steel-structure-torsion-design
Steel Structure Torsion Design: Warping & Open Sections

A 3D line drawing of an open I-section steel girder on a white background, exaggerated to show warping distortion of the flanges under torque, with red torque arrows and a warping normal-stress sketch, in a clean engineering teaching style.
Most steel design is taught as if every beam bends about its strong axis and every column goes straight down. Real buildings twist—when a crane runs off-center, a monosymmetric rafter carries an eccentric canopy load, or an open-web joist frame sways. Steel structure torsion design is the chapter engineers skip until a member shows diagonal cracks or a column twists at the base. It is the difference between assuming pure bending and asking what happens when the load does not hit the shear center.
This guide walks through the two torsion theories (Saint-Venant free torsion versus constrained warping torsion), why open I-sections are weak in torsion, what the shear center means for eccentric loads, when torsional or torsional-flexural buckling controls, and the practical fixes. Overall stability and lateral-torsional buckling are covered in the stability article, and connection forces in the connection article—see steel structure stability design and steel structure connection design. This one is about torsion as its own load path, the core of steel structure torsion design.
When Does a Steel Member Actually Twist?
Torsion enters a steel frame through three doors. First, a load that does not pass through the shear center—a crane wheel set off the web centerline, a canopy or eyebrow rafter loaded off its top, a duct or pipe hung from one side of a purlin—becomes a torque T = P·e. Second, inconsistent end restraint: a column fixed at the base and pinned at the top cannot warp freely, so any twist is locked into normal stress. Third, singly symmetric sections—a channel (C), a T, a single angle—have their shear center off the centroid, so even a load "on the centroid" is automatically eccentric.
Torsion is a coupled problem, not a simple twist. The rotation angle θ(x), the warping displacement of the cross-section, and the warping normal stress σ_ω all travel together along the member length. That coupling is exactly why hand calcs on a bent beam miss it: the software quietly assumes every load is concentric. Thin-walled members are the most exposed case—see steel thin-walled member design—but even ordinary hot-rolled I-girders twist when the eccentricity is large enough. The reason torsion is overlooked is practical: closed tubes are naturally stiff in torsion, open I-sections have almost no Saint-Venant stiffness, and ordinary portal software hides warping stiffness by default.
Two Torsion Theories — Saint-Venant vs Warping
There are two ways a section can resist twist, and a real member uses a blend of both.
Saint-Venant (free) torsion is the textbook case: the cross-section is free to warp (to distort in its own plane) as it twists, so it develops only shear stress—no axial normal stress. The torque is carried entirely by the Saint-Venant torsion constant J: T = G·J·θ′. Here G is the shear modulus and θ′ the rate of twist. For a closed hollow section J is large; for an open thin rectangle, J ≈ (1/3)·Σ b·t³—proportional to t cubed, which is why a thin open section is so weak.
Warping (constrained) torsion is the reality at connected ends. A stiffener, end plate, or embedded base stops the section from warping, so the twist now develops axial normal stress across the section—a "double moment" B—and an additional warping stiffness E·Cw (the warping constant). For an open I-girder this warping term is significant, and the total torque is T = GJ·θ′ − ECw·θ′′′. At a fully fixed end the warping term dominates; at a free end it vanishes. This is why end detail drives torsion response far more than it drives bending response—a core lesson of steel structure torsion design.
The section shape decides which theory dominates. A closed tube (circular or rectangular hollow section) has a large J and small warping contribution: it twists like a torsion spring, cheaply. An open I-section has a tiny J, so almost all the torque is resisted by warping stiffness, which demands a warping check at the ends. Material grade does not rescue a bad section here—compare Q235 vs Q355 steel strength for bending, but torsion stiffness comes from geometry.
| Section Type | J (torsion const., relative) | Cw (warping const., relative) | Typical Use |
|---|---|---|---|
| Closed round tube | High | Low | Poles, masts, torsion-critical struts |
| Closed rectangular tube | High | Low-ish | Columns, canopy posts |
| Open I / H-section | Very low (∝ t³) | Significant | Beams; torsion needs checking |
| Channel / C-section | Very low | Moderate | Purlins, girts (check TFB) |
Relative values for comparison; absolute J and Cw must be computed for your section (software or AISC tables). See AISC Design Guide 09: Torsional Analysis.
| Aspect | Saint-Venant (Free) Torsion | Constrained (Warping) Torsion |
|---|---|---|
| Cross-section warp | Free | Restrained at ends |
| Stresses | Shear only | Shear + axial (warping normal) |
| Stiffness term | G·J | E·Cw added |
| Where dominant | Long closed sections | Fixed-ended open sections |
| Hand calc? | Simple T = GJθ′ | Needs software / closed-form |
Free torsion produces shear only; warping torsion adds normal stress and a double moment. Open thin-walled members usually need the constrained check. See AISC Steel Construction Manual.
Shear Center & Eccentric Loads
The shear center (center of twist) is the point through which a transverse load can be applied without causing any twist. On a doubly symmetric I-section the shear center coincides with the centroid, so a load on the web centerline bends without twisting. On a channel, T-section, or any singly symmetric shape, the shear center lies outside the section—out beyond the web of a C-channel. Any load applied away from the shear center by eccentricity e applies torque T = P·e, in addition to the bending moment.
In practice the eccentricities accumulate from routine details. A crane rail installed a few millimeters off the web centerline plus a single-side maintenance walkway creates a continuous torque along every crane girder—see overhead crane steel building. A canopy or eyebrow roof loads a rafter off its top flange. A purlin hung with a single-sided duct or a light fixture is loaded off its shear center. A curtain-wall mullion carries glass pressure off its centroid. Each is small; together they twist the member that the model only bends. The deflection consequences are visible as excess lateral sway—see steel structure deflection control.
The remedy sequence is: first move the load back to the shear center (re-center the rail, symmetrically hang services, put the canopy load on a dedicated column). Second, if eccentricity cannot be avoided, run a constrained-torsion check for warping stress and double moment at the ends. Third, brace or stiffen as needed.
| Load Source | Typical Eccentricity | Resulting Action | Mitigation |
|---|---|---|---|
| Crane rail offset from web | 15–25 mm (0.6–1 in) | Torsion + horizontal load | Re-center rail; add braking truss |
| Canopy / eyebrow roof | Load off rafter top | Torsional moment + twist | Dedicated canopy columns |
| Single-side walkway / duct | Off purlin shear center | Torsion on purlin/girt | Symmetric hanging; brace |
| Curtain-wall mullion | Glass pressure off centroid | Torsion on mullion | Hollow section; check Cw |
Typical eccentricities by detail; exact torque = P times e. Consult our engineers for your load list.
Is Your Frame Being Designed As If Every Load Hits the Shear Center?
Crane rails, canopy loads, and asymmetric purlin hangings quietly twist members that the base model only bends. Send us your eccentric load list, and our engineers will add warping torsion checks, not just a thicker web.
Torsional Buckling & Torsional-Flexural Buckling
Torsion is not only a stress problem—it can also trigger a buckling mode. A compression member can buckle by twisting about its own longitudinal axis at a load below its bending buckling load. Torsional buckling (TB) is pure twist; torsional-flexural buckling (TFB) is bend-and-twist coupled together, common on singly symmetric thin-walled sections.
Which mode governs depends on the section. Doubly symmetric hot-rolled H-columns usually fail in flexural buckling, not torsional buckling—unless end warping restraint is weak, or the section is very thin. Singly symmetric channels, single-angle struts, and double-angle T-struts are genuinely at risk: their shear center off the centroid couples twist into any compression. Cold-formed purlins and wall girts are checked for TFB under the AISI system. A real example: a 5 t overhead crane workshop where the original model treated the crane girder as simply supported; a second look found the rail gauge offset 25 mm (1 in) from the web centerline plus a single-side maintenance walkway, giving a continuous torque the hand calc missed. Adding a top-chord braking truss and centering the rail cured the twisting at roughly 3% of frame steel weight—a cheap fix once steel structure torsion design is taken seriously.
The analysis point: do not confuse lateral-torsional buckling (LTB) of a bent beam—covered in the stability guide—with torsional or torsional-flexural buckling of a compression member. General finite-element software handles both; portal design software often needs a manual check for open thin-walled columns, which also interacts with P-Δ effects in steel structure second-order analysis. Purlins are checked for this in steel purlin system design.
| Cross-Section | Governing Buckling Mode | Checked By |
|---|---|---|
| Doubly symmetric H column | Flexural buckling (usually) | AISC Chapter E |
| Channel / single-angle strut | Torsional-flexural buckling | AISC E + TFB formula |
| Cold-formed purlin / girt | Torsional-flexural buckling | AISI S100 |
| Bent I-beam | Lateral-torsional buckling | AISC Chapter F |
| Thin-walled open column | Torsional buckling possible | FE / Design Guide 09 |
Mode selection by section symmetry; exact critical loads per AISC Specification Chapter E. consult our engineers for open thin-walled columns.
Torsional Bracing & Practical Fixes
The torsion fix hierarchy is the same as the stress-reduction one: make the load concentric, brace the twist, or change the section.
To give ends warping restraint, add transverse stiffeners, use a rigid end-plate moment connection, or embed the column base—each stops the section from warping and activates the warping stiffness that open sections rely on. Intermediate lateral bracing shortens the unbraced torsional length and raises the critical load. The single most effective change is also the simplest: replace an open I-section with a closed round or square tube, which has a large J and resists torsion with almost no added stress.
Field retrofits follow the same logic. A cracked crane girder gets a top-chord braking truss that transfers the eccentric wheel torque away from the girder. A single-angle brace is replaced by a double-angle T or a square tube. An eccentric canopy gets its own posts so the portal column no longer carries combined bending plus torsion. Bracing strategy is detailed in steel building bracing system, and crane-building specifics in overhead crane steel building. Bring in a structural engineer when open thin-walled columns are tall relative to their width, eccentric loads exceed about 10% of axial force, or crane capacity is above roughly 10 t.
Cost & When Torsion Matters
Torsion checking itself costs engineering hours, not steel. The material levers are cheap relative to failure: switching an open H-section to a closed tube typically adds on the order of 10%–30% per section (verify with current pricing), and adding a braking truss or torsional bracing adds about 2%–5% of frame steel weight. The projects where torsion actually controls are heavy-crane workshops, cantilevered canopies, curtain-wall mullions, cold-formed wall girts, and open lattice towers. A symmetric single-bay portal warehouse rarely has torsion in the governing load case. Cost structure is broken down in steel building quote breakdown, and section choice by weight class in light steel vs heavy steel structure.
Conclusion
Steel structure torsion design reduces to one cause and one geometry: a load off the shear center, constrained at the ends, on a section that may be weak in twist. Open I-sections have almost no Saint-Venant stiffness and rely on warping; closed tubes shrug off torsion; torsional-flexural buckling is most dangerous on singly symmetric thin-walled struts. Listing eccentric loads and writing end warping restraint into the connection drawings at the design stage is an order of magnitude cheaper than reinforcing a twisted member in service. Make steel structure torsion design a pre-fabrication check, not a post-crack repair.
Tired of Guessing Whether Your Frame Twists?
We model torsion the way it actually happens—eccentric crane loads, asymmetric canopies, and warping constraints at the ends—instead of assuming pure bending. Tell us your eccentric loads.
🏭 Explore: Steel Workshop · Steel Factory
Case Example
A heavy-crane workshop built for a Southeast Asian manufacturer, 7,800 m2 (about 84,000 ft2) on 24 m (79 ft) bays, developed a twist problem when an off-center 20 t (22 tonne) crane rail met an asymmetric canopy on monosymmetric rafters. Hand calcs missed it until a girder showed diagonal cracking. The fix switched the torsion-critical rafters from open H-sections to rectangular hollow sections and added a horizontal braking truss at roof level. Peak twist rotation fell from 0.018 rad to 0.004 rad, and the cracking did not propagate. Adding the braking truss cost about 3% of frame steel weight, against roughly 25% to upsize every rafter. The eccentric-load cause and end warping restraint are the levers; see overhead crane steel buildings and steel structure connection design for how torsion forces enter the joints.
Reference Links
- AISC 360 Specification for Structural Steel Buildings
- ASCE 7 Minimum Design Loads and Associated Criteria for Buildings and Other Structures
- ISO 12944 Corrosion protection of steel structures by protective paint systems
About the Author
Senior Structural Engineer
With over 20 years of hands-on experience in steel structure design and prefabricated building engineering, our in-house senior structural engineer has personally contributed to more than 500 steel building projects—including warehouses, industrial factories, aircraft hangars, agricultural buildings, and commercial structures. The focus is on translating design codes such as AISC 360, ASCE 7, and Eurocode 3 into buildable, cost-effective steel solutions that balance structural performance, fabrication efficiency, and total project cost.
Learn more about our engineering team
Frequently Asked Questions
What is the difference between Saint-Venant and warping torsion?
Saint-Venant (free) torsion lets the cross-section warp freely, so it produces only shear stress. Warping (constrained) torsion happens when ends are fixed by stiffeners or end plates and cannot warp, adding normal stresses and a "double moment." Open thin-walled members usually need the constrained-torsion check.
Why are open I-sections bad in torsion?
Their Saint-Venant torsion constant J is very small (roughly J ≈ ⅓·Σ b·t³ for thin rectangles), so most of the load must be carried by warping stiffness. A closed hollow section (tube) has a far larger J and resists torsion with little added stress.
What is the shear center and why does it matter?
The shear center is the point through which a transverse load can be applied without causing torsion. On a channel or monosymmetric section it lies outside the section. If your crane rail, canopy, or hanging load sits off that point, every load becomes a torque T = P·e.
What is torsional-flexural buckling?
It is a failure mode where a compression member buckles by bending and twisting at the same time, common on singly symmetric thin-walled sections (channels, single-angle struts). Hot-rolled doubly symmetric H-columns rarely control this way, but cold-formed purlins and wall girts must be checked.
How do I fix torsion in an existing steel beam?
First try to move the load back to the shear center. If that is not possible, add a braking truss or lateral torsional bracing, switch to a closed tube section, or provide end stiffeners that constrain warping. A 2%–5% frame steel increase usually solves it—far cheaper than post-crack repair.
steel-structure-vibration-control
steel-member-local-stability