FEA Best Practices for Structural Engineers

Finite Element Analysis is a powerful tool, but it’s also a tool that can produce convincing-looking wrong answers. After years of using FEA on projects ranging from residential buildings to F1 facilities, here are the practices that help me get reliable results.

The Cardinal Rule

Never trust a result you can’t verify by other means.

Before diving into mesh refinement and solver settings, remember: FEA is a tool to refine your understanding, not replace engineering judgment. Always have a back-of-envelope check ready.

Model Setup Fundamentals

Choosing the Right Element Type

For building structures, common choices:

Mesh Quality Matters

Poor mesh quality is the most common source of FEA errors. Guidelines:

Aspect ratio: Keep elements roughly square (ratio < 3:1)

Mesh refinement: Finer mesh near:

• Load application points
• Support locations
• Geometric discontinuities
• Areas of high stress gradient

Convergence check: Run the analysis with progressively finer meshes until results stabilize (within 5% typically acceptable).

Mesh study example:
Mesh 1: 500mm elements  → Peak stress: 245 MPa
Mesh 2: 250mm elements  → Peak stress: 267 MPa
Mesh 3: 125mm elements  → Peak stress: 278 MPa
Mesh 4: 62.5mm elements → Peak stress: 281 MPa

Mesh 3 to 4 change: 1.1% - convergence achieved

Boundary Conditions

The most critical and error-prone aspect of FEA modeling:

Supports

• Pinned support: Restrains translation, allows rotation
• Fixed support: Restrains all degrees of freedom
• Roller: Restrains perpendicular translation only

Reality check: Does your boundary condition represent actual behavior?

• “Fixed” columns on pad foundations aren’t truly fixed
• “Pinned” beam connections have some rotational stiffness
• Soil-structure interaction affects foundation behavior

Loading

Load application:

• Point loads: Create stress concentrations - use bearing plates
• Distributed loads: More realistic but check tributary areas
• Pattern loading: Required for slabs per EC2

Load combinations:

• Generate all relevant combinations before analysis
• Use envelope results for design
• Check that critical combination makes physical sense

Common Pitfalls

1. Stress Singularities

FEA will predict infinite stress at:

• Point loads on shells
• Re-entrant corners
• Support points

These aren’t real - they’re mathematical artifacts. Solutions:

• Apply loads over finite area
• Use fillet radii at corners
• Interpret results away from singularity
• Use stress averaging

2. Hourglassing

In reduced integration elements, hourglass modes can develop:

• Zero-energy deformation patterns
• Results look plausible but are wrong

Prevention:

• Use full integration elements
• Enable hourglass control
• Check for suspicious deformation patterns

3. Shear Locking

Thin elements with low-order shape functions can be overly stiff:

• Beam elements that don’t bend properly
• Shell elements that behave like rigid plates

Solutions:

• Use elements with shear deformation
• Increase polynomial order
• Use specialized thin element formulations

4. Overconfidence in Linear Analysis

Linear analysis assumes:

• Small displacements
• Linear material behavior
• No contact or gaps

When to go nonlinear:

• Displacements > L/50
• Stresses approaching yield
• Stability concerns (P-delta, buckling)
• Connection behavior with gaps

Validation Strategies

Back-of-Envelope Checks

Before any FEA, estimate:

Beam deflection: δ = 5wL⁴/384EI (simply supported, UDL)

Natural frequency: f = (π/2L²)√(EI/m) (simply supported)

Axial stress: σ = P/A

If FEA differs by more than 20% from hand calc, investigate why.

Reaction Checks

Sum of reactions must equal applied loads:

• ΣFx = Applied Fx
• ΣFy = Applied Fy
• ΣFz = Applied Fz

Most software reports equilibrium - always check this.

Deflected Shape Review

Does the deformation make sense?

• Symmetric loading should give symmetric deformation
• Supports should have zero displacement (in restrained direction)
• No unexpected movement

Benchmark Problems

Validate your modeling approach on problems with known solutions:

• Simply supported beam under UDL
• Plate with central point load
• Cantilever with end load

Project-Specific Examples

Flat Slab Analysis (Kidbrooke Village)

For RC flat slabs with complex column layouts:

1. Model full floor plate, not isolated bays
2. Use shell elements with appropriate thickness
3. Apply pattern loading as per EC2
4. Extract design moments using Wood-Armer method
5. Verify against equivalent frame method

Portal Frame Analysis (F1 Facilities)

Large-span portal frames with crane loading:

1. Model frame as 2D plane stress or 3D line elements
2. Include realistic connection stiffness
3. Apply crane loads at multiple positions
4. Run P-delta analysis for stability
5. Check deflections against functional requirements

Transfer Structure (High-Rise)

Heavy transfer beams supporting multiple floors:

1. 3D model with solid elements for transfer
2. Full dead load from above before superimposed
3. Construction sequence analysis
4. Check bearing stresses at supports
5. Long-term deflection considerations

Software-Specific Tips

Tekla Structural Designer

• Good for code checking of standard structures
• Limited mesh control - verify critical areas in dedicated FEA

SCIA Engineer

• Excellent mesh control and result visualization
• Watch for default settings on European codes

Robot Structural Analysis

• Powerful but steep learning curve
• Custom result combinations require setup

ETABS

• Optimized for buildings
• P-delta analysis well-implemented

Conclusion

FEA is indispensable for modern structural engineering, but it requires disciplined application:

1. Plan your analysis before starting the software
2. Simplify intelligently - complexity doesn’t equal accuracy
3. Validate relentlessly - hand calcs, reactions, deformed shape
4. Document assumptions - for review and future reference
5. Question results - if they seem wrong, they probably are

The best analysts I know spend more time thinking than clicking. Invest in understanding your structure before investing in mesh refinement.