FEM Validation of Analytical Methods for Laminated Glass

The Benchmark Study

Galic et al. (2022) performed a comprehensive validation study comparing multiple design methods against physical experiments. The test specimen was a laminated glass beam: 500 × 100 mm, composed of 5 mm glass + 0.76 mm PVB (Trosifol UltraClear) + 5 mm glass, tested in four-point bending at temperatures from −20°C to 50°C.

The study compared five approaches:

1. EN 16612 analytical method (stiffness families)
2. SCIA Engineer glass addon (stiffness families)
3. Dlubal RFEM RF GLASS (stiffness families and manual G input)
4. Wolfel-Bennison analytical method (actual G from TTS)
5. Abaqus SIMULIA (Prony series + WLF, full viscoelastic FEM)

SCIA Engineer: Stiffness Families

SCIA Engineer’s glass addon implements EN 16612 using stiffness families to characterise the interlayer. The user selects a family number (0, 1, or 2) and the software assigns a fixed ω value for each load condition.

The interlayer is described by a single integer (the family number), not by its actual G(t,T) function. No direct shear modulus input is possible in the standard addon.

Result: approximately 64% error versus experimental deflection measurements. The error is systematic and conservative — SCIA always overpredicts deflection because the family-based ω underestimates the actual coupling.

Dlubal RFEM: Manual G Input

Dlubal RFEM with the RF GLASS addon offers two modes:

Mode 1: Stiffness families (≈ 60% error)

Same approach as SCIA — same conservative error.

Mode 2: Manual G input (< 1% error)

RFEM allows the user to directly specify the interlayer shear modulus G at the design condition. When the correct G value (from TTS characterisation) is entered:

This proves definitively that the FEM solver is not the problem. The accuracy bottleneck is the interlayer data.

Abaqus SIMULIA: Full Viscoelastic FEM

Abaqus implements a full 3D viscoelastic analysis using the TRS (Thermo-Rheological Simplicity) material model. The input consists of:

• Prony series coefficients {g_i, τ_i} for the normalised shear relaxation
• WLF constants C₁, C₂, T₀ for temperature dependence
• Instantaneous elastic properties: G₀ and ν

The solver computes the interlayer response at each time step and temperature, capturing the full time-dependent stress redistribution within the laminate.

Result: Galic et al. report results “in nearly 100% agreement with the experiment” — < 1% error across the full temperature range.

Accuracy Comparison

Implications for Practice

For structural engineers

• Do not trust stiffness family results blindly. They overestimate deflection by ~61%, leading to unnecessarily thick and expensive glass.
• If your software allows manual G input (Dlubal, Strand7, SOFiSTiK), use actual G(t,T) from TTS characterisation — the accuracy improvement is dramatic.
• For standard rectangular panels, the Wolfel-Bennison analytical method with actual G gives 3% accuracy — no FEM needed.
• Reserve full viscoelastic FEM (Abaqus-level) for complex geometries, point-fixed glazing, or research validation.

For the industry

The investment should be in material characterisation (Prony series databases with actual G(t,T) data), not in more sophisticated FEM solvers. FRACTAN’s multi-vendor interlayer database provides exactly the data that transforms a 61%-error calculation into a 3%-error calculation.