Weld versus Bolt: Cost, Capacity and Inspection Compared on One Joint

Ask a detailer whether to weld or bolt a connection and you usually get a shop-culture answer: this fabricator welds everything, that erector bolts everything. Both can be right, because the honest answer is not one comparison but three, and they seldom agree. One asks which joint carries more load. One asks which costs less to make. One asks which costs more to prove is sound. A weld can win the first, lose the second and change the third depending on where the work happens.

The cleanest way to see all three at once is to stop arguing in general and design the same joint both ways. So that is what we do here: one plate, one factored reaction, checked once with a bolt group and once with a fillet weld to AISC 360, with the demand taken from a real CalcSteel analysis rather than assumed. Then we line the two up on capacity, on cost and on inspection, and let the numbers say where each belongs.

The demand is not invented. A simply supported beam of 8 m span carries a factored uniform load of 50 kN/m, and the CalcSteel engine returns an end reaction of 200 kN, exactly the closed-form wL/2. That reaction is the shear the end connection has to deliver into the support, so V_u = 200 kN is the number both fasteners must beat.

The connecting element is the same in both designs: a 10 mm plate, 200 mm deep, in A36 steel (F_y = 250 MPa, F_u = 400 MPa). In the bolted design the plate is drilled and bolted to the support; in the welded design the same plate is fillet welded to it along its two vertical edges. Keeping the plate identical is what makes the comparison fair: only the fastener changes.

Take M20 A325 bolts (Group A) in single shear, with the threads in the shear plane, the common and conservative assumption. AISC 360 Table J3.2 gives a nominal shear stress F_nv = 372 MPa (54 ksi), and the bolt area is A_b = πd²/4 = 314 mm². The nominal shear per bolt is F_nvA_b, and with the resistance factor φ = 0.75 the design strength is φR_n = 87.7 kN per bolt.

Dividing the demand by that step, 200 / 87.7 = 2.3, so we need 3 bolts, giving a group design shear of 3 × 87.7 = 263 kN, a 76 percent utilization. Bolt shear governs the fastener here: the plate bearing and tear-out ceiling is 144 kN per bolt, well above 87.7 kN, so the bolts fail before the holes do. Those plate checks still have to be made, and they are covered in the bearing and tear-out guide; they are a property of the plate, not of the choice to bolt.

Now weld the same plate to the support with a fillet along each of its two vertical edges, a total weld length L_w = 400 mm. With E70XX electrodes (F_EXX = 483 MPa, 70 ksi) and the reaction running parallel to the weld axis, the design strength per unit length is φ(0.60 F_EXX)(0.707 w), where w is the leg size. Solve that for the demand and the joint needs a leg of only 3.2 mm.

Here is the twist the tables hide: you cannot use it. AISC 360 Table J2.4 sets a minimum fillet of 5 mm for a 10 mm plate, driven by cooling rate, not by demand, so the code floor governs. A 5 mm fillet over 400 mm delivers 307 kN, a 65 percent utilization, and you get that capacity for free. The maximum leg on a 10 mm edge would be 8 mm, and the base metal (plate shear rupture along the welds, 720 kN) never governs. See the fillet weld strength guide for the directional-strength increase when the load runs across the weld instead of along it.

Line the two designs up and the surprise is how little separates them: 263 kN bolted, 307 kN welded, both clearing the 200 kN demand with margin. On this ordinary joint, capacity is not the discriminator. Either fastener works, and picking on strength alone would be picking on noise.

Capacity does start to decide at the extremes. When the demand climbs and the plate is thick, a weld stays compact where a bolt group would need many rows and a longer plate, so the weld wins on geometry. When the material is thin, coated or awkward to reach with an electrode, bolts win. But for the broad middle of everyday shear connections, both land in the same place, which is exactly why the real decision moves to the next two columns.

The two fasteners do not build capacity the same way. Bolts come in discrete steps: 87.7 kN, 87.7 × 2, 87.7 × 3, and so on. You cannot buy 2.3 bolts, so you round up to 3 and overshoot the demand. Welds are a continuous dial: leg size and length trade against each other smoothly, so in principle you tune the weld to the load.

In practice the weld has its own floor. The 3.2 mm leg the demand asks for is below the 5 mm minimum fillet, so the dial is pinned at its low end and the joint is oversized by the code, not by you. The lesson runs both ways: on a lightly loaded joint the minimum fillet makes the weld cheap and strong by default, while a heavily loaded joint rewards the weld's smooth scaling, since a bigger leg or a longer run adds capacity without adding a single hole or bolt.

It is tempting to think choosing a fastener settles the connection. It does not. Bearing, tear-out, block shear and net-section rupture are limit states of the plate, and the plate is common to both designs. Bolt the plate and you still check bearing at each hole and block shear through the bolt group. Weld the plate and the holes are gone, but gross shear yield and shear rupture of the plate, and the strength of the base metal at the fusion line, are still on the list.

So switching from bolts to welds moves the weakest link, it does not delete the chain. The detail checks live in their own guides: bolt bearing and tear-out, block shear and net versus gross section, and prying action when the joint carries tension rather than pure shear. The point for this comparison is only that the element checks are shared overhead, paid whichever fastener you pick.

Cost, not strength, is what usually decides, and it is not a single number, it flips with where the work happens. In the shop a weld is cheap: the piece is fixtured flat, the process can be semi or fully automated, and there is no drilling pattern to coordinate between two parts. In the field the balance inverts. A bolt goes in with a wrench in minutes, by a crew that does not need a welding certification, a power source, shielding gas, edge preparation or a tent against wind and rain.

That is the old fabricator rule with a reason behind it: weld in the shop, bolt in the field. Field welding is the expensive corner of the matrix, and it is expensive before a single inspection happens, because of access, position, weather and the certified labor it needs. On the material side the trade is smaller and points the other way: bolts add hardware you buy by the box, welds add consumables and energy, so per joint the fastener hardware is rarely what tips the decision. Labor and location are.

The third comparison is the one estimates forget. A snug-tight bolt, which is all a typical shear connection needs, is verified by a visual check: the plies are in firm contact and the nut is on. A pretensioned or slip-critical bolt needs a verified installation method, turn-of-nut, direct-tension-indicator washers, twist-off bolts or a calibrated wrench, per the RCSC Specification and AISC 360 Chapter N. Either way the evidence is on the surface, cheap to see and cheap to repeat.

A weld is a different bill. Every weld gets a certified visual inspection to AWS D1.1, and beyond that the demand category calls up nondestructive testing: magnetic particle or dye penetrant for surface flaws, ultrasonic or radiographic testing for complete-penetration groove welds. That means a qualified inspector, more time, sensitivity to weather and access, and defects that can hide inside the metal where only NDT finds them. The joint that was cheapest to weld in the shop can become the one that is most expensive to prove, and that cost is real even though it never shows on the erection drawing.

Put the three columns together and a working rule falls out. Reach for bolts when:

• the work is in the field and erection speed matters;
• the joint may be taken apart later (temporary works, phased erection, future extension);
• hot work is a hazard, near galvanizing, coatings, or flammable contents;
• the material is thin or coated and you want to avoid distortion and burn-through.

Reach for welds when:

• the work is in the shop, where welding is cheapest and best controlled;
• the joint must be compact or stiff, a moment connection, a stiffener, a built-up member with no room for a bolt group;
• bolt access is poor or a clean, hole-free appearance is wanted.

Two nuances override the rule. Under fatigue, welds introduce stress concentrations and demand careful detailing, while a properly pretensioned slip-critical bolt can be the calmer choice. And never share one load path between bolts and welds: they have very different stiffness, so they do not draw load in proportion to their strength, and one arrives first. Let each element be all-bolted or all-welded.

The bolt side of the comparison hides a second decision: how much to pretension. Snug-tight and slip-critical are not the same bolt, and the target pretension sets both the clamping force and the installation method an inspector will later verify. Use the calculator below to see how bolt grade and diameter drive the pretension and the tightening torque, then map that back onto the inspection column above: the more pretension the joint relies on, the more the installation has to be proven.

The worked joint is in AISC 360, but the comparison is code-independent, only the notation changes. In EN 1993-1-8 a bolt in shear is F_v,Rd = α_v f_ub A / γ_M2, the same F_nvA_b idea with the safety carried by a partial factor instead of φ. A fillet weld uses the directional or the simplified method, with a design shear strength f_vw,d = f_u / (√3 β_w γ_M2), and the correlation factor β_w plays the role of the electrode match.

ABNT NBR 8800 follows the same LRFD-style format as AISC, with resistance factors on bolt shear and on weld metal. The upshot is that the three-column decision, capacity, cost, inspection, is identical across codes; what moves between them is the exact factor and the minimum-fillet table, not the logic of when to weld and when to bolt.

• Sizing the weld from demand and forgetting the floor. The load asked for 3.2 mm, but the 5 mm minimum fillet governs. Specifying the smaller leg is a code violation, not an optimization.
• Sharing one load path between bolts and welds. Their stiffnesses differ, so they do not share load by strength; the stiffer path is overloaded first. Make each element all-bolted or all-welded.
• Confusing snug-tight with pretensioned. Pretensioning every bolt wastes money on joints that only need snug; leaving a slip-critical or fatigue joint snug is unsafe. Match the class to the demand.
• Field welding coated or galvanized steel without prep. Zinc and paint make porosity and fumes; the weld that passed on bare plate can fail on a coated one.
• Inspecting the wrong thing. Ordering ultrasonic testing on a fillet weld, or skipping pretension verification on a slip-critical joint, spends the inspection budget where it does not buy safety.