Structural Bracing: the Braced versus Unbraced Criterion Behind the Code

Structural bracing is the set of members that give a frame its sideways strength and stiffness. Gravity a frame handles naturally, beams and columns are already in line with it. Horizontal load is the problem: wind, earthquake, crane surge and notional out-of-plumb forces all push the floors sideways, and a bare beam-and-column grid with simple (pinned) connections has nothing to stop that push. It is a mechanism. Bracing is what closes it.

There are only three ways to resist that sideways push: bend the columns (a moment frame), fill a bay with a wall (a shear wall or infill), or triangulate a bay with diagonals (a braced frame). Diagonal bracing is the one this guide is about, and it is the most efficient of the three, because a diagonal resists the push in pure axial force instead of bending. Axial members are stiff and light; bending members are neither.

The whole subject comes down to a single, testable idea: a brace must complete a load path (a strength job) and it must be stiff enough to keep the drift inside the code limit (a serviceability job). Get both and the frame is braced. Miss either and you have a frame that either fails a check or, worse, sways enough to crack cladding and frighten occupants while every member still passes its own stress check. We will size a real brace against both criteria on the CalcSteel engine.

Before sizing anything, place diagonal bracing among its alternatives, because the code treats them differently and so does the cost.

• Moment frame. No diagonals. The joints are rigid, and the columns and beams resist sway by bending. It leaves the bay open for doors and glass, but it is flexible and steel-hungry: you often pay for stiffness (drift), not strength.
• Shear wall or infill. A concrete core or a braced masonry panel. Very stiff, but it blocks the bay and ties the steel design to another trade.
• Braced frame. Diagonals triangulate the bay. Stiff, light and cheap, at the cost of a blocked bay line. This is where structural bracing lives: X (cross) bracing, single diagonals, chevron (inverted V) and K bracing, plus the eccentric and buckling-restrained variants used for seismic ductility.

The dividing line the code cares about is sway versus non-sway. A moment frame is a sway frame: its columns can lean, so their buckling (effective) length is long. A braced frame is a non-sway frame: the diagonals hold the joints in place, so the columns buckle over their true length. That single distinction, sidesway prevented or not, is the criterion behind half of what follows.

Start with statics, because the brace force is decided by geometry alone, not by stiffness. Push the top of the bay with a story shear V = 100 kN. In a braced bay with simple (pinned) beam-to-column connections, the columns are vertical two-force members: a vertical member cannot resist a horizontal push, so it carries none of the shear. The only member that can is the diagonal.

The diagonal runs from a base corner to the opposite top corner. With a bay 6.0 m wide and 4.0 m tall its length is the hypotenuse, √(6² + 4²) = 7.211 m, inclined at 33.69° from horizontal, so cos θ = 6 / 7.211 = 0.832. Only the horizontal component of the diagonal's axial force resists V, so:

P · cos θ = V  →  P = V / cos θ = 100 / 0.832 = 120.2 kN

That is the load path in one line: a 100 kN horizontal push becomes 120.2 kN of axial force in the diagonal, and its vertical component, P · sin θ = 66.7 kN, is delivered into the columns as extra axial. The flatter the brace, the larger P: at 20° the same shear would need 106 kN, at 45° only 141 kN over √2. Geometry sets the demand before any section is chosen, which is exactly why this number is the first thing to check by hand.

Codes do not phrase bracing as one rule, they phrase it as three demands that a brace has to meet at the same time. Read together they are the criterion:

1. Strength, the load path. The brace and every element in the path (diagonal, its connections, the collector beams, the columns below) must carry the brace force P = V / cos θ. AISC 360 chapters D (tension) and E (compression), Eurocode 3 clause 6, and NBR 8800 all size the member for this axial force.
2. Stiffness, the drift. The braced frame must not sway more than the serviceability limit, typically H/400 to H/500 under wind (ASCE 7 Appendix CC, AISC Design Guide 3). Drift protects cladding, partitions and comfort, and it is a limit no strength check contains.
3. Stability, the effective length. Because bracing makes the frame non-sway, the columns it braces may be designed with a braced effective-length factor K near 1.0 instead of the sway value that runs from about 1.2 to well above 2. The brace itself must also survive load reversal, so its own compression buckling governs, not its tension yield.

The deep reason bracing works is the third point. A sway (moment) frame lets its column tops translate, and a column whose top can move sideways buckles at a fraction of the load of one whose top is held. Bracing holds that top. In the AISC alignment charts this is the jump from the sidesway-uninhibited nomograph (K ≥ 1) to the sidesway-inhibited one (K ≤ 1). Everything below is that abstract criterion turned into numbers.

Translate the three demands into the three checks an engineer actually runs on the diagonal.

Strength. In tension, φP_n = 0.9 F_yA_g (AISC D2, gross-section yield). In compression, φP_n = 0.9 F_crA_g with F_cr from AISC E3, which collapses with slenderness. Both must exceed P = V / cos θ.

Stiffness. A single diagonal's contribution to lateral drift is Δ = P L_d / (E A cos² θ). This is what has to stay under H/400. Note it depends on area A, not on strength: a stronger steel grade does nothing for drift, only more area or a steeper brace does.

Stability and slenderness. Codes cap the slenderness of the diagonal: KL/r ≤ 200 for compression members and about 300 for tension-only members (AISC and Eurocode agree closely). This is the check that most often decides the section, because a diagonal long enough to cross a bay is naturally slender. For an X brace, the tension diagonal restrains the crossing point, so the compression diagonal buckles over half its length, which is the entire reason X bracing is so common. We derive that in the dedicated steel cross bracing guide.

Here is the demonstration. One bay, 6.0 m by 4.0 m, pinned bases, HEB 160 columns and an IPE 330 beam, S355 steel, under a 100 kN story shear. We run it twice on the CalcSteel finite-element engine and read the eave drift and the column moment.

The bare moment frame drifts 254 mm. That is H/16, seventy-six times over the H/400 serviceability limit, a frame no cladding would survive, and its columns are working in bending, 200 kN.m each. Add one SHS 90x90x5 diagonal, a 17 cm² tube weighing about 13 kg/m, and change nothing else, and the drift drops to 3.3 mm, H/1195. The column moment falls to zero: the diagonal now carries the shear as axial force, so the columns are pure struts again. A brace that weighs a fraction of one column took the frame from a gross serviceability failure to a check passed with a factor of thirty to spare. That is what structural bracing buys, and no amount of column upsizing in the moment frame is as efficient.

Now confirm the load path the statics predicted. In the braced model the engine reports the diagonal axial force as 120.18 kN. The hand value is V / cos θ = 100 / 0.832 = 120.19 kN. The two agree to 0.01 percent, which is the signature of a determinate load path: with simple connections the diagonal carries the entire story shear and nothing leaks into column bending.

The engine also confirms the vertical share. The diagonal's 66.7 kN vertical component appears as added axial in the columns, and the beam carries 50 kN of axial as it drags the far column into the brace. None of this is guesswork: it is equilibrium, and it is why the very first thing to do on any brace is to read P = V / cos θ off the geometry and sanity-check whatever the model returns against it. If the model disagrees with V / cos θ by more than a percent or two, the connections are not as pinned as the drawing claims, and frame action is stealing shear.

The whole braced-versus-unbraced story is one experiment you can run yourself. In the portal frame calculator below, build the 6.0 m by 4.0 m bay with HEB 160 columns and apply the 100 kN lateral load, then read the drift, that is the unbraced case, H/16. Now shrink the columns and watch the moment frame get worse, or stiffen them and watch how much steel it costs to reach H/400 by bending alone. Compare that against the 3.3 mm a single small diagonal delivered above.

It is the same finite-element engine used for the worked numbers, free, and with no login for the math.

Stiffness was easy, the tube delivered H/1195. So what sizes the brace? Strength under reversal. Take the demand P = 120.2 kN and check the SHS 90x90x5 (A = 17 cm², r = 34.8 mm, S355) three ways.

Read top to bottom, this is the lesson of structural bracing. In tension the tube is barely stressed, 22 percent, steel is enormously strong in tension. But wind blows both ways, so the diagonal must also work in compression, and as a single 7.211 m diagonal its slenderness is KL/r = 207, past the 200 code limit, and it buckles at 61.5 kN, a 195 percent failure. The fix is the criterion the code enforces: make it an X, so the tension diagonal braces the crossing point and the compression diagonal buckles over half its length. Slenderness drops to 104 and φP_n jumps to 241.7 kN, 50 percent used. Same tube, same load, different buckling length. The brace was never sized by its tensile strength or by drift, it was sized by its compression stability.

The drift and the vanished column moment are two faces of the same effect. In the unbraced moment frame the column top can translate, so the column is a sway member: its effective-length factor K runs from 1.2 to over 2, and it carries 200 kN.m of bending as it leans. In the braced frame the diagonal holds the top in place, so the column becomes a non-sway member, K near 1.0, with no lateral moment at all. Its axial capacity rises and its demand simplifies.

That is why bracing is worth so much more than its own weight. It does not just carry the wind, it shortens the buckling length of every column it holds and strips the sway moment out of them. A slender column that would need to be heavy to survive as a sway member can be light once a brace makes it non-sway. The alignment-chart jump from sidesway-uninhibited to sidesway-inhibited, and the connected topics of the effective length K factor and drift limits, are the code machinery for exactly this trade.

Sizing the brace for tension only and forgetting reversal. The tension check passed at 22 percent. If you stop there and the wind reverses, the same diagonal is in compression and buckles. Either design a tension-only pair (an X, with a diagonal for each direction) or check the single diagonal in compression.

Connecting to the elastic demand, not the capacity. The diagonal, its gusset, the bolts or welds, the collector beam and the columns below all carry the brace force. Codes ask you to design the load path for the brace's expected strength (in seismic work, its capacity), not merely the 120 kN elastic demand, so the brace yields before its connections break.

Assuming the brace point is real. A brace is only as good as the node it connects to. A diagonal framing into a flexible beam mid-span, or a connection that slips before it bears, does not hold the frame: the drift comes back and the effective length grows. The related brace stiffness and strength requirement (AISC Appendix 6) is the code's answer to this: a brace must deliver a minimum stiffness, not just a minimum force.

Out-of-plane buckling. The compression diagonal can buckle out of the plane of the frame, where nothing braces its midpoint. For a real X, the out-of-plane length is often the full diagonal even when the in-plane length is halved. Check both planes.

Flat is expensive. A very shallow brace (small angle to horizontal) needs a large axial force for the same shear, and it drifts more. Where headroom allows, a steeper brace is a lighter brace.

• Bracing has two jobs, judged at the frame: complete the load path (strength, P = V / cos θ) and hold the drift (stiffness, H/400 to H/500). A brace that passes one and fails the other is not adequate.
• The load path is geometry. On the worked bay, 100 kN of story shear became 120.2 kN of axial, and the engine matched V / cos θ to 0.01 percent. Check this by hand before you trust any model.
• Bracing is astonishingly efficient. One 17 cm² diagonal took the bay from 254 mm of drift (H/16, fail) to 3.3 mm (H/1195), a 76 times reduction, and removed 200 kN.m from each column.
• Compression stability sizes the brace, not tension or drift. The tube was 22 percent used in tension but failed at 195 percent as a full-length compression diagonal (KL/r = 207); crossed into an X (KL/r = 104) it passed at 50 percent.
• Bracing makes columns non-sway. It drops their effective length from a sway K near 2 to a braced K near 1 and strips out their sway moment, which is why the brace is worth far more than its weight.