Welding Flaws: The Complete Guide to Causes, Detection, Codes, and Prevention

Table of Contents

Welding Flaws vs. Discontinuities: What Actually Counts as a Defect

You strike the arc and lay down a bead that looks textbook-perfect. The toes are clean, the ripple pattern even, and nothing tells you to reach for a grinder. Then the radiograph comes back with dark spots you never could have seen from outside. The weld that looked good, wasn’t.

If that sting feels familiar, you have already met the central truth of weld quality: appearance and integrity are two different things, a lesson Good and Bad MIG Welds drives home repeatedly. Closing the gap starts with vocabulary, because in an inspection report a casual word like “flaw” can decide whether a joint is accepted or torn out.

Discontinuity vs. Defect: The Line That Decides Everything

A discontinuity is any interruption of the typical structure of a material, the exact wording in AWS D1.1/D1.1M:2020, which defines a discontinuity as “an interruption of the typical structure of a material” and notes a discontinuity is not necessarily a defect. A defect, by contrast, is a discontinuity that exceeds the acceptance limits written into the applicable code and is therefore rejectable.

Porosity is always a discontinuity. Whether it becomes a defect depends on how much there is, where it sits, and what the governing standard allows. The distinction is not about presence. It is about permission.

Quick Note: A discontinuity is a fact. A defect is a judgment made against a code limit. Two identical welds can land on opposite sides of that line under different acceptance criteria.

Why “Flaw” Is a Catch-All — and Why Words Matter

“Flaw” is the everyday umbrella term covering everything from a harmless surface ripple to a crack running through a pressure boundary. It is useful in conversation and dangerous in documentation. An inspection report that labels something a “flaw” without type, size, location, and code reference tells a welder almost nothing about what to fix.

The most practical way to organize imperfections is by family. ISO 6520-1 groups fusion-weld imperfections into six families: cracks, gas cavities, solid inclusions, lack of fusion and penetration, imperfect shape and dimensions, and miscellaneous imperfections. Those six buckets frame the technical sections ahead.

Severity Depends on Type, Size, Location, and Loading

Not every discontinuity threatens a joint equally. A small buried pore in a low-stress, statically loaded member may never matter. The same pore at a fatigue-critical detail under cyclic load is a different story. Inspectors weigh service conditions, not just size.

Failure data supports that emphasis. An HSE review of 47 failure cases (RR1215) found crack-like or volumetric flaws in 49% of cases and poor or undersized weld geometry in 45%. Geometry and fusion, not exotic surprises, dominate.

Where do these problems start? ASME classifies the causes of welding defects as 41% poor process conditions, 32% operator error, 12% wrong technique, 10% incorrect consumables, and 5% bad weld grooves. Almost none of that is mysterious, and most is controllable through disciplined Welding Practice.

What’s Ahead in This Guide

From here the guide moves in a deliberate order: first the code framework that governs acceptance, then a close look at each failure family, from fusion and penetration flaws to porosity, cracks, inclusions, profile problems, and distortion. The inspection methods that find these flaws come next, followed by root causes, prevention, and repair decisions.

Why This Matters on the Floor

For fabricators, welders, and engineers, this vocabulary is not academic. A fabricator who understands acceptance criteria can question an unclear rejection instead of absorbing unnecessary rework. A welder who knows the difference between a discontinuity and a defect can fix the cause rather than chase the symptom. An engineer who can read an inspection report knows when to accept, when to repair, and when to re-engineer the joint. That shared language is where good welds begin.

The Code Framework: ISO 6520, ISO 5817, AWS D1.1, and ASME Section IX

Here is the uncomfortable truth about welding flaws: none are rejectable on their own. A flaw becomes a defect only when a code says so. That distinction sits at the heart of every inspection report, and it is why two welders can argue for an hour about the same undercut and both be technically right. Before you judge any discontinuity, you need to know which standard governs the job. The ecosystem splits into two jobs: classifying flaws, and setting the limits that decide acceptance.

ISO 6520: The Classification System

ISO 6520-1 is the dictionary. It defines and numbers every known geometric imperfection in metallic welding, grouping them into categories such as cracks, cavities, solid inclusions, lack of fusion, and shape imperfections. Those numbers are international shorthand. When an inspector references “100” for a crack or “2014” for a blowhole, they are speaking ISO 6520. The standard itself doesn’t set limits; it just gives everyone a shared vocabulary.

ISO 5817: Quality Levels B, C, and D

Where 6520 says what a flaw is, ISO 5817 says how much is allowed. The current edition, ISO 5817:2023, covers fusion-welded joints in steel, nickel, titanium, and their alloys and assigns three quality levels designated B (highest), C, and D. It applies to material thicknesses of 0.5 mm and above and covers fully penetrated butt welds and all fillet welds. Level B is strict, D is the most permissive, and the choice is normally driven by what the joint must survive in service.

AWS D1.1 and D14.4: Structural Steel and Machinery

In North America, AWS D1.1 governs structural steel. It bundles classification, workmanship limits, and inspection into one document, and it defines acceptance by discontinuity type, size, and location. D1.1 is also specific about timing: visual inspection of steels may begin once completed welds have cooled to ambient temperature, not while they are still hot. For machinery and equipment rather than structures, AWS D14.4/D14.4M:2019 sets its own acceptance criteria, NDT requirements, and weld defect repair rules. Same weld, different rulebook. Both codes assume a properly qualified welder — the limits only mean something if the person running the arc is competent.

ASME Section IX: Qualifying Procedure and Personnel

ASME Section IX is frequently misread as an acceptance standard. It is not. Section IX qualifies welding procedures and welders; it proves a method can produce sound welds. The mechanics are concrete: QW-150 tensile tests require weld tensile strength above 58 ksi for an A36 baseline, and QW-160 guided-bend tests allow no open discontinuity in the weld or HAZ greater than 1/8 in. A qualified procedure doesn’t certify a specific weld — it confirms the process, materials, and parameters can meet the required properties, which is why procedure and welder are qualified separately. Turning those numbers into consistent, repeatable results comes down to deliberate welding practice.

Workmanship vs. Fitness-for-Service

Two philosophies decide how strict those limits are. Workmanship criteria set fixed, geometry-based limits — a maximum undercut depth or porosity count — regardless of stress. Fitness-for-service criteria, such as those in API 579 and BS 7910, instead ask whether an actual flaw will grow to failure under real loading. FFS is common in pressure equipment and aging assets; workmanship rules dominate everyday fabrication.

Because the codes differ, acceptance is code-relative: the same flaw can pass one standard and fail another. A little porosity that ISO 5817 level D tolerates might be rejectable under a stricter D1.1 clause or a contract calling for level B. A bead can look perfect — see any good and bad MIG welds guide — and still break a code limit you cannot see. That is why the first question in any flaw debate is never “how big is it?” but “what does the code on this job actually allow?”

Quick Note: Never accept or reject a weld on looks alone. The governing code — not the inspector’s opinion — defines the limit. Confirm the standard before the first arc, not after it.

Fusion and Penetration Flaws: Lack of Fusion, Incomplete Penetration, and Overlap

A weld can look perfectly dressed on the outside and still fail the one test that matters: did the metal actually fuse? This family of flaws — lack of fusion, incomplete joint penetration, and overlap — all break the same rule. The molten weld metal never formed a metallurgical bond with the base metal or the previous pass, so the two pieces sit side by side instead of becoming one.

Lack of Fusion: The Bond That Never Formed

Incomplete fusion is characterized by a lack of fusion between the base metal and the weld metal. It shows up in three places: sidewall fusion, where the bead fails to tie into the joint face; interpass fusion, where new metal won’t bond to the pass beneath it; and root fusion, where the first pass doesn’t melt into the root.

The physics come down to heat and arc placement. If the arc doesn’t melt the base metal, no bond can form — the filler simply freezes against cold steel. That happens when heat input is too low, travel speed runs ahead of the puddle, the electrode angle points the arc away from the sidewall, or a contamination film (oil, mill scale, oxide) sits between the metals. Magnetic arc blow can also push the arc off course on DC work.

AWS Inspection Trends reports that for large single-pass fillet welds the biggest problems are incomplete fusion at the root and slag inclusions. That’s telling: the flaws hide exactly where you can’t see them.

Incomplete Joint Penetration

Incomplete joint penetration is the cousin that lives at the root. The weld metal never extends through the full joint thickness. In a double-groove weld it’s typically caused either by not back gouging deep enough or by a back-gouge geometry with insufficient depth-to-width ratio for adequate access.

Root causes stack up fast: too-small a root gap, a thick root face, low current, a fast travel speed, or a joint design that simply leaves no room for the arc to reach bottom. Poor fit-up and poor access are the quiet killers.

Overlap: Weld Metal Without a Bond

Overlap is a buildup of weld metal on top of the base metal without fusion. The bead rolls over the toe and looks like it’s sitting on the plate, but there’s no tie-in underneath. It usually traces to travel speed too fast for the current, an incorrect electrode angle, or a weaving technique that lets the puddle run ahead of the arc. For a deeper dive, see our guide to the overlap welding defect, and the difference between overlapping as a technique vs. a defect.

Quick Note: A fusion flaw that passes visual inspection is not a pass. If the arc never melted the base metal, no amount of good bead appearance will create a bond.

Why Planar Flaws Are More Dangerous

Porosity and slag are volumetric — rounded or blobby — and they interrupt the section gently. Lack of fusion, incomplete penetration, and overlap are planar: sharp, crack-like, with a tiny tip radius. Under tensile load the stress concentrates at that tip, and the flaw can propagate like a crack. That’s why codes treat them as the most serious class of defect.

Diagnosis and Remedies

These flaws resist visual inspection because they sit inside the joint or tuck under the toe. That’s why ultrasonic testing and phased array, which find planar defects, matter more here than radiography. Once found, the fix is usually straightforward — though repairs add real cost and rework.

Increase heat input and reduce travel speed so the sidewall actually melts.
Remove the defective portion and re-weld with improved fit-up and access.
Correct the electrode angle and control arc blow.
Clean oil, scale, and oxide films before striking the arc.
Pull parameters from a qualified WPS instead of guessing at the machine.

Getting the current right is half the battle. Our 3/16 E7018 amperage guide shows how the wrong setting produces exactly these cold, non-fusing beads.

Porosity and Gas-Related Flaws: Pinholes, Blowholes, Wormholes, and Piping

Porosity is simply gas that was trapped in weld metal as it solidified. Something released gas into the molten pool, the pool froze before the bubbles could rise out, and what remains is a hollow void. Unlike a crack, porosity rarely announces itself. It shows up on a radiograph, or you discover it when a leak test fails.

Because the shape and distribution of those voids point straight back to a cause, it pays to know the vocabulary.

The Forms Porosity Takes

Scattered porosity: individual pores spread randomly across the weld. Usually contamination entering the arc.
Clustered porosity: a concentrated group of pores in one small region.
Linear porosity: pores aligned in a line, often tracking the root or a fusion boundary.
Wormholes: elongated, sometimes interconnected channels that can run deep into the weld.
Blowholes: larger, more rounded cavities.
Piping porosity: elongated pores, frequently found at the weld surface.

Elongated piping porosity is a special case worth remembering. It is confined to the face reinforcement of groove welds or the convex face of fillet welds, and on many jobs it is a direct signal that the electrode extension is too short. AWS guidance suggests that increasing electrode extension, which lowers amperage, and slightly reducing wire feed speed will eliminate it.

Which Gas Is Actually Trapping Itself?

The culprit changes with the material. In aluminium and stainless steel, hydrogen dominates, largely because those processes are so sensitive to moisture. In carbon steel, hydrogen, nitrogen and oxygen all contribute, with carbon monoxide from deoxidation reactions adding to the mix. Nitrogen tends to be a problem when the arc is exposed to air, which is why gas coverage matters so much, while oxygen shows up as oxide inclusions or as carbon monoxide when it meets carbon in the pool.

Hydrogen’s favourite route into the pool is moisture, and moisture hides in rust, mill scale, oil, grease, paint and condensation. In shielded metal arc welding, wet or improperly stored E7018 electrodes that were never held in a holding oven are a classic source. Long arc length, and skimped restarts where the puddle is not properly re-established, only add to it.

Beyond contamination, the shielding side of the process has its own failure points:

Wrong shielding gas or an incorrect flow rate
Leaks in the gas line, regulator or torch body
Turbulent flow that pulls air into the gas stream
Excessive arc length that lets the atmosphere reach the pool
Damp flux or wire in flux-cored and submerged arc welding

A worn or wrong-sized contact tip can disturb gas coverage just as badly, which is why contact tip troubleshooting is worth treating as routine maintenance. If you run gasless MIG with flux-cored wire, remember that the flux generates its own shielding, so a damp or damaged wire is a direct porosity risk. Piping porosity that runs the length of the weld face is a giveaway that the arc was stretched and the gas shielding broke down.

Why It Matters Mechanically

Every pore removes load-bearing metal, so the effective cross-section shrinks and the joint carries less load than the drawing assumes. Toughness suffers too. In one published study, the average Charpy impact toughness of welds with porosity defects fell by about 38% in the heat-affected zone compared with defect-free welds. On thin sections the loss can be enough to fail a bend or tensile test.

And because connected pores form continuous paths through the metal, porosity is one of the most common reasons a leak or pressure test fails. For pressure-retaining work, that leak path can matter more than the strength loss. Pores also act as stress raisers, and clusters near the surface can seed fatigue cracks under cyclic loading.

Quick Note: Porosity is usually a process-hygiene problem, not a technique problem. Clean the joint, dry the consumables, check the gas path, and most of it disappears. The fix is almost always the same: clean metal, dry consumables, and stable gas flow.

Cracks: Hot Cracking, Hydrogen-Induced Cold Cracking, Crater Cracks, and Lamellar Tearing

Cracks are the flaw family welders fear most. Unlike porosity or undercut, a crack is a sharp, stress-concentrating discontinuity that can propagate with almost no warning. A cracked weld can fail in service.

Cracks fall into two groups: hot cracks that open as the metal solidifies, and cold cracks that appear hours later.

Hot Cracking (Solidification Cracking)

Hot cracks open at temperatures near the solidus, while the last liquid films between growing grains are still present. As the weld shrinks, those films tear and cannot heal. Hot cracking is most associated with austenitic stainless steels, fully austenitic weld metal, and deep, narrow weld pools.

The standard resistance mechanism is chemistry, not technique. Controlled delta-ferrite content in austenitic stainless weld metal disrupts the continuous low-melting grain-boundary films that cause cracking. That is why 308L and 316L fillers are formulated to leave a few percent ferrite, while fully austenitic grades like 310 are trickier to weld. Deep, narrow pools also worsen segregation, so wider beads and better fit-up help.

Hydrogen-Induced Cold Cracking

Cold cracking — also called delayed cracking or hydrogen cracking — needs three things at once: hydrogen, a hard microstructure, and tensile stress from restraint. Remove any one and the crack cannot form. This is why it appears in the heat-affected zone of hardenable steels and in high-carbon, high-strength weld metal.

The word “delayed” is critical. Hydrogen diffuses slowly through the lattice, so cracks may not appear until hours after welding. TWI’s guidance on hydrogen cracking is blunt: inspection should not be carried out until an appropriate delay time has passed after welding is completed. Test too early and you sign off a weld that cracks overnight.

Quick Note: If a weld passes inspection in the morning and fails by the afternoon, suspect hydrogen. Delay time matters as much as the NDT method.

The money at stake is enormous. A 1973 estimate put UK industry costs from welding-related manufacturing problems at £260 million annually, with at least £40 million arising from repairing hydrogen-induced cracks and £140 million from service failure due to fatigue and brittle fracture, as reported in TWI’s repair statistics. EWI/PRCI work on cellulosic electrodes documented two cases requiring removal of many pipeline welds at a cost of over $4 million, and an offshore lay barge project where repair costs exceeded $10 million. In those cracked cellulosic welds, measured carbon equivalents reached up to 0.66 wt%, well above expectations, alongside weld carbon above 0.20 wt%.

Crater Cracks, Toe Cracks, and Lamellar Tearing

Cracks come in telling orientations and locations:

Crater cracks form at the end of a bead where the welder stops abruptly, leaving a shallow, star-shaped shrinkage void that often seeds larger cracks.
Transverse cracks run across the bead, usually in high-restraint, high-carbon joints — a hallmark of hydrogen cracking.
Longitudinal cracks run along the weld axis, common in hot cracking and centerline segregation.
Toe cracks sit in the heat-affected zone at the weld toe, where restraint and a hard HAZ meet the stress concentration of the profile.
Lamellar tearing is a step-like crack in the base plate, parallel to the rolled surface under the weld. It happens when through-thickness stress meets inclusions in plate with poor Z-direction ductility, common in thick, restrained T-joints.

The hardened steel welding guide explains the HAZ hardness and preheat logic behind many of these, which pairs well with this section: Can You Weld Hardened Steel? What Works & What Doesn’t.

Prevention: The Three Enemies of Cracking

Prevention targets hydrogen, hardness, and restraint together.

Preheat slows cooling, softening the HAZ and giving hydrogen time to escape.
Low-hydrogen consumables — properly dried E7018 or flux-cored wires — keep diffusible hydrogen low. Electrodes left in a damp shop reabsorb moisture and undo the benefit.
Controlled interpass temperature keeps heat input in the right band: enough to soften, not so much that you damage toughness.
Carbon equivalent control limits hardenability. Lower CE steels tolerate faster cooling without forming martensite.
Reduced restraint through good fit-up, balanced sequences, and letting joints move helps stress relax. Steady welding practice also limits stress-raising profile flaws.

Get these right and cracks become rare rather than routine. Skip them and, as the cost figures show, you will pay for it later — often after the weld has passed visual inspection.

Solid Inclusions and Profile Flaws: Slag, Undercut, Insufficient Throat, and Oversized Beads

Some welding flaws are hidden events trapped inside the metal. Others are right there in the bead shape, visible to anyone holding a fillet gauge. This section covers both: the solid, non-metallic material that gets frozen into the weld, and the geometric flaws that quietly rob a joint of its strength.

Solid Inclusions: Slag, Flux, Oxide, and Tungsten

A solid inclusion is any foreign material that ends up locked inside the weld metal. The usual suspects are slag, flux, oxides, and, in TIG work, fragments of tungsten electrode.

Slag inclusions occur most often in flux-based processes such as SMAW, FCAW, and SAW. The most common cause is simple and entirely avoidable: failure to completely remove slag from one pass before making another. The molten pool simply flows over the leftover slag and seals it in. Isolated inclusions tend to appear when rust or mill scale is present on the base metal, and sharp notches in joint boundaries or between weld passes act as perfect traps that promote slag entrapment.

On a radiograph, trapped slag shows up as dark lines or irregular patches. The telltale sign is a pair of parallel “wagon track” lines at the root-pass toes, which is slag sitting exactly where the root fused into the sidewall.

Quick Note: Good interpass cleaning is not optional busywork. On multi-pass flux welds, a wire brush and chipping hammer between every pass is the single cheapest defect-prevention tool in the shop.

Why Flux Processes Trap Slag

Any process that generates a slag layer is a candidate for entrapment. That is why SMAW, FCAW, and SAW demand deliberate technique. Poor groove access, a bead sequence that stacks weld metal into a corner it cannot reach, and running current too low all make it harder for slag to float out to the surface where it can be removed. High-deposition automated flux welding magnifies the problem, since a deep SAW bead can bury slag the operator never sees. For context on managing this at scale, see our guide to robotic submerged arc welding. Fit-up matters just as much: tight gaps and awkward geometry limit access and trap slag, which is a constant theme in weld tube fittings work.

Profile Flaws: Undercut and Insufficient Throat

Profile flaws reduce the amount of sound metal carrying the load.

Undercut — a groove melted into the base metal along the toe of the weld that is left unfilled. It creates a sharp notch that concentrates stress and acts as a corrosion initiation site, especially in cyclic or wet service.
Insufficient leg or throat size — the weld is simply too small for the design load, so it fails below its rated capacity.
Excessive convexity — a high, humped bead hides lack of fusion at the toes and wastes filler without adding usable throat.
Poor bead profile and uneven cap — irregular ripples and an inconsistent cap signal unstable travel speed, wrong angle, or current swing.

Undercut deserves special attention because it is so common. Per AWS CWI classification it ranks among the most frequently cited visual weld defects, alongside overlap, porosity, cracks, and arc strikes. Correct technique, proper current, and a steady travel speed prevent most of it.

What the Code Actually Allows

Here is a point that surprises many welders: undercut and slag inclusions are frequently tolerated up to defined limits in a code, unlike cracks and lack of fusion, which are generally rejectable outright. Acceptance criteria such as those in ISO 5817 set specific thresholds for depth, length, and frequency. The metalworking failures documented in HSE research often trace back to defects that slipped past those limits.

That distinction matters for repair decisions. A small, code-compliant inclusion may be acceptable as-is; a crack of any size is not. Reading the acceptance criteria before reaching for the grinder saves time and prevents unnecessary rework.

Distortion, Residual Stress, Arc Strikes, and Spatter

Not every defective weld hides inside the metal. Some of the most expensive flaws sit right on the surface, or warp a finished assembly so far out of tolerance that it never fits. Distortion, residual stress, arc strikes, and spatter are all classified as miscellaneous imperfections under ISO 6520-1, which is the point: they are geometric and metallurgical imperfections rather than internal cavities. Mechanical damage — dents, gouges, and chipped edges from handling — belongs in the same family, because it removes sound metal and leaves a stress riser behind.

Distortion and Residual Stress

When you weld, you heat a small zone far hotter than the surrounding plate. That hot metal expands, but the cold metal around it holds it back. As the weld cools it tries to shrink, and again the restraint resists. The result is locked-in residual stress, and often visible distortion: bowing, angular change, buckling, or a panel that oil-cans between its stiffeners.

The fix is not simply turning the amperage down. Distortion is controlled by balanced welding sequences, correct tacking, restraint, and managing heat input overall rather than lowering amperage alone. Techniques that spread heat evenly include:

Backstep welding — laying short beads against the direction of travel so each cools into a partly finished weld.
Skip welding — jumping around the joint to let sections cool.
Balanced sequences — welding opposing sides alternately, or from the centre outward.
Strong tacking and fixturing — holding parts until the weld is complete.

Post-weld straightening, whether by flame, press, or peening, can bring parts back, but it works on the symptom. The better play is controlling heat before the arc ever starts. A well-planned bench and jig setup makes this far easier, as covered in our guide to planning and setting up a welding shop.

Quick Note: Residual stress is not just a dimensional problem. Restrained shrinkage can drive hydrogen cracking when it combines with hard microstructures and diffusible hydrogen, a mechanism seen in failures involving chromium-plated steel with a carbon equivalent of 0.6 — documented in TWI’s hydrogen cracking analysis.

Arc Strikes

An arc strike is where the electrode touches the work outside the joint and leaves a small mark. It looks trivial, but the arc quenches that spot almost instantly into a hard, brittle microstructure, and it creates a sharp notch that concentrates stress. On thick or high-strength steel an arc strike can be the initiation point for a crack. Even a small strike on a fatigue-loaded member deserves inspection, since cracks often begin exactly there. Treat stray strikes as discontinuities: grind them out, inspect, and where the code requires, re-weld.

Spatter

Spatter is mostly cosmetic, but not always. On stainless and corrosion-resistant work, globules can trap contaminants and become corrosion sites, and many specifications limit it. It also fuses to fixtures and threads. Anti-spatter, correct polarity, and clean parameters cut it down — the same discipline that keeps your PPE and sightlines intact, as our protective eyewear guide explains.

The Downstream Cost

The consequences of these flaws compound. HSE lists the net effects of poor weld quality as fatigue, corrosion, fracture, and yielding. Residual stress lowers fatigue life, a hard arc strike feeds brittle fracture, and spatter on stainless promotes corrosion. None of them show up on a radiograph, which is exactly why standards treat them seriously even when a weld passes every volumetric test.

Detection and Inspection: Visual, PT, MT, RT, UT, and Phased Array

Every flaw you cannot see is a flaw that has to be found another way. Inspection methods fall into two broad camps: those that look at the surface and those that see inside the metal. Choosing the right one comes down to the flaw type, the material, and the joint geometry.

Visual Inspection: The Cheapest First Line

Roughly 80 percent of weld quality problems are visible to a trained eye before any expensive test gear comes out. Visual inspection catches undercut, overlap, spatter, arc strikes, and gross profile problems, and it costs almost nothing beyond time and a good light. Before you reach for instruments, learn what a sound bead actually looks like in your process. Our guide to good and bad MIG welds breaks down the surface cues, and deliberate welding practice is what makes those cues second nature.

Surface Methods: PT and MT

Liquid penetrant testing (PT) is a low-cost method widely applied to locate surface-breaking defects in non-porous materials. It works on ferrous and non-ferrous metals alike, making it useful on aluminium and stainless where magnetic methods fail. A coloured or fluorescent dye seeps into a crack or pinhole, then bleeds back out to reveal its shape.

Magnetic particle inspection (MT) is restricted to ferromagnetic materials but reaches further: it can reveal slag inclusions and cracks just below the surface. Push deeper than that and you need radiographic or ultrasonic methods.

Volumetric Methods: Radiography and Ultrasonics

Radiographic testing (RT) passes X-rays or gamma rays through the weld onto film or a digital detector. It excels at volumetric defects such as porosity and slag because those change the path length of the beam and show as darker or lighter regions.

Here is the limitation every inspector must respect: a lack-of-fusion plane running parallel to the radiographic beam may not register at all. The beam passes straight along the gap with almost no attenuation change. Ultrasonic testing (UT) sends high-frequency sound into the weld instead and detects that planar, crack-like geometry far better, because a flat reflector oriented across the beam returns a strong echo. For critical work, UT and phased array are the alternatives when RT comes back clean but suspicion remains.

Quick Note: A passing radiograph is not proof of a sound weld. Planar flaws aligned with the beam are RT’s blind spot, so specify ultrasonic or phased-array testing when lack of fusion or cracks are the concern.

Phased Array and Modern Methods

Phased-array ultrasonic testing steers and focuses the sound beam electronically, producing a cross-sectional image of the weld. It maps flaw location, depth, and orientation far more precisely than a single-probe UT scan, and it is increasingly used on pipeline girth welds and thick-section work. Emerging in-situ sensors and automated monitoring now watch arc parameters in real time, flagging drift before a flaw is ever deposited.

Timing, Coverage, and Documentation

When you inspect matters as much as how. Hydrogen-induced cold cracking can appear hours or days after the weld cools, so final inspection on susceptible steels should wait, often 24 to 48 hours after completion. Inspecting too early is how delayed cracks slip through.

Coverage is the other quiet failure point. The UK Health and Safety Executive found insufficient percentage coverage of inspection in 26 percent of the failure cases it reviewed. Sampling that skips the highest-risk locations defeats the purpose.

Repair rates are a useful reality check. TWI survey data puts average repair rates for the oil and gas and power sectors at 1 to 3 percent, with peak rates up to 25 percent in specific locations and up to 55 percent logged in some cases.

Finally, document everything: method, standard, extent of coverage, acceptance criteria, results, and any repairs. A traceable record is what turns an inspection into evidence.

Inspection does not create quality. It verifies it, and it tells you where your process is leaking. Match the method to the flaw, respect the timing, and cover what matters.

Root Causes, Prevention, and Repair: What Actually Reduces Weld Flaws

Finding a flaw is the easy part. Stopping it from happening again is where real quality lives. The UK Health and Safety Executive studied welding failures across industry and found the most prevalent causes were not exotic metallurgy. They were inadequate welder supervision (26%), insufficient inspection coverage (26%), and joints that were never designed for ease of fabrication. Dig one level deeper and the highest-ranking process sub-cause was welding parameters that were incorrectly specified or simply not understood by the welder.

That is a systems problem, not an equipment problem. Most flaws trace back upstream: a WPS nobody read, a gap forced closed, a fitter who left, or a joint drawn with no room for a torch. The pattern holds across shops, because failures cluster around communication and preparation, not welder talent. Hand a competent welder a clean, well-fitted joint, a clear procedure, and correct settings, and rejects become rare.

Root Causes You Can Actually Control

Unspecified or misunderstood parameters – the leading sub-cause. If amperage, travel speed, and polarity live only in a welder’s head, flaws follow. Document them.
Poor fit-up – gaps, mismatched edges, and tacked-up joints that force the welder to compensate with heat and speed.
Missing preheat – especially on thick or high-carbon sections, where hydrogen and fast cooling invite cracking.
Weak supervision – no one checking that the procedure is followed before the weld cools.
Thin inspection coverage – sampling that misses the joints most likely to fail.
Designs that ignore access – if a welder cannot reach the root, they cannot fuse it.

Short Note: Supervision and inspection coverage are management levers, not shop-floor luxuries. When 26% of failures trace to each, the fix is scheduling qualified inspectors, defining hold points, and verifying parameters at the arc rather than trusting the paperwork.

Qualification and Training: The Preventive Layer

Prevention starts with proof that the process works. Under AWS B2.1 and ASME Section IX, a Welding Procedure Specification (WPS) is backed by a Procedure Qualification Record (PQR) – the documented basis for proving a weld can meet required mechanical properties before production begins. Welders are then individually qualified to that procedure. ISO 3834 extends the same discipline into a documented shop-wide quality system, covering material traceability, inspection planning, and record-keeping. None of it works without trained people, which is why consistent welding practice matters as much as paperwork.

Qualification is not one-and-done. A welder’s qualification is limited by position, thickness, and process; change the essential variables and the qualification may no longer cover the work. Periodic re-testing and continuous production monitoring keep it meaningful.

Quick Note: A welder can be perfectly qualified and still produce defects if the machine is set differently from the procedure. Lock your equipment setup – check quick connect welding leads for clean, consistent connections.

The Correct Repair Workflow

Repair means removing a non-acceptable defect, re-welding, and reinstating the original geometry of the component, according to TWI. Rushing any step creates a new flaw.

Locate and size the flaw with NDT.
Excavate it fully – grind or gouge until sound metal is reached.
Confirm removal with repeat inspection.
Re-weld to procedure, not to habit.
Re-inspect the repaired area.
Reinstate geometry and finish the surface.

TWI notes the biggest perceived drivers of repair rates are the welder’s skill, the weld’s location or accessibility, and poor fit-up before welding – the very root causes you prevent upstream.

Repairs are often more crack-prone than the original weld because the surrounding structure restrains contraction more tightly. Preheat and hydrogen control matter even more the second time around, as TWI’s work on hydrogen cracking explains.

The Cost of Poor Quality – and Over-Repair

Flaws cost far more than repair time. Rework, scrap, warranty claims, and schedule delay compound quickly, and a single reject can halt an entire fabrication sequence. Catching a flaw at the station costs a fraction of catching it after coating, shipping, or installation. Yet over-repair is its own problem: cutting out sound metal, adding heat cycles, and repeatedly re-welding can weaken a joint more than a small, assessed flaw ever would. Document the repair, work to the procedure, and repair only what the code rejects.

Your Practical Playbook: Preventing, Finding, and Fixing Welding Flaws

Welding flaws rarely announce themselves. The bead can look flawless while the radiograph tells a different story. What follows is a repeatable routine that closes that gap, built on the causes, codes, and detection methods covered throughout this guide. Treat each stage as a checkpoint, and the odds of a defect slipping through drop sharply.

Before You Strike an Arc

Verify materials and consumables. Confirm the filler matches the base metal, check certification, and make sure MIG or flux-core wire is dry and clean. Learn how MIG welding wire composition affects the deposit.
Fit-up and cleanliness. Bad gaps, poor alignment, and oily edges cause lack of fusion and porosity before you ever light up. Clean the joint like the weld depends on it, because it does.
Review the procedure. Read the WPS, and know the governing code before welding, not after. If a parameter is unclear, resolve it now rather than mid-pass.
Preheat deliberately. For high-carbon-equivalent steels, lack of preheat has been repeatedly documented as a cause of hydrogen-induced cracking, making controlled preheat non-negotiable.

A tidy, well-planned shop reduces variability. If you are setting one up, this guide to planning a safe welding facility is worth the read.

During Welding

Hold the parameters on the sheet; chasing changes mid-pass invites defects.
Control travel speed, since it governs bead profile and penetration.
Clean slag and spatter between passes so inclusions cannot form.
Manage heat input and sequencing to limit distortion and residual stress.
Watch the arc, not just the puddle; hesitation and wandering both leave marks.

Small setup details matter more than most welders expect. A worn contact tip or the wrong polarity can sabotage an otherwise sound procedure, so replace consumables on schedule instead of waiting for a failure.

After the Last Pass

Inspect visually first. Most surface defects, undercut, and overlap, show up to the naked eye.
Respect the clock. Hydrogen cracking can appear hours later, so time delayed NDT correctly.
Choose NDT to match the flaw. RT, UT, PT, and MT each catch different things, so pick the method that sees your likely defect.
Document everything. Records make trends visible and turn one bad weld into a lesson.

Quick Note: ISO 3834 sets quality requirements for fusion welding of metallic materials, with Part 2 covering comprehensive requirements and Part 1 defining selection criteria. Use it to decide how much control your job truly needs.

The Mindset That Prevents Trouble

AWS D1.1 reminds us that a discontinuity is not necessarily a defect. Acceptance must always be judged against the governing code, never personal opinion. That is why scattered fine porosity is often acceptable while clustered or linear porosity signals a systematic problem and is judged severely. Same flaw family, very different verdicts. Keep the standards close and the ego in check.

Highest-Value Habits by Process

That list is short on purpose; these few habits neutralize the most common failure modes in each process.

SMAW: consistent arc length and interpass cleaning.
GMAW: correct gas flow, clean liner, and stable stick-out.
FCAW: mind polarity and slag removal.
GTAW: clean tungsten, correct polarity, and patience.
SAW: flux handling and travel consistency.

Where This Is Heading

Automated monitoring, digital WPS platforms, and in-situ sensing are moving quality control closer to the arc itself. Systems like robotic submerged arc welding already log parameters in real time, catching drift before flaws form.

None of this replaces skill, but it rewards welders who build habits early. Sharpen yours with our guide to building real welding skills, then browse more preventive, detection, and repair articles across weldgallery.com.

Frequently Asked Questions

What is the difference between a welding discontinuity and a welding defect?

A discontinuity is any interruption of the typical structure of a material, such as a void or a lack of fusion, and AWS D1.1 states it is not necessarily a defect. A discontinuity becomes a defect only when it exceeds the acceptance limits of the applicable code, making it rejectable.

What are the most common welding flaws?

The most frequently encountered flaws are lack of fusion, lack of penetration, porosity, slag inclusions, cracks, undercut, overlap, spatter, and distortion. HSE research found crack-like or volumetric flaws in 49% and poor or undersized geometry in 45% of the failures it reviewed.

Can porosity ever be acceptable in a weld?

Yes, in many codes scattered fine porosity within defined size and count limits is acceptable. Clustered or linear porosity is judged more severely because it indicates a systematic cause, and any visible porosity in a pressure-retaining joint is usually unacceptable.

What causes lack of fusion in a weld?

Lack of fusion is caused by insufficient heat input, travel speed too fast for the current, incorrect electrode angle, poor joint access, arc blow, or contamination films between passes. The fix is increasing heat input, slowing travel, and improving fit-up and access.

How are welding flaws detected?

Visual inspection is the first line, followed by liquid penetrant and magnetic particle testing for surface and near-surface flaws. Radiography detects volumetric flaws such as porosity and slag, while ultrasonic and phased-array testing are better for planar flaws like lack of fusion.

What is the difference between hot cracking and cold cracking?

Hot cracking occurs during solidification and is most associated with austenitic stainless steels and deep narrow pools. Cold or hydrogen-induced cracking happens after cooling when diffusible hydrogen, hard microstructures, and restraint combine, and it can be delayed, so inspection timing matters.

What do ISO 5817 quality levels B, C, and D mean?

ISO 5817 assigns three quality levels for fusion-welded joints: B is the highest or most stringent requirement, while C and D represent progressively more permissive limits. The appropriate level depends on loads, temperature, corrosion exposure, and pressure requirements.

How much does weld repair cost?

Repair costs vary widely, but the consequences are significant. TWI reports average repair rates of 1-3% in oil and gas and power work with peaks up to 25% and occasionally 55%, while EWI documented pipeline cases where weld removal and rework exceeded $4 million and $10 million.

How do I prevent welding flaws?

Work to a qualified welding procedure, verify base metal and consumables, clean the joint, control heat input and travel speed, use preheat and low-hydrogen consumables for high-carbon-equivalent steels, clean slag between passes, and inspect with the correct method at the correct time.

Is undercut always a defect?

No. Many codes permit limited undercut up to defined depth, length, and member-type limits, especially in statically loaded structures. Undercut in primary members under cyclic loading is much more strictly limited because it acts as a stress raiser and corrosion initiation site.