Module 3 · Lesson 3.1
From building to model
Every modelling decision throws something away. The skill is knowing what, and where it gets picked up again.
Why this matters
Open an analysis package and you can build something that looks convincingly like the building: members with real section shapes, rendered in three dimensions, rotating on screen. It is easy to believe that what you are looking at behaves like the structure. It does not. Almost everything that governs a steel design — local buckling, warping, joint flexibility, connection eccentricity — is invisible to that model, and has to be handled somewhere else.
By the end of this lesson you should be able to
- Trace a load from cladding to foundation
- Separate the physical member from the analytical element
- State what a beam element cannot represent
- Decide when a sub-model is needed
What you should already know
- Structural forms and load paths (Structural Analysis Fundamentals, Module 1)
- Actions and combinations (Module 2)
- Section properties, and the idea of a shear centre
The load path in a single-storey steel building
A steel building is unusually legible: you can normally stand inside one and see the entire load path.
Vertical load — snow, services, self-weight:
sheeting → purlin → rafter → column → base → foundation
Each step is a change of member and usually a change of direction. The sheeting spans a metre or two; the purlin spans a bay; the rafter spans the building; the column takes it to the ground.
Horizontal load — wind on the side of the building:
cladding → side rail → column → frame action or bracing → base
Wind on the gable, which is the one people forget:
gable cladding → gable post → roof bracing → eaves → vertical bracing → base
That third path is why a building needs bracing in the roof plane as well as in the walls. Wind on the end of the building arrives at roof level with nowhere to go unless there is something in the roof to carry it back to a braced bay.
A load path is not complete until it reaches the ground. A path that stops at 'the roof structure' has not been traced.
Try it
What does this model actually contain?
Element type decides what CAN be represented. Analysis type decides what is actually looked for. The two are independent, and getting either wrong quietly removes a failure mode from the design.
Element type
Analysis type
Beam element (6 degrees of freedom per node)
Almost always, for frames and members in bending and compression.
Captures
- Axial force
- Bending about both axes
- Shear
- St Venant torsion
Misses — must be handled elsewhere
- Local plate buckling — the element has no plates
- Warping torsion and bimoment, unless a 7th degree of freedom is provided
- Cross-section distortion
- The real stiffness of the joints at its ends
- Could this model represent local buckling?
- no — it has no plates
- Does this ANALYSIS look for buckling?
- no
- So is local buckling found here?
- NO
- Warping torsion represented?
- no — needs a 7th DOF
- Joint stiffness represented?
- only if you model it explicitly
- Where local buckling is actually handled
- cross-section classification, outside the analysis
This element has no plates, so local buckling cannot appear however the analysis is run. That is not a defect — it is why cross-section classification exists, and why it is done outside the analysis entirely.
Things worth trying
- Choose a beam element and a non-linear analysis. Local buckling still cannot appear — the element has no plates to buckle. No amount of analysis sophistication substitutes for classification.
- Choose shell elements and a linear analysis. Now the model COULD represent local buckling and the analysis does not look. This is the more expensive mistake, because it feels thorough.
- Switch to a truss element and read what it misses. Load applied between nodes produces bending the element cannot carry, so it is silently lost. Truss elements are for triangulated members loaded at their nodes.
- Tension-only elements make the analysis non-linear and give the structure a different stiffness in each direction of load — which surprises people reading deflection results.
- Note the last readout row. Whatever you choose, local buckling is handled by classification. The model is not where that question gets answered.
Physical member and analytical element
A physical beam is 8 m of steel with an end plate at each end, bolted to a column flange. The analytical element is a line between two nodes.
Three differences matter.
Length. The element usually runs centreline to centreline, so it is longer than the physical member — sometimes by half a column depth at each end. That is conservative for bending and unconservative for nothing, so it is generally left alone. But for a short member between two deep columns it can be a significant overstatement.
Position. The element runs along the centroidal axis. The real force may not. An angle bolted through one leg carries its load along the bolt line, which is offset from the centroid — so the real member has bending that the model never reported.
Ends. The element has whatever end condition you specified. The real connection has a rotational stiffness somewhere between zero and infinity, and the next lesson is about what happens when the two disagree.
The model is a hypothesis about the structure. Every hypothesis needs to be stated, and the ones that are never stated are the ones that cause trouble.
Worked example
The eccentricity a centreline model removes
Given
- An angle bracing member, 100 × 100 × 10, carrying 250 kN in tension
- Bolted through one leg only, with the bolt line 45 mm from the member's centroid
- Modelled as a truss element on the centroidal axis
Find
What the model reports, what the member actually sees, and where the difference goes.
Assumptions
- The bolt line is taken as the line of action of the force
- Eccentricity in one plane only, for clarity
Predict first
An engineer models a portal frame with shell elements instead of beam elements, and runs a linear analysis. Will the model now predict local buckling of the rafter web?
Practice
A bracing member carries 180 kN and is connected with the bolt line 38 mm from its centroid. What moment does that eccentricity produce, in kNm?
Practice
The same 38 mm eccentricity, but the force rises to 400 kN. What is the moment now, in kNm?
Practice
A beam is modelled centreline-to-centreline between two 300 mm deep columns. The clear span between column faces is 7.4 m. What span does the model use, in m?
Check yourself
Where is local buckling of a steel section accounted for in a normal frame design?
Summary
- Trace every load path to the GROUND — the gable path is the one usually missed
- A beam element carries axial, bending, shear and St Venant torsion, and nothing else
- It knows nothing about plates, so local buckling lives entirely in classification
- It knows nothing about warping unless a seventh degree of freedom is provided
- Element type decides what CAN appear; analysis type decides what is looked for
- Centreline modelling removes eccentricity that the real detail has
- Eccentricity moment scales with the force, so it does not fade as the member works harder
This is educational material. It uses simplified examples to teach principles, and must not be relied on for real design or safety-critical decisions. Module overview and checkpoint