Skip to content
  1. Blog
  2. Standards

Effective Length Factor K for Steel Columns - Braced vs Unbraced Frames to CSA S16

CSA S16 has no sway chart for K. Find K for a braced steel column with Annex G, on a 6.0 m axially loaded W250x67 that fails at K = 1.0 and passes at 0.85.

Calculations referenced in this post

Beam-column Pro report

Beam-column resistance of W shapes, HSS and channels to CSA S16 Cl. 13.8

Open the calculator →
On this page
Effective length factor K: braced X-frame and unbraced frame with right-pointing arrow

TL;DR

  • S16 has no sway chart. In an unbraced frame, P-Delta and notional loads go into the analysis and in-plane checks use K = 1.0.
  • For buckling in a braced frame, Annex G gives K between 0.5 and 1.0 from G = sum of column I/L over sum of girder I/L at each end.
  • A 6.0 m W250x67 with a pinned base and stiff girders gets K = 0.85, which lifts Cr from 2050 to 2220 kN, about 8 percent.
  • Refine K only on the axis that governs. Brace the weak axis at mid-height instead of third points and it drops to 1950 kN, whatever K is.

You'll learn

  • Why K is the distance between inflection points, and why S16 recommends 0.65 and 0.80 instead of the theoretical 0.5 and 0.7
  • How CSA S16-19 Cl. 10.3 splits K between in-plane bending (K = 1.0 with sway in the analysis) and buckling (K from end restraint)
  • How to get K for a column in a braced frame from the Annex G equation, including base and far-end girder adjustments
  • When refining K pays, and when the other axis makes it pointless

A W250x67 column, 6.0 m tall on a pinned base, fails a 2150 kN axial check at K = 1.0 and passes at K = 0.85, as long as it stays axially loaded. That 0.85 isn’t from a table. It comes from the stiffness of the girders at the top of the column, and in a braced frame that stiffness is worth about 8 percent of the column’s capacity.

Most of us learned K from a US textbook, with two alignment charts and K values of 1.5 or 2 in moment frames. CSA S16 works differently. S16-19 Cl. 10.3 picks K by how the column fails. In-plane bending gets K = 1.0 with the sway carried by the frame analysis, buckling gets K from the end restraint, and S16 has no sway chart at all.

K is the distance between inflection points

Annex F describes KL as the length of a pin-ended column with the same capacity as the real one. A pin carries no moment, so any point where the buckled column’s curvature flips acts like a pin. That’s an inflection point, and the distance between two of them is the effective length. So the familiar values are sketches you can check, not numbers to memorize.

A column fixed at both ends bows to one side, but its curvature reverses a quarter of the way in from each end, and those are the inflection points. The pin-ended stretch in the middle is half the length, so K is 0.5. A flagpole, fixed at the base and free at the top, bends into a quarter of a sine wave, and you have to mirror it below the base to finish the half-wave. That makes the equivalent pin-ended column twice as long, so K is 2.0, and because capacity falls with the square of KL, the flagpole carries a quarter of the pin-ended load.

Three vertical members with downward arrows and dashed curves; labels K = 1.0, K = 0.5, K = 2.0

In-plane bending or buckling - which K S16 wants

Cl. 10.3.1 sends you to one of two clauses depending on how the member would fail. A beam-column that fails by bending in its own plane uses its actual length, K = 1.0, as long as the frame analysis already includes the sway. A column that fails by buckling, whether straight axial buckling or lateral-torsional buckling out of plane, gets K from the restraint at its ends.

That split is what replaces the sway chart. In a moment frame, S16 asks you to put the sway into the analysis instead of into K. As the frame drifts, the gravity loads act through that sideways movement and add moment, which is the P-Delta effect. You capture it with a second-order analysis, or with the U2 factor that amplifies the sway results of a first-order one. You also add notional lateral loads of 0.005 times the factored gravity load at every storey, which stand in for columns that aren’t quite plumb. The sway then shows up as larger moments and axial loads, and the overall member strength check in Cl. 13.8.2 uses those with K = 1. Use a K of 1.5 as well and you’ve counted the sway twice.

So braced versus unbraced decides how you analyze the frame, not which chart you read. Cl. 13.8.1 calls a frame braced when its bracing makes it at least five times as stiff against sway as the moment connections alone. Anything less is unbraced, even with braces in it.

Annex G for a column in a braced frame

Annex G covers the buckling case in a braced frame, where the column ends can rotate but can’t move sideways. The stiffer the girders are compared with the column, the less the joints rotate and the closer K gets to the fixed-ended 0.5. Flexible girders leave the column close to pin-ended, at 1.0. The annex turns that into a number in four steps.

  1. At each end of the column, find \(G = \Sigma (I_c/L_c) / \Sigma (I_g/L_g)\), counting only members rigidly connected at that joint and bending in the plane of buckling. I is each member’s moment of inertia and L its length, with c for columns and g for girders.
  2. At the base, use G = 10 for a column that sits on the footing without a rigid connection, and 1.0 for a base rigidly attached to a properly designed footing (G.4).
  3. Adjust any girder whose far end is known. For no sway, multiply its I/L by 1.5 when the far end is hinged and by 2.0 when it’s fixed (G.5).
  4. Solve the G.3 equation for K, or read it off the nomograph in Figure G.1, S16’s alignment chart for sidesway prevented.

Step 3 is where imported habits bite. Many references tell you to halve a girder’s stiffness when its far end is pinned, but that factor belongs to the sway chart, where the girders bend in double curvature. The no-sway nomograph assumes single curvature instead, the girder’s floppiest shape. A hinged far end is actually stiffer than that (3EI/L against 2EI/L), and a fixed one stiffer again, which is where the 1.5 and 2.0 come from.

Line graph of effective length factor K versus G, with red and dark curves and a red marker

Neither curve gets near 0.5. Even with stiff girders a pinned base keeps K close to 0.7, and a “fixed” base at G = 1.0 stays above about 0.63.

Worked example - a 6.0 m ground-storey column

The column is an interior W250x67 in a two-storey braced building. Cl. 10.3.1 measures a bottom storey from the top of the base plate to the centre of the girders above, which here is 6.0 m. W460x52 girders spanning 9.0 m frame into the flanges with moment connections on both sides, and a W250x67 continues 4.0 m to the roof. The base is a four-bolt plate on a pedestal, not a rigid connection, so G = 10.

Is this a beam-column or an axially loaded column? The spans and loads match on both sides, so the girder moments balance and the column sees almost pure axial load. That makes it a buckling case under Cl. 10.3.3, and Annex G applies about the strong axis, if pattern live load leaves the moment negligible. Struts tied back to the bracing hold the weak axis at third points, so its unbraced length is 2.0 m with K = 1.0.

Calculation

Effective length and axial resistance of a 6.0 m W250x67 column in a braced frame

CSA S16-19 10.3.3CSA S16-19 Annex GCSA S16-19 13.3.1.1

Braced frame meeting the Cl. 13.8.1 five-times sway stiffness test, balanced girders so axial load only, 345 MPa steel, n = 1.34; weak-axis struts designed as braces to S16-19 Clause 9 and restraining twist, so Lz = 2.0 m and torsional buckling does not govern

Inputs
9 input values\(C_f = 2150\) kN, \(L = 6000\) mm, \(F_y = 345\) MPa ...
\(C_f\)2150kNFactored axial load from the governing NBCC combination
\(E\)200000MPaElastic modulus
\(L\)6000mmTop of base plate to centre of level 2 girders
\(F_y\)345MPaYield strength
\(A\)8550mm²W250x67 area
\(I_x\)104 x 10^6mm⁴W250x67 strong-axis moment of inertia
\(r_x\)110mmW250x67 strong-axis radius of gyration
\(r_y\)51.0mmW250x67 weak-axis radius of gyration
\(I_g\)212 x 10^6mm⁴W460x52 girder
Derivation 6 steps
Step 1: G at the top joint\(G_U = 0.92\)
\[G_U = \frac{I_c/L_c + I_c/L_{c,above}}{2\,I_g/L_g}\]
\[= \frac{104 \times 10^6/6000 + 104 \times 10^6/4000}{2 \times 212 \times 10^6/9000} = \frac{43.3}{47.1}\]
\(G_U = 0.92\)Both girders have columns at their far ends, so no G.5 factor applies.
Step 2: G at the base\(G_L = 10\)
\[G_L\]
\(G_L = 10\)Base plate on a pedestal, not rigidly connected.S16-19 G.4
Step 3: K from the Annex G equation\(K = 0.85\)
\[\frac{G_U G_L}{4}\left(\frac{\pi}{K}\right)^2 + \frac{G_U + G_L}{2}\left(1 - \frac{\pi/K}{\tan(\pi/K)}\right) + \frac{2\tan(\pi/2K)}{\pi/K} = 1\]
\[\text{solved numerically with } G_U = 0.92,\ G_L = 10\]
\(K = 0.85\)The root is 0.854; 0.85 changes Cr by under 0.2 percent.S16-19 G.3
Step 4: Strong axis at K = 1.0\(C_r = 2050\ \text{kN}\)
\[\lambda = \frac{KL}{r_x}\sqrt{\frac{F_y}{\pi^2 E}},\quad C_r = \frac{\phi A F_y}{(1 + \lambda^{2n})^{1/n}}\]
\[= \frac{6000}{110}(0.01322) = 0.721,\quad C_r = \frac{0.9(8550)(345)}{(1 + 0.721^{2.68})^{1/1.34}}\]
\(C_r = 2050\ \text{kN}\)S16-19 13.3.1.1
Step 5: Strong axis at K = 0.85\(C_r = 2220\ \text{kN}\)
\[\frac{KL}{r_x} = \frac{0.85 \times 6000}{110} = 46.4\]
\[\lambda = 46.4(0.01322) = 0.613,\quad C_r = \frac{2655}{(1 + 0.613^{2.68})^{1/1.34}}\]
\(C_r = 2220\ \text{kN}\)
Step 6: Weak axis, braced at third points\(C_r = 2360\ \text{kN}\)
\[\frac{KL}{r_y} = \frac{1.0 \times 2000}{51.0} = 39.2\]
\[\lambda = 39.2(0.01322) = 0.518,\quad C_r = \frac{2655}{(1 + 0.518^{2.68})^{1/1.34}}\]
\(C_r = 2360\ \text{kN}\)
Checks
Section not slenderS16-19 Table 1PASS
flange b/t 6.5 against 10.8, web h/w 25.4 against 36.1 Demand / capacity0.70
Strong axis with K = 1.0FAIL
\(C_f \le C_r\) : 2150 kN against 2050 kN Demand / capacity1.05
Strong axis with Annex G, K = 0.85PASS
\(C_f \le C_r\) : 2150 kN against 2220 kN Demand / capacity0.97
Weak axis, 2.0 mPASS
\(C_f \le C_r\) : 2150 kN against 2360 kN Demand / capacity0.91
Watch outAnnex G is for buckling only. Once the column carries real moment, the overall member check uses K = 1.0, and here that is 2150 kN against 2050 kN, a fail before the moment term is added.
Different inputs? Beam-column runs this check with yours. Check your own column at its K →

The default K = 1.0 fails this column by 5 percent, and the girders it already has make up the difference. Annex G lifts it from 2050 to 2220 kN, enough to pass as an axially loaded column.

Columns outside a frame - Figure F.1

For an isolated column with no girders to count, Figure F.1 gives six ideal cases, but its theoretical values assume restraints that real connections only approximate. Real fixed ends rotate a little, so S16 raises K in three of the six cases.

Bottom endTop endTheoretical KS16 recommended K
FixedFixed0.50.65
FixedPinned0.70.80
PinnedPinned1.01.0
FixedRotation fixed, free to sway1.01.2
FixedFree2.02.0
PinnedRotation fixed, free to sway2.02.0

Where the nomograph stops paying

That 8 percent only counted because the strong axis governed. Move the weak-axis struts from third points to a single strut at mid-height and the weak-axis length goes to 3.0 m. Its slenderness climbs to 58.8 and Cr drops to 1950 kN, below either strong-axis value.

Now the weak axis governs, and the column fails at 2150 kN whatever K you pick on the strong axis. So check the slenderness on both axes before you refine either one.

The refinement also fades in stocky columns. Below a slenderness of about 30, Cr sits close to the section’s full factored yield capacity, and the difference between K = 0.85 and 1.0 is about 2 percent.

The G.3 equation assumes every column in that part of the frame reaches its critical load at the same time, which G.2 calls conservative. A base whose fixity you can’t pin down, a stepped crane column or a spring support doesn’t fit the nomograph. For those, a buckling analysis in your frame software beats a guessed G, and running the frame with both pinned and fixed bases shows whether the base guess matters. Annex G is informative, so a reviewer may ask about your G, and the slenderness can’t exceed 200 whatever K you use (Cl. 10.4.2.1).

So in a braced frame, the girders you’ve already paid for can buy back the last 5 to 10 percent of a slender column’s capacity, but only on the axis that governs. That stops holding once the column carries real moment, because the overall member check then uses K = 1.0. In this example that’s 2150 kN against 2050 kN, a fail before the moment term is even added, and pattern live load with one span empty is the usual source. The Cl. 13.8.2 check on a column with base moment shows that side, and how much rotational restraint a base or girder connection gives decides whether your G is credible. Then run your own column with its K on each axis .

Topics
Share LinkedIn X

Disclaimer: This blog post is for informational purposes only and should not be taken as specific engineering advice. Always consult the latest edition of the National Building Code of Canada and relevant CSA standards for your projects.

Quick answers.

No. CSA S16-19 Annex G gives only the sidesway-prevented nomograph. For in-plane behaviour, Cl. 10.3.2 takes K = 1.0 provided P-Delta effects and notional loads are included in the frame analysis (Cl. 8.4), so sway shows up as amplified forces and moments rather than as a K above 1.

Annex G.4 suggests G = 10 for a column base that sits on but is not rigidly connected to the footing, and G = 1.0 for a base rigidly attached to a properly designed footing. Smaller values need an analysis to back them up.

CSA S16-19 Cl. 13.8.1 classifies a frame with bracing as braced when its sway stiffness is at least five times that of the same frame with only its moment connections. Otherwise it is unbraced.

Next up for you.

Guide to Joint Fixity in Structural Modelling

Move beyond pin or fixed assumptions in steel frame analysis: CSA S16 connection criteria and how to model semi-rigid joints.

One short email when a new post goes up. NBCC, CSA and the design checks behind them.