[
With no UK furnace big enough to conduct a test Dr Gordon Cooke, fire consultant to the Performa Systems testing programme, sponsored by Eurobond Laminates, identifies the key points in the development of a professional assessment. He shows that fire resisting steel faced sandwich panels with non-combustible rock wool cores can be used for large fire walls in industrial and storage buildings.
Fire resistance tests according to BS 476 Part 22 can only be made in enclosed furnaces, and the maximum size furnace available in the UK will only accommodate a 3m square partition wall assembly. The Performa system, on the other hand, would have to deliver a fire resistance of one, one-and-a-half, and two hours for fire walls incorporating, a) panels spanning 9m horizontally to any height and, b) panels spanning 8m vertically without intermediate supporting steel work. A professional assessment was thus required.
The key points in the development of the required assessment were:
– the factors affecting the fire resistance of steel faced sandwich panels with non-combustible cores substantially larger than that tested in the British Standard (BS) fire resistance test;
– the reasons for making a large scale ad hoc fire test; and
– how the large scale test contributed to the formulation of the Performa assessment.
It should be noted that constructions of other structural materials may be prone to loss of fire resistance when the element is large. For example tall brick walls, especially cantilever constructions, and tall board-faced metal stud firewalls can become potentially unstable in fire due to out-of-plumb effects. Thus care must be taken in the design and the assessment of the fire resistance of large elements of construction.
Standard fire resistance tests
Two fire resistance tests were successfully conducted at Warrington Fire Research Centre. These were on vertical assemblies of 1150mm wide Performa panels of 100 and 150 mm thickness and each assembly attained more than a two-hour rating. The furnace test gave essential information on which to base a professional quantified assessment of fire resistance for elements substantially larger than the fire tested specimen (an oversize element is an example of an ‘extended application’ in CEN terminology and requires careful assessment).
The BS 476 fire resistance test on a firewall is a severe test of the stability, integrity and insulation characteristics of a fire separating wall. It simulates a fully developed fire after flashover and is not to be confused with several of the ad hoc room-and-crib tests in which flashover does not occur, or occurs for only a short time, and it should not be compared with real fire scenarios. In the fire resistance test, the furnace combustion gas temperature reaches approximately 1200 degrees C at the end of the two-hour exposure and is accompanied by a high level of radiant heat.
The most important aspect of the the fire resistance of non-combustible steel faced panel construction is integrity of panel joints i.e. the ability to prevent gaps from forming through which flames and hot gases may pass. Thermally induced deformations in large elements can greatly affect integrity. Since, in a sandwich panel assembly the fire exposed steel face may be expected to delaminate early in a real fire (and within the first five minutes of a standard fire resistance test) causing opening of unstitched panel joints, the important factor is integrity of the joints in the unexposed steel face.
It should be noted that steel-faced panels having combustible cores may need their joints stitched (usually by rivetting), in order to prevent access of air which allows consumption of the core by flaming combustion rather than by pyrolysis. Once the core is involved in flaming combustion it reduces in thickness and has a reduced insulating effect. With a mineral fibre core it does not matter if the joints open up in the exposed steel face, as the core material is unaffected.
Deformations of long panels
Any differential bowing of adjacent panels could cause unstitched panel joints to dislocate in the unexposed face, causing a loss of integrity. The magnitude of bowing of the unexposed steel face (which is fixed at its ends in horizontal spans) is much greater in a 9m span than in a 3m span. The author has developed equations which can be used to predict both a) thermal bowing along a steel stud due to a temperature gradient across the section and b) the bowing which occurs when a panel face is heated along its entire length, while its ends are position-fixed.
Before attempting a large scale fire test it was thought prudent to erect three 9m long panels spanning horizontally between robust vertical columns, and then remove one steel face from the three panels to establish any instability effect when only one steel face (the unexposed face) was supporting the 150mm thick rock wool core. This would simulate the loading condition after the fire exposed face had delaminated. This room temperature test proved that the assembly was stable and the end fastenings were unaffected.
However this test did not show what happened when the unexposed steel face slowly rose in temperature and bowed, as in a furnace test exposure or a real fire. In Performa panels spanning horizontally, the steel faces at the ends are rigidly attached via 5mm diameter steel fastenings to the supporting steel columns. When restrained in this way, any rise in temperature due to heat flowing through the rock wool would be accompanied by bowing. 150mm thick Performa panels experienced a temperature rise on the unexposed face of nominally 80 degrees C at the end of a two hour exposure in the standard fire resistance test. This was calculated to cause a mid-span bow of nominally 180mm in a 9m span, but less for smaller spans.
Ideally, any large scale ad hoc test should impose a temperature rise on the unexposed face of nominally 80 degrees C during the period of maximum fire severity on the exposed face. This could be achieved, for example, with low voltage electrical heating tapes applied to the three lowermost three panels.
If all panels bowed to the same amount, no panel joint dislocation could occur. But the lowermost panel cannot freely bow along its lowermost edge since, in normal practice, it is contained within a light steel channel section fixed to the concrete floor slab.
The potential weakness in integrity is therefore along the joint between intermediate panels and the lowermost and the uppermost panel in the assembly. As the upper panel bows, it drags the upper edge of the lowermost panel and rotation about point A occurs with potential loss of integrity at point B. In the Performa system, the bowing is limited by using a core thickness that keeps the temperature rise of the unexposed face as low as possible, and the joint at B is stitched on both sides as an additional precaution.
Another thermal distortion effect that has been examined in the Performa system is in-plane bowing, when there is a temperature difference across the face of a panel. I
Large scale ad hoc fire test
To check the bowing behaviour and stability of a large Performa panel assembly, a full scale fire test was deemed necessary. Many fire test simulations were considered but it was clearly not possible to have an enclosure fire because of the large size. Nor was it practical to have a fire which, as in the furnace test, continued to burn for two hours, unless oil or gas was employed. The obvious choice of fire load was timber – the classic choice of fire test scientists and engineers and a choice that would lead to an environment friendly fire producing very little particulate matter. Although, as pointed out above, it would be desirable to impose a temperature rise on the unexposed face of approximately 80 degrees C, and this could be done with, for example, an assembly of electrical heating tapes applied over the lowermost three panels, it would not be possible to arrange this within the timescale. Besides, the tapes would need insulating to prevent heat loss and the assembly of tapes and insulation would obscure sight of the lower panel joints – the joints of principal interest.
An assembly of panels each spanning 9m horizontally and stacked to give a 10m height was assembled in the open air, fixed as in practice to a robust rectangular frame made of steel I-sections. A row of timber cribs each 1m square and 1.25m high using 50mm square sticks were spaced roughly 250mm apart and placed along the bottom to produce a ‘line fire’. This represented a fire load of 30MJ: this is six times the fire load used in the Factory Mutual large corner test for panels and linings (Class 4880), and was expected to give a severe fire. Measurements were made of deflections and temperatures of the exposed and unexposed steel faces.
In an extended application assessment, one has to assume that the whole wall face is exposed to a uniform fire, as in the 3m square BS furnace test specimen. In the furnace test the exposed face is delaminated over the whole area, and if the ad hoc test was to be meaningful, the whole of the 10m high face of panels would similarly have to be delaminated. This means that the dead load of the unexposed steel face becomes that of the steel face and rock wool core which remains bonded to it. This could add to the loading on the lowermost panels in the stack and instability bearing in mind that the composite sandwich action has been destroyed. The flame height, calculated using fire engineering guidance in BS 7974 (Application of fire engineering principles to the design of buildings Part 1: 2003 Initiation and development of fire within the enclosure of origin (sub-system 1)), was expected to be around 9m, so the exposed steel face of panels above 7m level were intentionally debonded after installation, and before the fire test, using a hot diffuse gas flame torch operated from a cherry picker.
The fire lasted approximately 90 minutes from ignition to the point of major decay – a point, admittedly, difficult to define. There was a cross wind blowing during the test assessed to be force 4 on the Beaufort scale. However, once all the cribs had become involved in fire the strong upward flames mostly dominated over the effect of wind.
As predicted, the fire exposed steel face delaminated early in the ad hoc fire test, and underwent massive bowing as it attained a temperature of more than 1300 degrees C over the lower two panels. However, again as predicted, the unexposed steel face remained cool and, although the lowermost panel was subjected to severe distortion caused by upward bowing of the base perimeter steel beam (a beam that would not be present in a normal site installation), the panel joints remained engaged and the rock wool core, similarly subjected to very high local temperatures, remained in place. This shows that distortion did not dislodge the protecting core material despite the very severe fire. The test proved good joint integrity was achieved and this was because the unexposed steel face temperature was kept low and uniform, which was achieved by the excellent elevated-temperature insulation properties of rock wool.
As a result of this development and test programme, the Performa firewall system has been assessed to have a two-hour fire resistance for 9m horizontal spans and for 8m vertical spans. In the case of horizontal spans, the lowermost interpanel joint needs stitching to prevent joint dislocation due to the differential bowing described above. Similarly, vertical panels need to be free to bow in the fire condition and this is allowed in the design of panel head detail, and a special joint at the boundary of the vertical panel assembly. The panels do, of course, require support from beams or columns having at least the same fire resistance as the panels.
Conclusions
A large scale ad hoc fire test has been successfully conducted to examine the effect of a severe fire on the integrity of the unexposed steel face of a 10m high assembly of panels, spanning 9m horizontally. The qualitative and quantitative results of the severe test, together with fire safety engineering calculations, has enabled an assessment of fire resistance to be made for large assemblies requiring up to two hours fire resistance.
It should be noted that the purpose of the ad hoc fire test was not to attempt to simulate a particular fire, as in practice fires may be more severe at different parts of the assembly depending on a wide range of fire parameters. The fire could be more severe at the top of
a fire wall when a ‘crown’ fire occurred, as in a high-rack storage building, or the fire could be more severe at the base, as in the case of a local fire on the floor immediately next to the panels. Further, the severity in
a fire compartment would be different for different fire load density and ventilation conditions.