Citation

Wald F, Chlouba J, Uhlir A, Kallerova P, Stujberova M. Fire Safety J. 2009; 44(1): 135-146.

Copyright

(Copyright © 2009, Elsevier Publishing)

DOI

10.1016/j.firesaf.2008.05.002

PMID

unavailable

Abstract

aDepartment of Steel and Timber Structures, Czech Technical University in Prague, Czech Republic bDepartment of Steel Structures, Slovak Technical University, Slovak Republic The paper reports on an experimental programme to investigate the global structural behaviour of a compartment in the three-storey steel frame building in a plant of the Mittal Steel Ostrava exposed to fire before demolition. The research project of the Czech Technical University in Prague was focussed on the examination of the temperature development within the various unprotected structural elements and its connections, the corresponding distribution of horizontal forces and the behaviour of the laterally unrestrained beams during the natural fire. The experiment also allowed studying of the heating of external elements, the influence of connection in a wall of sandwich panels, the temperature development in light timber-based panel and the degradation of the timber concrete composite element. Before the compartment fire, a local fire was prepared to verify the models of the temperature development in an unprotected column. The comparisons to the simplified calculations by European standards are included in the text to show their strong and weak points in prediction of temperatures of gas and structural elements during fire. Keywords: Structural engineering; Steel structures; Fire test; Full-scale tests; Fire design; Eurocodes; Connections The Cardington Laboratory of the Building Research Establishment, BRE, has provided the opportunity to carry out several research projects that included full-scale fire tests in a former airship hangar with dimensions 48 m×65 m×250 m. The laboratory comprises three experimental buildings: a six-storey timber structure, a seven-storey concrete structure and an eight-storey steel structure. The steel test structure is a steel-framed construction using composite concrete slabs supported by steel decking in composite action with the steel beams. It has eight storeys of 33 m and is 5 bays wide (5×9 m=45 m) by 3 bays deep (6+9+6=21 m) in plan. Seven large-scale fire tests at various positions within the experimental building were conducted. The first test performed in Cardington was a restrained beam test involving a single 305×165×UB40 composite beam section supporting the seventh floor of the building. A gas-fired furnace with 24 m2 was used to heat the beam to 875 °C. The second test, a plane frame test, involved heating a series of beams and columns across the full width of the building. Again, a gas-fired furnace of area 53 m2 was used to heat the steelwork to approximately 800 °C. The BRE corner compartment test, area 54 m2, was the first natural fire carried out in the Cardington Laboratory, representing a typical office fire: timber cribs were used to provide a fire load of density 700 MJ m−2. The compartment walls were constructed using fire-resistant board and the northern boundary was formed by constructing double-glazed aluminium screens. All columns were protected up to and including the connections. It was observed that the fire development was largely influenced by the lack of oxygen in the compartment [8]. The reported steel temperature reached 903 °C with gas-measured temperature 1000 °C. The fourth test, the BS corner compartment test with area of 70 m2, also used timber cribs to provide a fire load of density 788 MJ m−2. In this test, both the perimeter beams and the columns were fire protected with the internal beam unprotected. Load-bearing concrete blocks were used for the compartment walls. The fifth test was the largest compartment test in the world with area 342 m2. The compartment was designed to represent a modern open-plan office, 18 m×21 m. The compartment was bounded by fire resistant walls. The main aim was to investigate the ability of a large area of composite slab to support the applied load once the main beams had failed. Consequently, all the beams had no fire protection and all columns were fire protected. Again, the ventilation conditions governed the fire severity. In the demonstration test, the sixth large fire test with area 136 m2, unlike in the previous tests, real furniture, desks, chairs, filing cabinets and computer terminals were used to provide the fire load of density 788 MJ m−2. The ventilation was provided by windows and blank openings. The beams were unprotected while the columns were protected. This test was characterised by a rapid rise in temperature representing a severe fire scenario. The gas temperature reached the highest measured temperature during the compartment tests 1150 °C and was measured secondary beam deflection of 610 mm. The structural integrity fire test, large test no. 7, was carried out in a centrally located compartment of the building, enclosing a plan area of 11 m×7 m on the fourth floor, see Ref. [7], on 16 January 2003. The opening of 1.27 m high and 9 m length simulated an open window to ventilate the compartment and allowed the observation of the element behaviour. The ventilation condition was chosen to produce a fire of the required severity in terms of maximum temperature and overall duration. The applied load was simulated using 1100 kg sandbags applied over an area of 18 m×10.5 m on the floor immediately above the fire compartment. The sandbags represented 100% of the permanent actions, 100% of the variable permanent actions and 56% of live actions. Wooden cribs were used to provide a fire load of density 700 MJ−1 m−2. The applied load was designed to fail the floor, based on analytical and FE simulations, but the collapse was not reached. During this fire test, high quality data on the distribution of temperatures within the main structural members were collected, and particularly connections, protected columns in unprotected structure and composite slab. In the compartment gas was measured the maximal temperature of 1108 °C, on the unprotected primary beam 1057 °C, on the unprotected secondary beam 811 °C, on the beam-to-column header plate connection 839 °C and on the beam-to-beam fin plate connection 908 °C. The behaviour of the fin plate and header plate connection and the concrete slab were studied. The internal forces in the structure during the fire were measured by strain gauges on columns. A local fire test, which was performed on 15 June 2006 and a compartment fire test, on 16 June, on a structure of Ammoniac Separator II in the company Mittal Steel Ostrava were designed to complete the data of previous compartment tests and particularly to learn more about the connection temperatures and the internal forces in a structure, see Ref. [9]. The behaviour of the restrained beams, the heating of external elements, the temperature distribution on the sandwich panels, the light timber-based panels, and the timber concrete elements were studied, when exposed to the compartment fire. The localised fire was focussed on approval of models on the heating of an unprotected column. The building of Ammoniac Separator II was composed of a three-storey steel structure with the composite slabs, steel beams of hot rolled sections IPN160, IPN180 and IPN300, according to EU19-57, connected by the beam-to-beam and beam-to-column header plate connections stiffened by the diagonal wind bracings. The internal size of the fire compartment was designed 3.80×5.95 m with a height of 2.78 m. The structure of the enclosure was made from the light silicate and ceramic bricks. An opening of 2.4×1.4 m ventilated the room during the fire test. The doors and columns were equipped by the fire insulation by boards. The mechanical load on the floor above the fire compartment was composed of the dead and life load. The life load was simulated by about 1 m of water, which was placed into 26 steel barrels and 50 plastic boxes equally distributed on the floor. One box was stored on each barrel and the rest of the boxes were placed at the ends of the floor. The barrels and boxes were placed on the timber pallets and thermally insulated from the floor by 50 mm of a mineral wool (Fig. 1). The unwrought wooden cribs 50×50 mm of length 1 m of softwood dried to moisture till 13% with the average measured unit mass of 506 kg m−3. For the compartment fire, the cribs were placed into eight piles, see Fig. 2, and created the fire load density per unite area of 1039 MJ−1 m−2, which represents the fire load of 58.5 kg m−2. A pile consisted of 13 rows with 10 cribs each plus 2 bars on the top, which meant 132 bars per pile. The simultaneous ignition of piles was reached by their connection by the steel thin-walled channels filled by a mineral wool and penetrated by paraffin. The channels were located on the second layer of cribs connected by four piles together (Fig. 3). The gas temperature in the fire compartment was measured by four thermocouples located 0.3 m below the ceiling, marked in Fig. 4 as TGi. Two thermocouples were placed in front of the fire compartment, 0.5 and 1 m from the front wall. On the structure were located six thermocouples, and on joints seven more, marked as TCi. The position of thermocouples on the lower flange of beams at their mid-span is documented in Fig. 5. Two thermo-imaging cameras and seven video cameras scanned the experiment. Two video cameras were installed behind the thermo-resistant glass in the additionally prepared windows in the compartment internal wall. The development of the gas temperatures shows, see Fig. 4 and Table 1, that at the beginning of the fire, till about 30 min, the gas was warmer in the front part of the compartment, by about 200 °C (thermocouples TG3 and TG4), compared to the back of the fire compartment. During the fully developed fire, after 30 min, the highest temperatures were recorded in the back of the fire compartment. The maximal temperatures were read at thermocouples TG1 and TG2: 1050 and 1032 °C, respectively. In the front part, 920 °C was measured only on thermocouple TG3 (Fig. 6). Fig. 7 shows the comparison of the predicted gas temperature in the fire compartment by parametric fire curve according to EN 1993-1-2: 2005 Annex A, see Ref. [11], to the measured values. The parametric curve is derived for values b=1059 J m−2 s−1/2 K−1; O=0.0398 m1/2; Γ=1.188; qt=238 MJ m−2; . The predicted temperature was 997 °C in 71 min and the maximal measured temperature was 1050 °C in 62 min. Fig. 8 shows that the beam lower flange temperatures correspond to the beam positions in the fire compartment. The front beam, thermocouple TC16, reached the maximum temperature of its lower flange of 775 °C compared to the secondary beam in the back of the fire compartment of section IPN180 with the measured maximal temperature of 970 °C, TC2. The temperature of the beams is compared to the average temperature of the gas measured by thermocouples TG1, TG2, TG3 and TG4. In the step-by-step procedure, the steel temperature is calculated by the simple principle where the heat is brought in/brought out by the member surface and the member volume is heated/cooled. The temperature of the unprotected inner steel structure is in Eq. (4.25) in standard [11] given by: In Fig. 10 is presented the visualisation of the observation by the right hand side thermo-imaging camera, which was focused into left side of the primary beam and its connections. Two cameras were used to scan the beams and their connections. The readings were recorded every 30 s. The results were calibrated to the visible thermocouples, e.g. in Fig. 10 the point SP1 to the thermocouple to TC1. During the fire, the fumes often shrouded the observed objects, but the views with clear visibility from two cameras enabled to read the temperature in reasonable time steps, see Table 2. During the heating, the beam web, marked SP2, in the mid-span in 22 min was warmer by about 36 °C and the upper flange, marked SP3, colder in 54 min by about 26 °C, compared to the lower flange. The lower flange at the end of primary beam close to the connection D2, marked SP5, was colder by about 122 °C compared to the mid-span. This is about 18%, and the factor 0.88 used in Annex D of Ref. [11] predicts the temperature very well. Fig. 11 shows the connection temperatures measured during the experiment compared to the gas and beam temperatures in the mid-span temperatures. In Fig. 12 and Table 3, the temperature differences inside the connection are demonstrated. The standard for fire safety of steel structures EN 1993-1-2: 2005, see Ref. [11], for joints recommends usage of the same fire protection as for the adjacent structure. Alternatively, it provides the prediction of the temperature distribution within the connection, the reduction of the material properties of connectors by elevated temperature, and the analysis of the structure using the component method described in EN 1993-1-8: 2005, see Ref. [12]. For the temperature development in the connection, there are predicted two analytical methods. In the step-by-step method, the temperature is calculated as an element, where the heat is brought in/brought out by the member surface and the member volume is heated/cooled. The geometrical characteristic of the section is the section factor Am/V of the steel parts of which the joint is composed. The temperature of a joint may be assessed using the local section factor Am/V, the value of the parts forming the joint. As simplification, it is possible to consider uniform temperature distribution within the section and to take into account the biggest Am/V value of the steel parts connected into the joint. The temperature of either beam-to-beam or beam-to-column connection covered with a concrete slab can be determined from the temperature of the beam flange in the middle of the span. It is assumed that the temperature of the particular parts of the connection depends directly only on the distance from the lower edge of the connected beam and indirectly on the prediction of the temperature of the lower flange, usually calculated by the step-by-step procedure. If the height of the beam is smaller or equal to 400 mm, e.g. hk400 mm, the temperature in the height hk of the beam is given in Annex D Eq. (D.5) in Ref. [11] by: In Fig. 13, the predicted temperature of the beam is compared to the column header plate connection by applying the section factors of the header plate, Am/V=105 m−1, and in case of calculation from the beam lower flange temperature in the mid-span to the measured temperature in the header plate close to the lower bolt, TC8. The calculation by section factor is conservative in this case. The calculation from mid-span gives lower values during cooling compared to measured values. The results of calculations from the prediction of the gas temperature by the parametric fire curve according to EN 1993-1-2: 2005 Annex A, see Ref. [11], is demonstrated in Fig. 14. The maximal predicted temperature was 796 °C in 65 min and the measured was 691 °C in 71 min. Fig. 15 shows the comparison of the prediction of the temperature in the beam-to-beam connection, which was calculated by section factor of the connected beam, Am/V=200 m−1, to the measured gas temperature. The prediction is conservative during the heating phase and in the maximal temperature. The calculation based on the measured temperature of the beam's lower flange in the mid-span gives a lower temperature compared to the measured one. The reduction 1.0 instead of 0.88 in Eq. (2), e.g. An opening of size 2.4×1.4 m ventilated the experimental fire compartment, see Fig. 17, during the compartment fire. In front of the opening, plate of 0.95×0.3 m and thickness 10 mm at a distance of 2.0 m was suspended on thermo-protected hangers. The steel temperature was measured by two thermocouples placed horizontally in the centre of the plate 250 mm from the edges. The external plate measured temperatures were compared to the gas measured temperatures during the compartment fire and are summarised in Fig. 18 and Table 4. The lower temperatures were recorded by the thermocouple closer to the edge of the opening, e.g. J1. The temperature of the external plate exposed to the fire Tm in K may be predicted by a simplified model utilising the heat balance, see Ref. [13] and Annex B in Refs. [10] and [11], based on which a design is applied for the prediction of the maximal temperature only, The local fire was designed to evaluate the analytical model of the heat transfer into the ceiling and to evaluate analytical prediction models of the column temperature exposed to the natural fire. Close to the heat source, a steel column of section HEB 200 was arranged, and the column was fixed to the secondary beam, see Fig. 20. The measured temperatures during the localised fire are shown in Fig. 21. Table 5 and Fig. 22, and document the development of the gas and steel temperatures during the local fire. Temperatures were measured on the unexposed part of the column at height 1880 mm (TC15), 2080 mm (TC14) and 2580 mm (TC13) as well as in the gas, at thermocouples TG13, TG14 and TG15, and are shown in Fig. 23. The predicted temperatures according to Annex C of EN-1991-1-2 [10] Eq. (C1) to (C9) are set against measured temperatures of the secondary beam in Fig. 24. The thermocouple TC1 was located in the middle on the lower flange of the section I 300 with section factor Am/V=131 m−1. The dominant input into the model is the rate of heat release, which was for such a pile set as 1250 kW m−2, according to the experience of the Cardington tests, see Ref. [14]. The difference reached 70 °C as a maximum to the safe side. The prediction shows high sensitivity to the estimation of the rate of heat release. In the case of primary beams, where the thermocouples TC2 and TC3 were located in the middle on the lower flange of the IPN180 sections with section factor Am/V=180 m−1, see Fig. 25, no difference in predicted and measured values of the maximal temperatures was observed. The shift of the curve of the predicted temperature to the measured one is expected to reflect the small size of the fire compartment, which caused a slower cooling of the structure due to the radiation of heat from the walls. The results show a good accuracy of the prediction of the temperature of the horizontal structural elements heated by local fire, even in a compartment of such small size. The data on the distribution of temperatures within the structure and in particular on connections were collected during a full-scale fire test, which was carried out at the Ammoniac Separator II in the company Mittal Steel Ostrava in June 2006. In this paper, the measured temperatures are summarised and a comparison is made with the simplified analytical methods given in Eurocodes for calculating the temperature in the gas, steel members and connections. From these comparisons it can be concluded that: The methods for calculating the compartment temperature by the parametric fire curve given in Annex A of EN 1991-1-2 compare well with the measured data. The incremental analytical models allow presumption of temperatures of the unprotected beams with a good accuracy. Calculating the temperature of the beam-to-column connection from the measured gas temperature in the fire compartment, based on the mass of the connection parts according to Annex D of EN 1993-1-2, is conservative during the heating phase. A calculation based on the bottom flange temperature of the supported beam is less conservative. The prediction of the temperature of the beam-to-beam connections using the measured gas temperature in the fire compartment, based on the mass of the connection parts, is conservative during the heating phase. The calculation based on the bottom flange temperature of the supported beam may be improved by factor 1.0 instead of 0.88. The work was prepared with support of the Czech Grant Agency GAČR 103/07/1142. This outcome has been achieved with the financial support of the Ministry of Education, Youth and Sports of the Czech Republic, project No. 1M0579, within activities of the CIDEAS Research Centre.