- Email: [email protected]
Temperatures and thermal boundary conditions in reverse channel connections to concrete filled steel sections during standard and natural fire tests
Fire Safety Journal 78 (2015) 55–70
Contents lists available at ScienceDirect
Fire Safety Journal journal homepage: www.elsevier.com/locate/firesaf
Temperatures and thermal boundary conditions in reverse channel connections to concrete filled steel sections during standard and natural fire tests Tomáš Jána a,n, Yong C. Wang b, František Wald a, Kamila Horová a a b
Department of Steel and Timber Structures, Faculty of Civil Engineering, Czech Technical University in Prague, Thákurova 7, 166 29, Prague 6, Czech Republic School of Mechanical, Aerospace and Civil Engineering, University of Manchester, Manchester M13 9PL, UK
art ic l e i nf o
a b s t r a c t
Article history: Received 18 March 2015 Received in revised form 30 July 2015 Accepted 15 August 2015
The paper presents fire test and numerical simulation results of temperature distributions in reverse channel connections to concrete-filled tubular columns during standard and natural fire tests. The experiments included a furnace fire test with the composite frame subjected to increasing fire temperature according to the ISO 834 standard time–temperature curve and two natural fire tests in a full-scale structure including a cooling phase. These fire tests examined the effects of fire protection, different connection locations and different connection dimensions. Temperatures at different connection components were recorded. The recorded connection component temperatures were simulated by using the SAFIR fire engineering software to calibrate the thermal boundary conditions in different connection components. From the numerical simulation results, it has been concluded that radiation to the inner surfaces of the reverse channel and the adjacent part of the column tube is only from the gas volume bounded by these surfaces. A lower flame emissivity value than the standard value should be used. For simplicity, a value of εf ¼ 0.1 is proposed. Also a lower value of the convective heat transfer coefficient of 10 W/m2K can be used in the connection area. & 2015 Elsevier Ltd. All rights reserved.
Keywords: Reverse channel connection Concrete filled tube Heat transfer Fire test Temperature calculation Flame emissivity Numerical model SAFIR
1. Introduction Composite steel and concrete columns – with a concrete filled steel tube or partially-encased open cross-section – are becoming more frequently used in load bearing structures because of their high fire resistance and structural efficiency. In particular, concrete filled steel tubes require no formwork and additional steel reinforcement in concrete may be dispensed with in some cases. However, due to the difficulty of access inside the steel tube, making connections to concrete filled tubular columns is generally more complex. As a result, the fin plate connection is commonly used. Although this type of connection is simple and easy to fabricate, its mechanical performance may not be sufficient in many situations. Directly connecting an endplate to the tube requires specialist bolting technologies such as flowdrill or blind bolts. One relatively new connection type [1–4], which combines good mechanical performance, easy accessibility and moderate fabrication cost, is the reverse channel connection, as illustrated in Fig. 1. This type of connection consists of an end plate (welded to the beam n
Corresponding author. Tel.: þ 420 224 354 762; fax: þ 420 233 337 466. E-mail address: [email protected] (T. Jána).
http://dx.doi.org/10.1016/j.firesaf.2015.08.002 0379-7112/& 2015 Elsevier Ltd. All rights reserved.
and then bolted to the web of a channel) and a channel whose legs are welded to the column tube. The depth of the channel can be varied to accommodate any conventional flexible, flush or extended endplate connection. When a structure is exposed to a fire attack, connections are key components of the structure, particularly concerning the survivability (robustness) of the structure, as demonstrated by the World Trade Center collapse [5] and the structural fire tests at Cardington [6], Ostrava [7] and Mokrsko [8]. However, contemporary European standards EN 1993-1-2 [9] and EN 1993-1-8 [10] provide very limited information on quantifying the connection thermal and mechanical behavior at elevated temperatures during a fire. In other standards and guidelines around the world, there is no or very simple approach [11,12]: when a structure is fire protected, a thickness of the fire protection applied to connections should be equal or greater. During a fire, the behavior of the structure changes as a result of material property degradations and changes of the internal forces due to restrained thermal deformations and large structural deformations [13]. During the heating phase of a fire, exposed beams extend. If this extension is constrained by the adjacent structure, a relatively large compressive axial force develops in the
56
T. Jána et al. / Fire Safety Journal 78 (2015) 55–70
Fig. 1. Schematic of a reverse channel connection of a beam to a concrete-filled tubular column.
beam. Restraint to the thermal bowing of the beam also generates an additional hogging bending moment in the connection. At high temperatures, when vertical structural deformations are very big, axial restraint to the beam results in tension forces in the beam developing catenary action [14]. Connection components may fracture during the catenary action stage. Tension forces may also be generated in the beam and in the connection during the cooling stage. A limited number of fire tests using the reverse channel connection [1] have demonstrated its excellent performance compared to other types of connection to tubular columns. In order to develop methods to quantify the connection behavior, temperatures in different connection components should be obtained. Against this background, the European RFCS COMPFIRE project was implemented, see [15]. An important work package of the project was to determine connection temperatures. This was achieved by fire tests, detailed numerical simulations to broaden the scope of the fire tests and the development of analytical equations to calculate different component temperatures that may be incorporated into design methods such as EN 1993-1-2 [9]. This paper presents detailed results of the fire tests and a limited number of numerical simulations to determine the thermal boundary conditions to different components of the reverse channel connection. The results of this paper are used to develop a design calculation method.
2. Fire tests The fire tests included real fire scenarios in a full-scale experimental structure conducted by the Czech Technical University in Prague and a standard furnace fire test on a smaller composite steel and concrete frame structure at the University of Manchester, United Kingdom. 2.1. Furnace fire test at the University of Manchester A small frame was constructed with several connection arrangements. This frame was then placed inside a fire testing furnace with the directional natural gas burners, measuring about 2 m (depth) by 3.5 m (width) and 3.5 m (height). Fig. 2(a) shows the planar arrangement of the test frame inside the furnace and thermocouple locations. Fig. 2(b) shows a sketch of the test frame with connection reference notations. The ISO 834 standard fire temperature-time curve was followed in the fire test.
Fig. 2. University of Manchester furnace fire test: (a) furnace ground plan; (b) axonometric view of a test specimen with connection notations.
All steel beams were UB 305x165 40 and the steel tubular section filled with C25/30 normal weight concrete was either CHS 244.5 8 or SHS 250 8. A 100 mm thick C25/30 normal weight concrete slab was placed on top of the steel frame. the steelwork was not fire protected and no additional gravity loading was applied to the frame. table 1 gives an overview of the geometry of the monitored connections on the experimental structure. 140 thermocouples were used to record the temperatures in the reverse channel connection components, the concrete slabs and the beams at mid-span. Each thermocouple was pushed in the drilled hole with the depth that was equal to the half of the plate thickness. In order to achieve deeper hole in plates with the thickness less than 10 mm, the hole was drilled through the short weld made on the surface of the plate. The temperature in the furnace was measured by six standard plate thermometers, see Fig. 3, and the average temperature of these six plate thermometers was used as the reference temperature to follow the standard temperature–time curve. The heating phase took two hours and data recording continued for one more hour to obtain temperature data during cooling. During the cooling phase, electric fans were utilized to faster cool the specimen. The average temperature drop in the furnace was about 10 °C/min. This corresponds to the slope of the parametric temperature curve in the cooling phase for fire compartments of common administrative and residential buildings (the shape parameter Γ ¼2.5).
T. Jána et al. / Fire Safety Journal 78 (2015) 55–70
57
Table 1 Configuration of connections in the Manchester furnace fire test; dimensions in [mm]. Designation of connection
Beam cross-section
Column cross-section
End plate (depth/width/thickness)
Reverse channel
Bolts
A1 A2 B2-lower B2-upper B3-lower B3-upper C1 C3
UB 305x165 40
SHS 250 8 SHS 250 8 CHS 244,5 8 CHS 244,5 8 SHS 250 8 SHS 250 8 CHS 244,5 8 CHS 244,5 8
333, 4/200/20
UKPFC 230x90 32
6 x M20
Fig. 3. Measured gas temperatures at different locations in the furnace.
Fig. 4. Positions of thermocouples in the C1 reverse channel connection. Fig. 5. Temperature distributions in the reverse channel connection: (a) comparison between different components and (b) comparison of bolt temperatures.
Fig. 4 shows a typical reverse channel connection and thermocouple locations. The thermocouples were placed on the beam web (TC37 to TC39) at a distance of 200 mm from the endplate, in the center of the shaft of the upper, middle and lower bolts (TC31 to TC33), on the endplate (TC34 to TC36), on the reverse channel (TC40 to TC42) and the steel tube of the concrete-filled tubular column (TC43). Fig. 5(a) presents a typical example of temperature distributions in the reverse channel connection components. Higher temperatures were logically measured in the beam web, at a location far from the connection. However, the different connection components (reverse channel, end plate, bolts) experienced similar temperatures. The tube temperature was lower due to the heat sink effect of the infill concrete. Fig. 5(b) shows the temperatures measured in three different bolts (i.e. top, middle and bottom). They are also very close to each other. Similar temperature developments were recorded in the other connections (central, corner, edge). Fig. 6 shows a comparison of beam web temperatures measured on different locations of the structure. Differences in the
Fig. 6. Measured beam web temperatures at different locations of the structure.
58
T. Jána et al. / Fire Safety Journal 78 (2015) 55–70
temperatures of the structure are mainly caused by the different gas temperatures across the furnace. 2.2. Fire tests in a full-scale structure at the Czech Technical University in Prague An experimental two-storey composite steel and concrete structure with dimensions of 10.4 13.4 m2 and a height of 9 m, representing part of an administrative building, see Fig. 7, was constructed for the COMPFIRE project. Fig. 7(a) gives the serial sizes of the beams and columns. The tubular columns were un-
protected but filled with concrete. Fig. 8 shows the construction of the upper and lower composite floors. Table 2 shows the main dimensions of the reverse channel connections. The load bearing structure was partially fire protected. The PROMASPRAY F250 spray, a mixture of mineral fibers and cement binders with low density, was used as fire protection. On the second floor, the connections of central beams with the span of 9 m were covered with 20 mm of fire protection in order to obtain temperature differences in the unprotected and fire protected connections of the unprotected structure. In addition, cross bracings and secondary columns were fire protected because of
Fig. 7. Structure of the fire test building at CTU: (a) ground plan; (b) vertical section; (c) picture of the building.
Fig. 8. Construction of composite floors: (a) upper slab; (b) lower slab.
T. Jána et al. / Fire Safety Journal 78 (2015) 55–70
59
Table 2 Configurations of connections of the building fire test at CTU; dimensions in [mm]. Designation of connection First fire test (second floor) A2-B2 to A2 A2-B2 to B2 B2-C2 to B2 B2-C2 to C2 Second fire test (first floor) A2-B2 to A2 A2-B2 to B2 B2-C2 to B2 B2-C2 to C2
Beam cross-section
Column cross-section
End plate
Reverse channel
Bolts
Designed thickness of fire protection
IPE 270
TR 245/8
165/160/8
sheet 165/200/8 UPE 160 UPE 120 sheet 135/160/8
4x M16
20 mm
4x M12
165/160/8
sheet 165/200/8
4x M16
200/180/8
sheet 200/220/8 UPE 180
4x M20
60 mm
IPE 220
IPE 270
135/120/8
TR 245/8
IPE 330
stability of the bearing structure and cladding. On the first floor, 60 mm sprayed fire protection was again applied on the chosen connections, secondary columns and cross bracing. Due to the expected achievement of the membrane action of the composite floor during the second fire test, all edge beams and corner columns were also fire protected. The reverse channel connections were protected by a length of 250 mm measured from the edge of the column in all cases. The composite steel and concrete columns were unprotected. The measurements involved 97 thin thermocouples of type K to measure the gas temperatures (TG, 20 pcs., diameter 3 mm), temperatures at the mid-span of the beams (TB, 8 pcs., diameter 2 mm), at various locations in the reverse channel connections and the fin plate connections (TC, 50 pcs., diameter 2 mm), the composite slabs (TS, 7 pcs., diameter 2 mm) and the columns (TSG, 12 pcs., diameter 2 mm) for each fire test. 7 standardized plate thermometers were also used to obtain the adiabatic surface temperatures. Fig. 9 shows the thermocouple locations. Two fire tests were carried out, one on 6th September 2011 and one on 15th September 2011. Timber cribs were used as the fire source for both fire tests. For the first fire test on the second floor, fire load was created by 8 3 piles of wooden cribs of 50 50x1000 mm3 as shown in Fig. 10. All spruce cribs were dried to a moisture content of 12%. Each pile consisted of 6 layers with 7 cribs, a total of 42 cribs. The total volume of the timber cribs was 2.52 m3, to give a fire load density (per floor area) of 9.9 kg/m2 or 173.5 MJ/m2. A linear ignition source placed on the left-hand side was made of a thin-walled channel filled with mineral wool and wetted by paraffin. The ventilation of the fire compartments was
Fig. 9. Position of thermocouples for both building tests (first fire tests/second fire test).
enabled by window openings with dimensions of 5x2 m2 In the case of the second fire test on the first floor, 28 wooden piles were distributed on the ground area of the fire compartment in accordance with Fig. 11(a). Each pile consisted of 11 layers with 10 cribs. The total volume of wood was 7.65 m3, giving a fire load per floor area of 30.1 kg/m2 or 525 MJ/m2. All piles were ignited at the same time by the aid of the thin-walled channel filled with mineral wool and paraffin, see Fig. 11(b). Fig. 12(a) and (b) shows the recorded gas temperatures at different locations inside the fire compartment, 300 mm under the ceiling slab, during the first fire test. A maximum gas temperature of 979 °C was recorded at 26th min of the fire test. Because there are only small differences between the temperature measured by the thermocouple of type K and calculated adiabatic surface temperature [16] based on the measurement of the plate thermometers, see Fig. 12(c), the temperatures from thermocouples can be used to calculate heat transfer to the structure. Comparison of the gas temperature results with prediction models is in [17]. The thermal response of the structure depended on the spatial development of the fire, see Fig. 13. The development of the gas temperature during the second fire test corresponded to the fire scenarios with flashovers in the fire compartment of a typical administrative building. As illustrated in Fig. 14(a) and (b), in the initial stage of the fire, higher gas temperatures were observed in the front of the compartment (temperature differences of up to 200 °C at 30th min). In contrast, after the full development of the fire, the area with higher gas temperatures was shifted to the rear of the fire compartment (temperature differences of about 150 °C at 50th min). A maximum gas temperature of 1002 °C was reached in the rear of the fire compartment at 46th min of the fire test. During the cooling period, temperatures were distributed more uniformly. The temperatures from thermocouples can be used to calculate heat transfer to the structure because they are in good agreement with the adiabatic surface temperatures, see Fig. 14(c). As can be seen from the comparison of prediction models in Fig. 14(d), both the parametric fire curve [18] and the zone model [19] give the good results for the heating phase and maximum temperature, but the zone model underestimates the gas temperature during the cooling phase. The thermal response of the whole structure during the second fire tests is demonstrated in Fig. 15. 2.2.1. Temperatures of unprotected connections Two typical examples of temperature distributions in the unprotected reverse channel connection and connected members are presented for discussion. Fig. 16 shows the beam-to-central column connection B2-C2/B2 on the second floor. There was no fire protection to the structure. The reverse channel was made of UPE 120 (flange width 60 mm, flange thickness 8.0 mm, web width 120 mm, web thickness 5.0 mm), with a height of 135 mm. Both flanges of the channel section were welded to the column with a
60
T. Jána et al. / Fire Safety Journal 78 (2015) 55–70
Fig. 10. Provision of fire load: (a) distribution on the floor; (b) preparation of ignition.
Fig. 11. Position of timber piles on the 1st floor of the experimental building: (a) ground plan with the scheme of ignition; (b) ignition.
half V weld (throat thickness 8 mm). The end plate of the beam was made of 8 mm plate 135 mm depth x 120 mm width, welded to the web of the IPE 220 beam with a fillet weld (throat thickness 4 mm) on both sides of the web and connected to the reverse channel web by four M12 bolts. Fig. 16 also shows the locations of six thermocouples on the connection. Fig. 17 compares the measured temperatures in different parts of the connection, including the temperature of the bottom beam flange at mid-span. It is clear that the connection temperatures were significantly lower than the temperature of the beam flange. The column tube and weld of the reverse channel flange and the column tube had lower temperatures due to the heat sink effect of the infill concrete. However, the other connection component temperatures (lower bolt, reverse channel flange) were quite similar, with the maximum difference being less than 60 °C. Fig. 18 shows the beam-to-central column reverse channel connection A2-B2/B2 without fire protection on the first floor and the positions of seven thermocouples on this connection. The reverse channel was cut out of a square tube TR 200/200/8 (flange width 86 mm, web width 200 mm, flange and web thickness 8.0 mm), with a height of 165 mm. Both flanges of the channel section were welded to the column with a half V weld (throat thickness 8 mm). The end plate was made of 8 mm plate with a depth of 165 mm and a width of 160 mm, welded to the web of the beam IPE 270 with a fillet weld (throat thickness 4 mm) and
connected to the reverse channel by four M16 bolts. Fig. 19(a) compares the temperatures measured at different connection components with the beam lower flange temperature. The temperature distributions were very similar to those shown in Fig. 17. The connection component (reverse channel flange and bolt) temperatures were lower than the temperature of the beam lower flange, but they were close to each other. Fig. 19 (b) compares the temperatures in different bolts and it shows similar temperatures. The conclusions of these two fire tests with regard to temperature distributions in different connection components are the same as those from the University of Manchester furnace fire test. 2.2.2. Temperatures of fire protected connections Temperatures were also measured on fire protected reverse channel connections. Two typical examples are presented here. Fig. 20 shows the fire protected beam-to-central column connection A2-B2/B2 on the second floor. The reverse channel was made of UPE 160 (flange width 70 mm, flange thickness 9.5 mm, web width 160 mm, web thickness 5.5 mm), with a height of 165 mm. Both flanges of the channel section were welded to the column with a half V weld (throat thickness 9.5 mm). The end plate was made of 8 mm plate, 165 160 mm2 in size, welded to the web of the beam IPE 270 with a fillet weld (throat thickness 4 mm) on both sides of the web and connected to the reverse channel by
T. Jána et al. / Fire Safety Journal 78 (2015) 55–70
61
Fig. 12. Gas temperatures at the first fire test: (a) measured temperatures at the front part of the compartment; (b) measured temperatures at the back part of the compartment; (c) comparison with the adiabatic surface temperature.
Fig. 13. Measured bottom flange temperatures at mid-span of the beams from the first fire test.
four M16 bolts. The locations of thermocouples are similar to both cases presented above, see Fig. 20(a). The connection was protected for a length of 250 mm measured from the edge of the column. The thickness of the fire protection was variable between 30 and 60 mm, as shown in Fig. 20(b). Fig. 21 compares the component temperatures (lower bolt, reverse channel flange) of the protected connection. Again, the connection component temperatures were very similar, with the maximum difference of less than 50 °C. The comparison between the results in Fig. 21 and those in Fig. 17 indicates that under a similar gas temperature in the fire compartment and a similar lower flange temperature of the connected (unprotected) beam, the temperatures in the protected connection components were by about 300 °C lower than the unprotected connection component temperatures. This suggests that the length of the fire protection used in the test was sufficient for significantly
reducing the heat conduction from the unprotected beam. Fig. 22 shows the fire protected connection B2-C2/B2 on the first floor in the second fire test. The reverse channel was cut out of a square tube TR 220/220/8 (flange width 98 mm, web width 200 mm, flange and web thickness 8.0 mm), 200 mm high. The end plate was 8 mm thick, 200 mm in depth and 180 mm in width, welded to the web of the IPE 330 beam with a fillet weld (throat thickness 4 mm) on both sides of the web and connected to the reverse channel by four M20 bolts. Fig. 22(a) shows the locations of five thermocouples. The thicknesses of the fire protection measured before the second fire test are shown in Fig. 22(b). The temperatures measured in the fire protected reverse channel connection B2-C2/B2 during the second fire test are shown in Fig. 23. From the comparison with the temperatures in the similar connection without fire protection from Fig. 19(a), it is clear that the applied fire protection with an average thickness of 60 mm was able to reduce the temperatures in the connection by 400 °C. This leads to a significant increase in the connection fire resistance, see [20]. Once again, the temperatures in the reverse channel, in the endplate and in the bolts, may be considered to be the same. The tube temperature was lower due to the heat sink effect of the filled concrete. These conclusions are consistent with those of [21], which are based on results of fire tests of the simple frame of one steel beam and two composite columns in the furnace. This means that the temperature distribution in the reverse channel connection is notably affected by neither the presence of the concrete slab above the connections, nor the connection of more beams to the column at the same area, nor the change in the fire scenario.
3. Numerical model The fire tests described in the previous section have been
62
T. Jána et al. / Fire Safety Journal 78 (2015) 55–70
Fig. 14. Gas temperatures at the second fire test: (a) measured temperatures at the front part of the compartment; (b) measured temperatures at the back part of the compartment; (c) comparison with the adiabatic surface temperature; (d) comparison with the predicted gas temperatures.
Fig. 15. Measured bottom flange temperatures at mid-span of the beams from the second fire test.
Fig. 17. Comparison of measured temperatures in different components of the unprotected reverse channel connection B2-C2/B2 with the beam bottom flange temperature at mid-span.
Fig. 16. Position of thermocouples in the unprotected reverse channel connection B2-C2/B2.
T. Jána et al. / Fire Safety Journal 78 (2015) 55–70
63
Fig. 18. Position of thermocouples in the unprotected reverse channel connection A2-B2/B2.
tube of the composite column outside the connection area for the Czech Technical University tests (hc ¼ 35 W/m2K, εm ¼0.3). In the furnace test, there was no measurement of the column temperature outside the connection area. The PROMASPRAY F250 fire protection was used in the fire tests performed by CTU. This fire protection consists of 80% mineral wool fibers and 20% Portland cement [24]. The density of the material was assumed to be ρp ¼264 kg/m3[25]. This value corresponds with the density of sprayed mineral fibers introduced in [26]. Based on [26], the specific heat of the fire protection material was assumed to be cp ¼1050 J/kg K. Because the influence of the thermal capacitance (density multiplied by specific heat) of the lightweight fire protection materials on the resultant steel temperature is very small [27], the density and the specific heat value can be taken as constant. On the contrary, it is more important to obtain accurate temperature dependent thermal conductivity values. The temperature dependent thermal conductivity of the fire protection based on the mineral wool is according to [27], using the following equation:
⎛ T ⎞3 ⎟ λmineral _ wool = 0.03 + 0.2438 ⎜ ⎝ 1000 ⎠
(1)
where T is in Kelvin. Eq. (1) was modified so that it fits with the declared thermal conductivity of 0.043 W/mK at 24 °C by the manufacturer’s datasheet [25] for PROMASPRAY F250. The final thermal conductivity used in the numerical simulations was
⎛ T ⎞3 ⎟ λ p, T = 0.037 + 0.2438 ⎜ ⎝ 1000 ⎠
Fig. 19. Measured temperatures in the unprotected reverse channel connection A2B2/B2: (a) comparison between different connection components; (b) comparison of bolt temperatures.
modeled using the specialist fire engineering software SAFIR [22]. Level of spatial discretization used for the models of the connections is shown in Figs. 31, 34 and 37. The purpose of this simulation was to check the sensitivity of the simulation results to different assumptions of thermal boundary conditions so as to identify a suitable thermal boundary condition for calculating connection temperatures. 3.1. Material thermal properties The measured density of concrete was ρc ¼2250 kg/m3. The other thermal properties (thermal conductivity, density and specific heat) of steel and concrete were according to Eurocodes EN 1993-1-2 [9] and EN 1994-1-2 [23]. To demonstrate the applicability of these standard values, Fig. 24 shows the comparison between the measured and simulated temperatures of the steel
(2)
Fig. 25 shows the graphical representation of the dependence of thermal conductivity on temperature in degrees Celsius. The comparison of the measured and simulated beam temperatures in Fig. 26 demonstrates the applicability of the above values. 3.2. Thermal boundary condition In numerical heat transfer simulations of the University of Manchester test, the gas temperature was assumed equal to the measured temperature from the closest TG thermocouple in the furnace. Because the gas temperatures were not measured directly at the connections during the Czech Technical University tests, the average of the measured temperatures from the thermocouples around to the monitored connection was assumed as the gas temperature in the numerical model. The unexposed surface was in contact with air of a constant temperature of 20 °C. Because the convective heat transfer conditions were not measured, the design values of the convective heat transfer coefficient hc ¼25 W/m2K for surfaces subjected to the standard fire condition and hc ¼35 W/m2K for the natural fire condition according to EN 1991-1-2 [18] were used. However, recent study [28]
64
T. Jána et al. / Fire Safety Journal 78 (2015) 55–70
Fig. 20. Fire protected reverse channel connection A2-B2/B2: (a) position of thermocouples; (b) thicknesses of fire protection.
Fig. 21. Comparison of measured temperatures in different components of the fire protected reverse channel connection A2-B2/B2.
Fig. 23. Comparison of measured temperatures in different components of the fire protected reverse channel connection B2-C2/B2 with the beam bottom flange temperature in mid-span.
Fig. 22. Fire protected reverse channel connection B2-C2/B2: (a) position of thermocouples; (b) thicknesses of fire protection.
T. Jána et al. / Fire Safety Journal 78 (2015) 55–70
65
Fig. 24. Comparison of measured and simulated temperatures of the steel tube of the composite column.
Fig. 25. Thermal conductivity of PROMASPRAY F250 fire protection material in relation to temperature.
Fig. 27. Influence of different steel surface emissivity values on the beam bottom flange temperature: (a) University of Manchester test; (b) CTU test.
of the lower flange at mid span of the beam indicate, that the real surface emissivity of the steel is lower in these cases, see Fig. 27 (a) for the University of Manchester test and Fig. 27(b) for the CTU test. Base on the equation of the Lumped Capacitance Method for unprotected steel members in [9], the inverse calculation of the surface emissivity can be expressed as
εm =
Fig. 26. Comparison of the measured and simulated temperatures of the web of the fire protected beam C1-C2 at mid-span (the second building fire test).
suggests that it is appropriate to consider lower values of the convective heat transfer coefficient for the connection area due to a slower fire gas movement. For the surfaces in contact with the ambient temperature, the convective heat transfer coefficient was hc ¼ 4 W/m2K. According to EN 1993-1-2 [9], the design value of the surface emissivity of carbon steel is εm ¼0.7 and according to EN 1992-1-2 [29], the surface emissivity of concrete is εm ¼0.7. However, the comparisons between the calculated and measured temperatures
ca ⋅ ρa ∂θa ⋅ Am / V ∂t
− hc ⋅( θg − θa )
4 4⎤ ⎡ Φ⋅εf ⋅σ⋅⎣⎢ ( θg + 273) − ( θa + 273) ⎥⎦
(3)
where ca and ρa are the specific heat capacity and the density of steel, Am/V is the section factor of the lower beam flange, ∂θa is the increment of the steel temperature in time ∂t , hc is the convective heat transfer coefficient (hc ¼35 W/m2K for a natural fire and hc ¼25 W/m2K for a standard fire), θg is the measured gas temperature, θa is the measured temperature of the lower beam flange, Φ is the configuration factor (taken as 1.0 because the steel surface is in direct contact with the fire), εf is the fire gas emissivity for the whole furnace or compartment (taken as 1.0), s is the Stephan-Boltzmann constant ( ¼5.67 10 8 W/(m2K4)). Using the inverse calculation for the several beams in each fire test, the surface emissivity of the steel sections was specified as ε m ¼ 0.5 in the case of the University of Manchester furnace test and εm ¼ 0.3 in the case of the CTU tests. The surface emissivity is considered as a constant value, but it seems to be sufficient to predict the temperature of the structure. The calculated lower flange temperatures based on the obtained values of the surface emissivity are in good agreement with the measured temperatures, see Fig. 27. On the contrary, the change of the convective
66
T. Jána et al. / Fire Safety Journal 78 (2015) 55–70
according to EN 1991-1-2 [18], is εf ¼ 1.0. This is based on the assumption that the fire depth is very large. A lower flame emissivity should be used for the inner surface of the reverse channel connection, because the depth of the fire gas is small, see Fig. 28. The total emissivity of hot gas εf depends on the coefficient of radiation K and the thickness of gas, as follows [30]:
εf = 1 − e−KL
(4)
where L is the mean beam length of the gas. For gas – smoke from wood-based materials, the coefficient of radiation is K ¼0.8 [31]. For any general geometry of the gas volume, the mean beam length can be approximately expressed as
L=C
4V A
(5)
where V is the total volume of the gas, see Fig. 28(a), A is the area on which heat is radiated, see Fig. 28(b) and C is a correction factor (taken as 0.9) which shall be applied for the optically thick volume of the gas [32]. On the basis of the above equations, the fire gas emissivity inside the reverse channel for each reverse channel connection surface used in the simulations was calculated and the results are given in Tables 3 and 4.
4. Calibration of the numerical model 4.1. Unprotected connections To demonstrate the accuracy of the assumptions mentioned in Section 3 for thermal boundary conditions and fire protection thermal properties, the numerical simulation results are compared to the measured temperatures for three examples. One for the unprotected connection from the University of Manchester furnace test and two for the connections (unprotected and fire protected) from the CTU natural fire tests. More comparisons of numerical and test results can be found in [33].
Fig. 28. Volume (a) and area (b) for the calculation of the mean beam length of the gas layer between the reverse channel and the column tube.
heat transfer coefficient has small effect on the resulting temperatures of the beam flange (particularly at the high temperatures). The surface emissivity of the lower side of the composite slab, which was formed by zinc-coated trapezoidal sheets, was ε m ¼0.23, according to [27]. The recommended design value for the emissivity of the fire,
4.1.1. Comparison against the University of Manchester furnace fire test Fig. 29 shows the effects of using different convective heat transfer coefficients and emissivities of fire on component temperatures of the connection B2-lower. For the simulation, the measured temperature of thermocouple TG5 was used as the gas temperature outside as well as inside the reverse channel connection. The simulated temperature-time developments of the end plate and the reverse channel flange are compared with the temperatures measured during the University of Manchester furnace fire test. It is clear that the numerical results agree with the
Table 3 Fire gas emissivity inside the reverse channel connections from the University of Manchester test. Connection ID
V [m3]
A [m2]
C [dimensionless]
L [m]
K [dimensionless]
εf [dimensionless]
B2-lower, B2-upper, C1, C3 A1, A3, B3-lower, B3-upper
3.379 10 3 5.534 10 3
19.82 10 2 18.97 10 2
0.9 0.9
6.14 10 2 10.5 10 2
0.8 0.8
0.0479 0.0806
Table 4 Fire gas emissivity inside the reverse channel connections from the CTU tests. Connection ID
V [m3]
A [m2]
C [dimensionless]
L [m]
K [dimensionless]
εf [dimensionless]
B2-C2/B2 A2-B2/B2, A2-B2/A2
0.651 10 3 1.44 10 3
4.06 10 2 8.50 10 2
0.9 0.9
5.77 10 2 6.08 10 2
0.8 0.8
0.0451 0.0475
T. Jána et al. / Fire Safety Journal 78 (2015) 55–70
67
Fig. 30. Comparison between measured and simulated temperatures of the beam web and the column tube for the B2-lower connection in the University of Manchester furnace fire test.
Fig. 29. Influence of different emissivities of fire and convective heat transfer coefficients on the temperatures of the B2-lower connection in the University of Manchester furnace fire test: (a) end plate; (b) reverse channel flange.
test results best when the convective heat transfer coefficient is hc ¼10 W/m2K and the flame emissivity inside the reverse channel is calculated according to the approach described above (because of the same steel surface as on the beam, the surface emissivity ε m ¼0.5 is unchanged). Change in the convective heat transfer mainly affects low component temperatures and there are biggest differences in this range when using hc ¼ 25 W/m2K. Different geometry of the structure and different surrounding flow conditions in the connection area can be the reasons for using the lower value of the convective heat transfer coefficient [30]. Same boundary conditions (with hc ¼10 W/m2K) are also used for the simulation of component temperatures in Fig. 30. The agreement between the simulation and measured temperatures is excellent for all four connection components, including both the heating and cooling phases. During the heating phase, the maximum temperature difference between the measured and calculated values is not more than 30 °C. During the cooling phase, the maximum temperature difference is about 50 °C. Fig. 31 presents the temperature profile at 60th min. The uniformity of the temperature distribution in the reverse channel and the endplate is clearly shown. 4.1.2. Comparison against CTU building fire tests Figs. 32 and 33 show the comparison of the simulated and measured temperatures of the components of the connection A2B2/B2. The gas temperature in the numerical model was equal to the average temperature of thermocouples TG23, TG24, TG29 and TG30. As shown in Fig. 32, the simulation results agree with the
test results best when the convective heat transfer coefficient is hc ¼5 W/m2K and the flame emissivity inside the reverse channel is used according to Table 4 (steel surface emissivity is εm ¼0.3). Same boundary conditions are used for the simulation in Fig. 33. The predicted temperatures are in good agreement with the measured temperatures, especially during the heating phase. Larger temperature differences, of up to 100 °C, can be observed in the cooling phase. In general, the temperatures from the numerical model were higher than the measured temperatures. This difference was caused by a change in the gas flow close to the connection after the loss of integrity of the composite slab above this connection at 54th min of the second compartment fire test. The visualization of the temperature distribution on the three dimensional model of the reverse channel connection at 57th min is shown in Fig. 34. 4.2. Fire protected connections The calibration of the numerical model of a fire protected reverse channel connection is performed for connection A2-B2/B2 to the central column in the first compartment fire test. The gas temperature in the numerical model was equal to the average temperature of thermocouples TG3, TG4, TG9 and TG10. In the case of the fire protected connection, the effect of using different convective heat transfer coefficients on component temperatures is negligible, see Fig. 35. The results in Figs. 35 and 36 show that the temperatures obtained from the numerical model are in good
Fig. 31. Simulation results of the temperature field in the B2-lower connection at 60th min of the furnace fire test.
68
T. Jána et al. / Fire Safety Journal 78 (2015) 55–70
agreement with the measured temperatures. The simulated temperatures of the beam web in the connection area, the end plate, the lower bolt and the reverse channel flange differ from the measured temperatures by a maximum of 20 °C. Fig. 37 shows the visualization of the temperature field of the same connection at 29th min of the fire test. A uniform temperature distribution may be assumed for the connection design.
5. Summary and conclusions This paper has presented temperature distributions measured in reverse channel connections to concrete filled hollow tubes in
Fig. 34. Simulation results of the temperature field in the connection A2-B2/B2 at 57th min of the second building fire test.
Fig. 32. Influence of different emissivities of fire and convective heat transfer coefficients on the temperatures of the unprotected connection A2-B2/B2 in the second building fire test: (a) lower bolt; (b) reverse channel flange.
Fig. 35. Influence of different convective heat transfer coefficients on the temperature of the end plate and the beam web close to the end plate of the fire protected connection A2-B2/B2 in the first building fire test.
Fig. 33. Comparison between measured and simulated temperatures for the unprotected reverse channel connection A2-B2/B2 in the second building fire test.
Fig. 36. Comparison between measured and simulated temperatures of the lower bolt and the reverse channel flange for the fire protected connection A2-B2/B2 in the first building fire test.
T. Jána et al. / Fire Safety Journal 78 (2015) 55–70
69
for a real compartment fire exposure respectively were shown to give more accurate temperature calculation results. For simplicity, a value of hc ¼10 W/m2K may be used.
Acknowledgments The research was supported by funding from the European Community’s Research Fund for Coal and Steel (RFCS) under the Grant agreement no. RFSR-CT-2009-00021 (COMPFIRE) and the European Social Fund within the framework of realizing the project “Support of Inter-sectoral Mobility and Quality Enhancement of Research Teams at the Czech Technical University in Prague”, CZ.1.07/2.3.00/30.0034, period of the project's realization 1/12/ 2012–30/6/2015.
Fig. 37. Simulation results of the temperature field in the fire protected reverse channel connection A2-B2/B2 at 29th min of the first building fire test.
one standard furnace test and two realistic building fire tests. The fire tests were simulated using the SAFIR fire engineering software. The main objective of the numerical simulations was to determine the relevant thermal boundary conditions for the connections. The following conclusions can be drawn: (1) In the unprotected connections, temperature of the bolts, end plate, reverse channel and welds of the end plate and the beam web may be considered as uniform and lower than the beam temperature. Temperature of welds of the reverse channel flange and column tube may be considered lower than the reverse channel flange temperature. The temperature of the column tube and the adjacent welds is largely influenced by the concrete core in the composite column, which takes heat from the connection during the heating phase of fire. (2) The temperature distribution in the reverse channel connection is notably affected by neither the presence of the concrete slab above the connections, nor the connection of more beams to the column at the same area, nor the change in the fire scenario. (3) The temperature distribution in the all components of the protected connections may be considered uniform in the connection design. (4) The connection location within the fire compartment has no influence on the connection temperatures, except for local fire gas temperatures. (5) If the connection is fire protected but the connected beam is not, the heat conduction from the unprotected beam to the connected connection is negligible if there is a small length of fire protection to the steel beam adjacent to the connection. (6) For predicting connection component temperatures, the fire gas volume enclosed by the reverse channel and the column surface should be used to calculate the gas emissivity. This value (typically less than 0.1) is much lower than the recommended fire gas emissivity in EN 1991-1-2. For simplicity, a value of εf ¼0.1 may be used. (7) Convective and radiative heat transfer conditions were not measured at the fire tests, therefore the steel surface emissivity was obtained by the inverse calculation from the measured temperatures and the convective heat transfer coefficient was assumed as the standard value. But the calibration of the numerical model clearly implies that the convective heat transfer coefficient in the connection region should be lower than those recommended in EN 1991-1-2 (25 and 35 W/m2K respectively for a standard furnace fire and a real fire exposure). The values of hc ¼ 10 W/m2K for a furnace fire exposure and hc ¼5 W/m2K
References [1] J. Ding, Y.C. Wang, Experimental study of structural fire behaviour of steel beam to concrete filled tubular column assemblies with different types of joints, Eng. Struct. 29 (12) (2007) 3485–3502. [2] S. Elsawaf, Y.C. Wang, P. Mandal, Numerical modelling of restrained structural subassemblies of steel beam and CFT columns connected using reverse channels in fire, Eng. Struct. 31 (2011) 1217–1231. [3] S. Elsawaf, Y.C. Wang, Methods of improving the survival temperature in fire of steel beam connected to CFT column using reverse channel connection, Eng. Struct. 32 (2012) 132–146. [4] S. Elsawaf, Y.C. Wang, Behaviour of restrained structural subassemblies of steel beam to CFT column in fire during cooling stage, Eng. Struct. 46 (2013) 471–492. [5] FEMA (Federal Emergency Management Agency), World Trade Center Building Performance Study: Data Collection, Preliminary, Observations, and Recommendations, Washington, DC, 2002. [6] F. Wald, L. Simões da Silva, D.B. Moore, T. Lennon, M. Chladná, A. Santiago, M. Beneš, L. Borges, Experimental behaviour of a steel structure under natural fire, Fire Saf. J. 41 (7) (2006) 509–522. [7] F. Wald, J. Chlouba, A. Uhlíř, P. Kallerová, M. Štujberová, Temperatures during fire tests on structure and its prediction according to Eurocodes, Fire Saf. J. 44 (1) (2009) 135–146. [8] F. Wald, P. Kallerová, J. Chlouba, Z. Sokol, M. Strejček, Fire test on an administrative building in Mokrsko: Summary of the Results from the Fire Experiment, Printing house Česká technika, CTU in Prague, 2010. [9] EN 1993-1-2, Eurocode 3: Design of Steel Structures, Part 1.2: General Rules – Structural Fire Design, CEN, European Committee for Standardization, Brussels, 2005. [10] EN 1993-1-8: Eurocode 3:, Design of steel structures-Part 1–8: Design of joints, CEN, European Committee for Standardization, Brussels, 2005. [11] International Building Code, Falls Church, International Code Council (ICC), 2006. [12] J.P. England, S.A. Young, M.C. Hui, N. Kurban, Guide for the Design of Fire Resistant Barriers and Structures, Building Control Commission, Melbourne, 2000. [13] Y.C. Wang, The importance of considering whole structural behaviour in fire, in: Proceedings of the 3rd Structures in Fire Seminar, Singapore and Hong Kong, international conference, invited presentation, 2002. [14] Y. Yin, Y.C. Wang, Analysis of catenary action in steel beams under fire conditions, Prog. Steel Build. Struct. 6 (3) (2004) 1–8. [15] L. Simões da Silva, A. Santiago, F. Lopes, M. Veljkovic, T. Heistermann, N. Iqbal, F. Wald, T. Jána, B. Davison, I. Burgess, S.S. Huang, G. Dong, Y. Wang, P. Mandal, M. Jafarian, M. Skořepa, P. Velda, G. Koutlas, COMPFIRE-Design of composite joints for improved fire robustness: final report, Research Programme of the Research Fund for Coal and Steel, Brussels (2013). [16] U. Wickström, R. Jansson, H. Tuovinen, Validation fire tests on using the adiabatic surface temperature for predicting heat transfer, SP Technical Research Institute of Sweden (2009). [17] K. Horová, T. Jána, F. Wald, Temperature heterogeneity during travelling fire on experimental building, Adv. Eng. Softw. 62–63 (2013) 119–130. [18] EN 1991-1-2, Eurocode 1, Actions on structures-Part 1-2: General actions – Actions on structures exposed to fire, CEN, European Committee for Standardization, Brussels, 2002. [19] J.F. Cadorin, J.M. Franssen, D.I. Pintea, L.G. Cajot, M. Haller, J.B. Schleich, OZone V2.2.5 University of Liege and Arcelor Profil Luxembourg Research Centre, 2006. [20] T. Jána, F. Wald, Reduction of connection resistance during Veselí fire tests, J. Struct. Fire Eng. 6 (1) (2015) 21–27. [21] J. Ding, Y.C. Wang, Temperatures in unprotected joints between steel beams and concrete filled tubular columns in fire, Fire Saf. J. 44 (1) (2009) 16–32. [22] SAFIR (software), version 2011.b.3, ArGEnCo, University of Liege, 2011. [23] EN 1994-1-2, Eurocode 4, Design of Composite Steel and Concrete Structures,
70
[24] [25] [26]
[27] [28]
T. Jána et al. / Fire Safety Journal 78 (2015) 55–70 Part 1.2: General Rules – Structural Fire Design, CEN, European Committee for Standardization, Brussels, 2005. s Promat, PROMASPRAY F250: Fire protective spray on structures, Safety data sheet, revision date: 01/06/2015. s Promat, PROMASPRAY F250: Fire protective spray on structures, Technical data sheet. R.M. Lawson, G.M. Newman, Structural fire design to EC3 & EC4, and comparison with BS 5950, Technical Report, SCI Publication 159, The Steel Construction Institute (1996). Y.C. Wang, I. Burgess, F. Wald, M. Gillie, Performance-Based Fire Engineering of Structure, CRC Press, Boca Raton, 2012. X.H. Dai, Y.C. Wang, C.G. Bailey, Temperature distributions in unprotected steel connections in fire, in: Y.C. Wang, C.K. Choi (Eds.), Proceedings of 3rd
[29]
[30] [31] [32] [33]
International Conference on Steel and Composite Structures, Taylor & Francis, 2007, pp. 535–540. EN 1992-1-2: Eurocode 2, Design of concrete structures-Part 1-2: General rules – Structural fire design, CEN, European Committee for Standardization, Brussels, 2004. U. Wickström, Heat Transfer in Fire Technology, Luleå Tekniska Universitet (2012). D. Drysdale, An Introduction to Fire Dynamics, 3rd Edition, John Wiley & Sons, Chichester, 2011. M.Q. Brewster, Thermal Radiative Transfer and Properties, John Wiley & Sons, Inc., 1992. T. Jána, Teplota přípoje U profilem a čelní deskou při požáru, PhD. Thesis, Czech Technical University in Prague (2014).
Contents lists available at ScienceDirect
Fire Safety Journal journal homepage: www.elsevier.com/locate/firesaf
Temperatures and thermal boundary conditions in reverse channel connections to concrete filled steel sections during standard and natural fire tests Tomáš Jána a,n, Yong C. Wang b, František Wald a, Kamila Horová a a b
Department of Steel and Timber Structures, Faculty of Civil Engineering, Czech Technical University in Prague, Thákurova 7, 166 29, Prague 6, Czech Republic School of Mechanical, Aerospace and Civil Engineering, University of Manchester, Manchester M13 9PL, UK
art ic l e i nf o
a b s t r a c t
Article history: Received 18 March 2015 Received in revised form 30 July 2015 Accepted 15 August 2015
The paper presents fire test and numerical simulation results of temperature distributions in reverse channel connections to concrete-filled tubular columns during standard and natural fire tests. The experiments included a furnace fire test with the composite frame subjected to increasing fire temperature according to the ISO 834 standard time–temperature curve and two natural fire tests in a full-scale structure including a cooling phase. These fire tests examined the effects of fire protection, different connection locations and different connection dimensions. Temperatures at different connection components were recorded. The recorded connection component temperatures were simulated by using the SAFIR fire engineering software to calibrate the thermal boundary conditions in different connection components. From the numerical simulation results, it has been concluded that radiation to the inner surfaces of the reverse channel and the adjacent part of the column tube is only from the gas volume bounded by these surfaces. A lower flame emissivity value than the standard value should be used. For simplicity, a value of εf ¼ 0.1 is proposed. Also a lower value of the convective heat transfer coefficient of 10 W/m2K can be used in the connection area. & 2015 Elsevier Ltd. All rights reserved.
Keywords: Reverse channel connection Concrete filled tube Heat transfer Fire test Temperature calculation Flame emissivity Numerical model SAFIR
1. Introduction Composite steel and concrete columns – with a concrete filled steel tube or partially-encased open cross-section – are becoming more frequently used in load bearing structures because of their high fire resistance and structural efficiency. In particular, concrete filled steel tubes require no formwork and additional steel reinforcement in concrete may be dispensed with in some cases. However, due to the difficulty of access inside the steel tube, making connections to concrete filled tubular columns is generally more complex. As a result, the fin plate connection is commonly used. Although this type of connection is simple and easy to fabricate, its mechanical performance may not be sufficient in many situations. Directly connecting an endplate to the tube requires specialist bolting technologies such as flowdrill or blind bolts. One relatively new connection type [1–4], which combines good mechanical performance, easy accessibility and moderate fabrication cost, is the reverse channel connection, as illustrated in Fig. 1. This type of connection consists of an end plate (welded to the beam n
Corresponding author. Tel.: þ 420 224 354 762; fax: þ 420 233 337 466. E-mail address: [email protected] (T. Jána).
http://dx.doi.org/10.1016/j.firesaf.2015.08.002 0379-7112/& 2015 Elsevier Ltd. All rights reserved.
and then bolted to the web of a channel) and a channel whose legs are welded to the column tube. The depth of the channel can be varied to accommodate any conventional flexible, flush or extended endplate connection. When a structure is exposed to a fire attack, connections are key components of the structure, particularly concerning the survivability (robustness) of the structure, as demonstrated by the World Trade Center collapse [5] and the structural fire tests at Cardington [6], Ostrava [7] and Mokrsko [8]. However, contemporary European standards EN 1993-1-2 [9] and EN 1993-1-8 [10] provide very limited information on quantifying the connection thermal and mechanical behavior at elevated temperatures during a fire. In other standards and guidelines around the world, there is no or very simple approach [11,12]: when a structure is fire protected, a thickness of the fire protection applied to connections should be equal or greater. During a fire, the behavior of the structure changes as a result of material property degradations and changes of the internal forces due to restrained thermal deformations and large structural deformations [13]. During the heating phase of a fire, exposed beams extend. If this extension is constrained by the adjacent structure, a relatively large compressive axial force develops in the
56
T. Jána et al. / Fire Safety Journal 78 (2015) 55–70
Fig. 1. Schematic of a reverse channel connection of a beam to a concrete-filled tubular column.
beam. Restraint to the thermal bowing of the beam also generates an additional hogging bending moment in the connection. At high temperatures, when vertical structural deformations are very big, axial restraint to the beam results in tension forces in the beam developing catenary action [14]. Connection components may fracture during the catenary action stage. Tension forces may also be generated in the beam and in the connection during the cooling stage. A limited number of fire tests using the reverse channel connection [1] have demonstrated its excellent performance compared to other types of connection to tubular columns. In order to develop methods to quantify the connection behavior, temperatures in different connection components should be obtained. Against this background, the European RFCS COMPFIRE project was implemented, see [15]. An important work package of the project was to determine connection temperatures. This was achieved by fire tests, detailed numerical simulations to broaden the scope of the fire tests and the development of analytical equations to calculate different component temperatures that may be incorporated into design methods such as EN 1993-1-2 [9]. This paper presents detailed results of the fire tests and a limited number of numerical simulations to determine the thermal boundary conditions to different components of the reverse channel connection. The results of this paper are used to develop a design calculation method.
2. Fire tests The fire tests included real fire scenarios in a full-scale experimental structure conducted by the Czech Technical University in Prague and a standard furnace fire test on a smaller composite steel and concrete frame structure at the University of Manchester, United Kingdom. 2.1. Furnace fire test at the University of Manchester A small frame was constructed with several connection arrangements. This frame was then placed inside a fire testing furnace with the directional natural gas burners, measuring about 2 m (depth) by 3.5 m (width) and 3.5 m (height). Fig. 2(a) shows the planar arrangement of the test frame inside the furnace and thermocouple locations. Fig. 2(b) shows a sketch of the test frame with connection reference notations. The ISO 834 standard fire temperature-time curve was followed in the fire test.
Fig. 2. University of Manchester furnace fire test: (a) furnace ground plan; (b) axonometric view of a test specimen with connection notations.
All steel beams were UB 305x165 40 and the steel tubular section filled with C25/30 normal weight concrete was either CHS 244.5 8 or SHS 250 8. A 100 mm thick C25/30 normal weight concrete slab was placed on top of the steel frame. the steelwork was not fire protected and no additional gravity loading was applied to the frame. table 1 gives an overview of the geometry of the monitored connections on the experimental structure. 140 thermocouples were used to record the temperatures in the reverse channel connection components, the concrete slabs and the beams at mid-span. Each thermocouple was pushed in the drilled hole with the depth that was equal to the half of the plate thickness. In order to achieve deeper hole in plates with the thickness less than 10 mm, the hole was drilled through the short weld made on the surface of the plate. The temperature in the furnace was measured by six standard plate thermometers, see Fig. 3, and the average temperature of these six plate thermometers was used as the reference temperature to follow the standard temperature–time curve. The heating phase took two hours and data recording continued for one more hour to obtain temperature data during cooling. During the cooling phase, electric fans were utilized to faster cool the specimen. The average temperature drop in the furnace was about 10 °C/min. This corresponds to the slope of the parametric temperature curve in the cooling phase for fire compartments of common administrative and residential buildings (the shape parameter Γ ¼2.5).
T. Jána et al. / Fire Safety Journal 78 (2015) 55–70
57
Table 1 Configuration of connections in the Manchester furnace fire test; dimensions in [mm]. Designation of connection
Beam cross-section
Column cross-section
End plate (depth/width/thickness)
Reverse channel
Bolts
A1 A2 B2-lower B2-upper B3-lower B3-upper C1 C3
UB 305x165 40
SHS 250 8 SHS 250 8 CHS 244,5 8 CHS 244,5 8 SHS 250 8 SHS 250 8 CHS 244,5 8 CHS 244,5 8
333, 4/200/20
UKPFC 230x90 32
6 x M20
Fig. 3. Measured gas temperatures at different locations in the furnace.
Fig. 4. Positions of thermocouples in the C1 reverse channel connection. Fig. 5. Temperature distributions in the reverse channel connection: (a) comparison between different components and (b) comparison of bolt temperatures.
Fig. 4 shows a typical reverse channel connection and thermocouple locations. The thermocouples were placed on the beam web (TC37 to TC39) at a distance of 200 mm from the endplate, in the center of the shaft of the upper, middle and lower bolts (TC31 to TC33), on the endplate (TC34 to TC36), on the reverse channel (TC40 to TC42) and the steel tube of the concrete-filled tubular column (TC43). Fig. 5(a) presents a typical example of temperature distributions in the reverse channel connection components. Higher temperatures were logically measured in the beam web, at a location far from the connection. However, the different connection components (reverse channel, end plate, bolts) experienced similar temperatures. The tube temperature was lower due to the heat sink effect of the infill concrete. Fig. 5(b) shows the temperatures measured in three different bolts (i.e. top, middle and bottom). They are also very close to each other. Similar temperature developments were recorded in the other connections (central, corner, edge). Fig. 6 shows a comparison of beam web temperatures measured on different locations of the structure. Differences in the
Fig. 6. Measured beam web temperatures at different locations of the structure.
58
T. Jána et al. / Fire Safety Journal 78 (2015) 55–70
temperatures of the structure are mainly caused by the different gas temperatures across the furnace. 2.2. Fire tests in a full-scale structure at the Czech Technical University in Prague An experimental two-storey composite steel and concrete structure with dimensions of 10.4 13.4 m2 and a height of 9 m, representing part of an administrative building, see Fig. 7, was constructed for the COMPFIRE project. Fig. 7(a) gives the serial sizes of the beams and columns. The tubular columns were un-
protected but filled with concrete. Fig. 8 shows the construction of the upper and lower composite floors. Table 2 shows the main dimensions of the reverse channel connections. The load bearing structure was partially fire protected. The PROMASPRAY F250 spray, a mixture of mineral fibers and cement binders with low density, was used as fire protection. On the second floor, the connections of central beams with the span of 9 m were covered with 20 mm of fire protection in order to obtain temperature differences in the unprotected and fire protected connections of the unprotected structure. In addition, cross bracings and secondary columns were fire protected because of
Fig. 7. Structure of the fire test building at CTU: (a) ground plan; (b) vertical section; (c) picture of the building.
Fig. 8. Construction of composite floors: (a) upper slab; (b) lower slab.
T. Jána et al. / Fire Safety Journal 78 (2015) 55–70
59
Table 2 Configurations of connections of the building fire test at CTU; dimensions in [mm]. Designation of connection First fire test (second floor) A2-B2 to A2 A2-B2 to B2 B2-C2 to B2 B2-C2 to C2 Second fire test (first floor) A2-B2 to A2 A2-B2 to B2 B2-C2 to B2 B2-C2 to C2
Beam cross-section
Column cross-section
End plate
Reverse channel
Bolts
Designed thickness of fire protection
IPE 270
TR 245/8
165/160/8
sheet 165/200/8 UPE 160 UPE 120 sheet 135/160/8
4x M16
20 mm
4x M12
165/160/8
sheet 165/200/8
4x M16
200/180/8
sheet 200/220/8 UPE 180
4x M20
60 mm
IPE 220
IPE 270
135/120/8
TR 245/8
IPE 330
stability of the bearing structure and cladding. On the first floor, 60 mm sprayed fire protection was again applied on the chosen connections, secondary columns and cross bracing. Due to the expected achievement of the membrane action of the composite floor during the second fire test, all edge beams and corner columns were also fire protected. The reverse channel connections were protected by a length of 250 mm measured from the edge of the column in all cases. The composite steel and concrete columns were unprotected. The measurements involved 97 thin thermocouples of type K to measure the gas temperatures (TG, 20 pcs., diameter 3 mm), temperatures at the mid-span of the beams (TB, 8 pcs., diameter 2 mm), at various locations in the reverse channel connections and the fin plate connections (TC, 50 pcs., diameter 2 mm), the composite slabs (TS, 7 pcs., diameter 2 mm) and the columns (TSG, 12 pcs., diameter 2 mm) for each fire test. 7 standardized plate thermometers were also used to obtain the adiabatic surface temperatures. Fig. 9 shows the thermocouple locations. Two fire tests were carried out, one on 6th September 2011 and one on 15th September 2011. Timber cribs were used as the fire source for both fire tests. For the first fire test on the second floor, fire load was created by 8 3 piles of wooden cribs of 50 50x1000 mm3 as shown in Fig. 10. All spruce cribs were dried to a moisture content of 12%. Each pile consisted of 6 layers with 7 cribs, a total of 42 cribs. The total volume of the timber cribs was 2.52 m3, to give a fire load density (per floor area) of 9.9 kg/m2 or 173.5 MJ/m2. A linear ignition source placed on the left-hand side was made of a thin-walled channel filled with mineral wool and wetted by paraffin. The ventilation of the fire compartments was
Fig. 9. Position of thermocouples for both building tests (first fire tests/second fire test).
enabled by window openings with dimensions of 5x2 m2 In the case of the second fire test on the first floor, 28 wooden piles were distributed on the ground area of the fire compartment in accordance with Fig. 11(a). Each pile consisted of 11 layers with 10 cribs. The total volume of wood was 7.65 m3, giving a fire load per floor area of 30.1 kg/m2 or 525 MJ/m2. All piles were ignited at the same time by the aid of the thin-walled channel filled with mineral wool and paraffin, see Fig. 11(b). Fig. 12(a) and (b) shows the recorded gas temperatures at different locations inside the fire compartment, 300 mm under the ceiling slab, during the first fire test. A maximum gas temperature of 979 °C was recorded at 26th min of the fire test. Because there are only small differences between the temperature measured by the thermocouple of type K and calculated adiabatic surface temperature [16] based on the measurement of the plate thermometers, see Fig. 12(c), the temperatures from thermocouples can be used to calculate heat transfer to the structure. Comparison of the gas temperature results with prediction models is in [17]. The thermal response of the structure depended on the spatial development of the fire, see Fig. 13. The development of the gas temperature during the second fire test corresponded to the fire scenarios with flashovers in the fire compartment of a typical administrative building. As illustrated in Fig. 14(a) and (b), in the initial stage of the fire, higher gas temperatures were observed in the front of the compartment (temperature differences of up to 200 °C at 30th min). In contrast, after the full development of the fire, the area with higher gas temperatures was shifted to the rear of the fire compartment (temperature differences of about 150 °C at 50th min). A maximum gas temperature of 1002 °C was reached in the rear of the fire compartment at 46th min of the fire test. During the cooling period, temperatures were distributed more uniformly. The temperatures from thermocouples can be used to calculate heat transfer to the structure because they are in good agreement with the adiabatic surface temperatures, see Fig. 14(c). As can be seen from the comparison of prediction models in Fig. 14(d), both the parametric fire curve [18] and the zone model [19] give the good results for the heating phase and maximum temperature, but the zone model underestimates the gas temperature during the cooling phase. The thermal response of the whole structure during the second fire tests is demonstrated in Fig. 15. 2.2.1. Temperatures of unprotected connections Two typical examples of temperature distributions in the unprotected reverse channel connection and connected members are presented for discussion. Fig. 16 shows the beam-to-central column connection B2-C2/B2 on the second floor. There was no fire protection to the structure. The reverse channel was made of UPE 120 (flange width 60 mm, flange thickness 8.0 mm, web width 120 mm, web thickness 5.0 mm), with a height of 135 mm. Both flanges of the channel section were welded to the column with a
60
T. Jána et al. / Fire Safety Journal 78 (2015) 55–70
Fig. 10. Provision of fire load: (a) distribution on the floor; (b) preparation of ignition.
Fig. 11. Position of timber piles on the 1st floor of the experimental building: (a) ground plan with the scheme of ignition; (b) ignition.
half V weld (throat thickness 8 mm). The end plate of the beam was made of 8 mm plate 135 mm depth x 120 mm width, welded to the web of the IPE 220 beam with a fillet weld (throat thickness 4 mm) on both sides of the web and connected to the reverse channel web by four M12 bolts. Fig. 16 also shows the locations of six thermocouples on the connection. Fig. 17 compares the measured temperatures in different parts of the connection, including the temperature of the bottom beam flange at mid-span. It is clear that the connection temperatures were significantly lower than the temperature of the beam flange. The column tube and weld of the reverse channel flange and the column tube had lower temperatures due to the heat sink effect of the infill concrete. However, the other connection component temperatures (lower bolt, reverse channel flange) were quite similar, with the maximum difference being less than 60 °C. Fig. 18 shows the beam-to-central column reverse channel connection A2-B2/B2 without fire protection on the first floor and the positions of seven thermocouples on this connection. The reverse channel was cut out of a square tube TR 200/200/8 (flange width 86 mm, web width 200 mm, flange and web thickness 8.0 mm), with a height of 165 mm. Both flanges of the channel section were welded to the column with a half V weld (throat thickness 8 mm). The end plate was made of 8 mm plate with a depth of 165 mm and a width of 160 mm, welded to the web of the beam IPE 270 with a fillet weld (throat thickness 4 mm) and
connected to the reverse channel by four M16 bolts. Fig. 19(a) compares the temperatures measured at different connection components with the beam lower flange temperature. The temperature distributions were very similar to those shown in Fig. 17. The connection component (reverse channel flange and bolt) temperatures were lower than the temperature of the beam lower flange, but they were close to each other. Fig. 19 (b) compares the temperatures in different bolts and it shows similar temperatures. The conclusions of these two fire tests with regard to temperature distributions in different connection components are the same as those from the University of Manchester furnace fire test. 2.2.2. Temperatures of fire protected connections Temperatures were also measured on fire protected reverse channel connections. Two typical examples are presented here. Fig. 20 shows the fire protected beam-to-central column connection A2-B2/B2 on the second floor. The reverse channel was made of UPE 160 (flange width 70 mm, flange thickness 9.5 mm, web width 160 mm, web thickness 5.5 mm), with a height of 165 mm. Both flanges of the channel section were welded to the column with a half V weld (throat thickness 9.5 mm). The end plate was made of 8 mm plate, 165 160 mm2 in size, welded to the web of the beam IPE 270 with a fillet weld (throat thickness 4 mm) on both sides of the web and connected to the reverse channel by
T. Jána et al. / Fire Safety Journal 78 (2015) 55–70
61
Fig. 12. Gas temperatures at the first fire test: (a) measured temperatures at the front part of the compartment; (b) measured temperatures at the back part of the compartment; (c) comparison with the adiabatic surface temperature.
Fig. 13. Measured bottom flange temperatures at mid-span of the beams from the first fire test.
four M16 bolts. The locations of thermocouples are similar to both cases presented above, see Fig. 20(a). The connection was protected for a length of 250 mm measured from the edge of the column. The thickness of the fire protection was variable between 30 and 60 mm, as shown in Fig. 20(b). Fig. 21 compares the component temperatures (lower bolt, reverse channel flange) of the protected connection. Again, the connection component temperatures were very similar, with the maximum difference of less than 50 °C. The comparison between the results in Fig. 21 and those in Fig. 17 indicates that under a similar gas temperature in the fire compartment and a similar lower flange temperature of the connected (unprotected) beam, the temperatures in the protected connection components were by about 300 °C lower than the unprotected connection component temperatures. This suggests that the length of the fire protection used in the test was sufficient for significantly
reducing the heat conduction from the unprotected beam. Fig. 22 shows the fire protected connection B2-C2/B2 on the first floor in the second fire test. The reverse channel was cut out of a square tube TR 220/220/8 (flange width 98 mm, web width 200 mm, flange and web thickness 8.0 mm), 200 mm high. The end plate was 8 mm thick, 200 mm in depth and 180 mm in width, welded to the web of the IPE 330 beam with a fillet weld (throat thickness 4 mm) on both sides of the web and connected to the reverse channel by four M20 bolts. Fig. 22(a) shows the locations of five thermocouples. The thicknesses of the fire protection measured before the second fire test are shown in Fig. 22(b). The temperatures measured in the fire protected reverse channel connection B2-C2/B2 during the second fire test are shown in Fig. 23. From the comparison with the temperatures in the similar connection without fire protection from Fig. 19(a), it is clear that the applied fire protection with an average thickness of 60 mm was able to reduce the temperatures in the connection by 400 °C. This leads to a significant increase in the connection fire resistance, see [20]. Once again, the temperatures in the reverse channel, in the endplate and in the bolts, may be considered to be the same. The tube temperature was lower due to the heat sink effect of the filled concrete. These conclusions are consistent with those of [21], which are based on results of fire tests of the simple frame of one steel beam and two composite columns in the furnace. This means that the temperature distribution in the reverse channel connection is notably affected by neither the presence of the concrete slab above the connections, nor the connection of more beams to the column at the same area, nor the change in the fire scenario.
3. Numerical model The fire tests described in the previous section have been
62
T. Jána et al. / Fire Safety Journal 78 (2015) 55–70
Fig. 14. Gas temperatures at the second fire test: (a) measured temperatures at the front part of the compartment; (b) measured temperatures at the back part of the compartment; (c) comparison with the adiabatic surface temperature; (d) comparison with the predicted gas temperatures.
Fig. 15. Measured bottom flange temperatures at mid-span of the beams from the second fire test.
Fig. 17. Comparison of measured temperatures in different components of the unprotected reverse channel connection B2-C2/B2 with the beam bottom flange temperature at mid-span.
Fig. 16. Position of thermocouples in the unprotected reverse channel connection B2-C2/B2.
T. Jána et al. / Fire Safety Journal 78 (2015) 55–70
63
Fig. 18. Position of thermocouples in the unprotected reverse channel connection A2-B2/B2.
tube of the composite column outside the connection area for the Czech Technical University tests (hc ¼ 35 W/m2K, εm ¼0.3). In the furnace test, there was no measurement of the column temperature outside the connection area. The PROMASPRAY F250 fire protection was used in the fire tests performed by CTU. This fire protection consists of 80% mineral wool fibers and 20% Portland cement [24]. The density of the material was assumed to be ρp ¼264 kg/m3[25]. This value corresponds with the density of sprayed mineral fibers introduced in [26]. Based on [26], the specific heat of the fire protection material was assumed to be cp ¼1050 J/kg K. Because the influence of the thermal capacitance (density multiplied by specific heat) of the lightweight fire protection materials on the resultant steel temperature is very small [27], the density and the specific heat value can be taken as constant. On the contrary, it is more important to obtain accurate temperature dependent thermal conductivity values. The temperature dependent thermal conductivity of the fire protection based on the mineral wool is according to [27], using the following equation:
⎛ T ⎞3 ⎟ λmineral _ wool = 0.03 + 0.2438 ⎜ ⎝ 1000 ⎠
(1)
where T is in Kelvin. Eq. (1) was modified so that it fits with the declared thermal conductivity of 0.043 W/mK at 24 °C by the manufacturer’s datasheet [25] for PROMASPRAY F250. The final thermal conductivity used in the numerical simulations was
⎛ T ⎞3 ⎟ λ p, T = 0.037 + 0.2438 ⎜ ⎝ 1000 ⎠
Fig. 19. Measured temperatures in the unprotected reverse channel connection A2B2/B2: (a) comparison between different connection components; (b) comparison of bolt temperatures.
modeled using the specialist fire engineering software SAFIR [22]. Level of spatial discretization used for the models of the connections is shown in Figs. 31, 34 and 37. The purpose of this simulation was to check the sensitivity of the simulation results to different assumptions of thermal boundary conditions so as to identify a suitable thermal boundary condition for calculating connection temperatures. 3.1. Material thermal properties The measured density of concrete was ρc ¼2250 kg/m3. The other thermal properties (thermal conductivity, density and specific heat) of steel and concrete were according to Eurocodes EN 1993-1-2 [9] and EN 1994-1-2 [23]. To demonstrate the applicability of these standard values, Fig. 24 shows the comparison between the measured and simulated temperatures of the steel
(2)
Fig. 25 shows the graphical representation of the dependence of thermal conductivity on temperature in degrees Celsius. The comparison of the measured and simulated beam temperatures in Fig. 26 demonstrates the applicability of the above values. 3.2. Thermal boundary condition In numerical heat transfer simulations of the University of Manchester test, the gas temperature was assumed equal to the measured temperature from the closest TG thermocouple in the furnace. Because the gas temperatures were not measured directly at the connections during the Czech Technical University tests, the average of the measured temperatures from the thermocouples around to the monitored connection was assumed as the gas temperature in the numerical model. The unexposed surface was in contact with air of a constant temperature of 20 °C. Because the convective heat transfer conditions were not measured, the design values of the convective heat transfer coefficient hc ¼25 W/m2K for surfaces subjected to the standard fire condition and hc ¼35 W/m2K for the natural fire condition according to EN 1991-1-2 [18] were used. However, recent study [28]
64
T. Jána et al. / Fire Safety Journal 78 (2015) 55–70
Fig. 20. Fire protected reverse channel connection A2-B2/B2: (a) position of thermocouples; (b) thicknesses of fire protection.
Fig. 21. Comparison of measured temperatures in different components of the fire protected reverse channel connection A2-B2/B2.
Fig. 23. Comparison of measured temperatures in different components of the fire protected reverse channel connection B2-C2/B2 with the beam bottom flange temperature in mid-span.
Fig. 22. Fire protected reverse channel connection B2-C2/B2: (a) position of thermocouples; (b) thicknesses of fire protection.
T. Jána et al. / Fire Safety Journal 78 (2015) 55–70
65
Fig. 24. Comparison of measured and simulated temperatures of the steel tube of the composite column.
Fig. 25. Thermal conductivity of PROMASPRAY F250 fire protection material in relation to temperature.
Fig. 27. Influence of different steel surface emissivity values on the beam bottom flange temperature: (a) University of Manchester test; (b) CTU test.
of the lower flange at mid span of the beam indicate, that the real surface emissivity of the steel is lower in these cases, see Fig. 27 (a) for the University of Manchester test and Fig. 27(b) for the CTU test. Base on the equation of the Lumped Capacitance Method for unprotected steel members in [9], the inverse calculation of the surface emissivity can be expressed as
εm =
Fig. 26. Comparison of the measured and simulated temperatures of the web of the fire protected beam C1-C2 at mid-span (the second building fire test).
suggests that it is appropriate to consider lower values of the convective heat transfer coefficient for the connection area due to a slower fire gas movement. For the surfaces in contact with the ambient temperature, the convective heat transfer coefficient was hc ¼ 4 W/m2K. According to EN 1993-1-2 [9], the design value of the surface emissivity of carbon steel is εm ¼0.7 and according to EN 1992-1-2 [29], the surface emissivity of concrete is εm ¼0.7. However, the comparisons between the calculated and measured temperatures
ca ⋅ ρa ∂θa ⋅ Am / V ∂t
− hc ⋅( θg − θa )
4 4⎤ ⎡ Φ⋅εf ⋅σ⋅⎣⎢ ( θg + 273) − ( θa + 273) ⎥⎦
(3)
where ca and ρa are the specific heat capacity and the density of steel, Am/V is the section factor of the lower beam flange, ∂θa is the increment of the steel temperature in time ∂t , hc is the convective heat transfer coefficient (hc ¼35 W/m2K for a natural fire and hc ¼25 W/m2K for a standard fire), θg is the measured gas temperature, θa is the measured temperature of the lower beam flange, Φ is the configuration factor (taken as 1.0 because the steel surface is in direct contact with the fire), εf is the fire gas emissivity for the whole furnace or compartment (taken as 1.0), s is the Stephan-Boltzmann constant ( ¼5.67 10 8 W/(m2K4)). Using the inverse calculation for the several beams in each fire test, the surface emissivity of the steel sections was specified as ε m ¼ 0.5 in the case of the University of Manchester furnace test and εm ¼ 0.3 in the case of the CTU tests. The surface emissivity is considered as a constant value, but it seems to be sufficient to predict the temperature of the structure. The calculated lower flange temperatures based on the obtained values of the surface emissivity are in good agreement with the measured temperatures, see Fig. 27. On the contrary, the change of the convective
66
T. Jána et al. / Fire Safety Journal 78 (2015) 55–70
according to EN 1991-1-2 [18], is εf ¼ 1.0. This is based on the assumption that the fire depth is very large. A lower flame emissivity should be used for the inner surface of the reverse channel connection, because the depth of the fire gas is small, see Fig. 28. The total emissivity of hot gas εf depends on the coefficient of radiation K and the thickness of gas, as follows [30]:
εf = 1 − e−KL
(4)
where L is the mean beam length of the gas. For gas – smoke from wood-based materials, the coefficient of radiation is K ¼0.8 [31]. For any general geometry of the gas volume, the mean beam length can be approximately expressed as
L=C
4V A
(5)
where V is the total volume of the gas, see Fig. 28(a), A is the area on which heat is radiated, see Fig. 28(b) and C is a correction factor (taken as 0.9) which shall be applied for the optically thick volume of the gas [32]. On the basis of the above equations, the fire gas emissivity inside the reverse channel for each reverse channel connection surface used in the simulations was calculated and the results are given in Tables 3 and 4.
4. Calibration of the numerical model 4.1. Unprotected connections To demonstrate the accuracy of the assumptions mentioned in Section 3 for thermal boundary conditions and fire protection thermal properties, the numerical simulation results are compared to the measured temperatures for three examples. One for the unprotected connection from the University of Manchester furnace test and two for the connections (unprotected and fire protected) from the CTU natural fire tests. More comparisons of numerical and test results can be found in [33].
Fig. 28. Volume (a) and area (b) for the calculation of the mean beam length of the gas layer between the reverse channel and the column tube.
heat transfer coefficient has small effect on the resulting temperatures of the beam flange (particularly at the high temperatures). The surface emissivity of the lower side of the composite slab, which was formed by zinc-coated trapezoidal sheets, was ε m ¼0.23, according to [27]. The recommended design value for the emissivity of the fire,
4.1.1. Comparison against the University of Manchester furnace fire test Fig. 29 shows the effects of using different convective heat transfer coefficients and emissivities of fire on component temperatures of the connection B2-lower. For the simulation, the measured temperature of thermocouple TG5 was used as the gas temperature outside as well as inside the reverse channel connection. The simulated temperature-time developments of the end plate and the reverse channel flange are compared with the temperatures measured during the University of Manchester furnace fire test. It is clear that the numerical results agree with the
Table 3 Fire gas emissivity inside the reverse channel connections from the University of Manchester test. Connection ID
V [m3]
A [m2]
C [dimensionless]
L [m]
K [dimensionless]
εf [dimensionless]
B2-lower, B2-upper, C1, C3 A1, A3, B3-lower, B3-upper
3.379 10 3 5.534 10 3
19.82 10 2 18.97 10 2
0.9 0.9
6.14 10 2 10.5 10 2
0.8 0.8
0.0479 0.0806
Table 4 Fire gas emissivity inside the reverse channel connections from the CTU tests. Connection ID
V [m3]
A [m2]
C [dimensionless]
L [m]
K [dimensionless]
εf [dimensionless]
B2-C2/B2 A2-B2/B2, A2-B2/A2
0.651 10 3 1.44 10 3
4.06 10 2 8.50 10 2
0.9 0.9
5.77 10 2 6.08 10 2
0.8 0.8
0.0451 0.0475
T. Jána et al. / Fire Safety Journal 78 (2015) 55–70
67
Fig. 30. Comparison between measured and simulated temperatures of the beam web and the column tube for the B2-lower connection in the University of Manchester furnace fire test.
Fig. 29. Influence of different emissivities of fire and convective heat transfer coefficients on the temperatures of the B2-lower connection in the University of Manchester furnace fire test: (a) end plate; (b) reverse channel flange.
test results best when the convective heat transfer coefficient is hc ¼10 W/m2K and the flame emissivity inside the reverse channel is calculated according to the approach described above (because of the same steel surface as on the beam, the surface emissivity ε m ¼0.5 is unchanged). Change in the convective heat transfer mainly affects low component temperatures and there are biggest differences in this range when using hc ¼ 25 W/m2K. Different geometry of the structure and different surrounding flow conditions in the connection area can be the reasons for using the lower value of the convective heat transfer coefficient [30]. Same boundary conditions (with hc ¼10 W/m2K) are also used for the simulation of component temperatures in Fig. 30. The agreement between the simulation and measured temperatures is excellent for all four connection components, including both the heating and cooling phases. During the heating phase, the maximum temperature difference between the measured and calculated values is not more than 30 °C. During the cooling phase, the maximum temperature difference is about 50 °C. Fig. 31 presents the temperature profile at 60th min. The uniformity of the temperature distribution in the reverse channel and the endplate is clearly shown. 4.1.2. Comparison against CTU building fire tests Figs. 32 and 33 show the comparison of the simulated and measured temperatures of the components of the connection A2B2/B2. The gas temperature in the numerical model was equal to the average temperature of thermocouples TG23, TG24, TG29 and TG30. As shown in Fig. 32, the simulation results agree with the
test results best when the convective heat transfer coefficient is hc ¼5 W/m2K and the flame emissivity inside the reverse channel is used according to Table 4 (steel surface emissivity is εm ¼0.3). Same boundary conditions are used for the simulation in Fig. 33. The predicted temperatures are in good agreement with the measured temperatures, especially during the heating phase. Larger temperature differences, of up to 100 °C, can be observed in the cooling phase. In general, the temperatures from the numerical model were higher than the measured temperatures. This difference was caused by a change in the gas flow close to the connection after the loss of integrity of the composite slab above this connection at 54th min of the second compartment fire test. The visualization of the temperature distribution on the three dimensional model of the reverse channel connection at 57th min is shown in Fig. 34. 4.2. Fire protected connections The calibration of the numerical model of a fire protected reverse channel connection is performed for connection A2-B2/B2 to the central column in the first compartment fire test. The gas temperature in the numerical model was equal to the average temperature of thermocouples TG3, TG4, TG9 and TG10. In the case of the fire protected connection, the effect of using different convective heat transfer coefficients on component temperatures is negligible, see Fig. 35. The results in Figs. 35 and 36 show that the temperatures obtained from the numerical model are in good
Fig. 31. Simulation results of the temperature field in the B2-lower connection at 60th min of the furnace fire test.
68
T. Jána et al. / Fire Safety Journal 78 (2015) 55–70
agreement with the measured temperatures. The simulated temperatures of the beam web in the connection area, the end plate, the lower bolt and the reverse channel flange differ from the measured temperatures by a maximum of 20 °C. Fig. 37 shows the visualization of the temperature field of the same connection at 29th min of the fire test. A uniform temperature distribution may be assumed for the connection design.
5. Summary and conclusions This paper has presented temperature distributions measured in reverse channel connections to concrete filled hollow tubes in
Fig. 34. Simulation results of the temperature field in the connection A2-B2/B2 at 57th min of the second building fire test.
Fig. 32. Influence of different emissivities of fire and convective heat transfer coefficients on the temperatures of the unprotected connection A2-B2/B2 in the second building fire test: (a) lower bolt; (b) reverse channel flange.
Fig. 35. Influence of different convective heat transfer coefficients on the temperature of the end plate and the beam web close to the end plate of the fire protected connection A2-B2/B2 in the first building fire test.
Fig. 33. Comparison between measured and simulated temperatures for the unprotected reverse channel connection A2-B2/B2 in the second building fire test.
Fig. 36. Comparison between measured and simulated temperatures of the lower bolt and the reverse channel flange for the fire protected connection A2-B2/B2 in the first building fire test.
T. Jána et al. / Fire Safety Journal 78 (2015) 55–70
69
for a real compartment fire exposure respectively were shown to give more accurate temperature calculation results. For simplicity, a value of hc ¼10 W/m2K may be used.
Acknowledgments The research was supported by funding from the European Community’s Research Fund for Coal and Steel (RFCS) under the Grant agreement no. RFSR-CT-2009-00021 (COMPFIRE) and the European Social Fund within the framework of realizing the project “Support of Inter-sectoral Mobility and Quality Enhancement of Research Teams at the Czech Technical University in Prague”, CZ.1.07/2.3.00/30.0034, period of the project's realization 1/12/ 2012–30/6/2015.
Fig. 37. Simulation results of the temperature field in the fire protected reverse channel connection A2-B2/B2 at 29th min of the first building fire test.
one standard furnace test and two realistic building fire tests. The fire tests were simulated using the SAFIR fire engineering software. The main objective of the numerical simulations was to determine the relevant thermal boundary conditions for the connections. The following conclusions can be drawn: (1) In the unprotected connections, temperature of the bolts, end plate, reverse channel and welds of the end plate and the beam web may be considered as uniform and lower than the beam temperature. Temperature of welds of the reverse channel flange and column tube may be considered lower than the reverse channel flange temperature. The temperature of the column tube and the adjacent welds is largely influenced by the concrete core in the composite column, which takes heat from the connection during the heating phase of fire. (2) The temperature distribution in the reverse channel connection is notably affected by neither the presence of the concrete slab above the connections, nor the connection of more beams to the column at the same area, nor the change in the fire scenario. (3) The temperature distribution in the all components of the protected connections may be considered uniform in the connection design. (4) The connection location within the fire compartment has no influence on the connection temperatures, except for local fire gas temperatures. (5) If the connection is fire protected but the connected beam is not, the heat conduction from the unprotected beam to the connected connection is negligible if there is a small length of fire protection to the steel beam adjacent to the connection. (6) For predicting connection component temperatures, the fire gas volume enclosed by the reverse channel and the column surface should be used to calculate the gas emissivity. This value (typically less than 0.1) is much lower than the recommended fire gas emissivity in EN 1991-1-2. For simplicity, a value of εf ¼0.1 may be used. (7) Convective and radiative heat transfer conditions were not measured at the fire tests, therefore the steel surface emissivity was obtained by the inverse calculation from the measured temperatures and the convective heat transfer coefficient was assumed as the standard value. But the calibration of the numerical model clearly implies that the convective heat transfer coefficient in the connection region should be lower than those recommended in EN 1991-1-2 (25 and 35 W/m2K respectively for a standard furnace fire and a real fire exposure). The values of hc ¼ 10 W/m2K for a furnace fire exposure and hc ¼5 W/m2K
References [1] J. Ding, Y.C. Wang, Experimental study of structural fire behaviour of steel beam to concrete filled tubular column assemblies with different types of joints, Eng. Struct. 29 (12) (2007) 3485–3502. [2] S. Elsawaf, Y.C. Wang, P. Mandal, Numerical modelling of restrained structural subassemblies of steel beam and CFT columns connected using reverse channels in fire, Eng. Struct. 31 (2011) 1217–1231. [3] S. Elsawaf, Y.C. Wang, Methods of improving the survival temperature in fire of steel beam connected to CFT column using reverse channel connection, Eng. Struct. 32 (2012) 132–146. [4] S. Elsawaf, Y.C. Wang, Behaviour of restrained structural subassemblies of steel beam to CFT column in fire during cooling stage, Eng. Struct. 46 (2013) 471–492. [5] FEMA (Federal Emergency Management Agency), World Trade Center Building Performance Study: Data Collection, Preliminary, Observations, and Recommendations, Washington, DC, 2002. [6] F. Wald, L. Simões da Silva, D.B. Moore, T. Lennon, M. Chladná, A. Santiago, M. Beneš, L. Borges, Experimental behaviour of a steel structure under natural fire, Fire Saf. J. 41 (7) (2006) 509–522. [7] F. Wald, J. Chlouba, A. Uhlíř, P. Kallerová, M. Štujberová, Temperatures during fire tests on structure and its prediction according to Eurocodes, Fire Saf. J. 44 (1) (2009) 135–146. [8] F. Wald, P. Kallerová, J. Chlouba, Z. Sokol, M. Strejček, Fire test on an administrative building in Mokrsko: Summary of the Results from the Fire Experiment, Printing house Česká technika, CTU in Prague, 2010. [9] EN 1993-1-2, Eurocode 3: Design of Steel Structures, Part 1.2: General Rules – Structural Fire Design, CEN, European Committee for Standardization, Brussels, 2005. [10] EN 1993-1-8: Eurocode 3:, Design of steel structures-Part 1–8: Design of joints, CEN, European Committee for Standardization, Brussels, 2005. [11] International Building Code, Falls Church, International Code Council (ICC), 2006. [12] J.P. England, S.A. Young, M.C. Hui, N. Kurban, Guide for the Design of Fire Resistant Barriers and Structures, Building Control Commission, Melbourne, 2000. [13] Y.C. Wang, The importance of considering whole structural behaviour in fire, in: Proceedings of the 3rd Structures in Fire Seminar, Singapore and Hong Kong, international conference, invited presentation, 2002. [14] Y. Yin, Y.C. Wang, Analysis of catenary action in steel beams under fire conditions, Prog. Steel Build. Struct. 6 (3) (2004) 1–8. [15] L. Simões da Silva, A. Santiago, F. Lopes, M. Veljkovic, T. Heistermann, N. Iqbal, F. Wald, T. Jána, B. Davison, I. Burgess, S.S. Huang, G. Dong, Y. Wang, P. Mandal, M. Jafarian, M. Skořepa, P. Velda, G. Koutlas, COMPFIRE-Design of composite joints for improved fire robustness: final report, Research Programme of the Research Fund for Coal and Steel, Brussels (2013). [16] U. Wickström, R. Jansson, H. Tuovinen, Validation fire tests on using the adiabatic surface temperature for predicting heat transfer, SP Technical Research Institute of Sweden (2009). [17] K. Horová, T. Jána, F. Wald, Temperature heterogeneity during travelling fire on experimental building, Adv. Eng. Softw. 62–63 (2013) 119–130. [18] EN 1991-1-2, Eurocode 1, Actions on structures-Part 1-2: General actions – Actions on structures exposed to fire, CEN, European Committee for Standardization, Brussels, 2002. [19] J.F. Cadorin, J.M. Franssen, D.I. Pintea, L.G. Cajot, M. Haller, J.B. Schleich, OZone V2.2.5 University of Liege and Arcelor Profil Luxembourg Research Centre, 2006. [20] T. Jána, F. Wald, Reduction of connection resistance during Veselí fire tests, J. Struct. Fire Eng. 6 (1) (2015) 21–27. [21] J. Ding, Y.C. Wang, Temperatures in unprotected joints between steel beams and concrete filled tubular columns in fire, Fire Saf. J. 44 (1) (2009) 16–32. [22] SAFIR (software), version 2011.b.3, ArGEnCo, University of Liege, 2011. [23] EN 1994-1-2, Eurocode 4, Design of Composite Steel and Concrete Structures,
70
[24] [25] [26]
[27] [28]
T. Jána et al. / Fire Safety Journal 78 (2015) 55–70 Part 1.2: General Rules – Structural Fire Design, CEN, European Committee for Standardization, Brussels, 2005. s Promat, PROMASPRAY F250: Fire protective spray on structures, Safety data sheet, revision date: 01/06/2015. s Promat, PROMASPRAY F250: Fire protective spray on structures, Technical data sheet. R.M. Lawson, G.M. Newman, Structural fire design to EC3 & EC4, and comparison with BS 5950, Technical Report, SCI Publication 159, The Steel Construction Institute (1996). Y.C. Wang, I. Burgess, F. Wald, M. Gillie, Performance-Based Fire Engineering of Structure, CRC Press, Boca Raton, 2012. X.H. Dai, Y.C. Wang, C.G. Bailey, Temperature distributions in unprotected steel connections in fire, in: Y.C. Wang, C.K. Choi (Eds.), Proceedings of 3rd
[29]
[30] [31] [32] [33]
International Conference on Steel and Composite Structures, Taylor & Francis, 2007, pp. 535–540. EN 1992-1-2: Eurocode 2, Design of concrete structures-Part 1-2: General rules – Structural fire design, CEN, European Committee for Standardization, Brussels, 2004. U. Wickström, Heat Transfer in Fire Technology, Luleå Tekniska Universitet (2012). D. Drysdale, An Introduction to Fire Dynamics, 3rd Edition, John Wiley & Sons, Chichester, 2011. M.Q. Brewster, Thermal Radiative Transfer and Properties, John Wiley & Sons, Inc., 1992. T. Jána, Teplota přípoje U profilem a čelní deskou při požáru, PhD. Thesis, Czech Technical University in Prague (2014).