The result that looks wrong
Here is a C12X20.7 modelled in Section Designer exactly as it is rolled, with flanges sloped 2 in 12, root fillets and rounded flange tips, with every result compared with the AISC Steel Construction Manual, 16th Edition, Table 1-5.
| Property | AISC Manual | Real shape, Section Designer | Difference |
|---|---|---|---|
| Area, A | 6.08 in² | 6.079 in² | −0.02% |
| Ix | 129 in⁴ | 129.0 in⁴ | +0.01% |
| Iy | 3.86 in⁴ | 3.862 in⁴ | +0.06% |
| Sx / Zx | 21.5 / 25.6 in³ | 21.50 / 25.58 in³ | within 0.1% |
| Sy / Zy | 1.72 / 3.47 in³ | 1.721 / 3.461 in³ | within 0.3% |
| rx / ry | 4.61 / 0.797 in | 4.607 / 0.797 in | within 0.1% |
| Torsion constant, J | 0.369 in⁴ | 0.375 in⁴ | +1.7% |
| Shear centre behind the web, e0 | 0.870 in | 0.690 in | −21% |
| Warping constant, Cw | 112 in⁶ | 87.3 in⁶ | −22% |
Nine properties agree to a fraction of a percent. Two are a fifth out. Is the software wrong?
What Cw and e0 measure
When an open section such as a channel twists, its cross-section does not stay flat. The flanges bend in their own plane in opposite directions, so points along the section move along the member's length by different amounts. This out-of-plane movement is warping. If the supports or the rest of the member stop the section warping freely, the flanges carry extra bending stresses and the member is stiffer in twist. The warping constant Cw measures how much: it sums the square of the warping displacement over the area, and the flange tips, which move the most, count by far the most.
The shear centre is the point a transverse load must pass through for the section to bend without twisting. For a channel it lies outside the section, behind the web, at a distance e0. A load applied through the web is off the shear centre, which is why channels twist under it unless they are braced.
Both values feed the lateral-torsional and flexural-torsional buckling checks, so the difference is worth understanding.
How AISC works out Cw and e0
For channels, the Manual takes A, I, S, Z and r from the real rolled shape, but e0 and Cw from thin-walled open-section theory. The section becomes three flat plates along their centrelines, with the flanges at their average thickness. For the C12X20.7, with d = 12 in, bf = 2.94 in, tw = 0.282 in and tf = 0.501 in:
Centreline dimensions. b′ = bf − tw / 2 = 2.799 in and h′ = d − tf = 11.499 in.
Shear centre. e0 = 3 b′² tf / (6 b′ tf + h′ tw) − tw / 2 = 1.010 − 0.141 = 0.869 in (AISC: 0.870).
Warping constant. Cw = (tf b′³ h′² / 12) (3 b′ tf + 2 h′ tw) / (6 b′ tf + h′ tw) = 111 in⁶ (AISC: 112).
Both formulas land on the table values, which shows how the table was made. The last 1% most likely comes from AISC working with unrounded dimensions.
Why the real shape warps less
A rolled channel's flanges are not flat. They slope, so there is more steel near the web and less at the tips, even though the average thickness is the same 0.501 in. The tips are exactly where warping displacements are largest, and Cw grows with the square of them, so taking steel away from the tips lowers Cw sharply: from 112 to about 87 in⁶, 22% less.
The same shift moves the shear centre. With less steel out at the tips, the flanges carry a smaller share of the shear flow, and the shear centre moves towards the web: 0.690 in instead of 0.870 in.
An independent check
To make sure this is not a quirk of one program, the same outline was solved with sectionproperties, the open-source finite element section solver many engineers use:
| Property | AISC (thin-walled) | sectionproperties | Section Designer |
|---|---|---|---|
| e0 | 0.870 in | 0.690 in | 0.690 in |
| Cw | 112 in⁶ | 87.37 in⁶ | 87.32 in⁶ |
The two solvers agree within 0.06%. As a check on the method, both give the thin-walled answers when the channel's walls are made very thin (1 mm), which is where those formulas hold.
Model B: the shape the formulas assume
If the difference really comes from the flange slope, a channel with flat flanges of the average thickness should reproduce the table's Cw. It does, and it shows the other side of the trade:
| Property | AISC Manual | Flat flanges, Section Designer | Difference |
|---|---|---|---|
| Cw | 112 in⁶ | 112.4 in⁶ | +0.4% |
| J | 0.369 in⁴ | 0.359 in⁴ | −2.7% |
| e0 | 0.870 in | 0.813 in | −6.6% |
| A | 6.08 in² | 6.217 in² | +2.3% |
| Ix | 129 in⁴ | 133.6 in⁴ | +3.6% |
| Iy | 3.86 in⁴ | 4.835 in⁴ | +25% |
Cw now matches, but A, Ix and especially Iy move away, because this shape has extra steel at the flange tips that the real channel does not. e0 still falls a little short of 0.870 in, because the thin-walled value is measured on centrelines and a solid model of any thickness lands slightly inside it. sectionproperties again gives the same numbers: Cw = 112.4 in⁶ and e0 = 0.813 in.
What this means in practice
- No single model matches every AISC value for a channel, because the Manual works the properties out with different methods. That is not a modelling error.
- For the real section's behaviour, such as warping stresses or a detailed torsion analysis, Cw ≈ 87 in⁶ and e0 ≈ 0.69 in from the real shape are the more accurate values.
- For AISC 360 checks of lateral-torsional and flexural-torsional buckling, use the tabulated J = 0.369 in⁴ and Cw = 112 in⁶. The Specification is applied with the tabulated values. Model B reproduces that Cw if you need it from the app.
- Check what your software reports. A finite element section solver gives the real-shape value. If it does not match the table, look at the shape it solved before you suspect the numbers.
Section Designer always solves the solid section by finite elements. It has no thin-walled mode, so a real-shape model gives about 87 in⁶ and only a flat-flange model gives about 112 in⁶. Expect the same kind of gap on other rolled channels with sloped flanges, and check the shape before you compare.
Try it yourself
Both models are Example 4 on the Section Designer verified examples page, with every value compared and both models to download or open in the app. Section Designer works out J, Cw and the shear centre by finite elements for any section you build, so you can try the same comparison on your own channels.