How the two differ in kind
A bolt sits in a hole that has been made deliberately larger than the bolt. That clearance is not slack workmanship; it is the tolerance the entire erection process runs on, and it is why a beam cut a couple of millimetres long still lands on a column that is a couple of millimetres out of plumb. A weld does the opposite. Filler and parent metal fuse, and when the joint cools there is no interface left to have a tolerance — the two pieces are one piece. Nearly every practical difference between these two connections is that single fact working its way outward through the drawing office, the shop and the site. It even changes what the word capacity means. A bolt group has four things competing to fail: the shank in shear, the steel crushing in front of the hole, the material ahead of the hole tearing out to a free edge, and the member's own section once the holes have taken material out of it. The smallest of them is the answer, which is why so many bolted connections turn out to be governed by the plate rather than by the fastener, and why buying a stronger bolt often buys nothing at all. A weld has an effective throat of fused metal, and then the base metal at the fusion face behind it, which governs a great many joints on its own.
The practical divider is not strength but location, and it is the reason the trade's standing default is shop weld, field bolt. Welding wants everything a fabrication shop has and a steel frame in the air does not: the work turned by a positioner so the weld runs downhand instead of overhead, still air so nothing blows the shielding gas off a gas-shielded arc, electrodes and flux kept dry because hydrogen is a cracking mechanism rather than a housekeeping preference, preheat that can be applied and measured on thick material, and an inspector who can physically reach both sides of the joint. Bolting wants almost none of that. It wants access, a crew and a wrench, and it is largely indifferent to cold, wind and rain. Erection sequence then compounds the split. A piece landed and pinned with erection bolts is stable within minutes and the crane moves to the next lift; a field-welded joint holds the crane, or a set of temporary steel and guys, until it has been welded, cooled and inspected. Across a frame with hundreds of connections, that is not a detail — it is the programme.
Then there is the asymmetry nobody enjoys discussing, which is how each connection is proved. A bolted joint largely tells you what it is. Snug-tight is a visual judgement about plies in firm contact. Where pretension is required, the verification method leaves its evidence sitting on the finished joint: match marks turned through the specified angle, a direct tension indicator squashed to its gap, the spline sheared cleanly off a tension-control bolt. A weld tells you nothing of the sort. A perfectly formed cap can sit over porosity, slag inclusions or a wholesale lack of fusion at the root, and finding those means ultrasonic or radiographic testing by a separately qualified trade, working to a procedure written before the arc was ever struck. That is not an argument against welding — welded joints are made soundly by the thousand every day — it is an argument for deciding early, because a welding procedure, a qualified welder and a testing regime are a cost and a programme item, not something to discover during erection. Which is the honest shape of this comparison: it has no winner, but it has four questions that settle any given joint. Where will this connection actually be made? Is there a back face to get a wrench onto, or is the member closed? Does the design want this joint rigid, or does it want it to rotate? And will anything here ever need to come apart?
The factors that actually differ
| Bolted connection | Welded connection | |
|---|---|---|
| How the load crosses | Through the shank of a discrete fastener, or by friction between clamped plies if the joint is pretensioned and slip-critical. Four limit states compete — bolt shear, bearing in front of the hole, tearout to the nearest edge, and the connected member's own net section — and the smallest one is the connection. | Through the effective throat of fused metal, which for an equal-leg fillet is a shade over seven-tenths of the leg size, and then straight on through the base metal at the fusion face. The base metal governs plenty of joints, especially where thin material is being welded with a generous fillet. |
| Tolerance built into the joint | Designed slack. A standard hole is clearance-drilled over the bolt, oversized and slotted holes give more where movement has to be absorbed, and between them they swallow the ordinary reality of column plumb, beam length and a holding-down bolt set where the concrete crew set it. | Fit-up is the job. Root gaps and fillet legs are specified with tolerances of a millimetre or two, and a gap outside them is not fixed by putting in more weld — it is a remedial procedure, an increased weld size or a re-cut, all of them decided by someone other than the person holding the torch. |
| Where the work wants to happen | Anywhere. Cold, rain, wind and awkward positions slow a bolting crew down without changing what the connection is, and the tools are a wrench and access. | In a shop, on a positioner, in the flat position, out of the wind, with dry consumables and controlled preheat. Field welding is possible everywhere and comfortable nowhere: shelters against wind for gas-shielded processes, hot-work permits and a fire watch in an occupied building, and out-of-position work by a welder qualified for that position. |
| Proving it was done right | Largely observable after the fact. Snug-tight is visual; pretension is verified by an agreed method — turn-of-nut match marks, calibrated wrench, a squashed direct tension indicator, a sheared tension-control spline — and most of that evidence stays visible on the joint for anyone who walks it later. | Visual inspection reaches the surface only. Internal soundness needs volumetric testing — ultrasonic or radiographic on complete-penetration welds, magnetic particle or dye penetrant for surface-breaking flaws — against a qualified procedure, by a qualified welder, with a paper trail. The inspection is a trade of its own and needs its own access. |
| Section shape and access | Needs to reach both faces: a wrench on one side and something holding the nut on the other. A closed hollow section refuses that outright, which is why bolting a tube usually means welding an end plate to it first, or reaching for specialist one-sided blind fasteners. | Single-sided by nature, which is exactly why tubular trusses, hollow-section nodes and skewed diagonals are welded. Odd angles that would need a bent plate or a fabricated fitting to bolt are just a different cut and a different weld preparation. |
| Heat, distortion and residual stress | A cold operation. Nothing in the assembly moves that the crane did not move, and the geometry after bolting is the geometry the erector set. | Weld metal shrinks as it cools and drags the joint with it — angular distortion across a fillet, shortening along a long seam, and locked-in residual stress at every run. In the shop this is managed with sequence, jigs and deliberate pre-set. In a frame being welded closed in the air it can pull columns out of plumb while the crew watches. |
| What it costs the member | Every hole removes material. A tension member has to be checked on its net section and on the block shear path through the bolt pattern, so the connection can quietly cost the member a slice of the capacity it was chosen for. The steel itself only has to be drillable or punchable. | Takes nothing away — the gross section can be developed. In exchange the steel has to be genuinely weldable: carbon equivalent, thickness-driven preheat and, where a weld pulls on a plate through its thickness rather than along it, the risk of lamellar tearing along the rolling-direction inclusions. |
| Fatigue and load reversal | A correctly pretensioned slip-critical joint transfers load by friction with no slip and no stress raiser at the transfer, which is why it is the standard answer under crane runways, on machine supports and anywhere the force reverses. Bearing joints under reversal need the movement thought about, not assumed away. | Every weld toe is a stress concentration, and the fatigue codes classify welded details accordingly rather than treating them as parent material. The class can be improved — grinding the toe, removing backing, easing the transition — but that is deliberate work specified in advance, not a bonus. |
| Taking it apart afterwards | Unbolts. Temporary works come down, a frame can be extended, a member can be swapped, and a structure detailed for deconstruction can genuinely be deconstructed rather than demolished. | Comes apart with a torch, and cutting it destroys the ends of the members it joined. Reversibility is a property a connection either has or does not, and only one of these two has it. |
Which one, and when
Choose bolted connection when…
- The joint is being made on site. Weather, welding position, preheat control and inspection access all point the same way, and a bolting crew works in conditions no welder should be asked to.
- The connection has to absorb erection tolerance — column plumb, beam length, a base plate landing on holding-down bolts that are where they are.
- The load reverses or cycles: crane runways, machine bases, braced bays taking wind both ways. A pretensioned slip-critical joint is the fatigue-friendly answer and a weld toe is not.
- Something here is meant to change. Temporary steel, a frame designed to be extended, plant that will be replaced, a building intended one day to be taken apart rather than knocked down.
- The design wants the joint to rotate. A single vertical line of bolts through a shear tab gives a simple connection the flexibility its analysis assumed it had.
Choose welded connection when…
- The work is happening in the shop, where the position, the environment, the consumables and the inspector are all already under control.
- The member is a closed hollow section, or the geometry is skewed — no back face for a wrench, no pair of parallel plies to clamp.
- The joint has to be continuous and rigid, or has to develop a tension member's full gross section without losing anything to holes.
- Space is tight. Bolts need edge distances, spacing, and a plate large enough to hold all of it; a weld needs only a fusion face, so a welded joint is almost always the more compact one.
- The connection is a repair or a strengthening on existing steel where there is no clean way to introduce a bolt group into what is already there.
Now run your own numbers
This page holds no prices on purpose — a national average is wrong for almost every real project. Quantify both options with your dimensions and your local quotes.
Frequently asked questions
- Which one is stronger?
- It is the wrong question, and it is worth saying so plainly, because both connections get designed until they carry the load and neither has a general capacity advantage. What differs is which limit state governs and how much of the answer sits in the plates rather than in the fastener or the weld metal. A bolt group is the lowest of shear through the shank, bearing in front of the hole, tearout to the edge and the member's own net section — so a stronger bolt frequently buys nothing, because the plate was the limit. A weld is the throat, and then the base metal behind it, which governs a great many joints on thinner material. Note what the two calculators here actually return, because they are not the same kind of number and treating them alike will cost you. The bolt tool has already applied the resistance factor, so what it shows is a factored design capacity — the governing per-bolt value multiplied by the count — but only across the checks it makes, with block shear, net section and eccentricity declared out of scope. Do not discount it a second time. The weld tool gives the NOMINAL weld strength, which has had no factor applied at all and still needs one, plus the connected base metal checked separately. Use both to see which limit state is close, not to sign off a connection.
- Why is 'shop weld, field bolt' the standard approach?
- Because every advantage welding has lives in the shop, and every advantage bolting has lives on site. In the shop the work can be turned so the weld runs downhand, the air is still so shielding gas stays where it is meant to be, consumables stay dry, preheat can be applied and measured, and the inspector is already there with access to both sides of the joint. On site none of that is true, and all of it has to be bought back with enclosures, position-qualified welders, hot-work permits and testing access. Bolting, meanwhile, is close to indifferent to where it happens. Then there is the crane. A piece landed and pinned with erection bolts is stable in minutes and the hook moves on; a field-welded joint holds the crane or a set of temporary steel until it has been welded, cooled and inspected. Multiply that across a frame and the default stops looking like tradition and starts looking like arithmetic.
- Which one should make a moment connection?
- Both can, and this particular question has real history behind it. Welded moment frames were the assumed answer in seismic regions until the 1994 Northridge earthquake, where beam-flange-to-column welds fractured brittlely across a large number of buildings — connections that had been treated as prequalified by convention rather than demonstrated by testing. What came out of the investigation reshaped practice: connection geometries that must be tested to demonstrate the rotation they claim, notch-tough filler metals specified for demand-critical welds, real attention to backing bars and the shape of the weld access hole, and — importantly here — bolted moment connections such as the extended end plate and the bolted flange plate taking their place as tested, prequalified options in their own right. In ordinary non-seismic work the choice is more about labour location: a welded moment connection concentrates the work in the shop, while a bolted one buys field speed at the price of a larger end plate, more bolts and the steel needed to carry them.
- Is field welding ever the right call?
- Yes, and pretending otherwise would be dishonest. Heavy column splices, tubular and skewed nodes that simply cannot be bolted, connections where the bolted equivalent would need a plate that does not fit in the space available, and repairs or strengthening of existing steel where there is no clean way to introduce a bolt group — all of these are routinely field-welded, and correctly so. What has to be planned for is everything the shop was quietly providing. Weather enclosures and windbreaks, welders qualified in the positions the joint actually presents rather than the ones that were convenient to test, preheat control on thick material, consumable storage that keeps hydrogen out, a hot-work permit and fire watch if the building is occupied or finished, physical access for the inspector and their equipment, and the fact that the erected piece stays on the crane or on temporary steel until the weld has cooled and passed. None of that is a reason to avoid field welding. It is a reason to price and programme it before it appears on a drawing.
- How do the costs actually differ?
- Structurally rather than numerically, which is the only honest way to answer it. Weld cost is dominated by deposited metal and the labour hours to deposit it, and there is a geometric trap in that: the cross-sectional area of an equal-leg fillet grows with the SQUARE of the leg size while its capacity grows only linearly with it. Specifying a leg one size larger than the calculation needs is therefore far more expensive than it looks, and two modest fillets on both sides of a plate usually beat one heavy one on a single side. Codes also set minimum fillet sizes tied to the thickness of the thinner part joined, so there is a floor you cannot design below. Bolt cost is per hole and per fastener: in a shop with a CNC drill line the holes are close to incidental, and the fastener assembly is a purchased item with a purchased-item price. Where the two genuinely diverge is location — bolting costs roughly the same in the shop and in the air, welding costs a great deal more in the air — and then afterwards, because a welded joint is decisively the more expensive one to change. Every alteration is hot work, undoing it damages the members it joined, and any protective coating has to be made good again.
- Do bolts have to be tightened to a specific tension?
- It depends on the joint type, and this is probably the most commonly muddled thing on a steel site. A snug-tight bearing joint needs only the plies brought into firm contact — the bolt bears on its shank, and pretension is not part of the calculation. A pretensioned joint requires a specified minimum bolt tension, installed by an accepted method: turn-of-nut from snug, calibrated wrench, a direct tension indicator, or a tension-control bolt whose spline shears off at tension. A slip-critical joint requires that same pretension AND a faying surface of a specified class, because the load crosses by friction before the shank ever touches the side of the hole — which makes paint, mill scale, dirt, grease and ice on the mating faces structural items rather than housekeeping. One thing worth fixing in your head regardless: the specification is tension, not torque. Torque is only an indirect and lubrication-sensitive way of estimating tension, which is precisely why the other verification methods exist.