Horrible Welds: What Causes Them, What They Really Cost, and How to Fix (or Avoid) Them

Marcus had been welding for eleven years when the X-ray report came back on a Thursday afternoon and ended his week: 19 of 40 pipe joints rejected, all of them porosity. He stood in the shop, looking at two rows of welds that looked beautiful from the outside—smooth beads, even ripple, no undercut—and realized the ugliest welds in the building were the ones nobody could see. That single report triggered a three-day shutdown, a full re-inspection, and a very uncomfortable conversation about the regulator he’d been told was “fine” six months earlier.

Horrible Welds: Why Bad Welds Happen and Why They Cost So Much

Marcus’s rejection report is a story nearly every fabrication shop can retell in some form. Two rows of pipe, beads so even they could have doubled as a training poster, and an X-ray film that disagreed with every one of them. Nineteen joints condemned for porosity nobody could see from the outside. The report never asked how the weld looked. It asked what was inside it.

That’s the hard truth behind horrible welds. On the shop floor, the phrase has almost nothing to do with ugly. It’s working shorthand for weld defects and discontinuities: the flaws that slip past the eye, hide beneath a smooth cap, and wait patiently for load, pressure, cycles, or time to finish the job. A weld can be the best-looking thing in the building and still be the most dangerous.

What “Horrible” Actually Means When We Talk About Welds

Welding is one of the most widely used joining processes on earth, and it is inherently a process of controlled defects. Every arc introduces discontinuities. The metallurgy, the heat, the cooling rate, the shielding, the operator’s hands—all of it guarantees that a weld is never mathematically perfect. The real job is not producing a flawless weld. It’s keeping flaws below defined acceptance limits.

The catalog is longer than most people expect. AWS B1.11, the Guide for the Visual Inspection of Welds, lists what inspectors hunt for: porosity, incomplete fusion, incomplete joint penetration, undercut, underfill, overlap, lamination, seams and laps, cracks, slag inclusions, excessive reinforcement, convexity and concavity, arc strikes, spatter, melt-through, weld size, and surface oxidation. Read that list twice. Half of it is invisible from the outside, and much of it can survive years of service before announcing itself. If you want to train your eye on something simple first, our breakdown of good and bad MIG welds covers the visual clues that separate a sound bead from a suspicious one.

Experienced Welders Make Horrible Welds Too

Here’s the part that surprises people outside the trade: defect rates are not concentrated among beginners. A 2020 study of boiler fabrication published in the IOP Conference Series examined 12,499 pieces and found an overall defect rate of 21.6%, with porosity alone responsible for 724 defects and undercut for 613. A 2019 power-plant piping project in Indonesia reported 22.44% of welded joints rejected, with porosity the dominant cause at 41.75% of all rejections. Both were professional operations staffed by qualified welders.

The difference between a horrible weld and an acceptable one is rarely talent. It’s usually a system: schedule pressure that rewards speed over procedure, a regulator diaphragm that quietly failed six months ago, filler metal stored in a damp corner, fit-up gaps nobody measured, or a machine drifting out of range while the operator trusts memory instead of a gauge. That’s why deliberate welding practice and a training culture matter more than raw hand speed—and why most of this guide is about causes rather than blame.

Codes Give “Horrible” a Number

Because “bad” needs a definition, the industry leans on codes. ISO 5817 defines weld quality levels B, C, and D for fusion-welded joints, giving fabricators a shared language for how horrible is too horrible. AWS D1.1 governs structural steel, ASME BPVC Section IX governs procedure and welder qualification, and API 1104 governs pipeline work. The same 3/16-inch undercut can pass on a handrail in one shop and be an automatic reject on a pressure vessel in the next—and both decisions can be correct.

What You’ll Find in This Guide

Over the next sections we move in a deliberate order: the terminology that decides whether something is an indication, a discontinuity, or a defect; a two-part field guide to the full defect taxonomy, starting with gas and arc-related flaws and then the fusion, cracking, and inclusion defects that cause catastrophic failures; the real root causes across machine, method, material, measurement, environment, and people; how inspection catches these things, from visual examination through NDT and welder qualification; what rework truly costs in labor, schedule, safety, and liability; repair methods that hold instead of hiding the problem; a process-by-process comparison of SMAW, GMAW, FCAW, GTAW, and SAW defect profiles; and a prevention playbook you can put to work Monday morning.

Throughout, we lean on primary sources rather than folklore: AWS standards, ISO and ASME codes, and safety material from OSHA and NIOSH. Welding hazards are well documented—fume, arc radiation, burns, and ergonomic strain included—and the same agencies that track injuries make clear how much of this is preventable with procedure, equipment care, and discipline.

Indication, Discontinuity, or Defect? The Vocabulary That Decides Whether a Weld Is ‘Horrible’

Ask ten welders what a “horrible weld” is and you’ll get ten answers. One points at uneven ripple, another shrugs off porosity because “it’s only a bracket,” and a third wants anything with undercut cut out. They’re all seeing real things, but using three different words as if those words meant the same thing. That mix-up is where shop-floor arguments, wasted rework, and scrap begin.

Before causes and fixes, get the language straight: pass versus reject rarely depends on how a weld looks, but on what a specific code requires.

Three Words, Three Different Meanings

The terms build on each other, and blending them is a fast way to look like a beginner in a QC meeting:

• Indication: a signal from an inspection method. That’s all. A penetrant indication could be a crack, a scratch, spatter, or a machining mark. The method is saying “look here,” not “this is bad.”
• Discontinuity: a break in the ideal, uninterrupted weld. Porosity, slag, undercut, misalignment, overlap, incomplete penetration—all discontinuities. They interrupt the intended uniformity of the joint, and they exist in virtually every weld ever made at some microscopic scale.
• Defect: a discontinuity that exceeds the acceptance criteria for that application. It’s a rejectable discontinuity—one the applicable code, the engineer, and the service conditions have collectively declared unacceptable.

Read that again, because it flips how most people think: every defect is a discontinuity, but most discontinuities are not defects. A weld full of them can still be perfectly acceptable work.

The Code Is the Referee

Every judgment about “horrible” has to name a rulebook:

The numbers inside those documents are where it gets interesting. AWS B1.11, the visual inspection guide, shows how much is settled with trained eyes and simple gauges. Under AWS D1.1 visual criteria, complete-joint-penetration groove welds in butt joints transverse to computed tensile stress shall have no visible porosity—zero tolerance, because that’s the direction the stress comes from. AWS also calls incomplete fusion unacceptable when it exceeds 1 inch in length, or an aggregate exceeding 8% of the weld length in short welds. Hard, measurable lines, not opinions.

ISO 5817 grades by consequence instead. At quality level B, lack of fusion is not permitted at all; at levels C and D, short imperfections are allowed within defined limits. Porosity behaves the same way: level D permits up to 2.5% in a single-layer projected area, tightening to 1.5% at level C and 1% at level B. Same weld, same X-ray, different specified level—opposite verdicts.

Same Flaw, Two Verdicts: A 3/16-Inch Undercut

Picture a 3/16-inch undercut along the toe of a fillet weld.

On a handrail, it’s a cosmetic blemish. The member is oversized, loads are light and effectively static, and a small loss of throat thickness costs nothing. Grind it out if the customer cares, then sign it off.

Move that weld to a pressure vessel carrying hot fluid through thousands of pressure and thermal cycles, and the undercut becomes a sharp notch cut into the load path by the arc. It concentrates stress at the toe, thins the pressure boundary, and gives a fatigue crack its perfect birthplace—while trapping corrosion in a spot no inspector will ever see again. Different service, different answer, and this one is reject.

Acceptance Is Defined by Intended Service

When an engineer writes acceptance criteria, they’re answering one question: what happens if this weld fails? The variables that push a discontinuity from tolerable to fatal include:

• Loading type: static loads forgive far more than cyclic ones, which find every notch and grow cracks from it.
• Corrosion: salts, acids, and moisture turn tiny crevices into active corrosion cells.
• Pressure: stored energy means a pinhole becomes an explosion, not a drip.
• Temperature: cold service promotes brittle fracture; hot service drives creep and oxidation.
• Consequence of failure: a cracked gate hinge versus a cracked gas line, same geometry, different universe.

That’s why visual judgment alone is unreliable: two welds can look identical and sit at opposite ends of the scale because one carries a water line and the other carries a bridge. To sharpen your eye for structurally sound versus merely pretty, see our guide to good and bad MIG welds, and our breakdown of the overlap welding defect shows how one geometry change becomes code rejectable. The Michigan Tech welding defects webinar is a solid free primer on the same distinction, and it pairs well with the structured approach in our welding practice guide.

The Takeaway

“Horrible” is not a weld property. It’s a verdict produced by comparing a discontinuity against acceptance criteria written for a specific service. Learn to separate the signal from the finding, and the finding from the judgment, and you’ll stop scrapping parts that were fine—and stop shipping parts that weren’t.

The Horrible Welds Field Guide, Part 1: Gas and Arc-Related Discontinuities

Most horrible welds announce themselves two ways: trapped gas, or heat and arc that misbehaved before the metal froze. Both families appear in this first stretch of the field guide, and none are mysterious once you know what feeds them.

Porosity: The Gas-Trap Family

Porosity is a cavity-type discontinuity formed by gas entrapment during solidification. The gas almost always traces back to contamination or a shielding-gas problem, making porosity the most common red flag in a rejection report. Pores are spherical, so they concentrate stress less sharply than a crack or slag line. The real issue is what they signal: parameters, technique, consumables, gas shielding, or fit-up out of control.

Inspection reports split porosity into four patterns, each pointing at a different cause:

• Scattered porosity — random isolated pores. Often damp electrodes or light surface contamination.
• Clustered porosity — a local pocket, often where the arc started or restarted.
• Linear porosity — pores lined up along one path, commonly tracing a root gap or a seam of dirt.
• Piping (wormhole) porosity — elongated voids along the weld axis. Codes list this separately from elongated porosity so the note points at the right correction.

The usual suspects: damp electrodes, dirty base metal, lost shielding gas, drafts, excessive gas flow that creates turbulence instead of a blanket, and the wrong arc length. On the wire side, a worn or clogged contact tip and a dirty nozzle wreck shielding as surely as an empty cylinder. Under AWS D1.1, visible porosity in non-critical groove and fillet welds is limited to a sum not exceeding 3/8 inch per linear inch, with individual pores capped near 3/32 inch in fillet work. International work uses the tighter tiers of ISO 5817.

Undercut: The Groove You Didn’t Fill

Undercut is a groove melted into the base metal along the weld toe that the filler never filled back in. It shows up worst at the toes in vertical and overhead work, caused by too long an arc, excessive travel speed, a bad work angle, or too much current and heat. Under load it cuts throat thickness and leaves a sharp notch right where bending and fatigue stresses peak. Under common pipeline criteria, undercut deeper than 1/32 inch — or 12.5% of pipe wall thickness, whichever is smaller — is a defect. The fix is boring: shorten the arc, slow travel slightly, correct the angle, and add fill.

Burn-Through, Spatter, Arc Strikes, and Sugaring

Burn-through (melt-through) is a hole blown through the root, most common on thin material, wide root gaps, and high current with no backing. Drop the amperage, tighten fit-up, speed up travel, and add backing when the job allows.

Spatter is molten metal stuck around the weld: rarely a strength issue, but a corrosion, coating, and appearance problem that costs cleanup time. High current, long arc, wrong polarity, and a damp electrode all make it worse. Arc strikes are more serious, creating hard, brittle spots outside the joint that can nucleate cracks; most codes reject them outright, so grind and inspect them.

On stainless, “sugaring” is the rough, oxidized, blackened surface left when shielding or heat control fails. Chromium burns off at the surface, taking corrosion resistance with it. Correct polarity and heat control fixes most cases.

Reinforcement Problems: Excess, Underfill, Convexity, and Overlap

Too much weld is still a defect. Excessive reinforcement creates a stiff, heavily notched crown that draws stress, and many codes cap crown height. Underfill and concavity shrink the cross-section below what the design intended. An overlap defect happens when molten metal rolls over the base metal without fusing to it: the toes look smooth but hide a cold, unfused edge. All four trace back to travel speed, work angle, wire feed, and heat input, so the fix is usually a settings change plus technique work.

Gas and arc defects are the loud ones: visible, smelly, and easy to diagnose. The next part covers the quiet killers, fusion, cracking, and inclusion defects that pass the naked eye and still fail under load.

The Horrible Welds Field Guide, Part 2: Fusion, Cracking, and Inclusion Defects

Part 1 covered flaws you can see; this part covers the ones that look right and aren’t. Lack of fusion, cracks, and trapped inclusions hide under an attractive cap pass, and they fail suddenly in service rather than sagging first. Surface porosity rarely ruins a component; an unfused sidewall can.

Lack of Fusion and Incomplete Joint Penetration

Lack of fusion (LOF) is where weld metal never bonded to the fusion faces, or one bead never bonded to the one below it. Three versions show up constantly:

• Sidewall LOF — the bead kisses the joint face without melting into it.
• Interbead LOF — the next pass rides on the previous one without fusing.
• Root LOF — the first pass misses the root face entirely.

Fusion needs heat, dwell time, and the filler aimed where the metals should merge. Travel too fast, park the arc on the puddle instead of the joint face, or run too little current, and the bead sits beside the parent metal instead of in it. That often comes from amperage set from memory: a 3/16 E7018 amperage perfect for a flat groove can be wrong vertical. Mill scale, oil, and rust make it worse.

Incomplete joint penetration (IP) is LOF’s cousin: the weld doesn’t reach through the full thickness, so the effective throat is thinner than the drawing demands. Oversized root faces, tight root gaps, and low-current root passes create it. Radiography finds it; the eye rarely does. See the MTU welding defects webinar material.

High-Low Misalignment: The Invisible IP Generator

Hi-lo is the offset between two pipe or plate edges at the joint. The cap can look immaculate while the inside is stepped, because internal misalignment creates incomplete penetration even when the visible pass looks fine. Poor fit-up, tacking without alignment clamps, out-of-round pipe, or thermal movement cause it. Only a hi-lo gauge tells the truth.

Cracking: The Automatic Reject

Cracks are defects under AWS acceptance criteria regardless of size; structural and pressure work effectively allow none, which is why they appear so bluntly in AWS D1.1. Stress concentrates at the crack tip, and a crack grows rather than heals.

Hot cracking (solidification and crater cracks) happens while the weld is still hot. Deep, narrow centerline beads with high sulfur or phosphorus are the usual recipe, and a concave crater is the classic tell. Fix it with better bead shape, crater fill, and less restraint.

Cold cracking is the sneaky one. Hydrogen-induced cracking appears after welding finishes, sometimes up to 48 hours later. Moisture in electrodes, damp flux, oil, and humid air feed hydrogen into the arc, which diffuses into the heat-affected zone and builds pressure under restraint. That’s why preheat, post-heat, and dry electrodes matter — and why crack-sensitive steels deserve thought about welding hardened steel.

Toe and underbead cracks start in the HAZ, often parallel to the weld toe — the ones you find Monday after welding Friday. Lamellar tearing happens in thick plate when nonmetallic inclusions tear under through-thickness shrinkage stress. ISO 5817 treats all of these across severity levels, so read the level before accepting or rejecting under ISO 5817.

Inclusions and Distortion Cracks

Slag inclusions occur when flux gets trapped between passes, usually when interpass cleaning is rushed; stick and flux-cored work are the usual suspects. Each trapped channel is a stress riser and a corrosion site, and aligned slag lines are rejectable. Tungsten inclusions belong to GTAW: the electrode dips into the puddle, current runs too high, or the wrong polarity pushes heat into the tungsten. The torch behavior covered in TIG welding polarity shows which settings to change.

Distortion-induced fit-up cracking, also called restraint cracking, happens when parts are welded in the wrong sequence. Stress locks in, a later weld adds more, and a crack opens at an already-stressed spot. The fix is sequencing and balanced welding, not bigger clamps.

Case: A Fillet Weld That Passed Visual and Failed UT

A fabricator ran a half-inch fillet on a structural T-joint. Even legs, a clean toe line, no undercut — visual passed it. Ultrasonic testing found interbead lack of fusion along most of its length. The welder had traveled vertical-down on a joint the procedure called vertical-up, and interpass temperature was never checked.

Why “Grind It and Weld Over It” Makes It Worse

Washing over a defect doesn’t remove it — it buries it and hides it from the next inspector. Grinding a shallow dish over a crack leaves the crack tip intact and lets it grow into the new metal. Only full excavation to sound metal and re-welding to the procedure holds.

Why Welders Produce Horrible Welds: Root Causes in the Machine, Method, Material, and Human

A rejected weld is a symptom, not the disease. The cause rarely lives at the end of the torch. Root-cause analysis means chasing the weld backward until you find the variable that actually moved. A fishbone diagram gives you six branches to search: machines, methods, materials, measurement, environment, and people.

Six Branches, One Question: What Actually Changed?

The fishbone works because it forces you past the obvious. Map each branch against your defect data and the picture sharpens fast:

• Machines — worn drive rolls, voltage drifting under load, unstable generator power, clogged nozzles, a contact tip oval after a thousand hours of wire.
• Methods — no written settings, torch angle left to memory, travel speed improvised, WPS compliance treated as paperwork.
• Materials — mill scale, oil, paint, moisture on the plate, filler metal stored open in a humid shop.
• Measurement — no calibrated gauges, no defect tracking, no baseline for what “good” looks like.
• Environment — wind, drafts, poor lighting, awkward positioning a jig could have solved.
• People — training gaps, fatigue, and the pressure that makes shortcuts feel reasonable.

Machines and Materials: The Quiet Failure Points

A machine that drifts 15 amps between jobs still strikes an arc, which is exactly why nobody notices. Connectors corrode, cables fray, and the arc gets unstable just enough to cost you penetration at the root. Regular maintenance of machines and connectors remains one of the least glamorous and most effective fixes in a shop. A worn contact tip wanders the wire off-center, throwing the arc sideways and producing lack-of-fusion that looks fine from outside.

Material culprits are mundane. Mill scale, cutting oil, primer, and condensation all release gas or refuse to fuse. A damp E7018 is not just a handling problem: moisture breaks down in the arc and feeds hydrogen into the weld pool, the fuel behind hydrogen-induced cracking. Keeping filler dry and electrode ovens running isn’t bureaucracy; it’s metallurgy.

Method, Measurement, and Environment

Method failures hide in the gap between what the procedure says and what the shop does. When a job gets hot, travel speed climbs, weave width shrinks, and joint prep gets rushed. Measurement failures are subtler: if you can’t measure a defect, you can’t manage it, and a plant that doesn’t track defects by type and location is flying blind.

Environment is the branch most often dismissed and most often guilty. Wind strips shielding gas before it reaches the pool, turning good technique into scattered porosity; a $30 windbreak has rescued more weld schedules than most capital purchases. Poor lighting and cramped positioning don’t cause the defect directly — they cause the inconsistency that does. Setting up properly, per the fundamentals in our guide to planning a welding facility, prevents a whole class of defects.

People: The Blame Trap

Blaming the welder is the fastest conclusion available and usually the weakest. A Lean Six Sigma project on turnaround welds began with a 3.66% average butt weld repair rate blamed on wind exposure and insufficient training, but the completed root-cause analysis overturned the obvious suspect. The dominant inputs were a failed shielding-gas regulator diaphragm and drift in the torch-angle setup fixture, not operator technique. The welder had been blamed for a machine that was quietly failing.

That pattern repeats. A gage R&R study on visual inspection returned 41% study variation, meaning nearly half of what the plant called process variation was actually inspector disagreement. Six Sigma research on boiler welding named welder skill, shielding or flux gas, and welding conditions as the three critical variables driving defects — two of the three machine and process variables, not the human.

Pareto: Fix the Vital Few

Not every cause deserves equal attention. Pareto analysis says a small handful of inputs drives most of your defects. Sort rejections by type and location for a month and the pattern usually surfaces: porosity at one station, lack of fusion on one fixture, cracking after one material change.

Corrective actions that hold up under audit are consistent: periodic maintenance of machines and connectors, material cleanliness control, structured welder training, and verified parameters rather than recalled ones. Porosity causes are almost always actionable at that level — clean the base metal, store filler properly, and check for gas leaks before blaming anyone’s hands. Ongoing skill building, as covered in our guide to building real welding skills, prevents defects at the source instead of catching them at inspection. And because defect-driven rework multiplies fume exposure hours, OSHA’s welding hazard guidance exists for exactly that reason.

Fix the system and the welder stops being the variable you manage. Fix the welder alone, and the same defect returns next quarter.

How Horrible Welds Get Caught: Visual Inspection, NDT, and Real Acceptance Testing

A horrific weld only becomes a “horrible weld” when it’s measured against a standard and fails. Until then it’s just metal. So the real question isn’t whether a bead looks pretty, but how a shop catches the flaws hiding inside the joint. That means inspecting before, during, and after welding, then confirming with nondestructive testing (NDT) when the consequences of failure are high.

Visual Examination: The Cheapest Test You Already Own

Visual examination (VT) is the most underrated tool in the shop. It costs nothing extra and catches most problems before they become permanent. AWS B1.11 organizes visual examination into three timeframes: before welding, during welding, and after welding. Each stage catches different defects.

• Before welding: joint preparation, fit-up, root gap, cleanliness, and hi-lo misalignment. Catching a wide root gap here costs seconds; catching it after welding costs hours.
• During welding: arc stability, travel angle, interpass cleaning, and whether the welder is actually following the procedure. The bead is still fixable at this point.
• After welding: bead profile, undercut, overlap, spatter, and surface cracking.

Good visual inspection needs more than sharp eyes. Inspectors require documented visual acuity (near-vision testing is standard), adequate lighting, and the right gauges: an undercut gauge, a hi-lo gauge for internal misalignment, fillet weld gauges, and a magnifier for crack detection. If you’re a welder checking your own work, treat lenses and visibility as inspection equipment, not just safety gear — the details matter, as covered in our guide to protective eyewear for welding.

Certification matters too. AWS QC1 is the standard for AWS Certification of Welding Inspectors — the credential behind the Certified Welding Inspector (CWI) designation. But even a certified inspector’s eyes aren’t perfectly repeatable. A gage R&R study on visual inspection found 41% measurement variation, which is why serious shops calibrate inspectors against boundary samples and welded reference specimens before trusting any defect data.

NDT: Seeing What the Eye Can’t

When internal soundness matters, visual inspection isn’t enough. AWS B1.10 details the nondestructive examination methods used to detect internal discontinuities the eye can’t see. Each method has a specific job:

Radiographic testing is the method that detects internal porosity and fusion defects in pipe and pressure work — which is exactly why “looks good” is not a quality metric. A bead can be flawless on the surface and still be hiding porosity, as our comparison of good and bad MIG welds explains from the surface point of view. Add hardness testing for hydrogen-cracking risk in the heat-affected zone and ferrite checks for austenitic stainless and duplex work, and you have a reasonably complete picture.

Destructive testing is the other half of the story: bend, tensile, and Charpy impact tests verify that a procedure is sound rather than a specific production weld. These are used for procedure qualification, not routine inspection.

Sampling vs. 100% Inspection

Not every weld gets X-rayed. Structural standards like AWS D1.1 use a risk-based sampling approach, inspecting a percentage of welds and escalating when a rejection appears. Pressure piping and pipeline girth welds often run 100% RT or UT because the consequence of failure is catastrophic. The rule of thumb: higher consequence means higher inspection frequency and more redundant methods stacked together.

The Paperwork Behind the Weld: WPS, PQR, and Welder Qualification

Inspection verifies reality; documentation proves it was planned. A procedure qualification record (PQR) captures the actual variables and test results of a qualified weld. From that, an engineer writes a welding procedure specification (WPS) telling the welder exactly what to do. Then welder performance qualification under ASME BPVC Section IX (or ISO 9606-1) verifies that a given welder can produce sound welds to that qualified procedure. Requalification is triggered by process or essential variable changes — a detail many shops miss. Building those skills is a long game, and our guide to welding practice covers how real competence is developed.

Horrible welds don’t get caught by opinion. They get caught by a documented procedure, a calibrated inspector, the right NDT method for the failure mode, and paperwork that proves the whole chain was controlled. Without those, “it looks fine” is the most expensive sentence in the shop.

The Real Cost of Horrible Welds: Rework, Schedule, Safety, and Liability

Ask a shop owner what a rejected weld costs and you will usually get an hourly rate. Ask a finance director and you get something else entirely: lost capacity, slipped milestones, doubled inspection, and a customer who suddenly samples every shipment. The real price of horrible welds lives in the gap between those two answers, and that gap is where quality budgets are either funded or quietly cancelled.

None of it lands as a single line item. It accumulates, joint by joint, until someone finally adds it up.

Direct Costs: Every Rejected Weld Is Paid For Twice

The visible bill starts at the joint itself. Excavating a defect almost always takes longer than making the weld did the first time, because you are grinding or gouging to sound metal, cleaning the cavity, re-fitting, and welding it again under worse conditions: restricted access, preheated steel, and a supervisor watching the whole time.

• Labor hours: excavation, cleanup, re-weld, plus the inspection and paperwork that trail every repair.
• Consumables: grinding discs, gouging carbons, filler metal, shielding gas, and the contact tips and nozzles that wear faster on rework.
• Re-inspection: repaired joints are typically re-tested to the original acceptance criteria, sometimes at 100% even when the rest of the lot was sampled.
• Repeat NDT charges: radiography and ultrasonic testing are billed per joint and per visit, so a second shoot costs nearly as much as the first.

When the defect is something like an overlap welding defect, the excavation runs deeper, heat input climbs, and the odds of introducing a second flaw go up. Rework has a habit of begetting more rework.

Indirect Costs That Never Make It Onto the Job Ticket

• Schedule slippage: rework consumes bay time that was already promised to the next job.
• Overtime: recovery is usually bought with weekend shifts at premium rates.
• Customer containment: if suspect parts shipped, you may pay to sort, re-test, or recall them on the customer’s floor.
• Lost capacity: every hour spent repairing is an hour not earning, while fixed costs keep ticking.
• Reputation and payment: sustained nonconformance stalls approvals, tightens customer surveillance, and can delay the certificates that release payment.

A single rejected pressure joint can hold an entire hydro test, and that test may hold an entire shipment. That is the part finance notices even when the shop floor does not.

The Safety Bill Nobody Budgets For

Rejected welds mean repeat grinding and welding, and each of those operations adds fume, noise, burn risk, and ergonomic exposure to the workforce. Welders absorb metal fumes, ultraviolet radiation, hot spatter, and repetitive strain during the extra hours a repair demands, and NIOSH exposure limits make clear those risks scale with time on the tool, not with how good the finished bead looks. Then there is the far heavier cost: a weld that fails in service does not simply destroy hardware, it can injure or kill the people who trusted it.

What Defect Reduction Is Actually Worth

The encouraging part is that repair rate is measurable, and measurable problems can be turned into dollar values.

Those results come from the kind of systematic study published in welding rejection-rate research, and the pattern is consistent: defect reduction is not a cost center. It is a profit project with a faster payback than most capital equipment.

Compliance, Audits, and Legal Exposure

Repeat nonconformance is a certification problem, not just a production problem. Sustained defect rates can trigger customer audits, jeopardize welder and procedure qualifications, and put approved-supplier status at risk. And if a weldment fails in service while the records show documented nonconformance that was never corrected, the conversation stops being about warranty and starts being about liability.

Repairing a Horrible Weld: Procedures That Actually Hold

Repair is where a lot of shops quietly make things worse. The defect gets ground a little, covered with a fast pass, and pushed out the door — and six months later it comes back as a crack, a leak, or a rejected shipment. A repair that actually holds follows a fixed sequence, and cheating the order is how you turn one bad weld into three.

Step 1: Stop, Quarantine, and Characterize the Defect

Before anyone picks up a grinder, pull the part from production and tag it. Then characterize the defect: what type is it, how big, where exactly, and in which orientation relative to stress? A linear indication running transverse to a tension load is a very different problem from scattered porosity in a low-stress web. Orientation tells you whether the flaw is likely to grow.

This is also the moment to fix the measurement system. In one controlled porosity project, the team couldn’t trust its own defect counts until gage R&R was brought from 41% down to 14.2% — only then did the reduction work mean anything. If two inspectors can’t agree on what “reject” looks like, no repair data is real.

Step 2: Find the Cause Before You Fix the Weld

Repairing a symptom without touching the cause just reschedules the failure. The documented causes of failed lack-of-fusion repairs are improper work angle, incorrect travel angle, and parameters set too cold for the joint. If those stay the same, the re-weld fails the same test. Confirm the procedure, the settings, and the consumable before the arc restarts.

Step 3: Excavate to Sound Metal

Excavation means grinding, gouging, or arc-air removal of the defect, and “to sound metal” is the operative phrase. Remove the visible surface only, and the tail of a crack or a pocket of porosity sits just under your new bead. The standard repair for porosity and lack of fusion is grinding out the affected area and re-welding once the root cause has been resolved. For lack of fusion at a fillet weld root, that means complete removal of the weld down to the root, followed by re-welding with correct technique and parameters. Bring the joint back toward its original profile so the replacement pass isn’t simply filling a canyon.

Hydrogen Control: Dry, Warm, and Patient

Cracks and porosity often trace back to moisture. Damp electrodes and stored-wet flux are direct contributors to hydrogen cracking and porosity, so low-hydrogen electrodes come out of a holding oven, not a damp toolbox. Preheat slows the cooling rate on crack-sensitive steels, and post-heat or controlled cooling lets hydrogen diffuse out before it embrittles the heat-affected zone. Filler selection matters too: match the base metal and the procedure, and don’t substitute whatever is on the shelf. Getting the current and electrode handling right is part of the same discipline — the numbers for common sticks are covered in this guide to 3/16 E7018 amperage settings and technique.

Why “Wash Over” Repairs Fail

A cover pass over an unfused or cracked area hides the flaw under sound-looking metal. The underlying discontinuity is still there, now shielded from casual inspection and growing under load. Codes like AWS D1.1 and ISO 5817 treat cracks and lack of fusion as unacceptable for exactly this reason — they don’t heal, they propagate. Grinding it and welding over it doesn’t fix a joint; it conceals one. Similar logic applies to surface flaws like overlap, where the mechanical notch survives any cosmetic bead laid on top.

Re-Weld, Re-Inspect, and Know When to Scrap

After re-welding, inspect to the same criteria that rejected the part in the first place. Many codes limit repeated repairs or require engineering approval, because each thermal cycle degrades the surrounding metal. A second or third repair in the same spot is a warning sign, not routine.

Remember the hidden bill: grinding plus re-welding adds fume, noise, and ergonomic load, all documented occupational hazards per OSHA and NIOSH. That’s why protective eyewear and ventilation aren’t optional on rework. Sometimes the cheapest, safest move is scrapping the part and rebuilding it clean — especially when the defect is deep, the geometry is thin, or the service is critical. A $200 part is always cheaper than a failure in the field.

Which Processes Produce the Most Horrible Welds? SMAW vs. GMAW vs. FCAW vs. GTAW vs. SAW

When a part fails inspection, the finger usually points at the welder. Often the process is just as guilty. Like weld tube fittings, the fusion method must match the joint. Some processes invite certain defects more than others, so process selection is one of the best defenses against horrible welds.

1. Shielded Metal Arc Welding (SMAW/Stick)

Portable and affordable, stick is where many welders learn welding practice. With correct electrode settings, such as the 3/16 E7018 amperage, it produces sound, code-compliant welds. Its flux-coated electrode makes slag inclusions the biggest weakness, though. Skip interpass cleaning and that slag becomes a defect. Poor electrode storage also introduces hydrogen, leading to delayed cracking.

2. Gas Metal Arc Welding (GMAW/MIG)

Done right, MIG welding is efficient, and its historical rise came from high deposition rates. But shielding gas is the weak point. Lose coverage, damage contact tips, or weld in a draft and porosity appears. A study measured GMAW fume at a median particle diameter near 149 nm, and NIOSH research estimates 60% greater total lung deposition than SMAW at equal concentrations, with about three times the particle surface area. Those ultrafine particles reach the alveoli, so respiratory protection is non-negotiable.

3. Flux-Cored Arc Welding (FCAW)

Compared in pieces like “Is Flux Core the Same as MIG?”, FCAW delivers the convenience of gasless MIG welding, and self-shielding suits outdoor work. The trade-off: slag inclusions return when cleaning is neglected, and correct polarity for flux core is essential. FCAW fume particles run around 352 nm — larger than GMAW, yet still under one micron and inside the respirable fraction as noted by NIOSH.

4. Gas Tungsten Arc Welding (GTAW/TIG)

TIG is unmatched for precision on specialty material, from DC welding aluminium to stainless. Getting TIG polarity and the AC or DC question right prevents its trademark flaws: tungsten inclusions and sugaring. It is also the most skill-sensitive process here; steady hands are needed to avoid overlap and lack of fusion.

5. Submerged Arc Welding (SAW)

Built for heavy plate and long seams, SAW is a powerhouse, and automation such as robotic submerged arc welding boosts repeatability. Flux handling is the weak link: contaminated, damp, or poorly reclaimed flux brings inclusions and porosity. It is not beginner-friendly, but in controlled code work its defect rates can be very low.

Beyond the Big Five

Beyond these five, resistance welding suffers expulsion and laser welding demands immaculate fit-up. In high-mix automation, repeatability problems in robotic cells still cause fusion defects, which matters for intricate work like the Zapp Plus 2 jewelry welder. Whatever the technology, fume and safety oversight should never slip, especially for at-risk welders covered in welding while pregnant guidance.

At-A-Glance: Defect Profiles

Industry data consistently lists porosity among the most common weld defects. Add that fume particles are typically under one micron per NIOSH, and the verdict is clear: the process isn’t the welder’s victim, nor the reverse. It is a system choice, and choosing wisely — alongside resources like how to use a wire welder — prevents horrible welds before the first arc is struck.

The Prevention Playbook: Common Mistakes That Create Horrible Welds, and Final Thoughts

Every defect in this guide has a cause, and every cause has a control. The difference between a shop that fights porosity every week and one that rarely sees a rejection isn’t talent or luck. It’s a system. Here’s the one worth building, followed by the mistakes that quietly undo it.

The 12-Point Prevention System

1. Qualify the procedure and the welder before production. A written WPS backed by a PQR, plus welder performance qualification, is the baseline. Standards like AWS D1.1 and ISO 5817 exist precisely so nobody has to guess what “good” means.
2. Clean the joint like it matters, because it does. Mill scale, oil, paint, rust, and moisture are porosity and cracking fuel. A minute with a grinder beats an hour with an arc gouge.
3. Store and handle consumables correctly. Electrodes in heated ovens, flux kept dry and sealed, wire spools covered. Don’t assume the package protects it — check the basics on MIG welding wire composition and shelf life.
4. Control shielding gas with flow meters at the torch, not just the regulator. A failed regulator diaphragm creates gas-flow decay you can’t see from the bench. Measure where the gas actually arrives.
5. Set and verify parameters with written settings, not memory. Voltage, amperage, travel speed, and polarity change with thickness and position. Keeping a card at the machine removes a whole class of error.
6. Jig and fixture to hold torch angle and joint alignment. Fixed tooling holds what hands cannot, and it prevents the internal high-low misalignment that hides incomplete penetration. For power source and setting fundamentals, see the complete MIG setup guide.
7. Build wind and draft protection. In a documented defect-reduction project, inspecting and correctly using wind protection was the single top-rated solution for reducing weather-driven weld defects.
8. Train and re-test, including eyesight checks. Standard welder training and annual eyesight testing were among the highest-rated solutions in that same project. Protect your vision in the process with the right welding eyewear, and keep skills sharp through deliberate welding practice.
9. Inspect early and inspect deliberately. Visual checks before, during, and after welding catch most problems cheaply. Calibrated boundary samples keep judgement consistent between inspectors.
10. Track defects by type and location with Pareto analysis. A chart tells you which three problems cause most of your scrap — usually far fewer than you’d guess.
11. Fix the cause, not the symptom. Use root-cause methods proven in welding: fishbone diagrams, Pareto charts, FMEA, five-why analysis, and PDCA or DMAIC cycles.
12. Treat repair rate as a managed KPI. Welding rejection rate and NDT backlog turn quality into numbers leadership can act on, proving improvement with data instead of opinion.

Common Mistakes That Keep Horrible Welds Coming Back

Blaming the welder first. Welder skill, shielding gas, and welding conditions are consistently the three highest-impact variables in welding quality studies — and only one of those belongs to the person under the hood. When training and gas supply are weak, no amount of pressure on the operator fixes the outcome.

Welding over cracks. Cover passes hide a flaw instead of repairing it. The crack propagates through the new metal and returns bigger.

Skipping the root cause. Re-welding the same joint with the same settings, gas, and consumables predictably reproduces the same defect. Fix the variable, then repair the weld. Poor ventilation during all that rework also raises fume exposure — the OSHA welding hazards guidance covers controls that belong in every shop’s plan.

Using “looks good” as a spec. Appearance is not acceptance. Every code defines measurable limits, and a nice-looking bead can still fail ultrasonic testing. Set up a safe, purpose-built workspace — see this guide on planning and setting up a welding shop — so inspection and control have a home.

Final Thoughts

Marcus didn’t lose 19 joints to incompetence. He lost them to a regulator he’d been told was fine, a measurement nobody took, and a system that let a hidden variable run unchecked for six months. That’s the real lesson behind horrible welds: they’re almost never a single person’s failure. They’re the output of a process, and processes can be measured, corrected, and held steady.

Fix the gas, clean the joint, qualify the procedure, inspect early, and chase causes instead of blame. Do that consistently, and the ugliest welds in your shop become the ones you never had to make.

Frequently Asked Questions

What is the most common cause of horrible welds?

Across multiple published case studies, porosity is the single most frequently reported weld defect—accounting for roughly a third to over 40% of rejections in boiler and piping projects. Its usual root causes are contamination (oil, paint, moisture, rust) and disruption of shielding gas or flux coverage. In practice, poor cleaning and shielding-gas control cause more rejections than a lack of hand skill.

Can a horrible weld be repaired, or does the part need to be scrapped?

Most defective welds can be repaired by excavating the full extent of the defect to sound metal, fixing the root cause, and re-welding to the original procedure and acceptance criteria. However, some codes limit repeated repairs, and heat-sensitive or highly restrained joints may need to be scrapped because re-welding increases cracking risk. Grinding over a defect and covering it with a new pass is not a repair—it hides a flaw that will surface later.

Is a good-looking weld automatically a good weld?

No. Visual appearance only confirms the weld’s surface condition. Internal defects such as lack of fusion, incomplete penetration, slag inclusions, and internal porosity can exist under a perfectly uniform cap pass and are only found with ultrasonic or radiographic testing. Codes like AWS D1.1 and ASME B31.3 allow visual acceptance of some surface conditions precisely because internal inspection methods carry the real burden of proof.

What is the difference between a discontinuity and a defect?

A discontinuity is any interruption in the typically uniform structure of a weld, such as a pore, an inclusion, or a small undercut. A defect is a discontinuity that exceeds the acceptance criteria for the specific application. That means the same imperfection can be acceptable in a decorative railing and unacceptable in a pressure vessel, which is why the applicable code and service conditions must be defined before production starts.

How do inspectors find defects that cannot be seen?

Volumetric and surface NDT methods do the work: liquid penetrant and magnetic particle testing find surface and near-surface flaws, ultrasonic testing detects internal planar flaws like lack of fusion and cracking, and radiographic testing detects internal porosity, inclusions, and penetration issues. Which method is used is usually dictated by the fabrication code and a risk-based inspection plan.

Which welding process produces the cleanest welds?

GTAW (TIG) generally produces the cleanest, most controlled weld with the least spatter and fume, which is why it dominates thin-wall stainless, aerospace, and critical root passes. That does not make it defect-free—tungsten inclusions, sugaring from shielding loss, and lack of fusion are its characteristic failures. Process choice should match material, thickness, position, code requirements, and fume exposure rather than reputation alone.

Can I weld over rust, mill scale, or paint to save time?

You can, but you will almost certainly produce a defect. Rust, mill scale, oil, paint, and galvanized coatings break down under arc heat and introduce hydrogen, oxygen, and other contaminants into the molten pool, which produces porosity and raises cracking risk. Cleaning the joint takes seconds; chasing a rejection through grinding, re-welding, and re-inspection costs hours.

Why does welding fume matter when talking about weld quality?

Because bad welds create more fume. Every rejected weld must be ground out and re-welded, doubling or tripling the airborne exposure for the same finished part. OSHA regulates welding fume constituents closely—chromium(VI) has a permissible exposure limit of 5 micrograms per cubic meter as an 8-hour time-weighted average—and IARC has classified welding fumes as carcinogenic to humans. Better weld quality is genuinely a health intervention.

What is a normal weld repair rate, and why should a shop track it?

Reported rates vary widely by industry, but many fabrication operations target a non-compliance or repair rate around 2% to 5%, while studies have documented real projects running from roughly 3% to over 22% before improvement work began. Tracking rejection rate by defect type and location turns quality into a measurable KPI and lets teams target the vital few causes instead of guessing.