Weld Imperfections and How NDT Finds Them

A weld is the zone where the parent metal melts and resolidifies. Metallurgically it is the most highly stressed region of the joint and the one most prone to imperfections. Which imperfection a weld contains matters, but where it sits, which way it is oriented and what material it sits in matter just as much: those three things decide the detection method. Hunting a surface-breaking crack and a buried slag inclusion with the same technique means reaching for the wrong tool. This article walks through the imperfections you meet most often in welding. For each one it explains how it forms and why a particular NDT method (VT/PT/MT/RT/UT/PAUT/TOFD) is the right choice — in terms of orientation, location and material.

1. The logic behind imperfection–method matching

NDT method selection comes down to three questions. Is the imperfection at the surface, just below it, or in the volume? Surface and near-surface imperfections belong to VT/PT/MT; volumetric ones belong to RT/UT/PAUT/TOFD. How is it oriented? Planar, steeply oriented imperfections (cracks, lack of fusion) return a strong echo when the sound beam strikes them square on, but stay invisible on a radiograph unless the beam runs parallel to them. Volumetric, rounded imperfections (porosity, slag) produce clear contrast on a radiograph. What is the material? MT works only on ferromagnetic materials. In austenitic stainless, duplex and nickel-based welds, the anisotropic, columnar-dendritic grain structure bends the beam, attenuates it heavily and raises the noise floor, which makes conventional UT difficult. Aluminum does not belong in that group: it is a low-attenuation material and takes UT well — aerospace tests it ultrasonically as a matter of routine. It simply cannot be tested by MT, because it is not ferromagnetic. These three axes come back in every imperfection below.

A word on wording. This article says imperfection for a discontinuity classified to ISO 6520-1, and defect only for one that fails the acceptance criteria. Anything you have found but not yet evaluated is an indication. Every defect is a discontinuity; not every discontinuity is a defect — the acceptance criteria decide which is which.

2. Porosity

Why it forms. Gas trapped in the molten pool that cannot escape before solidification — usually hydrogen, nitrogen or carbon monoxide (CO). Damp covered electrodes, oily or rusty surfaces, insufficient shielding gas and an excessively long arc are the typical causes. It shows up as isolated, scattered or clustered pores, or as elongated piping (wormhole) porosity.

Best method: RT (radiography). Porosity is a volumetric, rounded imperfection. The cavity absorbs less radiation than the surrounding metal, so it appears on film or on a digital detector as sharp, round dark spots. It has no preferred orientation, so radiography's classic weakness — missing a planar imperfection that lies parallel to the beam — does not apply. That is why porosity is the imperfection RT catches most comfortably. Surface-breaking pores also show up under VT and PT, but only RT (or volumetric UT) reveals how porosity is distributed through the thickness. Under UT, a round pore scatters sound in all directions: you see individual low-amplitude echoes, or, in a dense cluster, a back-wall drop. Quantifying it is far less intuitive than with RT.

3. Slag inclusion

Why it forms. In slag-generating processes (SMAW, FCAW, SAW), the welder does not clean the slag off completely between passes, or a narrow, badly shaped groove traps it in the pool. Most often it sits along the fusion line between passes, in elongated lines.

Best method: RT. Slag is volumetric as well, but its density differs from that of the weld metal, so on a radiograph it usually gives dark indications with irregular, more angular edges than porosity. Its location (between passes) and its volumetric nature both suit RT. Volumetric UT/PAUT will also find slag, but because slag scatters sound, it produces wandering, unstable echoes that are hard to size. Surface methods (PT/MT) only help when the slag breaks the surface; on buried slag they do nothing. That is why RT stays the first choice.

4. Lack of fusion (LOF)

Why it forms. The molten metal does not fuse properly with the parent metal or with the previous pass — sidewall lack of fusion, or inter-run lack of fusion. Low heat input, a wrong torch angle, a narrow groove and excessive travel speed are the main causes. It is a planar imperfection, usually oriented parallel to the sidewall of the joint, and it is a dangerous one.

Best method: UT/PAUT. Because LOF is planar, it creates a real problem for radiography: unless the beam runs closely parallel to the fusion plane, the imperfection is virtually invisible on the film. This is exactly where RT is least reliable. Ultrasonically, the unfused face acts as a reflector perpendicular to the beam and returns a strong echo. PAUT sweeps the beam electronically through a range of angles, which markedly improves the chance of catching sidewall lack of fusion; a sectorial scan (S-scan) covers the sidewall at the ideal angle. So for LOF the primary method is UT, preferably PAUT — RT alone is not considered sufficient.

5. Lack of penetration (LOP / incomplete penetration)

Why it forms. The weld metal fails to reach the root face; the root pass stays incomplete. Too wide a root face, too narrow a root gap, low current or high travel speed cause it. It leaves a regular, planar void along the root line.

Best method: RT and UT together. Because LOP forms a regular line at the root, a beam aligned with the root plane shows it very clearly on the radiograph as a sharp, straight, dark line — a case where RT is strong. On pipe that can be reached from one side only, RT is the classic way to assess root penetration. But when you need to measure the height of the imperfection, UT/PAUT/TOFD takes over; TOFD in particular sizes root imperfections accurately using diffracted signals. TOFD is not flawless, though: it has a lateral (surface) wave dead zone near the scanning surface and another dead zone near the back wall. If the root falls into either one, back it up with PAUT. Where the root is accessible, VT and PT are useful complements.

6. Cracks: hot, cold and hydrogen cracking

Cracks are the least acceptable and the most dangerous imperfection; most codes allow no crack of any size.

Hot cracking (solidification and liquation cracking): During solidification, low-melting-point films left at the grain boundaries cannot withstand the shrinkage stress. Sulfur and phosphorus contamination and heavy restraint against shrinkage both raise the risk. The two sub-types differ in location: solidification cracking occurs in the weld metal, along the centerline and in the crater; liquation cracking occurs not in the weld metal but in the heat-affected zone (HAZ) next to the fusion line — in the partially melted region — where grain boundary films remelt. Lumping both under "centerline cracking" is wrong.
Cold (delayed) cracking — HACC (hydrogen-assisted cold cracking): In hardenable ferritic/martensitic steels — those that can form martensite, with a high carbon equivalent — it appears hours to days after welding. Three factors must coincide: diffusible hydrogen + a hard (martensitic) microstructure + tensile stress. Damp electrodes, thick sections welded without preheat and a high carbon equivalent are the typical causes; it is most often found in the HAZ. The literature also uses HAC and HICC for it. One caution: HIC (hydrogen-induced cracking) in corrosion terminology is the blistering and stepwise cracking that occurs in wet H₂S (sour) service. It needs no applied stress and must not be confused with post-weld delayed cold cracking.

Best method. For surface-breaking cracks in ferromagnetic steel, MT is the first choice: apply the magnetic flux perpendicular to the crack and the leakage flux gives it away. Since you cannot know the crack direction in advance, magnetize the part in two perpendicular directions (longitudinal and transverse, for example) so you do not miss a crack lying parallel to the flux. MT also picks up cracks just below the surface, and it tolerates a rough weld surface better than PT does. On non-magnetic material — austenitic stainless (300 series) and aluminum — MT does not work, so use PT. (Martensitic 410/420, ferritic 430 and duplex stainless steels are ferromagnetic, and MT applies perfectly well to those. Do not let the word "stainless" talk you out of MT.) For buried or internal cracks, UT/PAUT is the primary method: the crack is planar and steeply oriented, so it returns a strong echo. RT shows a crack only when the beam is closely aligned with the crack plane; otherwise it misses it. And because hydrogen cracking is delayed, test only after the waiting period the applicable code requires. AWS D1.1, for instance, requires final NDT on quenched-and-tempered steels (Q&T, e.g. A514/A517) no sooner than 48 hours after the weld is completed; other codes and steels set different periods. Test too early and the weld can come out clean, with the imperfection appearing afterwards.

7. Undercut

Why it forms. A groove-shaped notch at the weld toe (or the root toe) where the parent metal has melted away and has not been filled with weld metal. Excessive current, a long arc, a wrong torch angle and too high a travel speed cause it. It is surface-breaking, it concentrates stress through its notch effect, and it initiates fatigue cracks.

Best method: VT. Undercut is at the surface. Assess it directly by visual testing, with good lighting, a proper viewing angle and depth measured with a weld gauge. VT is the primary method. Because a suspect, sharp-tipped undercut may hide the start of a crack, back it up with MT on ferritic steel or PT on non-magnetic material. On a radiograph, undercut can appear as slight darkening at the weld edge, but its depth cannot be measured reliably. That is why RT is not primary here.

8. Overlap

Why it forms. Weld metal flows over the parent metal without fusing to it, creating an unfused lap at the toe or at the root. Low heat input, excessive deposition, a wrong angle and a dirty surface cause it. The unfused line underneath the overlap forms a sharp notch, a discontinuity in its own right. (Note: some sources treat "cold lap" as a synonym for overlap, while others use it for inter-run or surface lack of fusion. The two should not be equated outright.)

Best method: VT plus a surface method. Overlap is an imperfection of weld surface geometry, so find it first with VT — the unfused protrusion stands out in profile. Where the unfused line underneath breaks the surface, MT (ferromagnetic) or PT (non-magnetic) gives a linear indication. On a radiograph, overlap rarely images reliably, because the fusion line lies parallel to the beam. So for overlap, surface methods lead.

9. Summary: which method for which imperfection?

Porosity → RT (volumetric, rounded, no orientation); VT/PT if surface-breaking.
Slag → RT (volumetric, between passes); UT/PAUT as a complement.
Lack of fusion (LOF) → PAUT/UT (planar, perpendicular to the beam; RT misses it).
Lack of penetration (LOP) → RT (clear line at the root) + UT/TOFD (sizing).
Cracks → surface-breaking: MT (steel and magnetic stainless grades) / PT (austenitic stainless and aluminum); buried: PAUT/UT; for HACC, respect the waiting period.
Undercut → VT (measured with a gauge) + MT/PT.
Overlap → VT + MT/PT.

The general rule: RT for volumetric, rounded imperfections; UT-PAUT for planar, steeply oriented ones; VT/MT/PT for surface imperfections. RT and UT complement each other — one is superior on porosity and slag, the other on cracks and fusion problems.

10. Related standards

ISO 6520-1: Classification and numbering of imperfections in fusion welds (the basis for the definitions used here).
ISO 5817: Quality levels (B/C/D) for imperfections in fusion-welded joints in steel, nickel and titanium; beam welding (electron and laser) is outside its scope.
ASME BPVC Section V: NDT methods — Article 2 (RT), Article 4 (UT/PAUT/TOFD), Article 6 (PT), Article 7 (MT), Article 9 (VT).
RT: EN ISO 17636-1/-2 (film and digital) for technique + ISO 10675-1 for acceptance.
UT: EN ISO 17640 (technique) + ISO 11666 (acceptance).
PAUT: EN ISO 13588 (technique) + ISO 19285 (acceptance).
TOFD: EN ISO 10863 (technique) + ISO 15626 (acceptance).
PT: EN ISO 3452 (technique) + ISO 23277 (acceptance).
MT: EN ISO 17638 (technique) + ISO 23278 (acceptance).
VT: EN ISO 17637 (visual testing of welds).
ASME BPVC Section IX / AWS D1.1: Welding and NDT requirements, depending on the construction code.
ISO 9712: Qualification of NDT personnel (Level I/II/III).

Acceptance criteria always come from the construction or in-service code the weld is tested to.

From the field

The most common reason people miss an imperfection in the field is that they picked the wrong method. Looking for sidewall lack of fusion with radiography alone is like sighting straight down a planar imperfection and calling the weld clean — the weld is not clean; you simply cannot see it. The second trap is timing. On high-strength steel, shooting MT the moment the arc stops and moving on will miss delayed hydrogen-assisted cold cracking (HACC). Respect the waiting period the code sets: AWS D1.1, for example, calls for a minimum of 48 hours on Q&T steel. The third trap is material. "Stainless" does not mean non-magnetic. MT will not work on austenitic (300 series) stainless or aluminum, so switch to PT there — but martensitic, ferritic and duplex stainless steels are ferromagnetic, and skipping MT on those throws away a method that works. Knowing the imperfection is half of knowing how to look for it: an inspector who understands why it forms also knows where and how to look.

This article is for educational purposes and does not replace the official standard, the code or the written test procedure — the final decision always rests with the code in force and the responsible inspector.

Take it to the field: To keep these NDT methods, standard references and field steps in your pocket — completely offline and free — take a look at the Doawise NDT Guide app.

At DoaWise we carry out weld testing across the full range of methods — from VT through RT, UT and PAUT/TOFD — to international standards, detecting and sizing every imperfection with the right method and producing recordable, auditable results.