رفتار حرارتی و ساختاری در مقیاس کامل ساختمان کامپوزیت موضوع آتش شدید محفظه

This paper presents a numerical investigation of the thermal and structural results from a compartment fire test, conducted in January 2003 on the full-scale multi-storey composite building constructed at Cardington, United Kingdom, in 1994 for an original series of six tests during 1995–1996. The fire compartment's overall dimensions were 11 m×7 m with one edge at the building's perimeter, using largely unprotected steel downstand beams, and including within the compartment four steel columns protected with cementitious spray. The compartment was subjected to a natural fire of fire load 40 kg/m2 of timber, in common with the original test series, but the composite slab forming its ceiling was subjected to a uniform applied load of 3.19 kN/m2, which is higher than the original. Numerical modelling studies have been performed using the numerical software FPRCBC to analyse temperature distributions in slabs, manual Eurocode 3 Part 1.2 calculations for beam temperatures, and Vulcan to model the structural response to thermal and mechanical loading. These are compared with the quite comprehensive test data, and a series of cases has been analysed in order to develop a comprehensive picture of the sensitivity of the behaviour to different assumed conditions. The comparison between the modelling of basic cases and the test results shows very good correlation, indicating that such modelling is capable of being used to give a realistic picture of the structural behaviour of composite flooring systems in scenario-related performance-based design for the fire limit state. The extended sensitivity studies show the influence of extra protection to the connection zones of primary beams, and the effects of different vertical support conditions at the perimeter of the fire compartment. The effect of incomplete overlapping of the reinforcing mesh in the slab, which is believed to have occurred in one region, is also considered.

Over the past decade, the fire engineering design of steel structures has developed considerably at its leading edge. The traditional prescriptive approach has placed reliance simply on limiting the temperature achieved by any structural component when subjected to a prescribed “Standard Fire” heating regime [1] up to the required fire resistance time, but taking no account of its position within the building or its loading condition. It has therefore been normal practice to protect steel beams and columns by using prescribed thicknesses of fire protection materials, whose purpose is simply to limit the steel temperatures to these values in the specified fire resistance time. More recent design rules [2] and [3] have been based on the behaviour of individual structural elements in loaded furnace tests, again subject to the Standard time–temperature curve. In two important respects these approaches fall short of realistic approaches to reality: 1. They do not incorporate heating of structure by natural post-flashover fires, whose characteristics include both growth and decay phases and are controlled by the fire load, ventilation and compartment properties. This has only recently [4] and [5] been introduced to design documentation with limitations on allowable compartment sizes. 2. They do not consider the effect of interaction and load-sharing between elements, which is particularly important in composite construction where slabs and beams in a floor interact completely as an integrated system. In the standard test an isolated unprotected beam can only be expected to achieve 15–30 min’ fire resistance if it carries any reasonable amount of load. There is a wealth of published test data from standard furnace fire tests on isolated elements; this information is of limited use in validating numerical modelling due to the drawbacks of the standard test as discussed. Tests on more complicated full-scale structural assemblies have been rare due to the costs and the complexity of such projects. Early work on fire testing of sub-assemblies was undertaken by Kruppa [6] and Rubert and Schaumann [7] which were followed later by full-scale testing on a loaded plane steel framework subject to a natural fire [8]. Accidental fires in composite steel-framed buildings, for example the well documented Broadgate fire [9] in London, have given strong indications that the performance of complete structural systems in fire is much better than is suggested by the standard test on isolated members. This poses the question of whether it is necessary to protect the structural steelwork to the extent demanded either by prescriptive documents [10] or by member-based design based on the standard fire test. In order to understand the behaviour of such continuous systems, and potentially to use this to advantage in fire engineering design of structures, it is clear from the cost of testing that the basic route must be via validated numerical modelling rather than tabulated data based on full-scale testing. However, in order to provide scientifically monitored data for validation of numerical analysis approaches, it was necessary to perform some testing on a full-scale structure constructed using current systems and site practices. The construction of the eight-storey steel framed building [11] (Fig. 1) at the Cardington Laboratory of BRE, to a design which was representative of contemporary medium-rise office buildings, was undertaken in 1994. This building was designed as a non-sway frame to BS5950 standards, to a specification which included bracing in a central core and two stair-wells, providing the necessary resistance to lateral wind loads. The main steel frame was designed for gravity loads only. The floor construction incorporated a steel decking topped with a light-weight in situ concrete composite floor. The erection and fabrication of the structure was performed in accordance with normal British site practice.Six full-scale fire tests were conducted inside the building at various locations (Fig. 2) during 1995 and 1996. The tests were extensively instrumented with thermocouples, strain gauges, and displacement transducers. The imposed load in all these tests was 2.66 kN/m2. The total load on the floor system was 5.48 kN/m2, representing a load ratio on the secondary composite beams of 0.44. This load was applied using sand bags which were distributed as evenly as possible on the floors. These tests have been well documented [12] and [13] and the results have since been useful in verifying numerical models [14] and [15]. The key data about each of this original series of tests are reiterated in Table 1. Perhaps the key point overall is that, while design codes based on isolated member tests in standard fire conditions [2] and [3] give the critical steel temperature for runaway failure of unprotected secondary beams in the region of 670 °C for all the tests at this load ratio, and the temperatures experienced were in all cases above this level, no structural failure was observed in any of them. Deflections of the floor system in most cases exceeded the normal furnace-test limit of span/30span/30 during the heating phase, reducing somewhat as cooling took place. Subsequent numerical analyses of the tests [14] and [15] have shown that this unexpected resilience of the floor system derives from combinations of: • Three-dimensional flexural bridging due to continuity of the slabs, both directly above fire compartments and with surrounding cool areas. This happens at all stages of deformation, being more pronounced when curvatures are biaxial and reverse over lines of support, but tends to be the dominant mechanism at low deflections. • Catenary tension in both slabs and composite beams when deflections reach high levels and the curvatures are predominantly uniaxial. For this action to provide substantial support the edges of the zone in catenary must be capable of resisting the largely horizontal tensile forces generated. • Tensile membrane action in slabs at high deflections. This is a self-equilibrating action [16] in which an individual slab in biaxial curvature generates a multi-directional tension field in its central region, which is resisted by a ring of compression around its edge zone. The necessary condition for this action to be effective is that vertical support to the edges of the slab should be maintained, and that its aspect ratio should be reasonably “square”.A seventh fire test was carried out on the building in January 2003. The main emphasis in undertaking this work was to collect more detailed data on the performance of typical beam-to-column and beam-to-beam frame connections in fire, a considerable spur to the work being the documented failures [17] of connections (admittedly, very different in detail) in some buildings of the New York World Trade Center complex on 11 September 2001, and subsequent recommendations for investigation. This test also provides further opportunities both to investigate the behaviour of the composite floor system in the multi-storey building and to use the results to check the capability of specialised numerical modelling software. The program Vulcan has been developed progressively over many years at the University of Sheffield to perform large-deflection global structural modelling of the behaviour of steel-framed, composite and concrete structures in fire conditions. It has been used to model the six original Cardington fire tests, as well as in more generic research studies [18], [19] and [20] intended to give insights into the behaviour of composite framing systems subject to fires, and in design analysis. Although so far the comparisons with data, particularly from the original test series, give confidence that the analytical approach of the software gives an accurate picture of the structural behaviour in fire, it depends on an adequate representation of many different structural properties and their interaction under complex conditions of temperature, loading and deformation. It is therefore imperative that any scientifically monitored fire test data should be used to indicate the reliability of the modelling, and any needs for development. This paper presents comparisons of the results of “Test 7” on the Cardington composite frame with the results of thermo-structural numerical modelling using Vulcan. Three-dimensional non-linear analyses of the fire compartment and surrounding cold structure have been carried out, using measured temperatures where available, but in other cases using either simplified temperature calculation or numerical thermal analysis of structural components.

This paper has presented the results from a series of thermal and thermo-structural numerical analyses of Fire Test 7 in a fire compartment of size 11 m×7 m including a complete bay of the column grid of the multi-storey composite frame in the BRE Cardington laboratory. The fire compartment was subjected to a fire load of 40 kg/m2, and the slab above the compartment was loaded uniformly with an imposed load of 3.19 kN/m2. The objectives of the analyses were: • To test the accuracy with which the behaviour can be predicted using the global analysis software Vulcan, which is based on a relatively coarse grid of geometrically non-linear beam-column and shell elements and concrete constitutive relationships with a smeared cracking assumption. • To check the sensitivity of predictions to the effects of support assumptions, which inevitably have to simplify the considerable complications which occur due to the design details of a real structure. Six different cases have been analysed. The results show that this type of modelling predicts the deflection behaviour with considerable accuracy, especially considering the uncertainties embodied in representing the biaxial properties of concrete at high temperatures, and using analysis which does not consider the formation of discrete cracks. All the cases modelled follow very much similar deflection patterns; even with boundary assumptions which neglect all edge support provided by the vertical wind-posts are relatively accurate away from the local vicinity of this edge. As the representation of the vertical restraint at these points becomes more realistic, firstly by providing total vertical support, and then by allowing slip at the slotted holes, up to a reasonable estimate of the slip provided, the deflection profile near to the edge becomes progressively more accurate. Accurate thermal modelling of the structure, particularly in this case by obtaining an accurate representation of the temperatures in the over-sprayed 1 m zones at the outer ends of the primary beams, has a fairly major effect compared with weakening of the slab at the location of the observed longitudinal crack, suggesting that the deflection pattern is more a function of temperature distribution than of structure loading and strength. The load-carrying mechanism is largely one of tensile membrane action at high deflection in all cases, and it is interesting to compare the principal traction plots of Figs. 25 (Vulcan 3) and 31 (Vulcan 6). Although very similar tensile membrane action distributions are shown overall in these, the tensile vectors in the short-span direction are noticeably smaller in the crack region in Fig. 31 than in Fig. 25. There is clearly enough redundancy in the slab system for the force paths to be re-routed away from the major crack without any large change in the observed structural stability. This is logical, since a run-away deflection, after the formation of the major longitudinal crack, would require pull-in of the bay edges perpendicular to the direction of this crack. The continuity of the slab across these edges ties it into a large area of slab which surrounds it on three sides and therefore has considerable stiffness and load capacity to resist this horizontal pull-in. Clearly this will not be true under all circumstances, particularly in corner bays where some of the pull-in is only resisted by columns. From the viewpoint of design, where it is desired to make use of tensile membrane action in ensuring fire resistance of a composite floor, it worth remembering the large longitudinal and transverse cracks in the concrete slab, and their association with very high deflections. With deflections of the order of span/10 there is a considerable shortening of the distance between adjacent edges of the slab. Even if the strain implied by this were tolerable, concrete is a brittle material in tension, and the shortening will inevitably localise into a very few discrete cracks rather than “smearing” as is assumed in the modelling. This may constitute an integrity failure of the slab as a separating element, even if no structural instability is caused.( Fig. 32)