Section unit difference is the gap between how a material behaves as a small coupon and how it performs once it is built into a full-scale structural member. Ignoring that gap has caused premature failures from warehouse roofs to bridge decks.
Designers who treat the “section” (the real built piece) as identical to the “unit” (the idealized material specimen) unknowingly inherit stress concentrations, residual stresses, and boundary effects that can slash capacity by 30 %. The following sections break down where the difference originates, how to quantify it, and how to design around it without over-conservatism.
Micro-Meso-Macro: Where the Divergence Begins
At micro scale, a 25 mm tensile coupon contains only a few thousand grains; a 300 mm deep beam has millions. The larger population allows extreme-orientation grains to cluster, dropping yield by 8–12 MPa in mild steel.
Meso-scale differences appear at the heat-affected zone. A unit test coupon is machined away from weld seams, so it never sees the 200 °C/mm gradient that locks in 150 MPa tensile residual stress.
Macro-scale effects include flange curling and shear lag. A 1 m wide plate in a built-up girder can lose 15 % of its effective area because edge fibers unload into the web, something the unit test never warns about.
Residual Stress: The Hidden Load Case
Residual stress is locked-in before live loads ever arrive. In a rolled W14×90, the flange tips sit at 0.3 Fy in compression while the web center is equally in tension, creating a built-in bending moment that eats 25 % of the nominal plastic hinge capacity.
Cutting a window in the web releases that moment; strain-gauge surveys show 600 µε instantaneous rebound. Designers can reclaim capacity by specifying thermal stress relief at 650 °C for 30 min, a step that adds only $150 to a 20 ft beam.
Finite-element models that ignore residual stress over-predict buckling strength by 18 % on average. A simple fix is to apply an initial imperfection of L/1000 plus a triangular residual pattern from the ECCS catalogue; correlation with lab tests improves to within 3 %.
Size Effect on Fracture Toughness
A 10 mm Charpy specimen may show 80 J at −20 °C, yet a 50 mm thick flange fractures at 35 J under the same temperature. Thicker sections constrain plastic flow, triaxial tension escalates, and the critical flaw size drops from 12 mm to 3 mm.
ASTM E1921 defines a master curve shift: for every doubling of thickness, T₀ moves 8 °C warmer. A bridge member designed for −30 °C service must therefore use steel rated for −46 °C if the flange exceeds 75 mm.
Specifying thermo-mechanically controlled process (TMCP) steel with 50 % refined ferrite grain refines the fracture probability by an order of magnitude without adding alloying cost.
Shear Lag Reduction Factors That Actually Work
Shear lag slashes net section efficiency when only part of a wide member connects. For a 250 mm wide gusset welded to the middle third of a flat bar, the effective area drops to 0.7 Ag under static load and 0.55 Ag under fatigue.
AISC Specification Eq. D3-1 gives a blanket 0.75 factor, but that is median; finite-element parametrics show the real factor varies from 0.48 to 0.92 depending on connection length ratio a/W. A quick design chart with a = 1.2 W (L 150 mm) pushes the factor to 0.87, saving two bolt rows.
Slotted holes plus seal welds restore 95 % efficiency by letting shear flow redistribute. The extra welding time adds 12 min per connection, offset by removing four bolts.
Buckling: When Plate Slenderness Meets Reality
Unit tests report local buckling at b/t = 12 for 350 MPa steel, yet a built-up box column with the same ratio buckles at b/t = 9 because longitudinal residual stress adds 0.15 Fy compressive pre-load. Reduce allowable slenderness by 20 % whenever longitudinal welds are present.
Post-buckled strength reserves differ. A stub column test shows 15 % post-critical strength, but in a 12 m long member second-order effects erase that reserve; design to the tangent modulus load, not the stub squash load.
Using corrugated webs eliminates local buckling entirely; 2 mm thick folded plate at 60 mm pitch carries 450 kN/m shear without stiffeners, cutting steel weight by 22 % on tested crane girders.
Composite Action: Slip That Changes Strain Distribution
In a push-out unit test, studs develop 90 kN each at 2 mm slip. In a 30 m bridge girder, that same stud sees only 60 kN because slab curling reduces normal stiffness 25 %, shifting neutral axis upward 8 % and raising bottom flange stress 12 %.
Partial shear connection at 70 % is safe in units, yet in long spans the cumulative slip exceeds 6 mm, unlocking a moment redistribution that can over-top the slab by 20 mm under live load. Specify minimum 85 % connection for spans over 25 m.
Longitudinal deck reinforcement can be dropped one bar size when headed studs are arranged in staggered pairs at 300 mm spacing; full-scale tests show the fatigue life doubles by breaking joint lines.
Fire Resistance: When Section Size Alters Temperature Rise
A 150 × 150 × 8 mm SHS reaches 620 °C in 18 min in a 100 MW fire, losing 60 % strength. A 400 × 400 × 16 mm SHS of the same Hp/A needs 38 min because the thicker wall conducts heat away faster, keeping the mid-plane at 510 °C.
Intumescent coatings rated for 90 min on small furnace specimens delaminate on 25 mm plate due to differential expansion. Specify epoxy intumescent with 2 % flexibility additive and anchor nozzles at 300 mm centers to prevent spalling.
Concrete filling adds 55 min to the rating but shifts failure to the reinforcement. Use 2 % steel fiber to keep the core integral when cover spalls; full-scale furnace tests show no explosive failure.
Fatigue: Notch Effects Multiply at Joints
A machined coupon endures 2 × 10⁶ cycles at 240 MPa. The same steel, now a cope-hole detail in a girder, fails at 70 MPa because the 2 mm cold-cut notch introduces Kt = 4.2. Grinding the cope to a 6 mm radius restores 130 MPa capacity.
Post-weld toe burrs 0.3 mm high cut fatigue life by 60 %. A 45 s TIG dressing pass melts 0.5 mm and extends life to 1.8 × 10⁶ cycles, cheaper than upgrading to Category C steel.
Ultrasonic impact treatment (UIT) introduces −250 MPa compression to 4 mm depth; tests on 20 mm cover plates show 3 × life extension even at 1.5 × design stress range.
Corrosion in Crevices: Section Geometry Dictates Rate
A 1 mm gap between doubler and web traps 0.8 mL/cm² chloride solution, creating an oxygen differential cell that corrodes at 0.4 mm/year. Seal welding the edge drops the rate to 0.05 mm/year by excluding the electrolyte.
Hot-dip galvanizing a unit sample adds 100 µm zinc, but in a 300 mm built-up box the kettle cannot fluidize interior corners; zinc thickness drops to 40 µm and gives only 15 years life. Specify internal metallizing after closure welding instead.
Adding 0.3 % copper to weathering steel shifts the protective patina from 3 years to 9 months in marine zones, saving repainting cycles on exposed fascia girders.
Dynamic Amplification: Mass Changes Response
A single beam unit weighs 50 kg; the assembled floor system is 5 t. The added mass drops the natural frequency from 45 Hz to 8 Hz, walking into the human sensitivity range. A 2 % damping ratio in the unit becomes 5 % after composite slab and ceiling, cutting acceleration 55 %.
Finite-element modal updating shows that non-structural partitions add 18 % stiffness but only 8 % mass, raising frequency 0.7 Hz. Include them in serviceability models to avoid over-design with needless tuned-mass dampers.
For footbridges, specify 4 kN/m² accidental load in dynamic analysis rather than static; this captures crowd synchronization and keeps peak acceleration below 0.5 m/s² without extra steel.
Practical Checklist for Design Offices
Run two parallel models: one with ideal material curves, one degraded by 0.3 Fy residual stress and 0.9 fracture toughness. If capacity ratio drops below 0.9, refine the detail before the first shop drawing is issued.
Request mill certificates for 1/4 thickness position from the flange center; the outer 10 mm cools fastest and holds 40 MPa higher yield, information that lets you trim 5 % weight from heavily loaded columns.
Keep a library of past field strain measurements; benchmarking new designs against real data exposes over-optimism faster than any code commentary.