Beam self-weight is the load you must guess before you can compute it, and the allowance is how takedowns break the circle: estimate the member’s weight, size the member under everything including the estimate, then replace the estimate with the actual and confirm the loop closes. StructLoads keeps that loop honest by carrying self-weight as an explicit, updateable line in each member’s ledger rather than a percentage lost inside a lump, because the two failure modes of self-weight are both bookkeeping: allowances that never get replaced with actuals, and slab-and-beam geometry counted twice where the two overlap. The numbers are small per beam, half a kilonewton to a few kilonewtons per metre, and decisive in aggregate: the structure’s own skeleton is typically 15 to 30 percent of everything the columns carry.
The chicken-and-egg, and the standard escape
Sizing needs loads; loads include the member being sized. The traditional escapes are rules of thumb, and they remain good openings. For steel, a floor beam’s weight in kN/m sits almost always between 0.5 and 1.5, light rolled sections near the bottom, heavy ones near the top, and the classic first guess of 10 percent of the beam’s applied load rarely embarrasses anyone on ordinary spans. For concrete, the estimate is geometry times density: a downstand web of width b and depth-below-slab h weighs 24·b·h kN/m, so a 300 by 450 web is 3.2 kN/m, and the span-to-depth habits that set h make the guess nearly deterministic. Transfer members break the rules of thumb, their sections are set by force, not span, and a transfer beam’s self-weight can reach 10 to 20 kN/m and a meaningful share of its own design moment, which is why the iterate-and-replace discipline exists.
The iteration itself converges fast because self-weight is a minor term with a major paper trail: guess 1.0 kN/m, size a beam that turns out to weigh 0.9, and the second pass changes nothing but the record; guess 5 for a transfer and land at 12 and the second pass changes the design. One pass of replacement is almost always enough, two always, and the criterion for stopping is explicit: the weight assumed in the ledger equals the weight of the section on the drawing. What must never happen is the third state, a sized structure whose takedown still carries the openings guesses, because every downstream number, reactions, column loads, footings, inherits the stale term.
| Member type | Opening allowance | Typical resolved weight |
|---|---|---|
| Steel secondary beam | 0.5-0.8 kN/m | 0.4-0.7 kN/m |
| Steel primary beam | 1.0-1.5 kN/m | 0.8-1.5 kN/m |
| Concrete downstand, 300x450 below slab | 24·b·h = 3.2 kN/m | per final section |
| Steel or concrete transfer | 5-20 kN/m, force-sized | often revises the design |
The double-count and the gap: where the geometry bites
Concrete construction hides the classic self-weight error at the slab-beam junction. The slab’s self-weight is computed over its full area; the beam’s weight is then added along its line; and if the beam’s weight was taken as the full rectangle, width times total depth including the slab thickness, the overlap, the slab sitting inside the beam’s depth, has been paid for twice. The honest convention is one of two: beam weight as the downstand only, 24·b·(D − t) with the slab covering the junction, or beam as the full rectangle with the slab area netted down by the beam footprint. Either works; mixing them inflates the dead load by the overlap, roughly 0.3 to 0.7 kN/m per beam line, three to eight percent on a typical frame’s dead load once every line on every floor carries it, a bias that then rides the multi-storey accumulation into the columns as pure fiction. The mirror error, counting the downstand but forgetting the slab thickening at wide band beams, biases the other way; consistency per project, stated once in the calculation record, kills both.
Steel’s version is milder but real: the structural steel ledger should include what rides the beam that the section tables do not list, fire protection at 0.1 to 0.4 kN/m for heavy casings, connection material, and stiffeners on cranked or heavily loaded members, conventionally swept into a few percent on the section weight rather than itemized. The number is small; the habit of deciding where it lives, and writing that down, is the point, the same dead-load honesty that governs every dead-versus-live split.
What self-weight does at the scale of the building
Per beam, self-weight is minor: a 0.7 kN/m secondary under a 25 kN/m floor load is under 3 percent, invisible inside any single member’s margin. In aggregate it stops being minor, because the structure carries itself all the way down: slabs, beams, columns, and walls together commonly make up 15 to 30 percent of a concrete building’s total gravity load, and the column at the bottom of a ten-storey stack carries ten floors of framing weight regardless of whether anyone remembered the beams. The takedown consequence is directional: errors in member self-weight are systematic, not random, the same convention applied to every line of every floor, so a 5 percent bias in the frame’s weight arrives at the foundations as a 1 to 1.5 percent bias in everything, always the same sign. Small, but it is exactly the kind of quiet systematic error that makes an independent takedown check disagree with the original by an amount too consistent to be noise, and chasing that disagreement back to a double-counted junction is a rite of passage.
Software’s role here is bookkeeping, not magic: StructLoads computes member self-weight from the drawn geometry and density, which retires the stale-allowance failure automatically, the weight updates when the section does, and leaves the engineer one duty the tool cannot take, choosing the junction convention and applying it to what gets drawn. A model whose beams are drawn full-depth over an uninterrupted slab has the double-count built into its geometry, and no amount of computation launders a drawing error into a load truth.
Columns, walls, and the rest of the skeleton
The same allowance discipline extends to the vertical members, with gentler stakes and one sharper trap. Column self-weight per storey is geometry again: a 400 by 400 reinforced concrete column is 3.8 kN per metre of height, 13 kN per 3.5 m storey, small against the hundreds or thousands of kilonewtons it carries, and the standard practice of adding each storey’s column weight at that storey, rather than pretending the column is weightless, keeps the accumulation honest for the price of one line per level. Walls scale the same arithmetic up: a 200 mm concrete core wall is 4.8 kPa of elevation, and a tall core’s own weight is a large, entirely computable number that belongs in the takedown from the first pass, not an allowance at all.
The trap is the columns’ change of section. Takedowns built on a single assumed column size carry the top storeys’ slender reality and the bottom storeys’ heavy reality equally wrongly, in opposite directions, and the honest ledger steps the self-weight as the sections step. It is the vertical twin of the beam loop: assume, size, replace, and the stopping criterion is identical, the weight in the ledger is the weight on the drawing, storey by storey. Skeleton weight is the one load the engineer controls completely; it deserves to be the one load the record gets exactly right.
A worked example: closing the loop on one floor
Take one bay of a concrete floor, illustrative round numbers throughout: 8 m primaries at 6 m centres, 200 mm slab (4.8 kPa), finishes 1.5 kPa, live 3.0 kPa. The takedown opens with an allowance: primaries guessed at 4.0 kN/m. Applied load per primary: (4.8 + 1.5 + 3.0) x 6 = 55.8 kN/m, plus the 4.0 guess is 59.8, and sizing lands on a 400 by 600 section, downstand 400 by 400 below the slab: actual weight 24 x 0.4 x 0.4 = 3.8 kN/m. Replace 4.0 with 3.8, nothing downstream moves by more than a rounding, the loop closes in one pass, and the ledger states the section it assumed.
Now the counter-example that pays for the article: the same bay with the beam entered as the full 400 by 600 rectangle, 5.8 kN/m, over an unreduced slab. The junction’s 400 by 200 overlap, 1.9 kN/m, is now counted twice, 3.4 percent of the primary’s total load, and every primary on every floor of a ten-storey frame carries the same fiction downward: the interior columns arrive at the foundation roughly 1 percent heavy, systematically, and the independent checker who nets the junction correctly will sit at a stubborn, uniform 1 percent disagreement until someone finds it. The error was never dangerous, conservative, even, but it was never a decision either, and the worked takedown that documents its junction convention up front is the one whose numbers two engineers can reproduce to the decimal.
Key takeaways: beam self-weight in takedowns
Open with an allowance, 0.5 to 1.5 kN/m for steel floor beams, 24·b·h for concrete downstands, 10 percent of applied load as the blunt fallback, size the member, then replace the guess with the actual until ledger and drawing agree; transfer members, force-sized and heavy, are where the loop genuinely iterates. Choose one slab-beam junction convention, downstand-only or netted slab, and apply it everywhere, because the double-count is small per line and systematic everywhere. Self-weight is 15 to 30 percent of a building’s gravity in aggregate: minor per member, decisive in the columns, and worth exactly one explicit line per member in the record.
Quick answers
How do you allow for beam self-weight in a load takedown?
Open with an estimate, 0.5 to 1.5 kN/m for steel floor beams, 24 times web width times downstand depth for concrete, or 10 percent of the applied load as a first pass, size the member including the estimate, then replace it with the sized section’s actual weight and rerun until the ledger matches the drawing, one pass for ordinary beams, more for transfers. StructLoads computes self-weight from drawn geometry, so the update happens when the section does.
What does a typical beam weigh per metre?
Steel secondaries run 0.4 to 0.7 kN/m and primaries 0.8 to 1.5, with fire casing adding 0.1 to 0.4 where it applies. Concrete downstands follow geometry at 24 kN/m³: a 300 by 450 web below the slab is 3.2 kN/m, a 400 by 400 is 3.8. Transfer members break the bands, force-sized sections reach 10 to 20 kN/m, which is why their self-weight genuinely participates in their own design loop.
How do you avoid double-counting the slab inside the beam?
Pick one convention and state it: either the beam contributes only its downstand below the slab, with the slab’s area load running uninterrupted over the junction, or the beam is the full rectangle and the slab area is netted by the beam footprint. Mixing them counts the overlap twice, roughly 0.3 to 0.7 kN/m per line, a few percent of frame dead load, systematically, and a model drawn full-depth over an unreduced slab has the error built into its geometry.
Does beam self-weight really matter to the totals?
Per member, barely: a secondary’s weight is under 3 percent of its load. In aggregate, decisively: the structural skeleton is 15 to 30 percent of a building’s gravity, every storey of it accumulating down the columns, and self-weight errors are systematic rather than random, the same bias on every line, every floor. That is why a small junction double-count surfaces as a stubborn uniform disagreement between independent takedowns rather than as noise.
When should you not settle for the standard allowance?
On transfer beams and anything force-sized, where self-weight reaches 10 to 20 kN/m and a real share of the member’s own design moment, so the guess-size-replace loop must actually iterate; on long spans, where the weight term compounds with span squared in the moment; and at handover, where an unreplaced allowance anywhere is a defect: the takedown’s final state must carry actual section weights, because every reaction, column, and footing downstream inherits whatever the ledger still assumes.