Building a Steel Mezzanine Deck Over a Working Floor

Where the Legs Land

A warehouse slab is a floor, not a foundation. It was designed for a distributed load over a prepared sub-base, usually 150 to 200 mm thick, often with a single layer of mesh or steel fibre, and cut into panels by saw joints on a grid that has nothing to do with yours. You are about to hand it four to six concentrated loads per bay. Find out what is there before you drill: core it under ASTM C42 where the client will allow a core, scan it first for mesh, conduits and any post-tensioning, and open a trial hole at one location to see the sub-base rather than trusting the as-built.

Two checks decide whether the slab takes the column. The bearing pressure under the plate is one, against the concrete's allowable bearing stress in ACI 318 — a larger plate is nearly always cheaper than any other fix. Punching shear through the slab, together with the bearing capacity of the ground beneath it, is the other, and that is what usually sends you to a broken-out pad footing instead. Keep the plate clear of saw-cut joints and free edges, because both make the shear cone one-sided and both are where the slab was already planning to crack.

The holding-down bolts have their own governing document. Post-installed anchors into an existing slab are designed under ACI 318's anchoring provisions, and in a thin slab the concrete breakout cone, not the steel of the anchor, is very often what limits it. Edge distance, anchor spacing and the member depth all enter that calculation, so a bolt pattern that fits the plate is not automatically a bolt pattern that fits the slab. Where the numbers do not work, a deeper pad or a cast-in solution is the answer; a longer bolt into the same 150 mm is not.

Practicalities on the day. The building grid is not square and the slab is not level, so survey the base positions and pack the plates rather than discovering a 20 mm dip when the last column is plumb. Set the levelling shims where they will stay accessible until the grout goes in. And accept that column bases in a live warehouse get hit — a bollard or an impact-protected base detail at the aisle-facing legs costs less than the first repair, and considerably less than the investigation that follows one.

The Span Between Beams Is the One People Feel

The deck decides the secondary beam spacing, not the other way round. Pick a profile and gauge — 38 mm (1.5 in) rib is the common floor deck, with 51 mm and 76 mm profiles where the spans stretch — and its span table hands you a spacing. Read which span condition the table is quoting, because single-span, double-span and triple-span values for the same sheet are materially different, and a sheet that is continuous over three supports on the drawing must actually be laid that way on the deck.

There are two deflection questions, and mixing them up is the classic error. The construction-stage one is the deck under wet concrete plus the crew and their equipment, and the SDI floor deck standards cap it as a fraction of the clear span with an absolute ceiling as well — take that figure from the ANSI/SDI standard the deck is supplied to rather than assuming it. The in-service one is the finished floor under live load, against the deflection limits in the IBC's structural chapter. Sagging deck also holds more concrete than the drawing allowed, and that extra depth is extra dead load that deflects it further, which is why ponding gets checked rather than eyeballed.

The effective moment of inertia in that check is not something you derive. Cold-formed sheet has an effective width that changes with stress level, and the corrugation geometry, embossments and gauge are all specific to the product, so the value comes from the manufacturer's published table or the SDI's for that exact profile — never from a section calculation on the drawn shape. Any deflection number computed from a made-up I is arithmetic pretending to be engineering.

The deck is also the diaphragm that ties the whole floor together and delivers wind and seismic load into the braced bays. That capacity comes from the fastening pattern: the support fasteners and the side-lap connections between sheets, sized against the SDI Diaphragm Design Manual. Puddle welds through sheet steel fall under AWS D1.3 rather than the D1.1 the frame is welded to, and a decker who button-punches side laps that were specified as screws has quietly changed the diaphragm without changing the drawing.

Finally, what sits on top. A composite slab acts with the deck through embossments and gives the stiffest, quietest floor; a non-composite topping on form deck is simpler and heavier for the same performance; a bare deck with a structural panel or chipboard mezzanine board is fastest to build, lightest, and the one that transmits every dropped tote to the people below. That choice belongs with the client: it sets the floor's acoustic character for the next twenty years.

Beams Are Unbraced Until the Deck Is Down

A beam sized on its full plastic moment is relying on being restrained along its compression flange, and the drawing quietly assumes the deck does that. Before the deck arrives, the same beam is unbraced over its entire span with a decker walking along the top flange. AISC 360's Chapter F provisions define the unbraced length below which lateral-torsional buckling takes nothing off the capacity, and comparing that length against the erection condition is a two-minute check that occasionally changes the temporary works plan.

Restraint arrives in an order. Secondary beams brace the primaries once connected, provided the connection restrains the flange rather than simply hanging off the web, and the deck braces the secondaries once fastened rather than once laid loose. Between those moments the frame is at its most vulnerable, so temporary bracing and a defined sequence belong on a drawing rather than in somebody's head. Camber is rarely worth specifying at these spans, and where it is, AISC 303 sets the tolerance you will actually receive. Deflection still accumulates: a bare frame that is dead level goes down under its own deck and topping, and again under load. Agree the tolerance with the client before the topping goes on, and expect the floor to read a few millimetres low at midspan once it is doing its job, which is the frame behaving, not failing.

Everything Above Arrives at a Handful of Bolts

Every kilogram the deck collects travels along a secondary, into a primary, and then into six or eight bolts through a cleat. Three things can give way there and they are genuinely different failures: the shanks shearing, the steel in front of a hole crushing as the bolt presses into it, and the block of plate between an end hole and the free edge tearing out. AISC 360 handles all three, and which one governs depends as much on plate thickness and edge distance as on the bolt schedule.

The reason this deserves a check rather than a look-up is that the connection changes on site. A cleat gets trimmed to clear an existing pipe run or a stiffener nobody drew, and the end bolt now has a new edge distance. A hole gets slotted so the erector can pull a bay into line, and the slot moves the connection into a different bearing case with a lower coefficient. Neither of those goes back through the design office, and both move the governing limit state — usually onto the plate, where a bigger bolt buys nothing.

The schedule's suffixes matter too. Whether the threads fall in the shear plane is a real difference in capacity that depends on grip length being right, and nobody verifies that from the floor. Snug-tightened, pretensioned and slip-critical are three different installations under the RCSC specification with three different inspection regimes, and a joint drawn as slip-critical does not become one because the bolts feel tight. If the design relies on friction rather than bearing, the faying surface condition is part of the specification and the paint system has to respect it.

During erection the rule is fixed and worth quoting exactly: OSHA's steel erection standard, 29 CFR 1926 Subpart R, does not let the load come off the hoisting line during final placing of a solid web structural member until it is secured with at least two bolts per connection, of the size and strength on the erection drawings, drawn up wrench-tight. That is the minimum for stability, not the completed connection — the balance goes in and gets torqued before anything is stored on the bay above.

Record what you installed. Bolt grade and diameter, hole type, whether the connection was pretensioned and by what method, and the calibration record for whatever was used to do it. On a mezzanine, that file is what an inspector, an insurer or the next occupier's engineer will ask for, and the connections will be behind cladding and services by the time anyone asks.

1. Set and plumb the columns on their packed plates, with holding-down bolts torqued and the levelling shims still accessible.
2. Land the primary beams, two bolts per connection wrench-tight before the crane or lift releases each one.
3. Run the secondaries in the sequence on the erection drawing, keeping every completed line braced before starting the next.
4. Complete and tighten every connection to the specified method, then grout the bases once the frame is plumb and surveyed.
5. Lay and fasten the deck, including the side laps, so the diaphragm the frame relies on actually exists.
6. Pour or fix the topping, then install the permanent guarding before anything else is carried onto the floor — temporary fall protection stays up from the first landed beam until the permanent rail replaces it, not from this step onwards.

It Passes Every Deflection Check and Still Feels Wrong

Mezzanines are the textbook bouncy floor, and the reasons are structural rather than accidental. There are no partitions, no ceiling, no raised floor and often no topping, so the damping available is at the bare-structure end of the range rather than the fitted-out end. Spans are long because the client wanted the columns out of the way. The mass is low. Then someone puts a pack bench on it, and every footfall in the aisle arrives at the bench as movement.

AISC Design Guide 11 is the document that covers this properly. The useful shorthand is that floors with a fundamental frequency low enough to be resonated by the first harmonic of a walking pace behave very differently from stiff floors that respond impulsively to each step, and a mezzanine frequently lands in the awkward region between them. Be honest about what a frequency check gives you: it is a screen, and the guide's full acceleration assessment needs the effective weight of the vibrating floor panel, which no single beam's properties can supply.

The fixes are stiffness and mass, not strength. Shortening the span or deepening the section moves frequency far more effectively than upgrading the grade of steel, which moves it not at all. A concrete topping adds both mass and damping and is the single most reliable cure. Where machinery is the source rather than people — a conveyor, a wrapper, a compressor on the deck — the excitation is at a fixed frequency and the answer may be isolation at the machine rather than anything done to the floor.

The Edge, the Gate and the Way Down

The edge is where a mezzanine hurts people, and it is governed twice by two different rule sets. During erection it is a construction fall-protection problem — OSHA 29 CFR 1926.502(b) fixes guardrail height and the loads a top rail and midrail must take, and Subpart R covers the steel erection around it. Once the building is handed over it is a workplace, and the permanent guarding comes under OSHA's walking-working surface rules in 29 CFR 1910.28 and 1910.29, with the building code's own guard requirements applying in parallel.

Pallet gates are where the thinking usually stops too early. Whatever type is chosen — swing, up-and-over, sliding — the test is that there is no position in its cycle where a person can stand at an open edge. Up-and-over gates pass that test by geometry, which is why they have become the default; a single swing gate does not, no matter how disciplined the operator is. The gate's own landing area also takes the heaviest point load on the entire floor, since that is where a full pallet is set down, and it deserves a local check rather than the field figure.

Then there is what falls off the floor rather than who falls off it. Toe boards, mesh infill or solid panels below the handrail stop stock going over the edge onto the pick face below, and where anything can drop onto a walkway the answer is a physical barrier or an exclusion, not signage. Keep the pallet gate's drop zone out of any circulation route below and mark it on the layout drawing so racking never migrates into it.

Getting off the floor is a code question long before it is a fabrication one. The number of stairs, the maximum travel distance to reach one and the stair geometry itself come from the means of egress chapter of the IBC and, for a workplace, OSHA's stairway rules, and they are the requirements most often discovered late — after the layout has committed the only free wall to racking. The mezzanine's own status matters here too: the IBC's mezzanine provisions limit its area relative to the room below and set conditions on openness, and a mezzanine that breaches them is a storey, with everything that follows for exits, fire resistance and building height.