Welding Defects Undercut: Causes, Acceptance Limits, and How to Fix It

The grind wheel was singing along the toe of a fillet weld when Dave, the fabricator I was shadowing that winter, held up his hand and killed the torch. “Look,” he said, tapping the metal with a gloved finger. A thin, dark line ran the whole length of the weld where the bead met the plate — not spatter, not slag, just a narrow groove bitten into the base metal. “That’s undercut,” he said. “The inspector will fail this in about four seconds, and by then I’ll have spent forty minutes grinding it out and re-welding it. When I should have just slowed down.”

I’ve thought about that moment hundreds of times since, because undercut is exactly that kind of defect. It doesn’t look dramatic. It doesn’t crack or leak or crumble in your hands. It’s a shadow line at the edge of a weld — and yet it is one of the most common reasons a weld gets rejected, one of the most misunderstood causes of fatigue failure in welded structures, and one of the easiest problems to prevent once you understand what the arc is actually doing to the corner of your joint.

If you have landed here searching for welding defects undercut, you are probably in one of three situations. You failed an inspection and need to understand what the inspector saw. You are studying for a welder qualification or a CWI exam and need the acceptance criteria to stick. Or you are simply tired of grinding out your own welds and want to stop creating the problem in the first place. This guide is written for all three readers, and it goes deeper than a definition: we will cover what undercut is at a metallurgical level, the exact code limits used in the industry, why it forms in each common welding process, how to measure it, how to repair it legally, and how to prevent it with settings and technique you can apply on your next pass.

What Exactly Is Undercut in Welding?

Undercut is a groove melted into the base metal adjacent to the weld toe — or, at the root of a joint, along the weld root — that is not filled in by weld metal. In plain terms, the arc ate away part of the plate and the filler never replaced it. The result is a notch: a sharp, local reduction in thickness sitting exactly where the joint carries the highest stress.

The formal definitions reinforce this. AWS D1.1, Structural Welding Code — Steel, treats undercut as a profile discontinuity and judges it visually against depth and length limits in its visual inspection acceptance criteria. AWS B1.11, Guide for Visual Inspection of Welds, devotes an entire section and illustration set to undercut because it is one of the discontinuities examiners are trained to spot first. The classic trade description — the one taught in nearly every welding textbook — is that undercut is “a groove melted in the base metal adjacent to the weld toe or weld root and left unfilled by weld metal.”

Three details in that sentence matter:

• Base metal, not weld metal. Undercut is damage to the plate or pipe itself. You cannot fix it by depositing more filler over a cold, untouched groove; the groove has to be addressed deliberately.
• Toe or root. Most people picture the visible line along the edge of the bead on the surface. Undercut can also occur at the root side of a joint, where you often cannot see it at all without gouge-and-inspect or NDE.
• Left unfilled. The defect is defined by the absence of metal. That is why travel speed and manipulation are such common culprits — the pool moved faster than the filler could follow.

Undercut is sometimes confused with a handful of neighboring discontinuities, so it is worth drawing the line sharply. Overlap is the opposite failure: weld metal spills over the base metal without fusing, creating a lip instead of a groove. Cold lap (also called lack of fusion at the toe) is filler that sat down on unfused base metal — again, excess metal where there should have fusion, not a void. Underfill is a shortfall of weld metal across the whole face of the joint, not a notch at its edge. Lack of penetration is a root problem, not a toe problem. Undercut is uniquely a notch at the boundary, and notches are exactly what fatigue cracks love.

The Two Main Types of Undercut

Welders and inspectors generally sort undercut into two categories, and the distinction changes both how it forms and how dangerous it is.

Longitudinal Undercut

Longitudinal undercut runs parallel to the weld axis along the toe of the bead. This is the version almost everyone recognizes: a continuous or intermittent dark line where the weld meets the plate. It forms when the arc is directed at the base metal edge rather than into the joint, or when the pool wets the plate and then moves on before filler metal arrives to replace what was melted.

Because it tracks the weld toe — the exact location of the highest stress concentration in a fillet or groove weld — longitudinal undercut is the type most closely associated with fatigue. Test data going back decades show that sharp toe notches dominate the fatigue life of welded joints: the notch raises the local stress well above the nominal design stress, and a crack initiates there long before the average stress in the member would suggest.

Transverse Undercut

Transverse undercut runs across the weld, perpendicular to its axis, usually at the crater end of a bead or across a stop-start point. It forms most often when the arc is broken off incorrectly — pulled away while still at full current, leaving a crater that solidifies short of the surrounding plate height — or when an interpass boundary is cleaned aggressively.

Transverse discontinuities are treated more cautiously by inspectors than longitudinal ones because a crack-like orientation across the loading direction is worse than one running along it. A crater that solidifies with a notch in it is also a frequent origin of crater cracks, which is why the standard advice is to hold the arc at the crater, back up slightly, and let it fill before breaking the arc.

Root Undercut

There is a third condition worth naming because it shows up on pipe and butt joints: undercut at the root bead, on the inside of the joint. It forms when the root pass is too hot, too fast, or aimed at the lands instead of between them. Root undercut is invisible from the outside and is one of the reasons root passes are often radiographed or inspected with borescopes on critical work. If your procedure calls for back gouge and inspection, an overlooked root notch will be found — and it will be far more expensive to correct than it would have been to prevent.

Why Undercut Is a Serious Defect, Not Just a Cosmetic Flaw

The most common dismissive comment about undercut is “it’s just cosmetic.” That comment is wrong often enough to be dangerous.

Start with geometry. A weld toe is already a stress raiser because the stiff weld metal forces strain into the softer adjacent plate. Adding a sharp notch multiplies that effect: the smaller the notch radius, the higher the stress concentration. A 1/32 in (0.8 mm) deep undercut with a sharp root can push local stress several times the nominal applied stress. In fatigue-loaded structures — cranes, bridges, off-road equipment, press frames, trailer frames, pressure vessel nozzles — that is precisely where cracks start.

Second, undercut reduces the load-carrying cross section. In a fillet weld sized by leg length, the effective throat is already the smallest dimension. Biting a groove into the toe reduces it further, sometimes below the minimum effective throat specified by the drawing or the code. An inspector with a fillet gauge can fail a weld for this reason alone, even if the weld looks strong.

Third, undercut acts as a stress-corrosion and corrosion-initiation site. In corrosive environments, the groove collects moisture and deposits, shields the surface from passivation, and concentrates electrochemical attack. A notch that would be tolerable in a dry shop becomes a pit-and-crack incubator on an outdoor structure.

And fourth, undercut is a gateway defect: it is the visible symptom of a parameter problem that is usually also causing invisible damage. The same high current and fast travel that produce undercut are also producing excessive penetration, possible lack of fusion at the root, and a harder, more brittle heat-affected zone. Fixing the line you can see often means fixing the condition you cannot.

How Undercut Forms: The Metallurgy in Plain Language

To prevent undercut, it helps to picture what the arc is doing at 6,000°F and above, thousands of times per second.

The arc column is a concentrated plasma stream. Its heat is not spread evenly across the joint; it is focused where the arc root meets the work, and it shifts as the electrode or gun moves. When the arc sits against an edge of base metal — the toe of a fillet, the land of a groove, the corner of a lap joint — it melts that edge. Surface tension and the shielding gas flow then shape the molten pool. If filler metal arrives in time and wetting is good, the melted notch refills and solidifies flush. If the pool moves on before filler arrives, or if the filler lands in the middle of the pool instead of at the notch, the groove freezes as it was: an unfilled notch.

Four physical variables decide the outcome:

• Heat input at the toe. Higher current, longer arc (higher voltage in a CV process), and a slower arc travel over a given spot all deliver more energy into the plate edge.
• Filler delivery rate. Wire feed speed, electrode deposition rate, and travel speed must balance. If the plate melts faster than metal is added, you get a deficit.
• Pool wetting and surface tension. Contaminants — mill scale, oil, rust, paint, moisture — interfere with wetting. A pool that will not wet the plate will not fill a notch in it.
• Gravity and position. In vertical and overhead positions, gravity pulls the pool away from the toe. The pool sags, the notch at the upper toe stays open, and the filler drops into the middle of the bead instead.

This is why undercut is fundamentally a balance defect rather than a setting defect. There is no single “anti-undercut amperage.” There is only the relationship between heat at the edge, metal delivered to the edge, and the speed with which the pool moves away.

How to Spot Undercut: Inspection Basics

Visual inspection is the first line of defense, and the ANSI/AWS framework begins visual examination before any other method. The equipment list is short: adequate lighting, a magnifier, a fillet-weld gauge or an undercut gauge, and a trained eye.

What you are looking for:

• A continuous or broken dark line at the weld toe that reads as a shadow even after wire brushing — undercut holds darkness because it is a physical groove, while a surface stain wipes away.
• A notch you can catch with a fingernail or the point of a scribe. If the tip of a scribe drags into the line, you have depth worth measuring.
• Crater notches at bead ends — small crescent dips where the arc was broken.
• Any place where the weld profile steps down below the plate surface instead of tapering smoothly into it.

Measurement is done with a dedicated undercut gauge (a small template with graduated steps, referenced in AWS B1.11’s illustration set) or, on many shop floors, with a depth-rule or a modified fillet gauge. The procedure is simple: lay the gauge body across the weld and plate so the reference edge bears on undisturbed base metal, then read the step that matches the bottom of the groove. Always measure at the deepest point, and measure more than once — undercut depth varies along a bead, and code limits apply to the maximum.

Two inspection habits separate professionals from guesswork. First, wipe the weld. Grinding dust and soot make every line look deeper than it is. Second, light the weld from a low angle. Raking light exaggerates shadows and makes shallow notches obvious; direct frontal light hides them. It is the same trick photographers use to reveal surface texture, and it works on weld toes.

For root undercut you cannot see, the shop moves to other methods: gouge-and-grind a sample, use a borescope or fiberscope on pipe, or apply dye-penetrant testing to reveal the notch profile after sectioning. Radiography shows root undercut as a dark line at the joint root on film, and experienced film readers distinguish it from lack of penetration by the way the density fades at the edges.

Acceptance Criteria: How Much Undercut Is Too Much?

This is the question that sends most people to the code, and the answer depends on the standard governing your work. The most widely cited limits come from AWS D1.1’s visual inspection acceptance criteria for butt and fillet welds. The current D1.1 criteria for undercut, expressed for material thickness, are these:

Read that table twice, because it contains a trap. Under D1.1, for thin material the general depth limit is 1/32 in — but there is a provision allowing deeper undercut (up to 1/16 in) over short accumulated lengths, and conversely, short shallow undercut still has a total-length budget. Inspectors do not just measure depth; they add up how much of the bead is affected. A weld with 1/64 in undercut running its entire length can be rejected when a few short stretches of slightly deeper notch would pass.

Other standards draw the line differently. ASME pressure-work codes use their own visual acceptance tables with metric-friendly limits; API pipeline specs impose their own; many architectural and tubular-steel specs incorporate AWS D1.1 by reference; and individual employers or customers may impose more restrictive criteria than any code minimum. A customer specification that says “no undercut” is legal, common, and enforceable — code limits are maximums, not entitlements.

A Quick Word on Fatigue and Why Designers Care So Much

If you ever wonder why inspectors seem disproportionately fixated on a hairline groove, look at fatigue data. Decades of welded-joint fatigue testing — the basis for the detail categories in AWS D1.1’s fatigue provisions and in structural design codes worldwide — show that weld-toe discontinuities dominate failure. A plate with a sharp notch at the toe of a fillet weld can fail in fatigue at a fraction of the stress a smooth, well-profiled weld of the same size would carry.

This is also why “just grind it smooth” is a legitimate engineering answer for undercut. Grinding the toe to a smooth, generous radius removes the sharp notch root, drops the stress concentration, and in many cases improves fatigue category. Fabrication shops that build fatigue-critical structures routinely specify toe grinding as a mandatory post-weld treatment — not to hide a defect, but because the geometry of a ground toe genuinely performs better. The trick is that the grinding must remove the notch completely and leave a smooth transition; a half-ground undercut with a sharp bottom remaining is worse than useless, because it looks acceptable and still cracks.

What Causes Undercut? The Root Causes Behind Every Instance

Ask ten welders what causes undercut and you will get ten answers, because the defect has many entry points. But nearly every instance reduces to one of six root causes, and learning to diagnose by cause — rather than by guessing at settings — is what lets you fix it on the first try.

A diagnostic habit worth building: look at where the undercut sits before touching a single dial. Undercut on both toes points to heat or speed. Undercut on one toe only points to angle or manipulation. Undercut only at stops and starts points to crater technique. Undercut scattered with porosity points to cleanliness. Once you can name the category, the correction is usually obvious.

Parameter Causes: Amperage, Voltage, and Travel Speed

Amperage Too High

Excess current is the classic textbook cause of undercut, and it earns that reputation. More amps means a hotter, wider arc that melts a bigger crater out of the plate. In stick welding, this shows up quickly: the bead goes flat and wide, the ripples spread, spatter increases, and a distinct groove opens along both toes. On 1/8 in E6010 or E7018 running near the top of the recommended range, undercut can appear within a few inches if travel speed does not increase to match.

The correction is not always to drop amps, though. Sometimes amps are correct for penetration and the real problem is that the welder is moving too slowly for that heat. Reduce amps in modest steps (25–50 A for stick, a few volts and matched wire speed for MIG) while watching penetration; you do not want to trade undercut for lack of fusion. A better first move is often to keep the amps and speed up travel, then fine-tune.

Travel Speed Too Fast

Speed is the cause people forget because it feels like the safe choice — “less heat in the part.” But a pool that outruns its filler leaves a trench. Fast travel produces a bead that is narrow, tall, and stringy, with ripples stretched far apart and a continuous undercut line at the toes. It is extremely common in production work where a welder is chasing a schedule, and on thin sheet where the welder is trying to avoid burn-through.

Watch your ripples as you go. Even, closely spaced ripples mean the pool and filler are in balance. If the ripples start to elongate and the bead crown rises above the plate while the toes cut in, you are outrunning the filler. Slow down slightly, or increase deposition (wire speed, electrode size) to keep up.

Arc Length and Voltage Too Long

In stick welding, arc length is controlled by the welder’s hand; in MIG (GMAW-CV) and flux core, it is controlled by voltage. Both matter. A long arc spreads heat over a wider area, loses shielding gas coverage, increases spatter, and digs a wash-out groove at the toe. Excessive voltage in short-circuit MIG produces the characteristic “bird-track” spatter and a flat, wide bead with undercut; insufficient voltage gives a stubby, tall bead — usually without undercut, because the pool is starved rather than sprawling.

Keep arc length roughly the diameter of the electrode core for stick work, and dial voltage to produce a smooth, quiet arc with a bead about 1.5 to 2 times the wire diameter in width. If you hear harsh crackling and see wide spatter arcs, back the voltage down.

Technique Causes: Angles, Arc Direction, and Manipulation

Work Angle

The work angle — the position of the electrode or gun relative to the joint faces — decides where the arc points. Point it at a plate edge and you will melt that edge. On a T-joint, a work angle leaning too far toward one side puts the arc root on that plate’s toe and cuts a groove there while the opposite side sits cold. This is why single-sided undercut is almost always an angle problem, not a settings problem.

The fix is geometric: keep the arc pointed into the corner of the joint, split the heat between both members, and let the filler bridge the toes. On a fillet weld, that usually means bisecting the joint angle with a slight bias toward the heavier plate, then making small weaving motions that dwell at the toes just long enough to fill them.

Travel Angle and Push vs. Pull

Travel angle — the drag or push angle from vertical — also shifts the pool. A steep drag angle (too much lean back) piles the pool behind the arc and can leave the leading toe unfilled; excessive push can drive the pool ahead of the filler and wash out the trailing edge. Most processes favor a 5–15 degree angle: push (forehand) for spray and short-circuit MIG on flat sheet for good visibility and gas coverage, drag (backhand) for flux core and most stick welding because the slag follows the pool.

If you are seeing undercut on the toe behind your direction of travel, try reducing your travel angle toward vertical and slowing slightly. If it is on the toe ahead of travel, check that your work angle is splitting the joint and that you are not steering the arc onto the leading edge.

Weave Width and Dwell

Weaving is where undercut is won or lost on wide passes. A weave that zips across the joint without pausing deposits its metal in the middle and leaves the toes starved — the classic “middle-heavy” bead with notches on both sides. The correct rhythm is to pause briefly at each toe, letting the pool wet the plate edge and the filler fill it, then sweep across the middle. The pause should be a fraction of a second — long enough to fill, short enough to avoid melting a bigger crater.

On vertical-up welds, the weave is even more critical because gravity is working against you. Keep the weave narrow (generally no wider than about 2.5 to 3 times the electrode diameter for stick), carry a slight shelf of molten metal above the arc, and ensure each side-to-side motion deposits filler at the toe before moving on. Excessive weave width on vertical-up is one of the fastest ways to produce deep undercut on the upper toe.

Craters and Restarts

Many undercut failures are not along the bead at all — they are at the end of it. Breaking the arc straight out of a crater leaves a depression that solidifies below the surrounding surface, often with a notch at its rim. Build the habit of holding the arc at the crater for a beat, taking a tiny back-step, and letting the filler mound slightly before you break away. On restarts, overlap the previous bead by about a half electrode length and blend the profile so there is no step.

Undercut Causes by Welding Process

Shielded Metal Arc Welding (Stick)

In SMAW, the welder controls everything: arc length, angle, speed, and rod manipulation. Undercut comes from amperage set too high for the position, arc length stretched while running uphill, travel speed racing to keep the pool from sagging, and electrode angle that aims the arc at the plate. Cellulosic rods like 6010 run deep and aggressive and will undercut readily on high amps; low-hydrogen 7018 is more forgiving but will still notch a toe if the weave dwells too long in the middle.

Practical fix for stick: verify your amperage against a chart for rod size and thickness, keep a tight arc, and use a slight side-to-side motion with a real pause at each toe. On 3/16 in E7018 work, correct amperage, polarity, and technique all matter together — settings alone never tell the whole story, which is why technique guides for that electrode emphasize travel speed and manipulation as much as the dial. If you are running 7018 and still cutting toes, slow down and let the flux do its job.

Related reading: 3/16 E7018 Amperage, Settings, Polarity and Technique covers the full parameter picture for that electrode.

Gas Metal Arc Welding (MIG)

GMAW has three modes and each produces undercut differently. In short circuit, excessive voltage is the prime suspect: the arc lengthens, heat spreads, and the bead washes out at the toes. In globular, the erratic transfer and high current often cut grooves on both sides. In spray transfer, the high heat and fluid pool can undercut if travel is too fast or the stick-out is wrong — though spray’s good wetting helps fill.

Wire feed speed must be matched to voltage. If you raise volts without raising wire speed, you get a long arc and undercut; raise wire speed without volts and you get stubbing and lack of fusion. Travel speed has to be tuned to the pair. On thin sheet, welders often compensate for burn-through risk by speeding up — and then wonder why every seam has a groove. The better answer on thin material is to lower the parameters and keep the speed steady.

If you are sorting out MIG behavior on real projects, it helps to see side-by-side examples of weld quality differences, and Good and Bad MIG Welds: How to Tell the Difference shows exactly the profiles that separate a filled toe from a notched one.

Flux Cored Arc Welding (FCAW)

Flux core runs hot, penetrates hard, and is very prone to undercut in vertical and overhead positions if the weave is too wide or the amps are set for flat-position work. Self-shielded FCAW (no gas) tends toward higher parameters and heavier deposition; gas-shielded dual-shield runs even hotter. The slag also hides the toes — undercut can be concealed until you chip and brush, which is one reason FCAW welds get rejected after the welder has already moved on.

For FCAW, reduce parameters when you leave flat position, narrow the weave, and keep a molten shelf above the arc on vertical-up. Polarity matters too — most flux core runs DCEP, but running the wrong polarity changes penetration and deposition enough to change your defect profile, so confirm the machine’s polarity before chasing any undercut problem.

Gas Tungsten Arc Welding (TIG)

TIG undercut comes from amperage too high for the material, travel too fast for the puddle, a tungsten angle aimed at the edge, and — very commonly — filler rod technique: dabbing the rod into the middle of the puddle instead of feeding it at the toe. Without filler arriving at the edge, the melted plate notch stays open. On thin sheet, a torch that lingers too long at a corner will melt the edge away faster than you can add rod.

TIG is the easiest process in which to see the mechanics of undercut clearly, because nothing hides the pool. If you watch the toe as you move, you can literally see the groove open when the rod misses the edge. The correction is mechanical: aim the arc at the joint, feed the rod continuously or rhythmically at the leading toe, and slow your travel until the puddle fills flush every time.

Material and Joint Factors That Make Undercut More Likely

Some conditions stack the deck against you before you strike an arc:

• Thin material. High heat that would be harmless on 1/2 in plate will blow an edge away on 16 gauge. Thin stock punishes any excess parameter instantly.
• High thermal conductivity. Aluminum pulls heat out of the pool quickly, tempting welders to raise current — which then melts the edge faster than the puddle can be controlled. Aluminum undercut is almost always a heat-and-speed problem.
• Joint geometry. Lap joints and T-joints naturally expose an edge for the arc to bite. Sharp inside corners concentrate heat. J-grooves with narrow lands leave little margin.
• Position. Vertical and overhead force you to fight gravity, which pulls the pool away from the upper toe. Overhead is the hardest: parameters must stay low and the puddle small.
• Fit-up gaps. A wide root opening means more filler volume needed; if deposition does not increase to match, the joint consumes heat without receiving metal.
• Surface contamination. Mill scale, rust, oil, and paint block wetting. The pool skates over the contamination, the filler cannot bridge to the plate, and irregular undercut follows.
• Preheat and interpass temperature. Very hot base metal keeps the pool fluid longer and lets it sag away from the toes; too cold, and the pool freezes before it can wet and fill.

Prevention by Process: Practical Settings Discipline

Prevention is mostly discipline, not talent. This checklist covers the habits that eliminate undercut in normal production work.

Universal habits for every process: clean the joint, confirm your parameters against a chart or procedure before the first pass, watch your ripples as you weld rather than staring at the arc, inspect the toes at every stop, and keep a mental list of what you changed and why. If you are setting up equipment from scratch in a new space, the complete beginner’s guide to using a wire welder walks through the machine-side decisions that feed directly into bead profile control.

Inspecting Weld Toes After the Bead Is Down: A Field Procedure

The moment undercut becomes a problem you can fix cheaply is the moment it freezes — not after the part is in service. Professional shops run a short post-weld inspection loop that takes minutes and catches nearly every rejectable defect before money is spent shipping or painting.

Here is the loop to run after every pass on critical work:

1. De-slag and brush. Chip, wire-brush, and wipe the entire bead. Undercut is invisible under slag and soot. Do this before measuring anything.
2. Rake-light the toes. Hold a work light at a low angle along the bead and sweep both toes. Every shadow line you can see is a candidate for measurement.
3. Measure depth. At the deepest-looking point of each shadow line, use an undercut gauge. Record the reading. If it is at or near your acceptance limit, mark the spot with a soapstone tick for repair later.
4. Check length. For a rejectable depth, measure the accumulated length affected. Remember that D1.1-style limits cap not just depth but how much total length is allowed at the deeper end.
5. Check craters and starts. Look at every stop and start for crater notches, and measure the deepest ones the same way.
6. If the joint is root-critical, plan NDE. For root undercut — pipe, tube, backing-removed butt joints — supplement visual with dye penetrant, radiography, or borescope per your WPS or specification.

Two tools make this loop accurate rather than approximate: a step-style undercut gauge (AWS B1.11 pictures one in its inspection tools section) and a fillet weld gauge for throat dimension. Both are inexpensive, and both pay for themselves the first time they keep a rejectable weld from leaving the shop. Good lighting is the third tool — a headlamp or low floodlight set at 30 degrees or less to the plate surface.

How to Repair Undercut Correctly (and When You Cannot)

When undercut is rejectable, someone has to fix it. The correct repair depends on the configuration, but there is a firm rule at the top: undercut is repaired by removing it and refilling — or by blending it — never by simply smearing more weld metal over a dirty notch.

Grinding the Undercut Out

For shallow undercut, the accepted repair is to grind it out to leave a smooth, clean groove of sound base metal, then weld it back up with a few small passes. Rail-grind along the toe to remove the notch completely — you want to see clean, bright, defect-free base metal in the bottom of the groove you grind. Then wash in a stringer or two with slightly lower parameters than the original pass, watching that the new weld wets smoothly into the plate with no fresh toe notch. After welding, remove any reinforcement grind marks so the surface is smooth, and re-inspect to the same criteria.

Welding Over the Top

A common mistake is to run a reinforcement cap over an undercut bead and call it done. If there is still a notch-down-to-plate at the toe and the cap widens the previous profile, you now have an irregular profile with a hidden stress raiser — and usually you still have undercut at the cap’s own toes. In most code work, cosmetic coverage is not a repair; the defect must be removed or the repair weld must fully fill the void left by removal. When a repair cap is genuinely called for (for example, on a groove weld whose crowning is slightly low), grind the toes smooth first so the cap covers a clean surface, then blend afterward.

When Repair Is Restricted or Not Allowed

Avoid and repair restrictions exist in several cases:

• Galvanized or coated structures. Repair welding may be prohibited in the coating, or may require stripping the coating first because re-heating zinc makes the situation worse.
• Hardened or quenched-and-tempered steels. Codes and WPSs frequently forbid more than a limited number of repair cycles, because each repair reheats a heat-affected zone that has already been refined.
• High-restraint joints. Repairs on heavily restrained members crack from contraction stress; preheat and low-hydrogen filler are mandatory in these repairs.
• Hardfacing and cladding. Repairing a notch in a wear layer by welding over it may simply transfer the notch to a new toe; manufacturers often recommend grinding and monitored re-deposit.
• Brittle or prequalified-window valleys. If the original weld used a narrow procedure window, the repair must be done under the same approved WPS. “Quick fix” rods or out-of-range parameters are not allowed.

The controlling document always wins. Before repairing, check the code section on rework: many allow repair of weld imperfections by grinding and rewelding, but impose conditions on how deep you can grind, how many times you can re-weld the same area, and what temperature control is required.

Undercut vs. the Defects People Mistake It For

New welders lose marks on tests for calling one defect another. This table is the one I wish someone had handed me before my first qualification test.

Notice what all of these have in common: they are profile and fusion discontinuities, not defects of filler chemistry. The practical takeaway is that the welder who controls heat input, travel speed, and filler timing all at once controls nearly the entire defect list.

Mistakes That Create Undercut (and the Welders Who Keep Making Them)

After years of watching weld tests and shop floors, the same habits keep producing notched toes. Read these as a confession list, not a trivia list:

• Turning up the amps to “make sure it penetrates.” Then moving faster to stop the sag — which is exactly how you create a groove at the toe. Penetration and undercut are both heat effects; you cannot fix one with the same tool that causes the other.
• Creeping along the joint toward the end of the rod. As electrodes short, the welder drifts the angle to hold the arc, and the arc swings onto the near plate edge. That is why stick undercut often worsens in the last quarter of the rod.
• Only reading the drawing for size, not for position. Flat settings used in vertical or overhead are a guaranteed way to notch toes.
• Watching the puddle but not the toes. If you watch only the center of the pool, you will see a beautiful molten puddle while both toes are silently cutting in. Let your eyes sweep the full width of the pool edge.
• Grinding “before the inspector gets here.” Blending an undercut without refilling it just hides the cross-section loss. A smooth removed notch still has less effective throat than the print requires.
• Ignoring craters. Every stop is a potential failure point, and craters are the easiest defect to be credited with — the fix costs ten seconds.

Expert Insights, Real Numbers, and Useful Facts

A few facts and figures help anchor what undercut means in practice:

• Code depth limits are small on purpose. D1.1-style acceptance allows only ~1/32 in undercut on material under 1 in thick before length-budgeted exceptions apply. A notch the width of a credit card edge is already at the reject line.
• Weld profile, not chemistry, drives most service failures. A widely repeated observation in welding-inspection training is that profile discontinuities at the toe — undercut and sharp transitions — cause more service failures of welded joints than internal flaws like slag or porosity. The visible groove beats the hidden inclusion in the failure statistics.
• Fatigue categories reward toe quality. Structural design codes grade welded details by fatigue categories; improving the toe condition through grinding or a smoother profile can move a joint up a category and extend its calculated life dramatically. Grinding the toe of an eccentric bracket weld is one of the highest-value quality investments a fabricator can make.
• Undercut on a groove weld’s root is often a preheat or joint-geometry symptom. When the root keeps notching on the inside, welders frequently chase current when the real fix is a land-width adjustment, a root-gap change, or a controlled preheat that firms up the root pass.
• The dirtier the joint, the more undercut you will accept without noticing. Contamination kills wetting, and poor wetting is the second half of the undercut equation. This is why cleaning is listed as a process factor rather than a courtesy.

The single most useful mindset shift is to treat undercut as thermodynamics, not luck. The arc is moving energy into the base metal faster than the filler can replace it, or the filler cannot wet the surface because of angle, contamination, or position. Solve the energy-and-wetting balance and the notch disappears, process by process.

Common Questions About Undercut, Answered

Is undercut always a rejectable defect?

No. It is rejectable when it exceeds the acceptance criteria of the governing code or specification. Small, short stretches of shallow undercut pass under D1.1-style rules, and many other specs have their own bands. The trap is that “passes D1.1” is not the same as “acceptable for this customer” — stricter criteria are common and enforceable.

Can you weld over undercut?

Simply running another bead over it is not a proper repair in most code work. The correct procedure in nearly all cases is to remove the undercut to sound base metal (usually by grinding) and then refill, or to blend it where blending is code-acceptable. Never seal a notch with weld metal that does not fuse into the plate face below it.

Does grinding undercut remove the defect?

It removes the notch geometry, which is the dangerous part, but it also removes base metal thickness unless you refill. In many acceptance rules, a fully ground-out, smooth-blended toe with adequate remaining thickness and throat is acceptable because the stress raiser is gone. Where depth or thickness is at play, refill to restore cross section.

Which welding process undercuts the most?

Easily-driven processes produce the most undercut in practice: high-current MIG spray and flux core on out-of-position work are very common sources in production, and stick with cellulosic rods undercuts readily on high amps. TIG rarely undercuts, but only because the welder controls the rod timing by hand — it undercuts instantly when the rod misses the toe.

Does too fast cause undercut, or too slow?

Both, at opposite ends. Travel too fast leaves a groove because the filler never reaches the edge. Travel too slow, combined with high current, overheats and melts a wide crater that the pool cannot fill. Undercut is a window between too fast and too slow — which is another way of saying it is a balance defect.

What is the quickest check for undercut I can do without a gauge?

A fingernail, a scribe point, or the corner of a flat feeler gauge dragged along the toe will catch every depth worth worrying about. If the tip drops into a distinct groove, get the gauge and measure it properly. It is not a substitute for measurement — but it is a fast screen that never needs batteries.

Building a Shop That Rarely Sees Undercut

Undercut defects are environment-shaped. A shop that gets undercut under control usually did not buy better machines — it changed behaviors. The differences, in order of impact:

1. Procedures with real parameters. WPSs that list amperage, voltage, travel speed guidance, electrode size, and positions for each joint kill 80% of parameter-driven undercut before the arc starts.
2. Standardized torch and travel angles. Writing the angle guidance on the procedure card and enforcing it removes the angle-driven notch.
3. Self-inspection before the inspector. A two-minute toe inspection after every critical beat, by the welder, before signing off.
4. Cleaning culture. Wire brush, flapper disc, or chemical prep is treated as part of the welding procedure, not a suggestion.
5. Repair discipline. A clear, documented, code-compliant repair path so nobody is tempted to “just run another pass over it.”

None of these require expensive equipment. A shop that plans its facilities properly — good ventilation, good lighting, clean staging — makes them easier to execute. OSHA’s construction welding and cutting regulations (29 CFR 1926 Subpart J) set the baseline expectations for ventilation and protection when welding, and NIOSH publishes exposure guidance on welding fumes that every shop safety review should reference. If you are setting up or upgrading a workspace, the way your floor is laid out affects quality as much as the dials do: planning and setting up a safe welding shop covers the physical layout decisions that keep welders consistent.

Final Thoughts: The Ten-Second Fix and the Lifelong Habit

Let me bring you back to Dave’s welding table, because his point has stuck with me. He was not a talker, and he was not a theorist. He said the entire problem of undercut in one sentence: “The arc eats and the puddle feeds, and you’re just the guy who has to keep the two in balance.”

That is it. When the arc out-eats the puddle — through heat, speed, angle, or contamination — the plate bites back with a groove. When the puddle feeds the toes — with the right parameters, a real dwell at each edge, and a clean joint — the weld makes full contact with the plate.

So the ten-second fix is this: the next time you see a shadow line at a weld toe, stop, ask whether heat was too high, speed too fast, the angle aimed at the plate, the pool too dirty to wet, or the crater left unfilled — and write the answer down. Ten seconds of diagnosis once means fewer grinding sessions for years. The lifelong habit is even simpler: make your inspection loop automatic. De-slag, rake-light, gauge, mark, repair, re-weld. The welders who never have to apologize for undercut are not the ones with better machines. They are the ones who look at their toes.

Check your welds at every stopping point, and if you are new to MIG and want a controlled way to practice those profiles, the beginner’s wire welder guide linked earlier is a sensible place to start. If you already run flux core and you see a groove, check your polarity setting before you touch the dials — it changes everything downstream of the gun.

Authoritative Sources (Outbound References)

• AWS D1.1/D1.1M:2025 Structural Welding Code — Steel, visual inspection acceptance criteria for undercut: https://pubs.aws.org/Download_PDFS/D1.1-D1.1M-2025-AMD1_PV.pdf
• AWS B1.11M/B1.11:2015, Guide for Visual Inspection of Welds (undercut illustrations and measurement): https://pubs.aws.org/Download_PDFS/B1.11M-B1.11-2015-PV.pdf
• OSHA 29 CFR 1926.353, Ventilation and Protection in Welding, Cutting, and Heating: https://www.osha.gov/laws-regs/regulations/standardnumber/1926/1926.353
• OSHA Occupational Chemical Database — Soldering/Welding Fumes: https://www.osha.gov/chemicaldata/1033
• CDC/NIOSH Pocket Guide to Chemical Hazards — Welding Fumes: https://www.cdc.gov/niosh/npg/npgd0666.html
• Miller Electric, Welding Safety & Health Guide (Volume II — fume extraction and respiratory protection): https://www.millerwelds.com/-/media/inriver/miller-safety-guide-vol-ii-interactive.pdf
• Lincoln Electric, Safety Data Sheet guidance referencing ANSI Z49.1 and NFPA 51B: https://www.lincolnelectric.com/assets/US/EN/MSDS_lib/ZLE_SDS_NA-EN-200000000146.pdf