How Strong is a Weld? Unveiling the Strength Secrets

The customer wanted a 2,000-pound trailer hitch receiver welded to the tongue of a utility trailer, and on paper everything looked fine. The shop that built it was run by two guys who had been welding farm equipment for twenty years, and nobody doubted them. They tacked the 3/8-inch plate in place, dialed the machine to a setting that had worked on a hundred previous jobs, and ran what honestly looked like a decent bead — a nice stack of dimes, good color, no visible porosity, a little heavy on the reinforcement. When the hitch was loaded the first time with a skid steer on a sled, the weld held. On the second trip over a frost-heaved county road, the bead cracked along the toe on the driver’s side, the plate peeled open, and the tongue dropped. Nobody was hurt, but the customer’s machine took $4,000 in sheet metal damage and the shop ate the repair.

Here is what killed that weld, and it was not the welder’s hand. The plate was 3/8 inch thick and had been joined to the tube with a fillet that measured 5/32 of an inch on one leg and closer to 3/16 on the other — undersized for the load, and worse, fused only partially into the corner of the tube because the wire feed had been set for thicker material and the arc never dug into the root. The calculation everyone had in their heads was the old shop shorthand: a quarter-inch fillet carries about 4,800 pounds per inch, so six inches of weld means nearly 30,000 pounds of capacity, which is “way more than 2,000 pounds.” That shorthand is only half true. It assumes a full, clean, properly fused throat. What they actually had was closer to half a throat with a sharp notch at the toe where fatigue cracks love to start, and the cyclic loading from road vibration did the rest.

I have spent enough time cutting apart failed welds to know this story is not unusual. It is the default outcome whenever “strength” is treated as a property of the weld bead instead of a property of the whole joint — geometry, base metal, filler, penetration, profile, and loading history all at once. A weld is not a glue line with a number on it. It is a small casting fused into a larger structure, and its strength depends on which of those things you got right and which one you got wrong. The good news is that the whole subject is learnable, and most of it comes down to a handful of numbers you can hold in your head.

So let’s answer the question properly. This guide walks through what “strength” actually means when an engineer writes it on a drawing, where the numbers for common steels and fillers come from, how joint geometry turns a material property into a load capacity you can calculate, why the welder’s technique matters more than most people expect, how defects subtract from that capacity, what testing actually proves, and which codes govern the work. By the end you should be able to look at a fillet on your bench, measure it, and estimate how many pounds per inch it will carry — and, more usefully, know when the honest answer is “get it engineered.”

Key Takeaways

• Strength is not one number. Tensile strength, yield strength, impact toughness, and fatigue life are four different properties, and a weld can pass one while failing another.
• E70 means 70,000 psi minimum tensile — a guaranteed floor for the deposited metal, not a guarantee about your finished joint. That is why reading rod designations matters.
• Geometry usually decides the answer. Fillet weld load runs through the throat, which is 0.707 × leg size — not the leg you can see with a tape measure.
• Working rule of thumb: a full 1/4-inch fillet carries roughly 3,700 pounds per inch at allowable stress and about 7,400 pounds per inch at ultimate with an E70 filler.
• Defects subtract directly. Lack of fusion cuts effective throat proportionally, undercut notches the toe, and porosity or inclusions remove area while destroying fatigue life.
• Process matters less than execution. Stick, MIG, and flux core all produce code-quality welds with matched fillers; what differs is penetration, deposition rate, and how forgiving the process is of technique.
• Critical work gets engineered. Anything overhead, pressurized, cyclically loaded, or human-supporting belongs under a written procedure and a code such as AWS D1.1.

What “Strength” Actually Means for a Weld

When someone asks how strong a weld is, they almost always mean tensile strength — the peak load the metal can carry before it pulls apart. It is a real property and the one printed on electrode classifications and material certificates, but it is the first of four that matter and often the least relevant in the field. Structures rarely fail at the exact maximum load the material could theoretically survive; they yield and deform, crack under repeated loading far below yield, or shatter under a sudden impact at low temperature. Each failure mode maps to a different measurable property, and a welding procedure that is excellent for one can be mediocre for another.

Tensile strength is the maximum engineering stress a material carries before fracture, expressed in ksi or MPa. In a joint test the interesting question is not just whether the weld metal meets its number, but where the specimen breaks. If a transverse tensile coupon of an E70-filled joint in A36 plate fails in the base metal away from the weld, that is the ideal result — it proves the weld and its heat-affected zone outran the parent material. If it breaks in the weld metal, the filler was undermatched or the deposit was contaminated. Codes generally require the assembly to develop the base metal’s minimum specified tensile strength, so an A36 joint must average at least 58 ksi across the section rather than the bead itself hitting some magic figure.

Yield strength is where the metal stops springing back and starts permanently deforming, and for structural design it is the more useful number. Allowable-stress design is built on it: take the yield point, divide by a safety factor of roughly 1.5 to 1.67 under AISC, and keep working stress below the result. A36 yields at 36 ksi no matter how good your weld is, so a 70 ksi filler in an A36 joint does not make the joint 70 ksi strong — the base metal governs, and the assembly starts deforming at whatever load pushes A36 to 36 ksi across its net section. That is the most common misunderstanding in shop conversations about weld strength, and it is why overmatching the filler so often buys you nothing.

Impact toughness is the capacity to absorb energy during a sudden, fast fracture, measured with a Charpy V-notch test and reported in foot-pounds or joules. Welding changes the microstructure near the fusion line, and coarse-grained heat-affected zones can turn brittle: a joint that stretches ductilely at room temperature can shatter at 0°F. That is why low-temperature and impact-loaded service may require impact-tested consumables, low-hydrogen electrodes kept dry in a holding oven (see rod oven builds and rod redrying guidance), controlled heat input, and sometimes preheat.

Fatigue strength is the fourth property, and for most real-world weld failures it is the one that actually governs. Fatigue is damage accumulation from repeated loading: every cycle grows a microscopic flaw at a stress concentration a little further, until a joint loaded well below yield eventually cracks. How long that takes depends far more on joint geometry and weld-toe profile than on filler classification — a smooth, ground, blended toe can survive tens of millions of cycles at a stress range where an as-welded toe with sharp undercut fails in a fraction of that. Nothing about the 70,000 psi filler number changes this. Take one idea from this section: static strength and fatigue strength are different conversations, and winning one says almost nothing about the other.

How Base Metal Strength Is Specified

Structural steel is sold against a specification, and that specification defines the numbers you are allowed to use in a calculation. The most common mild structural steel in North America is ASTM A36, with a tensile range of 58–80 ksi and a minimum yield of 36 ksi. Notice it is a range, not a single value: a heat of A36 plate might test at 62 ksi or 78 ksi and both are compliant, which is exactly why design codes use the guaranteed minimum of 58 ksi rather than whatever the mill certificate happened to show. Anyone who designs to the upper end of that range is building on a number the steel is not obligated to deliver.

Higher-strength low-alloy steels changed the landscape in the 1960s and never gave it back. ASTM A572 Grade 50, now the default for most structural shapes, delivers 50 ksi yield and 65 ksi minimum tensile in a lighter section than A36 would require, and it welds with consumables that match it. Weathering steels like A588 and A242 are chosen for corrosion resistance as much as strength. Pipe grades under ASTM A106 (60 ksi yield for Grade B) and pressure vessel plate under SA-516 Grade 70 (70 ksi tensile) each come with their own welding requirements, and pressure vessel work sits under the ASME Boiler and Pressure Vessel Code rather than AWS D1.1, with tougher rules about procedure qualification and impact testing.

The practical lesson is that you cannot select a filler or size a weld until you know what the base metal is guaranteed to deliver, and that information is on the drawing or the material certificate — not in the metal’s appearance. Mild steel that “looks like A36” might be A572, might be a foreign equivalent with different sulfur and phosphorus, or might be pressure vessel plate that is not legal in a structural application without the right paperwork. Reading drawings is a real skill with real consequences: see reading welding blueprints for how joint symbols communicate all of this, and why some joints are grooved for why the symbol changes the required penetration.

Filler Metal Matching: E70 Means 70,000 psi

Every stick electrode sold in the United States carries an AWS classification code, and the first two digits are the strength. The first digit is minimum tensile strength in tens of ksi, the second indicates usable positions, the last digits identify coating and current — so a 6013 reads 60,000 psi minimum tensile, all positions, rutile coating for AC or DC, while a 7018 reads 70,000 psi, all positions, low-hydrogen iron powder for DC or AC. That first digit is a guaranteed floor, not an average: a good E7018 deposit often tests at 75–80 ksi, but you may only count on 70. Wire classifications follow the same logic — ER70S-6 is a 70 ksi solid MIG wire and E71T-1 is a 70 ksi all-position tubular wire. For the full walkthrough see welding rod numbers explained.

Overmatching means using filler stronger than the base metal, and in structural work it is the default and usually the right call. Welding A36 (58 ksi minimum tensile) with an E70 electrode overmatches by roughly 20 percent, buying margin against the small strength loss that dilution, contamination, and thermal cycling cause. It also shifts the plastic hinge away from the weld and into the base metal, where you want deformation during an overload — ductile base metal stretching is far more forgiving than weld metal cracking. The costs are real: the weld becomes the stiffest, strongest region of the joint, concentrating strain in adjacent base metal and in some fatigue-critical details shortening life, while a strong rapidly cooled deposit on a restrained joint raises the risk of hydrogen-induced cracking.

Undermatching means deliberately using filler weaker than the base metal, and it sounds wrong until you see why anyone does it. In high-strength steels — A572 Grade 65, quenched-and-tempered plate, anything in the 80–100 ksi class — a matching filler would be expensive and harder to weld without cracking. An undermatched, more ductile deposit acts as a fuse: it deforms first, redistributes stress, and keeps the weld from being the brittle link, which is common in pressure vessel nozzles and creep service. The trade-off must be checked — the joint’s capacity is now governed by the weaker filler, so the throat has to be sized up to compensate. It is not something you pick casually; it requires a designer who has run the numbers.

Where people get into trouble is assuming the classification guarantees the finished joint. It guarantees the properties of a correctly deposited, correctly cooled specimen — nothing more. Dilution shifts chemistry, excessive current burns off alloying elements, and a weld that is half lack-of-fusion carries far less than 70,000 psi worth of anything. This is why procedure qualification exists, and why knowing how to choose a welding rod includes understanding what the rod will and will not promise once it is in your joint.

Weld Joint Types and Geometry: Where the Number Comes From

The reason a material property becomes a load-carrying capacity is geometry, and this is the part most casual welders never see. A welded joint transfers load through a specific cross-section of weld metal, and that cross-section — not the bead’s width, not its appearance, not its classification — is what you size. The four basic joint types are butt (square or grooved), corner, lap, and tee. Butt and corner joints usually take groove welds; lap and tee joints almost always use fillets, and fillets are where the throat-thickness rule takes over.

For a fillet weld, load is assumed to act on the throat — the shortest distance from the root of the joint to the face of the weld. On a standard equal-leg fillet, the throat equals the leg size multiplied by the sine of 45 degrees, which is 0.707. So a fillet that appears to be 1/4 inch on the outside has a structural throat of 0.707 × 0.25 = 0.177 inches, and that 0.177 is what carries the load. This is the single most important arithmetic fact in fillet weld design, and it is also the most frequently ignored. Put it another way: an equal-leg fillet is only about 71 percent as strong per unit length as a full-throat square edge of the same apparent size, which is why every sizing rule of thumb works backward from the throat.

Now put a real number on it. The allowable stress on a fillet weld per AWS D1.1 and AISC practice is 0.30 × F_EXX — thirty percent of the filler’s nominal tensile strength — acting in the plane of the throat. For an E70 filler that is 0.30 × 70,000 = 21,000 psi. Multiply by the throat area per inch of weld length: a 1/4-inch fillet has a 0.177-inch throat per inch of length, so capacity is 21,000 × 0.177 ≈ 3,710 pounds per inch of weld at allowable stress. The pattern holds across sizes — about 1,850 pounds per inch for an 1/8-inch fillet and about 7,430 for a 1/2-inch fillet. You will also see roughly 4,800 pounds per inch quoted for a quarter-inch fillet, which comes from working in ultimate rather than allowable terms; the two conventions differ by about a factor of two, so always state which one you are using, and for critical work run it per code or hire it out.

Sizing rules of thumb are useful because they compress this into a working habit. The classic shop guidance is that a fillet’s leg should be about three-quarters of the thickness of the thinner part joined — a 1/4-inch fillet on 3/8-inch plate — capped so the leg never exceeds the thickness of the thinner member. AWS D1.1 additionally requires groove weld effective throat to equal the thickness of the thinner joined part unless the design calls for otherwise, and sets minimum fillet sizes that scale with the thickness of the thicker part. Excessive fillet size is not free: a huge bead on thin plate pulls in more heat, distorts the part, and pushes the fusion boundary into a coarse-grained, crack-sensitive zone. For the hands-on side, techniques for strong joints and dialing in your settings cover the details, while full penetration versus fillet welds shows where each family belongs.

Groove Welds, Penetration, and Weld Volume

Fillets are simple arithmetic; groove welds are where “is it full penetration” becomes the entire conversation. A groove weld fills a prepared space between beveled, chamfered, or squared edges. The standard preparations are square groove (no bevel, used on thin material), single-V, double-V or double-bevel, and U-grooves for very thick plate where a V would demand an unreasonable volume of metal, plus J-grooves common in pipe and heavy pressure work. Each choice trades preparation effort against weld volume: a double-V roughly halves the deposited metal for a given thickness compared with a single-V, but requires welding from both sides.

Full penetration means the weld metal runs through the entire thickness from root to face, so the whole cross-section of the base metal participates in carrying load. This is what you need when the joint is the critical element — cranes, trailer hitches like the one in our opening story, moment connections, anything in bending. AWS D1.1 credits a groove weld with full effective throat only when it is made in a groove reaching the entire thickness (or properly backed or rooted), it meets visual acceptance, and it passes the required nondestructive examination. In practice the back side either shows a visible root bead — the tell-tale of true penetration — or you use a backing strip, a consumable insert, or a root-pass technique that guarantees fusion.

Partial penetration groove welds are legal and common, but they are explicitly designed with a reduced effective throat, typically the groove depth plus any specified penetration, which can be as little as half the plate thickness. D1.1 permits them in statically loaded structures with a defined minimum throat and restricts them in cyclically loaded applications unless designed for it. This is a favorite source of quiet structural weakness: a bevel that only went halfway, an operator who “captured the root” by eye, and a drawing that assumed full penetration. The numbers are unforgiving — a full-penetration butt weld on 1/2-inch plate has an effective throat of 0.500 inches, while a partial weld reaching only 0.300 inches carries 40 percent less load no matter how the bead looks on top. Weld volume also drives cost, distortion, and heat input, which is why engineers reach for partial penetration and intermittent welds wherever full development of the plate is not actually required.

Penetration is ultimately a procedure question, not a design assumption. The blueprint symbols tell you what was intended — see reading welding blueprints — but only a procedure qualification record proves that your machine, position, and technique actually deliver it, which is exactly the gap a full penetration versus fillet comparison is meant to close.

Stress Concentrations and Weld Profiles

If tensile strength tells you how much load the metal can take in a laboratory coupon, stress concentration explains why a real weld fails at a fraction of that load. A stress concentration is any geometric discontinuity that forces internal stress lines to crowd together: a notch, a sharp toe, an undercut, an abrupt change in thickness, a spatter globule in the wrong place. In a welded joint the two biggest offenders are the weld toe — where the face of the weld blends into the base metal — and the weld root, where penetration stops. Both act like pre-started cracks, and both are why fatigue life is governed by geometry rather than filler classification.

Quantify it roughly: a smooth, well-blended weld toe has a stress concentration factor of perhaps 1.3 to 1.5, meaning local stress is 30–50 percent higher than the nominal calculation suggests. Add a sharp toe with undercut — a small groove melted into the base metal at the edge of the bead — and the factor can jump above 2.0, which means stress at the toe could exceed the material’s yield point while average stress in the member is still at half of allowable. The metal at that notch yields, reverses on unloading, and starts a fatigue crack that grows with each cycle. Undercut is not a cosmetic defect; it is a crack starter that also removes base metal from the load path, doing damage twice.

Weld profile is therefore a fatigue design variable, not an appearance preference. Codes distinguish as-welded toes (sharp, abrupt, worst fatigue category) from ground or peened toes (blended smooth, one or two categories better) and dressed profiles. AWS D1.1’s fatigue provisions and the AISC category tables assign different allowable stress ranges to these profiles: a transverse fillet loaded across the weld might be one category as welded and a better category if ground and blended, which can double or triple the allowable stress range for a given design life. Improving profile is one of the cheapest fatigue-life investments available, and creating a bad one — undercut, overlap, a sharp reinforcement, a crater crack at the stop — is one of the cheapest ways to destroy one.

The same physics applies to weld reinforcement: a bead standing proud above the plate creates a step that load must flow around, and the sharper the step the worse the concentration. This is why codes cap reinforcement near 1/16 inch for groove welds and require it blended flush in fatigue-critical or pressure applications, and why an over-built weld can be structurally worse than a slightly smaller, correctly profiled one. Process mistakes that produce bad profiles — dragging the arc, wrong travel speed, wrong stickout, reversed polarity — are exactly what this list of costly flux-core mistakes warns about, loose spatter adds its own notches if it is not cleaned off, and travel direction changes both penetration and toe shape, which is why push versus pull technique and spatter control belong in a strength conversation at all.

The Welder’s Actual Contribution: Technique, Settings, and Cleanliness

Everything discussed so far — grades, filler classifications, throat geometry — is decided on paper before the arc is struck. The welder’s job is to make the real deposit match the assumption, and the gap between the two is where welds fail. Cleanliness comes first because it is the cheapest thing to get right and the most commonly skipped: mill scale, rust, oil, paint, galvanized coatings, and moisture all generate gas or slag in the arc, and gas that cannot escape the solidifying puddle becomes porosity. Rust and moisture also introduce hydrogen, which produces no immediate defect at all — it diffuses into the heat-affected zone and can cause delayed cracking hours after a weld looked perfect. Grind to bright metal, degrease, and burn off galvanizing deliberately rather than accidentally, since zinc makes both a porous weld and a hazardous fume (see welding galvanized steel), and handle the fumes as the health issue they are: OSHA and NIOSH set the ventilation expectations, while welding fume toxicity covers what those fumes contain.

Settings are the second lever, controlling penetration, deposition rate, and heat input all at once. Too little amperage or voltage for the material thickness means the arc never melts the root, and you get a bead sitting on the plate with a cold, unfused interface underneath — lack of fusion, the most dangerous common defect because it looks fine from outside and subtracts directly from effective throat. Too much input burns through thin material, over-dilutes the deposit, coarse-grains the heat-affected zone, and can burn off alloying elements so the filler no longer delivers its classified strength. Every process has published starting windows for exactly this reason — flux core amperage charts and wire speed and voltage charts — because guessing is how people end up with cold lap, and getting polarity right changes arc stability and penetration rather than just appearance.

Direction and technique shape the profile. Pushing gives shallower penetration and a smoother bead; dragging digs deeper with more reinforcement, which is why flux core is almost always dragged while MIG on thin material is often pushed. Vertical-up gives the deepest weld on thick plate because the previous pass supports the puddle, but it demands correct travel speed and arc length — vertical-up settings and horizontal settings differ for metallurgical reasons, not tradition. Work angle, arc length, and stickout all move the fusion boundary, and a long stickout on MIG drops amperage at the arc, quietly turning a properly fused weld into a partially fused one. Stops and restarts matter too: lifting straight off leaves a shrinkage crater that codes reject outright, and uncleaned interpass slag becomes slag inclusions, just another form of reduced cross-section. That discipline separates a weld that measures out at its design throat from one that merely looks like it does, which is the practical answer to avoidable porosity.

Testing Methods: What Actually Proves a Weld

There is no single test that tells you “this weld is strong,” and the honest answer to “how do I know?” is layered: visual inspection finds the obvious, nondestructive examination finds the hidden, and mechanical testing proves the procedure. Nondestructive methods inspect the actual weld in the structure without damaging it; destructive testing sacrifices a sample to measure properties directly. Most code work uses both, and each method has a blind spot that matters.

Visual inspection (VT) is the foundation and catches more real defects than all other methods combined, because most rejects are visible to someone who knows what acceptance criteria apply: wrong size, undercut, overlap, surface porosity, misalignment, bad profile, crater cracks. It needs only good lighting, a mirror, a weld gauge, and an inspector. Liquid penetrant (PT) wicks dye into surface-breaking discontinuities and draws it back out with a developer, revealing cracks and porosity open to the surface — cheap and portable, but blind to anything below the surface. Magnetic particle (MT) magnetizes the part so iron particles pile up at flux leakage from a crack, finding surface and near-surface flaws roughly the top eighth inch, but only in ferromagnetic materials, so it is useless on stainless, aluminum, or cast iron.

Ultrasonic testing (UT) sends high-frequency sound into the weld and reads the reflections: a lack of fusion, slag inclusion, or crack appears as an echo at a known depth. It covers the full thickness of heavy sections, needs no radiation, and is extremely sensitive to planar defects, but it is heavily dependent on technician skill and calibration. Radiographic testing (RT) — X-ray or gamma — produces film showing the interior as a shadow, so porosity and inclusions show up beautifully as dark spots and the record is permanent. Its limits are the radiation hazard and licensed operators, difficulty detecting planar flaws oriented parallel to the beam, and the cost of field radiography. For broader references on inspection practice, TWI publishes excellent technical material on NDE methods and their limitations.

Destructive tests are how a procedure earns its credentials. A transverse tensile coupon is pulled until it breaks; AWS D1.1 requires the assembly to develop at least the minimum specified tensile strength of the base metal, and the break location shows which element governs. Guided bend tests — root, face, and side bends — stretch a strip 180 degrees around a mandrel, exposing lack of fusion or porosity the tensile test could hide. A Charpy V-notch test measures toughness for low-temperature service, and macro-etching cuts and etches a cross-section so you can see actual penetration, throat, and profile with your own eyes.

Codes and Standards: AWS D1.1, WPS, and PQR Without the Legalese

Codes convert engineering knowledge into minimum requirements you can inspect against, and in North American structural steel the dominant document is AWS D1.1, the Structural Welding Code for Steel. D1.1 is not law by itself — it becomes binding when a jurisdiction, an owner, or a contract adopts it by reference, which is how most building and bridge codes work. It covers preheat, filler selection, joint design, welding positions, quality requirements, inspection, and the mechanical tests a procedure must pass. Related AWS documents carve out specific territory: D1.3 for sheet steel, D1.5 for bridges with tougher fracture-critical provisions, D1.6 for stainless, D1.2 for aluminum, D17.1 for aerospace. Pressure work lives under ASME BPVC Section IX and pipeline transmission under API 1104, each with its own philosophy about what qualifies.

Three acronyms carry most of the confusion, and they nest cleanly. The WPS (Welding Procedure Specification) is the recipe: process, filler class, polarity, current and voltage ranges, travel speed, shielding gas and flow, joint geometry, position, preheat and interpass temperature, and required technique. The PQR (Procedure Qualification Record) is the proof the recipe works — a test plate welded to those parameters and subjected to tensile, bend, and often impact testing, with results documented. The WPQ (Welder Performance Qualification) is the welder’s ticket: it qualifies the individual to weld using a specific qualified procedure in specific positions, demonstrated by welding a test assembly and passing visual and bend examination. A welder qualified on one WPS is not automatically qualified on every procedure — the ranges of process, position, thickness, and diameter are defined, and stepping outside them requires requalification.

Prequalification is the pragmatic middle ground that keeps routine work moving. AWS D1.1 Clause 4 lets you use certain common joints and processes — the usual butt, corner, tee, and lap configurations in flat and horizontal positions with conventional electrodes — without performing a full procedure qualification, provided you stay inside the specified ranges for preheat, position, and material. It is essentially the code saying these combinations are well understood enough that new test data is not demanded each time. Move to vertical or overhead on thick material, change processes, alter preheat conditions, or work on fracture-critical members, and formal qualification returns.

The honest trade-off: full compliance costs money and time, and on a backyard gate none of it applies. People get hurt in the middle zone — a shop lifting beam, a structural modification, a highway trailer, a railing someone will lean on. Those are exactly the jobs where “I’ve been doing this twenty years” stops being sufficient and a documented procedure becomes the difference between an insurance claim being paid and denied. Both Miller’s resource library and Lincoln Electric’s technical library publish procedure notes that map onto D1.1 requirements, and for the practical groundwork see solid fundamentals, deliberate practice at home, and non-negotiable safety discipline.

Flux Core vs MIG vs Stick: Strength on the Same Joint

This is one of the most argued questions in every shop and one of the easiest to answer honestly: if all three processes are correctly applied with a matched 70 ksi filler, fully fused and properly sized, the finished joints are essentially equivalent in static strength. E71T-1 flux core, ER70S-6 MIG, and E7018 stick all guarantee a 70,000 psi minimum deposit, and AWS D1.1 treats them identically for allowable-stress calculations. What differs is the difficulty of achieving that number consistently — the penetration profile, deposition rate, heat input, and how each process behaves out of position, in the wind, on dirty metal, and on a Wednesday afternoon when the operator is tired. That distinction is what the flux core versus MIG comparison and the broader MIG vs TIG vs stick breakdown are really about.

Solid-wire MIG with 75/25 argon-CO2 gives clean, predictable, low-spatter welds with excellent control on thin material and easy out-of-position capability at moderate deposition rates; its weakness is that the shielding gas is a blanket and a breeze blows it away. Flux-cored arc welding splits into two things people lump together: self-shielded FCAW (the gasless wire) puts its shielding in the flux and handles wind and dirt better, while dual-shield FCAW adds external gas for high deposition and deep penetration. Gasless flux core runs on electrode-positive polarity, drags, and produces more spatter and slag, and it tends to lay good face reinforcement with shallower root fusion than a properly set MIG weld on clean plate. Stick remains the field standard because it is indifferent to wind, works on rusty plate with the right electrode, gives full penetration with 7018 or 6010 root passes, and needs no bottle or liner — at the cost of lower deposition efficiency and constant rod changes.

Where process choice genuinely changes strength is in penetration achievable per pass and the heat-affected zone it creates. A deeply penetrating process on a poorly prepared joint can compensate for a missing bevel; a shallow one cannot, so the same “weld” made by two processes can have completely different effective throats on a joint that was never grooved. That is the practical answer to whether arc welding beats MIG — stick is not inherently stronger, but the electrode choices available in stick cover a wider range of joint conditions. The same logic runs through TIG versus MIG and choosing between stick and MIG. For process background, the flux core versus MIG showdown goes deeper, as does stick versus MIG head to head, while Britannica’s welding overview covers how the technology developed.

Common Myths: “The Weld Is Stronger Than the Base Metal”

The most repeated claim in welding is also the most conditional. An E70 deposit at 70,000 psi minimum tensile welded into A36 plate at 58,000 psi minimum really is about 20 percent stronger per square inch of metal, so in that narrow, literal sense the claim is true and it is why overmatching is the structural default. But the joint as a whole is not stronger than the base metal, because capacity is the smaller of throat area times allowable stress versus plate area times allowable stress — and an undersized fillet on a wide plate is almost always the weak link. The myth gets worse when it becomes reassurance: “the weld is stronger than the metal” often means “you cannot overload this weld,” which is false in the presence of any defect, notch, or cyclic load. A weld with 30 percent lack of fusion is not stronger than anything.

Second myth: more weld is always safer. Oversizing a fillet beyond what the joint requires costs heat, causes distortion, enlarges the heat-affected zone, and can push the fusion line into a coarse-grained, crack-prone region — while doing nothing to raise the joint’s rated capacity, which was set by the throat the design called for. Third: a bead that looks like stacked dimes is a strong bead. Appearance correlates with consistency, and consistency with settings and technique, but a beautiful weave laid over a cold root is a perfectly presented failure waiting for load. Fourth: flux core is weaker than MIG, or stick weaker than both — matched fillers land in the same place, and stick versus MIG often shows stick winning on penetration in dirty conditions where a MIG weld would porosity up. Fifth: the strongest filler available makes the strongest joint — an 80 or 100 ksi filler in A36 buys nothing static, adds cost, and can make cracking more likely.

For the record, strength is not always the governing requirement. E7018’s legitimate applications, E6013’s legitimate applications, and the 6011 versus 6013 comparison each describe cases where the nominally “weaker” rod is the correct engineering choice because toughness, penetration, cleanup, or position was the real requirement — and that is the mark of someone who actually understands weld strength rather than reciting it.

Practical Strength Examples: The Numbers You Can Use

Let’s build a couple of worked examples so the arithmetic stops being abstract. Example one: a bracket welded to a column with two 4-inch fillet welds in a lap configuration carrying a 6,000-pound vertical load shared equally. Each weld runs 4 inches, so total length is 8 inches. With a quarter-inch fillet in E70 filler, capacity is roughly 3,710 pounds per inch × 8 = 29,680 pounds, nearly five times the applied load. Drop the fillet to 3/16 inch and you still have 2,790 × 8 = 22,320 pounds; at 1/8 inch, 1,850 × 8 = 14,800 pounds, still more than double. This is why weld sizing feels generous on light structures — and why the caveat matters: that math assumed the load is truly shared, the plate does not tear out at the toe, the weld is fully fused along its length, and there is no fatigue or impact in the loading.

Example two: a lifting lug with a full-penetration groove weld, 3/8-inch plate, E70 filler, carrying 10,000 pounds in tension. The effective throat equals the plate thickness at 0.375 inches, so capacity is about 21,000 psi × 0.375 ≈ 7,875 pounds per inch of width at allowable stress. A lug 2 inches wide develops roughly 15,750 pounds — comfortably above 10,000. But if that weld were partial penetration reaching only 0.200 inches, capacity drops to 4,200 pounds per inch and the same 2-inch lug yields 8,400 pounds, so it fails. Same bead, same filler, same welder, 47 percent less capacity. Penetration is not an academic variable; it is half the answer to how much load a weld carries. Working with stick specifically? Rod size selection for weld strength and choosing rod diameter show how consumable size interacts with the throat you deposit, while rod types and their uses maps the common options to applications.

Sizing Fillets for Load: A Working Method

Sizing a fillet weld for a known load is a four-step calculation you can do on the back of a napkin, and doing it deliberately is what separates fabrication from guessing. Step one: determine the actual load on the weld group and how it distributes — tension, shear, or bending, and whether the geometry makes it share equally, since eccentric loads put most of the force on the ends. Step two: choose the stress basis — for an E70 fillet under AWS D1.1 and AISC practice, roughly 21,000 psi on the throat at allowable level, or about 42,000 psi at ultimate if you are checking failure rather than serviceability. Step three: divide load by allowable stress for the required throat area, divide by 0.707 to convert throat to leg, then divide by available weld length. Step four: round up to a practical size and check it against the three-quarters-of-thickness heuristic.

Worked: a 5,000-pound shear load carried by 10 inches of weld with E70 filler. Required throat = 5,000 ÷ 21,000 = 0.238 square inches total, or 0.0238 square inches per inch of length; required leg = 0.0238 ÷ 0.707 = 0.034 inches. That absurdly small answer tells you the real design driver is not throat shear at all — it is plate tear-out, buckling, connection eccentricity, fatigue, or the code minimums for production welds, since D1.1 sets minimum fillet sizes by the thickness of the thicker part, typically 1/8 inch through 1/4-inch material and 3/16 inch for 1/4- to 1/2-inch. Outcomes like this are common and useful, because they show when the arithmetic says the weld was never going to be the problem. At the other end, high loads or short weld lengths push the required size up, and the better fix is usually more weld length or a groove weld rather than brute fillet size: doubling the leg doubles the throat, but the deposited volume grows fourfold, which is why heavy connections favor full penetration over sheer bead size.

Frequently Asked Questions

How much load can a 1/4 inch fillet weld carry per inch?

Roughly 3,700 pounds per inch at allowable stress with an E70 filler — a 0.177-inch throat (0.707 × 0.25) at 21,000 psi — and about 7,400 pounds per inch at ultimate. You will also see 4,800 pounds per inch quoted for a quarter-inch fillet, which comes from an ultimate-stress convention, so the two figures differ by about a factor of two. Always ask which basis a number uses, and remember both assume a sound, fully fused weld shared evenly along its length.

Is a weld always stronger than the base metal?

No. Overmatched filler such as E70 in A36 plate is indeed stronger per square inch of deposited metal, but the assembled joint is governed by geometry, fusion quality, and loading. A full-penetration groove usually fails in the plate, a small fillet fails in the fillet, and a weld with lack of fusion or sharp undercut can fail below either. The claim describes a correctly made, correctly sized joint in a static test, not any bead you happen to lay down.

Is 7018 actually stronger than 6013, and does it matter for my project?

E7018 guarantees 70,000 psi against 60,000 psi for E6013 — about 17 percent more — plus a low-hydrogen coating that delivers better toughness and ductility. For structural, cyclic, low-temperature, or highly restrained joints that difference matters, which is why codes default to low-hydrogen electrodes. For a light, static, non-critical repair on mild steel it rarely governs, and 6013’s forgiveness usually wins. See 7014 versus 7018 and picking a mild steel rod.

Is flux core welding as strong as MIG?

Yes, when both use a matched 70 ksi filler with correct settings and full fusion — the finished joint is mechanically equivalent, and gasless flux core has the advantage in wind and on imperfectly clean metal. The differences are practical rather than fundamental: flux core drags and leaves more slag, MIG needs its gas blanket but gives a cleaner bead. What actually decides strength is whether fusion was achieved, which is why getting flux core settings right matters more than the process name.

Do I always need a full penetration weld?

No, but you need it wherever the joint must develop the full strength of the connected parts — cranes, lifting lugs, trailer hitches, moment connections, and most cyclically loaded details. Partial penetration is permitted in many static applications when the effective throat is explicitly designed, and intermittent fillets are normal for secondary members. The danger is assuming penetration you did not achieve: a drawing calling for full penetration with an unfused root has silently become a partial penetration joint that nobody calculated.

What size fillet weld should I use on quarter-inch plate?

The usual rule of thumb is a leg of about three-quarters of the thickness of the thinner part — roughly 3/16 inch on 1/4-inch plate — capped so the leg never exceeds the thickness of that thinner part. AWS D1.1 also imposes production minimums that scale with the thickness of the thicker joined part, commonly 1/8 inch through 1/4-inch material and 3/16 inch above it, so the code minimum often governs rather than the strength calculation.

How can I test a weld’s strength at home?

You cannot measure tensile strength without a machine, but you can verify geometry and fusion cheaply: gauge the actual throat with a weld gauge, bend a sample coupon in a vise and inspect the fracture face for penetration and porosity, cut and etch a cross-section to see the fusion line, or load-test a duplicate to destruction if the consequences are contained. For the visual side, these common defects make a good checklist.

How much undercut is acceptable?

AWS D1.1 generally limits undercut to about 1/32 inch (0.8 mm) in depth for members not subject to cyclic loading, with tighter requirements as the application gets more demanding, and any undercut in a fatigue-critical detail is disproportionately damaging because it notches the exact spot cracks start. Practically, if you can catch a fingernail in it, treat it as a defect to grind out rather than a number to measure down to the limit.

When should I hire an engineer instead of calculating it myself?

Whenever failure could cause serious injury and the detail is not covered by a routine prequalified procedure: anything that lifts or suspends a load overhead, anything pressurized, anything people lean on or walk under, anything welded to a vehicle or trailer used on public roads, and anything subject to vibration or repeated loading. Also hire one when the base metal is unknown or high-strength, when you are modifying a structural member, or when an insurer has asked for documentation.

Why does a weld fail in fatigue when the load was never close to its rating?

Because fatigue is not about peak load. It is about repeated cycling well below yield, with cracks initiating at notches — sharp toes, undercut, spatter, abrupt terminations — and growing a little with every cycle. A weld rated for thousands of pounds statically can crack in the hundreds of pounds range if the geometry is poor and the loading is cyclic, which is why a joint that survives one overload can fail after months of vibration. The countermeasures are geometric: blend the toes and eliminate undercut.

Conclusion: Back to the Trailer Tongue

The hitch that peeled off that winter failed for reasons that were all knowable in advance. The fillet was undersized for the load and for the calculation being used to justify it. The root had not fused, so the effective throat was smaller than the measured leg suggested. The toe was sharp, so the vibration from every mile of rough road concentrated there instead of spreading through the joint. Three separate failures — geometry, fusion, and profile — and each one alone would have been enough to explain the crack. None of them were visible in the photograph everyone looked at when they said the weld looked good.

That is the real answer to how strong a weld is. It is as strong as the throat you actually created, fused to the base metal you actually cleaned, profiled to the standard the loading actually demands, and verified by the inspection the consequence of failure actually warrants. The filler classification gives you a floor for the metal; the joint geometry gives you a number you can calculate; the welding technique determines whether reality matches the calculation; and testing determines whether you were right. Miss any one of those links and the number in your head is fiction, no matter how many years of experience produced it.

If you take nothing else from this guide, take the habit of asking four questions before you strike an arc on anything that matters: what is the base metal’s guaranteed strength, what effective throat does this joint need, what will make this deposit actually achieve it, and how will anyone know afterward that it did. Answer those honestly and you will build joints that hold up — not because someone told you the weld was stronger than the metal, but because you can show your work. And for anything that lifts a load over someone’s head, run the numbers through a professional working to a current code, because the margin you are trusting was never meant to be estimated.