● THE WELDING KNOWLEDGE HUBGUIDES / REVIEWS / COMPARISONS
WELD DEFECTS & INSPECTION

Concavity in Welding: Causes, Code Limits, Inspection, and How to Fix a Concave Weld

By sohagarbin October 14, 2026 35 min read

Table of Contents

What Is Concavity in a Weld? A Plain-Language Definition

You finish a fillet weld that looks smooth and clean. The ripples are even, the toes tie in, and you are ready to move on. Then you slide a gauge across the face and the needle tells a different story: the center of the weld sags below the line joining the two toes. That gentle hollow is concavity, and it is one of the quietest ways a weld can lose strength.

Concavity is simple to define. It is a weld face or root surface that dips below the theoretical plane connecting the weld toes. On a fillet weld, picture a straightedge laid across the two toes; if the weld surface curves below that straightedge, you have a concave profile. It is the opposite of convexity, where the face bulges above that same plane.

The Shape of the Problem: Below the Line

Concavity is a shape imperfection rather than a crack or a hole, but shape controls how a joint carries load. Industry references group it with other geometric imperfections that change a weld’s cross-section instead of its internal soundness, as this review of fillet weld shape imperfections explains.

Concavity, Convexity, and a Flat Face

Run a straightedge across the toes and you get three possible outcomes:

  • Flat face: level with the plane through the toes.
  • Convex face: bulges above that plane, adding bulk but no usable throat.
  • Concave face: dips below the plane, removing throat the design counted on.

A flat profile is closest to the ideal for a fillet; convexity is tolerated but wastes filler, while concavity eats into strength.

Why Throat Thickness Is What Actually Counts

A fillet weld does not carry load through its legs; it carries it through its throat, the shortest distance from the weld root to the face. For a 90-degree joint, the theoretical throat equals 0.707 times the leg size. Weld a leg of 3/8 in (9.5 mm) and the design credits roughly 0.265 in (6.7 mm) of throat.

That is the number the engineer used to size the joint. AWS D1.1 credits only the theoretical throat in design, and surplus convexity adds no strength. Concavity does the opposite: it removes throat, so a weld can pass a casual glance and still fall short of the capacity the drawing assumed.

Face Concavity vs. Underfill

Concavity shows up in two places. On a fillet weld, it describes the face. On a groove weld, the same shortfall below the surrounding base metal is called underfill. Both leave the joint thinner than designed, and both are treated as geometric defects rather than internal flaws.

Quick Note: A concave fillet and a convex fillet can both look “smooth.” Smoothness tells you nothing about throat thickness — only a gauge or a profile measurement does.

What This Guide Covers

Concavity rarely appears by accident. It comes from heat input, travel speed, joint fit-up, and technique, and the process you run shapes which fix works. This guide walks through what causes it, what the major codes allow, why root concavity behaves differently from face concavity, how to measure it, and how to prevent or repair it.

If you are still building the fundamentals, our guide to welding practice and to telling good from bad MIG welds will help you read a bead before a gauge does. For process-specific fixes, see our notes on contact tip wear in GMAW and overlap as a defect, since many of the same parameters drive both.

Bottom line: concavity is a shape you cannot judge by eye alone. If it sits within code limits, the weld stands. If it drops the throat below what the design needs, it has to be repaired. Use the rest of this article as a checklist before you sign off a weld or call for a grinder.

Concavity vs. Convexity, Underfill, Undercut, and Root Concavity

On the shop floor these terms get used almost interchangeably. In an inspection report, they mean something very specific. The fastest way to understand concavity is to compare it to its mirror image, then separate face-side problems from root-side ones.

Two Sides of One Line: Convexity and Concavity

Convexity and concavity are mirror images measured against the same theoretical flat-face triangle — the imaginary straight line joining the two weld toes. A face that bulges above that line is convex. A face that sags below it is concave. Same reference line, opposite direction.

The eye is easily fooled. A wide, smoothly blended convex bead can look strong, yet it adds weight and can mask a shallow throat; a neat concave face can look tidy yet starve the joint. Profile only matters in relation to the line you draw between the toes.

ISO 6520-1, the international classification of weld imperfections, treats them as separate numbered items: excessive convexity is reference 503, while root concavity is reference 515. On a butt weld the equivalent overfill gets a different label entirely — excess weld metal, reference 502. So 502 is the butt-weld version of too much reinforcement, and 503 is the fillet-weld version of the same overfill habit. Shape imperfections like these are among the most common geometric defects inspectors flag.

Underfill, Undercut, and Overlap: Neighbours, Not Synonyms

Underfill applies to groove welds, where the face should finish flush with — or slightly proud of — the surrounding base metal. When deposited metal sits below that level, you have underfill, also called insufficient reinforcement. It is not concavity: concavity describes the curve of the face, while underfill describes the face sitting below the plate surface.

Undercut is a groove melted into the base metal along the weld toe and left unfilled — a notch in the parent metal, not the weld. Overlap is the opposite behaviour at the same toe: molten metal rolls over the surface without fusing. Both are taken seriously because they create stress concentrations. For a deeper look, our guide to the overlap welding defect covers causes, detection, and repair.

Face-Side Concavity vs. Root-Side Concavity (Suck-Back)

A concave face sits on the side you welded. Root concavity — often called suck-back — appears on the far side of an open-root joint, where the root bead sags inward as the pool freezes. Familiar in pipe, tube, and open-root work, it is shaped by fit-up, as our weld tube fittings guide shows.

Crucially, root concavity is not lack of penetration. In a genuine lack-of-penetration defect the weld never reaches full depth. With root concavity the weld reaches full depth — the molten pool simply sags inward before it solidifies, thinning the root. That difference changes both the diagnosis and the fix. If you are still building an eye for profile, a session on good and bad MIG welds trains it fast.

A Simple Way to Name What You See

Before you reach for a standard, match the feature to its location:

Feature Where it sits What it means
Concavity Face, below the toe line Throat thickness at risk
Excessive convexity (503) Face bulging above the toe line Excess metal, poor profile
Excess weld metal (502) Butt-weld face above plate level Overfill
Underfill Groove-weld face below plate surface Insufficient reinforcement
Undercut Groove in base metal at the toe Notch, stress riser
Root concavity (515) Far side of an open-root joint Root thinning, not LOP

Quick Note: A slightly concave fillet face is actually permitted by most codes, provided the minimum weld size is met considering both leg and throat. A concave groove-weld face is treated differently — it is usually read as underfill, because that face is required to finish flush or positive.

Learn to name the defect first, and the standard you reach for next will finally make sense. Whether you are self-checking practice welds or signing off an inspection report, precise language is half the job.

Why Concavity Matters: Throat Thickness, Strength, and Fatigue

A weld is a shape, but an engineer sees a number. The throat dimension is where the strength lives, and concavity quietly steals from it. On a fillet weld, the face should reach the line joining the two toes. When it sags below that, the effective throat shrinks. On statically loaded joints the safety margin often absorbs it. On cyclically loaded joints, or on partial joint penetration (PJP) groove welds sized around the theoretical throat, that missing millimeter is the difference between a passing joint and a rejected repair.

Throat Thickness: The Design Credit You Can Lose

PJP groove welds are the most vulnerable. The engineer calculates capacity assuming a theoretical throat measured from the root to the weld face. A concave face curves inward, so the actual throat is less than nominal, and the load-carrying capacity drops with it. A weld that passes a visual once-over can still be undersized where it matters—a good-looking bead is not the same as a sound one, a theme we explore in good and bad MIG welds. That is why AWS D1.1 and related references tie throat measurement directly to acceptance rather than to appearance.

AWS D1.1 gives inspectors a practical order of operations: measure the throat first. If the throat meets the requirement, a concave fillet face can pass as a profile condition. If the throat comes up short, the weld fails regardless of how smooth the face looks. That single rule is why welders should think less about how a bead looks and more about whether the measured throat clears the drawing. Getting that measured throat right, pass after pass, starts with disciplined welding practice and clean contact tips that keep the arc stable.

Fatigue: Shape Controls Stress Concentration

Under cyclic loading, profile is anything but cosmetic. Irregular weld geometry concentrates stress at the weld toe, a well-documented fatigue initiation site. A concave face creates a notch-like transition where the weld meets the base metal, and repeated stress cycles can open a crack there long before the parent metal is fully stressed. Weld-toe grinding and profile control exist precisely because geometry, not just size, drives fatigue life. TWI’s review of shape imperfections grades these toe transitions on sharpness and depth for exactly this reason: a smooth, shallow hollow behaves very differently from an abrupt notch.

When Static Loads Buy You Some Slack

For statically loaded structures the risk is lower, so a minor concave fillet on a bracket that only sees occasional, gentle loading usually keeps adequate margin. Gross concavity is still a red flag, though, because it rarely arrives alone. It tends to signal excess travel speed, low heat input, or shaky technique—the same conditions that produce incomplete fusion, overlap welding defects, and other hidden problems in fillet welds. Reading concavity as a process symptom is usually more useful than judging it as a standalone flaw.

Pressure Piping and Pipelines: Thickness Is the Limit

On pressure piping and pipelines, concavity is assessed against minimum wall or design thickness, not shape alone. ASME B31.1 permits root-side concavity only when the resulting weld thickness is at least equal to the thickness of the thinner member and the contour is smooth without sharp edges. That is a thickness rule wearing a shape requirement—which is why an inspector reaches for a gauge and a wall-thickness reading, not just a visual. Field references such as SSAB’s guide to root concavity make the same point: thickness first, contour second.

Why Concavity Foils Inspection

Concavity also makes nondestructive examination harder. A curved or hollow face scatters ultrasonic signals and muddies radiographic interpretation, which is exactly where lack of penetration or fusion may be hiding underneath. The geometry that defines the defect can help conceal it—one more reason inspectors treat deep concavity as a stop-and-check condition.

Quick Note: Measure the throat before you judge the profile. If the throat clears the drawing, a concave fillet face can pass; if it does not, no amount of smooth blending saves it.

The Causes of Weld Concavity: Process, Parameters, and Technique

Every concave weld shares the same basic story: a molten pool that was too large, too hot, or too unsupported to hold its shape, so it sagged and drew inward before it froze. The cause behind that story, though, changes with the process, the parameters, and the operator’s hand. Reading concavity correctly starts with working backwards — cause, to mechanism, to control.

Face Concavity in Fillet and Groove Welds

On a fillet or groove face, concavity usually means the bead spread wider and thinner than the joint can support. Four things do most of the damage:

  • Travel speed too high. Rushing deposits less metal per unit length, so the pool can’t bridge the toes and falls into a hollow. This is the classic trigger in flat and horizontal fillets.
  • Voltage too high for the amperage. Excessive voltage widens and flattens the arc, spreading the bead and thinning the face.
  • GMAW spray transfer in short, fast passes. Spray gives high heat and fluid metal; combined with quick passes, it lays a bead that never builds a crown.
  • Low deposition relative to joint size. An undersized weld on a wide gap will always sit below the toe line.

Torch angle feeds into all of these. Too much push angle spreads heat ahead of the pool; too much drag angle starves the toes. Watch contact-tip condition too — a worn or spattered tip distorts the arc and bead shape, a point covered in our guide to contact tip welding.

Root Concavity and Suck-Back

Root concavity is a different animal. SSAB describes it as shrinkage of the weld pool in the through-thickness direction — the root simply contracts and pulls back as it solidifies, leaving a dip below the parent metal surface. Melting of the root pass by the second pass can produce it too, when heat from the fill pass reaches through and re-melts metal that had already bridged the gap.

What pushes the pool into that state?

  • Root gap too small (insufficient penetration and a fragile bridge) or too large (a pool with nothing to hold it).
  • Oversized root face, which demands more heat to fuse and lets the pool hang.
  • Missing or weak backing-gas purge, so the hot root oxidises and sags.
  • Excessive heat input or travel speed on the root pass.

Therness notes root concavity is frequently associated with TIG welding and usually traces back to poor preparation, especially a root gap that is too small or too large. The TWI review of shape imperfections makes the same point: geometry and heat control decide whether the root bulges, sits flat, or sucks back. SSAB’s remedy is straightforward — keep the root face to a maximum of 1.5 to 2 mm and adjust the torch angle to reduce heat input.

Quick Note: A sagging root and a sagging face look similar on a gauge, but they come from opposite ends of the joint — treat them separately.

How Each Process Earns Its Concavity

GTAW/TIG: the manual pool control that makes TIG precise also makes it the process most tied to root concavity. Without a stable gap or good fit-up, the pool sags. Getting current and polarity right matters here too — see our guide to AC or DC for TIG welding steel.

SMAW/MMA: slag and slower freezing hide the pool, so operators overheat thin sections and run a long arc, flattening the face without noticing until the slag is chipped.

GMAW/MIG: spray transfer plus fast travel is the textbook face-concavity recipe; short-arc and pulsed runs rarely show it, but excessive inductance or a long stick-out can still flatten the bead.

FCAW/SAW: high deposition and deep, fluid pools react badly to an open root, where the driving arc can punch through and suck the bead back. See polarity for flux core welding.

The common thread is control, not luck. Match electrode size and current to the joint, keep travel speed and voltage in step, verify fit-up and root face before the arc starts, and practise on scrap until the pool tells you where it’s heading. Our complete guide to welding practice walks through that routine step by step.

Acceptance Criteria: What AWS D1.1, ISO 5817, ASME, and API 1104 Allow

A concave weld isn’t automatically a defect—it’s rejectable only when a governing code says so. Here’s the catch that trips up fabricators: a weld can look “off” and still pass, or look smooth and fail hard. Everything hinges on which standard governs your work, what the joint carries, and how the inspector measures the sag. “It depends on the code” isn’t a cop-out—it’s the whole answer.

Let’s walk through the four standards you’ll meet most often on real jobs.

AWS D1.1: Concave Fillet Faces Are Allowed

Under AWS D1.1:2025, the Structural Welding Code—Steel, a fillet weld may have a concave face. The code does not restrict fillet concavity as long as the minimum weld size is maintained considering both the leg and the throat. That last clause is what inspectors hammer on. A deep concave face can satisfy the leg dimension at the toes while cutting the actual throat below the design assumption. If the effective throat falls short, the weld fails—no matter how tidy it looks.

Where D1.1 draws a hard line is on the opposite problem: excessive convexity. Table 7.9 Schedule C sets maximum convexity based on weld face width:

Weld Face Width Maximum Convexity
Up to 5/16 in (8 mm) 1/16 in (2 mm)
Over 5/16 in up to 1 in (25 mm) 1/8 in (3 mm)
1 in (25 mm) and above 3/16 in (5 mm)

For groove welds, D1.1 is stricter on the low side. Underfill is not permitted on complete joint penetration (CJP) welds—the face must sit flush with the base metal or show positive reinforcement within limits. A groove weld face dipped below the plate surface gets flagged immediately, whereas a concave fillet face might sail through if the throat holds.

Quick Note: D1.1’s tolerance for concavity isn’t a licence to under-weld. Always verify the throat, not just the leg—that’s the measurement that decides pass or fail.

ISO 5817: Concavity Scaled to Thickness

Where AWS works in inches and simple schedules, ISO 5817:2023 takes a proportional approach. It defines three quality levels—B, C, and D—where B is the highest requirement on the finished weld and D the lowest. Your contract or drawing tells you which level applies.

For root concavity on thickness over 3 mm, the limits scale with material thickness (t):

  • Level D (moderate): h ≤ 0.2t, capped at 2 mm—a smooth transition required.
  • Level B (stringent): h ≤ 0.05t, capped at 0.5 mm—a smooth transition required.

That smooth-transition requirement matters as much as the depth number. A steep, abrupt hollow is treated more harshly than a shallow, blended one because it concentrates stress. If you’re chasing a specific level, read the acceptance table carefully rather than guessing from memory.

ASME B31.1 and B31.3: Pressure Piping

For power and process piping, ASME B31.1 (power piping) and B31.3 (process piping) set the rules. They evaluate concavity against reduced-wall-thickness and reinforcement criteria tied to the pipe’s design pressure and temperature. The logic is unforgiving—pressure boundary welds can’t tolerate a hollow that weakens the wall below the calculated minimum, so an inspector cross-checks concavity against the allowable reinforcement and undercut envelope rather than applying a single number.

API 1104: Pipeline Girth Welds

For transmission pipelines, API 1104 governs girth welds, including root concavity (suck-back). Its acceptance criteria are written around fracture mechanics and radiographic/visual evaluation, with limits that vary by wall thickness and imperfection category. Because pipelines operate under pressure and cyclic loading, API 1104 treats a deep root hollow seriously—particularly high-low and lack-of-penetration combinations.

Why the Governing Code Wins

Notice the pattern. AWS D1.1 is generous on fillet concavity but strict on CJP underfill. ISO 5817 grades severity by quality level and thickness. ASME and API frame everything around service conditions. No single “concavity limit” exists across the industry, so the first question on any suspect weld should be: which code applies?

Get the governing standard right before you pick up a gauge or a grinder. If you’re building habits that keep these defects out of a joint, spend time on deliberate welding practice and brush up on how to tell good and bad welds apart—those instincts make code checks faster. You can also review the industry’s broader acceptance guidance for shape imperfections to see how different codes converge on one core principle: throat thickness and load path, not cosmetics, decide whether a concave weld passes.

Root Concavity (Suck-Back) in Pipe and Open-Root Welding

Fillet concavity costs a little throat thickness. Root concavity is a different animal. On pipe, pressure vessels, and pipeline girth welds, the root sees product, pressure, and fatigue cycles first — and suck-back hollows it from the inside. Bend tests don’t always catch it; neither does an ambient-temperature hydrostatic test. That’s why inspectors rank it among open-root welding’s most consequential defects.

What Root Concavity Actually Looks Like

On a proper open-root pass, the underside should show a bead that fills to the parent-metal surface, sitting barely convex. In a concave root, the pool sags inward, leaving the finished profile below the pipe’s inside surface — a groove or “sucked-back” trough where metal should be. Severe cases read as a channel running along the seam’s underside, with sharp transitions at the sidewalls.

The geometry matters: root thickness carries hoop stress in thin-wall pipe. Sag the root inward and you thin the wall exactly where tensile stress peaks. Per SSAB’s defect reference, concavity reduces the effective throat and acts as a stress raiser — hence far tighter limits than convexity.

Why the Purge Is the Leading Control

Backing gas purge is the leading control. Without a stable purge holding pressure behind an open root, the molten pool sags under its own weight as it solidifies. On stainless and other reactive alloys, argon or an argon-hydrogen mix shields the hot root from oxidation and presses back against gravity and surface tension so the pool freezes flat.

Quick Note: Purge pressure is a balance, not a “more is better” setting. Too little and the pool sags inward; too much and you blow the root bead outward, thin the root, or cause porosity. Set it by procedure and verify with a purge dam and an oxygen meter — not by ear.

Hold time counts as much as pressure. Stop the purge the instant the arc stops and the last few millimetres of still-molten root can drop before freezing. Keep gas flowing until the root falls below the sag temperature — several seconds minimum on small pipe, longer on heavy wall.

Root Gap, Land, and Heat Input

A root gap or land wider than the procedure allows leaves less parent metal to support the root pool, raising the risk of inward draw. The open root is held by surface tension, the land, and the purge; widen the gap and you remove that support. A thin land is equally poor — too little metal underneath for the root to key into.

Excessive root-pass heat or slow travel increases sag too: more amps make the pool larger and more fluid, and slower travel gives gravity more time to work. Root passes want enough heat for fusion and penetration, but not so much that the pool outruns the purge’s ability to hold it — a core welding practice skill on pipe work.

GTAW vs. Consumable-Electrode Roots

GTAW is common for root passes: it’s a controlled, low-deposition process with independent arc and filler control. AWS D10.11, the recommended practice for root pass welding of pipe without backing, cites GTAW for that precise control — dial heat down and freeze quickly. Getting TIG polarity and arc control right on DCEN is central to a clean, flat root. Consumable-electrode processes like GMAW, FCAW, and SMAW deposit faster but run hotter and more fluid, making open roots more sag-prone — a key reason flux-cored and stick roots need tighter technique. See how DCEN vs. DCEP polarity shifts heat balance.

Detecting Root Concavity: Accessible vs. Not

When the root side is accessible, visual inspection to ISO 17637 with a root or undercut gauge is direct and reliable — you measure trough depth against the parent surface. On most pipeline girth welds the root side isn’t accessible, so radiographic testing takes over. RT reads the thinned profile directly: the sagging root shows as a darker band on the film, giving the inspector its depth and length. API 1104 sets the acceptance limits for those girth welds.

Either way, the fix comes back to three controls: a stable purge with adequate hold time, a root gap and land within procedure, and heat input matched to the root’s ability to freeze flat.

How to Measure and Inspect for Concavity

Concavity is a geometry problem, and geometry is measured with a gauge — not guessed by eye. Whether you are a welder checking your own work or a building real welding skills mindset (see our welding practice guide) applies, a few seconds with the right tool settles the pass/fail question far more reliably than a squint and a shrug. Let’s walk through the practical methods for fillet and groove welds, then the tooling and inspection codes behind them.

Measuring a Concave Fillet Weld

For fillet welds, the governing dimension is the actual throat, not the leg and not the depth of the sag. Here is the sequence:

  1. Use a calibrated fillet weld gauge or a bridge cam gauge to read the leg length.
  2. Measure from the root of the joint to the deepest point of the concave face — that is your actual throat.
  3. Compare it to 0.707 × the specified leg size (the theoretical throat of a mitre fillet).

This is the CWI rule worth tattooing on your hood: measure throat first. If the actual throat meets or exceeds 0.707 times the specified leg, a concave face passes. If the throat is short, the weld fails on throat — the concavity is just the visible evidence of it — and it must be built up with additional passes to restore the throat thickness.

A bridge cam gauge is particularly useful here because it reads both the throat and the leg in one setting, and its pointed arm can reach down into a scooped face. A simple fillet gauge set with concave/convex scales does the same job on straight sections.

Groove Welds: Underfill vs. Reinforcement

For groove welds the measurement datum changes. Reinforcement and underfill are both measured from the base metal surface on both sides — never from the weld face itself. If the weld face sits above the base metal plane, that is reinforcement (capped by a maximum limit). If it sits below, that is underfill, an under-thickness condition. A concave groove face is simply reinforcement that dips below the datum on part of the cap.

Quick Note: Not every fillet weld gauge set includes convexity or concavity scales. Many basic sets only read leg and throat. If your set lacks a concave scale, measure the actual throat against 0.707 times the leg — that comparison is all you need for acceptance.

Visual Inspection: The Primary Route

Concavity is an external profile condition. Per AWS D1.1 Clause 8.9 and Table 8.1 item 4, weld profiles must conform to Clause 7.23 for both statically and cyclically loaded nontubular connections. Internationally, ISO 17637 governs visual testing of fusion welds, and TWI’s review of shape imperfections reinforces that face geometry is a visual acceptance matter.

Good lighting, clear magnification, and a clean surface are prerequisites — the same care that keeps you safe behind protective eyewear keeps your eyes reliable for inspection. Consistent technique from your good and bad MIG welds habits pays off at the inspection bench too.

Why Volumetric NDT Won’t Help (Mostly)

Because concavity lives on the surface, radiographic and ultrasonic testing have nothing to report on a concave face. Volumetric NDT looks inside the joint; a scooped cap is fully in view already. The exception is root concavity (suck-back), where access is limited and radiography becomes the primary route to reveal the internal profile. For surface geometry, visual testing plus gauges remains the workhorse.

Document your readings, match them to the applicable acceptance table, and let the numbers — not the shine of the weld — decide.

How to Prevent Concavity: Parameters, Fit-Up, and Technique

Concavity almost never appears because of one dramatic mistake. It creeps in when several small variables drift at the same time: travel speed slows, the tip walks too far from the work, a root gap opens up, purge flow sags. Prevention is therefore a playbook, not a single setting. Control the variables before the arc starts, then watch the ones that move while you weld.

Parameter Control: Match the Machine to the Joint

On manual fillet welding, travel speed is the parameter most likely to drift — and the most common cause of profile problems. Fast enough to keep the pool fluid, slow enough to fill the joint, and steady enough that the toes wet out. A travel speed too slow for the deposition rate dumps more metal than the joint can absorb, and the face crowns up into convexity. Too fast flattens the pool, starves the toes, and leaves the face sagging below the line joining them.

Voltage is the other half of the equation. Run voltage too high relative to amperage and the arc spreads, the pool turns thin, and the face hollows. Run it too low and you get the opposite failure. Match wire feed speed, current, travel speed, and joint size as one set — not four independent dials. A cold pool that never spreads to the toes before it freezes piles metal in the middle, producing convexity from too little heat rather than too much filler. On open-root work, match root-pass current and travel speed to the qualified window, not the hotter settings you use for cap passes.

Quick Note: If the face looks hollow, resist the urge to simply crank amperage. Fix travel speed and arc length first; heat usually follows.

Contact-Tip Distance and Torch Angle

Contact-to-work distance (stick-out) controls how much current actually reaches the pool. Let it grow and deposition drops, the pool cools, and the profile wobbles. Keep it consistent and re-check the nozzle and tip for wear between passes; a worn bore changes the arc before any setting does. Torch angle matters too. Keep the wire angled to push metal into the toes rather than pile it in the center, and hold a steady work angle so both legs of the fillet get equal heat. For proper technique fundamentals, this welding practice guide is worth working through, and this piece on contact tip selection and maintenance covers the consumables side.

Fit-Up: Gaps, Root Face, and Tacking

Fit-up decides how much metal a joint can hold. Hold root gap to the qualified tolerance and treat any widening trend as a fit-up problem to correct before welding — not something to bridge with extra wire. Follow SSAB’s guidance and keep the root face around 1.5 to 2 mm so the root pass has something to key into. Prevent the gap from closing during tacking: uneven tack spacing, or tacks that pull the joint tight, leave you chasing a moving gap down the seam.

Purge and Technique on Open-Root Work

Confirming purge gas flow and hold time before the root pass starts removes the largest single cause of root concavity. Verify it every time, not just the first joint. Then control the weave: steady oscillation, consistent pauses at the toes so the pool wets out, and deliberate tip manipulation on the root to keep penetration even. For polarity choices that affect arc behavior, see this guide to TIG welding polarity.

A Quick Prevention Checklist

  • Set wire feed speed, current, and travel speed as one matched set.
  • Hold voltage in balance with amperage — no thin, spread arc.
  • Keep contact-to-work distance constant; inspect the tip.
  • Check root gap and root face (≈1.5–2 mm) before tacking.
  • Verify purge flow and hold time on every open root.
  • Use qualified root-pass settings, not cap-pass heat.
  • Pause at the toes; keep the weave steady.

The causes of shape imperfections are well documented, and nearly all of them are controllable before and during the pass. Nail these habits and the gauge will sit flat where it should.

Repairing a Concave Weld: When to Fix and How to Do It Right

Not every concave weld is a defect. When the measured throat thickness still satisfies the design and the governing code, and the face is smooth with no undercut or underfill beyond allowance, the weld can often be accepted as-is. The deciding question is simple: does the joint still carry the load the engineer credited it with? If yes, document it. If no, it must be repaired.

Acceptable vs. Must-Repair

Shape imperfections are judged by whether they reduce the effective throat, not by how they look. A shallow, uniform dip that leaves the required throat intact is a profile condition. A concavity that eats into the throat is a deficiency, because the joint is now weaker than the design assumed. External references such as TWI’s review of shape imperfections make the same distinction, and the code limits in AWS D1.1 and ISO 5817 draw the line between acceptable and rejectable. The exact numbers vary by application, but the principle is consistent.

Fillet Welds: Restore the Throat, Not Just the Face

A classic trap is treating a concave fillet as a cosmetic problem. If the measured fillet throat is less than 0.707 times the specified leg, the weld is deficient and must be built up with additional passes to restore the full throat dimension. Simply running a light cap over the dip to make it look fuller does not fix anything; the throat stays short and the strength stays compromised. The correct repair deposits real weld metal with the same process and parameters that produced the original pass. An undersized throat is a strength defect, not a cosmetic one.

Groove Welds and Underfill

Underfill on a groove weld is corrected by adding weld metal before the joint is accepted. The goal is to bring the face to flush or to positive reinforcement within the code’s limits — not to leave a hollow that reduces the cross-section. Blend the added metal smoothly into the toes so no new undercut or notch is created, and confirm the allowable reinforcement limits in the governing standard before you weld.

Quick Note: Grinding is the standard correction for excessive convexity, but it is not the fix for concavity. You cannot grind metal back into a weld — concavity needs added metal, not removed metal.

Do It By Procedure

Repairs must conform to the original qualified welding procedure and be reinspected against the same acceptance criteria. Dig out the area, clean it back to sound metal, remove slag and any contamination, and confirm preheat and interpass requirements before you restart. A repair that isn’t tied to a qualified WPS is just guesswork with a stinger. For consistent technique and parameter control, see our guides to building real welding skills, to 3/16 E7018 amperage settings, and to running a wire welder. On critical work, the same care with consumables and setup shown in contact tip maintenance pays off during repair.

Step-by-Step Repair Sequence

A disciplined repair follows the same order every time:

  1. Confirm the defect against the acceptance criteria and record the location.
  2. Clean the area and remove slag, spatter, and contamination.
  3. Deposit additional weld metal per the qualified WPS to restore the throat or face.
  4. Blend the toes and allow controlled cooling.
  5. Reinspect to the same standard and document the result.

Document and Reinspect

After welding, re-measure the throat or face, confirm the profile meets the drawing and code, and record the repair in the inspection report. Post-repair checks use the same gauge and the same criteria as the first inspection. For structural or pressure work, cite the governing standard — such as AWS D1.1 — and confirm the profile limits before signing off. Good records protect you if the joint is questioned later.

Note: The most common mistake is covering a problem rather than restoring the design throat. A pretty cap over a weak throat passes the eye test but fails the intent of the code.

Common Mistakes, Key Takeaways, and Final Thoughts

After all the gauges, schedules, and code clauses, the mistakes that trip people up are remarkably consistent.

Five mistakes that quietly cost throat

  • Trusting a smooth finish. A weld can look clean and still be undersized, especially on a fillet where the throat hides behind the legs. It looks good, so it must be good — until a gauge lands on it.
  • Judging a fillet by leg size alone. Legs can match the drawing while the throat still sags below the effective value you need. Profile, undercut, and undersizing are checked together — a weld that satisfies the convexity schedule but carries heavy undercut at the toes still fails visual inspection.
  • Grinding the hollow away. Grinding removes metal, and on a concave fillet that only deepens the throat deficit. The cure is deposition, not abrasion.
  • Ignoring root concavity. It hides on the far side of a pipe or open-root joint, so it is easy to overlook. Root concavity is almost always a purge, fit-up, or heat-input issue rather than a surface one.
  • Borrowing the wrong criteria. Pipeline acceptance under API 1104 is its own regime; a structural criterion from AWS D1.1 should never be carried into it. Even static joints must pass profile inspection, because gross convexity or concavity can signal trapped slag, poor fusion, or a process running cold.

Quick Note: Profile acceptance is a proxy for process control. The shape of the face tells the inspector how your parameters — current, travel speed, and arc length — were actually running.

When you find concavity, resist the urge to sand it flat or grind it back to a prettier profile. Neither adds the metal the joint is missing. Whether the governing standard is a structural one like AWS D1.1, a pipeline code like API 1104, or ISO 5817 for a European fabrication, the rule to fix in your head is simple: match the measurement to the code on the drawing, then decide pass or repair.

Key takeaways

  • Concavity is a throat problem, not just a looks problem.
  • A slightly concave fillet can be perfectly legal when the governing standard allows it and throat thickness is maintained.
  • Root concavity is usually a purge, fit-up, or heat issue — not something you repair at the surface.
  • Inspection only counts when it matches the standard on the drawing.

Note: A concave cap that meets code is not automatically a defect, and a fat, convex cap hiding lack of fusion is far more dangerous than a legal hollow. Judging by eye alone gets this backwards every time.

Final thoughts

Profile is the honest report card of your process. Chase a tidy face and the throat usually follows: steadier travel angle, consistent amperage, and disciplined fit-up. None of this is guesswork — the shape imperfections that concavity belongs to are well documented in reviews such as this TWI round-up, and the limits themselves live in published standards.

Sharpen those fundamentals in our guide to welding practice, and compare concavity with a neighbouring flaw in our breakdown of the overlap welding defect — see our guide to good and bad MIG welds for the visual tells. Concavity is not a character flaw — it is feedback. Read it, measure it, adjust, and you will weld flatter, stronger joints for the rest of your career. It is one of the most fixable imperfections in welding, and mastering it sharpens everything else you do.

Frequently Asked Questions

Is a concave weld always a defect?

No. A slightly concave fillet weld face is permitted by codes such as AWS D1.1 as long as the minimum weld size, considering both leg and throat, is achieved. It only becomes a defect when the concavity reduces the effective throat below the required value, or when it is underfill on a groove weld that must be flush or positively reinforced.

What causes concavity in a weld?

Face concavity is usually caused by high travel speed, voltage too high relative to amperage, or short fast spray-transfer passes. Root concavity, or suck-back, is caused by shrinkage of an unsupported molten pool, poor root gap or root face fit-up, and inadequate backing gas purge on open-root joints.

How do I measure concavity on a fillet weld?

Use a fillet weld gauge or a bridge cam gauge to measure the actual throat from the root to the deepest point of the concave face. Compare that measurement to 0.707 times the specified leg size; if the measured throat is short, the weld fails even if the leg size looks correct.

What is the difference between concavity and underfill?

Concavity describes a fillet weld face that dips below the theoretical plane joining the toes, and it can be acceptable if throat is maintained. Underfill specifically describes a groove weld face lying below the adjacent base metal surface, and AWS D1.1 does not permit underfill on full-penetration groove welds.

What causes root concavity (suck-back) and how is it prevented?

Root concavity happens when a molten root pool sags inward as it freezes, often because backing gas purge is inadequate, the root gap or land is too wide, or root-pass heat and travel speed are wrong. It is prevented mainly by verifying purge flow and hold time, holding fit-up to the qualified tolerance, and matching root-pass parameters to the procedure.

Can I just grind a concave weld to fix it?

No. Grinding is the correct repair for excessive convexity because it removes excess metal, but grinding a concave weld removes even more throat and makes the problem worse. A concave or undersized weld must be repaired by depositing additional weld metal to restore the required throat or reinforcement.

What do AWS D1.1 and ISO 5817 say about concavity limits?

AWS D1.1 permits concave fillet faces as long as the minimum weld size including throat is maintained, but does not permit underfill on groove welds. ISO 5817:2023 sets root concavity limits by quality level and thickness, such as h <= 0.2t with a 2 mm cap at moderate level D and h <= 0.05t with a 0.5 mm cap at stringent level B.

Does radiographic testing detect face concavity?

No. Face concavity is external surface geometry, so volumetric methods such as radiographic or ultrasonic testing have nothing to report on it. Visual inspection with gauges is the primary route for face concavity, while radiography is the primary route for root concavity where the root side cannot be seen.

Authoritative Sources & Further Reading

STAY CURIOUS. STAY INFORMED.

Explore the latest articles and insights on WeldGallery.