Broken Weld: Why Welds Crack and Fail, How to Inspect Them, and How to Fix It Right

A fabricator once told me about a weekend job he was proud of: a handrail welded to a shop stair, ground smooth, painted, done. Six months later a worker leaned on it and the whole rail broke away at the weld — one clean, brittle snap that had been growing invisible for months. That is the unsettling truth about a broken weld: it rarely announces itself until the joint is already carrying no load at all.

A Handrail That Gave Way: What a Broken Weld Actually Means

A fabricator once told me about a weekend job he was proud of: a handrail welded to a shop stair, ground smooth, painted, done. Six months later a worker leaned on it and the whole rail broke away at the weld — one clean, brittle snap that had been growing invisible for months. That is the unsettling truth about a broken weld: it rarely announces itself until the joint is already carrying no load at all.

If you have landed here, you are probably in one of three situations. You found a crack, a separated joint, or a suspicious ring around a weld bead, and you want to know what it means. You are worried about safety. Or you need to fix it in a way that passes inspection and actually holds. All three bring you to the same question: why did this weld break, and what do I do now?

This article answers that completely — failure modes, root causes, inspection, repair to code, and prevention.

Is Every Flaw a Broken Weld? Not Quite

Before diagnosing anything, it helps to get the vocabulary straight, because the welding world draws a line most people outside it do not see.

A discontinuity is any interruption in the normal structure of a weld. Porosity, a small overlap, a slightly undersized bead — these are discontinuities. A defect is a discontinuity that fails to meet the acceptance criteria of the governing code. A broken weld is one that has lost its ability to do its job: cracked through, separated from the base metal, or deformed under load it was designed to carry.

The distinction matters. A small amount of porosity in a decorative gate weld may be perfectly acceptable; the same porosity in a pressure-bound pipe weld may be cause for rejection. Context, code, and service determine the verdict. As the American Welding Society puts it, not every flaw is a failure — but some flaws always are.

How a Broken Weld Shows Itself

Failures do not all look the same. Knowing the common presentations helps you describe what you are seeing to an inspector or contractor:

• Hairline cracks. Fine lines on the weld surface, often no wider than a strand of hair. They may run along the bead (longitudinal), across it (transverse), or radiate from the crater at a stop point. Under bright light and magnification they read as sharp, dark lines — not smooth ripples.
• A ring around the bead. That halo tracing the edge where the weld meets the base metal is often a toe crack or a sign of lack of fusion. It can look like nothing more than a shadow or a paint crack until you wipe it clean.
• A joint that simply separated. Sometimes the bead stays intact but detaches from the base metal, or the joint snaps apart entirely. This is common in brittle failures and fatigue, where a crack grew internally until the remaining metal could not hold the load.
• Leaking or weeping. In pipe and pressure work, the first symptom is often not visual at all — a hiss, a fizzing bubble test, or a slow seep that was never there before.
• Deformation or movement. A rail that flexes where it should not, a bracket that has shifted, a gap that was not there when the part was built. Metal does not move without a reason.

The unifying point: a broken weld is a symptom, not a standalone event. The crack you can see is the final chapter of a story that began with design choices, heat input, filler selection, fit-up, cooling rates, and load cycles. Weld it again the same way, and you will get the same result.

That is why the causes matter. The American Welding Society’s Inspection Trends feature on weld cracking from May 2024 identifies thermal stress, hydrogen embrittlement, rapid cooling, high restraint, and brittle microstructures as the common drivers behind cracking — five forces, all invisible in the finished bead.

Why “Broken Weld” Is a Safety Search, Not Just a Repair Search

People search for a broken weld because something is wrong right now, and the stakes feel high for good reason. Welding, cutting, and brazing carry documented hazards — OSHA identifies metal fume exposure and ultraviolet radiation among them — which means the repair itself is hazardous work, even before you consider the failed structure’s instability. A rail that already gave way can give way again while you are grinding on it. Plan the repair as hazardous work, not a quick fix.

Our guides on welding PPE and MIG welding safety precautions cover the gear and practices that keep repair work safe.

Here is what comes next, in order:

1. The failure modes — every way a weld breaks, and how each one looks.
2. The real root causes — metallurgical, procedural, design, and human.
3. Where broken welds show up in the real world, and what they cost.
4. How to inspect, from a flashlight to full nondestructive testing.
5. How to repair the weld the right way, step by step, to code.
6. Codes, documentation, and the legal side of weld repair.
7. Prevention, repair-versus-replace decisions, and a final checklist.

Start by reading the failure in front of you accurately — because everything downstream depends on getting that diagnosis right.

How Welds Actually Break: The Failure Modes You Need to Recognize

When you are standing next to a damaged joint, the question is not “how do I cover this up”, but “how did this weld lose its strength?”. A broken weld is almost never the result of one blow. Instead, it is the end point of a crack, a defect, or a gradual loss of material that grew until the remaining metal could no longer carry the load. For those learning the trade or inspecting their first failure, it helps to break these down into two categories: the cracks that split the metal, and the other defects that allow those splits to begin.

Crack Families: The Splits That Spell Trouble

Cracks are the most dangerous failure mode, as they concentrate stress and can propagate without much warning. While no two breaks look identical, welders tend to run into the same families:

• Longitudinal cracks: Run roughly parallel to the weld bead. These often point to high shrinkage stresses or an overly large weld for the joint.
• Transverse cracks: Cut across the weld, perpendicular to its length. They are frequently tied to tensile stress pulling the joint apart.
• Crater cracks: Form at the very end of a weld bead where the puddle shrank. This is an easy one to prevent with proper crater filling, a skill taught in guides such as Flux Core Wire Welding Tips: Pro Techniques for Cleaner Stronger Welds.
• Toe cracks: Start at the edge (the toe) of the weld where it meets the base metal. These are heavily influenced by undercut, which creates a sharp groove for a crack to initiate from.
• Root cracks: Begin at the root of the joint, the hardest area to see and to access. They can go undetected until the piece is under full load.
• HAZ cracks: Appear in the Heat-Affected Zone (HAZ) next to the weld. These are often tied to rapid cooling, restraint, and trapped hydrogen.

Perhaps the biggest split to know is time-based: hot cracks form as the metal solidifies, while cold cracks (also called delayed or lag-time cracks) can form hours, or even days, after welding. As noted by the American Welding Society, cold cracks are associated with lack of preheat, high restraint, low temperature, and high hydrogen content in the weld deposit. These factors can be controlled with the correct procedure, especially when working with low-hydrogen electrodes or code-driven work like pipe welding procedures.

Non-Crack Defects That Lead to Breakage

Not every broken weld starts as a “crack”. Many start as discontinuities that weaken the cross-section or notch the metal. The most common culprits are porosity (gas pockets), slag inclusions (trapped flux), lack of fusion (unbonded metal), undercut (a groove at the toe), overlap/cold lap (weld metal rolled onto the surface without fusing), and undersized welds (too little reinforcement to carry the load).

These issues are not minor cosmetic flaws. Undercut, in particular, acts as a stress concentrator and a common fatigue crack initiation site. When combined with poor access, high restraint, or rushed technique, they set the stage for failure. This is why shop discipline, from safe setup to checking your parameters when running E7018 rods, matters just as much as the weld itself.

Ductile, Brittle, and Fatigue Fractures

Once a weakness reaches its limit, the joint breaks in one of three ways:

• Ductile fracture: The metal stretches, necks, and tears with noticeable deformation. This is the “safer” failure, as it usually gives warning.
• Brittle fracture: A sudden, low-energy snap with little to no necking. Brittle breaks are often triggered by low temperatures, high restraint, or a hard HAZ, and they are the kind that can be catastrophic.
• Fatigue failure: The most common in-service break. Repeated loading causes microscopic cracks to grow over time, often from stress risers. Over enough cycles, what looks intact will separate without yielding. As highlighted in fatigue and fracture research, poor weld design, inadequate procedure, or incorrect technique are primary drivers of structural welding failures.

At the end of the day, a broken weld is not bad luck. It is a symptom of a failure mode. By learning to spot cracks, recognize dangerous defects, and distinguish how a break occurred, you can stop chasing patches and start finding the root cause for a proper, code-compliant repair.

Let me count the current fragment words and figure out how to adjust.

Let me estimate. The current fragment is fairly long. Let me count roughly.

I’ll count carefully.

Paragraph 1: “Every broken weld tells a story, and the ending is always written long before the crack becomes visible. When a joint fails, the instinct is to blame the welder, the rod, or “bad metal.” The truth is messier. Cracks almost never have one cause. They have two, three, or four small decisions stacking up until the joint runs out of ways to cope.” ~65 words.

P2: “Welding engineers group those decisions into four families: metallurgical causes, procedural causes, design causes, and human causes. If you understand all four, you can walk up to a failed joint and diagnose it instead of guessing. As the American Welding Society’s coverage of weld cracking makes clear, cracking is a stress-plus-microstructure problem, and you usually have to address both to stop it.” ~63

H3 ~9
List items: ~85, 33, 30, 30 = ~178
H3 ~8
P: ~35
List: ~35, 27, 45, 45, 20 = ~172
Callout ~40
H3 ~11
P1 ~65
P2 ~70
H3 ~7
P ~70
P ~30
H3 ~4
P ~55
P ~55
P ~40

Total roughly 65+63+9+178+8+35+172+40+11+65+70+7+70+30+4+55+55+40 = 977? Let me sum: 65+63=128; +9=137; +178=315; +8=323; +35=358; +172=530; +40=570; +11=581; +65=646; +70=716; +7=723; +70=793; +30=823; +4=827; +55=882; +55=937; +40=977.

So it’s about 977 words, slightly over 912. Need to trim to between 608-912. Target ~850. So cut ~130 words. Actually safer to land around 830-870.

I need to keep same heading, same facts, all links. Links:
1. AWS weld cracks external
2. weldgallery welding-defects-undercut
3. Hobart Brothers welding fails
4. weldgallery electrodes
5. weldgallery how-to-set-up-a-mig-welder
6. weldgallery spot welder
7. weldgallery welding PPE 2026
8. OSHA welding cutting brazing
9. RapidDirect types of welding defects
10. OneStopNDT acceptance criteria

Keep all 10. Trim filler.

Let me rewrite, condensing some list item explanations and merging where possible, trimming human causes and multi-cause sections.

Let me write with an eye to ~850 words.

Draft:

Why Welds Break: The Real Root Causes Behind Every Failed Joint

Every broken weld tells a story, and the ending is usually written long before the crack becomes visible. When a joint fails, the instinct is to blame the welder, the rod, or “bad metal.” The truth is messier: cracks rarely have one cause. They have two, three, or four small decisions stacking up until the joint runs out of ways to cope.

Welding engineers sort those decisions into four families: metallurgical, procedural, design, and human. Understand all four and you can diagnose a failed joint instead of guessing. As the American Welding Society’s coverage of weld cracking makes clear, cracking is a stress-plus-microstructure problem — you usually have to address both to stop it.

Metallurgical Causes: What the Metal Does When Nobody Is Watching

• Hydrogen embrittlement. The classic cause of delayed cracking, also called cold cracking. Hydrogen sneaks in from moisture in electrodes, oil and rust on the plate, damp shielding gas, or plain humidity. As the weld cools, hydrogen migrates to high-stress zones and makes the steel brittle, so the crack can appear hours or even days later. That’s why same-day visual inspection is never enough. For a deeper dive, see our piece on welding defects like undercut and their acceptance limits.
• Rapid cooling. Weld metal and the heat-affected zone (HAZ) can cool fast enough to form hard, brittle microstructures like martensite. Thin sections, cold base metal, and strong wind all accelerate it.
• Brittle HAZ microstructure. The zone beside the weld never melts, but its grain structure changes permanently. High-carbon and high-strength steels are especially vulnerable.
• High restraint. On heavy plate or rigid fixturing the shrinking weld has nowhere to go, so residual stress climbs toward yield strength.

Procedural Causes: What Happened at the Machine

Procedural causes are the ones you control directly, which makes them the most fixable. The Hobart Brothers technical library’s review of real welding fails keeps circling the same suspects:

• Wrong heat input. Too hot burns out alloying elements and widens the HAZ. Too cold causes lack of fusion — one of the sneakiest crack starters, because it hides under a good-looking bead.
• No preheat. Welding cold, thick, or higher-carbon plate creates thermal shock that drives both hydrogen cracking and rapid-cooling brittleness.
• Contaminated or mismatched filler. Rusty rods or a consumable that doesn’t suit the base metal weaken the joint. Choosing the right electrode prevents a surprising number of failures — our guide to electrodes and how to choose them covers the basics.
• Wrong polarity or settings. Polarity changes penetration and deposition behavior, and the change is often invisible until the joint cracks. Unsure your machine is right? Start with our MIG welder setup walkthrough.
• Travel speed errors. Too fast starves the joint; too slow overheats and over-penetrates it.

Design Causes: The Joint Was Doomed Before the Arc Struck

Some welds break because of geometry, not skill. A weld at the corner of a stiffened joint creates a stress riser that concentrates load exactly where weld meets base metal. Sharp notches, abrupt transitions, and undersized legs do the same. Forced fit-up is another silent killer: tacking a joint closed under heavy gap pressure stores mechanical energy in the weld, and that energy has to release somewhere.

The costliest design mistake is placing a weld in a high-cycle fatigue zone — a bracket that flexes millions of times, a trailer crossmember that vibrates every mile. There the weld toe becomes the crack initiation point, and no amount of skill outruns fatigue. For how joining physics changes with process, our explanation of how spot welders work offers useful context on heat and load transfer.

Human Causes: The Process Nobody Verified

Behind most metallurgical and procedural causes sits a documentation gap: no written welding procedure specification (WPS), no welder qualification, no filler certification check, no interpass temperature control. An unqualified welder can lay beautiful beads that fail anyway, because welding is as much metallurgy as technique. Rushed or unsafe conditions also breed procedural shortcuts, which is why crews should keep up with proper welding PPE and shop safety habits. Per OSHA’s welding, cutting, and brazing guidance, uncontrolled hot work and poor environmental control are recognized hazard patterns in their own right.

The Multi-Cause Reality

Here is the uncomfortable part: a typical broken weld has at least three contributing causes. Add high restraint, skip the preheat, weld with a damp rod on a slightly oily joint, and put that weld in a flexing location. Any one alone might survive. Together, they guarantee failure.

The RapidDirect defect taxonomy and OneStopNDT’s acceptance criteria reference both teach the same lesson: treat a crack as a system symptom, not a single event. When production welds show repeated defects, the responsible move — and in many codes the required one — is to requalify the procedure, not just tell the welder to try harder. Fixing the welder alone buys you a few months before the same crack returns in the same place.

Word count estimate: let’s compute.

P1: ~62
P2: ~60
H3: 9
Metallurgical list: item1 ~72, item2 ~30, item3 ~28, item4 ~27 → 157
H3: 8
P: ~34
Procedural list: ~38, 26, 40, 37, 18 → 159
Callout: ~40
H3: 11
P1: ~63
P2: ~63
H3: 7
P: ~85
H3: 4
P1: ~48
P2: ~78

Sum: 62+60=122; +9=131; +157=288; +8=296; +34=330; +159=489; +40=529; +11=540; +63=603; +63=666; +7=673; +85=758; +4=762; +48=810; +78=888.

Hmm ~888, within 912 but close. Let me trim a bit more to ~860 for safety. Note HTML tags aren’t counted presumably; word count of visible text. Let me trim ~30 words.

Trim P2 intro slightly, trim human causes, trim last paragraph.

Let me finalize carefully, aiming ~850-880.

Actually, counting is tricky. Let me be a bit conservative and trim more, target ~820.

I’ll shorten the metallurgical item 1 a bit? No, keep facts. Trim “also called cold cracking” keep. It’s fine.

Trim final paragraph: “Fixing the welder alone buys you a few months before the same crack returns in the same place.” Remove — saves 18 words. Hmm but it’s a good closing. Keep but shorten: “Fixing the welder alone just buys time before the same crack returns.” (12 words).

Trim human causes paragraph a bit.

Let me just produce and be mindful. Also note duplicate target attributes on internal weldgallery links — original had target=”_blank” rel=”noopener” on internal links too. Keep same as original.

Final count target ~870. Fine.

Let me write it out.

Why Welds Break: The Real Root Causes Behind Every Failed Joint

Every broken weld tells a story, and the ending is usually written long before the crack becomes visible. When a joint fails, the instinct is to blame the welder, the rod, or “bad metal.” The truth is messier: cracks rarely have one cause. They have two, three, or four small decisions stacking up until the joint runs out of ways to cope.

Welding engineers sort those decisions into four families: metallurgical, procedural, design, and human. Understand all four and you can diagnose a failed joint instead of guessing. As the American Welding Society’s coverage of weld cracking makes clear, cracking is a stress-plus-microstructure problem, and you usually have to address both to stop it.

Metallurgical Causes: What the Metal Does When Nobody Is Watching

• Hydrogen embrittlement. The classic cause of delayed cracking, also called cold cracking. Hydrogen sneaks in from moisture in electrodes, oil and rust on the plate, damp shielding gas, or plain humidity. As the weld cools, hydrogen migrates to high-stress zones and makes the steel brittle, so the crack can appear hours or even days later. That is why a same-day visual inspection is never enough. For a deeper dive, see our piece on welding defects like undercut and their acceptance limits.
• Rapid cooling. Weld metal and the heat-affected zone (HAZ) can cool fast enough to form hard, brittle microstructures like martensite. Thin sections, cold base metal, and strong wind all accelerate it.
• Brittle HAZ microstructure. The zone beside the weld never melts, but its grain structure changes permanently. High-carbon and high-strength steels are especially vulnerable here.
• High restraint. On heavy plate or rigid fixturing, the shrinking weld has nowhere to go, so residual stress climbs toward the metal’s yield strength.

Procedural Causes: What Happened at the Machine

Procedural causes are the ones you control directly, which makes them the most fixable. The Hobart Brothers technical library’s review of real welding fails keeps returning to the same handful of suspects:

• Wrong heat input. Too hot burns out alloying elements and widens the HAZ. Too cold produces lack of fusion, one of the sneakiest crack starters because it hides under a good-looking bead.
• No preheat. Welding cold, thick, or higher-carbon plate creates a thermal shock that drives both hydrogen cracking and rapid-cooling brittleness.
• Contaminated or wrong filler. Rusty rods or a consumable that doesn’t suit the base metal weaken the joint. Picking the right electrode in the first place prevents a surprising number of failures — our guide to electrodes and how to choose them covers the basics.
• Wrong polarity or machine settings. Polarity changes penetration depth and deposition behavior, and the change is often invisible until the joint cracks. If you’re unsure your machine is set up correctly, start with our MIG welder setup walkthrough.
• Travel speed errors. Too fast starves the joint; too slow over-penetrates and overheats it.

Design Causes: The Joint Was Doomed Before the Arc Struck

Some welds break because of geometry, not skill. A weld placed at the corner of a stiffened joint creates a stress riser that concentrates load exactly where weld meets base metal. Sharp notches, abrupt transitions, and undersized legs all do the same thing. Forcing fit-up is another silent killer: tacking a joint closed under heavy gap pressure stores mechanical energy in the weld, and that energy has to release somewhere.

The most expensive design mistake is putting a weld in a high-cycle fatigue zone — a bracket that flexes millions of times, a trailer crossmember that vibrates on every mile. In those spots the weld toe becomes the crack initiation point, and no amount of welding skill can outrun fatigue. For how the physics of joining changes with process, our explanation of how spot welders work offers useful context on heat and load transfer.

Human Causes: The Process Nobody Verified

Behind most metallurgical and procedural causes sits a documentation gap: no written welding procedure specification (WPS), no welder performance qualification, no filler certification check, and no interpass temperature control. An unqualified welder can produce beautiful beads that fail anyway, because welding is as much metallurgy as technique. Rushed or unsafe conditions also breed procedural shortcuts, which is why crews should stay current on proper welding PPE and shop safety habits. Per OSHA’s welding, cutting, and brazing guidance, uncontrolled hot work and poor environmental control are recognized hazard patterns in their own right.

The Multi-Cause Reality

Here is the uncomfortable part: a typical broken weld has at least three contributing causes. Add high restraint, skip the preheat, weld with a damp rod on a slightly oily joint, and place that weld in a flexing location. Any one of those alone might survive. Together, they guarantee failure.

The RapidDirect defect taxonomy and OneStopNDT’s acceptance criteria reference both teach the same lesson: treat a crack as a system symptom, not a single event. When production welds show repeated defects, the responsible move — and in many codes the required one — is for the engineer to requalify the procedure, not just tell the welder to try harder. Fixing the welder alone simply buys a few more months before the same crack returns in the same place.

Where Broken Welds Show Up: Real-World Scenarios and What They Cost

Every broken weld has a story. Joints rarely fail under a controlled lab test; they fail on a trailer hauling a skid steer down the highway, on a handrail someone leans on, or inside a pipe no one inspects until it leaks. Knowing where these failures happen, and how they surface, is the fastest way to build the habit of spotting them early in your own work.

Scenario 1: The Cracked Trailer Frame Weld

Trailer frames live in fatigue hell. Every bump cycles the welds, and the stress concentrations at crossmember-to-rail joints are brutal. A fabricator I know found a crack in his dovetail ramp after one season by the simplest sign imaginable: paint flaking off in a thin line, with rust bleeding out behind it. Steel doesn’t rust from the surface like that unless something underneath is moving. That hairline was a fatigue crack walking through a fillet weld, invisible until the coating gave it away.

This is the classic fatigue story: a weld can look perfect on the surface and still be cracked internally or at the root after repeated loading. Fatigue dominates failures in cyclic service, which is why trailer, equipment, and vehicle welds deserve more inspection attention than static structures. If you haul with it, weld it with detail — proper leg sizes, smooth toes, no undercut. Our guide to undercut causes and acceptance limits explains why that one defect is so destructive in cyclic loading.

Scenario 2: The Handrail or Gate That Lets Go

Handrails and gates are the most common broken welds people actually see, because the loading is human and sudden. A gate sags; the latch stops catching; the post wobbles. Look closely and you’ll usually find a cracked root or a crater crack at the end of a bead where the welder pulled the torch away without filling the crater. On gate hardware, spot welds and short tack-heavy joints add another failure path — our explainer on how spot welders work covers why those joints fail differently than continuous fillets.

The consequence ranges from nuisance (re-weld a gate on a Saturday) to serious injury if a rail lets a worker fall from height. OSHA’s welding, cutting, and brazing hazard guidance treats load-bearing welded assemblies as safety-critical for exactly this reason.

Scenario 3: The Leaking Pipe Weld

Pipe welds announce themselves differently: they leak. Sometimes it’s a weeping pinhole found during a walkthrough; sometimes it’s a hiss that turns into a shutdown. Corrosive or high-temperature service makes it worse — a lack-of-fusion defect that would be tolerated in a handrail becomes a leak path in a pressure line. That’s why pipe work lives under strict procedures; see pipe welding procedures and inspection for how WPS, positions, and NDT fit together on real piping.

Scenario 4: Roll Cages and Suspension Brackets

Here the discovery method is often a noise — a tick, a clunk — or a visual check after an event. Suspension brackets crack at the weld toe first, then propagate until the part separates. Roll cages fail the same way, usually at a joint that was welded cold, with no preheat, on thin-wall tubing. Because these parts protect lives, the repair standard isn’t “looks good,” it’s “proven.” That means NDT, not eyeballs, and it means knowing rod behavior cold. If your shop runs 7018 on these joints, our 3/32 E7018 amperage guide and 5/32 E7018 settings guide cover the parameters that keep hydrogen low and toes wetted in.

Scenario 5: Structural Steel Cracking

On buildings and bridges, cracks often show up during scheduled inspection rather than dramatic failure — an inspector marking a crack at a beam-to-column connection, then tracing it back to a weld toe. AWS’s inspection trends coverage of weld cracks is worth reading for how these defects behave in real structures.

Scenario 6: Pressure Vessel Repairs

The most consequential broken welds live in boilers and pressure vessels. The National Board notes that ASME Section I requires full-penetration welds, and radiography isn’t always practical because of pipe diameter, wall thickness, or access. Pressure vessel weld repairs depend on the location and nature of the defect within the vessel, and that decision drives the whole inspection and documentation path. Worse, uncertified repairs can void compliance and force unplanned shutdowns, turning a small weld job into a site-wide cost event.

Notice the pattern: the cheaper the discovery method, the earlier the catch. A visual check costs nothing; a forced shutdown costs everything. That’s the argument for building inspection into every repair — and it’s a habit best learned before you ever strike an arc, starting with basics like proper welding PPE and clean, controlled setup. The next section walks through exactly how to inspect a suspected broken weld, from your eyes to full NDT.

How to Inspect a Suspected Broken Weld: From Visual Check to NDT

Finding a crack is one thing. Proving how far it runs is another. A weld can look fine on the surface and still be split deep at the root, which is why inspection is a sequence rather than a single glance. The good news: you don’t need a lab truck on day one. Almost every professional inspection starts the same way, with your own eyes and a good light.

Step 1: Visual Inspection (VT) — The Mandatory First Move

Visual testing isn’t the “cheap” option; it’s the required one. VT is the first inspection performed on every welded joint, and it precedes every other method. If an inspector is going to run ultrasonic testing on your pipe weld, they will still look at it first, because visual examination defines where the other methods need to focus. AWS and ASME practice both treat VT as the baseline acceptance check before any advanced examination is authorized.

Do it properly:

• Clean the joint. Wire-brush away spatter, slag, paint, and rust. You cannot find a hairline crack under a layer of mill scale.
• Use real lighting. A dim shop hides toe cracks. Raking light (shining from a low angle across the bead) throws shadows into tiny fissures and makes them pop.
• Magnify. A 10x inspector’s glass turns a “discoloration” into a legible crack. Many hairline failures are invisible at arm’s length.
• Check the whole joint, not just the bead. Look at the toes, the crater, and the heat-affected zone on both plates.

Acceptance criteria are not a matter of taste. Your governing code (AWS D1.1, ASME BPVC, API 1104) defines what is a rejectable crack versus an acceptable undercut, porosity, or surface profile. When in doubt, or when the joint is structural, pressure-retaining, or safety-critical, bring in a Certified Welding Inspector rather than guessing.

Step 2: Choosing the Right NDT Method

Nondestructive testing splits into two families, and the split matters more than the acronyms.

• Surface methods — Magnetic Particle (MT) and Liquid Penetrant (PT) — find cracks that reach the surface. MT uses magnetic fields and iron particles on ferrous steel; PT uses a dye that seeps into surface-breaking flaws and is then blotted out with developer. Both are fast, relatively inexpensive, and excellent for toe cracks, crater cracks, and fatigue cracking.
• Volumetric methods — Ultrasonic (UT) and Radiographic (RT) — look inside the weld. UT bounces sound waves off internal discontinuities and is superb at finding lack of fusion and root cracks. RT passes X-rays or gamma rays through the joint and produces a film image; it’s the classic method for pipeline and pressure-vessel work. Eddy current testing adds a surface-and-near-surface option, especially useful on non-ferrous or tubing where magnetism won’t work.

For pressure equipment, ASME Section V governs which examination methods are permitted and how they must be applied — an authorized inspector will reference it directly when a repair is signed off. Which method is “required” depends on the code, the service, and the joint, not on what equipment happens to be parked in your yard.

Reading the Results Honestly

Here’s the part most people miss: NDT results are statistical, not absolute. A pass means “no indication of significance was detected by this method, applied this way, by this operator.” Research published in 2022 in the International Journal of Fatigue used Bayesian methods to estimate both probability of detection and true defect size distributions, showing that detection reliability varies with flaw size and method — which is exactly why codes set sensitivity thresholds and require qualified procedures and personnel. If a UT report comes back clean but the joint keeps failing, the answer is often a different method, not a different welder.

Whatever method runs, safety governs the room: arc-flash eye protection, respirators for penetrant and dust exposure, radiation barriers for RT, and proper hot-work PPE. Our guides on what a welder wears and MIG welding safety cover the details, and OSHA’s welding, cutting, and brazing hazard guidance is the standard reference for shop controls. Also check undercut acceptance limits before you misjudge a profile flaw as a crack.

Note: If the joint is loaded, pressure-bearing, or part of a public structure, don’t inspect it yourself and hope. Document, stabilize the load, and escalate to qualified inspection before anyone leans on it again.

How to Repair a Broken Weld the Right Way: Step-by-Step

The instinct when you spot a crack is to grind it flush, run a hot pass over the top, and put the machine back in service. That instinct is exactly how a broken weld gets worse. A crack is not a surface blemish; it is a sharp tip crawling through load-bearing metal, and new weld metal laid across an unremoved crack tip does nothing but hide the evidence while the crack keeps growing underneath.

Repair welding is a documented, qualified process, not a shop favor. AWS D1.1 Clause 6.22 governs how repair welding is performed and what records must document it, and corrective-action records must trace three things: the defect, the disposition, and the qualified procedure used for the fix. The practical takeaway is blunt — a repair weld generally requires the same qualification, preheat, and inspection rigor as the original production weld.

The Repair Sequence, in Order

1. Stop the load. Shore, block, or unbolt the assembly and tag the joint out of service. A cracked member sitting in tension can let go violently the moment the first cut is made.
2. Remove the weld and every crack tip. Cut, gouge, or grind the weld metal out entirely and run past each visible crack end into sound metal. Root cracks often need back-gouging from the reverse side before the root is rewelded. Cracks frequently branch beyond what your eye can follow, which is why AWS has published dedicated guidance on identifying and removing weld cracks.
3. Verify complete removal. Wire-brush the groove, then run magnetic particle (MT) or liquid penetrant (PT) testing over the prepared area before any new metal goes in. Any linear indication means you are not finished cutting.
4. Review the WPS and PQR. Confirm filler classification, joint geometry, welding position, preheat requirement, and the qualified heat-input range before striking an arc.
5. Preheat where required. Measure with a contact thermometer, never by color. Preheat slows cooling, drives off surface moisture, and lowers the hydrogen cracking risk.
6. Control heat input and bead placement. Stay inside the qualified current band, keep passes small and layered, and avoid wide weaving that overheats the heat-affected zone. Getting the settings right matters: 3/32 E7018 amperage and polarity guidance shows how narrow the correct window really is.
7. Manage interpass temperature. Do not lay the next bead until the previous one has cooled below the procedure’s maximum. Rushing passes is one of the most common causes of re-cracking.
8. Apply post-weld heat treatment where specified. Stress relief or a hydrogen bake-out is not optional on heavy, highly restrained, or pressure-retaining joints.
9. Clean, dress, and re-inspect. Remove slag and spatter, blend the toes, and take the joint back through acceptance testing against the governing code.

What Not to Do

• Never weld over a crack. It seals the surface and guarantees the crack resumes under load.
• Never patch without gouging out. A cover plate or “scab” weld adds restraint and stress risers without removing the flaw.
• Never skip cleaning between passes. Slag trapped under the next bead becomes a lack-of-fusion defect waiting to initiate a crack.
• Never substitute filler casually. The wrong classification or a rusty, contaminated rod changes chemistry and toughness. If you are unsure what a consumable code means, start with electrode selection, numbers, and polarity explained.
• Never skip post-repair inspection because the weld “looks good.” Appearance says nothing about subsurface fusion.

On pipework and pressure equipment, the repair follows the pipe procedure and adds coupon and NDT requirements — see pipe welding procedures for WPS, positions, passes, and inspection. The National Board likewise expects documented repair procedures and authorized oversight on pressure-retaining parts, because the paperwork is what proves the joint was fixed to standard rather than merely filled in.

Codes, Standards, and the Legal Side of Weld Repair

The moment a weld cracks, the conversation stops being purely technical and becomes partly legal. Somebody has already written down what “acceptable” means, and it is almost never “it looks fine to me.” Every repair on a handrail, trailer, pressure pipe or pressure vessel is measured against a written standard. Knowing which one applies separates a repair that closes the file from one that opens a new problem.

Which Code Actually Governs Your Repair?

There is no single welding rulebook. The governing document depends on what the part does, where it lives, and who inspects it. The four you will meet most often:

The load-bearing principle in AWS D1.1 acceptance criteria is simple and non-negotiable: cracks are rejected unless the applicable code specifically permits them. There is no “small enough” crack. A hairline indication at a weld toe and a half-inch break get the same verdict — remove it.

ASME’s Section IX plays a different role. It is not an acceptance code; it is the rulebook for qualification — who is allowed to write a procedure, and which welder is allowed to run it. If your repair welder’s performance qualification expired, or covers a different process, position or thickness, the weld may be perfectly executed and still non-compliant on paper.

The Paper Trail: WPS, PQR, WPQR, and Performance Records

Four documents matter, and they stack on each other:

• WPS (Welding Procedure Specification) — the recipe: process, filler classification, polarity, heat input range, preheat, interpass temperature and positions. A detailed walkthrough lives in our pipe welding procedures guide.
• PQR (Procedure Qualification Record) — the evidence the recipe works: a test weld with documented mechanical results supporting every variable the WPS claims.
• WPQR (Welder Performance Qualification) — proof that a specific welder can run that WPS in that position.
• Production and inspection records — what was welded, when, by whom, and what the inspector accepted.

After a failure, these documents are the first thing anyone asks for. They show whether the design intended a low-hydrogen fill, whether the welder was qualified on the actual material, and whether the joint was ever inspected. A repair done without paperwork may pass a visual check today and still void an insurance claim tomorrow. The AWS D1.1 repair documentation requirements are worth reading before you strike an arc, not after.

Who Is Allowed to Declare a Weld Acceptable?

On structural work, that is typically an AWS Certified Welding Inspector (CWI). A CWI verifies preparation, preheat, bead placement, final profile and dimensions, then either accepts the weld or documents the reject. Destructive and volumetric examination follow the methods defined by the standard itself, which for ASME work means Section V’s examination techniques.

An inspector is not there to police you. A good inspection regime catches a root crack while it is one inch long, before it becomes a structural story. If inspection is new to you, our piece on undercut acceptance limits shows how acceptance criteria are written and applied in practice.

R-Stamps, Authorized Inspectors, and Jurisdiction

Pressure equipment changes everything. In most U.S. jurisdictions, an organization repairing boilers and pressure vessels must hold an R-Stamp issued under the National Board’s registration and repair rules. The stamp certifies a quality system, qualified procedures and authorized supervision — not merely skilled welders.

Alongside the stamp, repairs to coded pressure equipment usually require an authorized inspector from an accredited inspection agency to witness the work, review the Manufacturer’s Data Report, and sign off before the item returns to service. Piping carries similar obligations: gas-carrying systems bring purging, testing and documentation well beyond a visual check, which we cover in gas line welding rules.

None of this overrides basic worker protection. Repairs happen in tight, hot, awkward conditions, and OSHA’s welding, cutting and brazing standards set the floor for fire prevention, ventilation and shielding. On grinding and gouging duty, that means full welding PPE, not a compromised setup borrowed from another task.

Preventing Broken Welds: Design, Procedure, and Shop Habits

Every broken weld tells a story, and the ending was written long before the crack appeared. Most failures are designed in, procedurally invited, or humanly enabled, and cracked handrails, split trailer frames, and leaking pipe welds all point the same way.

Joint Design: Where the Battle Is Won or Lost

The strongest prevention step happens on paper, before any metal is cut. A fillet weld at a sharp corner or in a high-cycle fatigue zone loads the weld toe until a crack starts, however good the bead looks. AWS and AISC both stress that detail selection, not weld quality alone, drives fatigue life. Move the joint out of high-stress regions when possible; otherwise grind the toes smooth, blend transitions, and avoid abrupt section changes that act as stress risers.

Leg and throat size matter too. An undersized weld carries more load per unit area than intended and fatiques faster, while oversized legs on thin plate raise restraint and heat input and invite distortion and hydrogen cracking. Follow the code minimums with a sensible margin and resist the urge to beef it up without engineering approval.

Procedure Control: The WPS Is Your Insurance Policy

A written Welding Procedure Specification is not paperwork; it is the recipe that makes a sound weld repeatable. Following the WPS requirements for positions, passes, and interpass temperature keeps heat input inside the qualified range. Cranking amperage up to move faster, running a bigger wire than specified, or skipping preheat on a cold morning moves you outside that envelope, where cracking risk climbs.

Preheat deserves emphasis: slowing the cooling rate gives dissolved hydrogen time to diffuse out of the weld metal before it gathers at stress points and triggers a cold crack, which matters most in restrained thick sections where the base metal acts as a heat sink. Hobart Brothers’ welding fail analyses repeatedly identify skipped or insufficient preheat as a root cause in cracked production welds.

Consumable discipline is the other half of the hydrogen equation. Low-hydrogen electrodes such as E7018, identified by their EXXX-15 and EXXX-18 coating types, are built for low diffusible hydrogen, but their flux coatings are hygroscopic and drink moisture from shop air within hours. Rod ovens and portable quivers at the manufacturer’s recommended temperature keep them dry. Knowing electrode classifications and how to choose the right one prevents swapping a fast-freeze rod for a low-hydrogen rod because it welds nicer. Polarity and amperage matter: an E7018 run cold gives porosity and poor fusion, so proper E7018 amperage settings are procedure discipline, not preference.

For critical work, a hydrogen bake-out after welding drives out residual hydrogen that survived the weld cycle, standard practice for heavy structural and pressure-retaining welds.

Fit-Up, Cleanliness, and Sequencing

Sound welds start with sound joints. Poor fit-up — excessive gaps, mismatched edges, or forced fit-up acting like a built-in spring — creates stress before the arc is struck. Tacks should follow the production procedure; understanding proper tack spacing and settings prevents tacks that crack or fail to hold restraint.

Cleanliness is unglamorous but non-negotiable. Mill scale, rust, oil, and paint produce porosity, slag inclusions, and unstable arcs. Grind and wire brush the joint faces and remove slag fully between passes so inclusions cannot act as crack initiators.

Sequencing matters in restrained assemblies: welding one side completely before balancing the other locks in shrinkage stresses, while balanced, staggered passes let the assembly shrink incrementally. On thick sections, back-gouging the root and welding from the second side ensures full fusion and relieves trapped root stress, a technique borrowed from structural and pressure-vessel practice.

Invest in People: Training and Certification

Procedures only work when the operator follows them, so welders are qualified to the applicable code — AWS D1.1 as referenced in AWS inspection literature, for example — which verifies they can consistently produce sound welds to the written procedure in the positions they actually work. Certification is not a one-time event; continuity requirements mean welders maintain qualification through regular work or re-testing. A trained welder also values proper protective equipment, because comfort at the joint builds weld quality.

Inspection as Prevention, Not Policing

Find defects before they are buried under later passes or hidden under a coating: early visual and non-destructive testing catches undercut, porosity, and lack of fusion while the part is still on the bench and access is easy. Build inspection into the workflow rather than saving it for a final sign-off, and quality control becomes a feedback loop that stops the same defect repeating. Inspectors and welders working as a team rather than adversaries keep defect rates down.

Prevention comes down to sequence: design the joint honestly, follow a qualified procedure, keep metal and consumables clean, manage heat and restraint deliberately, use qualified people, and inspect early. None of it is exotic; it is the difference between a weld that serves for decades and one that lets go.

Repair vs. Replace, DIY vs. Pro: Costs, Comparisons, and Common Mistakes

Once a broken weld is confirmed, the question stops being technical and becomes economic: cut it out and re-weld, or replace the part? Call a certified contractor, or handle it yourself? The answer depends less on the size of the crack than on what the joint does, where it lives, and what happens if it fails again.

Repair or Replace: How to Decide

A repair is reasonable when the base metal is sound, the crack hasn’t extended into the parent material, the part isn’t in high-cycle fatigue service, and the original design can be restored with a qualified procedure. Replacement is the safer choice when:

• The component is fatigue-critical or load-cyclic. Trailer suspension brackets, roll cage nodes, crane lugs, and anything seeing repeated loading are the strongest candidates for replacement. A repaired joint in a fatigue zone often carries residual stress and a heat-affected zone that will crack again sooner than the original did.
• The crack keeps returning. Two failures in the same location is a design or procedure problem, not a welding problem.
• The base metal is contaminated, thinned, or bent. Overheated or gouged-out material can’t be trusted back into service no matter how good the new weld looks.
• The part is cheap relative to the risk. A $60 bracket isn’t worth an afternoon of gouging, NDT, and documentation.

Shop Weld vs. Field Weld, DIY vs. Pro

Moving a part to the shop almost always beats welding it in place: better position, better access, controlled preheat, and the option of NDT afterward. Field welding wins only when the part can’t be removed — pipelines, pressure vessels, building frames — and there the cost of access, rigging, and on-site inspection usually dwarfs the bead itself.

For DIY, the honest line is this: non-critical, non-code, low-load fabrications are fair game for a competent home welder, provided the safety basics are covered, starting with a solid grounding in welding PPE rather than treating it as an afterthought. Anything load-bearing, pressure-containing, or regulated belongs with a qualified welder working to a written procedure. The OSHA welding, cutting, and brazing hazards guidance is a useful reality check on what a field repair legally drags in — fire watch, ventilation, hot-work permits, and more.

What Actually Drives the Cost

Welders assume the arc time is the bill. It almost never is. Repair cost is dominated by four things:

The asymmetry is stark on pressure equipment: an uncertified repair can force a site shutdown, a National Board review, and full re-inspection, dwarfing the original weld cost many times over. The bead was cheap. The consequences weren’t.

The Mistakes That Send Welds Back

Across reported repair failures, the same handful of errors appear again and again:

1. Welding over an existing crack without gouging out the root cause. This is the single most common reason a repaired weld breaks again — the crack tip stays sharp, the original cause stays active, and the new bead just hides the evidence. AWS’s coverage of weld crack causes makes the point plainly: a crack is a symptom, and welding over it treats the picture, not the fire.
2. Skipping preheat on restrained, thick, or hydrogen-sensitive joints.
3. Wrong or contaminated filler — a rusty E7018 held overnight, wrong polarity, or a rod that doesn’t match the base metal. Electrode selection and polarity decisions matter more in repair than in new work, because repair joints are usually worse-fit.
4. No cleaning between passes, leaving slag inclusions that become the next crack origin.
5. Ignoring the governing code and improvising a procedure.
6. No inspection afterward — the repair gets painted and forgotten, exactly how the first failure grew invisible.

The practical takeaway: repair when the metal is sound, the service is gentle, and you can verify the result. Replace when the joint is fatigue-loaded, the failure repeats, or the paperwork cost of proving the repair exceeds the part. And never let a grinder and a paint gun stand in for an inspector.

Key Takeaways and Your Next Step After Finding a Broken Weld

That handrail from the opening story did not fail because of bad luck. It failed because a design decision, a shop habit, and an inspection step each gave a little, and nobody noticed until the joint stopped carrying load at all. That is the most useful idea in this entire article: a broken weld is a system failure, not a standalone event. Metal does not decide to crack on its own. Something in the chain of design, procedure, fit-up, welding, or inspection opened a door.

The good news is that the same logic works in reverse. If a break is a system failure, then fixing the system is what actually repairs it. A bead laid over the top of a crack treats the symptom. Everything below is about treating the cause.

Your Six-Step Sequence After Finding a Crack

1. Stop the load. Take the joint out of service, block the area, and remove pressure, weight, or traffic from the member. Nothing else on this list matters if the crack is still growing while you plan.
2. Verify what you have. Clean the area, use good light and magnification, then confirm the full extent with NDT. Any crack found in a structural weld should be treated as rejectable until it has been fully removed, repaired, and re-inspected, which is exactly how AWS inspection guidance frames crack acceptance.
3. Identify the governing code. Structural steel points to AWS D1.1, pressure equipment to the ASME BPVC and National Board repair rules, pipelines to API 1104. The code tells you what is allowed before you strike an arc.
4. Use a qualified procedure and a qualified welder. Correct filler, correct preheat, correct heat input, correct interpass temperature, written down and followed. If you need a refresher on what a WPS actually controls, our pipe welding procedures guide walks through it step by step.
5. Re-inspect. Repeat visual inspection and the NDT method the code calls for. A repair that was never re-checked is simply a new, undocumented crack waiting for a second appearance.
6. Document everything. Procedure, welder ID, consumable batch, preheat readings, NDT reports, and inspector sign-off. A complete paper trail protects the owner during audits, resale, and insurance claims, as AWS D1.1 repair documentation practice makes clear.

The Benchside Checklist

• Is the joint unloaded and the area secured?
• Has the crack been traced to both tips, not just where it is visible?
• Was VT paired with at least one surface or volumetric NDT method? Combining visual inspection with MT, PT, UT, or RT gives the most reliable picture of a failed joint.
• Is there a WPS, and does the welder hold a current qualification for that process and position?
• Were low-hydrogen consumables stored correctly, and was preheat verified with a temp stick or probe?
• Was the gouged area cleaned and is your welding PPE rated for grinding and gouging hazards, not just arc time?
• Did a qualified person sign off after the repair?

One more habit worth building: treat inspection as prevention rather than policing. A quick check while the joint is still accessible catches undercut, poor fusion, or a crater crack that would otherwise hide under paint for six months. If you want a working example of a defect that looks harmless and is not, read our undercut guide on causes, acceptance limits, and fixes.

If you are heading back into the shop after finding a broken weld, start with the load, end with the signature, and never skip the middle. Then keep learning: browse our related guides on MIG welding safety precautions and weld inspection right here on WeldGallery, so the next joint you finish is one nobody has to chase down later.

Frequently Asked Questions

What causes a weld to break?

Most broken welds trace back to cracking driven by thermal stress, hydrogen embrittlement, rapid cooling, high restraint, or a brittle heat-affected zone. Design flaws, contamination, wrong filler metal, skipped preheat, and unqualified welding practice are the usual upstream causes.

How do I tell if a weld is broken or just ugly?

Clean the joint and look for cracks, undercut, or separation; any crack at any size is a failure until proven otherwise. Cracks, lack of fusion, and heavy undercut need repair, while surface discoloration and minor cosmetic issues usually do not.

Can a broken weld be repaired, or does the part need replacing?

Sound base metal usually means the weld can be gouged out, rewelded to a qualified procedure, and re-inspected. Fatigue-critical, heavily deformed, or thin-gauge parts, and anything regulated like pressure vessels, often require replacement or an authorized repair.

Is it legal to weld over a cracked weld?

No — depositing weld metal over an unremoved crack is not a code-acceptable repair under standards such as AWS D1.1, which requires the crack to be completely removed first. Regulated equipment also requires a qualified procedure and, in many jurisdictions, an authorized inspector.

What NDT method is best for finding cracks in welds?

Magnetic particle and liquid penetrant testing are best for surface-breaking cracks, while ultrasonic and radiographic testing find subsurface flaws such as lack of fusion. Most inspections start with visual testing and then choose the NDT method based on the joint geometry and code.

Why did my weld crack after it already looked fine?

Cold or hydrogen cracking can appear hours to days after welding as the joint cools and hydrogen diffuses under restraint. That delay is why a same-day visual check alone is not a reliable sign-off for restrained or high-strength steels.

How much does a professional weld repair cost?

Cost depends on access, inspection requirements, downtime, and documentation rather than the weld bead itself. Field repairs on heavy equipment or pressure equipment cost far more than shop work because NDT verification and code paperwork dominate the invoice.

Are broken welds dangerous?

Yes — a failed structural weld can drop a load or a person with no warning, since cracks grow under load until the remaining metal can no longer carry it. Handrails, trailer frames, suspension parts, and lifting gear should be taken out of service immediately after a break is found.

How do I prevent welds from breaking in the first place?

Qualify your procedure, control heat input and preheat, use low-hydrogen consumables stored properly, keep joint surfaces clean, and weld with an appropriate sequence to manage restraint. Then inspect early with VT plus NDT before the weld is coated, buried, or loaded.