where a plant has inherited a spec that calls radiography on everything out of habit, an NDT programme audit and gap assessment tends to find the same avoidable cost sitting in the same place: RT ordered against a failure mode RT does not see well.

Orientation Is the Whole Argument

Radiography is a transmission method. It records how much energy survives the path through the part, so it responds to the thickness of material removed along that path. A spherical pore 1 mm across removes 1 mm of steel from the path and shows up. A lack-of-fusion plane 0.05 mm wide removes 0.05 mm from the path when it lies edge-on to the beam and disappears into film grain. Rotate that same plane so its width lies along the beam and it becomes visible — which is why radiographers chase source angles on suspected planar flaws.

Ultrasonics is a reflection method. It responds to acoustic impedance mismatch across a surface and to how much of the beam that surface throws back toward the receiver. A flat face perpendicular to the beam returns nearly everything. A face tilted away returns a fraction. A spherical pore scatters in every direction and returns a modest signal from any angle. So the two methods have opposite blind spots, and the blind spots are geometric rather than a matter of equipment quality.

The practical consequence is that neither method is a superset of the other, and specifications that treat one as a straight substitute produce escapes. Where the same joint carries both risks — a heavy-wall nozzle with fusion risk on the bevel and porosity risk in the cap — the defensible answer is UT for the volume and a surface method for the cap, not a single volumetric shot chosen on price. Getting that combination approved is routine work for an outsourced ASNT Level III.

Thickness and Geometry Set Hard Boundaries

AWS D1.1 draws the line explicitly. Clause 8.19.1 applies conventional ultrasonic testing to groove welds and heat-affected zones from 5/16 in. (8 mm) through 8 in. (200 mm) inclusive, and excludes tube-to-tube T-, Y- and K-connections from that procedure. Below 5/16 in. the near-field and the geometry of the entry surface stop conventional angle-beam work from resolving anything useful, and radiography takes the joint. Above 8 in. the code sends you to a qualified alternative rather than the standard clause.

ASME sets a different floor for a different purpose. Code Case 2235-13 permits ultrasonic examination in lieu of radiography on welds in material 1/2 in. (13 mm) or greater in thickness. Below that thickness there is no in-lieu-of route under that Case at all, so a thin-wall vessel weld that the drawing calls for RT stays an RT weld. The same Case fixes the examination volume: the weld plus the lesser of 1 in. (25 mm) or t on each side for material 8 in. (200 mm) or under.

Geometry then removes options that thickness left open. Nozzle-to-shell welds, set-through configurations, dissimilar-metal joints with a clad interface and small-bore socket welds all degrade ultrasonic coverage in ways a scan plan has to demonstrate rather than assert. Radiography degrades differently, through geometric unsharpness and source-to-film distance, which is why some awkward joints end up with both methods applied to different parts of the same weld. Neither problem is solved by buying a better instrument.

Radiation Safety Is a Schedule Cost, Not a Footnote

The regulatory numbers are specific. 10 CFR 20.1003 defines a radiation area as one where an individual could receive more than 0.005 rem in 1 hour at 30 cm, and a high radiation area as more than 0.1 rem in 1 hour at 30 cm. 10 CFR 20.1301 limits the dose in any unrestricted area from licensed operations to 0.002 rem in any one hour and 0.1 rem in a year to a member of the public. The radiographer surveys, ropes and posts the boundary against those figures before the source leaves the camera.

Personnel requirements are equally concrete. Under 10 CFR 34.43(a) no individual acts as a radiographer until they have received training in the listed subjects, completed a minimum of two months of on-the-job training, and been certified through a radiographer certification program by a recognised certifying entity — the NRC accepts ASNT's IRRSP for isotope radiography. Refresher safety training runs at intervals not exceeding 12 months, and a supervisor observes each radiographer during actual operations at six-month intervals.

None of this applies to ultrasonics. There is no exclusion boundary, no source transport, no radiation work permit, no adjacent trades stood down, and no night-shift premium bought simply to empty the area. On a live plant that difference is frequently larger than the whole inspection budget. Contractors tracking who holds which radiation credential against which job usually end up handling it inside NDT personnel certification tracking rather than a spreadsheet that expires quietly.

The Record: Film, Digital Images and Encoded Data

The strongest genuine argument for radiography is the artefact it leaves. A radiograph is a record a second party reopens in five years without the original technician, the original instrument or the original calibration block. Digital radiography and computed radiography keep that property and add archiving and image processing. In a dispute — a failed hydrotest, an insurance claim, a fitness-for-service argument — an image everyone can look at settles questions that a written interpretation reopens.

Manual ultrasonics has the opposite property. What survives is a report describing indications one person saw on a screen that no longer exists. That asymmetry is real, and it is why clients who have been burned once insist on radiography long after the technical case has moved on. It is also the asymmetry that encoded ultrasonics erases. Code Case 2235-13(e) requires the complete data set to be recorded in unprocessed form, with no gating, filtering or thresholding of the response from the examination volume.

So the honest comparison is three-way, not two-way: radiography, manual UT, and encoded PAUT or TOFD. Manual UT loses the records argument outright. Encoded ultrasonics matches it and adds through-wall depth information that a radiograph never carried. Where a corrosion or fitness-for-service case will later be built on the data, that depth record is worth more than the image — which is why fitness-for-service assessment under API 579 starts from encoded ultrasonic data rather than film.

What ASME Permits When You Swap UT for RT

Section VIII Division 1 authorises the swap in UW-51(a)(4), which requires the examination to meet Section VIII Division 2 paragraph 7.5.5 and points to Section V Article 4 Mandatory Appendix VIII for fracture-mechanics-based acceptance, used together with Mandatory Appendix IX. Section I permits the substitution through Code Case 2235, Code Case 2816, or PW-52.1 — the two Code Cases carrying fracture-mechanics acceptance criteria while PW-52.1 routes to Section V Article 4 Mandatory Appendix VII and workmanship-based criteria. Those are different acceptance philosophies reaching different verdicts on the same indication.

The technique restriction is the part that surprises people. Across Sections I, VIII Division 1 and VIII Division 2, the permitted examinations are limited to TOFD and phased array with computer-based data acquisition and analysis, using equipment mechanically mounted and guided on the examination surface by automatic or semi-automatic means. Manual straight-beam UT, manual angle-beam UT and manual phased array are not permitted as substitute techniques when UT replaces required RT.

Code Case 2235-13 spells out the rest of the burden. Paragraph (b) requires a documented scan plan showing transducer placement, movement and coverage together with beam angles, beam directions relative to weld centreline and the volume examined. Paragraph (c) requires a written procedure conforming to Section V Article 4, demonstrated on qualification blocks containing at least three flaws — one at the OD surface, one at the ID surface and one subsurface. Paragraph (f) restricts data analysis and interpretation to Level II or III personnel certified under their employer's written practice.

What AWS D1.1 Permits on Structural Steel

AWS D1.1 keeps radiography and ultrasonics in separate parts of the same clause. Clause 8 Part E covers Radiographic Testing and Clause 8 Part F covers Ultrasonic Testing of Groove Welds, with 8.15 governing extent of testing. Acceptance sits in Table 8.2 for statically loaded nontubular connections and cyclically loaded connections in compression, and Table 8.3 for cyclically loaded nontubular connections in tension — the familiar amplitude-and-length method, with the tension table markedly tighter.

The 2020 edition added a third route, retained in D1.1/D1.1M:2025, the 25th edition: Annex H (Normative) — Phased Array Ultrasonic Testing (PAUT). Annex H carries Table H.1 for PAUT essential variables, Table H.2 for PAUT acceptance criteria and Table H.3 for discontinuity classification, and it works from encoded linear scanning rather than hand rastering. Annex G governs qualification and calibration of UT units with approved reference blocks, and Annex N remains informative guidance on UT of welds by alternative techniques.

For a structural fabricator the practical reading is that PAUT is no longer an engineering-approved deviation under 8.8 but a named normative annex the code owns. That changes the conversation with an engineer of record from a request for concession to a citation. Fabricators putting crews through that transition normally run it as on-site corporate NDT training tied to their own mockups rather than a generic classroom course.

Where the Cost Actually Lands

On a per-weld basis, radiography on thin wall is usually the cheaper shot. A crew, a camera, a set of films, an exposure calculation and a reading. That number is the one that appears on quotes and the one that keeps RT on specifications that should have moved years ago, because it is the only cost anyone measures.

The costs that swamp it are schedule costs. Clearing a boundary means adjacent trades stop. Working the shot at night means a shift premium on the radiography crew and on everyone who has to stay. Source transport, security, dosimetry and area surveys carry administrative overhead that no per-weld figure captures. On a turnaround where the critical path runs through a weld, the difference between examining it with the unit live and examining it with a floor evacuated is measured in production, not in inspection.

Ultrasonics inverts the profile. Higher equipment and qualified-time cost per weld, near-zero interference cost. It also front-loads spending: procedure qualification, demonstration blocks, encoders and scanners, and technicians trained on the specific instrument. That front-loading is precisely what makes the decision an owner decision rather than an inspection-contractor decision. Owners weighing it against inspection interval and risk generally take it into risk-based inspection programme design rather than settling it weld by weld.

Where PAUT and TOFD Have Already Displaced RT

Heavy-wall pressure equipment moved first, because that is where the fracture-mechanics acceptance criteria pay. Radiographic acceptance rejects on indication length against a workmanship standard. The Code Case route accepts flaws against Tables 1, 2 and 3 by thickness band, using measured height and through-wall position — significantly more tolerant of a flaw that a fracture-mechanics analysis shows is harmless. Fewer unnecessary repairs on a thick weld is a large number, and repairs on thick welds carry their own risk of introducing worse defects.

Pipeline girth welds moved next, for the same reason plus mechanised zonal discrimination that reports which weld zone each indication sits in. Storage tank shell welds and structural steel followed as encoded PAUT acquired its own code home. In every one of those cases, the displacement happened when a code gave the technique a citation, not when the technology first worked — the equipment predated the acceptance by roughly a decade.

What has not displaced is thin-wall work, small-bore piping, and any joint where the client specification names radiography without an alternative clause. The full decision path — what ASME permits, what a client accepts, and what the qualification burden actually costs — is worked through in the companion piece on when phased array replaces radiography. Vessel and tank owners applying it in service usually pair it with API 510 pressure vessel inspector services.

Writing the Choice Into a Procedure That Survives Review

A method selection that lives in an inspector's head fails the first audit. Put it in the procedure: the construction code, the clause authorising the method, the expected defect family, the thickness range, the acceptance table and the personnel level allowed to interpret. Code Case 2235-13 models the format well — every requirement is a numbered paragraph with a demonstrable output, from the scan plan in (b) to the unprocessed data record in (e) to the Level II-or-III interpretation restriction in (f).

Personnel qualification is where these procedures most often fall over. The Code Case requires personnel to be qualified and certified in accordance with their employer's written practice, with ASNT SNT-TC-1A or CP-189 used as a guideline, and requires that the people who acquire and analyse data be trained on the specific equipment and take part in the procedure demonstration. A technician certified for conventional UT is not automatically qualified to interpret encoded phased array data, and an auditor will ask for the demonstration record.

If your organisation has no Level III to own that document, the gap is structural rather than administrative. Bringing in an interim NDT Level III closes it, and method-specific ASNT training delivered in the United States builds the bench underneath. Either way, decide the method against the defect, cite the clause, and stop letting habit choose between radiography and ultrasonics.

Request a consultation