Cold Bending of Laminated Glass: Challenges, Design, and Viscoelastic Effects
What Is Cold Bending?
Cold bending is a technique for producing curved glass panels by mechanically forcing initially flat glass sheets into a curved configuration at ambient temperature. Unlike hot bending, where glass is heated on a mould in a bending oven until it slumps to the desired shape under gravity, cold bending applies external constraints to elastically deform the glass into a curve after it has been laminated.
The process is straightforward in concept: a flat laminated glass panel is placed against a curved frame or mould, and mechanical fixings (clamps, bolts, or adhesive connections) hold it in the curved shape. The glass remains in the elastic regime throughout — it is not heated beyond ambient temperature and no permanent plastic deformation of the glass occurs.
Why Cold Bending?
Free-form architecture is a dominant trend in contemporary building design, with buildings completely clad in curved glass facades. The challenge is producing these curved panels at a reasonable cost while meeting structural design requirements. Cold bending offers several advantages over hot bending:
- No mould cost. Hot bending requires a custom steel mould for each panel geometry. In free-form facades where every panel has a different curvature, mould fabrication often dominates the panel cost. Cold bending uses standard flat glass.
- Better optical quality. Hot-bent glass can develop optical distortions from the bending process (roller wave, mould marks). Cold-bent glass retains the optical quality of flat float glass.
- Standard flat glass supply chain. Cold bending uses ordinary flat laminated glass panels, which can be sourced from any qualified laminator. No specialist bending equipment is needed.
- On-site or in-factory bending. Panels can be bent into their frames either in the factory or on-site during installation, providing flexibility in the construction sequence.
The trade-off: Cold bending introduces permanent elastic stress into the glass, which reduces the available stress capacity for external loads (wind, thermal, impact). The glass is effectively "pre-loaded" by the bending process. Understanding and quantifying this pre-stress is the central engineering challenge.
The Engineering Challenge
When a flat laminated glass panel is forced into a curved shape, the glass plies develop bending stress. In a monolithic panel, the maximum stress depends on the curvature, the glass thickness, and the Young's modulus of glass (E = 70-72 GPa). For a laminated panel, the stress state is more complex because the polymeric interlayer couples the glass plies to varying degrees depending on its shear stiffness.
The key complication is that the interlayer is viscoelastic: its stiffness depends on both time and temperature. This means the stress state in cold-bent laminated glass changes continuously over time:
- During bending (seconds to hours): The interlayer is relatively stiff. The laminate behaves close to the monolithic limit, and the bending stress is high.
- Immediately after reaching target curvature: The peak stress is reached. This is the most critical moment for glass breakage.
- Long-term relaxation (days to years): The interlayer gradually relaxes. Shear coupling between the glass plies decreases, and the stress decays toward the layered limit.
The stress does not simply follow the deformation history — the interlayer retains a memory of the entire loading path. This is the memory effect of viscoelasticity: the current stress depends not only on the current strain, but on the complete history of strain from the moment bending began.
Springback and Long-Term Relaxation
If a cold-bent laminated glass panel is released from its constraints, it does not remain in the curved shape. Two distinct phenomena occur:
Immediate elastic springback
The moment the constraints are removed, the elastic energy stored in the glass plies causes the panel to spring back toward its original flat shape. The degree of springback depends on the interlayer stiffness at the moment of release: a stiffer interlayer maintains more of the curvature.
Long-term viscoelastic recovery
After the initial springback, the panel continues to flatten over time as the interlayer relaxes. Experimental observations show dramatic differences between interlayer types: panels laminated with standard PVB can lose nearly half of their imposed rise within 9 months, while those with ionomer (SGP) interlayers retain most of their curvature.
This long-term recovery must be accounted for in design. If a facade panel is cold-bent and then fixed to the supporting structure, the constraint reactions (the forces holding the panel in shape) evolve over time as the interlayer relaxes. These reactions are not monotonic — they increase as the deformation is imposed, reach a peak, and then decrease during relaxation.
Optimal Deformation Shape
Research has shown that the optimal target shape for single-curvature cold bending is sinusoidal, not circular. This is a critical finding with direct practical implications:
Why sinusoidal is better than circular: A perfectly circular (constant curvature) deformation can only be achieved by applying concentrated moments at the panel ends. These end moments create stress concentrations in the interlayer that can promote delamination. A sinusoidal deformation shape, by contrast, produces no stress intensification even when the interlayer shear modulus is high. It also allows significant simplification of the governing equations.
In practice, the actual boundary conditions in installed cold-bent panels enforce a deformation that is only approximately circular. The absence of reported delamination failures in circular cold-bent panels installed in practice may be attributed to the fact that real constraints produce a shape that deviates from perfect circularity — inadvertently moving closer to the more favourable sinusoidal profile.
Interlayer Effects
The choice of interlayer material has a profound effect on the stress evolution in cold-bent laminated glass. Based on parametric studies comparing three representative PVB grades:
| Interlayer | Peak Stress Behaviour | Long-Term Behaviour |
|---|---|---|
| Acoustic PVB (soft) | Much lower peak stress — response stays closer to layered limit | Decays rapidly to layered limit |
| Conventional PVB | Intermediate — clear coupling effect during deformation | Gradual relaxation over days to weeks |
| Stiff PVB (structural) | Approaches monolithic limit during bending — highest peak stress | Maintains coupling initially, then decays after ~1-2 days to an inflection point |
Interlayer thickness effect
Counter-intuitively, thinner interlayers produce higher stresses in cold-bent laminated glass. This is because a thinner interlayer provides stronger shear coupling between the glass plies (the shear strain gradient across the interlayer is steeper), pushing the response closer to the monolithic limit. The hierarchy is preserved throughout the relaxation process: thicker interlayers always produce lower peak loads.
Temperature Effects
Temperature has a dramatic effect on cold-bent laminated glass through the temperature dependence of the interlayer's viscoelastic properties. The relaxation function shifts along the time axis according to the Williams-Landel-Ferry (WLF) equation.
Parametric analysis for conventional PVB at three temperatures shows:
- At 10 degrees C: The interlayer is stiff. The response is close to the monolithic limit throughout the observation period. Peak stress is high.
- At 20 degrees C: Intermediate behaviour with clear viscoelastic relaxation.
- At 40 degrees C: The interlayer is very soft. The response rapidly approaches the layered limit. Peak stress is significantly lower.
Practical implication: Since the peak stress is highest at low temperatures (when the interlayer is stiff), and this peak occurs at the moment the target deformation is first reached, a possible strategy to reduce peak stress is to increase the temperature of the laminated glass during the bending process — for example, by using heating blankets. This softens the interlayer during the critical bending phase, reducing the peak stress, while the long-term service temperature determines the relaxation behaviour.
Deformation Rate and Loading History
The rate at which the target deformation is applied significantly affects the peak stress. Comparing three loading histories for the same final curvature:
- Fast (parabolic ramp, 120 seconds): Highest peak stress — the interlayer does not have time to relax during bending.
- Medium (1 hour): Moderate peak stress — some relaxation occurs during bending.
- Slow (3 steps over 3 days): Lowest peak stress — the interlayer relaxes between each loading step.
Importantly, all three histories converge to the same long-term asymptotic stress. The deformation rate only affects the transient phase. This means that the bending process can be optimised to minimise peak stress without affecting the final service state.
Design Recommendations
| Parameter | Recommendation |
|---|---|
| Target shape | Sinusoidal preferred over circular — avoids stress concentrations at panel ends |
| Interlayer for stress control | Softer interlayers (standard PVB) produce lower peak stress during bending |
| Interlayer for shape retention | Stiffer interlayers (ionomer) retain curvature better long-term |
| Deformation rate | Slower bending reduces peak stress — consider multi-step loading over hours/days |
| Temperature during bending | Higher temperature reduces peak stress — consider heating blankets |
| Interlayer thickness | Thicker interlayer = lower peak stress (weaker shear coupling) |
| Glass type | Heat-strengthened glass preferred — higher allowable stress and better post-breakage behaviour |
| Minimum radius | Depends on glass thickness, interlayer, and allowable stress — requires case-by-case analysis |
Design tension: The interlayer choice involves a trade-off. A soft interlayer (PVB Clear) minimises bending stress but provides weak shear coupling for in-service loads (wind, thermal). A stiff interlayer (ionomer) provides excellent in-service performance but produces the highest peak stress during bending. The optimal choice depends on the governing design scenario: if the cold-bending stress is the limiting factor, use a softer interlayer; if in-service structural performance is critical, use a stiffer interlayer and manage peak stress through slower bending or higher temperature.
Modelling Approaches
Quasi-elastic approximation
The most common engineering approach treats the interlayer as a linear elastic material with a time-dependent secant shear modulus. At each time instant, the stress is calculated as if the current strain had been applied instantaneously. This approach is simple (no time integration required) and is codified in standards. However, it neglects the memory effect of viscoelasticity — the hereditary influence of the loading history on the current stress state.
For cold bending, the quasi-elastic approximation is reasonably accurate for predicting peak stress and long-term asymptotic stress (within 4-7% for stiff PVB, up to 15-20% for conventional PVB during the relaxation phase). It is acceptable for preliminary design but may miss important transient effects.
Full viscoelastic analysis
A rigorous approach solves the time-dependent integro-differential equations that describe the viscoelastic interlayer. This captures the full memory effect and provides accurate stress predictions throughout the entire deformation and relaxation history. Two main numerical frameworks exist:
- Prony series with FEM: The relaxation function is approximated by a series of decaying exponentials (Prony series), and the equations are solved using commercial FEM software (ABAQUS, ANSYS). This requires 20-30 Prony terms for accuracy across the full time range and can take minutes per simulation.
- Fractional calculus approach: The relaxation function is described by connected branches of power laws, and the constitutive equation uses fractional differential operators. This approach requires only 6-8 material parameters (vs 40-60 for a full Prony series), has a clear geometric interpretation on the bi-logarithmic plane, and can be solved in under a second using the L1 formula with variable time steps.
Research note: The fractional calculus approach to cold bending of laminated glass was developed at the University of Parma and published in Engineering Structures (2024). The model has been validated against ABAQUS simulations for multiple interlayer types and loading conditions, showing excellent agreement. Read the paper.
Model Your Cold-Bent Glass
Use our free tools to compute Prony series for any interlayer material and analyse the viscoelastic response under custom loading and temperature conditions.
Launch Simulation Tools