Frame Stiffness vs FEA

You’ve built a complex model, the render looks fantastic, and the software spits out pages of data. But deep down, there’s a nagging question: Can I trust these numbers?

With today’s software, it’s easy to become a software operator instead of a master modeler. An ineffective model gives a false sense of precision, hiding incorrect load paths or underestimating stresses. Our duty of care under bodies like PEO and EGBC demands we use these tools responsibly.

This post covers the fundamentals behind two engines in our digital toolbox: the Frame Stiffness Method (FSM) and the Finite Element Method (FEM). It shows when to trust your frame analysis implicitly and when to scrutinize your mesh.

The Frame Stiffness Method (FSM)

For most of us, FSM is the workhorse behind day-to-day analysis of beams, columns, and braces in programs like SAP2000, S-FRAME, and ETABS.

It works by idealizing our structure as a discrete system-a collection of 1D line elements (members) connected at discrete points (joints). The software builds a global stiffness matrix, solves for joint displacements under your applied loads, and then works backward to find the forces in each member.

But how does the software know how a member behaves between the joints? That’s where shape functions come in.

Shape Functions

Your software assumes every beam-column element deforms according to a prescribed mathematical recipe, its shape functions. For a standard 2D beam-column element, these are:

1. Axial Displacement: Assumed to be linear from one end to the other.
2. Transverse (Bending) Displacement: Assumed to follow a cubic polynomial (\(w(x) = C_0 + C_1x + C_2x^2 + C_3x^3\)).

This is the most important concept to grasp, because that cubic polynomial is special. It exactly describes shear and moment in a segment loaded only at its ends. The second derivative (related to moment) is linear. The third derivative (related to shear) is constant.

Why this matters: FSM provides the theoretically exact solution (relative to engineering beam theory) for a prismatic member subjected only to concentrated loads and moments at its joints. The shape function matches the real behaviour. Your software isn’t guessing-it’s giving you the answer for the idealized problem you’ve defined.

When Your Frame Model Gives You the Exact Answer

Say loads are applied only at the joints, like a typical truss, a braced frame with loads at floor levels, or a moment frame with column point loads. Your analysis is precise.

Common Mistake: Adding unnecessary nodes to “refine” the model. A classic case is a buried pipeline with measured settlements imposed at 20-meter intervals. Concerned about the spacing, the engineer added nodes between the measurement points and was surprised the results didn’t change. They mistook this for convergence. The software already had the exact solution for those point loads, so intermediate nodes change nothing. The real error was modelling a continuous soil load as widely spaced point loads.

For distributed member loads, your software uses superposition. It calculates the fixed-end forces, applies the opposite to the joints, solves the “exact” point-loaded global model, and adds the results. For standard load types, this process is also exact.

The Finite Element Method (FEM)

FSM is fantastic for skeletal structures. For shear walls, slabs, complex connections, or foundations, we model a continuum, and that’s the domain of the Finite Element Method.

FEM (or FEA for analysis) works by breaking a continuous surface or solid into a collection of small, simple pieces called “finite elements” (like triangles or quadrilaterals) connected at nodes.

FEM is (Almost) Always an Approximation

This is the fundamental shift: an FEA solution is an approximation. Its accuracy depends on two things:

1. The shape functions of the elements you choose.
2. The size and arrangement of those elements (your mesh).

Instead of FSM’s cubic polynomial, a basic element like a Constant Strain Triangle (CST) uses a linear polynomial. That assumes strain (and therefore stress) is constant across the element. It’s a poor choice for bending unless you use an incredibly fine mesh.

Mesh rule of thumb: Your FEA solution is only as good as your mesh. You must check that your results have converged. For aspect ratios and density, check out our guide on effective finite element meshing.

A Practical Guide to Getting Good FEA Results

If FEA is an approximation, how do we know it’s good enough? The answer is convergence.

Always Check for Convergence

Convergence is refining your mesh until the solution stops changing significantly. It’s non-negotiable for any serious FEA.

Here’s a simple workflow:

1. Build your model with a reasonably coarse mesh.
2. Run the analysis and record a key result you care about (e.g., maximum deflection, or peak stress at a corner).
3. Refine the mesh. A good rule of thumb is to halve the element size in the area of interest.
4. Run the analysis again and compare the result.
5. Repeat until the percentage change in your result is small enough for your engineering judgment.

Displacements Converge Faster Than Stresses

Different results converge at different rates, and this trips up many engineers. From fastest and most accurate to slowest:

1. Displacements
2. Strains & Stresses (which are derived from the derivatives of the displacement field)

Watch for: Converged deflections don’t mean converged stresses. A mesh that accurately predicts building drift -a critical serviceability limit state- might be dangerously coarse for the peak stresses needed to design reinforcement around an opening in a shear wall. If stress drives your design (as per CSA A23.3, for example), you need a much finer mesh in that region.

Meshing Best Practices & Common Pitfalls

• Aspect Ratios: Keep your elements as square-like as possible. Aim for an aspect ratio (longest side / shortest side) of less than 3:1. Long, skinny elements give biased, inaccurate results.
• Stress Gradients: Use a finer mesh where you expect stresses to change rapidly: re-entrant corners, around openings, near point loads, and at supports.
• Model Consistently: An engineer modeled a curved concrete silo wall with flat shell elements (a reasonable simplification) but applied the true, continuous radial pressure. That inconsistency introduced large, artificial bending moments at the facets and a 200%+ error in stress. A consistent model would have simplified the load into point forces at the nodes.

Common pitfall: Don’t Chase Singularities. A sharp re-entrant corner (like the inside corner of an L-shaped wall opening) is a stress singularity. Mathematically, the stress there is infinite, so refining the mesh at that point keeps raising the value and it never “converges”. That is an artifact of the model, since real materials would yield and redistribute the stress. Focus on the stress a small, realistic distance from the corner.

Advanced Considerations in Canadian Practice

Handling Concrete Cracking (CSA A23.3)

A linear elastic analysis of a concrete structure using gross section properties overestimates stiffness. That underestimates deflections and misdistributes moments. The note to CSA A23.3-24 Clause 9.5.1 says a finite element analysis should account for the effects of cracking.

In Practice vs. In Code: A full nonlinear analysis is rarely practical for building design, but the code demands we do better than gross properties. Applying stiffness modifiers to your members and shells (e.g., \(0.35I_g\) for beams, \(0.70I_g\) for columns, \(0.25I_g\) for slabs) gives a more realistic distribution of forces and a better estimate of deflections.

Including Shear Deformations (CSA S16)

Standard FSM beam elements (Euler-Bernoulli theory) neglect shear deformations. That’s fine for slender members, but for deep beams, link beams, or plate girders, shear can be a significant part of the deflection. Most software can include it (Timoshenko beam theory). For steel members, CSA S16 provides guidance on calculating the required shear area (\(A_v\)).

P-Δ Effects

First-order analysis neglects secondary moments from axial loads on a displaced structure (P-Δ and P-δ effects). CSA S16 mandates that stability effects be considered. A direct second-order analysis is the preferred and more reliable method, and it’s available in modern software. Don’t turn it off without a justified stability assessment.

Choosing the Right Modelling Approach

Build the simplest possible model that accurately captures the essential structural behaviour. That means actively challenging your model’s assumptions to gain real confidence in its predictions.

• For typical frame structures, the Frame Stiffness Method is efficient and often theoretically exact for the member forces you need for design. Trust it.
• For walls, slabs, and complex geometries, the Finite Element Method is a powerful, necessary tool. But it’s an approximation that demands your engineering judgment to verify its accuracy through careful meshing and convergence checks.

Your Next Steps:

1. The next time you model a simple frame, take a moment to identify why FSM is giving you an exact solution for the joint-loaded case.
2. On your next FEA model, perform a quick convergence study on one critical area. Halve the mesh size and see how much your peak stress changes. Is it what you expected?
3. Pull up a past concrete model and check if you accounted for cracking in your stiffness assumptions as required by CSA A23.3.