How Ultrasonic Waves Detect Flaws in Steel

Ultrasonic testing (UT) can detect flaws in steel because a crack, lamination or lack of fusion is a steel-to-air boundary that reflects about 99.996% of the sound energy striking it head-on. A straight 0° beam finds flaws lying parallel to the plate surface, and angle beams of 45° to 70° find the weld flaws a 0° beam misses.

That one rule, that a flaw is found when the beam meets it close to square, explains most of what UT can and can't do in structural steel. This guide applies it to the steel products a fabricator, erector or owner actually buys and inspects: plate, welds, forgings and castings. It covers laminations and the straight-beam lamination scan, the main weld discontinuities, hydrogen cracking, lamellar tearing, and the grain and attenuation limits that make coarse-grained steels and austenitic welds hard to test. Two worked examples show how an inspector turns an echo into a depth and a position.

The wave physics itself (velocity, wavelength, Snell's law, near field) lives in our pillar on the science behind ultrasonic testing, and the technique families (pulse-echo, PAUT, TOFD, immersion) are compared in types of ultrasonic testing in NDT. This page stays with steel and what goes wrong inside it.

How do ultrasonic waves detect flaws in steel?

Ultrasonic waves detect flaws in steel by reflection: a pulse travels through the metal at a known velocity, part of it bounces back from any boundary where the acoustic impedance changes, and the time of that echo gives the reflector's distance. A flaw detector shows the echoes on an A-scan, a plot of amplitude against time.

The size of the reflection depends on what sits on the other side of the boundary. Steel has an acoustic impedance of about 46 MRayl and air about 0.0004 MRayl, so an open crack, a lamination or an unfused sidewall returns almost everything that hits it squarely. A slag inclusion or a tungsten particle reflects less, because the impedance step is smaller. A tightly closed crack under compressive stress can pass much of the sound straight through it, which is why "tight cracks give the clearest echo" is wrong: only open crack faces are near-perfect reflectors.

Orientation matters as much as impedance. A smooth planar flaw acts like a mirror. If the beam strikes it square, the echo comes back to the probe; if the beam strikes it at a glancing angle, the echo leaves in another direction and the probe receives little or nothing. That is why inspectors choose the beam angle from the flaw they expect, not from the wave type. A 0° longitudinal beam is right for a lamination lying parallel to the rolled surface. A 45°, 60° or 70° shear beam is right for a fusion face or a crack running through the weld thickness.

An A-scan doesn't show a flaw's size or shape. The screen position gives the sound path; depth and surface position come from that path and the beam angle, and size comes from scanning the probe and noting where the echo drops (for example the 6 dB drop method). Amplitude is affected by orientation, roughness and distance as well as size, so a code rates amplitude against a reference reflector rather than treating it as a flaw dimension.

Key takeaway: UT finds a flaw in steel when the beam meets a strong reflector close to square; the angle is chosen for the flaw orientation you expect.

What flaws can UT find in steel plate, welds, forgings and castings?

UT finds internal planar flaws best (laminations, cracks, lack of fusion and lamellar tears) and volumetric flaws less reliably (porosity and rounded slag), provided the beam angle suits the flaw. The product form decides which flaws are likely and which scan finds them.

The table gives typical practice, not a code requirement. Which document applies, and whether UT is done at all, comes from the purchase order, the contract documents or the design code, as later sections explain.

Two flaw groups deserve a warning. Porosity scatters sound in many directions and gives small, wandering echoes; radiography is usually better at showing it. Surface-breaking toe cracks can be found by angle-beam UT, but magnetic particle testing is often faster and more sensitive for them on ferritic steel, which is why UT and magnetic particle examination are commonly specified together on structural welds.

Key takeaway: match the scan to the product: 0° for plate and forgings, angle beam for welds, and a second method where UT is weak.

How does a straight-beam lamination scan work?

A lamination scan sends a 0° longitudinal beam through the plate thickness and watches the back-wall echo; a lamination shows as an echo between the surface and the back wall, and a large one cuts off the back-wall echo entirely. Laminations lie parallel to the rolled surface, so the 0° beam meets them square, which makes this the one flaw a straight beam finds best.

Laminations form when inclusions, porosity or pipe from the ingot or slab are rolled flat into a sheet-like separation inside the plate. They are usually invisible from the surface. They matter in two ways. First, a lamination in the zone a weld will stress can open up under through-thickness load. Second, a lamination in the path of an angle beam blocks or redirects that beam, so a weld scanned from that side is not really tested.

What do the ASTM plate specifications cover?

ASTM publishes two widely used straight-beam plate specifications. ASTM A435/A435M (active edition A435/A435M-17, reapproved 2023) covers straight-beam, pulse-echo examination of rolled, fully killed carbon and alloy steel plates 1/2 in. (12.5 mm) and over. Its scope states its purpose plainly:

"It was developed to assure delivery of steel plates free of gross internal discontinuities such as pipe, ruptures, or laminations and is to be used whenever the inquiry, contract, order, or specification states that the plates are to be subjected to ultrasonic examination." (ASTM A435/A435M-17(2023), section 1.1, ASTM store page)

ASTM A578/A578M (A578/A578M-17, reapproved 2023) covers rolled carbon and alloy steel plates 3/8 in. (10 mm) and over "for special applications". Its scope says "the method will detect internal discontinuities parallel to the rolled surfaces" and that it provides three levels of acceptance standards, so the purchaser chooses how strict the plate must be. Both specifications require examiners qualified and certified to the latest SNT-TC-1A or an equivalent standard acceptable to the purchaser.

Neither specification applies automatically. Each is a purchase requirement: it governs only when the order or the governing specification calls it up. A mill test report that doesn't mention A435 or A578 means the plate was not ultrasonically examined to either one.

How does an inspector run the scan?

The scan is set up on the plate itself. The inspector couples a 0° probe on a sound area, adjusts gain to put the first back-wall echo at a set screen height, and scans along a grid or in continuous paths. Practitioners quoting A435 describe setting the back reflection between 50% and 75% of full screen height before scanning (secondhand, from an NDT.net forum thread; check the edition you are working to). Then the inspector watches for two signs:

• an intermediate echo between the initial pulse and the back wall
• a drop or total loss of the back-wall echo

The loss of back reflection is the stronger sign. A large lamination returns all the sound at its own depth, so no energy reaches the back wall. The inspector then maps the boundary by moving the probe outward until the back-wall echo returns, and marks the outline on the plate. The acceptance criterion in A435 is written in exactly those terms. Its abstract on the ASTM store page reads: "Any discontinuity indication causing a total loss of back reflection which cannot be contained within a circle is unacceptable." The circle's diameter is quoted in practitioner discussions as 3 in. (75 mm) or one half of the plate thickness, whichever is greater (A435 section 6.1 as quoted on an NDT.net forum thread; CONFIRM in the current edition).

Two traps catch new inspectors. A lamination produces repeated echoes at multiples of its own depth, and a careless reading treats the second multiple as a second, deeper flaw. And a clean back-wall echo only proves the area under the probe is sound; coverage depends on the scan pattern, so a coarse grid can step over a small lamination entirely.

Why do welding codes scan the base metal before the weld?

AWS D1.1 practice, as described by users of the code, includes a straight-beam scan of the base metal through which the angle beam will pass, done before shear-wave testing of complete joint penetration (CJP) welds. Older editions numbered the clause 6.25.5 or 6.26.5, and forum discussion says the scan has no acceptance criteria of its own: the results are reported, and they decide which side and which angles can test the weld (CONFIRM the clause in the 2020 or 2025 edition named in your contract). The point is practical. If a lamination sits in the beam path, the inspector must test from another face or with another angle, or report that part of the weld as not tested.

D1.1 also deals with mill-induced laminar discontinuities that show on cut edges and groove faces, and code users describe exploring those with straight-beam UT to find their true extent, with the Engineer deciding disposition (AWS forum discussion; CONFIRM in the contract edition). The number limits in that provision aren't reproduced here because no public source confirmed them.

Key takeaway: the 0° lamination scan is a plate-acceptance test under A435 or A578 when ordered, and a beam-path check before weld UT.

Worked example 1: a lamination check on 1.5 in. plate

This example shows how a 0° reading becomes a depth and an accept-or-report decision. It uses Evident's longitudinal velocity for 1020 steel, 5,890 m/s (0.2319 in./µs). The velocity of the actual plate should be checked on the plate itself, because it varies with alloy and processing.

Inputs

Step 1: depth of the reflector. Depth = v_L × t ÷ 2 = 0.2319 × 5.17 ÷ 2 = 0.60 in. below the scanning surface, about 40% of the thickness.

Step 2: identify the multiples. The echoes at 10.34 µs and 15.51 µs are exactly 2 × and 3 × 5.17 µs. On a depth-calibrated screen they read 1.20 in. and 1.80 in. The 1.80 in. reading is deeper than the 1.50 in. plate, which proves these are repeat echoes bouncing between the scanning surface and the lamination, not two more flaws. They are reported as one lamination at 0.60 in.

Step 3: map the area. The inspector moves the probe outward from the indication and marks where the back-wall echo comes back. Two areas are found:

For 1.5 in. plate, half the thickness is 0.75 in., so the 3 in. minimum governs. On plate thicker than 6 in., T ÷ 2 would govern instead.

Step 4: check the weld beam path. Suppose area B lies next to a groove weld to be tested with a 70° probe from this face. A 70° beam reaches 0.60 in. depth at a surface distance of 0.60 × tan 70° = 0.60 × 2.7475 = 1.65 in. from the probe exit point on the first leg. After reflecting off the back wall, it climbs back to 0.60 in. depth at (2 × 1.50 − 0.60) × 2.7475 = 2.40 × 2.7475 = 6.59 in. from the exit point. If the probe must sit where either of those crossings falls inside area B, the lamination will block the beam, and that weld zone should be tested from another face or angle or reported as not tested.

Area B passes the plate specification and can still compromise the weld inspection. That is why the base-metal scan is reported even when it has no acceptance limits of its own. The A435 circle size is quoted secondhand; confirm it, and any stricter A578 level named in the order, before using this example as a procedure.

Which weld discontinuities does angle-beam UT detect?

Angle-beam UT detects planar weld discontinuities best, chiefly lack of sidewall fusion, lack of root fusion, incomplete penetration and cracks, when the beam strikes them close to perpendicular; it is weaker on porosity and rounded slag, which scatter rather than reflect. Choosing the angle from the joint geometry is the core skill.

A refracted shear wave is used for most structural weld UT because it lets the probe sit on the parent plate beside the weld cap and still reach the whole weld volume, bouncing off the back wall into a second leg if needed. The beam angle is chosen so that it meets the most likely flaw plane square. For a sidewall lack of fusion on a bevel, the best angle is about 90° minus the bevel angle, so a 30° bevel is best tested with a 60° probe. A 70° probe on the same joint meets the fusion face 10° off square and can return a weak echo from a real flaw. That off-angle effect is the main reason AWS D1.1 asks for testing from both sides of the weld axis wherever mechanically possible, and why its procedure chart assigns different angles to different thickness ranges and weld zones.

D1.1 doesn't use DAC in the usual way. It rates each indication with d = a − b − c, where a is the indication level in dB, b is the reference level in dB (from the 0.060 in. side-drilled hole in the IIW-type block), and c is an attenuation factor tied to sound path. Evident's AWS tutorial defines c as "Material attenuation factor, (2x [soundpath length in inches – 1 in.]) expressed in dB" and says the inspector then compares the rating "to an acceptance/rejection table provided by AWS". The table bands depend on weld thickness, probe angle and whether the connection is statically or cyclically loaded, so a d value means nothing without those qualifiers. The rating method and its tables are covered in our article on AWS D1.1 weld inspection criteria, and the difference between examination and acceptance standards in ultrasonic testing acceptance criteria. Sidewall fusion problems are covered in depth in lack of fusion in welds.

Key takeaway: angle-beam UT is a planar-flaw method; pick the angle from the bevel, scan from both sides, and rate amplitude only against the code's table and its qualifiers.

Worked example 2: locating a weld indication

This example converts an angle-beam reading into depth, surface position and position in the weld cross-section, and checks the result against the plate thickness. It uses a shear velocity in steel of 3,250 m/s (0.1280 in./µs) and ignores the wedge delay, which the instrument zero removes.

Joint and setup

Indication 1: sound path 3.20 in.

1. Which leg? 3.20 in. is less than the 4.39 in. first leg, so the reflector is on the first leg.
2. Depth: 3.20 × cos 70° = 3.20 × 0.3420 = 1.09 in. This is less than the 1.50 in. thickness, so the geometry is possible.
3. Surface distance from the exit point: 3.20 × sin 70° = 3.20 × 0.9397 = 3.01 in.
4. Position across the weld: 3.50 − 3.01 = 0.49 in. short of the centerline, on the probe side.
5. Where is the fusion face at that depth? Half the root opening plus the bevel offset above the root: 0.125 + (1.50 − 1.09) × tan 30° = 0.125 + 0.41 × 0.5774 = 0.36 in. from the centerline.
6. Interpretation: the reflector plots 0.49 − 0.36 = 0.13 in. beyond the near fusion face, in the heat-affected zone. Given normal plotting tolerance, it could be sidewall lack of fusion or an HAZ crack. The next step is to test it from the other side of the weld and with a second angle (60° meets the 30° bevel square), and to size its length by scanning along the weld.
7. Time of flight: 2 × 3.20 ÷ 0.1280 = 50.0 µs round trip in the steel.
8. D1.1 attenuation factor: c = 2 × (3.20 − 1) = 4.4 dB. If the indication level a is 44 dB and the reference level b is 46 dB, then d = 44 − 46 − 4.4 = −6.4 dB, which is then looked up in the contract-edition table for 1.50 in. at 70° and the connection's loading. The class isn't stated here because the table bands weren't verified from a public source.

Indication 2: sound path 6.10 in.

1. Which leg? 6.10 in. is longer than the 4.39 in. first leg, so the reflector is on the second leg, after the back-wall bounce.
2. Depth: 2T − SP × cos θ = 3.00 − 6.10 × 0.3420 = 3.00 − 2.09 = 0.91 in. below the scanning surface. (Using the first-leg formula would give 2.09 in., deeper than the plate, which is the classic error.)
3. Surface distance: 6.10 × 0.9397 = 5.73 in. from the exit point, which is 5.73 − 3.50 = 2.23 in. past the centerline on the far side.
4. Is that in the weld? The weld cap's half-width at the top surface is 0.125 + 1.50 × 0.5774 = 0.99 in. The reflector plots 2.23 − 0.99 = 1.24 in. beyond the far weld toe, in the parent plate.
5. Interpretation: it isn't a weld flaw. A second-leg echo out in the far plate points to a base-metal reflector, most often a lamination at about 0.91 in. depth or a geometric feature such as an attachment. The 0° scan of that area settles it.

The arithmetic is simple, but each number depends on the zero, the true angle and the velocity having been checked on a reference block. A 2° error in the real beam angle moves both positions by several hundredths of an inch per inch of sound path, which can move an indication from the fusion line into the weld metal. The formulas are collected in our ultrasonic testing formulas page.

Why does hydrogen cracking need a delay before UT?

Hydrogen (cold) cracks can form hours or days after a weld cools, so a weld examined too soon can pass UT and crack afterwards; that is why some codes set a minimum delay before the final examination of high-strength steels. The cracks usually sit in the heat-affected zone, under the bead or at the toe, and are planar, so angle-beam UT finds them well once they exist.

Hydrogen cracking needs three things together: diffusible hydrogen from moisture, contaminated consumables or the shielding environment; a hard, susceptible microstructure in the HAZ or weld metal; and tensile stress from restraint. Preheat, low-hydrogen consumables and controlled heat input reduce the risk. They don't remove it entirely, especially in thick, highly restrained joints.

TWI's work gives the clearest data on timing. Richard Pargeter's study Evaluation of Necessary Delay Before Inspection for Hydrogen Cracks (presented at the 82nd AWS Convention in 2001 and published in the Welding Journal research supplement in November 2003) monitored multipass butt welds with mechanized UT. In 350 MPa C-Mn steel, the paper reports: "Growth of existing sub surface HAZ cracks was observed for up to 6 hours after welding in the high restraint welds." For C-Mn steels up to 450 MPa yield and 50 mm thick, it proposed minimum delays before UT of 12 hours for SMAW at heat inputs up to 2.4 kJ/mm, and 24 to 36 hours for higher-heat-input SAW. In a 690 MPa (HY100) steel, TWI's summary of the paper reports new cracks appearing for up to 64 hours and proposes a 72-hour delay.

Code delays differ, and TWI's FAQ lists several:

The D1.1 clause number has moved between editions; one D1.1:2025 study guide places the 48-hour delay for quenched and tempered steels at 8.15.4 (CONFIRM in the edition your contract names). The practical rule for a structural job is to put the delay in the inspection plan, record the weld completion time and the UT time on the report, and treat any requested shortcut on A514-type steels as an Engineer's decision, not an inspector's.

Key takeaway: for high-strength quenched and tempered steels, final UT waits at least 48 hours under D1.1; for ordinary structural grades the code allows UT once the weld has cooled, but heavy restrained joints still deserve caution.

What is lamellar tearing, and how does UT find it?

Lamellar tearing is a stepped crack that forms in the parent plate just below a weld, parallel to the plate surface, when weld shrinkage pulls on plate that has poor through-thickness ductility. It forms during or soon after welding, mostly at T-joints, corner joints and cruciform joints in thick, restrained structural steel.

TWI's Job Knowledge article puts it simply: "Lamellar tearing can occur beneath the weld especially in rolled steel plate which has poor through-thickness ductility." TWI lists three conditions that must all be present:

• the shrinkage strains "must act in the short direction of the plate", that is, through the thickness
• the fusion boundary lies roughly parallel to the plane of the inclusions
• the plate "must have poor ductility in the through-thickness direction"

The susceptibility comes from flattened non-metallic inclusions, mainly manganese sulfides, rolled into planes parallel to the surface. When the joint pulls on those planes, they decohere and link up by short vertical shear steps, which gives the tear its terraced look. TWI ties the risk to the plate's short-transverse reduction of area (STRA): steels with STRA over 20% are "essentially resistant to tearing", while steels below 10 to 15% STRA "should only be used in lightly restrained joints". The through-thickness tension test behind those numbers is ASTM A770/A770M (A770/A770M-03, reapproved 2025), which covers plates 1 in. (25 mm) and over and whose stated principal purpose is "to provide a measure of the resistance of a steel plate to lamellar tearing". A770 is a mechanical test of the plate, not an ultrasonic one; the designer specifies it when a joint will load the plate through its thickness.

UT enters at two points. Before welding, a straight-beam scan of the plate where the T-joint will land finds laminations and dense inclusion bands that would raise the risk. After welding, tears are internal, so TWI notes that "internal cracks require ultrasonic examination techniques". A 0° beam from the face opposite the attachment meets a tear's terraces square and is the natural first look. An angle beam aimed into the HAZ below the weld finds the vertical steps. TWI also warns that "there may be problems in distinguishing lamellar tears from inclusion bands", which is the core interpretive problem: both lie parallel to the surface at similar depths. Comparing a pre-weld lamination map with the post-weld scan is the most reliable way to separate them, because a tear appears or grows after welding while an inclusion band was there before.

Prevention is a design and procurement matter: specify plate with guaranteed through-thickness properties where the joint demands it, detail joints so the weld doesn't pull directly through the plate thickness, and, where needed, butter the susceptible surface with a low-strength weld metal before making the joint, a method TWI says "has been widely employed". TWI also notes that aluminum-treated steels with very low sulfur (below 0.005%) carry a low risk.

Key takeaway: lamellar tearing is a plate-quality problem revealed by welding; scan the plate before, scan below the weld after, and compare the two maps.

How are steel forgings and castings examined?

Forgings are examined mainly with a straight beam calibrated by reference blocks or the DGS method, adding angle beams for rings and hollow forgings, under ASTM A388/A388M; castings are examined under ASTM A609/A609M, whose coarser grain makes the job harder. Both practices apply only when the order or specification calls them up.

Forgings

ASTM A388/A388M (active edition A388/A388M-26) covers "the contact, pulse-echo ultrasonic examination of steel forgings by the straight and angle-beam techniques", and its straight-beam techniques include the DGS (distance gain size) method. Its abstract describes the surface preparation the method needs: round forgings machined to cylindrical surfaces for radial examination, ends machined square for axial examination, and faces of disk and rectangular forgings machined flat and parallel. Forging reduces and orients the as-cast grain, so most forgings transmit sound well, and the main targets are inclusions, internal bursts, hydrogen flakes and laps. The machined, parallel surfaces are what make back-wall monitoring meaningful: as with plate, loss of back reflection is a key sign.

DGS sizing reports a flaw as the equivalent diameter of a flat-bottomed hole that would give the same echo at the same distance. It is a reflector-equivalent figure, not a true flaw size, because a real flaw's roughness and tilt change its echo. A388 itself is a practice with no built-in acceptance level; acceptance comes from the forging specification or the purchase order.

Castings

ASTM A609/A609M (active edition A609/A609M-12, reapproved 2023) covers pulse-echo examination of "heat-treated carbon, low-alloy, and martensitic stainless steel castings". It contains two procedures: Procedure A calibrates on a series of flat-bottomed-hole test blocks and has supplementary angle-beam requirements, and Procedure B calibrates on the back reflection from solid calibration blocks. Its scope note warns that "ultrasonic examination and radiography are not directly comparable" and describes the practice as a complement to radiographic examination under Guide E94/E94M.

Castings are harder for three reasons. As-cast grain is coarser than wrought grain and scatters sound, so lower frequencies are used at the cost of resolution. Cast surfaces are rough and often not parallel, which disturbs coupling and the back-wall reference. And the common casting flaws (shrinkage, gas cavities, hot tears) are irregular, so their echoes are ragged and harder to size than a lamination's. For these reasons foundries and purchasers often combine UT with radiography on critical castings.

Key takeaway: forgings suit UT well when surfaces are machined; castings need lower frequencies, careful calibration and often radiography beside UT.

Why do grain size and attenuation limit UT in some steels?

Sound loses energy as it travels through steel, partly by absorption and partly by scattering at grain boundaries, and in coarse-grained steels the scattering is strong enough to hide small flaws in grain noise. It is wrong to say ultrasound in steel "loses very little energy": attenuation is the main physical limit on UT in castings, heavy sections, some heat-treated steels and austenitic welds.

How does grain scatter depend on frequency?

When the grain is small compared with the wavelength, scattering falls in the Rayleigh regime, where attenuation from scattering rises roughly with the cube of the grain diameter and the fourth power of frequency. Going from 2.25 MHz to 5 MHz in that regime raises the scattering loss by (5 ÷ 2.25)⁴ = 24.4 times. Lower frequency helps, but the longer wavelength reduces resolution and sensitivity to small flaws. That trade-off is why inspectors drop to 1 MHz or 2.25 MHz on castings and heavy forgings and use 4 or 5 MHz on fine-grained plate and thin welds.

The wavelengths are v ÷ f; the "typical use" column is common practice, not a code requirement.

How do codes allow for attenuation?

AWS D1.1 builds a fixed attenuation allowance into its rating: the c factor of 2 dB per inch of sound path beyond the first inch. That is a convenient approximation, not a measured property of your steel. Published work (Holloway, 2017, in the CINDE Journal) shows the 2 dB/in. figure can be about 6 dB off at 60° and 70° with sound paths near 150 mm (6 in.), enough to move an indication across a class boundary. Where the real material attenuates more than the reference block, a reflector deep in the part looks smaller than it is. A transfer correction, measured by comparing the same echo in the block and in the part, is how procedures deal with the difference, and a DAC or TCG built on blocks of the same material captures the real loss with depth (see our DAC curve article).

Heat treatment and composition matter for the same reason. Normalized or quenched-and-tempered steels with fine grain transmit well; heavy sections that cooled slowly, and some alloy and cast grades, scatter more. Velocity also shifts a little with alloy and temperature (carbon steel velocity falls about 1% per 55 °C or 100 °F, per Evident), which is why the velocity is calibrated on the actual material.

Why are austenitic welds so difficult?

Austenitic stainless steel welds, and dissimilar welds with austenitic filler, solidify into long columnar grains that grow along the heat-flow direction. TWI describes the result: "the welds typically display an austenitic structure, which consists of textured large grains", and "this results in high ultrasonic scattering associated with mode conversion effects, beam distortion" and "a variation of ultrasound velocity with direction and position in the weld". In practice the beam bends and splits as it crosses the weld, so the calculated flaw position can be wrong, and grain noise can swamp small reflectors.

Shear waves suffer most from this anisotropy, so austenitic weld procedures commonly use refracted longitudinal (compression) wave probes, often dual-element, at low frequency, and qualify the technique on a mock-up welded the same way as the job. ISO 22825 covers UT of welds in austenitic steels and nickel-based alloys and, like most method standards, includes no acceptance levels. Stainless structural welding in the US falls under AWS D1.6/D1.6M (Structural Welding Code – Stainless Steel, 2017 edition with a 2021 amendment, per the ANSI listing), not D1.1, so a D1.1 UT procedure shouldn't simply be carried onto a stainless job.

Key takeaway: grain scatter and attenuation are real limits; lower the frequency, correct for transfer and attenuation in the actual material, and qualify austenitic weld procedures on representative mock-ups.

Which codes call for UT of structural steel, and who sets acceptance?

AWS D1.1 doesn't by itself require ultrasonic testing; it sets how UT is done and how indications are judged when the contract documents call for it. The requirement to test, and how much, comes from the contract, the building code chain or the purchase order.

D1.1 assigns visual inspection to every weld, but nondestructive testing beyond visual must be specified in the contract and bid documents, along with its extent. For building steel, the common trigger is AISC 360 Chapter N, which, as code users quote it, calls for UT of complete joint penetration groove welds 5/16 in. and thicker in Risk Category III and IV structures and of 10% of such welds in Risk Category II, while AISC 341 calls for UT of CJP welds in the seismic force-resisting system (secondhand quotations; AISC's site couldn't be read for this edit, so CONFIRM in the editions your project cites). In D1.1 itself, 5/16 in. is simply the bottom of the thickness range its UT part covers, which is where the "5/16 in. rule" confusion comes from.

Other documents work the same way:

API 1104 is outside this article's structural scope, but it is often misquoted. It recognizes manual and mechanized UT: a 2014 ECNDT paper by Ghaemi notes that the 19th edition (1999) "revised its description of UT requirements to include mechanized systems". Its Annex A alternative (engineering critical assessment) acceptance criteria need flaw height measured with known accuracy, which in practice means zone-discrimination AUT. Where UT is mandatory on a pipeline, a regulator or company specification makes it so. Washington State's hazardous liquid pipeline rule, for example, requires new girth welds to be inspected "by radiography or automatic ultrasonic testing in accordance with API 1104" (WAC 480-75-460). For US gas pipelines, PHMSA's 49 CFR 192.7 incorporates the 21st edition of API 1104, not the current 22nd.

Key takeaway: codes describe how to test and judge; contracts, design codes and orders decide whether to test.

Who is qualified to perform steel UT?

In the US, UT personnel are usually certified by their employer under a written practice based on ASNT's SNT-TC-1A or the ANSI/ASNT CP-189 standard; neither ASNT nor AWS certifies the technician under SNT-TC-1A. BINDT's PCN scheme is a UK third-party certification under ISO 9712, used mostly outside the US.

ASNT's standards page describes SNT-TC-1A (2024 edition, with addendum) as "Employer guidance for in-house NDT personnel qualification and certification programs" and CP-189 (2024) as a standard that "Establishes minimum requirements for employer-based NDT personnel certification, building on SNT-TC-1A". In both, the employer's Level III writes the practice, sets training and experience hours, administers exams and certifies Level I, II and III personnel. ASNT does offer central certifications of its own, but SNT-TC-1A certification belongs to the employer. The ASTM plate specifications above require examiners qualified to SNT-TC-1A or an equivalent acceptable to the purchaser, and contracts often name a level (commonly Level II to perform and evaluate, Level I under a Level II to assist).

Under ISO 9712, a central certification body certifies the person, independent of the employer. BINDT runs that scheme in the UK as PCN. A PCN certificate is a legitimate qualification, but a US structural contract that names SNT-TC-1A or CP-189 is asking for employer certification, and the original article's claim that BINDT certification is "accepted in some US industries" offered no source and has been removed.

Weld inspector credentials are a separate track. A Certified Welding Inspector (CWI) under AWS QC1 decides weld acceptance on many D1.1 jobs, and an AWS Certified Associate Welding Inspector (CAWI) works under a CWI. Neither credential qualifies a person to perform UT. Our pages on UT Level II responsibilities and the ultrasonic testing level 1 role explain the levels in more detail.

Key takeaway: check that the UT technician holds an employer certification to the practice the contract names, and that weld acceptance is assigned to the right person.

What can't ultrasonic testing detect in steel?

UT misses flaws that lie parallel to the beam, flaws hidden in a dead zone or behind a lamination, small flaws lost in grain noise, and surface flaws better found by MT or PT. It also can't tell a flaw's type with certainty from amplitude alone.

• Wrong orientation. A planar flaw parallel to the beam returns almost nothing. Transverse weld cracks are missed by a normal scan across the weld, and a 0° beam won't see a vertical crack.
• Near-surface dead zone. A single-element contact probe can't resolve reflectors just under the contact surface while it recovers from the transmit pulse. Dual-element probes and delay lines reduce the problem, and MT or PT covers the surface itself.
• Shadowing. A lamination or a large flaw blocks the beam from anything behind it, as worked example 1 showed.
• Grain noise and attenuation. In castings, heavy slowly cooled sections and austenitic welds, small reflectors can be below the noise, and calculated positions can be wrong because the beam bends.
• Closed cracks. A crack held shut by compressive residual stress can transmit sound and give a weak echo.
• Geometry echoes. Weld caps, root reinforcement, backing bars and attachment edges all produce echoes that must be plotted to be dismissed. Experienced plotting is what separates a root geometry echo from incomplete penetration.
• Porosity. Spherical pores scatter sound; radiography usually characterizes porosity better.
• Sizing limits. Amplitude-based methods rate indications against a reference, which is not a flaw size. Height sizing needs tip-diffraction techniques such as TOFD, which have their own near-surface and back-wall dead zones.
• Coverage. UT tests only where the beam goes. A coarse grid, a skipped side or an untested zone behind a lamination leaves steel untested even when the report shows no indications.
• Thin material. Weld UT practices such as ASTM E164 start around 1/4 in.; below that, other methods are usually more practical.

These limits are why a structural inspection plan usually pairs UT with visual inspection and MT or PT, and why a report should state the areas that could not be tested and why.

Key takeaway: a clean UT report means "no relevant indications in the volume covered"; state the coverage and the limits with the result.

How EMA Industries applies this

EMA Industries performs AWS D1.1 structural weld UT, magnetic particle and penetrant testing, and NDT Level III services from Southern California, and mobilizes nationwide. On a structural project, that means confirming which document actually requires UT and to what extent, running the base-metal lamination scan before angle-beam testing of CJP welds, holding the required delay on high-strength steels, plotting indications against the joint geometry rather than reading depth off the screen, and stating the coverage and any untested zones in the report. Level III support covers written procedures, transfer and attenuation corrections, and reviewing contractor UT reports for an owner. See our ultrasonic examination services or contact EMA Industries to scope a job.

Related reading

• The science behind ultrasonic testing: velocity, wavelength, impedance, Snell's law and near field
• Types of ultrasonic testing in NDT: pulse-echo, angle beam, PAUT, TOFD and immersion compared
• Sound waves used in rail inspection: how the same beam-to-flaw rule maps onto rail probes and defects
• Shear wave ultrasonic testing: the refracted shear waves used for weld UT
• AWS D1.1 weld inspection criteria: static and cyclic acceptance

Frequently asked questions

Can a straight-beam probe find cracks in a weld?

Rarely. Most weld cracks and fusion flaws run through the thickness, roughly parallel to a 0° beam, so the beam glances off them. A straight beam is used next to the weld to find laminations in the base metal and to measure thickness; the weld itself is tested with angle beams, usually 45°, 60° or 70°, chosen to meet the expected flaw plane square.

How deep can ultrasonic testing see into steel?

In fine-grained wrought steel, longitudinal waves can test several feet of material; long forgings and shafts are examined end to end. The practical limit is set by attenuation and grain scatter rather than distance alone, so a coarse-grained casting may be hard to test through a few inches. For welds, practices such as ASTM E164 cover roughly 1/4 in. to 8 in. thickness.

What is the difference between ASTM A435 and A578?

A435 is a general plate specification for fully killed carbon and alloy steel plate 1/2 in. and thicker, with a single acceptance rule based on total loss of back reflection. A578 covers plate 3/8 in. and thicker "for special applications" and offers three acceptance levels, so the purchaser can choose a stricter standard. Both apply only when the order or specification calls them up.

Does AWS D1.1 require ultrasonic testing of every CJP weld?

No. D1.1 requires visual inspection of all welds, and any NDT beyond that, including UT, must be specified in the contract documents along with its extent. Building codes such as AISC 360 Chapter N and AISC 341 are the usual source of UT requirements for CJP welds, depending on risk category and seismic design.

Why is UT of stainless steel welds harder than carbon steel welds?

Austenitic weld metal grows long, oriented grains that scatter sound and bend the beam, and the sound velocity changes with direction through the weld. Shear-wave signals suffer most. Procedures use low-frequency refracted longitudinal-wave probes, often dual-element, qualified on a mock-up of the same weld, under documents such as ISO 22825 and the stainless structural code AWS D1.6.

Sources

• ASTM International, A435/A435M-17(2023), Straight-Beam Ultrasonic Examination of Steel Plates, store page (scope 1.1, 1.2 and abstract), accessed October 2026
• ASTM International, A578/A578M-17(2023), Straight-Beam Ultrasonic Examination of Rolled Steel Plates for Special Applications, store page, accessed October 2026
• ASTM International, A388/A388M-26, Ultrasonic Examination of Steel Forgings, store page, accessed October 2026
• ASTM International, A609/A609M-12(2023), Castings, Carbon, Low-Alloy, and Martensitic Stainless Steel, Ultrasonic Examination Thereof, store page, accessed October 2026
• ASTM International, A770/A770M-03(2025), Through-Thickness Tension Testing of Steel Plates for Special Applications, store page, accessed October 2026
• ASTM International, E164-24, Contact Ultrasonic Testing of Weldments, store page, accessed October 2026
• NDT.net forum, ASTM 435 discussion (quotation of A435 section 6.1 and setup practice; secondary)
• TWI, Defects: lamellar tearing (Job Knowledge 47)
• R. Pargeter, TWI, Evaluation of Necessary Delay Before Inspection for Hydrogen Cracks, AWS Convention 2001; Welding Journal 82(11), 2003
• TWI, FAQ: Which standards stipulate delay time before inspecting welds?
• TWI, Influencing parameters for ultrasonic inspection of austenitic welds (Core Research Programme project 1153/2021)
• Evident Scientific, Ultrasonic testing of welds to AWS D1.1 (indication rating definitions)
• Evident Scientific, Material sound velocities
• ASNT, Standards: SNT-TC-1A and ANSI/ASNT CP-189
• BINDT, Standards that affect your PCN certificate
• M. Ghaemi, Codes for Automatic Ultrasonic Testing (AUT) of Pipeline Girth Welds, ECNDT 2014
• Washington Administrative Code, WAC 480-75-460, Welding inspection requirements (Cornell LII)
• PHMSA, 49 CFR 192.7, incorporation by reference (eCFR)