- Email: [email protected]
Auto-extinction of engineered timber: Application to compartment fires with exposed timber surfaces
Fire Safety Journal xxx (xxxx) xxx–xxx
Contents lists available at ScienceDirect
Fire Safety Journal journal homepage: www.elsevier.com/locate/firesaf
Auto-extinction of engineered timber: Application to compartment fires with exposed timber surfaces ⁎
Alastair I. Bartletta, Rory M. Haddena, , Juan P. Hidalgoa, Simón Santamariaa, Felix Wiesnera, Luke A. Bisbya, Susan Deenyb, Barbara Laneb a b
School of Engineering, The University of Edinburgh, The King's Buildings, Mayfield Road, Edinburgh EH9 3JL, United Kingdom Arup, Fitzroy Street, London W1T 4BQ, United Kingdom
A R T I C L E I N F O
A BS T RAC T
Keywords: Compartment fires Heat transfer Extinction Performance-based design Firepoint theory
A series of compartment fire tests with multiple exposed timber surfaces have been undertaken to explore the effect of exposed timber on the fire dynamics and the potential for auto-extinction. A test with exposed wall and ceiling achieved auto-extinction after approximately 21 min. Firepoint theory is applied using temperature data at the charline, is shown to predict a mass loss rate dropping below the critical value at 20–21 min, and thus is successful in predicting auto-extinction. Additional uncertainties caused by delamination are explored, and recommendations for the use of auto-extinction in design are given.
1. Introduction Engineered timber products are continuing to increase in popularity, with factors such as sustainability, speed of construction, and aesthetics becoming ever more important. With the development of products such as cross-laminated timber (CLT) acting as structural wall and floor slabs, buildings are being constructed with the load-bearing structure formed predominantly from structural, engineered timber. The architectural aspiration for such buildings is often to have some of this structural timber exposed. The effect of exposed combustible compartment linings challenges many assumptions used in the assessment of compartment fire behaviour. One method to support robust design is to ensure that the timber linings auto extinguish after the compartment fuel load has been consumed. This requires an understanding of the energy balance at the char line, and an investigation of the critical factors at large scale. This presents an opportunity to explore the effects of multiple exposed timber surfaces within a compartment on the compartment fire dynamics. This paper focusses on the contribution of exposed timber surfaces to the fuel load during the cooling phase of a compartment fire, and interrogates the condition(s) under which the burning timber is expected to auto-extinguish. 2. Combustion of timber Timber begins to pyrolyse upon exposure to an external heat source, producing flammable and inert gases, tars, and a rigid,
⁎
carbonaceous char layer [1]. Flaming ignition of these flammable gases will occur when the generated mixture falls within the flammability limits, i.e. the air to fuel ratio falls within the correct range and the gases have sufficient energy. Smouldering ignition is also possible, but does not usually occur simultaneously with significant flaming combustion [1,2]. As a result, under flaming conditions, the char layer will continue to increase in thickness [1], reducing the rate of heat transfer to the virgin timber and resulting in a subsequent gradual decline in pyrolysis rate and hence mass flux of pyrolyzate [3]. 2.1. Extinction The “opposite” of flaming ignition is flaming extinction, which occurs when volatiles cease being produced in sufficient quantities to form a stable flame. This can be expressed in terms of the Damköhler number (the ratio of diffusion phenomena to kinetic timescales), which increases as a function of flame temperature and residence time (the duration the pyrolyzate remains in the reaction zone). Extinction will occur if the Damköhler number drops below a critical value, which can be achieved by reducing either the flame temperature or the residence time. Extinction can therefore occur due to external suppression (most commonly due to the application of water to cool the reaction zone); burnout of the pyrolysing fuel; or a reduction in the net energy supplied to the remaining unburned fuel. This will in turn result in a reduction in the mass of volatiles produced, which, if sufficient, will result in extinction. Due to the aforementioned decline in pyrolysis rate due to
Corresponding author. E-mail address: [email protected] (A.I. Bartlett).
http://dx.doi.org/10.1016/j.firesaf.2017.03.050 Received 15 February 2017; Accepted 27 March 2017 0379-7112/ © 2017 Published by Elsevier Ltd.
Please cite this article as: Bartlett, A.I., Fire Safety Journal (2017), http://dx.doi.org/10.1016/j.firesaf.2017.03.050
Fire Safety Journal xxx (xxxx) xxx–xxx
A.I. Bartlett et al.
subscripts
Nomenclature A Cp E ΔHc k H L m⋅ ′ ′ M q⋅′ ′ O Q S T W x X
area (m2) specific heat capacity (kJ/kg K) energy consumed per unit mass (kJ/g) heat of combustion (kJ/g) thermal conductivity (W/mK) height (m) latent heat (kJ/g) mass flux (g/m2s) molar mass (kg/kmol) heat flux (kW/m2) opening factor (m−1/2) heat release rate (kW) extinction parameter (kW/m2) temperature (°C) molecular mass (g/mol) position (m) volumetric concentration
a c CLT cr e f f g H2O i in int l n O2 p t v
Greek
v w
ϕ
critical ratio
∞
the same critical heat flux should not necessarily be expected. As with the experiments in [4] the critical mass flux for extinction increased with a reduction in oxygen concentration, but airflow was found to have no noticeable effect. In similar tests on Australian softwoods, Emberley et al. [7] found a critical mass flux of 4.0 g/m2s.
the build-up of a char layer, a burning timber sample has the potential to auto-extinguish naturally. To understand this condition in more detail, firepoint theory can be used to analyse the energy balance between the char and the virgin timber. 2.2. Firepoint theory
2.3. Compartment fire behaviour
Rasbash et al. [4] present firepoint theory as a means of determining ignition criteria for PMMA. They conducted a series of experiments on PMMA samples heated from above with a radiant panel to determine the effects of incident heat flux, air flow and oxygen concentration on the critical mass flux for ignition. In a range close to the critical heat flux for piloted ignition (12 kW/m2 to 19 kW/m2), the critical mass flux was found to increase as a function of heat flux, from about 3.8 g/m2s to 5.2 g/m2s; thereafter becoming independent of external heat flux. This initial variation was attributed to the variations in flame behaviour with lower heat fluxes. The effects of airflow around the sample were also investigated; an initial drop from around 5.3 g/m2s at 0lpm airflow to 3.2 g/m2s at 30 lpm, rising again to around 5.0 g/m2s at 60 lpm was observed. Reducing the oxygen concentration below ambient values resulted in a sharp increase in critical mass flux from around 3.3 g/m2s to 10.4 g/m2s at 19% O2. Rasbash et al. [4] concluded that firepoint theory may be used to determine if a material will continue to burn in the absence of a supporting heat flux:
S=(ϕ∆Hc − L v ) m⋅ cr" + q⋅e" − q⋅l"
air char CLT contribution critical external final flames gasification water vapour initial incoming air internal losses net oxygen pyrolysis total vaporisation ventilation wood ambient
Considerable research has been undertaken on compartment fires, the key aspects of which are summarised by Drysdale [8]. Wooden cribs are typically used as the fuel load, as they have been shown to be representative of typical room furnishings [9], and have good repeatability. Numerous correlations exist for the burning rate of cribs, and this is typically related to the opening factor of a compartment – calculated by:
O=
At Av Hv
(2)
where At is the total internal surface area excluding the floor and the opening, Av is the ventilation area, and Hv the ventilation height. Compartment fires can be said to follow three main stages: 1) the initial growth phase, in which burning is fuel-controlled; 2) the fullydeveloped, post-flashover phase, in which burning is ventilationcontrolled (Regime 1), and any excess fuel will be burned outside the compartment in an external plume; and 3) the decay phase, in which the burning will transition back to fuel-control (Regime 2) before extinction is achieved due to burnout of the fuel load.
(1)
where ϕ is the critical ratio of convective heat transfer to the heat of combustion of the volatiles, ΔHc is the heat of combustion of the solid, Lv is the heat of pyrolysis, and q⋅e′ ′ and q⋅l′ ′ are the external heat flux and heat losses respectively. If S > 0, the flame will be sustained, but if S < 0, extinction will occur. This equation has previously been applied to small-scale spruce/fir timber samples in a fire propagation apparatus (FPA) [5,6], where it was shown that timber samples will auto-extinguish when the mass flux of volatiles drops below 3.48 g/m2s (at ambient oxygen concentrations), and can be predicted using the firepoint equation. In the FPA setup, this value was reached at incident heat fluxes at or below 31 kW/ m2 – however it should be noted that in a compartment fire setup, heat losses, oxygen concentrations, and boundary conditions may differ, so
2.4. Compartment fires with combustible surfaces A compartment with exposed combustible surfaces such as timber presents an additional complexity due to the additional fuel load present. The existing empirical correlations were tested in situations where the fuel load was spread over the floor, and may not necessarily be valid when part of the fuel load is on the walls or ceiling. In a compartment with combustible surfaces, these will ignite, and the flames from these large exposed areas will produce additional heat which may increase the rate of burning of the compartment fuel load. The involvement of exposed combustible surfaces will also increase the total heat release rate, thus reducing the time to flashover. 2
Fire Safety Journal xxx (xxxx) xxx–xxx
A.I. Bartlett et al.
another was approximately 0.20 – the same as Crielaard [15]. All timber surfaces were constructed of CLT panels 100 mm thick with 5 uniform lamellae of thickness 20 mm. A single opening 0.8 m wide and 2 m tall was present in the front wall, however once the floor system was installed, the effective height was reduced to 1.84 m, corresponding to an opening factor of around 19 m−1/2. Four wooden cribs were placed in the compartments with a total mass of approximately 14 kg each. This corresponds to a fuel load of around 127 MJ/m2, which is relatively low compared to a typical office space with a mean design fuel load of 420 MJ/m2 [17], however since the decay phase was the area of interest, a short fully-developed compartment fire phase was desired.
Eventually, the fuel load in the compartment will burn out, and the combustible surfaces will be the only fuel remaining. At this stage, an energy balance can be performed, for example by using firepoint theory, to determine if the radiative exchange is sufficient for continued burning, until the walls or ceiling eventually burn through. The majority of CLT compartment fire tests to date have had either completely protected or completely unprotected timber surfaces [10– 13], and confirm expectations that exposed timber will contribute significantly to the heat release rate and accelerate flashover. Li et al. [14] conducted three tests with some timber surfaces exposed – one with one wall exposed, one with two opposite walls exposed (4.5 m×2.5 m), and one with two perpendicular walls exposed (4.5 m×2.5 m and 3.5 m×2.5 m), corresponding to view factors to the side walls of 0.18 and 0.19 with opposite and perpendicular walls exposed respectively and 0.15 to the back wall with perpendicular walls exposed. Both tests with two surfaces exposed experienced a secondary flashover after the onset of delamination, whereas the third autoextinguished. Crielaard et al. [15] tested scaled compartments with internal dimensions 0.5 m×0.5 m×0.5 m with 1, 2, or 3 exposed surfaces. The view factor from any one surface to another is 0.20, comparable to the side walls in the tests of Li et al. [14]. The test with one surface, and one of the two tests with two surfaces eventually reached auto-extinction. The other test with two surfaces and the test with three surfaces however, did not reach this state, with the two exposed surfaces reaching a secondary flashover as in [14] due to delamination. The three exposed surfaces continued burning and showed no sign of decreasing their heat release rate (HRR).
3.1. Test outcomes Of the two β tests carried out, β−2 failed to reach auto-extinction due to progressive delamination of CLT panels. Further details of the failure mechanisms are available in [18]. Test β−1 achieved autoextinction, and is therefore the focus of the current paper. 3.2. Instrumentation Inconel sheathed Type-K thermocouples were inserted into the exposed timber surfaces to record the internal temperature profiles and enable an analysis of the thermal gradients. 101 thermocouples were embedded in the back wall, and 101 in the ceiling, at 13 different locations in each surface. Thermocouples were inserted into pre-drilled holes at nominal depths of 5 mm, 10 mm, 15 mm, 20 mm, 25 mm, 30 mm, 35 mm, 40 mm, 50 mm, 60 mm, and 80 mm (high-density) or 10 mm, 20 mm 40 mm, 60 mm, and 80 mm (low-density) from the exposed surface (positioning shown in Fig. 2). Metallic discs of 10 mm diameter and coated to match the emissivity of charred timber were embedded into the timber surfaces at each group of thermocouples to allow measurement of surface temperature and an estimation of the incident heat flux. Additionally, six trees of 12 Type-K thermocouples were placed inside the compartment to provide gas-phase temperatures at various locations in the compartment. The tests were undertaken under a large extraction hood fitted with oxygen, carbon dioxide, and carbon monoxide analysers, enabling calorimetry to be performed and providing an estimate of the total heat release rate.
3. Experimental configuration A series of five full-scale CLT compartment fire tests was carried out at the BRE Burn Hall to explore the changes in fire dynamics and potential for auto-extinction due to exposed timber surfaces contributing to the fire. For the current paper, the two tests in which back wall and ceiling were exposed (configuration β) are explored (as shown in Fig. 1). All other surfaces were encapsulated with a system consisting of fire-rated gypsum plasterboard (rated Type F according to BS EN 520 [16]) and stone wool. The internal dimensions (excluding encapsulation and floor system) were 2.75 m long, 2.75 m wide, and 2.95 m high. With encapsulation and the floor system considered, the view factor from one surface to
Fig. 1. Experimental compartment.
3
Fire Safety Journal xxx (xxxx) xxx–xxx
2810 L L L 0 (a)
0
H
H
L
L
H
H
L
L
H
H
1375
Distance from front [mm]
Distance from floor [mm]
A.I. Bartlett et al.
2750
Distance from left wall [mm]
2750 L 1375 L L 0 (b)
0
H
H
L
L
H
H
L
L
H
H
1375
2750
Distance from left wall [mm]
Fig. 2. Thermocouple positioning in the exposed (a) back wall and (b) ceiling showing (H)igh and (L)ow thermocouple density.
The wood cribs were placed on a false floor system comprised of high density stone wool, then plasterboard, then MDF attached to a structure made of 50 mm timber joists, which provided an air gap. This false floor system was then placed on top of four load cells, which recorded the mass loss rate, enabling the crib heat release rate to be calculated independently of the CLT contribution. The same scales were used to record the mass of the exposed CLT panels before and after each test. This provided the approximate mass loss, allowing an estimate of the effective heat of combustion of the CLT. Thirteen bi-directional McCaffrey probes [19] were installed in the doorway to measure the gas velocities in and out of the compartment. Thermocouples were placed at the same height as each of the probes, along with two oxygen sensors at 0.4 m and 0.6 m from the top of the opening, enabling further calorimetry to be undertaken on the gases leaving the compartment, thus allowing internal heat release rate to be calculated independently [20]:
dering of the char. It is worth noting in the later stages of the test that the total heat release rate shown, as calculated through calorimetry, will be significantly affected by dilution of the collected gases, and as such may give an artificially low reading. The internal heat release rate calculations are also extremely sensitive to oxygen concentration in the opening, and may be over-estimating the internal heat release rate. No external flaming was observed after around 22 min, and thus from this point onwards internal heat release rate equals total heat release rate. Fig. 4 shows the gas-phase temperatures at a thermocouple tree in the centre of the compartment, also showing a decreasing trend from around 15 min. The solid-phase temperature profiles in Figs. 5 and 6 also show a decline from around 17–23 min, corresponding to the onset of extinction. The test was halted after 92 min, at which point water was applied to quench any continuing smouldering.
⎡ ⎤ ̇ = EO ⎢ XO2,∞ − XO2 ⎥ ṁ in MO2 (1 − X H O ∞) Qint 2 2 , Ma ⎣ (1−XO2 ) ⎦
Firepoint theory can be applied directly at the char-timber interface using the experimental results to determine if the relationship holds in large-scale applications. For direct application of Eq. (1), the material properties Lv, ΔHc, and ϕ must be determined, and the heat flux components calculated. The incident heat flux q⋅e′ ′ can be approximated from the temperature data at the timber surface, and heat losses q⋅′f ′ estimated by summation of the radiative losses, convective losses, and the conductive losses into the sample. For an initial analysis with fewer assumptions however, Eq. (1) can be generalised to provide a more fundamental solution to the firepoint equation. " S = q⋅"f + q⋅e" − q⋅l" − L v m⋅ cr (4)
4. Application of firepoint theory at the char line
(3)
where XO2 is the measured volumetric oxygen concentration in the outflow, taken as the average of both oxygen sensors, and min is the mass flow rate into the compartment. This method requires that the oxygen concentration in the outflow is uniform (or well-mixed), which is the case for fully-developed (post-flashover) fires. For well-ventilated fires this expression may result in an overestimation of the HRR due to the lower uniformity of oxygen concentration in the exhaust. Nonetheless, this expression provides an upper bound for the heat released inside the compartment because this concentration measurement is obtained at the top of the vent, where lowest concentrations of oxygen are expected. This calculation is expected to provide a slight overestimation of the HRR since a correction for inefficient combustion has not been applied due to the lack of reliable CO data.
Combining all the heat flux terms to give a net heat flux q⋅n′ ′, and setting S to zero to represent extinction gives: " q⋅n" = L v m⋅ cr (5)
3.3. Summary of results Flashover for Test β−1 took place at 8 min 33 s from ignition of the cribs, shortly after ignition of the exposed back wall. Flashover was defined by the rapid increase in HRR from 1550 to 6210 kW, as seen in Fig. 3, before a short period of burning accompanied by external flaming up to 15 min characterised by a HRR decreasing to 4500 kW. A rapid decrease in HRR and cessation of external flaming precedes autoextinction, which occurred after 21 min which correlates well with the external HRR shown in Fig. 3. This was observed as a reduction in external flaming, until all flaming was internal, at which point the area of burning wall started to reduce. The continuing internal HRR can be partially attributed to smoul-
Fig. 3. Heat release rate versus time for Test β−1.
4
Fire Safety Journal xxx (xxxx) xxx–xxx
A.I. Bartlett et al.
Gas Temperature [oC]
1400
respectively. This equation was applied to each set of thermocouple readings in the exposed wall and the exposed ceiling of Test β−1. The char depth was found through interpolation of the temperature profile [21]. The thermal gradients in the char and timber were calculated based on temperature data within the timber. A char conductivity of 0.25 W/mK was assumed [22], and a timber conductivity of 0.18 W/ mK [22]. Tewarson and Pion [23] experimentally determined values for the heat of gasification, Lg, for various solids using differential scanning calorimetry (DSC). For timber, they found a heat of gasification of 1.82 kJ/g. The heat of gasification includes the heat required to raise the solid to its pyrolysis temperature. Assuming a pyrolysis temperature of 300 °C [3,23] (noting that some pyrolysis will occur below this temperature, however the mass loss will be low and can be neglected), the heat of pyrolysis can be calculated from [20]:
260mm 460mm 660mm 860mm 1060mm 1260mm 1460mm 1660mm 1860mm 2060mm
1200 1000 800 600 400 200 0
0
10
20
30
40 50 60 Time [min]
70
80
90
Fig. 4. Gas temperature profile in the centre of the compartment for Test β−1.
Temperature [oC]
1200
L v =Lg −
9mm 20mm
800
40mm
600
60mm 80mm
400 200 0
0
10
20
30
40
50
60
70
80
90
Time [min] Fig. 5. Typical temperature profile in the back wall of Test β−1 as a function of distance from initial CLT surface.
Temperature [oC]
1200 800 600 400 200 0
0
10
20
30
40 50 60 Time [min]
70
80
90
The net heat flux at the char line (the area of interest, as is this is the boundary of the reaction zone) can be solved by a simple energy balance, such that the net heat flux will be equal to the energy conducted through the char layer less the energy losses through conduction into the timber. These heat fluxes can be obtained at a given timestep using a discretised version of Fourier's Law, along with the solid-phase thermocouple data, yielding (assuming a char formation temperature of 300 °C):
1 ⎛ Tc−300 300−Tw ⎞ −k w ⎜k c ⎟ Lv ⎝ ∆xc ∆x w ⎠
Cp dT
(7)
(8)
where the heat of combustion is assumed to be 17.5 MJ/kg (literature values typically range from 15 MJ/kg [4] to 20 MJ/kg [22]). It is evident from Fig. 7 that the energy balance approach and the calorimetry approach give similar values for the mass loss rate after the initial peak, and therefore these mass loss rates can be used to estimate time to extinction. There is significant variation in the peak values calculated due to variations in the effective heat of combustion during this burning period, and insufficient resolution in the temperature data to form an accurate thermal gradient. However, since it is the decay phase that is of interest, these discrepancies are unimportant for the present analysis. Making an initial assumption of ambient oxygen concentration, this gives an extinction time of around 21 min using firepoint theory for the wall, and around 20 min for the ceiling. Visual observations during the test showed that initial extinction occurred around 21 min, showing a very good correlation to the predictions of firepoint theory. Extinction occurred gradually over 3–4 min, with external flaming reducing and the flaming area on the internal surfaces gradually decreasing. This suggests that it may be possible to use firepoint theory to predict auto-extinction in a compartment with timber surfaces (in which delamination does not occur), with critical mass loss rates determined from bench-scale testing, given the right inputs for Eq. (1) – and obviously once considerable additional research has been undertaken to confirm this result.
Fig. 6. Typical temperature profile in the ceiling of Test β−1 as a function of distance from initial CLT surface.
m⋅ " =
∞
Q m⋅ " = CLT A∆Hc
Surface 5mm 9mm 16mm 20mm 25mm 30mm 35mm 40mm 50mm 60mm
1000
Tp
where Cp is the specific heat capacity of the timber (using temperature dependent values from Eurocode 5 [24]), and T∞ and Tp are the ambient and pyrolysis temperatures respectively. This gives a heat of vaporisation of around 1.1 kJ/g. The resulting mass loss rates are shown in Fig. 7 (dashed lines) along with the critical mass loss rates determined from bench-scale testing [5,6]. The values in Fig. 7 are calculated over several locations on each surface which had sufficient temperature resolution to apply Eq. (6). No obvious variation in mass loss rate was observed as a function of location. It can be seen from Fig. 7 that the method is inherently robust – due to the steep gradient of the mass loss curve, small changes in the experimental values of critical mass flux will not lead to major changes in the predicted time to extinction. To verify this method, the mass loss rate was also estimated from the calorimetry. The HRR of the CLT was calculated by subtracting the crib HRR (calculated from measured mass loss data) from the total HRR. This was converted to mass loss rate:
Surface
1000
∫T
5. Effects of delamination (6)
5.1. Test observations
where k is the thermal conductivity, T the temperature, and x the position, with subscripts c and w referring to the char and wood
It should be noted that the CLT heat release rate and mass loss rate 5
Fire Safety Journal xxx (xxxx) xxx–xxx
Mass loss rate [g/m2s]
A.I. Bartlett et al.
(a)
20
20
15
15
10
10
5
5
0
0
10
20 30 Time [min]
40 (b)
Mass loss rate
0
0
10
Calorimetry
20 30 Time [min]
40
Extinction
Fig. 7. Calculated CLT mass loss rate at the (a) back wall and (b) ceiling for Test β−1, with critical value from FPA testing [5,6] indicated.
5.2. Relation to firepoint theory
curves presented in previous sections are average values over the whole exposed surface. In reality, localised phenomena, such as delamination, could lead to increased mass loss rates over a small area. Delamination is a phenomenon in which the outermost lamella, or part thereof, detaches from the lamellae behind it, effectively reducing the char layer thickness at that position and thus increasing the net heat flux into the sample (and mass loss rate). After extinction was reached, a small area towards the right side of the back wall delaminated at around 27 min into the test, as shown in Fig. 8, leading to a locally increased mass loss rate due to an effective increase in the exposed fuel areas, and bringing the mass loss rate above the threshold for re-ignition. It is noteworthy that the delamination occurred locally, with only an area the width of one plank (about 200 mm wide) detaching, and thus avoiding a secondary flashover, which would likely have occurred if the entire lamella had fallen off at once. This localised burning continued for approximately 25 min, before again reaching local extinction. During this time, a second area of similar size delaminated on the left side of the back wall, burning for approximately 20 min before again extinguishing. Neither of these small burning regions had significant effects on the total heat release rate, and the general downward trends in HRR, solid-phase temperature, and gas-phase temperature continued.
Unfortunately, due to the small areas of re-ignition relative to the exposed surface area, this phenomenon was not captured at sufficient resolution by temperature data, and thus only a qualitative analysis is possible. Current knowledge of delamination and its causes is limited, and thus it is not currently possible to accurately predict whether (or when) it will occur. Applying the firepoint equation to a delaminated area becomes more difficult due to the complex boundary conditions present; the boundary to consider is the front of the underlying lamella. A partially delaminated lamella, as observed in Fig. 8, will partially shield the underlying timber from incoming radiation and also from radiation losses, whilst simultaneously re-radiating heat to the freshly exposed timber. This is a complex situation, the physics of which cannot be adequately captured by the current model. Further research is needed in particular on the causes of delamination. 5.3. Delamination-dominated fire dynamics In the duplicate Test β−2, the onset of delamination occurred before initial auto-extinction had been achieved, and the continued, localised delamination led to ignition of the second lamella, and autoextinction was not achieved, similar to the findings of [14,15]. The mass loss for Test β−2, as calculated by Eq. (8), is compared to that of Test β−1 in Fig. 9. Because of the delamination and the resulting, aforementioned changes in boundary conditions this caused, Eq. (6) is inappropriate for this scenario. In Fig. 9, times are adjusted such that t=0 is the time at which flashover occurred. It can be seen that both
Fig. 9. Comparison of calculated CLT mass loss rates determined from HRR for both βconfiguration tests.
Fig. 8. Localised re-ignition caused by local delamination of a single timber plank 30 min into the test.
6
Fire Safety Journal xxx (xxxx) xxx–xxx
A.I. Bartlett et al.
experiments follow similar trends for approximately the first 15 min, before the onset of delamination led to an increase in burning rate in Test β−2, and before the delamination and re-flashover cycle then repeats (twice actually). It can be seen from Fig. 9 that the mass loss rate never drops below around 6 g/m2s, and thus auto-extinction would not be expected based on the results in [5–7]. Further details of Test β−2 are given in [18].
[2] A.H. Buchanan, Structural Design for Fire Safety, John Wiley and Sons, Chichester, 2001. [3] I.S. Wichman, A. Atreya, A simplified model for the pyrolysis of charring materials, Combust. Flame 68 (1987) 231–247. http://dx.doi.org/10.1016/0010-2180(87) 90002-2. [4] D. Rasbash, D. Drysdale, D. Deepak, Critical heat and mass transfer at pilot ignition and extinction of a material, Fire Saf. J. 10 (1986) 1–10. http://dx.doi.org/ 10.1016/0379-7112(86)90026-3. [5] A.I. Bartlett, R.M. Hadden, L.A. Bisby, B. Lane, Auto-extinction of engineered timber: the application of firepoint theory, Interflam, Proceedings of the Fourteenth International Conference and Exhibition on Fire Science and Engineering, Interscience Communications, 2016, pp. 1381–1391. [6] A.I. Bartlett, R.M. Hadden, L.A. Bisby, B. Lane, Auto-extinction of engineered timber as a design methodology, in: Proceedings of the 14th World Conference on Timber Engineering, TU Wien, 2016. [7] R. Emberley, A. Inghelbrecht, Z. Yu, and J.L. Torero, Self-extinction of timber, Proceedings of the Combustion Institute, In Press, 2017, http://dx.doi.org/10. 1016/j.proci.2016.07.077. [8] D. Drysdale, An Introduction to Fire Dynamics, John Wiley and Sons, Chichester, 2011 (http://dx.doi.org/10.1002/9781119975465). [9] E. Butcher, J. Clarke, G. Bedford, A fire test in which furniture was the fuel, Fire Research Station, Fire Research Note No. 695, Borehamwood, UK, 26 p, 1968. [10] A. Frangi, M. Fontana, Fire performance of timber structures under natural fire conditions, in: Proceedings of the Eighth International Symposium, International Association for Fire Safety Science, pp. 279–290, 2005, http://dx.doi.org/10.3801/ iafss.fss.8-279. [11] T. Hakkarainen, Post-flashover fires in light and heavy timber construction compartments, J. Fire Sci. 20 (2002) 133–175. http://dx.doi.org/10.1177/ 0734904102020002074. [12] A. Frangi, G. Bochicchio, A. Ceccotti, M.P. Lauriola, Natural full-scale fire test on a 3 storey XLam timber building, in: Proceedings of the 10th World Conference on Timber Engineering, Miyazaki, Japan, 2008. [13] X. Li, X. Zhang, G. Hadjisophocleous, C. McGregor, Experimental study of combustible and non-combustible construction in a natural fire, Fire Technol. 51 (2014) 1447–1474. http://dx.doi.org/10.1007/s10694-014-0407-4. [14] X. Li, C. McGregor, A. Medina, X. Sun, D. Barber, G. Hadjisophocleous, Real-scale fire tests on timber constructions, in: Proceedings of the 14th World Conference on Timber Engineering, TU Wien, 2016. [15] R. Crielaard, J.-.W. van de Kuilen, T. Karel, G. Ravenshorst, P. Steenbakkers, A. Breunese, Self-extinguishment of cross-laminated timber, in: Proceedings of the 14th World Conference on Timber Engineering, TU Wien, 2016. [16] BSI, BS EN 520: Gypsum plasterboards. Definitions, requirements and test methods, British Standards Institution, London, 2004. [17] CEN, Eurocode 1. Actions on structures, Part 1-2: General actions – Actions onstructures exposed to fire, European Committee for Standardisation, Brussels, 2002. [18] R.M. Hadden, A.I. Bartlett, J.P. Hidalgo, S. Santamaria, F. Wiesner, L.A. Bisby, S. Deeny, B. Lane, Effects of exposed engineered timber construction systems on compartment fire dynamics, Fire Safety Science – in: Proceedings of the Twelfth International Symposium, International Association for Fire Safety Science, 2017. [19] B. McCaffrey, G. Heskestad, A robust bidirectional low-velocity probe for flame and fire applications, Combust. Flame 26 (1976) 125–127. http://dx.doi.org/10.1016/ 0010-2180(76)90062-6. [20] M.L. Janssens, Measuring rate of heat release by oxygen consumption, Fire Technol. 27 (1991) 234–249. http://dx.doi.org/10.1007/BF01038449. [21] A.I. Bartlett, R.M. Hadden, L.A. Bisby, A. Law, Analysis of cross-laminated timber charring rates upon exposure to non-standard heating conditions, Proceedings of 14th International Conference and Exhibition on Fire and Materials, Interscience Communications, 2015., pp. 667–681. [22] A. Inghelbrecht, Evaluation of the Burning Behaviour of Wood Products in the Context of Structural Fire Design, The University of Queensland, Queensland, Australia, 2014, p. 88 (MSc thesis). [23] A. Tewarson, R.F. Pion, Flammability of plastics—I. Burning intensity, Combust. Flame 26 (1976) 85–103. http://dx.doi.org/10.1016/0010-2180(76)90059-6. [24] CEN, Eurocode 5. Design of Timber Structures, Part 1-2: General. Structual fire design, European Committee for Standardisation, Brussels, 2004.
6. Conclusions and further work The research and analysis presented in this paper have demonstrated that application of firepoint theory, directly at the charline, can adequately describe the processes of auto-extinction of timber in compartments with large exposed areas of timber. To apply this concept in design, knowledge of the incident heat flux and heat losses, as a function of fuel load, is necessary to allow the energy balance to be calculated without experimental in-depth temperature measurements. These will be a function of surface temperature (for incident heat flux), exposed timber configuration (for view factor calculation for incident heat flux), and compartment geometry (for convective and radiative losses). Delamination of fire-exposed CLT has been shown to have a significant impact on the potential for achieving auto-extinction, thus increasing the risk of localised re-ignition or a global increase in burning rate, which may prevent auto-extinction completely. If the concept of auto-extinction is to be utilised in design, further work exploring the fundamental causes of (and demonstrated means to mitigate) delamination in fire is essential. Along with selecting the correct inputs to Eq. (1), understanding delamination and the resulting changes in boundary conditions for the application of firepoint theory will hopefully, with additional research, provide a means for architects’ aspirations of exposed timber surfaces to be realised. Acknowledgements The authors gratefully acknowledge funding from Arup and the University of Edinburgh to undertake the testing described herein, and from Arup and the EPSRC through iCASE Studentship 14220013. The authors would also like to acknowledge the support from KLH for supplying the timber, Rockwool International A/S for supplying stone wool insulation and SWG for supplying the fixings. The technical assistance from Nikolai Gerasimov, Timothy Putzien and Mark Fenton was much appreciated. Arup, BRE, IFIC Forensics, and The Royal Academy of Engineering are acknowledged for their generous, on-going support to Fire Safety Engineering research at the University of Edinburgh. References [1] F.L. Browne, Theories of the combustion of wood and its control, United States, Department of Agriculture Forest Service, Report No. 2136, Madison, WI, 72 p, 1958.
7
Contents lists available at ScienceDirect
Fire Safety Journal journal homepage: www.elsevier.com/locate/firesaf
Auto-extinction of engineered timber: Application to compartment fires with exposed timber surfaces ⁎
Alastair I. Bartletta, Rory M. Haddena, , Juan P. Hidalgoa, Simón Santamariaa, Felix Wiesnera, Luke A. Bisbya, Susan Deenyb, Barbara Laneb a b
School of Engineering, The University of Edinburgh, The King's Buildings, Mayfield Road, Edinburgh EH9 3JL, United Kingdom Arup, Fitzroy Street, London W1T 4BQ, United Kingdom
A R T I C L E I N F O
A BS T RAC T
Keywords: Compartment fires Heat transfer Extinction Performance-based design Firepoint theory
A series of compartment fire tests with multiple exposed timber surfaces have been undertaken to explore the effect of exposed timber on the fire dynamics and the potential for auto-extinction. A test with exposed wall and ceiling achieved auto-extinction after approximately 21 min. Firepoint theory is applied using temperature data at the charline, is shown to predict a mass loss rate dropping below the critical value at 20–21 min, and thus is successful in predicting auto-extinction. Additional uncertainties caused by delamination are explored, and recommendations for the use of auto-extinction in design are given.
1. Introduction Engineered timber products are continuing to increase in popularity, with factors such as sustainability, speed of construction, and aesthetics becoming ever more important. With the development of products such as cross-laminated timber (CLT) acting as structural wall and floor slabs, buildings are being constructed with the load-bearing structure formed predominantly from structural, engineered timber. The architectural aspiration for such buildings is often to have some of this structural timber exposed. The effect of exposed combustible compartment linings challenges many assumptions used in the assessment of compartment fire behaviour. One method to support robust design is to ensure that the timber linings auto extinguish after the compartment fuel load has been consumed. This requires an understanding of the energy balance at the char line, and an investigation of the critical factors at large scale. This presents an opportunity to explore the effects of multiple exposed timber surfaces within a compartment on the compartment fire dynamics. This paper focusses on the contribution of exposed timber surfaces to the fuel load during the cooling phase of a compartment fire, and interrogates the condition(s) under which the burning timber is expected to auto-extinguish. 2. Combustion of timber Timber begins to pyrolyse upon exposure to an external heat source, producing flammable and inert gases, tars, and a rigid,
⁎
carbonaceous char layer [1]. Flaming ignition of these flammable gases will occur when the generated mixture falls within the flammability limits, i.e. the air to fuel ratio falls within the correct range and the gases have sufficient energy. Smouldering ignition is also possible, but does not usually occur simultaneously with significant flaming combustion [1,2]. As a result, under flaming conditions, the char layer will continue to increase in thickness [1], reducing the rate of heat transfer to the virgin timber and resulting in a subsequent gradual decline in pyrolysis rate and hence mass flux of pyrolyzate [3]. 2.1. Extinction The “opposite” of flaming ignition is flaming extinction, which occurs when volatiles cease being produced in sufficient quantities to form a stable flame. This can be expressed in terms of the Damköhler number (the ratio of diffusion phenomena to kinetic timescales), which increases as a function of flame temperature and residence time (the duration the pyrolyzate remains in the reaction zone). Extinction will occur if the Damköhler number drops below a critical value, which can be achieved by reducing either the flame temperature or the residence time. Extinction can therefore occur due to external suppression (most commonly due to the application of water to cool the reaction zone); burnout of the pyrolysing fuel; or a reduction in the net energy supplied to the remaining unburned fuel. This will in turn result in a reduction in the mass of volatiles produced, which, if sufficient, will result in extinction. Due to the aforementioned decline in pyrolysis rate due to
Corresponding author. E-mail address: [email protected] (A.I. Bartlett).
http://dx.doi.org/10.1016/j.firesaf.2017.03.050 Received 15 February 2017; Accepted 27 March 2017 0379-7112/ © 2017 Published by Elsevier Ltd.
Please cite this article as: Bartlett, A.I., Fire Safety Journal (2017), http://dx.doi.org/10.1016/j.firesaf.2017.03.050
Fire Safety Journal xxx (xxxx) xxx–xxx
A.I. Bartlett et al.
subscripts
Nomenclature A Cp E ΔHc k H L m⋅ ′ ′ M q⋅′ ′ O Q S T W x X
area (m2) specific heat capacity (kJ/kg K) energy consumed per unit mass (kJ/g) heat of combustion (kJ/g) thermal conductivity (W/mK) height (m) latent heat (kJ/g) mass flux (g/m2s) molar mass (kg/kmol) heat flux (kW/m2) opening factor (m−1/2) heat release rate (kW) extinction parameter (kW/m2) temperature (°C) molecular mass (g/mol) position (m) volumetric concentration
a c CLT cr e f f g H2O i in int l n O2 p t v
Greek
v w
ϕ
critical ratio
∞
the same critical heat flux should not necessarily be expected. As with the experiments in [4] the critical mass flux for extinction increased with a reduction in oxygen concentration, but airflow was found to have no noticeable effect. In similar tests on Australian softwoods, Emberley et al. [7] found a critical mass flux of 4.0 g/m2s.
the build-up of a char layer, a burning timber sample has the potential to auto-extinguish naturally. To understand this condition in more detail, firepoint theory can be used to analyse the energy balance between the char and the virgin timber. 2.2. Firepoint theory
2.3. Compartment fire behaviour
Rasbash et al. [4] present firepoint theory as a means of determining ignition criteria for PMMA. They conducted a series of experiments on PMMA samples heated from above with a radiant panel to determine the effects of incident heat flux, air flow and oxygen concentration on the critical mass flux for ignition. In a range close to the critical heat flux for piloted ignition (12 kW/m2 to 19 kW/m2), the critical mass flux was found to increase as a function of heat flux, from about 3.8 g/m2s to 5.2 g/m2s; thereafter becoming independent of external heat flux. This initial variation was attributed to the variations in flame behaviour with lower heat fluxes. The effects of airflow around the sample were also investigated; an initial drop from around 5.3 g/m2s at 0lpm airflow to 3.2 g/m2s at 30 lpm, rising again to around 5.0 g/m2s at 60 lpm was observed. Reducing the oxygen concentration below ambient values resulted in a sharp increase in critical mass flux from around 3.3 g/m2s to 10.4 g/m2s at 19% O2. Rasbash et al. [4] concluded that firepoint theory may be used to determine if a material will continue to burn in the absence of a supporting heat flux:
S=(ϕ∆Hc − L v ) m⋅ cr" + q⋅e" − q⋅l"
air char CLT contribution critical external final flames gasification water vapour initial incoming air internal losses net oxygen pyrolysis total vaporisation ventilation wood ambient
Considerable research has been undertaken on compartment fires, the key aspects of which are summarised by Drysdale [8]. Wooden cribs are typically used as the fuel load, as they have been shown to be representative of typical room furnishings [9], and have good repeatability. Numerous correlations exist for the burning rate of cribs, and this is typically related to the opening factor of a compartment – calculated by:
O=
At Av Hv
(2)
where At is the total internal surface area excluding the floor and the opening, Av is the ventilation area, and Hv the ventilation height. Compartment fires can be said to follow three main stages: 1) the initial growth phase, in which burning is fuel-controlled; 2) the fullydeveloped, post-flashover phase, in which burning is ventilationcontrolled (Regime 1), and any excess fuel will be burned outside the compartment in an external plume; and 3) the decay phase, in which the burning will transition back to fuel-control (Regime 2) before extinction is achieved due to burnout of the fuel load.
(1)
where ϕ is the critical ratio of convective heat transfer to the heat of combustion of the volatiles, ΔHc is the heat of combustion of the solid, Lv is the heat of pyrolysis, and q⋅e′ ′ and q⋅l′ ′ are the external heat flux and heat losses respectively. If S > 0, the flame will be sustained, but if S < 0, extinction will occur. This equation has previously been applied to small-scale spruce/fir timber samples in a fire propagation apparatus (FPA) [5,6], where it was shown that timber samples will auto-extinguish when the mass flux of volatiles drops below 3.48 g/m2s (at ambient oxygen concentrations), and can be predicted using the firepoint equation. In the FPA setup, this value was reached at incident heat fluxes at or below 31 kW/ m2 – however it should be noted that in a compartment fire setup, heat losses, oxygen concentrations, and boundary conditions may differ, so
2.4. Compartment fires with combustible surfaces A compartment with exposed combustible surfaces such as timber presents an additional complexity due to the additional fuel load present. The existing empirical correlations were tested in situations where the fuel load was spread over the floor, and may not necessarily be valid when part of the fuel load is on the walls or ceiling. In a compartment with combustible surfaces, these will ignite, and the flames from these large exposed areas will produce additional heat which may increase the rate of burning of the compartment fuel load. The involvement of exposed combustible surfaces will also increase the total heat release rate, thus reducing the time to flashover. 2
Fire Safety Journal xxx (xxxx) xxx–xxx
A.I. Bartlett et al.
another was approximately 0.20 – the same as Crielaard [15]. All timber surfaces were constructed of CLT panels 100 mm thick with 5 uniform lamellae of thickness 20 mm. A single opening 0.8 m wide and 2 m tall was present in the front wall, however once the floor system was installed, the effective height was reduced to 1.84 m, corresponding to an opening factor of around 19 m−1/2. Four wooden cribs were placed in the compartments with a total mass of approximately 14 kg each. This corresponds to a fuel load of around 127 MJ/m2, which is relatively low compared to a typical office space with a mean design fuel load of 420 MJ/m2 [17], however since the decay phase was the area of interest, a short fully-developed compartment fire phase was desired.
Eventually, the fuel load in the compartment will burn out, and the combustible surfaces will be the only fuel remaining. At this stage, an energy balance can be performed, for example by using firepoint theory, to determine if the radiative exchange is sufficient for continued burning, until the walls or ceiling eventually burn through. The majority of CLT compartment fire tests to date have had either completely protected or completely unprotected timber surfaces [10– 13], and confirm expectations that exposed timber will contribute significantly to the heat release rate and accelerate flashover. Li et al. [14] conducted three tests with some timber surfaces exposed – one with one wall exposed, one with two opposite walls exposed (4.5 m×2.5 m), and one with two perpendicular walls exposed (4.5 m×2.5 m and 3.5 m×2.5 m), corresponding to view factors to the side walls of 0.18 and 0.19 with opposite and perpendicular walls exposed respectively and 0.15 to the back wall with perpendicular walls exposed. Both tests with two surfaces exposed experienced a secondary flashover after the onset of delamination, whereas the third autoextinguished. Crielaard et al. [15] tested scaled compartments with internal dimensions 0.5 m×0.5 m×0.5 m with 1, 2, or 3 exposed surfaces. The view factor from any one surface to another is 0.20, comparable to the side walls in the tests of Li et al. [14]. The test with one surface, and one of the two tests with two surfaces eventually reached auto-extinction. The other test with two surfaces and the test with three surfaces however, did not reach this state, with the two exposed surfaces reaching a secondary flashover as in [14] due to delamination. The three exposed surfaces continued burning and showed no sign of decreasing their heat release rate (HRR).
3.1. Test outcomes Of the two β tests carried out, β−2 failed to reach auto-extinction due to progressive delamination of CLT panels. Further details of the failure mechanisms are available in [18]. Test β−1 achieved autoextinction, and is therefore the focus of the current paper. 3.2. Instrumentation Inconel sheathed Type-K thermocouples were inserted into the exposed timber surfaces to record the internal temperature profiles and enable an analysis of the thermal gradients. 101 thermocouples were embedded in the back wall, and 101 in the ceiling, at 13 different locations in each surface. Thermocouples were inserted into pre-drilled holes at nominal depths of 5 mm, 10 mm, 15 mm, 20 mm, 25 mm, 30 mm, 35 mm, 40 mm, 50 mm, 60 mm, and 80 mm (high-density) or 10 mm, 20 mm 40 mm, 60 mm, and 80 mm (low-density) from the exposed surface (positioning shown in Fig. 2). Metallic discs of 10 mm diameter and coated to match the emissivity of charred timber were embedded into the timber surfaces at each group of thermocouples to allow measurement of surface temperature and an estimation of the incident heat flux. Additionally, six trees of 12 Type-K thermocouples were placed inside the compartment to provide gas-phase temperatures at various locations in the compartment. The tests were undertaken under a large extraction hood fitted with oxygen, carbon dioxide, and carbon monoxide analysers, enabling calorimetry to be performed and providing an estimate of the total heat release rate.
3. Experimental configuration A series of five full-scale CLT compartment fire tests was carried out at the BRE Burn Hall to explore the changes in fire dynamics and potential for auto-extinction due to exposed timber surfaces contributing to the fire. For the current paper, the two tests in which back wall and ceiling were exposed (configuration β) are explored (as shown in Fig. 1). All other surfaces were encapsulated with a system consisting of fire-rated gypsum plasterboard (rated Type F according to BS EN 520 [16]) and stone wool. The internal dimensions (excluding encapsulation and floor system) were 2.75 m long, 2.75 m wide, and 2.95 m high. With encapsulation and the floor system considered, the view factor from one surface to
Fig. 1. Experimental compartment.
3
Fire Safety Journal xxx (xxxx) xxx–xxx
2810 L L L 0 (a)
0
H
H
L
L
H
H
L
L
H
H
1375
Distance from front [mm]
Distance from floor [mm]
A.I. Bartlett et al.
2750
Distance from left wall [mm]
2750 L 1375 L L 0 (b)
0
H
H
L
L
H
H
L
L
H
H
1375
2750
Distance from left wall [mm]
Fig. 2. Thermocouple positioning in the exposed (a) back wall and (b) ceiling showing (H)igh and (L)ow thermocouple density.
The wood cribs were placed on a false floor system comprised of high density stone wool, then plasterboard, then MDF attached to a structure made of 50 mm timber joists, which provided an air gap. This false floor system was then placed on top of four load cells, which recorded the mass loss rate, enabling the crib heat release rate to be calculated independently of the CLT contribution. The same scales were used to record the mass of the exposed CLT panels before and after each test. This provided the approximate mass loss, allowing an estimate of the effective heat of combustion of the CLT. Thirteen bi-directional McCaffrey probes [19] were installed in the doorway to measure the gas velocities in and out of the compartment. Thermocouples were placed at the same height as each of the probes, along with two oxygen sensors at 0.4 m and 0.6 m from the top of the opening, enabling further calorimetry to be undertaken on the gases leaving the compartment, thus allowing internal heat release rate to be calculated independently [20]:
dering of the char. It is worth noting in the later stages of the test that the total heat release rate shown, as calculated through calorimetry, will be significantly affected by dilution of the collected gases, and as such may give an artificially low reading. The internal heat release rate calculations are also extremely sensitive to oxygen concentration in the opening, and may be over-estimating the internal heat release rate. No external flaming was observed after around 22 min, and thus from this point onwards internal heat release rate equals total heat release rate. Fig. 4 shows the gas-phase temperatures at a thermocouple tree in the centre of the compartment, also showing a decreasing trend from around 15 min. The solid-phase temperature profiles in Figs. 5 and 6 also show a decline from around 17–23 min, corresponding to the onset of extinction. The test was halted after 92 min, at which point water was applied to quench any continuing smouldering.
⎡ ⎤ ̇ = EO ⎢ XO2,∞ − XO2 ⎥ ṁ in MO2 (1 − X H O ∞) Qint 2 2 , Ma ⎣ (1−XO2 ) ⎦
Firepoint theory can be applied directly at the char-timber interface using the experimental results to determine if the relationship holds in large-scale applications. For direct application of Eq. (1), the material properties Lv, ΔHc, and ϕ must be determined, and the heat flux components calculated. The incident heat flux q⋅e′ ′ can be approximated from the temperature data at the timber surface, and heat losses q⋅′f ′ estimated by summation of the radiative losses, convective losses, and the conductive losses into the sample. For an initial analysis with fewer assumptions however, Eq. (1) can be generalised to provide a more fundamental solution to the firepoint equation. " S = q⋅"f + q⋅e" − q⋅l" − L v m⋅ cr (4)
4. Application of firepoint theory at the char line
(3)
where XO2 is the measured volumetric oxygen concentration in the outflow, taken as the average of both oxygen sensors, and min is the mass flow rate into the compartment. This method requires that the oxygen concentration in the outflow is uniform (or well-mixed), which is the case for fully-developed (post-flashover) fires. For well-ventilated fires this expression may result in an overestimation of the HRR due to the lower uniformity of oxygen concentration in the exhaust. Nonetheless, this expression provides an upper bound for the heat released inside the compartment because this concentration measurement is obtained at the top of the vent, where lowest concentrations of oxygen are expected. This calculation is expected to provide a slight overestimation of the HRR since a correction for inefficient combustion has not been applied due to the lack of reliable CO data.
Combining all the heat flux terms to give a net heat flux q⋅n′ ′, and setting S to zero to represent extinction gives: " q⋅n" = L v m⋅ cr (5)
3.3. Summary of results Flashover for Test β−1 took place at 8 min 33 s from ignition of the cribs, shortly after ignition of the exposed back wall. Flashover was defined by the rapid increase in HRR from 1550 to 6210 kW, as seen in Fig. 3, before a short period of burning accompanied by external flaming up to 15 min characterised by a HRR decreasing to 4500 kW. A rapid decrease in HRR and cessation of external flaming precedes autoextinction, which occurred after 21 min which correlates well with the external HRR shown in Fig. 3. This was observed as a reduction in external flaming, until all flaming was internal, at which point the area of burning wall started to reduce. The continuing internal HRR can be partially attributed to smoul-
Fig. 3. Heat release rate versus time for Test β−1.
4
Fire Safety Journal xxx (xxxx) xxx–xxx
A.I. Bartlett et al.
Gas Temperature [oC]
1400
respectively. This equation was applied to each set of thermocouple readings in the exposed wall and the exposed ceiling of Test β−1. The char depth was found through interpolation of the temperature profile [21]. The thermal gradients in the char and timber were calculated based on temperature data within the timber. A char conductivity of 0.25 W/mK was assumed [22], and a timber conductivity of 0.18 W/ mK [22]. Tewarson and Pion [23] experimentally determined values for the heat of gasification, Lg, for various solids using differential scanning calorimetry (DSC). For timber, they found a heat of gasification of 1.82 kJ/g. The heat of gasification includes the heat required to raise the solid to its pyrolysis temperature. Assuming a pyrolysis temperature of 300 °C [3,23] (noting that some pyrolysis will occur below this temperature, however the mass loss will be low and can be neglected), the heat of pyrolysis can be calculated from [20]:
260mm 460mm 660mm 860mm 1060mm 1260mm 1460mm 1660mm 1860mm 2060mm
1200 1000 800 600 400 200 0
0
10
20
30
40 50 60 Time [min]
70
80
90
Fig. 4. Gas temperature profile in the centre of the compartment for Test β−1.
Temperature [oC]
1200
L v =Lg −
9mm 20mm
800
40mm
600
60mm 80mm
400 200 0
0
10
20
30
40
50
60
70
80
90
Time [min] Fig. 5. Typical temperature profile in the back wall of Test β−1 as a function of distance from initial CLT surface.
Temperature [oC]
1200 800 600 400 200 0
0
10
20
30
40 50 60 Time [min]
70
80
90
The net heat flux at the char line (the area of interest, as is this is the boundary of the reaction zone) can be solved by a simple energy balance, such that the net heat flux will be equal to the energy conducted through the char layer less the energy losses through conduction into the timber. These heat fluxes can be obtained at a given timestep using a discretised version of Fourier's Law, along with the solid-phase thermocouple data, yielding (assuming a char formation temperature of 300 °C):
1 ⎛ Tc−300 300−Tw ⎞ −k w ⎜k c ⎟ Lv ⎝ ∆xc ∆x w ⎠
Cp dT
(7)
(8)
where the heat of combustion is assumed to be 17.5 MJ/kg (literature values typically range from 15 MJ/kg [4] to 20 MJ/kg [22]). It is evident from Fig. 7 that the energy balance approach and the calorimetry approach give similar values for the mass loss rate after the initial peak, and therefore these mass loss rates can be used to estimate time to extinction. There is significant variation in the peak values calculated due to variations in the effective heat of combustion during this burning period, and insufficient resolution in the temperature data to form an accurate thermal gradient. However, since it is the decay phase that is of interest, these discrepancies are unimportant for the present analysis. Making an initial assumption of ambient oxygen concentration, this gives an extinction time of around 21 min using firepoint theory for the wall, and around 20 min for the ceiling. Visual observations during the test showed that initial extinction occurred around 21 min, showing a very good correlation to the predictions of firepoint theory. Extinction occurred gradually over 3–4 min, with external flaming reducing and the flaming area on the internal surfaces gradually decreasing. This suggests that it may be possible to use firepoint theory to predict auto-extinction in a compartment with timber surfaces (in which delamination does not occur), with critical mass loss rates determined from bench-scale testing, given the right inputs for Eq. (1) – and obviously once considerable additional research has been undertaken to confirm this result.
Fig. 6. Typical temperature profile in the ceiling of Test β−1 as a function of distance from initial CLT surface.
m⋅ " =
∞
Q m⋅ " = CLT A∆Hc
Surface 5mm 9mm 16mm 20mm 25mm 30mm 35mm 40mm 50mm 60mm
1000
Tp
where Cp is the specific heat capacity of the timber (using temperature dependent values from Eurocode 5 [24]), and T∞ and Tp are the ambient and pyrolysis temperatures respectively. This gives a heat of vaporisation of around 1.1 kJ/g. The resulting mass loss rates are shown in Fig. 7 (dashed lines) along with the critical mass loss rates determined from bench-scale testing [5,6]. The values in Fig. 7 are calculated over several locations on each surface which had sufficient temperature resolution to apply Eq. (6). No obvious variation in mass loss rate was observed as a function of location. It can be seen from Fig. 7 that the method is inherently robust – due to the steep gradient of the mass loss curve, small changes in the experimental values of critical mass flux will not lead to major changes in the predicted time to extinction. To verify this method, the mass loss rate was also estimated from the calorimetry. The HRR of the CLT was calculated by subtracting the crib HRR (calculated from measured mass loss data) from the total HRR. This was converted to mass loss rate:
Surface
1000
∫T
5. Effects of delamination (6)
5.1. Test observations
where k is the thermal conductivity, T the temperature, and x the position, with subscripts c and w referring to the char and wood
It should be noted that the CLT heat release rate and mass loss rate 5
Fire Safety Journal xxx (xxxx) xxx–xxx
Mass loss rate [g/m2s]
A.I. Bartlett et al.
(a)
20
20
15
15
10
10
5
5
0
0
10
20 30 Time [min]
40 (b)
Mass loss rate
0
0
10
Calorimetry
20 30 Time [min]
40
Extinction
Fig. 7. Calculated CLT mass loss rate at the (a) back wall and (b) ceiling for Test β−1, with critical value from FPA testing [5,6] indicated.
5.2. Relation to firepoint theory
curves presented in previous sections are average values over the whole exposed surface. In reality, localised phenomena, such as delamination, could lead to increased mass loss rates over a small area. Delamination is a phenomenon in which the outermost lamella, or part thereof, detaches from the lamellae behind it, effectively reducing the char layer thickness at that position and thus increasing the net heat flux into the sample (and mass loss rate). After extinction was reached, a small area towards the right side of the back wall delaminated at around 27 min into the test, as shown in Fig. 8, leading to a locally increased mass loss rate due to an effective increase in the exposed fuel areas, and bringing the mass loss rate above the threshold for re-ignition. It is noteworthy that the delamination occurred locally, with only an area the width of one plank (about 200 mm wide) detaching, and thus avoiding a secondary flashover, which would likely have occurred if the entire lamella had fallen off at once. This localised burning continued for approximately 25 min, before again reaching local extinction. During this time, a second area of similar size delaminated on the left side of the back wall, burning for approximately 20 min before again extinguishing. Neither of these small burning regions had significant effects on the total heat release rate, and the general downward trends in HRR, solid-phase temperature, and gas-phase temperature continued.
Unfortunately, due to the small areas of re-ignition relative to the exposed surface area, this phenomenon was not captured at sufficient resolution by temperature data, and thus only a qualitative analysis is possible. Current knowledge of delamination and its causes is limited, and thus it is not currently possible to accurately predict whether (or when) it will occur. Applying the firepoint equation to a delaminated area becomes more difficult due to the complex boundary conditions present; the boundary to consider is the front of the underlying lamella. A partially delaminated lamella, as observed in Fig. 8, will partially shield the underlying timber from incoming radiation and also from radiation losses, whilst simultaneously re-radiating heat to the freshly exposed timber. This is a complex situation, the physics of which cannot be adequately captured by the current model. Further research is needed in particular on the causes of delamination. 5.3. Delamination-dominated fire dynamics In the duplicate Test β−2, the onset of delamination occurred before initial auto-extinction had been achieved, and the continued, localised delamination led to ignition of the second lamella, and autoextinction was not achieved, similar to the findings of [14,15]. The mass loss for Test β−2, as calculated by Eq. (8), is compared to that of Test β−1 in Fig. 9. Because of the delamination and the resulting, aforementioned changes in boundary conditions this caused, Eq. (6) is inappropriate for this scenario. In Fig. 9, times are adjusted such that t=0 is the time at which flashover occurred. It can be seen that both
Fig. 9. Comparison of calculated CLT mass loss rates determined from HRR for both βconfiguration tests.
Fig. 8. Localised re-ignition caused by local delamination of a single timber plank 30 min into the test.
6
Fire Safety Journal xxx (xxxx) xxx–xxx
A.I. Bartlett et al.
experiments follow similar trends for approximately the first 15 min, before the onset of delamination led to an increase in burning rate in Test β−2, and before the delamination and re-flashover cycle then repeats (twice actually). It can be seen from Fig. 9 that the mass loss rate never drops below around 6 g/m2s, and thus auto-extinction would not be expected based on the results in [5–7]. Further details of Test β−2 are given in [18].
[2] A.H. Buchanan, Structural Design for Fire Safety, John Wiley and Sons, Chichester, 2001. [3] I.S. Wichman, A. Atreya, A simplified model for the pyrolysis of charring materials, Combust. Flame 68 (1987) 231–247. http://dx.doi.org/10.1016/0010-2180(87) 90002-2. [4] D. Rasbash, D. Drysdale, D. Deepak, Critical heat and mass transfer at pilot ignition and extinction of a material, Fire Saf. J. 10 (1986) 1–10. http://dx.doi.org/ 10.1016/0379-7112(86)90026-3. [5] A.I. Bartlett, R.M. Hadden, L.A. Bisby, B. Lane, Auto-extinction of engineered timber: the application of firepoint theory, Interflam, Proceedings of the Fourteenth International Conference and Exhibition on Fire Science and Engineering, Interscience Communications, 2016, pp. 1381–1391. [6] A.I. Bartlett, R.M. Hadden, L.A. Bisby, B. Lane, Auto-extinction of engineered timber as a design methodology, in: Proceedings of the 14th World Conference on Timber Engineering, TU Wien, 2016. [7] R. Emberley, A. Inghelbrecht, Z. Yu, and J.L. Torero, Self-extinction of timber, Proceedings of the Combustion Institute, In Press, 2017, http://dx.doi.org/10. 1016/j.proci.2016.07.077. [8] D. Drysdale, An Introduction to Fire Dynamics, John Wiley and Sons, Chichester, 2011 (http://dx.doi.org/10.1002/9781119975465). [9] E. Butcher, J. Clarke, G. Bedford, A fire test in which furniture was the fuel, Fire Research Station, Fire Research Note No. 695, Borehamwood, UK, 26 p, 1968. [10] A. Frangi, M. Fontana, Fire performance of timber structures under natural fire conditions, in: Proceedings of the Eighth International Symposium, International Association for Fire Safety Science, pp. 279–290, 2005, http://dx.doi.org/10.3801/ iafss.fss.8-279. [11] T. Hakkarainen, Post-flashover fires in light and heavy timber construction compartments, J. Fire Sci. 20 (2002) 133–175. http://dx.doi.org/10.1177/ 0734904102020002074. [12] A. Frangi, G. Bochicchio, A. Ceccotti, M.P. Lauriola, Natural full-scale fire test on a 3 storey XLam timber building, in: Proceedings of the 10th World Conference on Timber Engineering, Miyazaki, Japan, 2008. [13] X. Li, X. Zhang, G. Hadjisophocleous, C. McGregor, Experimental study of combustible and non-combustible construction in a natural fire, Fire Technol. 51 (2014) 1447–1474. http://dx.doi.org/10.1007/s10694-014-0407-4. [14] X. Li, C. McGregor, A. Medina, X. Sun, D. Barber, G. Hadjisophocleous, Real-scale fire tests on timber constructions, in: Proceedings of the 14th World Conference on Timber Engineering, TU Wien, 2016. [15] R. Crielaard, J.-.W. van de Kuilen, T. Karel, G. Ravenshorst, P. Steenbakkers, A. Breunese, Self-extinguishment of cross-laminated timber, in: Proceedings of the 14th World Conference on Timber Engineering, TU Wien, 2016. [16] BSI, BS EN 520: Gypsum plasterboards. Definitions, requirements and test methods, British Standards Institution, London, 2004. [17] CEN, Eurocode 1. Actions on structures, Part 1-2: General actions – Actions onstructures exposed to fire, European Committee for Standardisation, Brussels, 2002. [18] R.M. Hadden, A.I. Bartlett, J.P. Hidalgo, S. Santamaria, F. Wiesner, L.A. Bisby, S. Deeny, B. Lane, Effects of exposed engineered timber construction systems on compartment fire dynamics, Fire Safety Science – in: Proceedings of the Twelfth International Symposium, International Association for Fire Safety Science, 2017. [19] B. McCaffrey, G. Heskestad, A robust bidirectional low-velocity probe for flame and fire applications, Combust. Flame 26 (1976) 125–127. http://dx.doi.org/10.1016/ 0010-2180(76)90062-6. [20] M.L. Janssens, Measuring rate of heat release by oxygen consumption, Fire Technol. 27 (1991) 234–249. http://dx.doi.org/10.1007/BF01038449. [21] A.I. Bartlett, R.M. Hadden, L.A. Bisby, A. Law, Analysis of cross-laminated timber charring rates upon exposure to non-standard heating conditions, Proceedings of 14th International Conference and Exhibition on Fire and Materials, Interscience Communications, 2015., pp. 667–681. [22] A. Inghelbrecht, Evaluation of the Burning Behaviour of Wood Products in the Context of Structural Fire Design, The University of Queensland, Queensland, Australia, 2014, p. 88 (MSc thesis). [23] A. Tewarson, R.F. Pion, Flammability of plastics—I. Burning intensity, Combust. Flame 26 (1976) 85–103. http://dx.doi.org/10.1016/0010-2180(76)90059-6. [24] CEN, Eurocode 5. Design of Timber Structures, Part 1-2: General. Structual fire design, European Committee for Standardisation, Brussels, 2004.
6. Conclusions and further work The research and analysis presented in this paper have demonstrated that application of firepoint theory, directly at the charline, can adequately describe the processes of auto-extinction of timber in compartments with large exposed areas of timber. To apply this concept in design, knowledge of the incident heat flux and heat losses, as a function of fuel load, is necessary to allow the energy balance to be calculated without experimental in-depth temperature measurements. These will be a function of surface temperature (for incident heat flux), exposed timber configuration (for view factor calculation for incident heat flux), and compartment geometry (for convective and radiative losses). Delamination of fire-exposed CLT has been shown to have a significant impact on the potential for achieving auto-extinction, thus increasing the risk of localised re-ignition or a global increase in burning rate, which may prevent auto-extinction completely. If the concept of auto-extinction is to be utilised in design, further work exploring the fundamental causes of (and demonstrated means to mitigate) delamination in fire is essential. Along with selecting the correct inputs to Eq. (1), understanding delamination and the resulting changes in boundary conditions for the application of firepoint theory will hopefully, with additional research, provide a means for architects’ aspirations of exposed timber surfaces to be realised. Acknowledgements The authors gratefully acknowledge funding from Arup and the University of Edinburgh to undertake the testing described herein, and from Arup and the EPSRC through iCASE Studentship 14220013. The authors would also like to acknowledge the support from KLH for supplying the timber, Rockwool International A/S for supplying stone wool insulation and SWG for supplying the fixings. The technical assistance from Nikolai Gerasimov, Timothy Putzien and Mark Fenton was much appreciated. Arup, BRE, IFIC Forensics, and The Royal Academy of Engineering are acknowledged for their generous, on-going support to Fire Safety Engineering research at the University of Edinburgh. References [1] F.L. Browne, Theories of the combustion of wood and its control, United States, Department of Agriculture Forest Service, Report No. 2136, Madison, WI, 72 p, 1958.
7