Weld Imperfections and Their Root Causes

A dark blot on a radiograph, or a thin groove glinting along a weld toe, is usually the end of the story. The real engineering story started much earlier — minutes before the arc was ever struck. A moisture-soaked electrode, a miscalculated heat input, a root pass that was never cleaned, a final crater filled in haste. What we call a discontinuity is almost never bad luck. It is the delayed consequence of a decision taken somewhere among four things: parameters, technique, material and joint preparation. This article looks at discontinuities not only from the NDT desk but from the tip of the welder's torch: how each one is born, which physical and metallurgical mechanism creates it, which acceptance criterion judges it, and which concrete field measure stops it before the arc even starts. Two documents frame everything that follows: EN ISO 6520-1:2007, which defines the imperfection types, and EN ISO 5817:2023 (4th edition, superseding the 2014 edition), which grades acceptance as B, C or D.

Note on editions: The numerical limits below come from EN ISO 5817:2023, Table 1 and Annex B, and they are reproduced here verbatim — an inspector's report cannot contain words like "mostly" or "generally". If your contract cites an older edition or a different code, verify the limits against the edition you actually use. The Foreword of the 2023 revision lists the main changes: editorial updates; use of the effective throat thickness aA in rows 1.3, 1.4, 1.16 and 3.2; changed or added figures for several rows (1.4, 1.5, 1.6, 1.11, 1.14, 1.16, 1.19, 2.12, 2.13, 4.1); a revised Table 1 row no. 4.1 (multiple imperfections); and removal of the old Annex B, replaced by the fatigue grading system. Most acceptance limits are unchanged from 2014.

1. Discontinuity, imperfection and quality level: getting the concepts straight

Start with the three words most often confused on site. A discontinuity is any deviation from the ideal homogeneous structure of the weld metal or the heat-affected zone (HAZ); on its own it is neither good nor bad. An imperfection is a discontinuity that has been defined and measured in the language of EN ISO 6520-1 and 5817. A defect is an imperfection that exceeds the acceptance criterion — one that requires rejection or repair. Put simply: every defect is a discontinuity; not every discontinuity is a defect — the acceptance criteria decide which is which. Mind the terminological nuance: in ISO language "imperfection" means any deviation from the ideal weld, so in practice it overlaps with discontinuity. The decisive, binding distinction is the one between imperfection and defect. The two-tier split that treats discontinuity and imperfection as separate stages is essentially AWS terminology. Either way, the bridge is the quality level: the inspector's answer to "is this pore a defect?" lies not in the existence of the pore but in the quality level being applied.

EN ISO 6520-1 sorts geometric imperfections in metallic materials into six main groups for fusion welding: cracks (100), cavities and porosity (200), solid inclusions (300), lack of fusion and penetration (400), imperfect shape and dimension (500), and miscellaneous imperfections (600). This numbering lets a test report speak one language — "503 excessive convexity" or "2011 gas pore". (Part 2 covers pressure welding; for fusion welding always write "-1".)

EN ISO 5817 then defines three quality levels for the types that matter in fabrication: D (moderate), C (intermediate) and B (stringent). D does not mean "low quality"; B is the most demanding requirement, the one that corresponds to a fully finished weld. The scope is not limited to steel: steel, nickel, titanium and their alloys, t ≥ 0.5 mm, beam welding (laser and electron) excluded. Two scope notes are routinely missed on site. First, the standard covers full penetration butt welds and all fillet welds; for partial penetration butt welds only the principles apply. Second, it is written primarily for direct visual testing; volumetric methods have their own acceptance standards. Aluminium and its alloys fall under EN ISO 10042, beam welding under EN ISO 13919-1/-2.

Three definitions you must know before reading any limit (ISO 5817:2023, Clause 3): Short imperfection: in a weld ≥ 100 mm long, the total imperfection length within the worst 100 mm section is ≤ 25 mm; in a weld < 100 mm long, ≤ 25% of the weld length. Many rows of Table 1 — 401-D, 4021-D, 5011/5012, 5013, 515 — rest on this definition. Without it, the h/l limits get misapplied in the field. Systematic imperfection: a series of imperfections repeating at regular intervals along the weld being tested, not acceptable even when each one stays within its size limit. Projected area: the projected area of the weld on the radiograph, used to calculate the porosity percentage; the percentage limits refer to this area.

The welder does not choose the level. Design and engineering set it in the WPS and the fabrication specification, according to the service condition and the type of loading. In steel structures this choice is not arbitrary: EN 1090-2 ties the execution class to the ISO 5817 level. The rough mapping is EXC1 → D, EXC2 → C, EXC3 → B, EXC4 → B plus additional requirements, but EN 1090-2 Table 17 gives the level row by row and allows a lower level for some imperfection types (undercut, excess weld metal and convexity, excess penetration can drop to D in EXC2). Always read the mapping from Table 17 of the edition you are using; never write the letter mapping into a report from memory.

For fatigue-loaded joints, Table 1 alone is not enough. ISO 5817:2023 gives the fatigue grading system in the informative Annex B: the columns C63, B90 and B125 correspond to FAT classes, and the designation takes the form ISO 5817-B90. Annex B does not apply to level D; B125 generally cannot be achieved in the as-welded condition, and fillet welds fall outside the scope of B125. There is no general EN 1090-2 clause saying "at least C for fatigue-loaded details"; the correct tools for fatigue are ISO 5817 Annex B and EN 1993-1-9.

One more critical bridge: an ISO 5817 quality level is not, in itself, an RT or UT acceptance criterion. Method and level selection runs through EN ISO 17635, radiographic acceptance through EN ISO 10675-1, ultrasonic acceptance through EN ISO 11666, phased array and automated UT application through EN ISO 13588 with acceptance in EN ISO 19285, penetrant and magnetic particle acceptance through EN ISO 23277 / EN ISO 23278, and visual testing through EN ISO 17637. Without that chain, the phrase "NDT surveillance" means nothing.

2. Porosity: gas that solidifies before it can escape

Porosity (group 200) forms when gas dissolved in the molten weld pool is trapped as the metal solidifies. Porosity is a volumetric discontinuity; its main effect is loss of effective cross section, and because its notch is blunt it is on its own less dangerous than planar imperfections (401, 402, cracks) and sharp surface notches such as undercut — note that undercut is not a planar discontinuity but a surface geometry notch, so do not lump the two together. That said, linear porosity (2014), clustered porosity (2013), elongated cavity and worm hole (2015 / 2016), surface pore (2017) and root porosity (516 — classified in group 500 in ISO 6520-1, yet a gas-driven root surface imperfection) are taken seriously. In every case, porosity signals that process control has been lost.

The root causes almost always point back to a gas source:

Moisture: moisture pick-up in a covered electrode or in submerged arc flux produces both porosity directly and diffusible hydrogen. In basic coverings and SAW flux, moisture routinely creates porosity through H₂/CO evolution during solidification; the same hydrogen causes hydrogen-induced cold cracking days later (see Section 8). These two outcomes are parallel — neither is secondary to the other. When you see porosity, question the hydrogen risk as well. For basic (low-hydrogen) electrodes, the common manufacturer and AWS A5.1/A5.5 ranges are 260–430 °C for 1–2 hours for re-drying and 120–150 °C in a holding oven; the binding documents are the manufacturer data sheet and the approved WPS — the figures above give the order of magnitude, not a site instruction. Permitted atmospheric exposure time varies with the electrode classification (AWS A5.1/A5.5 and the D1.1 exposure tables). The reverse error exists too: a basic electrode dried above the manufacturer's limit gives an unstable arc and more porosity. "The more you dry it, the better" is false.
Contamination and insufficient deoxidation: oil, paint, primer, rust, galvanized coating, cutting dross — all of them break down under heat and generate gas. A separate mechanism is CO porosity: if the pool lacks enough Si/Mn deoxidizer, welding over an oxidized or coated surface lets carbon combine with oxygen and evolve CO. The remedy is a wire or filler material with enough deoxidizer, backed by surface cleanliness. Clean the weld zone down to bright metal over an adequate width on both sides.
Loss of shielding: in GMAW and GTAW, low gas flow, wind or draught, a blocked nozzle or a leaking hose lets atmospheric nitrogen and oxygen into the pool. AWS D1.1 limits wind speed in the weld zone to about 5 mph (8 km/h) for gas-shielded processes (GMAW/GTAW/FCAW-G); above that, wind screens are mandatory (the clause number varies by edition, so confirm it in the edition you use).
Technique: an excessive arc length, a wrong torch angle and a very high travel speed also disturb the shielding. When gas escapes continuously and directionally, the pore does not stay spherical: it stretches with the solidification front and becomes a worm hole (2016).

Acceptance (ISO 5817:2023, Table 1):

2.3 — porosity area relative to the projected area (option a1): single pass D ≤ 2.5% / C ≤ 1.5% / B ≤ 1%; multi-pass ≤ 5% / 3% / 2%. Option a2 of the same row gives ≤ 2.5% / 1.5% / 1% relative to the fracture surface area, and applies only to production test pieces and qualification test specimens.
Single pore diameter: in butt welds d ≤ 0.4 s (max. 5 mm) / 0.3 s (max. 4 mm) / 0.2 s (max. 3 mm); in fillet welds d ≤ 0.4 aA (max. 5 mm) / 0.3 aA (max. 4 mm) / 0.2 aA (max. 3 mm).
2013 clustered porosity: cluster diameter dA ≤ 25 / 20 / 15 mm; systematic clustered porosity is not permitted at any level.
2014 linear porosity: h ≤ 0.4 s / 0.3 s / 0.2 s (max. 4 / 3 / 2 mm) and l ≤ s, max. 75 / 50 / 25 mm.
2.6 — 2015/2016 elongated cavity and worm hole: at D, h ≤ 0.4 s, max. 4 mm and l ≤ s, max. 75 mm; C and B tighten progressively.
1.18 — 516 root porosity: at D, acceptance depends on the application; not permitted at C and B.

3. Slag and solid inclusions: the price of poor interpass cleaning

Slag inclusion (301) occurs in slag-producing processes such as SMAW and SAW when a pass is deposited over slag left from the previous one. Slag is lighter than the metal and normally floats to the surface, but a convex bead profile, a narrow groove angle or a fast travel speed traps it in the side pockets. The same group covers tungsten inclusion (3041) and copper inclusion (3042).

These two metallic inclusions often get lumped together on site, yet their acceptance logic is exactly opposite. A tungsten inclusion is not an automatic rejection: in ISO 5817:2023, 3041 is judged by size under Table 1 row no. 2.10 "304 — metallic inclusion other than copper": h ≤ 0.4 s (max. 4 mm) / 0.3 s (max. 3 mm) / 0.2 s (max. 2 mm). It is an inclusion to avoid, but it is acceptable when it stays within the limit. (When evaluating by RT, remember that the separate provisions of EN ISO 10675-1 also apply.)

Copper inclusion deserves a specific warning. Copper transferred into the steel from a contact tip or a copper backing bar is not merely an inclusion — it initiates liquid metal embrittlement, or copper penetration cracking. In ISO 5817, 3042 is not permitted at D, C or B.

Preventive measures:

Remove the slag completely after every pass by grinding, wire brushing or chipping; pay particular attention to slag pockets at the weld toes.
Avoid convex, crowned passes; a slightly concave or flat profile lets the slag float out.
Set weaving and travel speed so that slag does not run ahead of the pool.
In GTAW, do not dip the tungsten into the pool. If you do, re-grind the electrode, remove the inclusion completely by grinding or milling, and verify with PT or RT where required. Carbon arc gouging is the wrong tool for this: it is far too aggressive and it risks carbon pick-up, especially in stainless steel.

4. Lack of fusion: heat was there, bonding was not

Lack of fusion (401) means that no metallurgical bond formed between the weld metal and the parent metal, or between passes. Mechanically it is extremely dangerous; it behaves like a sharp planar notch.

Get the radiographic physics right: a planar discontinuity produces contrast on the film only when the beam is aligned with the plane of the imperfection. A few degrees of misalignment are enough to miss it altogether. That is why angle-probe UT, and particularly phased array UT (PAUT), is preferred for sidewall lack of fusion; application follows EN ISO 13588, acceptance levels EN ISO 19285, and the method selection itself EN ISO 17635.

The root cause is nearly always insufficient heat input or wrong arc placement: low current with high travel speed; a torch angle that points the arc at the previous deposit rather than at the sidewall; poor pass sequencing and weaving that never reaches the wall; too low a voltage in GMAW short-circuit mode.

Prevention means staying inside the heat input range of the WPS, verifying sidewall melting visually — the welder should see the pool "licking" the groove face — and directing the arc into the fusion line with the correct work angle.

Acceptance (ISO 5817:2023, Table 1):

1.5 — surface lack of fusion (401): not permitted at D, C or B.
2.12 — internal lack of fusion (401/4011/4012/4013): accepted at D as a short imperfection — in butt welds h ≤ 0.4 s, max. 4 mm; in fillet welds h ≤ 0.4 a, max. 4 mm. Not permitted at C and B.
This row did not change in the 2014 → 2023 revision; only the explanatory figure changed. Believing that it "got stricter in 2023" leads an inspector to an unnecessary rejection or a wrong repair decision.
Micro lack of fusion (4014): accepted at D and C, not permitted at B.

5. Lack of penetration: the arc that never reaches the root

Two codes must be kept apart in the lack-of-penetration family. 4021 (incomplete root penetration) means the root area of a joint designed for full penetration was not fused to the intended depth. 402 (lack of penetration) is the general case where penetration falls short of the design depth, and it typically applies to partial penetration joints. In pipelines and pressure equipment, root imperfections translate directly into leakage and fracture initiation risk.

Root causes: too large a root face, too small a root gap or an insufficient groove angle; low heat input with high speed; an electrode too thick to enter a narrow root; magnetic arc blow.

Prevention means preparing the groove to the approved WPS, checking root gap and root face against tolerance before tacking, and holding the right current and speed on the root pass. (EN ISO 9692-1 gives recommended preparation geometries; the approved WPS is what binds.) On single-sided roots, two different tools must not be confused: purge gas prevents oxidation — on stainless and duplex work, verify the residual oxygen level with an analyser and take the threshold from the project specification — while ceramic or copper backing controls the root profile. If copper backing is used, keep the arc off the backing bar; copper pick-up means 3042.

Root protection also has an acceptance counterpart: in ISO 5817, 610 temper colours (visible oxide film) is a row of its own, and at all three levels it is left as "acceptance depends on the application (material, corrosion protection)". On stainless and duplex work this is the row that inspector and fabricator argue about most often; put the threshold — the colour chart and the accepted shade — into the project specification.

Other typical discontinuities of the same root area: 515 root concavity (suck-back), 516 root porosity, 509 sagging and 504 excess penetration. These tighten noticeably at C and B; 516 root porosity does not tighten — it is simply not permitted at C and B (at D, acceptance depends on the application).

Acceptance (ISO 5817:2023, Table 1):

1.6 — 4021 at the surface, only for single-sided full penetration butt welds: at D h ≤ 0.2 t, max. 2 mm (short imperfection); not permitted at C and B.
2.13 — 4021 internal, T-joints (fillet welds): at D h ≤ 0.2 a, max. 2 mm; not permitted at C and B. (In 2014 this T-joint row sat under code 402; in 2023 the code was moved to 4021 — a real change that supports the "mind the edition" argument.)
2.13 — 402 lack of penetration (partial penetration joints): at D h ≤ 0.2 s or 0.2 i, max. 2 mm; at C, h ≤ 0.1 s or 0.1 i, max. 1.5 mm; not permitted at B. Overlooking that 402 is still accepted at C within limits sends a perfectly sound weld into an unnecessary repair.

6. Undercut: the arc that eats the edge

Undercut (5011 continuous / 5012 intermittent) is a groove melted into the parent metal along the weld toe and not filled by weld metal. Because it concentrates stress, it is a crack initiation point in fatigue-loaded structures.

The root causes sit directly in technique: excessive current and arc length; high travel speed; a wrong work angle and no dwell at the edges of the weave; in the horizontal position, gravity pulling the pool downwards.

Acceptance (ISO 5817:2023, Table 1, no. 1.7):

t > 3 mm: D: h ≤ 0.2 t, max. 1 mm; C: h ≤ 0.1 t, max. 0.5 mm; B: h ≤ 0.05 t, max. 0.5 mm.
t = 0.5–3 mm: D: h ≤ 0.2 t; C: h ≤ 0.1 t; B: not permitted. (No absolute maximum is given in this band.)
In both bands the row requires a smooth transition, and imposes the short imperfection condition for D and C; undercut is not treated as a systematic imperfection in this row. Even when the measured h is within the limit, a groove with a sharp transition is not acceptable.

Prevention is a package, not a single fix: keep the current inside the WPS range, shorten the arc, reduce travel speed, pause briefly at the edges of the weave, and direct the filler material towards the toe.

7. Crater and crater cracks: the flaw of the last second

A crater pipe — 2025 (end crater pipe) when it stays open at the surface, 2024 (crater pipe) when it closes inside the weld metal — is a shrinkage cavity that forms when the arc is cut abruptly at the end of a pass and the pool solidifies before it fills. As the metal in that hollow contracts, the surrounding material has already solidified, so shrinkage stresses concentrate there and often produce a star-shaped crater crack (104). A crater crack belongs to the hot (solidification) crack family, and small as it is, it can propagate into the main weld.

Prevention is entirely in the welder's hands: bring the torch back and fill the crater before breaking the arc — on many GMAW and GTAW machines the "crater fill" function does this by ramping the current down — and extinguish the arc back inside the bead, not at its end point. (Backstep appears in this article as a bead sequencing technique in Section 9; do not confuse it with an arc extinction technique.)

Acceptance (ISO 5817:2023, Table 1): a crater pipe is not an automatic rejection; it is judged against numerical limits.

1.4 — 2025 surface end crater pipe: at D h ≤ 0.2 s (max. 2 mm) and d ≤ 0.3 s (max. 3 mm); at C h ≤ 0.1 s (max. 1 mm) and d ≤ 0.2 s (max. 2 mm); not permitted at B.
2.8 — 2024 internal crater pipe: at D h or l ≤ 0.2 t (max. 2 mm for t > 3 mm); not permitted at C and B.
Cracks (100) and crater cracks (104): not permitted at D, C or B. The only nuance is the micro crack (1001): accepted at D, while at C and B acceptance depends on an assessment of the parent metal's crack susceptibility. So even the "moderate" level D tolerates no crack — yet it accepts a crater pipe within numerical limits at D and C. Do not mix up these two rows.

8. Cracks (group 100): hot, cold, liquation, reheat and lamellar tearing

Cracks are the one imperfection family rejected at all three ISO 5817 levels. That is exactly why understanding the mechanism is not a bargaining position but an obligation.

Hot (solidification) cracking. The mechanism is not limited to the crater; it also appears in the body of the bead. Drivers: sulphur and phosphorus segregation; the centreline cracking tendency of narrow, deep beads with a depth-to-width ratio above 1; a shortage of delta ferrite in austenitic stainless steel. Prevention: parent and filler material low in S and P, a wider bead cross section that lowers the depth-to-width ratio, and filler material selected from the WRC-1992 (or Schaeffler) diagram targeting 3–10 FN delta ferrite in austenitic stainless steel. Fully austenitic stainless steels and Ni alloys are high-risk.

Liquation cracking. This is the form of hot cracking that appears in the HAZ: immediately beside the fusion boundary, low-melting-point grain boundary films partly liquate and separate under solidification shrinkage. It shows up especially in austenitic stainless steels, Ni-based alloys and high heat input multi-pass work. Prevention: limit heat input and interpass temperature, use parent metal with low S and P and controlled residual elements, and apply a thin-pass technique.

Ductility-dip cracking (DDC). In Ni-based welds this is a mechanism separate from solidification cracking: there is no melting. It occurs in the solid state, in a temperature range below the solidus, when grain boundary ductility collapses. It is the dominant damage mode in multi-pass, highly restrained Ni-based welds, and it is reduced by filler material chemistry that increases grain boundary tortuosity and by low-restraint design.

Hydrogen-induced cold (delayed) cracking. In C-Mn and low alloy steels this is the most expensive damage mechanism, and it can appear with a delay of typically up to 48 hours after the arc goes out; in high strength grades and thick sections the delay can reach 72 hours or more. That is why critical work specifies an NDT waiting period, and the duration comes from the applicable code (for example AWS D1.1 requires 48 hours for certain high strength steels; ASME B31.3 typically 24 hours; EN 1090-2 has its own clause on delayed testing after welding — read each from the relevant clause of the edition you use). Four conditions must coincide: (1) diffusible hydrogen, (2) a susceptible microstructure (hard HAZ, martensite), (3) tensile stress, and (4) a temperature below roughly 150 °C. Break any one of the four and the crack does not form; that is precisely the logic of prevention.

The tools: EN 1011-2:2001 Annex C (Method A and Method B) for preheat and interpass temperature calculation. Method A is built on the carbon equivalent CEV(IIW) = C + Mn/6 + (Cr+Mo+V)/5 + (Ni+Cu)/15 and is aimed mainly at C-Mn steels. Method B is built on CET = C + (Mn+Mo)/10 + (Cr+Cu)/20 + Ni/40 and is aimed mainly at fine-grained, high strength (TMCP/QT) low alloy steels. Method B is not a "thin section" method; the validity range given in EN 1011-2 Annex C, C.3 is d = 10–90 mm, CET 0.2–0.5%, HD 1–20 ml/100 g, heat input Q 0.5–4.0 kJ/mm, yield strength ≤ 1,000 N/mm², ISO/TR 15608 groups 1–4. Outside that range, fall back on the Method A / CEV approach; both methods carry limits on thickness, hydrogen and heat input.

On the diffusible hydrogen side, do not confuse two documents. EN ISO 3690 is only a measurement method and reports the result as HD in ml/100 g; the H5/H10/H15 class designations live in the consumable standards (EN ISO 2560, ISO 14341, ISO 17632, ISO 18275 and others), while EN 1011-2 additionally uses its own A–E hydrogen scale. For the HAZ hardness limit, work from the hardness table of EN ISO 15614-1: in the as-welded condition, 380 HV10 for steel groups 1–2 and 450 HV10 for group 3; after heat treatment, 320 and 380 HV10 respectively. (The standard draws its line between the heat-treated and the not heat-treated condition, not between quenched-and-tempered and other steels.) Interpass temperature has both a lower limit (cold cracking) and an upper limit (loss of toughness). Post-heating after welding can be applied to speed up hydrogen diffusion; it is not the same thing as PWHT.

Reheat (stress relief) cracking. Treating PWHT as a cure-all gets expensive on site. In Cr-Mo(-V) steels and stabilized austenitics (321/347), intragranular carbide precipitation in the coarse-grained HAZ hardens the grain interiors during PWHT, the relaxation of residual stress concentrates at the grain boundaries, and the crack is born during the heat treatment itself. So PWHT both removes hydrogen and lowers residual stress — but in these materials it can also be a source of cracking; the heat treatment cycle (ramp, temperature, soak) and the filler material selection must be assessed separately with that risk in mind.

Lamellar tearing. In thick-section T-joints and corner joints, under tensile loading in the through-thickness (Z) direction, tearing propagates along planes of aligned MnS inclusions. A code warning: this phenomenon has no reference number of its own in the main ISO 6520-1 classification table. It is treated as a cracking phenomenon and evaluated within group 100 according to its morphology. (A common error is to tag it as 105; in ISO 6520-1, 105 = group of unconnected cracks and 106 = branching crack.) Prevention: Z-quality steel to EN 10164 (Z15/Z25/Z35), joint geometry modified to reduce through-thickness tension, buttering, and parent metal with low sulphur.

9. Spatter, arc strikes, distortion and shape imperfections

Spatter (602; tungsten spatter 6021): droplets thrown from the arc pool that stick to the surface. Calling them harmless is wrong: in ISO 5817 Table 1, spatter is left as "acceptance depends on the application (material, corrosion protection)". For fatigue-loaded steel welds, informative Annex B, Table B.1, no. 1.23 applies: spatter is not permitted only in the B125 (FAT 125) column; in the C63 and B90 columns, the "depends on the application" provision of Table 1 governs, and Annex B does not apply to level D. On stainless steel, spatter is a corrosion initiation point. Root causes: unbalanced voltage and inductance, a long arc, dirty or moist gas, the wrong GMAW transfer mode. Prevention: set the parameters to match the transfer mode, choose a suitable gas mixture (the CO₂ content, for example), and use clean wire at the correct contact-tip-to-work distance.

Arc strikes (601): the discontinuity that looks trivial and yet gets rejected more often than any other in the field. An arc struck uncontrolled on the parent metal cools a very small volume very fast; the result is local martensite and a crack initiation site. In ISO 5817 it is not permitted at C and B, and at D only if the material properties are not affected. In practice: grind the mark away, check the loss of thickness, and re-test the area with MT or PT.

Distortion: the shrinkage and angular deformation caused by uneven heating and cooling. It is not a discontinuity but a dimensional imperfection — and one of the most expensive problems at assembly. Prevention: the minimum necessary heat input and weld size, a balanced and symmetrical welding sequence (backstep, skip or block sequencing), solid restraint and pre-setting, and avoiding oversized welds.

The shape imperfections in the same group, with their codes: 502 excess weld metal (defined for butt welds only), 503 excessive convexity (defined for fillet welds only), 504 excess penetration, 509 sagging, 511 incompletely filled groove, 512 asymmetric or incorrect leg length, 507 linear misalignment. Note that 504 and 509 are separate codes; merging them into "504 excess penetration / root sagging" is a reporting error. All of them are controlled through the WPS profile, fit-up tolerances and disciplined technique.

10. The root cause chain: parameters, technique, preparation, material

Put all the imperfection families above into one frame and four root-cause families emerge:

Heat input (parameters): too low, and you risk lack of fusion, lack of penetration and cold cracking; too high, and you get distortion, undercut, liquation cracking and — most critical of all — grain coarsening in the HAZ, loss of low temperature toughness (CVN), HAZ softening in TMCP/QT steels, and a disturbed ferrite/austenite balance in duplex stainless steel. That is why the WPS states heat input with both a lower and an upper limit. The WPS itself is defined to EN ISO 15609-1; the most common route is qualification by test (EN ISO 15614-1, with ISO 15613 for pre-production testing), but where the contract allows, ISO 15610 (approved consumables), 15611 (previous welding experience) and 15612 (standard procedure) are also valid routes.
Technique (welder skill): arc length, work and travel angles, weaving, crater filling, interpass cleaning. The certificate is EN ISO 9606-1; for fully mechanized and automatic welding, operator and setter qualification comes under EN ISO 14732 — in a shop with a robot or a SAW station, 9606 alone is not enough.
Preparation: groove geometry (EN ISO 9692-1 recommendations, with the approved WPS binding), root gap, surface cleanliness, electrode conditioning. The great majority of discontinuities are decided here, before the arc is struck.
Material and consumables: cracking susceptibility via CEV/CET, filler material compatibility, gas purity, moisture and hydrogen class.

The role of the quality management system (EN ISO 3834-2/-3/-4, with ISO 3834-1 governing which part to select) is precisely to make those four families documentable: the right WPS, a qualified welder, controlled consumables, recorded testing.

From the field

Here is a pattern I have watched repeat over the years: a porosity report lands, everyone checks the gas flow, and the real culprit turns out to be the basic electrode that arrived on site that morning in an open box. In one pipe spool shop we were getting root porosity bead after bead, with parameters that looked flawless on paper. Then we opened the electrode oven and found a dead thermostat. Everyone had assumed those electrodes were conditioned; in fact they had been quietly picking up moisture. Days of rejections traced back to one broken part. And porosity was only what we could see in that batch. The real risk was hidden: the delayed root cracking that the same hydrogen could cause days later. An inspector who closes a moisture report with porosity has read only half the story.

The second lesson repeats just as often: undercut and lack of fusion almost always come down to speed. In shops on piece rate, or under delivery pressure, welders creep the travel speed up without noticing. The dwell at the edges disappears and sidewall fusion weakens with it. Spend ten seconds looking at the bead profile and the weld toe with the naked eye before radiography — in other words, take the discipline of EN ISO 17637 seriously — and you will see half of what the film is about to show you. NDT finds an imperfection; profile control and parameter discipline prevent it. A prevented imperfection always costs less than a found one.

Related standards

As in any report that cites numerical limits and codes, the edition year binds. Where no year is given below, work from the edition cited in your contract.

EN ISO 6520-1:2007 — Classification of geometric imperfections in metallic materials — Part 1: Fusion welding.
EN ISO 5817:2023 (4th edition) — Fusion-welded joints in steel, nickel, titanium and their alloys (beam welding excluded) — quality levels (B/C/D) and the informative Annex B for fatigue (C63/B90/B125).
EN ISO 10042 (the edition you use) — Arc-welded joints in aluminium and its alloys — quality levels.
EN ISO 13919-1 / -2 — Quality levels for beam-welded joints; the split follows the material, not the beam type: -1 steel, nickel, titanium and their alloys, -2 aluminium, magnesium and their alloys. Both parts cover electron and laser beams.
EN 1090-2 (the edition you use, Table 17) — Execution of steel structures; execution class ↔ ISO 5817 level mapping, given row by row.
EN 1993-1-9 — Fatigue in steel structures; read together with ISO 5817 Annex B.
EN ISO 15609-1 / EN ISO 15614-1 (+ ISO 15610/15611/15612/15613) — WPS definition and procedure qualification routes.
EN ISO 9606-1 / EN ISO 14732 — Welder qualification; operator and setter qualification.
EN ISO 3834-1/-2/-3/-4 — Quality requirements for fusion welding and selection of the appropriate part.
EN ISO 9692-1 — Recommended joint preparation geometries for arc welding of steels.
EN 1011-2:2001 — Arc welding of ferritic steels; preheat (Annex C, Method A: CEV / Method B: CET), cold cracking control, the A–E hydrogen scale.
EN ISO 3690 — Determination of diffusible hydrogen in weld metal (a measurement method only; the H5/H10/H15 classes belong to consumable standards such as EN ISO 2560, ISO 14341, ISO 17632 and ISO 18275).
EN 10164 — Through-thickness properties (Z15/Z25/Z35) — against lamellar tearing.
EN ISO 17635 / 17637 / 17640 / 17636-1/-2 — NDT method and level selection, visual testing, UT, RT.
EN ISO 13588 / EN ISO 19285 — Automated and phased array ultrasonic testing (PAUT) application and acceptance levels.
EN ISO 10675-1 / 11666 / 23277 / 23278 — RT, UT, PT and MT acceptance levels.
EN ISO 2553 — Welding symbols; showing joint preparation and testing requirements on the drawing.
ASME BPVC Section IX (the edition you use) — Procedure and personnel qualification. Note: Section IX contains no acceptance criteria; the discontinuity limits sit in the construction code (for example Section VIII Div. 1 UW-51/UW-52, ASME B31.3 Table 341.3.2).
AWS D1.1 (the edition you use) — Structural welding code for steel; acceptance criteria, wind limits, electrode exposure tables, NDT waiting periods for delayed cracking.

Take it to the field: To keep the welding standards, procedure logic and field control steps in this article in your pocket — completely offline and free — take a look at the Doawise Welding Eng Guide app.

At DoaWise we carry out welding engineering and inspection services, from WPS/WPQR preparation to NDT surveillance, to international standards (EN ISO and ASME/AWS). We prevent discontinuities before they form on site, and when they do form we classify them in the language of EN ISO 6520-1 and evaluate them against the EN ISO 5817 quality level and the relevant NDT acceptance standard — producing recordable, auditable results.