Importing the CAD geometry, auto-meshing everything and hitting run is how a model ends up giving some spectacularly wrong answers. The temptation is real with a complex structure, a looming deadline, and powerful analysis software.
A systematic approach works on a simple frame or a complex shell structure.
Why Your Fancy Model Might Be Failing You
Modern software can handle thousands of elements without breaking a sweat. But here’s what those colourful stress contours won’t tell you: whether your model actually represents how the structure carries load.
An effective structural model does three things well:
Accurately represents the primary load path. If loads don’t flow through your model the way they’ll flow through the real structure, nothing else matters.
Provides sufficiently accurate results for design. Not exact results-sufficiently accurate.
Gives you insight into structural behaviour. It comes from understanding how the structure works.
Key Takeaway: A structure has never collapsed because design stresses were 5-10% off. Structures fail because critical load path elements weren’t recognized in the first place.
Step 1: Understand the Structural Form Before You Model It
Before you open any software, you need to answer a fundamental question: How does this structure carry load?
If you’re analyzing a shell structure, understanding shell theory lets you choose the right elements and spot the bending regions. If it’s a frame with shear walls, you need to know whether flexure or shear dominates the response. This isn’t academic-it determines what kind of elements you need, where to refine your mesh, and what results to check.
Consider a cylindrical silo under internal pressure. Away from the supports, it’s essentially pure membrane behaviour-hoop tension with no bending. Near the base, where fixity creates restraint, you’ll see significant bending that dies out within a characteristic “decay length.”
If you don’t understand this going in, you’ll either:
- Use shell elements where membrane elements would suffice (wasting effort)
- Miss the critical stress gradients near the base entirely (dangerous)
- Mesh uniformly when you should be refining locally (inefficient)
Step 2: Identify the Load Path-All of It
Every load applied to your structure needs a path to the ground. Your model needs to capture that path, or you’re not modelling reality.
For gravity loads in a building, this is usually straightforward: slab to beam to column to foundation. But what about lateral loads? How do forces transfer through floor diaphragms? Are your connections stiff enough to develop the moment resistance you’re assuming?
The load path question becomes critical when you’re dealing with:
- Lateral force-resisting systems where diaphragm action transfers loads to shear walls or frames
- Transfer structures where loads change direction or skip members (often requiring checks for shear deformation in deep members)
- Complex connections where forces must pass between different structural systems
Sketch the load path by hand before you build the model. If you can’t trace how gravity and lateral loads get to the foundation on paper, your computer model won’t figure it out for you.
Common Mistake: Modelling a diaphragm as rigid without checking whether that assumption is valid. Flexible diaphragms distribute loads very differently than rigid ones.
Step 3: Get Your Boundary Conditions Right
Boundary conditions might be the most consequential assumptions you’ll make. The difference between a fixed base and a pinned base can completely change your frame’s behaviour-moment distribution, lateral stiffness, drift, everything.
Real boundary conditions are almost never perfectly fixed or perfectly pinned. That column base you’re modelling as fixed? It has some rotation capacity. Your rigid foundation has some flexibility too.
What matters is whether they’re close enough, and how sensitive your answers are to them.
Practical guidance:
- Column bases: Unless you have a massive pedestal with significant embedment, “fixed” is probably optimistic. Consider running cases with different fixity assumptions.
- Foundation stiffness: For spread footings on competent soil, you can often assume rigid for gravity loads. For lateral analysis of taller structures, foundation flexibility can matter.
- Connections: If your frame analysis assumes rigid beam-column connections, make sure your connection design can actually deliver that rigidity.
Run the model with different boundary condition assumptions. If results are highly sensitive to that choice, refine your assumption or design conservatively for both cases.
Step 4: Know Your Model’s Underlying Assumptions
Every analysis method comes with assumptions baked in. Classical beam theory assumes plane sections remain plane. Shell elements assume certain thickness-to-span ratios. The stiffness method assumes linear-elastic behaviour.
Software inherits these assumptions whether you think about them or not.
Common assumption pitfalls:
Geometric simplification without consistent loading: If you simplify a circular cross-section to a polygon (say, for easier meshing), you need to simplify your loads the same way. Applying radial pressure to a polygonal section creates artificial bending moments at every corner-moments that don’t exist in the real circular section.
Element type mismatch: Using shell elements when membrane elements would suffice isn’t just inefficient-it can introduce spurious bending stresses. Conversely, using membrane elements when bending is significant will miss real stresses entirely.
Ignoring second-order effects: For slender structures under significant axial load, P-delta effects can substantially increase moments. Linear analysis won’t capture this.
Document your assumptions explicitly. When you hand off a model or return to it months later, you need to know what simplifications you made and why.
Step 5: Build Incrementally and Validate Continuously
Build the model in stages, each validated against your engineering judgment.
Start with hand calculations. For a simple frame, you can estimate base shears and overturning moments with a free-body diagram. For a continuous beam, you can sketch approximate moment diagrams. These become your reality check.
Then build the simplest computer model that captures essential behaviour. For a multi-storey building, this might be a stick model with lumped masses. For a shear wall building, it might be a 2D analysis of a representative wall line.
Add complexity only when needed. As you refine the model-adding 3D effects, diaphragm flexibility, foundation springs-each addition should make physical sense. If adding a feature dramatically changes your results, you need to understand why.
Check results at every stage. Can you explain why moments increased when you added that element? Does the deflected shape make physical sense? Are reactions in equilibrium with applied loads?
Key Takeaway: When something changes unexpectedly, you can trace the cause because you’ve added one thing at a time.
Sensitivity Analysis: Your Safety Net
Running a single analysis gives you a single answer. Sensitivity analysis tells you how much to trust that answer.
At minimum, investigate sensitivity to:
- Boundary conditions: What if that base isn’t as fixed as you assumed?
- Member stiffness: What if concrete cracks more than you expected? (Needed to satisfy CSA A23.3 stiffness requirements).
- Connection rigidity: What if that moment connection has some rotation?
- Load distribution: What if the actual load pattern differs from your assumption?
If your design is sensitive to any of these-meaning small changes in input cause large changes in output-you have two options. Refine your input (get better information). Or design conservatively for the range of possible outcomes.
Sensitivity analysis also validates your model. A well-behaved model should respond proportionally to reasonable changes in input. If changing foundation stiffness by 20% causes forces to jump by 200%, something is wrong-either with your model or with your understanding of the structure.
Documentation: Because You’ll Forget
Six months from now, you won’t remember why you modeled that connection as pinned or why you used a 200 mm mesh in one area and 500 mm in another.
Essential documentation includes:
- Sketches of the model concept showing member types, boundary conditions, and load paths
- List of assumptions with justification for each
- Material and section properties with sources
- Validation checks showing comparison with hand calculations or simpler models
- Sensitivity study results with conclusions about model robustness
This is how you and your reviewers verify the model suits its purpose.
Pro-Tip: Include a “model limitations” section that explicitly states what the model doesn’t capture. This prevents misuse and shows you’ve thought carefully about scope.
What Separates Effective Models from Dangerous Ones
It’s not element count. You can have 50,000 elements and still be wrong if you’ve missed the primary load path.
It’s not visual polish. Beautiful renderings mean nothing if boundary conditions are unrealistic.
It’s not software sophistication. A hand calculation that captures real behaviour beats a fancy FEA model that doesn’t.
Putting It Into Practice
Sketch first. Draw the structure, mark the load paths, note your boundary condition assumptions.
Estimate key results by hand.
Build the simplest model that captures essential behaviour. Validate it against your hand calculations.
Add complexity incrementally. Understand what each addition changes and why.
Run sensitivity studies. Know which inputs your design depends on most.
Document everything.
The goal of structural modelling isn’t to produce impressive output. It’s to understand how your structure behaves well enough to design it safely. Start simple, build systematically, check continuously, and never trust a result you couldn’t have estimated first.