Published 03.11.25
Until now, little has been known about how modern multi-pane low-energy windows behave in fully developed fires. Fire models have often assumed complete glass fallout. However, new furnace tests conducted at DBI by researchers from DTU and DBI now show clear differences: Double-glazed windows exhibited significant fallout, while triple-glazed panes largely remained in place.
In the history of building fires, windows have often been the weak point. They are exposed, fragile, and when they fail under heat, they create openings that feed a fire with oxygen and allow flames to escape. For decades, fire safety engineers have built their models around this assumption: Once a fire develops beyond the flashover point, windows will crack and fall out, leaving the full opening exposed.
That assumption may have worked in the past, when single-pane and simple double-glazed windows were the norm. But the building industry has changed. In the pursuit of energy efficiency, modern constructions are fitted with large, multi-pane low-energy windows. Triple glazing is no longer exotic but standard in much of Northern Europe. And while this has obvious benefits for energy performance and comfort, it raises new questions for fire safety.
Until now, those questions have remained largely unanswered. Very little experimental data has existed on how modern multi-pane windows actually perform when exposed to fire. This knowledge gap was the starting point for a joint project between DTU Civil and Mechanical Engineering and DBI – the Danish Institute of Fire and Security Technology.
To investigate, the researchers carried out a series of controlled furnace experiments at DBI. Complete window assemblies were mounted in large-scale furnace frames and exposed to two different fire curves representing post-flashover conditions: The traditional ISO 834 standard curve and a parametric fire curve designed to mimic more realistic room fires in energy-efficient buildings.
The programme consisted of eight separate tests of both double- and triple-glazed low-energy windows, including one smaller triple-glazed unit and several larger specimens. Throughout the experiments, the team monitored the progression of cracks in each pane, recorded the time until failure occurred, and measured how much of the glass surface area ultimately fell out.
- With two-pane windows we saw 15-53% of the glass area fall out, creating large openings for fire and smoke. With three-pane windows the loss was less than 2%, and often the glass stayed in place even though it cracked, explains Hjalte Bengtsson, PhD Fellow at DTU and lead author of the study.
The results also showed a relatively linear sequence of breakage — the time between the first, second and third pane cracking was roughly even. And even when temperatures increased, the amount of glass falling out did not rise significantly.
The findings underline a simple but important point: The assumption of total window fallout in fire models no longer matches reality.
- For decades, design models have simplified the problem by assuming complete window fallout, says Ian Pope, Fire Safety Researcher at DBI and co-author of the study.
- Our results show that this assumption is not valid for modern triple-glazed units. They behave differently, and that has consequences for how we predict fire development inside compartments.
This shift does not mean that modern buildings are inherently safer. Rather, it underlines that energy-efficient designs, with their tighter envelopes and limited ventilation, change the dynamics of fire development – and that can both mitigate and create new challenges for fire safety.
When glass stays in place, less oxygen reaches the fire. That alters the dynamics inside the room. Instead of rapid, high intensity burning, a fire may smoulder or burn at lower intensity for longer. For structural elements – especially combustible ones such as timber – this can change whether or not self-extinction occurs.
- Less ventilation can mean that fires develop differently inside the room. They may burn longer at lower intensity, which can change how structures respond – especially in modern timber buildings where self-extinction is an important factor, says Hjalte Bengtsson.
The observation that fallout does not necessarily increase with higher temperatures also points to a more complex relationship between thermal exposure and glass behaviour than previously assumed.
As Hjalte Bengtsson notes, the higher-temperature ISO tests represented a maximum exposure, while lower, more realistic fire curves might produce less damage.
Ian Pope adds that under sustained high heat, the glass layers may even begin to fuse together, which could help the pane maintain its integrity rather than fail completely.
What the study clearly demonstrates is the need to revisit how windows are represented in fire safety engineering. Models and design strategies that assume automatic fallout risk underestimating or misrepresenting how real fires develop in today’s buildings.
- This is a first step. We now have data that can help us rethink how we model windows in fire, so fire safety can be designed to reflect real fire behaviour more closely, says Hjalte Bengtsson.
Looking even further ahead, the researchers see the possibility that windows could become an active element in fire strategy rather than just a vulnerability.
As Ian Pope notes:
- With more knowledge, one could imagine windows being designed to play a deliberate role in controlling fire development – not only as a weak point, but as a managed part of the overall fire strategy, for example by being engineered to fail or open at specific stages of a fire.
The study is one of the first to explore the post-flashover fire performance of modern low-energy windows, and as such, it represents an important first step. While the tests were conducted under controlled furnace conditions and did not include factors such as coatings, seals or mechanical stress, they still provide some of the clearest experimental data to date on how double- and triple-glazed windows actually behave in severe fire exposure.
- It’s a simple setup, but it gives us valuable insight into how modern glazing systems perform when exposed to fire. We can now start refining both the experiments and the models, says Hjalte Bengtsson.
Building on this foundation, future research will examine larger windows, different glass types and more realistic fire dynamics — including how pressure, ventilation and duration interact with the layered structure of multi-pane glazing. Together, these efforts will help make fire safety engineering more evidence-based and better aligned with how fires behave in today’s energy-efficient buildings.
The study, ‘Experimental study of the post-flashover fire performance of multi-pane low-energy windows’ (2025), was carried out by researchers from DTU – Civil and Mechanical Engineering and DBI – Danish Institute of Fire and Security Technology. Lead author Hjalte Bengtsson (DTU) and co-author Ian Pope (DBI) conducted a total of eight furnace tests at DBI’s large-scale fire testing facilities in Denmark.
Each test involved complete window assemblies exposed to two fire curves representing post-flashover conditions: the traditional ISO 834 standard curve and a parametric curve designed to simulate more realistic room fires in energy-efficient buildings. Both double- and triple-glazed units were tested, including one smaller triple-glazed window and several larger specimens.
During each 30-minute exposure, the team systematically observed cracking patterns, the linear sequence of pane breakage, and the extent of glass fallout. The study recorded the time to first cracks in each pane and the percentage of total window area lost, revealing distinct differences between two- and three-pane configurations.
Contact
