Welding Defects: Causes & Prevention

Every common welding defect has a known cause, a known prevention, and a test that finds it. This page covers the major ones, porosity, undercut, incomplete fusion, cracking, distortion, spatter, and heat-affected zone damage, in that order: what creates each defect, what stops it, and which inspection method catches it. Arguments about weld quality end the same way every time. The applicable code decides what is acceptable, so acceptance runs through every section here. These defects apply across MIG, TIG, and Stick welding. The process changes which defect is most likely, not whether defects are possible.

One vocabulary distinction does most of the work in a defect dispute. ISO 6520-1 sorts fusion-weld imperfections into six groups: cracks, gas cavities, solid inclusions, lack of fusion and penetration, imperfect shape and dimensions, and miscellaneous imperfections such as spatter and arc strikes. An imperfection is any deviation from the ideal weld. It becomes a defect only when it exceeds the acceptance criteria of the applicable code, which for structural steel work in the United States is AWS D1.1/D1.1M. A single pore is an imperfection. Enough pores of the wrong size in the wrong place is a defect that fails inspection. A fabricator who rejects every visible imperfection over-inspects. One who accepts every imperfection under-inspects. The glossary covers the general terms used here.

The defect map: cause, prevention, detection

The table maps each defect to its cause, its prevention, and the test that detects it. The sections that follow take the defects one at a time and add what a table cannot hold: why each one matters, which materials and processes run the highest risk, and how acceptance is usually judged.

Porosity

Porosity is gas trapped in the solidifying weld metal: hydrogen, nitrogen, or oxygen absorbed by the molten pool and released as bubbles that cannot escape before the metal freezes. The gas tracks the material. Hydrogen dominates in aluminum and stainless; carbon steel picks up hydrogen, nitrogen, and oxygen. Much porosity sits below the surface.

The causes cluster into three families: contamination on the plate or filler (oil, paint, rust, moisture, primer applied too heavily), moisture in the consumables, and shielding failure. Shielding fails in both directions. Too little flow, a leak in the gas line, a draught across the joint, or a shop fan lets air in. So does too much flow, which turns laminar shielding into turbulence that drags air in with it. TWI’s job-knowledge work puts a number on the threshold: roughly one percent air entrainment in the shielding gas produces distributed porosity, and above about 1.5 percent, gross surface-breaking pores appear.

Prevention is clean bright metal in and beside the groove, dry electrodes handled as the supplier specifies, a leak-checked gas system with the flow verified at the nozzle, and shelter from wind. Detection is visual for surface pores and radiographic for subsurface porosity, which shows as rounded dark indications, with ultrasonic testing as the alternative. Acceptance under the applicable code limits pore size, count, and distribution: scattered fine porosity is often acceptable, while clustered or linear porosity is judged severely because it points to a systematic cause.

Undercut

Undercut is a groove melted into the base metal at the weld toe and left unfilled. It thins the section, and worse, it is a notch. Under cyclic loading the sharp geometry concentrates stress and becomes a fatigue crack initiation site. That is why a defect that looks cosmetic is treated seriously in dynamically loaded structures.

The causes are excess current or voltage, a long arc, travel speed too high for the filler to wash into the toe, and an electrode angle that drives heat into the edge instead of the joint. Prevention is the mirror image: lower current, shorter arc, slower travel, and technique that washes filler into the toes, with enough overlap between passes in multi-run welds. Detection is visual inspection with a gauge. An undercut gauge or bridge cam measures depth against the code limit, and root-side undercut, which the eye cannot reach, is found by radiographic or ultrasonic testing. On acceptance, a commonly quoted AWS D1.1 visual limit holds undercut to 1/32 in on material under 1 in thick, with limited exceptions to 1/16 in in less critical locations. The applicable edition and loading case govern, and fatigue-sensitive joints get the strictest limits.

Incomplete fusion and incomplete penetration

These two internal defects are related but distinct, and confusing them wastes diagnostic effort. Incomplete fusion (lack of fusion) is a bonding failure: deposited weld metal never melted into the sidewall of the groove or into the previous pass. Incomplete penetration (lack of penetration) is a depth failure: the bead never reached the root of the joint. Both are planar, sharp-edged discontinuities that concentrate stress and start cracks, and both can hide beneath a sound-looking bead face. That makes them the classic subsurface inspection problem.

Lack of fusion comes from heat input too low, travel too fast, a torch angle that aims the arc away from the sidewall, contaminated surfaces, and grooves that deny the arc access. Lack of penetration adds tight root gaps, insufficient current for the thickness, and misaligned fit-up. Prevention starts in design: a groove that gives the arc access, a root opening and bevel that suit the process (see weld joints), and parameters with enough heat to melt both sidewalls. On the floor it is correct angle, adequate current, stringer control at the root, and clean metal.

Detection favors ultrasonic testing, which reflects strongly off planar gaps. Radiography also finds these defects, but it can miss a lack-of-fusion plane that lies parallel to the radiation beam, so orientation matters when choosing between the two. Most codes treat significant lack of fusion or penetration as rejectable in structural work: the effective throat is not there.

Slag inclusions

Slag inclusions are nonmetallic slag trapped in the weld, the solid-inclusions group of ISO 6520-1, and a flux-process risk: stick, flux-cored, and submerged arc welding in multi-pass work. They form when interpass cleaning is skipped, when current is too low for the bead profile, or when the bead shape traps slag at the toes. Prevention is to brush and chip between passes and run parameters that deposit a convex bead washing into the toes.

Detection is radiographic, which shows trapped slag as dark lines or irregular patches, with parallel wagon-track lines at the root-pass toes a telltale sign, and ultrasonic testing as the alternative. Acceptance follows the same code-relative rule as porosity: an inclusion becomes a defect when its size or count exceeds the limits of the applicable code.

Hot cracking

Hot cracking, or solidification cracking, forms as the weld pool freezes. In susceptible chemistry, impurities, mainly sulfur and phosphorus, segregate to the grain boundaries and stay liquid after the grains have formed. The last liquid films then tear apart under shrinkage stress before they can close. The defect is most associated with austenitic stainless steels, fully austenitic weld metal, and deep, narrow pools.

Prevention is mostly chemistry and shape. Use filler metal with the right composition; a controlled delta-ferrite content is the standard resistance mechanism in austenitic stainless weld metal. Keep the pool wider than it is deep, and lower the restraint so the solidifying metal is not pulled apart. Detection is visual for surface-breaking cracks, with dye penetrant or magnetic particle testing to confirm tight surface cracks and radiographic or ultrasonic testing for internal ones. Cracks of any type hold a special status in every major code: they are not tolerable at any size. A hot crack is always excavated and re-welded, never monitored.

Hydrogen-induced cold cracking

Cold cracking is the defect that arrives late. It needs three factors acting together: hydrogen dissolved in the weld, a hard susceptible microstructure in the weld or heat-affected zone, and tensile stress, which thick, restrained joints supply. Remove any one and the crack does not form. The hydrogen comes from moisture: damp electrode coatings, flux, or shielding gas, plus oil, paint, condensation, or hydrated rust on the plate. A hard microstructure forms when hardenable steel cools too quickly.

The distinctive behavior is timing. Atomic hydrogen diffuses after the weld cools, so cracks can appear hours or even days after the part has passed initial inspection. That is why crack-sensitive work is inspected after a deliberate delay. The cracks typically sit in the heat-affected zone, under the bead or at the toe, where the eye may never see them.

Prevention attacks all three factors. Use low-hydrogen consumables, the E7018 class, kept dry in heated storage and handled within their exposure limits. Keep the joints clean and dry. Control preheat and interpass temperature per the welding procedure, and cool the weld under control. AWS D1.1 sets minimum preheat and interpass values by steel category and thickness in its preheat table, and its annex offers a hydrogen-control method for calculating preheat from hydrogen level and restraint. TWI’s hydrogen-control practice runs preheat as high as about 200C on thick, high-carbon-equivalent steel, with post-weld holds measured in hours (two to three hours on susceptible steel, longer for the most crack-sensitive combinations). Higher-carbon and low-alloy steels are the susceptible families. Plain low-carbon steel rarely needs these measures outside thick, restrained joints. Detection is magnetic particle or penetrant testing for surface-breaking cracks and ultrasonic testing below the surface, applied after the delay that lets the crack develop.

Distortion

Distortion is the shape change created by weld metal and adjacent base metal expanding when heated and contracting when cooled against the rest of the part. It takes familiar forms: transverse and longitudinal shrinkage, angular rotation around the weld, bowing along an assembly, and twisting. Thin sheet distorts most because it has the least stiffness to resist the shrinkage forces, which is why sheet metal fabrication treats welding sequence as a core process decision.

Prevention is shop practice, not code. Deposit the least weld metal the design allows; bigger is not safer, it is just more shrinkage force. Use the fewest passes, and intermittent welds where the joint permits. Balance the sequence about the assembly neutral axis, and back-step or skip long welds. Preset parts so shrinkage pulls them into true, and clamp or fixture against movement. These are the distortion-control techniques TWI and Lincoln Electric both teach as standard fabrication practice. Detection is dimensional: measure the welded assembly against the drawing, the same discipline that governs machining tolerances. A dimensionally wrong weldment is rejected on measurement, not on appearance. Mechanical or flame straightening can correct distortion, but prevention costs a fraction of correction.

Spatter

Spatter is the least structural defect and the most annoying: droplets of molten metal thrown from the arc and frozen onto the surrounding plate. A spattered weld can be perfectly sound, so spatter is judged by cost and by what it signals. It adds grinding labor, fouls fixtures and nozzles, interferes with coatings, and flags an unstable process that may be producing porosity or fusion problems at the same time.

Causes: voltage set too high, an arc too long, wire feed and travel out of balance, low inductance on short-circuit transfer, dirty or rusty plate, and shielding gas choice. Pure CO2 trades cheap gas for heavy spatter; argon-CO2 blends run cleaner. Prevention is parameter tuning to a stable arc, the right gas blend, correct stickout, and clean plate. Anti-spatter compounds protect fixtures but do not fix the process. Detection is visual, which is the point: a spatter problem shows in the first inch of a test bead, so it gets fixed before production, not discovered after.

Heat-affected zone problems

The heat-affected zone (HAZ) is the band of base metal the weld thermally changed without melting. Its problems are metallurgical rather than geometric, and three matter most.

Hardening: in hardenable steels the HAZ heats above its transformation temperature and cools fast, forming hard, brittle microstructure that cracks readily. Preheat and controlled cooling are the countermeasure, the same mechanism behind hydrogen-cracking prevention. Sensitization: in austenitic stainless steel, holding roughly 425 to 815C lets chromium carbides precipitate at grain boundaries, depleting chromium beside the boundary and opening the door to intergranular corrosion along the weld. Low-carbon L grades and stabilized grades resist it, a material decision made at design time, not in the shop (see welding materials). Grain coarsening: excessive heat input grows the grains and lowers toughness.

One related base-metal failure belongs here as a design edge case. Lamellar tearing is step-like cracking in rolled plate loaded through its thickness by the shrinkage of heavily restrained T and corner joints. Joint design, buttering the surface, or plate with guaranteed through-thickness ductility prevents it. Detection depends on the problem: hardness testing for hardening, corrosion testing for sensitization where service demands it, and joint review for lamellar-tearing risk before welding, because no surface test finds it after the fact.

Prevention: procedure and design

Defect prevention splits into two halves that meet in the welding procedure. Design decides the joints, the materials, and the weld sizes. The procedure and the welder control heat, cleanliness, and sequence.

On the design side, the biggest gains come from the cheapest decisions made earliest. Size welds to the load, not to the nerves: minimum-size welds that meet the drawing reduce distortion and cracking risk together. Choose joint geometry that gives the arc access, since fusion defects are often designed in as grooves the torch cannot reach. Specify materials with welding in mind: low-carbon or stabilized stainless for welded construction, plate with through-thickness ductility for restrained corner joints, and low-hydrogen consumable classes for thick or high-carbon-equivalent steel. Put fit-up requirements on the drawing so gap and alignment arrive controlled rather than assumed. Plan the welding sequence as part of design for manufacturing review before the first arc strike, because sequence is the main distortion lever and it is nearly free to change on paper.

On the procedure side, the controls are discipline. Weld to a qualified procedure with parameters inside its window. Clean the joint inside and out to bright metal. Keep flux processes brushed and chipped between passes. Handle low-hydrogen consumables per their storage rules, apply and verify preheat where the procedure calls for it, and fill craters at every stop. Repairs follow the same logic in miniature: assess with the right test, excavate the defect completely to sound metal by grinding or gouging, clean the cavity, re-weld under a qualified repair procedure, and re-inspect the repair itself. TWI’s hydrogen practice adds one counterintuitive rule for repairs: use higher preheat than the original weld, because a repair joint is more restrained than the original.

How each inspection method works

Each nondestructive method answers a different question, and matching the method to the defect risk is what makes inspection meaningful. Visual testing, the scope of ISO 17637, is the first line: undercut, spatter, overlap, cracks that reach the surface, profile and size problems, and fit-up errors before welding. Done with gauges, it is quantitative: an undercut gauge or bridge cam for depth, fillet gauges for leg size, not just eyes.

Dye penetrant testing finds surface-breaking defects only, but on any nonporous material, which matters for stainless and aluminum where magnetic methods cannot work. Magnetic particle testing finds surface and near-surface discontinuities, cracks above all, but only in ferromagnetic material. It is the standard check for hydrogen cracks at the toe and for verifying complete crack removal after repair grinding.

Radiographic testing images the interior and excels at volumetric defects: porosity and slag show as dark rounded or irregular indications, though a lack-of-fusion plane parallel to the beam can escape it. Ultrasonic testing, including phased array, is strongest for planar internal defects, lack of fusion, lack of penetration, and cracks, because it reflects off oriented interfaces that radiography can miss. The inspection methods overview covers the equipment and process detail behind each method.

Acceptance then belongs to the code: AWS D1.1 for structural steel, with ISO 5817’s quality levels B, C, and D as the common European framework, each mapping the same imperfections to graded acceptance limits.

Weld defect checklist

• Joint cleaned to bright metal inside the groove and beside it.
• Consumables correct for the base metal, dry, and inside their exposure limits.
• Gas system leak-checked, flow verified at the nozzle, draughts controlled.
• Fit-up checked against the drawing: root gap, land, alignment.
• Preheat and interpass minimums applied, measured, and recorded where the procedure requires them.
• Weld size and length as drawn, with no reinforcement added for reassurance.
• Welding sequence planned for balance before the first pass.
• Craters filled at every stop.
• Interpass cleaning done for every flux process pass.
• Visual inspection performed with gauges before any part leaves the station.
• NDT method matched to the defect risk: MT or PT for surface, RT or UT for internal.
• Hydrogen-crack-sensitive work inspected after an appropriate delay.

Common mistakes

• Judging welds by appearance: a clean, uniform bead can hide lack of fusion, and a rough bead can be sound.
• Over-welding for safety, which raises distortion, shrinkage, and cracking risk instead of strength.
• Chasing porosity through parameters while the real cause is a gas-line leak or a fan across the joint.
• Running low-hydrogen electrodes from an open box in damp conditions, which defeats their purpose.
• Welding thick or higher-carbon steel cold, with no preheat, in a restrained joint.
• Skipping interpass cleaning in multi-pass flux work and welding over slag.
• Stopping the arc abruptly and leaving an unfilled crater that cracks.
• Inspecting crack-sensitive welds immediately, before delayed hydrogen cracks have had time to appear.
• Grinding a defect visually “clean” without penetrant or magnetic particle confirmation that it is gone.