Undercut in Welding: Definition, Causes, Limits, and Fixes

A fabricator once told me the story of a set of crane-lug fillet welds that looked picture-perfect from three feet away, yet the inspector rejected every single one with the same two words: ‘undercut, reject.’ What the welder saw as a clean, washed-in bead edge, the inspector saw as a razor-thin notch melted into the base metal — a hidden fatigue crack waiting for the first ten thousand load cycles. That gap between ‘looks good to me’ and ‘code-compliant’ is exactly where this article lives.

What Is Undercutting in Welding? The Definition That Every Welder Needs

That inspector’s two-word verdict — “undercut, reject” — gets spoken in fabrication shops every single day. The welder sees a clean, washed-in bead edge. The inspector sees a razor-thin notch melted into the base metal. Both people are describing the same few tenths of a millimeter of missing metal, and the only things separating them are a precise definition, a gauge, and an acceptance table. Before we get to causes, limits, and repairs, let’s settle the question the whole article rests on: what exactly is undercut?

The AWS Definition of Undercut

The authoritative wording comes from AWS A3.0, Standard Welding Terms and Definitions, which defines undercut as “a groove melted into the base metal adjacent to the weld toe or weld root and left unfilled by weld metal.”

Read that sentence slowly, because three phrases inside it carry most of the weight:

• “A groove melted into the base metal” — undercut is not weld metal piled on top of the plate. It is the plate itself, melted away by the arc and never replaced.
• “Adjacent to the weld toe or weld root” — undercut always lives at the edge of the weld, at the boundary between the bead and the parent metal. It never forms in the middle of a sound weld face.
• “Left unfilled by weld metal” — the groove remains open after the pass is finished. If filler flows in and closes that valley, you no longer have undercut; you have something else, and something else has its own repair.

Physically, undercut looks like a sharp, dark crease running along the line where the weld face meets the base metal. A healthy weld profile blends gradually into the plate, almost soft at the transition. Undercut is the opposite: a visible valley beneath a hard toe line, often faintly discolored from arc heat, that catches a shadow the moment you rake a light across it. SSAB’s welding defect guidance describes it in exactly those terms — a groove in the base metal adjacent to the weld toe or root that the weld metal failed to fill.

Location matters as much as appearance. Inspection guidance from the Iowa DOT notes that undercut most frequently shows up on the vertical leg of a horizontal fillet weld or along the top edge of a horizontal groove weld — precisely the two places where gravity pulls the molten puddle away from the edge faster than filler can chase it back.

And this is not an exotic problem. TWI calls undercut probably the most common geometric shape imperfection in arc and gas welding. If you weld long enough, you will make it, find it, and have to decide what to do about it.

Discontinuity or Defect? The Distinction That Decides Your Rejection

Here is where a lot of welders get burned, because AWS draws a line that changes everything about how a shop handles a piece of undercut:

• A discontinuity is any interruption of the typical structure of a material — a porosity cavity, a slag inclusion, a zone of lack of fusion, or a groove at the weld toe.
• A defect is a discontinuity that, by its nature or accumulated effect, renders the part unable to meet minimum acceptance standards.

The consequence is simple and powerful: all defects are discontinuities, but not all discontinuities are defects. A shallow undercut on a static, non-cyclic connection, sitting inside the limits of AWS D1.1, is a documented discontinuity — acceptable as written, not a reject. That same notch on a cyclically loaded tension member, running longer than the code allows, becomes a defect, and the piece comes out. Knowing this distinction is the difference between a productive conversation about which acceptance table applies and a fruitless argument about whether the weld “looks fine.”

Why This Tiny Groove Earns So Much Attention

• Structurally, it behaves like a sharp notch — a stress concentrator parked exactly at the weld toe, which is where fatigue cracks in welded structures most like to begin. A groove that costs almost nothing in static strength can still cut fatigue life disproportionately.
• Commercially, it is one of the most common causes of weld rejection and rework. Grinding, filling, and re-inspecting cost time and money that a balanced parameter set would have saved before the arc was ever struck.
• By process, undercut is genuinely universal. It appears in shielded metal arc welding (SMAW), gas metal arc welding (GMAW), flux-cored arc welding (FCAW), submerged arc welding (SAW), and gas tungsten arc welding (GTAW) alike. The physics never changes: the arc melts base metal faster or wider than filler metal can refill. That is why prevention habits carry across processes and materials — the same toe-notching tendency shows up whether you are running 7018 on structural steel, chasing edges in aluminum MIG welding, or tuning arc characteristics in aluminum flux core welding on thin sheet.

Hold onto that measurement habit, because everything that follows builds on it: how undercut differs from underfill, overlap, and lack of fusion, what actually causes it in each process, what the acceptance numbers really are, and how to repair it the right way the first time.

Undercut vs Other Weld Discontinuities: Telling the Difference at a Glance

Defining undercut is one thing; spotting it correctly is another. Many weld discontinuities share a similar footprint at the surface, but the correction for each could not be more different. Grinding out what is actually a crack will only open it further, while adding filler over an undercut never fills the notch that was melted into the base metal. Identification always comes first.

The Most Common Lookalikes

Undercut is an unfused groove melted into the base metal. These are the imperfections most often confused with it:

• Underfill: The opposite of excess reinforcement. Instead of gouging base metal away, underfill is simply not enough weld metal to bring the face up to profile. It forms from low wire feed speed, poor travel control, or an undersized bead. The fix is adding qualified weld metal, never blending base metal away.
• Overlap (Cold Lap): Molten weld metal rolls over the parent metal at the toe or root without fusing to it. According to TWI, the causes are excessive weld metal, a shallow puddle, or an incorrect work angle. No groove is melted out; there is extra, unfused metal sitting on top. It usually requires removal, not capping.
• Lack of Fusion: A bonding failure between weld passes or between weld metal and base metal. It is often subtle, appearing as a tight, linear shadow with no visible crater. Detection needs more than a glance, and repair means gouging to sound metal before re-welding.
• Porosity: Small, round gas pockets trapped on or below the surface, caused by contamination, damp consumables, or shielding gas trouble. These are cavities, not elongated notches along the weld toe.
• Excessive Concavity or Convexity: Profile issues rather than notches. Excess convexity is a bulging face; excess concavity is a sunken one with little reinforcement. Both can affect throat measurements, but neither creates the abrupt, melted groove that defines undercut.
• Cracks: The most severe of the bunch. Cracks are planar, often branching, razor-sharp fractures that can propagate under load. As stated in AWS D1.1 Table 8.1, cracks carry zero tolerance in both static and cyclic connections, while undercut has measurable depth limits. The two are judged very differently for a reason.

At-A-Glance Comparison

Discontinuity	Location	Visual Appearance	Primary Cause	Typical Repair Logic
Undercut	Weld toe or root	Narrow, melted groove in base metal with an abrupt edge	Excess heat, fast travel, poor toe coverage	Blend to a smooth transition, or qualified fill if over allowable depth
Underfill	Weld face	Depressed face below the base metal surface	Insufficient filler deposition	Add qualified metal to restore reinforcement; never grind base metal
Overlap/Cold Lap	Weld toe or root	Rounded weld metal lapping over unfused base metal	Low heat, steep work angle, oversized puddle	Remove lapped metal to sound material; it cannot bond by capping
Lack of Fusion	Between passes or to base metal	Tight linear lack of coalescence, often hidden	Low amperage, poor manipulation, slag entrapment	Gouge to sound metal and re-weld with corrected parameters
Porosity	Face or sub-surface	Round gas pockets, sometimes clustered	Contamination or shielding defects	Judge by code; removal depends on size, distribution, service
Cracks	Any location, often toe or crater	Sharp, jagged, branching fracture	Restraint, hydrogen, stress	Zero tolerance in most structural codes; remove to sound metal and repair to a qualified procedure

All of these can be detrimental, but undercut holds a unique danger. As noted by Portland Community College’s AWS D1.1-based criteria, undercut creates a stress concentration point, a stress riser, that can become the starting point for cracking. That is why mistaking a crack for undercut, or the reverse, is so hazardous. On repair work using equipment like an Air Arc Welding and Carbon Arc Gouging setup, correct identification is mandatory before any gouging begins.

The practical rule: judge the origin before choosing the fix. Is metal missing from the base metal (undercut)? Is metal missing from the face (underfill)? Is unfused metal sitting on top (overlap)? Or is there a fracture (crack)? That answer decides whether the remedy is precise blending, qualified filler, full removal, or rejection. Even routine prep, such as using an aluminum cleaner for welding, can prevent porosity lookalikes, but no amount of cleanliness will fix the geometry of a true notch.

Types of Undercut: Toe Undercut, Root Undercut, and Where Each One Hides

Not every undercut appears in the same spot, and knowing where it forms is just as important as knowing how deep it is. While the root cause may be shared, each form of undercut carries its own inspection challenges, as well as its own effect on acceptance. With that being said, there are four physical forms that every welder should be able to identify on the shop floor.

Toe Undercut: The Most Visible Offender

Toe undercut is by far the most common form seen in fabrication. It takes shape as an unfilled groove melted into the base metal at the weld toe, which is the transition point where the face of the weld meets the parent metal. Instead of a smooth, gradual blend, a sharp notch is left behind where filler metal failed to refill the area.

This form is most often spotted on fillet and groove welds from the outside of the joint. For many structural welds, the toe is already the location that fatigue cracks attempt to initiate from. When paired with poor technique, this tiny notch becomes all the more concerning, especially for members under repeated loading.

Root Undercut: Hidden From Direct Sight

While the toe is out in the open, root undercut is far sneakier. This form develops along the root of the weld, gouging into the base metal at the very bottom of the joint penetration. On open root joints, it can sometimes be seen from the face side, however, its presence is much easier to miss.

Most critically, on single-sided pipe welds, undercut may also be found on the bore surface, out of direct sight, according to TWI. For pipe fabricators, this means a weld can look perfectly acceptable from the cap side, yet hold root undercut internally. For tight, code driven piping, this hidden form is one of the biggest reasons for bore inspections.

Undercut in Multipass Welds and Side Undercut on the Groove Face

Not all undercut is limited to just the first or final pass. In multipass welds, heat from subsequent beads can re-melt the toes of previous passes. As stated by TWI, undercut can occur at the toes of each pass of a multipass weld, where the trapped groove can result in slag becoming stuck in the undercut region. This can create a compounding issue, as trapped slag can go unnoticed before the next pass is laid down.

Alongside this, side undercut can form directly on the groove face itself. Rather than being at the outer toe of the joint, it is melted into the prepared bevel face between passes. This form is incredibly easy to cover over with the following bead if not inspected between passes, making interpass visual checks a necessity.

Longitudinal vs Transverse Undercut: Stress Direction Matters

The location may grab attention, however, the orientation is what often determines if it’s rejectable. Longitudinal undercut runs parallel to the length of the weld, whilst transverse undercut runs across the weld, perpendicular to its axis. This difference is vital, as tensile stress rarely pulls along the same path for every member.

This exact distinction is recognized by the 2025 AWS D1.5 Bridge Welding Code, which distinguishes undercut transverse to primary tensile stress from undercut parallel to tension or in compression, applying far tighter limits to the transverse case. In simpler terms, a small, longitudinal groove may be allowed in certain areas, whilst the same depth running transverse to primary tension can be drastically less tolerable.

What Causes Undercutting? A Process-by-Process Breakdown of the Root Causes

Undercut is, at its core, a heat-balance failure. The arc is a shared heat source: it has to melt a groove in the base metal and simultaneously deposit enough filler metal to refill that groove before it solidifies. When the arc melts the base metal faster or wider than the filler can chase it, the result is an unfilled channel cut into the plate at the edge of the weld. Every cause below comes back to one of three problems: too much heat, too little filler, or too little time.

The Five Dominant Causes

Most shop experience lines up with Kobelco’s published list of undercut causes: excessive welding amperage, too-fast electrode manipulation, too-long arc length or too-high arc voltage, inappropriate travel and work angles, and — on submerged arc welding — wire tracking too close to the groove face. TWI summarizes the same physics as a wide-spreading arc from high voltage combined with insufficient fill from low current or high travel speed. In other words, the arc is carving a wide path while the filler is arriving too slowly to close it.

• Travel speed too fast: the puddle gets no dwell time to wet out and fill the trailing edge.
• Amperage too high: the arc digs a deeper channel than the deposited volume can replace.
• Voltage or arc length too long: the arc cone widens, spreading heat sideways into the toe instead of straight into the joint.
• Wrong work or travel angle: the arc force is pointed at the toe rather than the joint, blowing the puddle away from the edge it should be filling.
• Weave without a pause: a weave that swings through the toes deposits metal in the center and leaves the edges starved.

SMAW (Stick Welding)

Stick electrodes undercut most often when the welder runs hot and fast with a long arc. The fix is concrete: keep the arc length roughly equal to the core-wire diameter, back the amperage off one dial step if the puddle is looking thin and watery, and hold a drag angle of about 10–15 degrees so the arc force points into the joint, not at the toe. If you are weaving, pause for a heartbeat at each toe so filler metal actually arrives there. Damp electrodes are a quiet contributor here too — they destabilize the arc and encourage erratic bead edges, so store low-hydrogen rods in a holding oven. If you are still refining stick fundamentals, our Advantages of Stick Welding guide covers arc control in depth.

GMAW and FCAW

Undercut in wire processes is almost always a wire-feed-balance problem. If voltage and travel speed are high while wire feed speed is low, the arc is wide and the filler supply is thin, which is exactly the TWI recipe. Raise the wire feed speed or drop the voltage, and slow the travel at the toes. Keep contact-tip-to-work distance consistent, since a wandering stick-out changes arc voltage even when the display does not. Push angle makes it worse on plate; a slight drag keeps the arc pointed at the root. Heat balance gets trickier on aluminum, where high thermal conductivity pulls heat away from the puddle fast — see our Aluminum MIG Welding guide for how travel speed and preheat interact there.

GTAW

TIG undercuts because arc voltage rises with arc length, so a “long” TIG arc spreads heat into the toe while the welder’s hand is still moving. Keep the tungsten close, reduce amperage (or use pulse if available), and add filler at the toe before the torch advances, not after. Slowing the travel rhythm and neutralizing the torch angle both help; SSAB’s recommended corrections — decrease travel speed, use small stops at the end of oscillation, use a neutral or small forehand angle, decrease voltage, and reduce heat input — apply directly to TIG work.

SAW and the Less Common Contributors

On submerged arc welds, high voltage lengthens and widens the arc under the flux, and a wire tracking too close to the groove face cuts the side wall instead of filling it. Reposition the wire toward the joint center, drop the voltage, and confirm the travel carriage is tracking true.

Two secondary causes deserve respect: arc blow and part magnetism, which push the arc to one side and scallop the opposite toe (counter it with ground placement, run length, or AC where the procedure allows), and filler metal that melts too slowly for the heat input being used, which is common when a hardfacing or alloy filler is run at plate-welding settings — our Alloy Steel Welding Electrodes selection notes explain that matching melt rate to amperage matters.

Why Undercut Matters: How a Small Notch Cuts Fatigue Life

Undercut rarely fails a structure the day it is welded, and that is exactly what makes it dangerous. A beam with a shallow groove at the weld toe will usually carry its full static load without complaint, which is why shop crews wave it off as cosmetic. The problem appears later, after thousands or millions of load cycles, when the notch does what notches do: concentrate stress, incubate a crack, and hand it a place to start.

The weld toe is a natural geometric discontinuity anyway. Load lines must turn as they enter the weld reinforcement, and that turn already raises stress locally. A smooth, blended toe raises it gently; a sharp undercut groove raises it abruptly, and metal at the root of that groove can be carrying several times the nominal stress of the surrounding plate.

Static Strength vs. Fatigue: Two Very Different Stories

At the depths codes permit, undercut has a modest effect on static strength and a severe, disproportionate effect on fatigue life. Removing 1 mm from the toe of a 12 mm fillet weld does not meaningfully reduce the area carrying a one-time overload, so yield and ultimate numbers barely move. Fatigue does not care about average area; it cares about the sharpest local feature, and a single crack initiating at an undercut toe can propagate under loads far below the material’s yield point.

This is why the codes covered in our main guide to undercut limits treat cyclic service far more harshly than static service, and why fatigue-sensitive fabrication gets tighter as the consequence of failure rises, from aerospace weldments down to pressure-retaining joints.

What the Fatigue Research Actually Shows

The published data is consistent enough to quote in a WPS review or an inspector’s rejection report:

• Fatigue analysis of butt-welded joints found that the reduction in fatigue strength caused by weld toe undercut is at least twice that of equivalent joints without undercut, when both are compared against flush-ground plate. Grinding smooth is the benchmark; undercut roughly doubles the penalty.
• A T-joint study by Bell and co-workers, later reviewed in the International Journal of Fatigue, concluded that the deleterious effect of undercut is dominated by the stress concentration factor rather than depth alone. Undercuts shallower than about 0.5 mm did not measurably reduce fatigue life.
• A Japanese fillet-weld study reported that once weld toe undercut depth reaches roughly 1 mm, fatigue strength can drop by approximately 30%.

Together, the findings draw a line. Sub-half-millimeter undercut is often structurally benign; approach a full millimeter and the penalty is hard to ignore. The transition is not a cliff, but it is steep, which is why acceptance tables hover around the 0.5 mm and 1 mm marks. For deeper background, TWI’s work on fatigue design rules for welded structures is the standard reference.

Why Codes Limit Depth Instead of Sharpness

If sharpness drives the stress concentration, why don’t codes specify a maximum notch radius? Blunt practicality. AWS states plainly that there are no practical means to measure the sharpness of an undercut notch in a fabrication shop, so codes use depth as an overall measure of undercut severity — a measurable proxy for something we cannot gauge directly. Depth correlates well enough with severity, reads with a fillet gauge in seconds, and keeps inspection objective. That reasoning is documented in AWS’s welding handbook guidance (AWS B1.11) and reflected in the acceptance tables of AWS D1.1/D1.1M-2025 with Amendment 1.

Why Cyclic and Tension Members Get Tighter Limits

The tightening follows from the research above. A member in primary tension under repeated loading has every bit of its cross-section working, and the weld toe sits perpendicular to the load path, the worst possible orientation for a notch. The crack that starts there opens under every cycle and grows. A statically loaded member only has to survive its maximum load once, and it still has most of its area intact behind the notch.

So codes apply tiered tolerances: deeper undercut allowed on statically loaded, thicker material; shallower and more length-restricted undercut on cyclically loaded, thin, or primary tension members. Fatigue cracks in welded structures overwhelmingly initiate at weld toes, and the logic is unavoidable. The small groove you can barely see is not a cosmetic blemish. It is the most common starting point for the failure mode that governs the design life of welded structures, from wave-loaded aluminum hull joints to crane lugs and bridge connections.

How to Inspect and Measure Weld Undercut Correctly

Finding undercut starts long before a gauge touches the weld. The first step is cleaning: wire-brush the bead and the surrounding toes until the weld boundary is clearly visible. On aluminum, that means going further and using the right aluminum cleaner for welding prep, because oxide film and staining hide the very line you are trying to judge. Remove spatter, slag remnants, and grinding dust, then dry the surface.

Light Is Your First Measuring Tool

Once the weld is clean, angle your light. Flat, head-on illumination flattens everything and makes a shallow groove disappear. Raking light held at roughly 30 to 60 degrees casts a shadow inside the notch, so even a 0.010 in groove shows up as a dark line along the toe. Walk your light up and down the bead and look from both directions; undercut on the far toe is easy to miss from a single viewpoint. This is also where bare eyes start to struggle: TWI warns that measuring undercut is genuinely difficult because mill scale, surface irregularity, and spatter look similar to the imperfection itself at these sizes.

The Gauges Shops Actually Use

• Fillet weld gauges (the “round and hex” set): the most common shop tool. The radius notch sits on the leg, and the depth blade drops into the groove to read undercut depth directly.
• V-WAC gauge: a versatile weld inspection gauge that reads reinforcement, underfill, undercut, leg length, and throat in one body.
• Bridge cam gauge: reads profile height and undercut depth on groove welds, and doubles as a bevel and root-opening gauge.
• Pit gauge / depth micrometer: used when you need a numeric depth on a small pit or a shallow toe groove.

The technique matters as much as the tool. Set the gauge body flat and stable on the base metal, then bring the probe in perpendicular to the groove, not at an angle. An angled probe reads shallow and lets a real rejectable undercut pass. Read at the deepest point along the groove, and check both toes of every pass.

Shop experience helps, but formal training structures it. AWS B1.11, Guide for the Visual Inspection of Welds, frames visual examination in three phases: prior to welding, during welding, and after welding. The in-process phase is where undercut is actually prevented, because a welder who checks the toe while the bead is still hot can fix it with the next pass. Per the AWS Pocket Handbook for Visual Inspection, visual inspection conducted by properly trained inspectors discovers the vast majority of the defects that would otherwise only be found later by more expensive nondestructive test methods.

When Visual Is Not Enough

Undercut is a surface discontinuity, so ultrasonic and radiographic testing are generally the wrong tools for it — UT and RT see internal lack of fusion or porosity, not a toe groove. NDT is warranted when the same region also needs subsurface evaluation, when the joint geometry blocks direct visual access, or when a code requires complementary examination. The practical rule: if you cannot physically see and gauge the toe, you cannot accept or reject it.

Timing and qualification also control the call. AWS D1.1 requires that for ASTM A514, A517, and A709 Grade HPS 100W steels, acceptance criteria be based on visual inspection performed not less than 48 hours after weld completion, because delayed cracking may still be developing. Under D1.1, visual inspection is performed by a qualified inspector, and the person making the acceptance decision must be designated by the employer and trained to the applicable code requirements — which is one reason aerospace welding programs spend so much time on inspection discipline.

How Much Undercut Is Acceptable? AWS D1.1, D1.5, ISO 5817, and API Limits

Finding undercut is only half the job. The inspector’s next question decides whether you repair it, rerun the weld, or sign off: how much is too much? Every major fabrication code answers with numeric depth limits, and those numbers swing with the loading condition, material thickness, and even the direction the groove runs relative to stress.

Codes do not say “no undercut allowed.” They say “undercut within these limits is a tolerable discontinuity, not a defect.” Knowing which limit table applies to your joint is the difference between arguing with an inspector and having a factual conversation with one.

AWS D1.1:2025 Table 8.1 — The Baseline for Structural Steel

In the United States, the governing reference for most structural steelwork is AWS D1.1/D1.1M:2025 as corrected by Amendment 1 (February 2026). For nontubular connections, Table 8.1, item (7), breaks undercut limits into three practical categories.

• Static loading, material under 1 in (25 mm) thick: undercut must not exceed 1/32 in (0.8 mm) in depth.
• Static loading, material 1 in (25 mm) thick or greater: undercut must not exceed 1/16 in (2 mm) for any length of weld. Thicker sections tolerate a slightly deeper notch because the same groove is a smaller fraction of the load-carrying cross section.
• Cyclic loading, nontubular connections: undercut transverse to the computed tensile stress in primary members is limited to 0.01 in (0.25 mm). All other cyclic members are allowed up to 1/32 in (1 mm).

That 0.01 in surprises people: it is roughly the thickness of a business card, and it exists because transverse undercut in a tension member sits exactly where fatigue cracks start. The 1/16 in allowance for thick, statically loaded plate is nearly ten times larger, which says everything about how codes treat cyclic fatigue versus static strength.

The Accumulated-Length Rule and the 0.16 Factor

Depth is not the only limit. For welds 12 in or longer in the thin-material static category, undercut up to 1/16 in deep is permitted only when its accumulated length does not exceed 2 in in any 12 in of weld. For welds shorter than 12 in, accumulated undercut no deeper than 1/16 in must not exceed weld length multiplied by 0.16, a proportional allowance added by Amendment 1 in 2026, since a fixed 2 in threshold would be meaningless on short welds. A 6 in weld gets roughly 0.96 in before it fails.

Both limits must pass at the same time. A weld can be rejected on either one.

AWS D1.5 Bridge Welding Code: The 2025 Rewrite

Bridge work runs under AWS D1.5, tougher than D1.1 because bridge members are almost universally fatigue-cycled. The 2025 edition reworked undercut tolerances, and as the AWS Welding Digest summary of the changes explains, it introduced a notable 1/8 in tolerance at stiffener and connection-plate corners, where geometry makes a perfect toe profile nearly impossible and where past codes forced costly cosmetic rework on structurally insignificant areas. The broader trend is a preference for blending allowable undercut smooth rather than welding over it.

BS EN ISO 5817: Quality Levels B, C, and D

European and international work typically follows BS EN ISO 5817, which classifies welds into three quality levels: B (stringent), C (medium), and D (lowest). For material above 3 mm thick, allowable undercut depth runs from roughly 0.5 mm at level B up to 1 mm at levels C and D, with undercut never exceeding 0.5 mm at the highest quality level. ISO also limits undercut length to a fraction of total weld length, echoing the AWS accumulated-length philosophy.

A weld rejected under ISO level B may be perfectly acceptable under level C, so always confirm which quality level the drawing or project specification invokes.

API 650 and Other Industry Context

Storage tank fabrication under API 650 permits more generous visual tolerances than structural codes, since shell and roof seams are largely statically loaded. Pressure vessels, aerospace, and marine work each carry their own acceptance tables. This is why a welder moving between industries, say from structural steel to aerospace welding, must re-learn the numbers rather than carry old habits along.

Code / Condition	Allowable Undercut Depth
D1.1 static, material < 1 in	1/32 in (0.8 mm)
D1.1 static, material ≥ 1 in	1/16 in (2 mm), any length
D1.1 cyclic, primary member, transverse to tension	0.01 in (0.25 mm)
D1.1 cyclic, all other members	1/32 in (1 mm)
ISO 5817, thickness > 3 mm	0.5 mm (level B) to 1 mm (levels C/D)
D1.5 2025, stiffener/connection-plate corners	1/8 in (3 mm) special tolerance

The Spec Always Wins

One final rule overrides everything above: the project specification is the contract. A bridge owner, refinery, or EPC contractor can write a spec stricter than any base code, and when they do, the stricter number governs. Good fabricators build acceptance criteria into the procedure documentation before the first arc is struck, so the welder never has to guess which table applies.

Treat these limits as something to design out through technique, not something to measure and argue about afterward. The next section shows you how.

How to Prevent Undercut: Settings, Technique, and Procedure Discipline

Undercut is rarely caused by one bad decision. It usually comes from several small ones stacking up: an arc running a little hot, a travel speed creeping upward, and a wrist that stops watching the toe. Prevention works the same way in reverse. Align the settings, angles, and procedure discipline, and the groove stops appearing. The goal is not simply to weld slower. It is to keep the arc, filler, and puddle balanced from the first inch to the last.

Treat the Settings as One System

Most undercut starts as a heat-balance problem: the arc melts base metal faster or wider than molten filler can refill it. Amperage, voltage, wire feed speed, and travel speed are not four independent knobs. Change one, and the others usually need a matching adjustment.

• Amperage: Stay within the WPS range. If the bead looks flat and the toes are washing away, reduce current slightly before adding more filler.
• Voltage or arc length: Excessive voltage widens the arc and spreads heat across the base metal. On GMAW and FCAW, bring voltage down in small steps and listen for a crisper arc.
• Wire feed speed: In GMAW and FCAW, WFS controls deposition. Increase travel speed without enough WFS and the puddle starves, leaving hollow toes.
• Travel speed: Rushing is the most common single cause. Slow enough for the puddle to wet both edges, but not so much that you build a tall, cold bead.

On thin material the window is narrow. Reduce amperage, shorten the arc, and watch heat input. Thin sheet overheats long before the weld looks obviously wrong.

Angle and Travel Technique

A slight drag angle of roughly 10 to 15 degrees helps weld metal flow back into the groove and fill the melted base metal. Pushing the puddle tends to throw heat ahead of the arc and leave the trailing toe starved, especially in vertical and overhead work.

On T-joint fillets, hold a consistent 45-degree work angle. That splits heat evenly between both legs and prevents preferential melting of one toe. If the work angle drifts toward the vertical member, that side overheats while the other gets too little fusion.

Bead style matters as much as angles. Stringers are easier to control at the toes because the puddle stays compact. When a weave is necessary, do not whip through the edges. SSAB recommends small stops at the end of each oscillation so the pool can fill the toe region before you move on. The pause should be brief but deliberate.

Process-Specific Habits That Save the Toe

• SMAW: Keep a short arc, use the correct electrode angle, and let the slag ride behind the puddle. On vertical-up work, watch the toes constantly. Our all-position welding rod guide covers electrode choices for these positions.
• GMAW and FCAW: Match WFS to travel speed and keep stickout stable. Do not raise voltage to smooth a bead that is really suffering from poor travel technique.
• GTAW: Control arc length with the pedal and add filler at the leading edge of the puddle. Taper current at stops, where undercut often appears.
• SAW: Check wire alignment before every joint. Wire tracking too close to the groove face melts the sidewall without filling it.

Procedure Discipline: The Part Welders Cannot Wing

A qualified WPS is the foundation. Use prequalified joint details where the code permits, and make sure the parameter ranges used on the floor match the procedure. AWS D1.1 requires welder and welding operator qualification for the specific process, position, and material being welded. That is not paperwork for its own sake. It is the structural backbone of any serious undercut prevention program.

Practice on the actual process and position before production. Coupons reveal whether your travel angle and speed suit vertical-up or overhead work. For arc blow, move the work clamp closer to the joint, clean the ground contact, and change the ground location or electrode angle. On thin material, keep heat input under control rather than relying on speed alone.

Then use in-process self-checks. After each pass, look at both toes before starting the next one. If a groove is forming, adjust the settings and technique immediately. Catching a shallow toe early is far cheaper than repairing it after inspection.

Undercut prevention is a habit, not a single setting. Balance the heat, watch the angles, pause at the toes, and follow the procedure. Do those four things consistently, and most undercut never gets a chance to form.

How to Repair Undercut the Right Way: Grinding, Welding, and When to Do Each

Once an inspector flags undercut, you have exactly two legitimate ways out: smooth it out with a grinder, or put qualified weld metal back into the groove. There is no third option, and “it’ll be fine” is not one of them. What surprises most welders is that the preferred route has recently flipped. For decades, the instinct was to bury the groove under a quick cosmetic bead. The 2025 rewrite of the AWS D1.5 Bridge Welding Code moved the industry the other way: for undercut at or below the allowable depth, grinding is now the prescribed treatment, not weld metal.

Route One: Blend Grinding

Blend grinding means using a abrasive wheel or burr to feather the sharp notch at the weld toe into a smooth, gradual transition between weld face and base metal. You are not removing the weld; you are removing the notch geometry, which is what actually drives fatigue cracking. The finished surface should show no visible tool marks or step, and you should be able to drag a fingertip from base metal over the toe without catching.

D1.5’s new rule comes with guardrails, and they matter:

• The ground surface must be smooth, with no gouges or ridges left by the tool.
• Grinding must not reduce the net cross section below 98% of the nominal area.
• At toes transverse to the direction of tension, the depth removed after grinding is capped by plate thickness, starting at 1/32 in for plates up to 1 in thick and rising to 1/8 in for plates over 2 in thick.

That last point is the real trade-off. Grinding kills the notch, but it also removes a thin slice of load-bearing throat. On a heavy-section member the loss is trivial; on a thin fillet it can matter, which is why the caps scale with thickness. For cosmetic aluminum work where blending is routine anyway, the same philosophy applies, and our alloy wheel welding guide shows how far grinding can safely go on a repaired section.

Route Two: Weld-Metal Fill

When undercut exceeds the allowable depth, exceeds the grinding caps, or simply sits somewhere grinding can’t reach cleanly, it gets welded. AWS CWI field guidance also flags a trap here: some undercut-looking valleys actually contain trapped slag inclusions from the original pass. Welding over that groove seals the defect inside a sound-looking bead, so deep or suspicious pockets must be gouged or ground out to sound metal and refilled, never covered.

The proper sequence looks like this:

1. Clean out the groove. Gouge or grind the undercut back to bright, sound metal. Dig out any slag, oxide, or spatter sitting in the valley. On aluminum, that means proper chemical and mechanical prep as covered in our guide to aluminum cleaner for welding, because a clean groove is non-negotiable.
2. Select a small, low-hydrogen electrode. A 1/8 in or 3/32 in E7018-class rod keeps heat input low and gives you control in a narrow groove.
3. Drop the amperage and run stringer beads. Fill the groove with one or two small stringers at reduced current, keeping the work angle pointed at the toe being filled.
4. Dress and re-inspect. Blend the new bead smooth, then measure it again against the original acceptance criteria.

The Four Classic Repair Mistakes

Almost every bad repair falls into one of four boxes:

• Welding over a dirty or slag-filled groove. The bead bridges the contamination instead of fusing to it, converting undercut into a lack-of-fusion defect plus a slag inclusion.
• Leaving a crater at the stop. Crater cracks and crater undercut are just new discontinuities manufactured during the repair.
• Creating fresh undercut on the opposite toe. Pushing too much current or travel speed into the fill melts a new notch on the far side of the bead.
• Skipping re-inspection. A repaired weld is not a finished weld.

That last one carries explicit code weight. Under AWS D1.1, repaired welds must be retested by the same method originally used and judged by the same acceptance criteria applied to the original weld. If it was visually inspected and gauged the first time, it gets visually inspected and gauged again. If it was ultrasonically tested, it goes back under the probe. The repair loop only closes when the second inspection passes, and the rework is documented in the welder’s or shop’s records.

Conclusion: Undercutting at a Glance and a Shop-Floor Checklist

Strip away the code tables and inspection jargon, and the undercutting welding definition collapses to one clean model: a groove of base metal that was melted away and never filled back in. It sits at the weld toe or root, measured in fractions of a millimeter or thousandths of an inch, and matters for where it sits, not how big it looks. Undercut is measurable, preventable and repairable — but in cyclic service that toe notch is one of the most reliable fatigue crack initiation sites in welded steel.

Everything else here — the AWS discontinuity-versus-defect distinction, toe versus root undercut, causes, gauges and acceptance limits — serves that model.

The Four Things to Remember

• Judged by depth, with tighter limits under cyclic tension. A groove that passes on a static member can fail outright on a fatigue-critical one; D1.1 and D1.5 treat primary tension members most strictly.
• Almost always a heat-balance and technique error. Travel too fast, amperage too high, arc too long, wrong angles, a weave that never pauses at the toes: the arc melts base metal faster than filler refills it.
• Preventable with discipline, not talent. Correct angles, controlled travel speed, stringer beads or paused weaves, and parameters treated as one balanced system.
• Repaired by blending or qualified weld fill, never by ignoring it. The 2025 D1.5 Bridge Welding Code rewrite tightened fatigue tolerances and formalized grinding as the preferred first treatment for allowable undercut.

The Ten-Point Shop-Floor Checklist

1. Parameters set per WPS. Amperage, voltage, wire feed and travel speed as one balanced system, not four knobs tweaked to look nicer.
2. Correct angles. A consistent drag angle of about 10–15 degrees for electrode-positive work, plus a steady 45-degree work angle on fillets.
3. Slowed travel at the toes. Ease off and let the puddle wash in and fill the edges instead of racing across.
4. Stringer beads or deliberately paused weaves. A weave that whips through the toes without a pause undercuts both sides.
5. In-process visual check. Watch the toe while the puddle is hot; a groove forming now is fixed by technique, not by a grinder later.
6. Gauge measurement. Clean the weld, rake light across at 30–60 degrees, and read depth perpendicular with a fillet gauge, V-WAC or bridge cam gauge.
7. Correct acceptance table. Static or cyclic, nontubular or tubular, thickness and any stricter project spec: confirm the governing row before passing anything.
8. Clean repair. Grind or gouge to sound metal, remove trapped slag, and fill with a small-diameter low-hydrogen stringer at reduced amperage.
9. Re-inspection. The repaired area is judged to the original acceptance criteria, not a relaxed “it’s only a repair” standard.
10. Documentation. Record the discontinuity, disposition, repair and re-inspection result; traced quality is what makes a good day defensible.

Why the Trained Eye Still Wins

Visual inspection finds the vast majority of weld defects, and undercut is a surface discontinuity. Ultrasonic and radiographic testing have their place, but on the average floor a qualified inspector, good lighting and a ten-dollar gauge remain your highest-value quality control tools.

That is why fatigue-critical fabrication keeps tightening these tolerances: as structures get lighter and service lives more demanding, a razor-thin toe notch works further against the fabricator. The gap between looks-good-from-three-feet and code-compliant is widening; the shops that thrive catch undercut before the inspector does.

Frequently Asked Questions

What is undercutting in welding?

Undercutting is a groove melted into the base metal adjacent to the weld toe or weld root and left unfilled by weld metal, as defined by AWS A3.0. It appears as a narrow channel along one or both edges of the weld bead and reduces the cross-sectional thickness of the base metal.

What is the difference between undercut and underfill?

Undercut is a notch cut into the base metal at the weld toe or root, while underfill is simply insufficient weld metal deposited in the joint, leaving the weld surface below the required reinforcement level. Undercut is a stress riser; underfill is a throat-thickness problem, and they are corrected differently.

How much undercut does AWS D1.1 allow?

Under AWS D1.1:2025 Table 8.1, statically loaded connections on material under 1 in thick are limited to 1/32 in depth, with limited accumulated-length exceptions up to 1/16 in, and material 1 in or thicker is limited to 1/16 in. For cyclically loaded connections, primary members stressed in tension are limited to 0.01 in and all others to 1/32 in.

Is undercut a defect or a discontinuity?

Undercut is always a discontinuity, but it only becomes a rejectable defect when it exceeds the acceptance criteria of the governing code or project specification. A 0.5 mm undercut might be acceptable on a static bracket and completely rejectable on a fatigue-critical tension member.

Does undercut weaken a weld?

Yes, but not mainly through static strength at code-allowed depths. Its biggest danger is fatigue: the notch concentrates stress at the weld toe, where most fatigue cracks in welded structures begin, and research shows undercut can more than halve fatigue strength relative to a flush-ground joint.

How do you measure undercut depth?

After cleaning the weld, place a fillet weld gauge, V-WAC gauge, bridge cam gauge, pit gauge, or depth micrometer perpendicular to the groove and read the depth below the base metal surface. Good raking light makes shallow undercut easier to see, but shallow depths around 0.01 in generally require a gauge rather than the naked eye.

Can undercut be repaired by grinding alone?

Often, yes. If the undercut is only slightly beyond the acceptance limit, blend grinding can smooth the weld toe transition and bring it into compliance, provided it does not reduce the net cross section below the code’s floor — AWS D1.5 uses a 98%-of-nominal-area rule. Deeper undercut must be ground clean and refilled with weld metal per a qualified procedure.

What is the single most common cause of undercut in production welding?

Travel speed that is too fast for the deposition rate is the most frequently cited cause, closely followed by excessive amperage or voltage and incorrect electrode angle. The arc melts a groove at the toe and moves on before filler metal can flow in and fill it.

Why do fatigue-loaded structures have stricter undercut limits?

Undercut acts as a sharp notch at the weld toe, and cyclic loading drives crack initiation from exactly those stress concentrators. Codes such as AWS D1.1 and AWS D1.5 therefore limit undercut to as little as 0.01 in on members carrying primary tensile stress, because depth is the practical proxy for notch severity.