Expeimental and Numeical Studies on Fie Whils K. Matsuyama, N. Ishikawa 2, S. Tanaka 2, F. Tanaka, Y. Ohmiya 2, and Y. Hayashi 3 Cente fo Fie Science and Technology, Tokyo Univesity of Science, 264, Yamasaki, Noda-shi, Chiba 278-85, Japan 2 Dept. of Achitectue, Faculty of Science and Technology, Tokyo Univesity of Science, 264, Yamasaki, Noda-shi, Chiba 278-85, Japan 3 Building Reseach Institute, Tachihaa, Tsukuba-shi, Ibaaki 35-82, Japan Abstact In case that conflagation such as uban fie occued, thee is case which the swil flow, so-called fie whils sometimes geneates. The geneation of fie whils is a ae but potentially catastophic fom of fies. In ode to contol of fie whils, it is impotant to specify the geneating mechanism. In this pape, expeimental and numeical studies ae epoted on whiling flames, which modeled fie whils. A popane bune was used as a fie souce of expeiment appaatus. The distibution of tempeatue and velocity of fie whil wee measued in expeiments. Moeove, the pediction of fie whils phenomenon was caied out by using CFD code. The CFD code, which is the FDS Ve.3. developed in NIST, was used fo numeical simulation.. Intoduction * In the lage fie like an uban fie, the fie whils that is the singula phenomenon might be often geneated, and big damage be caused. The fie whils cause a theat fo the suounding because eaction against combustion is stongly pomoted, the flame height pogesses by the swil, the fie bands ae scatteed widely. Fo example, Geat Kanto Eathquake (in Japan, 923) and Humbug lage ai aid Coesponding Autho- Tel.: +8-4-724-5; Fax: +8-4-723-9873 E-mail addess: kmatsu@s.noda.tus.ac.jp (in Gemany, 943) ae typical as the damage case with the fie whils. Although the phenomenon of fie whil epoted in pevious study, the main focus is instance concening damage [, 2]. The othe hand, the cause of geneation and behavio of fie whils wee eseached expeimentally and theoetically [3-5]. Howeve, the geneation mechanism and behavio of the fie whils depend on vaious paametes included contingency such as the meteoological condition. Theefoe, it is difficult to claify quantitatively and qualitatively. Moeove, Copyight Intenational Association fo Fie Safety Science
what stategy is effective to pevent geneation is not efeed. The pupose of this study is to claify the geneation mechanism and behavio of the fie whils qualitatively and quantitatively. Fist of all, the eductively scale model expeiments wee caied out. Moeove, the numeical simulation of fie whils phenomenon was caied out by using CFD code. sceen (calcium silicate boad) c 2 s fie souce (gas bune) 2 25 sceen (heat-esistant glass) [Plan] Figue. The schematic of expeimental model (unit: mm) 5 5 5 2. Expeimental 2. Outline of Expeiment In the lage-scale fie whils geneated in Geat Kanto Eathquake and the lage ai aid in Humbug, the extenal foce of global of the Coiolis foce and the weathe, etc. is thought to be one of the impotant factos to be geneated the fie whils. Howeve, because it was used a eductive scale model appaatus in this study, these extenal foces wee disegaded. The expeiments wee caied out on the Full- Scale Fie Laboatoy at the Building Reseach Institute (in Japan). 2.2 Expeimental Pocedue The swil flame was compulsoily geneated by setting up the sceen assumed to be a building in the fie souce suoundings as shown in Figue. Because it expeimented unde calm, the swil of the flame will be geneated fo the self-induction cuent to the flame. A systematic expeiment concening the following fou paametes, which ae ) HRR (Heat Release Rate) of fie souce, 2) width of the cleaance between sceens, 3) distance between cente of fie souce and sceen, 4) height of sceen, was caied out. And, the claification of the geneation limit and the behavio wee tied. Figue 2. The expeimental model 2.3 Expeimental Model and Measuement Items () Expeimental Model As fo the expeimental model, the sceens of the calcium silicate boads and the heat-esistant glass wee set up in suoundings of the gas bune as shown in Figue. Distance between the cente of fie souce and a sceen ( s ) and width of the space between sceens ( c ) ae changed by egulaly moving the location of the sceens. Moeove, height of the sceen ( h ) is changed by egulaly moving foundation whee the souce was put up and down. The popane gas bune was used fo the souce. The size of the bune was assumed to be 2 x 2 [mm]. (2) Measuement items and method The tempeatue distibution of whiling flames, the height of whiling flames, and flame hoizontal flow velocity
wee measued. The measuement method of each measuement item is shown in the following. The location of measuement stating point is made to the cente of the bune, and made the z axis in the vetical diection, the x axis and the y axis in the hoizontal diection of the bune suface. a) Vetical tempeatue distibution of whiling flames To measue vetical tempeatue distibution of the flame, themocouple-net installing 48 themocouples is put in the locations shown in Figue 3. The net has type-k themocouples (5 [mm]-diam.). As shown in Figue 3, vetical tempeatue distibution was measued. simultaneous multipoint. The schematic of the PIV system was shown in Figue 4. The measuement by this system used the following device. -Camea: MEGAPLUS Camea Model ES. (KODAK) -Lase: Pulsed YAG Lase2D (JAPAN LASER Inc.) -Softwae: Vid PIV (OXFORD LASERS Co.) sceen (calcium silicate boad) sceen (heat-esistant glass) MEGAPLUS camea themocouple-net lase system video camea sceen (calcium silicate boad) faundation faundation gas buna lase sheet Figue 4. The schematic of the PIV system PC gas bune data logge Figue 3. The schematic of the measuement system b) Height of whiling flame The height of whiling flames was measued fom the distict put the heatesistant glass as one of the sceens by using a video camea. Moeove, the swiling duation was analyzed fom the video image. c) Hoizontal flow velocity of flame The hoizontal flow velocity of flame was measued by using the PIV (Paticle Image Velocimety) system. The PIV system is a measuing device that can measue the velocity field in the twodimensional suface fom time displacement of the paticle imaging to a PC 2.3 Expeimental Condition Expeimental conditions ae shown in Table. A squae gas bune as the fie souce is used, a width is [m]. Table. Expeimental conditions Paticulas Conditions Q H.R.R, 5, 2 [kw] distance between cente of fie souce and a sceen width of the cleaance between sc eens The height of the sceen s [m]. - 6 (each.2) c [m].5,., h [m],.,.5 As shown Figue, the distance between the cente of fie souce and a
sceen ( s ) was applied to the ange of. to 6[m] (each.2[m]), the width of the cleaance between sceens ( c ) was applied to.5,., and [m], the height of the sceen ( h ) was applied to,., and.5 [m], and HRR of fie souce ( Q ) applied to, 5, and 2 [kw]. The heat elease ate obtains [kw] fom the equation accoding to similaity law between a eductive scale model and a eal lage scale as shown in the following [6]. 2 / 5 L f f () L S Q = Q S 2.4 Expeimental Results () Height of whiling flames The height of the flame including flame tip was measued fo 6 seconds evey.3 seconds, and the aveage of the measued value was defined with the height of whiling flames. The height of whiling flames in case that the width of the cleaance between sceens ( c ) was changed was shown in Figue 5, in case that the distance between the cente of fie souce and a sceen ( s ) was changed was shown in Figue 6, in case of changing the ange of height of the sceen ( h ) was changed was shown in Figue 7. The height of whiling flames was extended as the height of the sceen ises as shown in Figue 6. The height of the flame lowes because of close to the fee buning when the distance between the cente of fie souce and a sceen expands to some degee. Moeove, the height of the flame lowes because it was located fa away flow of the swil wind, when the distance of between fie souce and sceens extends, and the width of the cleaance naows. The height of the flame ises because the entainment ai fom the uppe pat of the sceens becomes small as the sceen ises, and that has the effect denied the swil wind fom the cleaance. That is, the height of whiling flames depends on the atio of entainment ai fom the uppe pat of the sceens and the cleaance. Moeove, the appeaance of the fie whils (compaison between fee buning and whiling flames) is shown in Figue 8. It should be noted that the height of whiling flames was stetched extemely compaed with fee buning. Flame Height [m].2 <Distance between the cente of fie souce and a sceen> s: [m] s: [m] s: [m].5. Width of the cleaance between sceens (c) [m] Figue 5. Height of whiling flames in case of changing the ange of width of the space between sceens (Q :5[kW], h :.5[m], s :-[m]) Flame hight [m].4.2 <Width of the cleaance between sceens> c:.5[m] c:.[m] c: [m] 2 3 4 5 Distance between the cente of fie souce and a sceen (s) [m] Figue 6. Height of whiling flames in case of changing the ange of distance of cente of fie souce and sceen (Q :5[kW], h :.5[m], c :.5-[m])
Flame height [m] <Distance betweem the cente of fie souce and sceen> s: [m] s: [m].5 2 Height of sceens (h) [m] Figue 7. Height of whiling flames in case of changing height of the sceens ( Q :5[kW], c :.5[m], s :, [m]) axis was shown in Figue 9. H / D 2 deceases as hc / 4s inceases. The above tendency is common to all the expeimental pattens. Moeove, the height of the whiling flame depends on the heat elease ate. (2) Tempeatue of whiling flames The distibution of vetical tempeatue was measued fo two minutes evey two seconds, and the aveage of the measued value was shown in Figue and as example. As a geneal tendency, the constiction in the flame coe was geneated by the effect on the whiling motion in a cetain height. Figue 8. Compaison of obseved flame of fee buning and whiling flame -..5.4.3.2... -..5.4.3.2... -..5.4.3.2... Height [m] -25 25-25 25-375 375-5 5-625 625-75 75-875 875- (Q :[kw]) ( Q :5[kW]) ( Q :2[kW]) a) h :.5[m], s :4[m], c :.5[m] H/D [-] Q : [kw] Q : 2[kW] Q : 5[kW]. hc/4s 2 [-] Figue 9. Height of obseved whiling flame As the expeimental esult, the height of obseved whiling flame fo taking 2 hc / 4s in the x axis and H / D in the y -..5.4.3.2... -..5.4.3.2... -..5.4.3.2... Height [m] -25 25-25 25-375 375-5 5-625 625-75 75-875 875- ( h :.5[m]) ( h :.[m]) ( h :[m]) b) Q :2[kW], s :[m], c :.5[m]
-..5.4.3.2... -..5.4.3.2... -..5.4.3.2... Height [m] -25 25-25 25-375 375-5 5-625 625-75 75-875 875- ( c :.5[m]) ( c :.[m]) ( c :[m]) c) Q :2[kW], s :[m], h :.5[m] Figue. The distibution of vetical tempeatue Hight fom floo level [m].4.2 [kw] 2[kW] 5[kW] 2 4 6 8 Tempaatue above ambient (ΔT) [K] Figue. Vetical tempeatue distibution of tajectoy of flame (3) Hoizontal flow velocity of flame The measuement esults of the hoizontal flow velocity distibution at [m] using by the PIV system ae shown in Figue 2. The whiling motion can be confimed. The density of flow velocity distibution in the vicinity of the cente has thickened as the distance between the cente of fie souce and a sceen expands in case othe conditions ae the same. That is meant that the powe of the whiling motion stengthens. This tendency can be confimed fom Figue 3, too. a) s :2[m] b) s :6[m] c) s :4[m] d) s :[m] Figue 2. Flow hoizontal velocity of flame at [m] height fom floo level (Q :2[kW], h :.5[m], c :.5[m]) Avaage Flow Velocity [cm/s] 5 4 3 2 Cente of swil Cente of swil h :.[m] h : [m] 5 2 25 3 Distance between the cente of fie souce and a sceen (s) [cm] Figue 3. Flow hoizontal aveage velocity of flame ( Q :2[kW], h :.5[m], c :.5[m]) 3. Numeical Simulations The effectively of the Computational Fluid Dynamics (CFD) field model as a tool that claified the physical mechanism of whiling flames was examined. In this study, Fie Dynamics Simulato (FDS) developed by McGattan et al. in National
Institute of Standads and Technology (NIST) was used as a simulation code [7]. Outline of the CFD models (tubulence, combustion, adiation, etc.) used in FDS [8] wee shown in the following. 3. Outline of FDS code The govening equations used in FDS consist of the consevation equations of mass, momentum, enegy, chemical species, and a state equation. ρ + ρu = t (2) u ρ + ( u ) u p = ( ρ ) g + f + τ (3) + t ( ρh) + ρhu = t Dp q + k T + Dt l ρ h ρd Y l l l (4) ρyl + ρylu = ρdl Yl + ml (5) t p = ρ TR Y i / M i (6) ( ) Whee, P is the Pandtl numbe and Sc is the Schmidt numbe. In this numeical simulation, the Pandtl numbe P and the Schmidt numbe Sc wee assumed as constant value. 3.2 Conditions of Simulation As shown in Figue 4, the egion enclosed with sceens is a calculation domain. The simulation time was assumed to be 2 seconds. As fo the x and the y axially, the equal intevals lattice was used. The width of one lattice cell is 2/3x -2 [m]. In this case, one side of the fie souce is divided into 3 at equal intevals. As fo the z axially (height), it was assumed that it divided into 8 at equal intevals. The height of one lattice cell is.625x -2 [m]. In the top opening and the space of the sceen, the condition of atmospheic open (pessue constant) was set. The outside tempeatue was assumed to be 2 [ ]. The viscous stess tenso in the motion equation is given by the following equation. 2 τ = μ 2defu ( u) I (7) 3 Whee, I is a unit matix. Moeove, Lage Eddy Simulation (LES) using a standad Smagoinsky model as a Sub Gid Scale model (SGS) [9] is pefomed in FDS. Themal and chemical species diffusion impoves by mixed effect of small eddies in tubulence flow field. In an LES, themal conductivity and chemical species diffusivity ae elated to the tubulent viscosity by μlesc p kles = (8) P μ LES ρ l, LES =. (9) Sc ( D) Velocimete Themocouple Figue 4. Schematic of the calculation domain. 3.3 Compaison between esults of si mulations by FDS and expeiments () Height of whiling flames s Fie souce
In Figue 5, the height of whiling flames obtained by FDS simulations and the expeiment wee shown in case that heat elease ate (Q ) was 2 [kw]. The aveage flame height, as shown following equation (), in case of fee buning that poposed by Heskestad [] was shown by the solid line in Figue 5. H D = 3.7Q *2 / 5.2. () Whee, H is the flame height [m], D is the chaacteistic of fuel size [m], * and Q is the non-dimensional heat elease ate [-]. The height of whiling flames obtained by the expeiment doubles compaed with the esults of the calculation by equation (). On the othe hand, the height of whiling flames simulated by FDS code was about a half compaed with the esults of calculation by equation (). In the expeiment, the tendency that the height of whiling flames expands was obseved as s extended. While, in the simulation by FDS code, the tendency to which the height of whiling flames shank was obseved as s extended. It is thought that the tendency is diffeent because buning behavio in the vicinity of the fie souce suface of both is geatly diffeent. L/D [-] 5 4 3 2 Expeimental Equation () FDS Q * [-] Figue 5. Aveage height of the flame. a) FDS code b) Expeiment Figue 6. Schematic aveage height of the flame. (2) Tempeatue of whiling flames As an example, in Figue 7, the vetical tempeatue distibutions of flames obtained by FDS simulations and the expeiments wee shown in case that heat elease ate (Q ) was 5 [kw]. As well as the above esult on the height of whiling flames, on the vetical tempeatue distibution of flames, the simulation esults by using FDS code wee lowe than the expeimental esults. In the expeiment, the adiation fom the fie souce influences the themocouples. Theefoe, it is necessay to conside adiation fom the fie souce in FDS simulations in ode to obtain the esults coesponding to the expeiment. -..5.4.3.2... -. s[m].5.4.3.2... -...5.4.3.2.. -..5.4.3.2... 高さ [m] - 25 25-25 25-375 375-5 5-625 625-75 75-875 875- a) s :.[m] b) s : [m] Figue 7. Aveage tempeatue of the flame. (FDS, Exp.)
(3) Flow velocity of whiling flames The esults of FDS simulation and the expeimental esults wee compaed fo hoizontal aveage flow velocity of whiling flames in intenal and suounding. As an example, the compaison, Q was 2[kW], s wa and h was [m], was shown in Figue 8. In this case, the aveage scala quantity at hoizontal flow velocity was.4452[m/s] (FDS), and 465[m/s] (expeiment). The hoizontal aveage flow velocity of FDS simulation was faste than that of the expeiment. A compaison of expeimental esults and the FDS simulation esults was shown in Figue 9 in case that Q was 2 [kw]. As fo the expeimental esults, the tendency to fast with an incease in s was obseved while FDS simulation esults wee almost constant. 3.4 Discussions In the above, the esults FDS simulation and the expeiment wee compaed on tempeatue, height, hoizontal flow velocity of whiling flames. Howeve, thee wee diffeences in the esults of both because of the influence of the adiation to the themocouple, the diffeence of the flame behavio in the vicinity of the fie souce suface and the influences of the eddy viscosity of the eddy flow, etc. Theefoe, it seems that it is necessay to impove the computing envionment and the sub-model of FDS code to solve these poblems. The concete example is shown in the following. ) To impove the influence of the eddy viscosity by Eddy flow, the Smagoinske constant should be changed. 2) To bing the buning behavio in the vicinity of the fie souce suface, the chaacteistic of fuel size should be educed. a) FDS b) expeiment Figue 8. Compaison of aveage hoizontal flow velocity of whiling flames ( Q :2[kW], s : [m], h :.5[m], c :.5[m], h :[m]) Avaage Flow Velocity [cm/s] 5 4 3 2 Cente of swil h : [m](fds) h :.[m](fds) h :.[m](exp.) h : [m](exp.) 5 2 25 3 Distance between the cente of fie souce and a sceen (s) [cm] Figue 9. Aveage flow hoizontal velocity of whiling flames ( Q :2[kW], h :.5[m], c :.5[m], h :[m]) 5. Conclusion A systematic expeiment was caied out by using the eductive scale model in ode to solve the geneation mechanism and popeties of the fie whils qualitatively and quantitatively in this study. As a esult, it was confimed in the following. - The tendency that the height of whiling 2 flames deceases as hc / 4s inceases was shown. - As fo the tempeatue of whiling flame, the constiction (position in which the tempeatue lowes) is caused in cetain height by the swil.
- The aveage hoizontal flow velocity of whiling flames shows the tendency to incease as the distance between the cente of fie souce and a sceen ( s ) extends. In addition, the esults of simulation by using FDS code and the expeiment wee compaed fo the height, tempeatue, and flow hoizontal velocity. Howeve, the enomous discepancy was caused in both. As a cause, it is thought that thee wee the following poblems fo FDS code. - It is thought that the tendency is diffeent because buning behavio in the vicinity of the fie souce suface of both is geatly diffeent. - It is necessay to impove the influence of the eddy viscosity by Eddy flow. - It is necessay to conside adiation fom the fie souce in FDS simulations in ode to obtain the esults coesponding to the expeiment.. S. Sohma, Tonado caused by conflagation, Jounal of Japan Association fo Fie Science and Engineeing, Vol.24, No.2, pp.9-29, 974 (in Japanese) 2. Cental Weathe station, Investigation epot of Geat Kanto Eathquake, pp.-6, 926 (in Japanese) 3. Y. Hasemi:Modeling of the flamecoed fie Whil, Jounal of Achitectue and Planning, Achitectual Institute of Japan, No. 352, pp.9-24, 985 (in Japanese) 4. K. Satoh and K. T. Yang: Simulations of Swiling Fies Contolled by Channeled Self-geneated Entainment Flows,Poceedings of the 5th Intenational Symposium on Fie Safety Science, pp.2-22, 997 5. H.W.Emmons and S.J.Ying: The fie whil, Poceedings of the th Combustion Symposium (Intenational), pp. 475-488, 967 6. Y. Hayashi, T. Saga, Expeimental Study on Tempeatue Distibutions in Fie-Induced Flows, Jounal of Envionmental Engineeing, Achitectual Institute of Japan, No. 566, pp.25-32, 23 (in Japanese) 7. URL http://fie.nist.gov/fds/ 8. McGattan, K. et al., Fie Dynamics Simulato (Vesion3)- Technical Refeence Guide, NISTIR6783, 22 9. J. Smagoinsky., Ciculation Expeiments with the Pimitive Equations, I. The Basic Expeiment. Monthly Weathe Review, 963. Heskestad. G., Luminous Height of Tublent Diffusion Flames, Fie Safety Jounal, Vol. 5, pp.3-8, 983 Refeences Nomenclatue C p : specific heat [J/kgK] C S : Smagoinsky constant [-] c : width of the cleaance between sceen D : chaacteistic of fuel size [m] diffusion coefficient [m 2 /s] g : acceleation of gavity [m/s 2 ] H : height of flame [m] h : height of sceen [m], enthalpy [J/kg] h : height of measuement point [m] k : themal conductivity [W/mK] I : adiation intensity [W/m 2 st] M : molecula weight [-] m l : poduction ate of l th species pe unit volume [kg/m 3 s] P : pessue [Pa] P : Pandtl numbe [-]
Q : heat elease ate [kw] * Q : non-dimensional heat elease ate [-] R : univesal gas constant [J/kgK] S C : Schmidt numbe [-] s : distance between the cente of fie souce and a sceen [m] t : time [s] T : tempeatue [K] Y : mass faction [-] f : extenal foce vecto [N/m 3 ] q : adiative heat flux vecto [W/m 2 ] u : velocity vecto [m/s] κ : absoption coefficient [-] ρ : density [kg/m density (kg/m3)] μ : viscosity coefficient [kg/ms] Subscipt f : full scale s : eductive scale