Experimental study on downward flame spread characteristics under the influence of parallel curtain wall

Accepted Manuscript Experimental study on downward flame spread characteristics under the influence of parallel curtain wall Weiguang An, Rongliang Pan, Qingxuan Meng, Hongya Zhu PII: DOI: Reference:

S1359-4311(17)32935-6 http://dx.doi.org/10.1016/j.applthermaleng.2017.08.174 ATE 11057

To appear in:

Applied Thermal Engineering

Received Date: Revised Date: Accepted Date:

3 May 2017 30 August 2017 31 August 2017

Please cite this article as: W. An, R. Pan, Q. Meng, H. Zhu, Experimental study on downward flame spread characteristics under the influence of parallel curtain wall, Applied Thermal Engineering (2017), doi: http:// dx.doi.org/10.1016/j.applthermaleng.2017.08.174

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Experimental study on downward flame spread characteristics under the influence of parallel curtain wall Weiguang An1, 2 *, Rongliang Pan1, Qingxuan Meng1, Hongya Zhu2 †, 1

Key Laboratory of Gas and Fire Control for Coal Mines, School of Safety

Engineering, China University of Mining and Technology, Xuzhou, Jiangsu, 221008, PR China 2

Key Laboratory of Building Fire Protection Engineering and Technology of MPS, Tianjin, 300381, PR China

* †

Corresponding author. Tel.: +86 15895201749. E-mail address: [email protected]. Corresponding author. Tel.: +86 18202201771. E-mail address: [email protected]. 1

Abstract In this paper, experimental methods and theoretical analysis are employed to investigate effects of parallel curtain wall on downward flame spread characteristics of insulation materials on building facade. As the curtain wall spacing (D) rises, the front surface of facade flame becomes more irregular. For small spacing, the flame stretching phenomenon is obvious and periodical change of the flame height is observed. The average flame height (Hf) first drops and then rises as the spacing increases. The variation of Hf is significant as D≤11.5cm while indistinctive change is observed for D>11.5cm. The average maximum flame temperature first rises and then drops as D increases. There is a power function relationship between internal surface temperature of the curtain wall and D. The total heat feedback from the curtain wall to the facade decreases exponentially with the rising of D. A formula is proposed to predict the radiant heat feedback from the curtain wall, which is more dominant than the convective heat feedback. As D rises, the radiant heat feedback decreases, while the ratio of convective heat feedback to the total first rises and then drops. With the increasing of D, the average flame spread rate first rises and then drops, which is attributed to the competition of negative effect and positive effect of the curtain wall.

Keywords: building facade fire; downward flame spread; curtain wall spacing; heat transfer

2

1. Introduction The insulation materials or decorative materials are widely applied on building façade nowadays. Some of these materials are combustible, such as extruded polystyrene (XPS) foam, expanded polystyrene (EPS) foam and polyurethane (PU) foam, inducing considerable fire risk [1]. The fire safety of building facade has aroused researchers’ concern [2, 3]. Moreover, a curtain wall is always installed parallel to the building façade. Under this situation, a vertical channel is formed between the curtain wall and the building façade. The flame spread behavior will be special if a fire occurs on the building façade under the influence of the curtain wall. Flame spread over solid fuel is a classical research topic, which has been investigated by numerous researchers. Their works were thoroughly reviewed in Refs. [4-6]. Concerning on flame spread over building façade materials, especially polystyrene (PS) foams, Zhang et al. [7] investigated the ambient pressure effects and sample scale effects on horizontal flame spread over XPS foam, while Huang et al. [8] studied influences of altitude and sample orientation on heat transfer for flame spread over PS foams. Jiang et al. [9] deduced the correlation between the flame spread rate of PS foam and the sample width. Hajdukovic et al. [10] investigated fire performance and flame spread behavior of external thermal insulation composite system facades with EPS foam and thin rendering However, studies in term of channel effects on flame spread behaviors are far from enough. Tang et al. [11], Yang et al. [12] and Yao et al. [13] investigated the fire behavior in the horizontal channel, rather than the vertical channel. Under the influence of the vertical channel, the simulation of 3

back-layering length in tunnel fire and smoke movement characteristics in a mine laneway fire were studied by Wang et al. [14] and Fan et al. [15]. However, their focus was the smoke instead of the flame. Experimental study and three-dimensional modeling of turbulent burning of vertical parallel walls were conducted by Tamanini et al. [16] and Wang et al. [17, 18]. Their experimental results agreed well with the numerical results. Nevertheless, their work merely involved burning except flame spread. Experimental study and simulation were conducted by Wasan et al. to investigate the effect of the spacing between the combustible slab and the curtain wall [19, 20]. With the rising of the spacing, the flame spread rate decreased while the flame temperature first dropped and then increased. However, the results of numerical simulation conducted by An et al [21] indicated that the flame spread rate first rose and then dropped with the increase of the spacing. In addition, the presence of the curtain wall promoted the formation of the vortex field, producing the high-temperature zone on both sides of the flame centerline. In brief, previous researchers had not reached to a consistent conclusion concerning on the influence of the curtain wall. The effect of the curtain wall on characteristics of window ejected flame was studied by some researchers [15-18]. Lee et al. [22] deduced a function to describe the relationship between the spacing (D) and the flame height. A characteristic length was also proposed. Different characteristic lengths corresponded to different formulas for obtaining the flame heat flux. These results were verified under a different 4

atmospheric pressure [23]. Tang [24] also proposed a characteristic length l and established a mathematic model to predict the correlation between the flame height and D/l. Moreover, non-dimensional models were proposed by Tang et al. [25] to correlate the flame heights with the sloping curtain wall angles. The change of heat flux upon the facade with sloping curtain wall angle showed similar behavior as flame height. A correlation concerning on the relationship between the flame height and the flame heat flux was deduced [25]. The above literature review demonstrates that previous works focus on window ejected flame. However, scanty studies were conducted to explore effects of curtain wall on the façade flame spread, which is a fundamental problem in fire safety of buildings. In the present work, experiments were conducted to investigate the downward flame spread characteristics (flame shape, flame height, temperature field, heat transfer, flame spread distance and flame spread rate) of extruded polystyrene (XPS) foam, which is a typical insulation material. The parallel curtain wall was installed, and the spacing between the curtain wall and the facade was varied to explore their influence. The mechanism involved in curtain wall effects on flame spread characteristics was revealed. 2. Experimental methods and material The experimental setup is shown in Fig. 1. The experimental setup included XPS sample, curtain wall, vertical facade, camera, infrared camera, linear igniter, K-type thermocouple sequence and heat flow meter. The length, width, and thickness of XPS sample were 60, 10, and 4 cm, respectively. The XPS samples without fire retardant 5

additive were used for tests. Their fundamental physical properties at 300 K are presented in Table 1 [26-28]. The curtain wall was made of the phenolic foam, which was the incombustible insulation material with a thermal conductivity of 0.028 W m-1K-1 (evaluated at 300 K).The vertical facade was made of foam gypsum, whose thermal conductivity was 0.057 W m-1 K-1 (evaluated at 300 K). Since the thermal conductivity of the gypsum board is very small, the flame heat conduction to the vertical facade and its influence on flame development are negligible. The height and width of the curtain wall and the facade were 120 cm and 90 cm, respectively.

(a)

(b)

Fig. 1 Experimental setup (a) side view; (b) top view. 1–XPS sample; 2 – vertical facade; 3 –curtain wall; 4 –heat flow meter; 5 –K-type thermocouple sequence; 6 –infrared camera and 7 – camera. Table 1 Some physical properties of XPS a Material

k /W m-1 K-1

ρ/kg m-3

c/J kg-1 K-1 6

Tig/K

p

XPS a

0.029

34

1210

700.35

96.5%

k- thermal conductivity, ρ- sample density, c- specific heat, Tig -ignition temperature, p- porosity.

Both the camera and the infrared (IR) camera were positioned on the side of the XPS sample. The effective pixels of the camera was 6.14 million (16:9), and its frequency was 25 frames per second. The camera was employed to record the whole process of flame spread, and more importantly, to obtain real-time flame front position and flame height. With the video process software (Premiere), the flame-spread video was transformed into frames of images. A program created by the software MATLAB was used to convert every RGB image to the binary image. Then the zone with pixels whose values were 1 was deemed as the flame zone. The lower boundary of the flame zone was the flame front, whose position was recorded using the program. This method was defined as the image processing method, which has been employed in previous works [7, 29, 30]. This method was also used to obtain the length of the flame zone, i.e. the flame height. In addition, the highest shooting frequency of the IR camera was 100 HZ and the pixel number was 384×288. Its measuring range was -20 ~ 1200 °C with 2% measuring error. The input parameters of IR camera, especially the infrared emissivity, are very important to set. Parameters like ambient temperature, relative humidity and distance between the IR camera and XPS sample were measured and set before each test. To obtain the infrared emissivity of XPS flame, the flame temperature was simultaneously measured by the K-type thermocouple and the IR camera before experiments. The emissivity in the measuring software of the IR camera must be adjusted to obtain a temperature curve whose value 7

and varied trend approached to that measured by the thermocouple. Then, the emissivity that corresponded to this temperature curve was selected as the input parameter of the IR camera. In the present work, the infrared emissivity selected is 0.80, which approached the value (0.81) employed in De Ris’s work [31]. Two K-type thermocouple sequences were employed in this work. One was installed on the internal surface of the curtain wall, while the other 2 cm away from the sample surface in the vertical channel. Each sequence had three thermocouples. Their heights relative to the sample bottom were 20 cm, 30 cm and 40 cm. The diameter and response time of the K-type thermocouple were 0.5 mm and 0.03 s, respectively. The temperature measuring range was 0 ~ 1000 °C, with an accuracy of ±2.2 °C. The type of the water-cooled heat flux meter was GTW-10-32-485A, which could measure the total heat flux, including the convective and the radiant heat flux. As shown in Fig. 1, three heat flux meters were positioned along the sample side to measure the heat flux feedback from the curtain wall. The heat flux meter was a cylinder, whose measuring surface was approximately flush with the front surface of the XPS sample. Their installed heights relative to the sample end were 15 cm, 30 cm and 45 cm. During the experiment, the scale of the XPS sample remained constant, although the spacing (D) between the curtain wall and the facade varied. The experimental conditions are listed in Table 2. The ambient temperature and pressure of experimental location were about 11 °C and 101.2 kPa, respectively. A linear igniter was used to ignite the top of the XPS slab at the beginning of each experiment, and 8

each test was repeated three times. Table 2 Experimental conditions Condition 1 number

2

3

4

5

6

7

D (cm)

9

11.5

14.5

17.5

21

27

6

3. Results and Discussion 3.1 Flame shape and flame height The flame spread process is recorded using the camera, and the typical side views of the flame shape under different test conditions are presented in Fig. 2. It is found that the flame height changes with the varying of the spacing between the facade and the curtain wall. The height of the flame zone adhering to the back wall (i.e. the vertical facade) is larger than that keeping away from the back wall. As the spacing rises, the front surface of the flame becomes more irregular. This may be attributed to the restriction effect of the curtain wall on the front air entrainment. The increasing of the spacing weakens the restriction effect, leading to the strengthening of the front air entrainment. Thus the front surface of the flame becomes more irregular.

9

6 cm

9 cm

11.5cm

14.5cm

17.5cm

21 cm

27 cm

Fig. 2 Flame shapes under the influence of curtain wall with different spacing Employing the image processing method, the real-time flame height was obtained. When the spacing is narrow, the flame stretching phenomenon is obvious and periodical change of the flame height is observed. The typical example is the 6 cm-spacing, i.e., the minimum spacing for test. Its real-time flame height is shown in Fig. 3. The mechanism involved is analyzed as follows. The flame height is relatively stable at the initial stage of downward flame spread. With the forward moving of the flame front, more and more molten XPS is produced and adheres to the facade. These molten XPS can be ignited by the pulsatile flame, and thus the flame height rises. The increasing of the flame height leads to larger temperature difference between the top and the bottom of the vertical channel. More air is induced for combustion, which magnified the chimney effect and the stretching effect induced by the curtain wall, causing further increase of the flame height. With the cyclic action of the above 10

mechanism, the flame height can reach up to 90 cm. Subsequently, the flame height becomes dropping as most of the molten XPS is consumed and the fuel supply decreases. The chimney and stretching effects will be weakened according to the above mechanism, causing further decrease of the flame height. The minimum value is about 15 cm.

100 90

Flame height/cm

80 70 60 50 40 30 20 10 0

50

100

150

200

250

300

350

Time/s

Fig. 3 Variation of the flame height during downward flame spread process (D= 6cm) The flame heights at different time were averaged to obtain the average flame height. The average flame heights under different test conditions are presented in Fig. 4, which indicates that the average flame height first drops and then rises with the increasing of the spacing. There is a critical spacing, i.e. D0 (D0 =11.5cm in this work). The variation of the average flame height is significant as D≤D0 while indistinctive change is observed for D>D0. This result is consistent with that of Tang [17] in terms of window ejected flame height. However, Tang [17] found that the flame height decreased slowly as the spacing rises for D>D0, which is different from the present result. The reason involved is explained as follows. The presence of the curtain wall induces the chimney effect and the restriction of air entrainment. When D≤D0, the 11

chimney effect is significant and may be dominant. The chimney effect is weakened as the spacing increases. Thus the flame height drops. When D>D0, the chimney effect may be not dominant while the dominant factor is the restriction effect of the curtain wall. The restriction effect leads to the decreasing of the oxygen supply, causing the incomplete combustion and dropping of the heat release rate. Therefore the restriction effect could decrease the flame height. With the increasing of D, this restriction effect is also weakened, and thus the flame height rises. However, the heat release rate of the combustion source in Tang’s work approximately did not change, and the chimney effect was dominant for all test conditions. Therefore the flame height dropped as the spacing rose in their study.

Average flame height/cm

34

32

30

28

26

24

22 5

10

15

20

25

30

Curtain wall spacing/cm

Fig. 4 The average flame heights under different test conditions 3.2 Temperature field and heat transfer analysis Employing the infrared camera, the temperature fields under the influence of curtain wall with different spacing were obtained. The typical temperature fields at 12

266 seconds are shown in Fig. 5.

6cm

9cm

11.5cm

14.5cm

17.5cm

21cm

27cm

Fig. 5 Temperature fields under the influence of curtain wall with different spacing Each frame of the infrared video concerning on the flame zone has a maximum temperature, which is defined as Tn. The average maximum flame temperature can be calculated using the following equation.

T 

T1  T2  T3  ...  Tn n

(1)

where n is the sequence number of the infrared video frame. The calculated results are plotted in Fig. 6, which indicates that the average maximum flame temperature first rises and then drops as the spacing increases. This trend may be attributed to two competitive effects of the curtain wall. One of the effects is that the curtain wall can promote flame temperature since there exists heat feedback from the curtain wall to the XPS flame. Moreover, the curtain wall could also inhibit the heat dissipation. The other effect is that the curtain wall restricts the front air entertainment, leading to the incomplete combustion and decrease of the heat release rate. This causes the dropping 13

of the flame temperature. Both the former and the latter effects are weakened with the increase of the spacing. The later effect is dominant for the smaller spacing, whereas the former is decisive for the larger spacing. Therefore the above trend occurs.

830

Maximum temperature/oC

average value 825 820 815 810 805 800 795 790 5

10

15

20

25

30

Curtain wall spacing/cm

Fig. 6 The average maximum flame temperature as a function of curtain wall spacing Employing the two sequences of thermocouples used in the experiments, the temperature of the internal surface of the curtain wall and the ambient temperature 2 cm away from the sample surface in the vertical channel were obtained. These results are depicted in Fig. 7, which demonstrates that the temperature decreases with the rising of the spacing. There is a power function relationship between the temperature and the spacing, as shown in Eq. (2).

Ts  aDb

(2)

14

curtain wall verticle channel Fitting line Fitting line

Temperature/oC

250

Equation

200

150

y = a*x^b

Adj. R-Square

0.99252

0.98115 Value

Standard Error

curtain wall

a

3681.51583

502.64866

--

b

-1.47811

0.06969

verticle channel

a

1252.48835

147.71555

--

b

-0.87716

0.05586

100

50

5

10

15

20

25

30

Curtain wall spacing/cm

Fig. 7 Internal surface temperature of the curtain wall and ambient temperature (2 cm away from sample surface in vertical channel) as a function of spacing Eq. (2) could be used to predict the radiant heat feedback from the curtain wall to the facade. The radiant heat flux can be calculated using Eq. (3).

   Ts4  T04  Fs qrw

(3)

where ε and σ denote flame emissivity and Stefan-Boltzmann constant, respectively. Ts and T0 are the internal surface temperature of the curtain wall and the surface temperature of virgin XPS, respectively. Fs is the radiant view factor, which is obtained using the following equation [27].   (1  W 2 )(1  H 2 ) 1/ 2  W  W (1  H 2 )1/ 2 tan 1 ln    2 2 2 1/ 2 2   1W  H (1  H )   Fs     WH  H  2 1/ 2 1 1 1  H (1  W ) tan (1  W 2 )1/ 2  W tan W  H tan H   

(4)

where W = W/(D-d) and H = H/(D-d). W, H and d denote the width, height and thickness of XPS slab, respectively. 15

Incorporating Eq. (2), (3) and (4), Eq. (5) could be obtained.   (1  W 2 )(1  H 2 ) 1/2  W 2 1/2 1  W (1  H ) tan  4 4b 4 ln  2  a D  T0    1  W 2  H 2  (1  H 2 )1/2    qrw    WH H   2 1/2 1 1 1  H (1  W ) tan (1  W 2 )1/2  W tan W  H tan H   

(5)

Employing Eq. (5), the radiant heat flux from the curtain wall to XPS front surface could be calculated. With the varying of D, the predicted curve of radiant heat feedback is shown in Fig. 8. The predicted value decreases with the rising of D, and

Predicted radiant heat feedback/kWm

-2

the decrease is more remarkable for smaller D.

10

8

6

4

2

0

0

5

10

15

20

25

30

Curtain wall spacing/cm

Fig. 8 Predicted radiant heat feedback versus curtain wall spacing The total heat feedback from the curtain wall, including the radiant heat feedback and the convective heat feedback, was measured using heat flux meters. The measuring surface of the heat flux meter is approximately flush with the front surface of the XPS sample, as shown in Fig. 1(a). Moreover, the flame adheres to the façade and its thickness is smaller than XPS thickness. Thus the angle between the flame surface and the measuring surface of the heat flux meter is larger than 180°, causing 16

the flame heat transferred to the meter is little and negligible. Therefore, almost all the heat feedback measured by the heat flux meter comes from the curtain wall, which comprises radiation heat feedback and convective heat feedback. The total heat feedback to different flame spread zones (flame zone, flame front and preheating zone)

Total heat feedback from curtain wall/kW m

-2

under the influence of varied spacing is shown in Fig. 9.

flame zone flame front preheating zone Fitting line

2.0 1.8 1.6 Equation

1.4

y = y0 + A*exp(R0*x)

Adj. R-Square

0.97804

0.97807 Value

1.2

flame zone

y0

1.0 flame front

0.8 preheating zone

0.6

0.04066

-- A

4.86219

0.80884

-- R0

-0.16879

0.02149

y0

-0.00355

0.05461

-- A

2.99523

0.40256

-- R0

-0.12899

0.01948

y0

0.01902

0.01726

-- A -- R0

0.4

0.99743 Standard Error

0.03677

2.2453

0.1084

-0.12277

0.00746

0.2 0.0 5

10

15

20

25

30

Curtain wall spacing/cm

Fig. 9 The total heat feedback to different flame spread zones versus spacing Fig. 9 demonstrates that the total heat feedback decreases with the increase of the spacing. The heat flux decreases progressively from the flame zone to the preheating zone. The heat flux difference also drops with the increase of the spacing. The non-linear curve fitting was conducted to obtain the correlation between the total heat feedback and the spacing. The obtained formula is q  y0  A exp(R 0 D)

(6)

The values of y0, A and R0 are presented in Fig. 9. Since the value of y0 approaches to zero, Eq. (6) can be changed to the following equation. 17

q  A exp(R 0 D)

(7)

The total heat feedbacks to different flame spread zones are averaged. The convective heat feedback could be obtained subtracting predicted radiant heat feedback from the average value of total heat feedback. The results are plotted in Fig. 10. The ratios of the convective heat feedback to the total one are also presented in Fig. 10, which shows the ratios first rise and then drop with the increasing of the spacing. For all test conditions in this work, the convective heat feedback is lower than the radiant one, which is the dominant heat transfer mode from the curtain wall to the facade. This is different from the results of Singh and Gollner [32], who concluded that the convective heat feedback is dominant compared with radiation. The reason is they focused on the flame heat feedback of upward burning rather than

10 0.40

9

0.35

8

0.30

7

0.25

6

0.20

5 4

0.15

3

0.10

2

0.05

1

0.00

0

-0.05

-1

-0.10 5

10

15

20

25

30

Ratio of convective heat feedback to the total

Heat feedback from curtain wall/kWm

-2

the heat feedback from curtain wall during downward flame spread process.

Curtain wall spacing/cm

Fig. 10 The radiant, convective heat feedback and the ratio of convective heat feedback to the total one for different curtain wall spacing 18

3.3 Flame spread distance and flame spread rate The real-time flame spread distance, i.e., the downward moving distance of the flame front on the sides, is obtained using the image processing method and shown in Fig. 11. At the early period, the downward flame spread of XPS is basically stable under the influence of the curtain wall. However, the flame spread acceleration is observed at later period of spread process. This may be attributed to the downward flowing of the molten XPS. Fig. 12 presents the camera and infrared images of downward flowing of molten XPS, which demonstrates that the XPS flowing can promote the downward moving of the flame front. Further the flame spread is accelerated. The curve fitting results demonstrate the flame spread distance follows a

Downward flame spread distance/cm

power law with time.

60

6cm 9cm 11.5cm 14.5cm 17.5cm 21cm 27cm

50

40

Equatio y = a*x^b

30

20

10

0 0

50

6

a

Standar Adj. Rd Error Square 0.1071 0.0097 0.9977

--

b

1.1145 0.01861

9

a

0.3519 0.05258 0.9832

--

b

0.8818 0.03529

11.5

a

0.1678 0.01726 0.9953

--

b

1.0176 0.02155

14.5

a

0.1067 0.01194 0.9937

--

b

1.0980 0.02366

17.5

a

0.1964 0.02619 0.9955

--

b

1.0041 0.02736

21

a

0.1118 0.01678 0.9908

--

b

1.1340 0.03173

27

a

0.1267 0.01516 0.9953

--

b

1.0927 0.02486

spacing

100

150

200

250

Value

300

Time/s

Fig. 11 Downward flame spread distance versus time under different test conditions and the linear fitting results

19

(a) Camera image

(b) Infrared image

Fig. 12 Downward flowing of molten XPS during flame spread (D=27cm) Although various shapes of flame front may occur during XPS downward flame spread, the flame front from the front view is approximately a straight line (horizontal) for most of the time. This is different from some thermosetting materials, e.g., wood. The phenomenon may be attributed to the significant shrinkage of XPS foam exposed to heat flux. The flowing of the molten XPS could also change the flame front shape. In addition, the difference between sample width (10cm) and sample thickness (4cm) is relatively small. Therefore, the spread rate on the sides is approximately equal to that in the middle of sample front. Dividing the total flame spread distance on the side by the total time, the average flame spread rate is obtained and presented in Fig. 13.

20

Average flame spread rate/cm s

-1

0.27 0.26 0.25 0.24 0.23 0.22 0.21 0.20 0.19 0.18 0.17 0.16 5

10

15

20

25

30

Curtain wall spacing/cm

Fig. 13 Plot of average flame spread rate as a function of spacing With the increasing of the spacing, the average flame spread rate first rises and then drops. The peak value is 0.251 cm s-1 when D=11.5cm. It should be pointed out that the nails were used to fasten the XPS board during each experiment. The melting XPS may adhere to the nail, which will reduce the downward flame spread rate. Since the melting material adhering to the nail was little, the influence of nails was not obvious. Moreover, the same nails were employed under different test conditions, and thus their effects were approximately the same. Therefore, the varied trend of vf with curtail wall spacing is unchanged by the nails. The varied trend of the average flame spread rate is explained as follows. The increase of the spacing of the curtain wall induces many effects on the façade flame spread. These effects could be divided into two categories. One is the positive effect which can promote flame spread. The other is the negative effect which can decrease the flame spread rate. Firstly the positive effect is introduced. De Ris [33] 21

and Bhattacharjee et al. [34, 35] proposed a formula to predict the downward flame spread rate of thermal thick solid, as shown in Eq. (8). V f ,thick ~ Vg

g  g cg T f  Tv 2 s  s cs Tv  T 2

(8)

Eq. (8) demonstrates that the flame spread rate is positvely correlated with the flame temperature (Tf). In addition, Tf is also positvely correlated with the flame radiant heat flux, which is a key factor determining the flame spread rate. Fig. 6 indicates that the flame temperature first rises and then drops with the increase of D. Thus the flame radiant heat flux and the downward flame spread rate may have the similar varied trend. As stated above, the downward flow of molten XPS increases the downward flame spread rate. With the rising of D, the occurrence probability of molten XPS downward flowing increases while the time from ignition to downward flowing becomes shorter, leading to the rising of the downward flame spread rate. Secondly the negative effect is illustrated. The chimney effect is weakened as D rises. Thus the buoyancy induced air flow rate (Vg) drops. The downward flame spread rate may also drops according to Eq. (8). The weakening of the chimney effect may also lead to the increased spacing between the flame and the façade, causing the decreasing of the convective flame heat flux. In addition, it has been demonstrated that the total heat feedback from the curtain wall to the facade decreases with the increasing of D. The positive effect is dominant for smaller D while the negative effect is decisive for larger D. Therefore, the average flame spread rate first rises and the drops with the increasing of the spacing of the curtain wall. 22

Conclusions The influence of parallel curtain wall on the characteristics of downward flame spread over XPS slabs were investigated in this work. A series of laboratory-scale experiments were conducted and the spread processes were recorded. Further, the flame shape, flame height, temperature field, heat transfer, flame spread distance and flame spread rate were obtained. The influences of curtain wall spacing on these flame spread characteristics were analyzed and the mechanism involved is revealed. The results of the study are summarized as follows: As the spacing (D) rises, the front surface of the flame becomes more irregular. For small spacing, the flame stretching phenomenon is obvious and periodical change of the flame height is observed. The average flame height first drops and then rises with the increasing of the spacing. This may be attributed to the coupled influence of chimney effect and restriction effect induced by the curtain wall. The variation of the average flame height is significant as D≤11.5cm while indistinctive change is observed for D>11.5cm. The average maximum flame temperature first rises and then drops as D increases. Two competitive effects, i.e., heat feedback and oxygen supply restriction effect of the curtain wall, cause above phenomenon. The temperature of the internal surface of the curtain wall and the ambient temperature 2 cm away from the sample surface decreases with the rising of D. There is a power function relationship between the temperature and D. The total heat feedback from the curtain wall to the facade decreases exponentially with the rising of D. A formula is proposed to predict the radiant heat feedback from the curtain wall, which is more dominant than the 23

convective heat feedback. As D rises, the radiant heat feedback decreases, while the ratio of the convective heat feedback to the total one first rises and then drops. The downward flame spread of XPS is basically stable at early stage of spread process while flame spread acceleration is observed at later period. With the increasing of D, the average flame spread rate first rises and then drops, which is attributed to the competition of the negative effect and the positive effect of the curtain wall. The results of this work are beneficial to the fire hazard assessment of building façade installed with a parallel curtain wall, which contributes to the fire safety design of building façade. Acknowledgements This research is supported by National Natural Science Foundation of China (No.51606215), National Key Research and Development Program of China (No. 2016YFC0802907), China Postdoctoral Science Foundation (No. 2017T100421 and No. 2016M601917), Postdoctoral Science Foundation of Jiangsu Province (No. 1601005C), the open fund of Key Laboratory of Building Fire Protection Engineering and Technology of MPS (KFKT2016ZD03) and the Natural Science Foundation of Anhui Province (No. 1608085QE113). This is also a project funded by the Priority Academic Program Development of Jiangsu Higher Education Institutions (PAPD), the Fundamental Research Funds for the Central Universities (No. 2017QNA02) and the Excellent Team of China University of Mining and Technology (No. 2015ZY002).

24

References [1] R. Tu, Y. Zeng, J. Fang, Y.M. Zhang, The influence of low air pressure on horizontal flame spread over flexible polyurethane foam and correlative smoke productions, Applied Thermal Engineering, 94 (2016) 133-140. [2] Y. Wang, Q. Wang, J. Sun, L. He, K. Liew, Influence of fire location on the thermal performance of glass facades, Applied Thermal Engineering, 106 (2016) 438-442. [3] K.H. Lu, J. Wang, L.H. Hu, Vertical Temperature Profile of Fire-induced Facade Thermal Plume Ejected from a Fire Compartment Window with Two Adjacent Side Walls, Applied Thermal Engineering, 113C (2017) 70-78. [4] D. Drysdale, An introduction to fire dynamics, John Wiley & Sons, 2011. [5] J.G. Quintiere, Fundamentals of fire phenomena, John Wiley & Sons, Ltd, Chichester, 2006. [6] I.S. Wichman, Theory of opposed-flow flame spread, Prog. Energy Combust. Sci., 18 (1992) 553-593. [7] Y. Zhang, X. Huang, Q. Wang, J. Ji, J. Sun, Y. Yin, Experimental study on the characteristics of horizontal flame spread over XPS surface on plateau, Journal of hazardous materials, 189 (2011) 34-39. [8] X. Huang, J. Zhao, Y. Zhang, Y. Zhou, Q. Wang, J. Sun, Effects of altitude and sample orientation on heat transfer for flame spread over polystyrene foams, Journal of thermal analysis and calorimetry, 121 (2015) 641-650. [9] L. Jiang, H. Xiao, Y. Zhou, W. An, W. Yan, J. He, J. Sun, Theoretical and experimental study of width effects on horizontal flame spread over extruded and expanded polystyrene foam surfaces, Journal of fire sciences, 32 (2014) 193-209. [10] M. Hajdukovic, N. Knez, J. Kolsek, F. Knez, Fire Performance of External Thermal Insulation Composite System (ETICS) Facades with Expanded Polystyrene (EPS) Insulation and Thin Rendering, Fire Technology, 53 (2017) 173-209. [11] F. Tang, L.J. Li, F.Z. Mei, M.S. Dong, Thermal smoke back-layering flow length with ceiling extraction at upstream side of fire source in a longitudinal ventilated tunnel, Applied Thermal Engineering, 106 (2016) 125-130. [12] D. Yang, R. Huo, X.L. Zhang, X.Y. Zhao, Comparison of the distribution of carbon monoxide 25

concentration and temperature rise in channel fires: Reduced-scale experiments, Applied Thermal Engineering, 31 (2011) 528-536. [13] Y. Yao, X. Cheng, S. Zhang, K. Zhu, H. Zhang, L. Shi, Maximum smoke temperature beneath the ceiling in an enclosed channel with different fire locations, Applied Thermal Engineering, 111 (2017) 30-38. [14] Y.F. Wang, X.F. Sun, S. Liu, P.N. Yan, T. Qin, B. Zhang, Simulation of back-layering length in tunnel fire with vertical shafts, Applied Thermal Engineering, 109 (2016) 344-350. [15] C.G. Fan, A.Y. Li, Y. Mu, F.Y. Guo, J. Ji, Smoke movement characteristics under stack effect in a mine laneway fire, Applied Thermal Engineering, 110 (2017) 70-79. [16] F. Tamanini, A.N. Moussa, Experiments on the Turbulent Burning of Vertical Parallel Walls, Combustion Science and Technology, 23 (1980) 143-151. [17] H.Y. Wang, P. Joulain, Numerical Study of the Turbulent Burning between Vertical Parallel Walls with a Fire-Induced Flow, Combustion Science and Technology, 154 (2000) 119-161. [18] H.Y. Wang, P. Joulain, J.M. Most, Three-Dimensional Modeling and Parametric Study of Turbulent Burning along the Walls of a Vertical Rectangular Channel, Combustion Science and Technology, 109 (1995) 287-308. [19] S.R. Wasan, P. Van Hees, B. Merci, Study of pyrolysis and upward flame spread on charring materials-Part I: Experimental study, Fire and materials, 35 (2011) 209-229. [20] S.R. Wasan, P. Rauwoens, J. Vierendeels, B. Merci, Study of vertical upward flame spread on charring materials-Part II: Numerical simulations, Fire and materials, 35 (2011) 261-273. [21] J. An, Y. Jiang, R. Qiu, Y. Wang, Numerical study of polyurethane foam fire between narrow vertical parallel walls, Journal of Safety Science and Technology, 8 (2012) 5-9. [22] Y.-P. Lee, M.A. Delichatsios, Y. Ohmiya, K. Wakatsuki, A. Yanagisawa, D. Goto, Heat fluxes on opposite building wall by flames emerging from an enclosure, Proceedings of the Combustion Institute, 32 (2009) 2551-2558. [23] L.H. Hu, F. Tang, M.A. Delichatsios, Q. Wang, K.H. Lu, X.C. Zhang, Global behaviors of enclosure fire and facade flame heights in normal and reduced atmospheric pressures at two altitudes, International Journal of Heat and Mass Transfer, 56 (2013) 119-126. [24] F. Tang, Studies on Facade Flame Behavior Ejected from Opening of a Building Compartment under Different External Boundary and Pressure Conditions, in, Vol. PhD, 26

University of Science and Technology of China, Hefei, 2013. [25] F. Tang, L.H. Hu, Z.W. Qiu, X.C. Zhang, K.H. Lu, Window ejected flame height and heat flux along facade with air entrainment constraint by a sloping facing wall, Fire Safety Journal, 71 (2015) 248-256. [26] W. An, L. Jiang, J. Sun, K. Liew, Correlation analysis of sample thickness, heat flux, and cone calorimetry test data of polystyrene foam, Journal of thermal analysis and calorimetry, 119 (2015) 229-238. [27] F.P. Incropera, Fundamentals of heat and mass transfer, John Wiley & Sons, Hoboken, 2011. [28] X. Huang, Study on flame spread characteristics of the typical external wall insulation material PS under different environments, in, Vol. Ph.D., University of Science and Technology of China, Hefei, 2011. [29] L. Jiang, C.H. Miller, M.J. Gollner, J.-H. Sun, Sample width and thickness effects on horizontal flame spread over a thin PMMA surface, Proceedings of the Combustion Institute, 36 (2017) 2987-2994. [30] L. Jiang, J.-J. He, J.-H. Sun, Sample width and thickness effects on upward flame spread over PMMA surface, Journal of hazardous materials, 342 (2018) 114-120. [31] J. De Ris, Fire radiation-A Review, Proceedings of the Combustion Institute, 17 (1979) 1003-1016. [32] A.V. Singh, M.J. Gollner, A methodology for estimation of local heat fluxes in steady laminar boundary layer diffusion flames, Combustion and Flame, 162 (2015) 2214-2230. [33] J.N. De Ris, Spread of a laminar diffusion flame, in:

Symposium (International) on

Combustion, Vol. 12, 1969, pp. 241-252. [34] S. Bhattacharjee, J. West, Determination of the spread rate in opposed flow flame spread over thick solid fuels in the thermal regime, Proceedings of the Combustion Institute, 26 (1996) 1477-1485. [35] S. Bhattacharjee, M.D. King, S. Takahashi, T. Nagumo, K. Wakai, Downward flame spread over poly(methyl)methacrylate, Proceedings of the Combustion Institute, 28 (2000) 2891-2897.

27

Table Captions Table 1 Some physical properties of XPS Table 2 Experimental conditions

28

Figure Captions Fig. 1 Experimental setup (a) side view; (b) top view. 1–XPS sample; 2 – vertical facade; 3 –curtain wall; 4 –heat flow meter; 5 –K-type thermocouple sequence; 6 –infrared camera and 7 – camera. Fig. 2 Flame shapes under the influence of curtain wall with different spacing Fig. 3 Variation of the flame height during downward flame spread process (D = 6cm) Fig. 4 The average flame heights under different test conditions Fig. 5 Temperature fields under the influence of curtain wall with different spacing Fig. 6 The average maximum flame temperature as a function of curtain wall spacing Fig. 7 Internal surface temperature of the curtain wall and ambient temperature (2 cm away from sample surface in vertical channel) as a function of spacing Fig. 8 Predicted radiant heat feedback versus spacing Fig. 9 The total heat feedback to different flame spread zones versus spacing Fig. 10 The radiant, convective heat feedback and the ratio of convective heat feedback to the total one for different spacing Fig. 11 Downward flame spread distance versus time under different test conditions and the linear fitting results Fig. 12 Downward flowing of molten XPS during flame spread (D=27cm) Fig. 13 Plot of average flame spread rate as a function of spacing

29

Highlights 

Stretching and periodical change of flame height is observed for small spacing.



The average flame height first drops and then rises as the spacing increases.



Radiant heat feedback from curtain wall is predicted and proved dominant.



Total heat feedback from curtain wall follows an exponential function of spacing.



As spacing rises, the average flame spread rate first rises and then drops.

30