Weld Defect Types: Identification & UT Detection Methods

Weld defect types fall into two families that set the method. Planar flaws (cracks, lack of fusion, incomplete penetration) are best found by ultrasonic testing (UT), and AWS D1.1 rejects a crack of any size. Volumetric flaws (porosity, slag) show best on radiography (RT) and are judged by size, as in ASME VIII's 1/4 t limit on a rounded indication.

The usual list of defects is cracks, lack of fusion, incomplete penetration, porosity, slag inclusions and the profile defects undercut and overlap.

This page is the hub for EMA Industries' weld defect library. Each defect gets a short summary and a link to its detailed guide; this page covers how to classify a flaw, which method finds it, which probe angle to use and how codes treat it, ending with rail welds, where FRA defines a "defective weld".

Contents: Planar vs volumetric · Defect table · Each defect · Probe angles · Worked example · Codes · Rail welds · Limits · FAQ

What is the difference between planar and volumetric weld defects?

A planar defect is essentially two-dimensional, like a crack or an unfused face; a volumetric defect has real depth in all three directions, like a gas pore or a slag pocket. The distinction matters because it decides both the inspection method and how hard the code is on the flaw.

A flat reflector sends sound back to the probe like a mirror when the beam hits it square on, so a tight crack can give a strong UT echo. The same crack is almost invisible on a radiograph when the X-ray beam crosses it at an angle, because it removes almost no material along the beam path.

A pore is the opposite case. It is a sphere of missing metal, so it darkens a radiograph whatever the beam direction, but a sphere scatters sound in every direction and only a small part returns to the probe. TWI's guidance on the two most common cases matches this: "Ultrasonic inspection is normally more effective than radiography in detecting lack of inter-run fusion imperfections" (TWI, lack of sidewall and inter-run fusion), while radiography is normally the better method for porosity (TWI, porosity).

Standards formalize the split. ISO 23279:2017 gives a procedure for characterizing UT indications in welds as planar or non-planar, using echo amplitude, echo dynamics and directional response, and ISO 17640:2018 sets the manual UT techniques and testing levels for fusion welds.

Key takeaway: Classify the expected flaw first. If it is planar, plan UT and aim the beam at its face; if it is volumetric, RT usually shows it better.

Which NDT method finds each weld defect type, and how do codes treat it?

Each weld defect type has a usual cause, a method that finds it most reliably, and a typical code treatment. The table summarizes them for steel fusion welds. Code values are drawn from secondary sources and must be checked against the edition named in your contract.

Sources for code values: AWS D1.1 study material and forum reproductions (Med, CONFIRM); ASME VIII Div. 1 Appendices 4 and 12 and ASME B31.3 344.6.2 (secondary, CONFIRM). Full comparisons are in our guides to AWS D1.1 acceptance criteria and UT acceptance criteria across codes, and method-by-method detail is in NDT welding inspection: methods, codes and standards.

Are distortion and arc strikes weld defects?

Distortion and arc strikes are not planar or volumetric discontinuities. The original version of this page listed them alongside cracks and porosity, which mixed categories.

ISO 6520-1:2007, which classifies geometric imperfections in fusion welds, groups them as cracks, cavities, solid inclusions, lack of fusion and penetration, imperfect shape and dimensions, and miscellaneous imperfections. Arc strikes and spatter fall in the miscellaneous group. Distortion is not a weld discontinuity at all; misalignment falls under imperfect shape, and overall distortion is controlled by the dimensional tolerances of the fabrication code.

An arc strike can still matter. The rapid quench can harden the metal and start a small crack, and that crack is then a planar defect judged as one.

What are the main weld defect types?

Six defect types account for most weld rejections: cracks, lack of fusion, incomplete penetration, porosity, slag inclusions and undercut.

Cracks

A weld crack is a fracture in the weld metal or heat-affected zone (HAZ). Hydrogen (cold) cracks form hours after welding in hard HAZ microstructures, so codes such as AWS D1.1 set delays before final inspection for some high-strength steels. Hot cracks form during solidification, often down the weld centerline, and crater cracks form in an unfilled weld end.

Buried cracks are found with angle-beam UT, and surface cracks with MT on ferritic steel or PT on any non-porous metal. Because a crack tip keeps concentrating stress, most structural and pressure codes reject cracks regardless of length. Rail fatigue cracking is covered in our guide to rail fatigue crack detection.

Lack of fusion

Lack of fusion is a failure of the weld metal to fuse with the sidewall or with a previous pass. It is planar and often lies along the bevel face, so a probe whose beam strikes that face near 90° gives the strongest echo, and a beam from the wrong side may miss it.

Our full guide, lack of fusion in welds, works through the bevel-to-probe angle rule and shows how an off-angle lack of fusion can rate as a pass under AWS D1.1.

Incomplete penetration

Incomplete (or lack of) penetration means the weld didn't reach through the joint root, leaving an unfused gap. It differs from lack of fusion: penetration is about depth into the joint, fusion is about bonding to a face. UT finds it as a root-zone reflector, but the inspector must separate it from the normal echo off the root bead or backing bar, which sits at a similar position.

Porosity

Porosity is gas trapped as the weld metal solidifies. It ranges from scattered fine pores to clusters, linear porosity along a root and elongated piping porosity. RT normally shows porosity more clearly than UT, and codes accept it up to size and aggregate limits. Crater pipe is a shrinkage cavity, not gas porosity.

The full guide, porosity in welding: prevention and detection, gives the AWS D1.1, ASME VIII, B31.3 and API 1104 limits with worked acceptance examples.

Slag inclusions

Slag inclusions are non-metallic flux residue trapped between passes or against the sidewall. TWI notes that slag "is normally seen as elongated lines either continuous or discontinuous along the length of the weld" and is readily identified on a radiograph (TWI, slag inclusions).

Because slag is elongated, UT can detect it, and it is easily confused with lack of fusion along the same sidewall. Our guide to slag inclusion in welding and ultrasonic testing covers how to tell the two apart.

Undercut, overlap and profile defects

Undercut is a groove melted into the base metal at the weld toe and left unfilled. Overlap is weld metal rolled onto the surface without fusing. Both are found by visual testing (VT) with gauges, and UT adds little because the toe geometry already reflects sound.

Undercut limits in AWS D1.1 depend on material thickness and on static or cyclic loading, so a bare number is never enough.

Key takeaway: Volumetric flaws are judged by size; planar flaws are usually judged by type. Find out which family the indication belongs to before applying any limit.

How is the UT probe angle chosen for a weld?

Choose the angle whose beam strikes the most likely planar flaw closest to 90°, then check that the sound path is short enough for the plate thickness. Standard refracted shear-wave angles are 45°, 60° and 70°, and AWS D1.1 requires the search unit to produce a beam within ±2° of the chosen angle (as quoted from earlier editions; confirm in yours).

Three rules drive the choice:

1. Fusion faces. For sidewall lack of fusion, the angle that meets the bevel face at normal incidence is about 90° minus the bevel angle. A 30° bevel calls for a 60° probe. ISO 17640 similarly requires one angle to examine the fusion faces at or near normal incidence (BINDT summary; Med).
2. Thickness. In thin plate, a 70° beam travels a short distance and covers the root and cap without excessive skip distance. In thick plate, the 70° sound path becomes long and attenuation grows, so lower angles (60°, 45°) are used for parts of the weld.
3. The code's scanning chart. AWS D1.1 assigns angles by thickness and by weld zone in its UT testing procedure chart (Table 6.7 in editions before 2020; check the number in your edition). Forum reproductions of older editions show procedures that mix angles by depth, for example 60° for the top quarter and 70° for the middle half and bottom quarter (Med, CONFIRM). A footnote allows further evaluation with whichever of 70°, 60° or 45° is nearest perpendicular to the suspected flaw.

D1.1 Part F calls for a straight-beam scan of the base metal the angle beam will pass through (confirm the clause in your edition), because a lamination there can hide a weld flaw or produce a false indication. Our guide to angle beam testing in UT explains wedge angles and refraction, and UT calibration covers setting the reference level.

Worked example: probe geometry and indication location

A depth that comes out greater than the plate thickness, or on the wrong leg, means the geometry was misread. These numbers show how probe angle and thickness interact, using leg sound path = t ÷ cos θ, full-skip distance = 2t × tan θ, and the D1.1 attenuation factor c = 2 × (sound path in inches − 1) dB.

The 2 in. rows show why thick welds are not scanned with 70° alone. The 70° beam needs about 11 in. of clear scanning surface for one full skip and loses roughly 6 dB more than the 45° beam to the time it reaches the far side. The 2 dB per inch approximation in D1.1 can itself be about 6 dB off at long 60° and 70° paths (Holloway, CINDE Journal, 2017), another reason to keep paths short.

Locating an indication in the 1.0 in. plate at 70°:

• Indication at a sound path of 2.20 in. That's less than the 2.92 in. leg, so it's in the first leg. Depth = 2.20 × cos 70° = 2.20 × 0.342 = 0.75 in. below the scanning surface. Surface distance from the beam exit point = 2.20 × sin 70° = 2.20 × 0.940 = 2.07 in.
• Indication at a sound path of 4.00 in. That's past 2.92 in., so the beam has bounced off the back wall and is in the second leg. Depth = 2t − (4.00 × 0.342) = 2.00 − 1.37 = 0.63 in. below the scanning surface. Surface distance = 4.00 × 0.940 = 3.76 in.

Both depths are inside the 1.0 in. plate. If the second indication had been read as a first-leg depth (4.00 × 0.342 = 1.37 in.), it would sit below the back wall, which is impossible. The depth and surface distance then tell the inspector whether the reflector lies on the bevel face (likely lack of fusion), at the root (incomplete penetration or root geometry) or in the weld center (porosity, slag or centerline crack).

Key takeaway: Probe angle selection is a geometry problem first. Pick the angle that faces the expected flaw, then confirm the sound path and depth fit the plate.

How do AWS D1.1 and ASME treat weld defect types?

Both codes reject cracks outright, but they treat other weld defect types differently, and neither code's examination rules are its acceptance rules. AWS D1.1/D1.1M:2025 is the current structural steel code (AWS announcement), though many contracts still cite the 2020 edition.

AWS D1.1 (structural steel).

• Visual acceptance is in Table 8.1 (2020 numbering), with static, cyclic and tubular columns. It prohibits cracks of any size and sets undercut, porosity and profile limits by thickness and loading.
• UT indications are rated d = a − b − c and classed A to D against tables that depend on weld thickness, probe angle and static or cyclic loading. There are no fixed dB bands.
• D1.1 does not by itself require UT. The contract, or a building code such as AISC 360, sets which welds get UT.

ASME (pressure equipment and piping).

• ASME BPVC Section V says how to examine; acceptance comes from the construction code, such as Section VIII or B31.3.
• Section VIII Div. 1 Appendix 12 UT rejects cracks, lack of fusion and incomplete penetration at any length. Other indications reject only if the amplitude exceeds the reference level and the length exceeds the thickness-based limit (secondary sources; CONFIRM).
• B31.3 344.6.2 judges UT indications by amplitude and length, so it doesn't reject lack of fusion regardless of length the way Appendix 12 does (Low-med; CONFIRM).

ASME weld ultrasonic inspection covers the Section V and VIII workflow, and the two acceptance guides linked above give worked ratings.

What weld defects occur in thermite and flash-butt rail welds?

Rail welds suffer the same planar and volumetric defect types as structural welds, but FRA defines when they become a "defective weld" and what the track owner must then do. Continuous welded rail is made by flash-butt welding rail into strings in a plant, then joining strings in the field, usually by thermite (aluminothermic) welding (continuous welded rail).

The FRA definition of a defective weld

Under 49 CFR 213.113(d)(6), a defective weld is:

"a field or plant weld containing any discontinuities or pockets, exceeding 5 percent of the rail head area individually or 10 percent in the aggregate, oriented in or near the transverse plane, due to incomplete penetration of the weld metal between the rail ends, lack of fusion between weld and rail end metal, entrainment of slag or sand, under-bead or shrinkage cracking, or fatigue cracking."

The same paragraph says weld defects may start in the head, web or base, and that a weld defect running lengthwise through the weld is treated as a split web for remedial action. The definition is sized against head area even when the flaw sits in the web or base.

Thermite (aluminothermic) welds

A thermite weld is cast in a mold between rail ends, so its defects are mostly casting defects. An AAR survey of revenue-service thermite weld failures (TTCI TD-00-013, Kristan, 2000) found slag inclusions or gas porosity in 40% of reported failures, mainly at the web/base fillet, fatigue from weld-metal run-out in 32%, hot tears in 15%, and smaller shares of pull-aparts, grinding burns and sand inclusions.

Among defects found by detector cars, the same survey found that 44% were fatigue fractures in the head starting at micro-defects in the weld, with shrinkage and slag or porosity at 12% each.

The same digest warns that its data is likely skewed because rail flaw detection "only scans the head and web of the rail, and not the majority of the base." Hand UT of the base, and MT of the collar and fillets, closes that gap. See aluminothermic weld inspection.

Flash-butt welds

A flash-butt weld is forged: the rail ends are heated by resistance and arcing, then upset together, and the molten metal and oxides are expelled as flash. Its typical defects are planar. A study of flash-butt welded rails (Shvets et al., Paton Welding Journal, 2023) describes "cold welding" from insufficient heating before upsetting, lack-of-fusion areas with smooth, flat fracture faces, and clusters of silicate inclusions ("matt spots") where oxides were not expelled.

These flat, tight defects lie in the transverse plane of the weld, so they behave like lack of fusion in plate: they can give weak echoes unless the beam meets them squarely. A TTCI phased array study also warned that "un-ground flashing surrounding the weld can cause reflections that must be distinguished from true defects" (TTCI TD-16-050, 2016), and showed a geometry echo from a healthy flash-butt weld. See flash-butt weld inspection.

What remedial action applies to a defective weld?

A transverse defective weld shares a row of the 213.113 remedial action table with transverse fissures, detail fractures and engine burn fractures.

Source: 49 CFR 213.113(c) remedial action table, read from the 2024 CFR on govinfo. A = a designated person visually supervises each operation over the rail. A2 = a designated person inspects it, then may allow up to 10 mph for up to 24 hours. C = joint bars within 10 days, with speed limits until and after they are applied. D = joint bars within 7 days, with speed limited to 30 mph until they are applied. E and H = joint bars bolted per 213.121(d) and (e), and speed limited to 50 mph. Each speed limit applies or the class maximum, whichever is lower.

Worked example. A hand UT technician sizes a transverse indication in a thermite weld at about 30% of the head area. That is above the 5% threshold, so it's a defective weld, and it falls in the 25 to less than 60% band: action D, joint bars within 7 days and no more than 30 mph until they're on. A lengthwise weld defect of 3 in. would instead be treated as a split web and sized by length. Codes and the full table are in FRA defect codes explained.

Key takeaway: In rail, a weld flaw becomes a regulated defect at 5% of head area individually or 10% in aggregate, and the size band sets the remedial action.

What weld inspection can't tell you

No single NDT method finds every weld defect type, and some flaws escape all of them. Plan the inspection around these limits:

• Tight planar flaws. A tight lack of fusion or cold weld with no gap can transmit sound and give little or no echo. Published work (Rihar, NDT.net) notes that pure lack of fusion with metal-to-metal contact can't be detected by any NDT method.
• Off-angle reflectors. A planar flaw tilted away from every beam angle used returns weak echoes and may rate as acceptable.
• Radiography and planar flaws. RT can miss cracks and lack of fusion that don't line up with the radiation beam.
• Geometry echoes. Root beads, backing bars, weld caps and rail weld flash all reflect sound. Separating them from defects needs plotting and sometimes a second angle.
• False calls. In the TTCI thermite survey, 24.4% of the 41 detector car finds examined contained no observable defect, so sound welds were removed. Hand UT verification before cutting reduces this.
• Rail base coverage. Head-contact rail flaw testing, conventional or phased array, can't inspect the rail base flanges, where many thermite weld defects start (TTCI TD-16-050).
• Detection is not acceptance. An indication is only a defect once the governing code, contract or 213.113 says it is. The examination standard doesn't decide that.

PAUT and TOFD add imaging and through-wall sizing but share the same physics. See phased array ultrasonic testing and PAUT vs TOFD for rail weld inspection.

How EMA Industries applies this

EMA Industries performs AWS D1.1 structural weld UT, MT and PT, rail weld inspection of thermite and flash-butt welds, and NDT Level III services, including written UT procedures that set probe angles and scanning plans for each joint type. EMA is based in Southern California and mobilizes nationwide. To plan weld or rail weld inspection, see our ultrasonic examination services or contact EMA Industries.

Related reading: Lack of fusion in welds · Porosity in welding · Slag inclusion in welding · AWS D1.1 acceptance criteria · Magnetic particle testing

Frequently asked questions

What are the main weld defect types?

The usual list is cracks, lack of fusion, incomplete penetration, porosity, slag inclusions, undercut, overlap and excess reinforcement. The first three are planar defects, porosity and slag are volumetric, and the rest are surface profile conditions. Arc strikes, spatter and distortion are sometimes added to such lists, but they belong to other categories.

Which NDT method is best for detecting weld defects?

It depends on the defect family. UT is usually best for planar flaws such as cracks and lack of fusion, RT for volumetric flaws such as porosity and slag, MT or PT for surface cracks, and VT for profile defects such as undercut. Most inspection plans combine VT with one volumetric method.

Is every weld discontinuity a defect?

A discontinuity becomes a defect only when it exceeds the acceptance criteria of the governing code or contract. Small, scattered porosity may be acceptable, while a crack of any size is usually rejected. In rail, a weld discontinuity becomes a "defective weld" under 49 CFR 213.113 at 5% of head area individually or 10% in aggregate.

Why does UT find lack of fusion better than radiography?

Lack of fusion is a thin, flat reflector. A UT beam striking its face near 90° returns a strong echo, while an X-ray beam crossing it at an angle passes through almost no missing metal and shows little contrast. That's why the probe angle is chosen to face the bevel.

What causes defects in thermite rail welds?

Most come from the casting process: slag or gas porosity, weld-metal run-out, hot tears, shrinkage and sand inclusions. An AAR survey (TTCI TD-00-013, 2000) found slag or porosity in 40% of reported thermite weld failures, mostly at the web/base fillet, an area that test cars largely don't scan.

Sources

• 49 CFR 213.113, Defective rails, eCFR (text read on Cornell LII and the 2024 CFR on govinfo), October 2026
• TWI Job Knowledge: Lack of sidewall and inter-run fusion
• TWI Job Knowledge: Porosity
• TWI Job Knowledge: Slag inclusions
• ISO 6520-1:2007, Classification of geometric imperfections in metallic materials, Part 1: Fusion welding, EVS catalog listing
• ISO 23279:2017, Ultrasonic testing: characterization of discontinuities in welds, EVS catalog listing
• ISO 17640:2018, Ultrasonic testing: techniques, testing levels and assessment, EVS catalog listing
• AWS D1.1/D1.1M:2025 release announcement, American Welding Society
• TTCI Technology Digest TD-00-013: thermite weld failures, July 2000
• TTCI Technology Digest TD-16-050: Characterizing rail flaws using phased array ultrasound, November 2016
• Shvets, Ziakhor and Kapitanchuk, "Features of formation and transformation of oxides in flash-butt welding of K76F rails", Paton Welding Journal 7 (2023)
• Rihar, lack of fusion detection, NDT.net
• Holloway, AWS D1.1 attenuation factor analysis, CINDE Journal, 2017