Welded Joints: How Engineers Glue Metal With Metal

Stand under a steel bridge, look up at a ship's hull, or open the bonnet of a car, and you are surrounded by joints that were once two separate pieces of metal and are now permanently one. No bolts, no rivets, no adhesive — just a line of slightly rippled metal where the parts meet. That line is a weld, and it carries some of the heaviest and most safety-critical loads engineers ever design for.

It is tempting to think of welding as gluing metal with metal, and as a first picture that is fine. But a good adhesive sticks two surfaces together; a weld does something more drastic. It melts the edges of both parts and a filler in between, lets them flow into a single molten pool, and freezes that pool into one continuous piece of metal. Done well, you cannot point to a boundary — the joint is as much a part of the structure as the steel on either side. Done badly, it becomes the weakest link in the whole machine. This article is about why both outcomes are possible from the same process.

Fusion: why a weld is not glue

The dominant family of welding processes is fusion welding, and the name is the whole idea. A concentrated source of heat — an electric arc, a gas flame, a laser, or an electron beam — raises a small region of the joint above the melting point of the metal. For structural steel that means roughly 1500 °C; the arc itself sits far hotter, often above 5000 °C, so the heat is intense and local. The molten metal from both parts merges, usually together with molten filler metal fed in as a wire or rod, into a shared pool. When the heat source moves on, the pool cools and solidifies as a single crystal structure that bridges the two parts. That continuity is what makes a weld stronger than any glued joint: the load does not have to cross a stuck interface, because there is no interface left to cross.

The most common arc processes carry initials worth knowing. SMAW (shielded metal arc welding, or "stick") strikes an arc from a flux-coated electrode and is rugged enough for site work. GMAW (gas metal arc welding, or "MIG") feeds a continuous bare wire and shields the pool with an inert gas, giving fast, clean workshop welds. GTAW ("TIG") uses a non-consumable tungsten electrode and a separate filler rod for the most precise welds. They differ in speed, cost, and finesse, but the metallurgy underneath is the same: melt, mix, freeze.

The anatomy of a weld

To talk about welds precisely, engineers name their parts. A joint where two plates meet edge-to-edge and the gap is filled flush is a butt weld; a joint where the weld sits in the inside corner of two parts meeting at an angle (usually 90°) is a fillet weld. The fillet is the workhorse of fabrication because it needs no special edge preparation — you simply lay metal into the corner.

Figure 1: A butt weld fills the gap between plate edges; a fillet weld fills the corner. The throat is the shortest path through the weld and carries the load.

Several terms recur in every code and drawing. The weld toe is the line where the weld face meets the base metal; it is where the geometry changes sharply, and as we will see it is where fatigue cracks almost always start. The root is the deepest point of the weld, at the back of the joint, where fusion must be complete or the joint is hollow. The leg of a fillet is the length of its side along each plate. And the single most important dimension is the throat — the shortest distance from the root through the weld to its face. The throat is the plane the metal would tear along, so it is the throat, not the visible size, that sets a weld's strength.

Where the strength comes from: the throat carries the load

For an equal-leg fillet weld of leg length z, the throat is the line across the corner triangle. Geometry gives the throat thickness a directly:

a = z × sin 45° = 0.707 × z

That factor of 0.707 is one of the most-used numbers in steel design. It says a fillet weld is only about 70% as thick through its load path as its leg length suggests, which is exactly why engineers size welds by throat. The force a weld can carry is then the throat area multiplied by the metal's allowable shear stress — fillet welds are conventionally checked in shear across the throat regardless of how the load is actually applied, because that is the plane that fails.

Figure 2: The throat is the weld's true load-bearing thickness. For an equal-leg fillet, throat a equals 0.707 times the leg z.

Worked example 1: sizing a fillet weld for a bracket

Suppose a steel bracket carries a steady downward pull of F = 60 kN and is attached to a column by two fillet welds, one along each side of a 120 mm long plate. We will use a typical allowable weld shear stress of τ_allow = 95 N/mm² (a representative design value for a common mild-steel electrode). What leg size do we need?

First, the total weld length resisting the load is two runs of 120 mm, so L = 240 mm. The required throat area follows from force divided by allowable stress:

A_throat = F / τ_allow = 60000 N / 95 N/mm² = 632 mm²

Spread over the 240 mm length, the needed throat thickness is:

a = A_throat / L = 632 mm² / 240 mm = 2.63 mm

Convert throat to leg by dividing by 0.707:

z = a / 0.707 = 2.63 / 0.707 = 3.7 mm

A welder cannot dial in 3.7 mm, so we round up to the next standard size: a 5 mm fillet weld. That choice gives an actual throat of 3.54 mm and an actual capacity of about 0.707 × 5 × 240 × 95 = 80.6 kN against the 60 kN demand — a comfortable margin, which is what you want, because a weld is harder to inspect than the plate around it.

Worked example 2: heat input and why it matters

The second example is about heat, not force, because heat is what separates a sound weld from a cracked one. The heat input H — the energy poured into each millimetre of weld — is set by the arc voltage V, the current I, and how fast the welder travels, v:

H = (V × I) / v

Take a MIG weld run at V = 26 V, I = 250 A, travelling at v = 6 mm/s:

H = (26 × 250) / 6 = 6500 / 6 ≈ 1083 J/mm ≈ 1.08 kJ/mm

That number, about a kilojoule per millimetre, is a typical figure for structural steel, and engineers care about it for two opposite reasons. Too little heat and the parent metal chills the pool so fast that it never fully melts the joint, leaving lack of fusion — a weld that looks fine on the surface but is barely attached underneath. Too much heat and the region beside the weld stays hot for too long, growing coarse, brittle grains and distorting the part. Heat input is the dial the welder turns between those failures, and on critical work it is specified in a written procedure, not left to feel.

The heat-affected zone: the weld's quiet weak spot

Here is the fact that surprises people new to welding: the weld metal itself is often the strongest part of the joint. The trouble lives just beside it. The narrow band of base metal that got hot enough to change its structure but never melted is the heat-affected zone (HAZ). It was, in effect, heat-treated by accident — heated to high temperature and then quenched by the cold metal all around it.

Figure 3: A weld is really three materials side by side — the melted fusion zone, the heat-affected zone whose grain structure changed without melting, and the untouched base metal.

In many steels this accidental quench makes the HAZ harder and more brittle than the plate it came from, and in some it forms a structure prone to cracking. The deeper irony is that the HAZ is right next to the weld toe, the same place where the geometry concentrates stress. A joint can therefore have perfect-looking weld metal and still fail through a brittle, stress-raised band a few millimetres to the side. Managing the HAZ — by controlling heat input, by preheating thick or high-carbon parts so they cool more gently, and by choosing weldable steels in the first place — is a large part of what serious welding engineering actually involves.

Residual stress and distortion: the cost of heating one spot

Heat one corner of a plate and it expands; the cold metal around it does not, so the hot region is squeezed. As the weld cools it tries to shrink back, but now the surrounding metal restrains it, so it freezes in a state of permanent tension. These locked-in forces, present with no external load at all, are residual stresses, and beside a weld they often reach the metal's own yield strength.

Two consequences follow. First, the part bends — distortion — as the shrinking weld pulls plates out of line, which is why a welded frame can come out of the jig visibly twisted. Welders fight this by tacking parts in sequence, by balancing welds on opposite sides, and sometimes by pre-bending parts the wrong way so they spring straight. Second, residual tension adds directly to service stress: a weld already pulled to near yield by cooling has little margin left for the load it was built to carry. This is one reason fatigue cracks favour welds, and it is why high-integrity structures are sometimes stress-relieved by heating the whole assembly in a furnace to let the locked-in forces relax.

How welds fail in service

Welds rarely fail by simply being overloaded in one pull; a properly sized weld has margin against that. They fail in subtler ways, and recognising the pattern is half of designing against it.

The first and most common is fatigue cracking at the weld toe. Every weld toe is a sharp change in shape, which concentrates stress, and it usually sits in the brittle HAZ under residual tension — three bad things in one spot. Under repeated loading a crack starts there and grows, which is why welded bridges and cranes are designed by fatigue rules far stricter than the static strength of the joint would suggest, and why smoothing the toe by grinding can dramatically extend a structure's life.

The second family is cracking from hydrogen and brittleness. Moisture or contamination releases hydrogen into the hot weld, which diffuses into the hard HAZ and triggers cracks hours or days after welding — the unnerving phenomenon of cold cracking, which is why electrodes are baked dry and joints are kept clean. The third family is internal flaws: lack of fusion where the pool never bonded to the parent metal, porosity where trapped gas left bubbles, and slag inclusions where non-metallic residue was buried in the bead. None of these show on the surface, which is why critical welds are checked by ultrasonic or radiographic inspection rather than by eye.

Rules of thumb and typical values

Working welding design leans on a handful of practical guides. Fillet leg sizes are specified in whole millimetres, with a common minimum near 3 mm; legs much above 8–10 mm are usually replaced by multi-pass welds because a single huge fillet pours in too much heat. A frequent shop rule keeps the fillet leg no larger than the thickness of the thinner plate it joins. Heat input for structural steel typically lands between about 0.8 and 2 kJ/mm, and thick or hardenable steels are often preheated to 100–200 °C so the HAZ cools slowly enough to stay tough.

Quantity	Typical range	Why it matters
Fillet leg size	3–10 mm	Below: too weak; above: too much heat, distortion
Throat / leg ratio	0.707 (equal-leg)	True load-bearing thickness of the weld
Heat input (steel)	0.8–2 kJ/mm	Low: lack of fusion; high: brittle, coarse HAZ
Preheat (thick/alloy steel)	100–200 °C	Slows cooling, avoids cold cracking

How the codes treat welds

Because a weld can be the most dangerous element in a structure, welding is one of the most heavily codified activities in engineering, and the codes share a common philosophy expressed in my own words here. They never trust a weld to a welder's skill alone. Instead they require a written welding procedure that fixes the process, filler, heat input, and preheat; they demand that the procedure be qualified by destructively testing sample welds; and they require the welders themselves to pass tests proving they can follow it. On top of that they set inspection levels — how much of each weld must be examined and by what method — according to how badly a failure would hurt. Structural steel codes also assign every weld detail a fatigue class based on its geometry, because the shape of the toe matters more than the strength of the metal for parts that see repeated load. The thread running through all of it is that a weld is only as trustworthy as the documented, tested, inspected process that produced it.

Engineering judgment

If you take three ideas away from this, make them these. First, design by the throat, not the leg — the shortest path through the weld is what carries the load, and the 0.707 factor is worth committing to memory. Second, respect the heat-affected zone and the weld toe; the joint often fails not in the weld but in the disturbed, stress-raised metal right beside it, especially under fatigue. Third, remember that a weld is a process, not just a shape — its strength is decided as much by heat input, cleanliness, preheat, and inspection as by the dimensions on the drawing. The best welded structures come from engineers who size the weld correctly and then control everything around it.

Welding turns a box of separate plates into a single load-path, which is an enormous capability and an enormous responsibility. Size your welds with margin, draw the toes where stress is low, and treat the procedure as part of the design — and the joint becomes indistinguishable from the structure it serves.

To keep building your design intuition, explore the calculators and explainers at enggtools.in/articles — the pieces on stress concentration and fatigue pick up where the weld toe leaves off.