Installing a Steel Beam

The Crack Arrives Long Before the Collapse

A rolled section chosen purely on bending stress and shear will carry its load with a fat margin and still hand you a ruined ceiling. Steel is stiff, never rigid. Every beam sags under load, and the only live question is how far, and whether the brittle material hung beneath it can follow that curve without splitting. Plaster skim, taped plasterboard joints, floor tile, fixed glazing and any door in a tight casing all give up at strains far below anything that troubles the steel.

Structural failure of a correctly sized beam in a house or a light commercial fit-out is close to unheard of. Nuisance failure is routine. The callbacks that cost real money — a hairline chasing itself across new plaster, a bifold that binds every winter, grout lines opening over a new opening — nearly always trace to a beam that satisfied capacity comfortably and quietly missed the deflection criterion it was really being judged against.

Treat the sag limit as the governing case from the first site visit. Capacity is a pass or fail the engineer settles on paper and nobody ever sees again. Deflection gets judged by the client, in raking light, from the sofa, for thirty years.

Pin the Deflection Limit Before Anyone Picks a Section

Two numbers govern, not one: deflection under live load alone, and total deflection including dead load and long-term effects. Ratios in the region of span/360 for live load and span/240 for total are the familiar starting point, but the binding values come from the deflection table in the building code adopted by the authority having jurisdiction — the International Building Code across most of the United States, the Eurocode route via BS EN 1993-1-1 in the UK, and different tables again elsewhere. Never carry a remembered ratio across a border.

Ratios on their own mislead over long spans. Span/360 on a 3 m opening is a few millimetres and nothing cracks; the same ratio on an 8 m span permits a sag that will open every joint in a plaster ceiling. Engineers commonly pair the ratio with an absolute cap in millimetres once the span passes a threshold, and finish manufacturers — tile adhesive, rigid sheet flooring, frameless glazing — publish movement tolerances of their own that can be tighter than anything the code demands.

Masonry above the beam is the harshest client of all. Brickwork bedded on a flexing beam cracks along the bed joints, usually well in from the ends, and that crack is permanent. Where a beam carries a wall rather than a floor, expect a tighter limit and expect to be asked what sits on top of it before a depth gets chosen. Mention the tiled bathroom before the section is sized, not after.

Know the Mass Before You Plan the Lift

Weight comes before every other booking on site. Mass per metre times length, plus end plates, stiffeners, bolts and any fire protection, sets the crane class or tells you the piece goes in through the window on rollers and shear legs. That same figure sizes the needles, the props, the sole plates and the temporary load path all the way down to something that can genuinely take it.

Bare steel in the air has none of the restraint the finished frame will eventually give it. A long, slender section picked at the wrong points will twist out of plane under nothing but its own weight, and lateral-torsional instability during erection has hurt more people than overload in service ever has. Use a spreader bar on long pieces, pick at the marked lift points, keep slings clear of any cambered zone, and never lift a cambered beam from a single midspan choker.

Camber Is a Timing Decision, Not a Correction

Camber buys back dead-load deflection before the load ever arrives. The beam is fabricated with a deliberate upward bow so that once slab, joists, screed and partitions are on it, the soffit reads flat and the ceiling below has nothing to follow. It works only against loads that are permanent and predictable — self weight, slab, finishes. It does nothing at all for live load, and specifying camber to solve a live-load problem simply hides a section that is too shallow.

Over-cambering is its own defect. A beam left crowned after the dead load lands will telegraph a hump through a screed, throw a level floor out and give the plasterer a ridge to chase for a week. Rolled and induced camber both carry tolerances set by ASTM A6/A6M Standard Specification for General Requirements for Rolled Structural Steel Bars, Plates, Shapes, and Sheet Piling, and heat cambering has a maximum working temperature governed by AWS D1.1 Structural Welding Code — Steel; go past it and the metallurgy changes on you.

Mark the crown. Paint an arrow and the word UP on the web at both ends before the beam leaves the yard, because a cambered section installed inverted doubles the sag instead of cancelling it and there is no cheap way back from that. Check the crown with a stringline on the ground, on packers, before the piece goes anywhere near the opening.

The Supports Sag Too

Net movement at midspan is the beam's own flexural deflection plus whatever each end settles. Engineers calculate the first; site tends to forget the second. A mortar bed compressing, a padstone crushing at a corner, a timber wall plate squashing across the grain, a supporting post shortening, a masonry pier bedding down — every millimetre of that adds directly to the sag the ceiling below has to absorb, and none of it appears anywhere in the beam calculation.

Bearing detail earns its keep at this point. Reactions arrive as a concentrated load on material far weaker than the steel, so spread them: a proper padstone or spreader plate, full and even contact across the entire bearing length, and no attempt to make up a hollow with mortar dabs. Where a beam lands on masonry, bearing length and local crushing are separate checks from the beam itself, and both have to be done rather than assumed.

Rotation at the ends multiplies what anyone actually sees. A beam sitting on a soft or partial bearing rotates as it loads, and the vertical movement measured at a wall line or a window head can exceed the calculated midspan figure by a visible margin. Solid bearings remain the cheapest stiffness available on the job.

Setting Day

Temporary works go in first, and they go in properly: needles through the wall at the spacing the temporary works design calls for, props on sole plates that spread onto something structural rather than a suspended floor, and a load path that continues without interruption to the ground. Prop the load that is genuinely there, including everything above the opening, and leave it standing until the permanent support has been packed solid — not merely until the beam is in the hole.

Dry-fit before committing to anything. Offer the beam up, confirm it sits down on both bearings without rocking, then take a level along the soffit and a measurement to a fixed datum at midspan. Record both; they are the baseline for every reading that follows. A beam touching at three points out of four will find the fourth later, under load, and that movement lands in the ceiling.

Pack with steel, never timber or dry rubble. Fully bedded steel shims, or non-shrink grout worked in until the bearing is solid across its whole area, gives a support that will not compress through the first heating season. Fill the entire bearing, not the middle of it. Where a padstone was cast in place, confirm it has reached the strength the design assumed before letting the beam sit on it.

Load It in the Right Order

Strike props in stages and measure between stages. Take a reading at midspan against the datum, ease the props partially, read again, release fully and read a third time. The drop recorded is the dead-load deflection actually happening, and it ought to land close to what was predicted. A measured drop well beyond prediction means something is wrong at a bearing or the load is not what anyone assumed, and finding that out before the plasterboard goes up costs almost nothing.

Nothing brittle gets fixed until the permanent dead load is on and the beam has taken its position. Sequence ceiling, screed and tiling to follow the steel rather than lead it. Where programme pressure forces finishes on early, say so in writing and expect the residual movement to show up as cracking — deflection does not politely wait because the trade order was inconvenient.

Partitions meeting the underside of a beam need a deflection head, built as detailed instead of screwed tight for speed. A head track with the specified gap, slotted fixings and a sealed but flexible closure lets the beam sag without turning a stud wall into an accidental prop. A partition that becomes a prop picks up load it was never designed for and cracks at every corner.

Reading a Serviceability Failure After the Fact

Diagnosis starts with the crack pattern. Bed-joint cracking in masonry over the opening, a diagonal running from the corner of a door casing, grout failing in a line parallel to the beam, a ceiling crack tracking along the beam line rather than across it — all point at the same cause. Doors binding on the top hinge side in winter and freeing in summer point instead at movement that is load- and moisture-driven rather than a single settlement event.

Bounce is a separate complaint with its own criteria. A long, shallow beam can satisfy every deflection ratio in the code and still feel alive underfoot, because human comfort responds to natural frequency and damping rather than to static sag. AISC Design Guide 11, Vibrations of Steel-Framed Structural Systems Due to Human Activity, covers the assessment. Depth rather than weight is the lever here: a deeper section cures bounce that a heavier shallow one will not touch.

Once a beam is built in, stiffness turns expensive. Realistic options narrow to welding plate to the flanges, bolting a second section alongside, introducing an intermediate post to halve the span, or abandoning the brittle finish and detailing a movement joint instead. Every one of those is worse than choosing a deeper section at the outset, which is the whole argument for dragging the depth conversation to the front of the job.

What the Beam Leaves in the File

Paperwork protects whoever installed it. Keep the mill certificate or declaration of performance for the section, the grade it was ordered to — ASTM A992/A992M Standard Specification for Structural Steel Shapes, ASTM A36/A36M Standard Specification for Carbon Structural Steel for plate in many jurisdictions, the EN grades elsewhere — and the certificates for the bolts, typically ASTM F3125 Standard Specification for High Strength Structural Bolts and Assemblies. Site welds want an inspection record against AWS D1.1 or the execution requirements of BS EN 1090-2, depending on where you are working.

Fire protection is the last thing on and the first thing left out of the weight figure. Intumescent thickness and board encasement both depend on the section factor of the specific shape, so the protection specification follows the final section choice and cannot be read off the nominal size. Apply it after any final packing or adjustment, then protect it from the trades that follow.

Leave the survey readings with the drawings — datum level, dry-fit reading, part-strike reading, full-strike reading, dated. If a crack turns up in year two, that short record separates a beam performing exactly as designed from one that moved because a bearing failed, and it settles an argument that is otherwise ruinously expensive to have.