Training UT Technicians for Weld Fabrication Acceptance

What fabrication UT is actually looking for

A weld that has never seen service fails in characteristic ways, and none of them are corrosion. Lack of fusion on a bevel face is the classic, and it is the flaw ultrasonics is uniquely good at and radiography is often poor at, because a tight planar lack of fusion parallel to the bevel presents almost no density change to a film but reflects a shear beam strongly when the angle is right. Incomplete penetration at the root produces a sharp, repeatable corner response. Slag inclusions give ragged, multi-peaked responses. Porosity gives scattered low-amplitude responses that move erratically with small probe motions.

Cracking in fabrication has its own signatures and its own causes: transverse cracking in high-strength weld metal, hydrogen-assisted cracking in the heat affected zone hours after welding cooled, crater cracks at stop-starts, and lamellar tearing in the base plate under highly restrained T and corner joints. A technician trained only on artificial notches in a calibration block will find the notch and miss the tear, because a lamellar tear sits in the parent material away from the weld and outside where a routine scan pattern looks.

This is why the scan plan matters more in fabrication than in service work. The examination must cover the weld volume plus the heat affected zone from enough directions and angles that flaws in any expected orientation are struck near normal incidence. That means both sides where access allows, more than one angle where thickness demands it, and a transverse scan for transverse cracking. A scan plan that covers the weld centreline and nothing else is the most common technical deficiency in shop procedures.

Which document supplies the accept-or-reject rule

Three separate documents govern any fabrication UT job and technicians must be able to name all three: the one that supplies the technique, the one that supplies the acceptance criteria, and the one that qualifies the person. For pressure work built to ASME Section VIII Division 1, the technique comes from ASME Section V Article 4, the acceptance comes from Mandatory Appendix 12 or from the alternative flaw-based criteria where ultrasonics is used in lieu of radiography, and personnel qualification comes through Section V Article 1 and the employer's written practice.

For process piping the referencing code is ASME B31.3, and the fluid service category — normal, category D, category M or severe cyclic — determines how much examination is required and how strict the limits are. A technician who applies normal fluid service criteria to a severe cyclic line has produced a report that is wrong even though every measurement in it is correct. For structural steel the governing document is AWS D1.1, which is self-contained: it supplies technique, calibration, the rating calculation and the acceptance tables in one clause.

The document that supplies none of this is ASME Section IX, and the misconception is stubborn. Section IX qualifies welding procedure specifications, procedure qualification records, welders and welding operators. It tells you whether the person and the procedure that made the weld were qualified. It says nothing about how to examine the weld and nothing about whether an indication is acceptable. A shop quality manual that cites Section IX for NDT personnel qualification will be written up in the first serious client audit.

AWS D1.1: the indication rating, step by step

D1.1 ultrasonics is amplitude-based and procedural, and once learned it is fast. The reference level b is established on an IIW type block using the standard side-drilled hole, adjusted so the response sits on a defined horizontal reference line on the screen. In the field, the indication is peaked and the gain adjusted to bring it to that same reference line; the gain reading at that point is the indication level a. The attenuation factor c is calculated from the sound path: take the sound path in inches, subtract one inch, multiply by two decibels.

The indication rating is d = a − b − c. That number is then taken to the acceptance tables, which classify severity into classes as a function of the weld thickness and the depth zone in which the indication lies, with separate and stricter provisions for cyclically loaded tension connections than for statically loaded connections. Class A indications are rejected regardless of length. The lower classes are accepted or rejected depending on indication length, so length determination is part of the acceptance decision rather than supplementary information.

Two mistakes recur. The first is using surface distance or depth in place of sound path when calculating c, or omitting the one-inch subtraction; either shifts the rating by several decibels, which is enough to reclassify an indication in either direction. The second is failing to re-establish the reference level when the probe, wedge or instrument changes, or when the wedge has worn. D1.1 also constrains probe angle selection by material thickness and requires operator qualification including a practical demonstration, both of which shop procedures routinely under-specify.

ASME Section V technique requirements a shop must meet

Section V Article 4 is prescriptive about the physical set-up, and shops that treat it loosely generate findings. The basic calibration block must be of a material acoustically similar to the part and of a thickness related to the part thickness, and it carries side-drilled holes at defined fractions of its thickness plus notches for angle beam calibration. The distance amplitude correction curve is constructed by peaking the response from the same reflector at several depths and joining the marks; the curve must be built from enough points to cover the examination range, and it must be verified at defined intervals during the examination and whenever anything in the set-up changes.

Temperature is a hard requirement, not a nicety. The calibration block temperature must be within a stated tolerance of the examination surface temperature, because sound velocity and wedge behaviour both change with temperature. In a fabrication shop this bites during winter, when a block stored in a heated office is used to calibrate for work on steel that has been sitting on an unheated shop floor, and again on post-weld heat treated components examined before they have fully cooled.

Surface condition is specified too. Article 4 places a limit on the roughness of the scanning surface, and the weld and adjacent base metal must be prepared so the probe maintains coupling and the beam enters predictably. Spatter, undercut ridges and mill scale all cost signal. A technician who scans over spatter and reports a clean weld has not examined it; they have demonstrated that the sound never got in. Training should include measuring what the shop's real surfaces do to the reference response, because arguing about it in the abstract convinces nobody.

DAC versus DGS in shop work

A DAC curve is empirical. You take the same reference reflector at increasing sound paths, peak each response, mark the screen and join the marks. The curve then describes how an identical reflector's response falls off with range in that material with that probe, and any indication is reported as a percentage of DAC at its own range. Its strength is that it is built in the material of interest with the actual probe, so attenuation and surface losses are baked in. Its weakness is that it needs a representative block for each thickness range, which is expensive when a shop makes a wide spread of thicknesses.

DGS, or AVG, replaces the block with theory. Probe-specific diagrams relate distance, gain and equivalent reflector size, so a response can be expressed as the diameter of the disc-shaped reflector that would produce it. One characterised probe covers a range of thicknesses without a family of blocks. The cost is precision in the inputs: the probe must be characterised, and a transfer correction must be measured between the reference block and the actual component to account for differences in surface condition and material attenuation. Skip the transfer correction and DGS silently under-reports.

For training purposes the important lesson is not which is better but that neither is a licence to convert a response into a flaw size and stop. A percentage of DAC and an equivalent reflector diameter are both comparisons to an idealised reflector. A real lack of fusion is planar, rough and oriented, and its response depends on how the beam strikes it. Amplitude-based methods are calibrated conservatism, not measurement. Where actual flaw height matters — as it does under flaw-based acceptance criteria — the shop needs tip diffraction sizing, TOFD or phased array, and technicians trained specifically in those.

Couplant, surface condition and temperature: the mundane rejections

ASME Section V restricts couplant contaminants on specific materials, and fabrication shops working mixed metallurgy get caught by it. Couplant used on nickel base alloys is limited in residual sulfur, and couplant used on austenitic stainless steel and titanium is limited in halide content, both to low parts-per-million levels. The reasons are metallurgical rather than procedural: sulfur promotes embrittlement in nickel alloys and halides drive chloride stress corrosion cracking in austenitic stainless. A shop that keeps one drum of general-purpose couplant and uses it on everything is creating a materials problem, not just a paperwork one.

Couplant choice also affects the examination itself. Viscosity has to suit the surface and the orientation — thin couplant runs off a vertical seam and the signal drops mid-scan, thick couplant introduces a variable delay path and skews a thickness reading. Whatever the shop uses for calibration must be the same as what it uses for examination, because the transfer characteristics differ and a calibration performed with one and an examination with another is not a valid pairing.

Temperature deserves the same discipline. Cold steel in an unheated bay, hot components after post-weld heat treatment, and wedges that soften and change refracted angle at elevated temperature all shift results. The written procedure should state the permitted temperature range, the correction or the equipment required outside it, and the verification frequency. In practice these mundane controls generate more rejected reports and repeat examinations in fabrication than any failure to interpret a signal correctly.

What a technician arriving from in-service inspection gets wrong

The first thing is scan discipline. An in-service technician is used to a grid: named locations, fixed points, repeat readings compared to history. Fabrication has no history and no grid. The weld is new, the geometry is whatever the drawing said, and coverage must be established from first principles — full skip and half skip distances calculated from the refracted angle and thickness, index point verified on the block, and a scan pattern that puts the beam through the entire weld volume and heat affected zone from the required directions.

The second is what counts as a finding. In-service work is dominated by wall loss and by service-induced damage in predictable locations, so the mental model is 'how much is left'. Fabrication work is dominated by planar flaws whose detectability depends entirely on orientation relative to the beam. A technician who scans one side of a single-vee weld at one angle and reports it clean has satisfied a habit, not a procedure. Lack of fusion on the far bevel face may be invisible from that side and obvious from the other.

The third is the acceptance document. Technicians arriving from refinery work often reach instinctively for API-derived thinking, and the shop is not governed by it. The weld in front of them is accepted or rejected against ASME Section VIII Division 1, ASME B31.3 for the applicable fluid service, or AWS D1.1 against the class tables. Retraining that reflex is a specific training objective, not something that resolves on its own, and it is best done on the shop's own drawings, procedures and rejected specimens rather than in the abstract.

Building the training around the codes your shop actually works to

Generic ultrasonic training produces technicians who understand the physics and stall at the acceptance decision. The training that works in fabrication is built backwards from the shop's contracts: which codes appear in the purchase orders, which fluid service categories the piping work covers, whether structural work is statically or cyclically loaded, whether any client requires ultrasonics in lieu of radiography with flaw-based acceptance, and what materials are in the shop. Those answers determine the specimen set, the calibration blocks, the scan plans and the content of the specific examination.

On-site delivery is usually the right format for a fabrication shop for exactly that reason. Training on the shop's own instruments, wedges, blocks, couplant and written procedures means the specific examination — the one SNT-TC-1A requires to cover the employer's equipment, procedures and acceptance criteria — is written against reality. It also lets the training use the shop's own rejected and repaired specimens, which are more instructive than any commercial flaw set because the technicians recognise the joints and the welders who made them.

Atlantis delivers ultrasonic testing training at Levels I, II and III, along with RT, MT, PT, ET, VT, phased array and TOFD, in classroom, on-site corporate and blended formats, prepared to ASNT SNT-TC-1A and ISO 9712. We also provide ASNT Level III consulting for shops that need a written practice built or brought up to the edition their clients require, examinations prepared and graded, procedures written against Section V and D1.1, or a Level III of record. Contact info@atlantisndt.com for a consultation or a quote.

Request a consultation