[american institute of aeronautics and astronautics 43rd aiaa aerospace sciences meeting and exhibit...

PARTIALLY PREMIXED FLAMES AND FIRE SAFETY IN SPACE: A REVIEW

Suresh K. ~ ~ ~ a n v a l * Department of Mechanical Engineering, University of Illinois at Chicago, Chicago, Illinois 60607-7022, USA

and Ishwar K . purit

Department of Engineering Scitwce and Mechanics, Virginia Polytechnic Institute and State University, Blacksburg, Virginia 24061, USA

A fire under partial-g conditions may burn in nonpremixed or partially premixed modes. In a partially premixed flame, both fuel and/or oxidizer are partially premixed with oxidizer and/or fuel, respectively. Unwanted fires can originate in a partially-premixed mode when a pyrolyzed or evaporated fuel forms an initial mixture with the ambient air which is fuel rich. In partially premixed flames, as g decreases, the high temperature regions, particularly the nonpremixed reaction zones, are cooled due to enhanced radiation. Consequently, the spatial distribution of the high temperature zone changes in pg to become consistent with the flame topography change. Although the peak flame temperature in a pg flame is lower, the radiative heat transfer originates from a larger overall flame volume with a somewhat higher average temperature. This occurs through the smoothening of temperature gradients in pg due to a widening of the flame volume and an increase in the relative importance of diffusion. An initially 1-g partially flame becomes rounder and broader during its evolution to a steady O-g flame. The inner premixed reaction zone reaches steady state rapidly, but the outer nonpremixed flame evolves over a longer duration, since it is diffusion limited. In addition, the heat release rate reaches steady state much faster than the temperature. In the absence of radiation the peak heat release rate for the O-g flame is only one half of that for the l - g flame. Both the maximum flame temperature and heat release rate become larger in the presence of a coflow, since it increases the advective oxidizer flux, thus enhancing the global reaction rate. The heights of the inner and outer reaction zones heights increase as the jet velocity is increased, since the fuel residence time, which is determined by the reaction time, remains constant. The difference between the 1- and 0-g flame heights is more significant at higher jet velocities. Liited partially premixed flames show a strong effect of equivalence ratio on the flame liftoff height. Nitrogen has a noticeable effect on the flame litoff height. However, it is less effective as a suppressant in p-g relative to 1-g. The influence of argon is far less substantial. There is a negligible difference in the liftoff height as a function of gravity for Ar-diluted flames. Using the same levels as for NZ (which produce moderate liftoff) a flame is blown out when C02 is used. However, COz is more effective in y g than at 1-g.

I. INTRODUCTION Research on fire safety in partial-g environments is driven by the consideration of safe human travel,

especially for prolonged space missions, and the lack of a fundamental understanding and knowledge base for spacecraft fires. To assure fire safety in space vehicles, a multi-pronged approach that focuses on the onset, detection, and suppression of fires is required. A fire under partial-g conditions may burn in nonpremixed or partially premixed modes. In a nonpremixed flame, the fuel and oxidizer are separated by and transported to the reaction zone, while in a partially premixed flame, both fuel and/or oxidizer are partially premixed with oxidizer and/or fuel, respectively.

Partially premixed flames contain a rich premixed fuel-air mixture in a pocket or stream, and, for complete combustion to occur, they require the transport of oxidizer £tom an appropriately oxidizer-rich (or fuel-lean) mixture that is present in another pocket or stream. Partial oxidation reactions occur in hel-rich portions of the mixture, and any remaining unburned fuel andor intermediate species are consumed in the oxidizer-rich portions. Partial

* Professor, Dept. of Mech. and Ind. Eng., 842 West Taylor St., Chicago, IL, 60607 Professor & Department Head, Eng. Sci. & Mech., Virginia Tech, Blacksburg, VA, 24061

American Institute of Aeronautics and Astronautics

43rd AIAA Aerospace Sciences Meeting and Exhibit10 - 13 January 2005, Reno, Nevada

AIAA 2005-1440

Copyright © 2005 by the American Institute of Aeronautics and Astronautics, Inc. All rights reserved.

premixing, therefore, represents that condition when the equivalence ratio in one portion of the flowfield is greater than unity, and in another section its value is less than unity. In general, the global equivalence ratio is in the range fuel-lean to stoichiometric. Thus, PPFs can be described as hybrid flames with the characteristics of both nonpremixed and premixed flames. Consequently, using partial premixing, one can exploit the advantages of both nonpremixed and premixed flames regarding safety, emissions control, and flame stability [1,2].

A detailed understanding of the structure of partially premixed flames is important from b'oth practical and scientific considerations. Unwanted fires can originate in a partially-premixed mode when a pyrolyzed or evaporated fie1 forms an initial mixture with the ambient air which is fuel rich [3,4,5]. Partially premixed flames occur in many applications including gas-fired domestic burners, industrial furnaces, Bunsen burners, and various energy-conversion devices. Partial premixing also occurs under other circumstances, such as in lifted flames [6,7], in turbulent combustion due to local extinction and reignition processes [6], and in practical spray systems due to the presence of locally fuel vapor-rich regions (that result from the vaporization of smaller drop1e:t.s). In addition, nonuniform premixing can result in partial premixing. Partially premixed combustion may also be encountered in spaceship fires and future space applications 13,41.

Figure 1.1: Comparison between the predicted heat-release contours (left) and an experimentally obtained image (right) for a representative partially premixed (triple) flame.

- - Figure 1.1 provides another illustration of a partially premixed

flame that contains three reaction zones, i.e., a triple flame. This flame was established on a Wolfhard-Parker slot burner, by introducing a he1 rich methane-air mixture from an inner slot and a file1 lean mixture fiom two symmetric outer slots [2]. The figure presents a comparison of the volumetric heat release rates, computed using a detailed CFD (Computational Fluid Dynamics) model [I, with the: experimentally measured C2*-chemiluminescence intensities. The three reaction zones that characterize a triple flame are clearly distinguishecl. Two premixed reaction zones (one fuel-rich and the other fuel-lean) form the exterior "wings" of the flame, and a nonpremixed reaction zone is established in between the two wings. While all three reaction zones are clearly visible in the computed heat release rate contours, the lean premixed reaction zone is not quite as evident in the experimental image, since the fuel lean chemistry is not captured by C2*-chemiluminescence [8]. All three reaction zones (that are often referred to as flames) merge at a "triple point". The loci of the "triple points" form a "triple line" in a planar configuration.

Triple flames have been commonly used to examine the various fundamental combustion phenomena, such as flame propagation characteristics [9,10,11], flame response to stretch [12], flame stabilization [13,14], and the effects of Lewis number and mixture fiaction gradients on flame propagation [9,10]. These flames have been shown to be particularly useful to investigate the liftoff characteristics of laminar [13] and turbulent nonpremixed flames [IS]. The lifted nonpremixed flames often have a triple flame structure at the base, and their stabilization location is determinedl by a balance between the flame propagation speed and the local flow velocity at the base [14]. Likewise, propagating flames in both laminar [16, 171 and turbulent jets [IS] have been shown to have a triple flame structure at their base. Under certain conditions a propagating triple: flame may lose one or both of its premixed wings and degrade into an "edge flame" ~ 9 1 .

Several recent studies [20,2 1,22,23,24,25] have demonstrated that compared to premixed and nonpremixed flames, PPFs are better suited for a comprehensive validation of chemistry models. This is due to the hybrid characteristics of these flames and the existence of multiple reaction zones that interact with each other in a synergistic manner. For instance, in a triple flame, interactions between the reaction zones involve the transport of heat and radical species from the nonpremixed zone to the rich and lean premixed zones, partially oxidized fuel species from the rich premixed to nonpremixed zone, and


molecular oxygen from the lean premixed to the nonpremixed zone. Moreover, the rich premixed zone is characterized by the fuel rich chemistry involving fuel pyrolysis and partial oxidation reactions, while the nonpremixed zone is characterized by the oxidation of partially oxidized fuel species, and the lean premixed zone by relatively weak fuel consumption reactions but strong H02 formation and consumption reactions [20]. Thus a PPF provides a more effective crucible for the validation and development of detailed chemistry models, since the dominant reaction pathways are different in each reaction zone. The existence of multiple reaction zones also has important implications for the emission characteristics of PPFs. Previous investigations [26,27,28,29] have shown that by changing the level of partial premixing, one can reduce the emission of both NOx and soot to levels below those from the corresponding premixed and nonpremixed flames.

1.1 Partially Premixed Flames and Fire Safety in Space Partially premixed combustion is important in space applications due to many considerations as outlined in

the following: unwanted fires can originate in a partially premixed mode when a pyrolyzed or evaporated fuel forms an initial fuel ,rich mixture with the ambient air. Moreover, flames in such incipient fires are often diluted with fully oxidized products, such as C02 and H20, as well as with partial oxidation products, such as CO and Hz. Relatively little is known concerning such fires in partial-g environments. Previous studies have shown that the structure of PPFs can be modified significantly by changing the level of buoyancy [30, 3 11. Therefore the parametric space under which partially premixed flames occur is important to characterize. Fires in space burn and provide enhanced radiation heat transfer to other nonburning but ilammable areas, even when the fires are nonsooting. Therefore, characterizing radiation heat loss from partially premixed flames in partial-g environments is important Flame extinguishment techniques commonly involve the suppression of a flame by using a diluent. The effect of various diluents on the structure and extinction characteristics of partially premixed flames under partial-g conditions is poorly understood at present. While the suppression of premixed and nonpremixed flames under partial-g conditions has been investigated by several researchers, that of partially premixed flames has not been characterized. It is important to characterize the liftoff behavior of partially premixed under partial-g conditions, since in many cases fires do not necessarily "sit" on surfaces; they can consist of lifted flames for various reasons, such as dilution, oxidizer starvation, high velocity pyrolyzed jets, etc. In this context, it is also important to investigate the effects of various diluents on the behavior of lifted flames prior to their blowout under partial- g conditions. If fires originate in space, it is important to characterize the conditions under which they propagate back towards the source and stabilize, and those conditions that are conductive to blowoff. Moreover, enclosed fires generally contain hot diluents (e.g., products of combustion). Therefore, flame propagation and stability under these circumstances is also important.

The subject of this review paper is partially premixed flames (PPFs) and their relevance to fire safety in space. The characteristics of burner stabilized PPFs in different configurations are discussed in Section 2, where we describe the detailed flame structure and the effects of various parameters on it. Section 3 focuses on the effects of gravity on the structure and liftoff characteristics of PPFs. The effects of various diluents and fire suppressants on the liftoff and extinction behavior of these flames are also discussed in this section. Concluding rernarks and future research dealing with fire safety in space are presented in the last section.

II. Burner-Stabilized Partially Premixed Flames There have been numerous theoretical, computational, and experimental investigations on burner-stabilized

PPFs in different configurations. Both steady and unsteady flames have been investigated. The major focus of these investigations has been to examine their structure and characterize the parametric space for theu occurrence. As discussed in the preceding sect~on, the burner-stabilized configuration is relevant to many practical combustion systems. This configuration is also important fiom the perspective of fire safety in both earth and space environment, since fires initiate in a partially premixed mode, and it is important to characterize the parametric space under which PPFs occur.

Previous studies dealing with burner-stabilized PPFs have generally employed counterflow and coflow configurations. The counterflow configuration is well suited for fundamental studies as it provides a well- characterized, one-dimensional flame for detailed measurements and simulations. It also facilitates the examination of detailed chemistry models for a wide range of parameters, as it allows independent variation of the level of partial

3 4merican Institute of Aeronautics and Astronautics

premixing (i.e., equivalence ratio), strain rate, and pressure. On the other hand, the coflow configuration is a more realistic representation of flames in practical combustion systems, and well suited to investigate a broader class of combustion phenomena, including flame stabilization, propagation, liftoff, extinction, and pollutant formation. In either configuration, one can establish a double flame (i.e., containing two reaction zones) by having a fuel rich stream in contact with an oxidizer stream, or a triple flame by replacing the oxidizer with a fuel lean mixture in the oxidizer stream.

Yamoaka and Tsuji [32,33,34] reported the first series of measurements in methane-air counterflow PPFs established in the forward stagnation region of a porous cylinder immersed in a uniform air stream. They observed that by varying the stoichiometry of the rich premixed mixture, two separate reaction zones or double flames may be established. Rogg et al. [35] also employed a porous+ylinder configuration for a numerical study of a methane- air PPF using detailed chemistry. An important observation from their study pertained to the relevance of laminar partially premixed flamelets to the modeling of turbulent PPFs. More recent studies [36,37,38,39,40,4 11 considered counterflow PPFs established in the flow field of two opposing jets. Measurements and simulations of methane-air [23,24,25,39] and n-heptane-air [22,27,28] PPFs were performed to characterize the effects of various parameters on their structure. A counterflow flame is established by igniting the fuel-air mixture formed by two opposing jets, one containing a fuel rich mixture and the other containing air or oxidizer. In this configuration, the flame is observed to contain two spatially separated reaction zones, namely a rich premixed zone on the fuel side and a nonpremixed zone on the air side. The flame structure can be controlled by independently varying the fuel stream equivalence ratio +, global strain rate a,, pressure, and adding diluents in the two streams.

Figure 2.1 presents images of counterflow partially premixed flames established at strain rates a, = 50 and 100 i' and fuel stream equivalence ratios I+ = 4,6,9, and 20 [42]. Fuel is introduced from the bottom nozzle. The double flame structure that is characterized by separate premixed and nonpremixed reaction zones hecomes visually more distinct as a, decreases and I+ increases, which also increases the separation distance A between the two zones. The premixed reaction zone is established on the fuel-side of the stagnation plane at the location x, where the local axial velocity V, equals the burning velocity SLSa of the stretched flame. Since SL.= increases as 4 is reduced, the premixed flame moves away from the stagnation plane toward the fuel nozzle to satisfy the condition SL, = V,. The nonpremixed flame is established on the oxidizer side at the location x, where the fuel and oxidizer fluxes meet in stoichiometric proportion. Therefore, the value of the reaction zone separation distance A increases as the fuel-side premixing approaches stoichiometric conditions. Increasing the strain rate has an opposite effect, since for larger flow velocities the location x, is pushed toward the stagnation plane. Soot is formed in the 4 = 20 and a, = 50 i' flame for which the premixed reaction zone has a yellow-orange luminosity. This luminosity fades as 41 is decreased, and is absent for the a, = 100 s" flames.

strain rate 50 - I strain rate 100 -'

Figure 2.1: Direct images of counterflow partially premixed flames established at a, = 50 and 100 d1 and fuel stream Q = 4,6,9, and 20.

4 ~lmerican Institute of Aeronautics and Astronautics

The peak flame temperature increases with increasing premixing of the fuel stream. The peak temperatures of the partially premixed cases in Figure 2.1 are -2100 K, which are lower than the adiabatic flame temperature of 2274 K. The value of A increases as 4 decreases, i.e., with greater partial premixing. The peak C02 and Hz0 concentrations are relatively insensitive to partial premixing. The CO peak is on the premixed side and it increase as larger amounts of oxygen are supplied by decreasing the fuel side equivalence ratio. Likewise, the increasing amount of fuel-side oxygen reduces the peak Hz concentration upstream of the premixed fraction zone. The increased oxidation with decreasing 4 also reduces the peak concentrations of the C2 species. This shows the influence of partial premixing on the formation of C2-related soot precursors. It is known that partially premixed flames be less sooting than their nonpremixed or premixed counterparts [29,43].

111. EFFECT O F GRAVITY ON FLAME STRUCTURE The effects of gravity on premixed and nonpremixed flames have been extensively investigated over the

last decade. Law and Faeth [44], Kono et al. [45], and Ronney [46] have provided detailed reviews of experimental and computational studies dealing with such 1-g and microgravity (p-g) flames under different configurations. However, the corresponding literature regarding PPFs under 1- and p-g conditions is sparse. We have shown that for double flames, i.e., PPFs containing two reaction zones, the absence of gravity increases the spatial separation between the reaction zones, since diffusive transport is enhanced relative to advection as buoyant entrainment of the oxidizer is eliminated [30]. These effects increase the effective flame volume in a p-g flame compared to its 1-g counterpart. In addition, the spatial characteristics of the inner premixed region were found to be mostly unaffected by the gravitational acceleration, while the outer nonpremixed zones exhibited significant differences. Another investigation indicated that the overall flame structure is determined by interactions between the three reaction zones, which can be influenced by changes in the mixture velocity, equivalence ratio, and gravitational acceleration [2]. While the inner rich premixed reaction zone is weakly influenced by gravity, the central nonpre:mixed and outer lean premixed reaction zones exhibit significant differences at 0- and 1-g.

3.1 Global Flame Structure We first consider flames for which buoyancy and momentum effects are comparable at l g so that the

Froude number (Fr) -1. Figure .3.1.1 presents images of double and triple PPFs obtained in (1) Ig and (2) pg near the end of a microgravity drop. Visually, the pg flames assume a quasi-steady shape within -200 ms after releasing the rig; however, their temperature and radiation fields keep evolving until the end of the drop. For the double flames depicted in Figs. 3.1.l(a) and (b) the momentum ratios of the inner to the outer flows are nearly identical. For both the cases, the flame contains two reaction zones - an inner rich premixed zone and an outer nonpremixed zone. The outer nonpremixed reaction zone moves farther outward in pg for both conditions and, visually, the intensity at its tip weakens. In addition, there is a small increase in the height of the inner rich premixed zone and a slight weakening of its tip intensity in pg. The absence of buoyancy diminishes the oxidizer flux into the inner reaction zone. Consequently, the outer nonpremixed reaction zone moves laterally outward to meet its oxidizer demand.

Figure 3.1.1: Direct images of the double ((a), (b)) and triple ((c), (d)) flames a t l g and pg (a) 4h = 2.38, Q,, = 0, Vh = 21.2 c d s , Vent = 32.3, (b) gh = 1.68, gout = 0, (c) oh = 1.68, oOut = 0.36;and (d) gh = 1.68, gout = 0.5; for (b), (c) and (d): Vh = 15.9 c d s , Vout = 25.8 c d s . (RP=rich premixed, NP=nonpremixed, LP=lean premixed).

5 4merican Institute of Aeronautics and Astronautics

The heights of the normal- and microgravity double flames presented in Fig. 3.1.l(a) are higher than those of the corresponding flames in Fig. 3.1.l(b). This is due to the larger fie1 flow rate in the former case. The flame with a larger inner equivalence ratio $in exhibits a more pronounced lateral spread. Bennett et a1 [47] have shown that a larger value of $in for a lg flame produces greater oxidizer entrainment, most of which is buoyancy driven. Therefore, as a lg flame transits to pg, the loss of the buoyancy-induced flux has greater consequences for flames with greater $in.

Figure 3.1.l(c) presents images of a PPF with a triple flame structure in l g and pg. A third reaction zone appears outside of the nonpremixed flame region as a result of lean premixing of the outer stream. This lean premixed zone spreads further outward in pg and reduces the influence of gravity on the inner reaction zones by producing a fluid dynamic shielding effect. Reducing the buoyancy leads to an increased separation between the three reaction zones. Increased gravitational acceleration makes the nonpremixed reaction zone more compact and spatially closer to the inner rich premixed zone. For instance, in Fig. 3.l.i(d) the pg nonpremixed zone is approximately 20% longer and about 30% larger (at the point of greatest width) than in lg.

Figures 3.1.2(a) and (b) present a comparison of the predicted heat release rates with the experimentally- obtained images for two representative PPFs in lg and pg, one with a double flame structure and tlhe other a triple flame. The intrusive thermocouple array is visible in the triple flame image (Fig. 3.1.2(b)). Both the simulated heat release rates and the chemiluminescence in the direct images along the sides of the inner (premixed)~ flame decrease due to stretch effects. The heat release rate progressively decreases along the outer (nonpremixed) flame at downstream locations. The nonpremixed reaction zone has a weak tip, and its reaction rates are larger in upstream regions. In addition, the measured and predicted shapes and widths of the two outer (nonpremixed and lean premixed) reaction zones are in good agreement for the triple flame.

(a) (b) \

Figure 3.1.2: Experimental (direct image) and numerically simulated topologies of double and triple flames in l g and pg; (a) double flame (+h = 2.0) and (b) triple flame (+i, = 2.0,4,,, = 0.35). For both cams Vh= 30 c d s , v,,= 30 c d s .

The measured and simulated temperatures for the flames depicted in Figs. 3.1.2(a) and (b) are presented in Figs. 3.1.3(a) and (b). Both sets of data, measured and simulated, exhibit a similar qualitative behavior in l g and yg. As g decreases, the high temperature regions, particularly the nonpremixed reaction zones, are cooled due to enhanced radiation. The spatial distribution of the high temperature zone changes in pg, which is consistent with the flame topography change. In both the flames the high temperature regions lie along the nonpremixed reaction zone.

In order to examine the transition of partially premixed flames flom Ig to pg, the transient temperature data recorded during the drop are presented in Fig. 3.1.4 at a transverse location x=5 rnm for the double flame of Fig. 3.1.2(a). This location intersects the outer nonpremixed reaction zone in lg (which widens in yg). The local temperatures experience an initial transient between 1.5 s (when the rig is released fiom lg) and 2 s in accord with


the rapid initial changes in the flame topology after the rig is released. However, there is a second more gradual temperature change in pg that does not reach a steady state during the drop. The flame topologies in lg, during the - . - lg to pg transition, and under quasi-steady pg conditions are also shown in the figure.

Figure 3.1.3: Measured and predicted temperature (K) contours in l g and pg for the (a) double flame, &2.0, and (b) triple flame +h=2.0, dut=0.36. For both cases Vh=VOut=0.3m/s.

The temperature near the flame base is unchanged during the drop, since the diminishing buoyancy least influences this location. However, higher axial positions exhibit a progressively larger temperature decline during the drop due to the outward spreading of the nonpremixed reaction zone and the increased significance of thermal radiation. The thermocouple time constant is of the same order as the duration of flame spreading (-0.2 s). The temperature decline during the initial transient period occurs during 0.5 s, but a gradual temperature decrease continues thereafter until the end of the drop.

Drop Begins Drop Ends 1900

1850 - y=13mm

1800 -+ y= l8mm

3 -+ y=23mm

2 1750 -+ y=28mm

2 - y=33mm e, 3 1700 b

1650

1600

1550

Time(s)

Figure 3.1.4: Transient temperature data for a double flame at eight vertical (y) locations at 5mm away from the plane of symmetry (i.e., x= Smm). $h = 2.0, 4out = O., Vh= 30 c d s , Vout= 30 cm/s (Insert: schematic of characteristic flame scales).

7 American Institute of Aeronautics and Astronautics

3.2 Radiation Effects Further insight can be provided by using a phenomenological or scale analysis that compares the advection

time T, and the diffision time ~d for a representative flame (schematically described in the insert in Fig. 3.1.4). The characteristic fuel and air velocities v are assumed to be equal. The transit or convective time is ~,=l,lv, and the diffusion time required for the reactants to be transported laterally through the flame is ~ d = Id 'ID,, For the flames considered, the characteristic length scales are l C 4 0 mm and id-5 rnm, v-0.3 mls, and the diffisivity D -1.3x10-~ m2/s so that T, -0.1 s and ' c ~ -1 S. This implies that the rapid temperature decline during the first 0.5 s occurs due to changes in the advection as the flame transitions from lg to pg. The diffusive flux adjusts to the corresponding changes in the flame structure, and produces the second more gradual transient. The value of r d is of'the order of the duration of the experiment.

The measured peak flame temperature for the double flame decreases by -0.7% and for the triple flame by -2.1% in pg. Although the peak flame temperature in a pg flame is lower, the radiative heat transfer originates from a larger overall flame volume with a somewhat higher average temperature. We attribute this to the smoothening of temperature gradients in pg due to a widening of the flame volume and an increase in the relative importance of diffusion.

The effect of microgravity on the flame radiation can be analyzed by defining a radiation cooling time TR. It is defined as the time required for a gas volume temperature to decrease fiom an initial value Tf, by an amount, dT, where TR = Tr/(dT/dt) o Tf/(QR/(pcp)). Here, QR = (Qgen-Q,) denotes the radiation heat transfer, p density, and c, the heat capacity. The radiation (fourth) Damktihler number compares the heat generated to the heat lost through radiation, i.e., DaR = Q,n/QI,,, - T~/T,. As DaR+l, the radiation heat loss becomes more significant. The value of r, is related to the time taken by the gas to flow through the radiating volume. If the characteristic velocity is to account for the contribution of buoyant acceleration, then v = vi, + vb = v(g), where vk and ub represent the velocities at the burner exit and the corresponding buoyancy-induced component, respectively.

It is possible to show that TR - l(X&k,T - 1)" = - f&,~'), where QW denotes the heat of reaction of the fuel,

and Xf the fuel mole fraction. Therefore, DaF: - T R h C - ~ ( g ) I YX~,T')), where T is a function of Xf. The relative radiative - loss increases with a decrease in the value of v(g) or at larger . values of T, as expected. An alternate exprea;sion for DaR - ; (vin/vin + vi,/vb) ~(x~,T) - (1 + ~ r " ) ~(x~,T-':)) illustrates its - variation with the Froude number. As Fr increases (during the

transition from lg to pg), the value of DaR decreases, i.e., the relative significance of radiation increases. For a constant inlet

- stream velocity, this formulation represents the fractional N - radiative loss as a function of g.

An optically-thin radiation model has been frequently used in combustion simulations by defining a sink term due to the radiation heat loss qd 1481. Here, the Planck mean absorption coefficient K, accounts for the absorption and emission from the participating gaseous species C02, HzO, CO and CH4. Its value is obtained by using a polynomial approximation to the experimental data provideid in Ref. [49].

0 1 For a numerical solution that employs a time-accurate rfmm) *' algorithm, two different approaches can be employed to obtain

a stable PPF at O-g. One is to ignite the mixture at the Figure 3.2.1: Of the beginning of a O-g and allow it to attain a steady heat rate and temprahlre contours 'Or a state structure. The other approach is to first simulate a steady O - g PPF at the conditions of & = 2.5, Vh = 0.3 m flame at l-g, and then change the gravitational acceleration to

kt = ' 9 and 'out = O* The heat "Ieme late m and continue the time marching until a steady O-g flame a Of loo kW m-39 the is obtained. The experimental analogs of both methods have

temperature contour has a value of 600 K. been used in drop-tower experiments and their differences have been discussed by Bahadori et al. [50]. They observed that laminar jet nonpremixed flames did not reach steady state as the temperature fields were still evolving at the end of the drop period. Urban et al. [5 11 and Lin et al. [52] conducted long-duration tests in the space shuttle and observed that nonpremixed soot-containing hydrocarbon flames were almost twice as long as p-g flames observed in ground-based drop-tower facilities. Similar trends may


be expected for PPFs and it is worthwhile to examine these transient aspects in order to understand the influence of changing gravitational acceleration.

Figure 3.2.1 presents the temporal evolution for a 0-g PPF in terms of a selected value of the heat release rate (100 kw/m3) and the temperature contour (600 K) when it is simulated using the second approach, i.e. when a stable 1-g flame is subjected to a sudden 0-g condition at t = 0. The flow conditions are & = 2.5, Vin = 0.3 m s-', &,, = 0, and V,, = 0. The specified heat release rate contour has a value of 100 kw/m3, while the temperature contour has a value of 600 K. The initial 1-g flame at t = 0 becomes rounder and broader during its evolution to a steady 0-g flame. The inner premixed react~on zone reaches steady state rapidly, but the outer nonpremixed flame evolves over a longer duration. In addition, the heat release rate reaches steady state much faster than the temperature. For example, the specified heat release rate contour at 0.4 s is near that at 1.8 s and 2.2 s, but the temperature contour at t = 0.4 s appears to be still developing. Since td = z,,J, the isotherm is influenced both by diffusion and radiation.

Figure 3.2.2 presents the flame structures in terms of the heat release rate and temperature contours for 1- and 0-g flames simulated with and without the radiation model. The flow conditions are the same as for the flame in Fig. 3.2.1. All flames exhibit a double flame structure. For the 1-g flames (Figs. 3. 2.2(a) and (b)), the heat release rate contours indicate that the heights of both the inner premixed and outer nonpremixed zones slightly increase when radiation is accounted for, implying a small increase in the chemical time. Radiation also decreases the temperature in the high-temperature regions, as indicated by the intersection of the 1400 K and 1800 K isotherms with the centerline. However, the overall effect of radiation on the structure of the 1-g flame is of less significance than on the 0-g flame. As indicated by the heat release rate contours (cf. Figs. 3.2(c) and (d)), when radiation is addressed, the heights of both the inner premixed and outer nonpremixed reaction zones increase by about 2 mm (>lo%). In addition, the heat release rate intensity near the flame tip decreases, and the region occupied by the 1800 K isotherm shrinks from that without radiation. The decreases in the flame temperature and the gbbal reaction rate due to radiation produce a relatively longer and thicker flame at 0-g flame. Radiation also decreases the thermal and mass diffusivities through reduced temperatures.

(a) 1-g no Rad. (b) 1-g with Rad. (c) 0-g no Rad. (d) 0-g with Rad.

J

15 10 5 0 5 10 15 15 10 5 0 5 10 15 r (mm) r (mm) r (mm) r Imm)

Figure 3.2.2: Double flame structure in terms of the heat release rate and temperature contours for 1- and O-g flames simulated with and without radiation. The flame conditions are the same as those in Fig. 4.

Figure 3.2.3 presents temperature profiles along the centerline and along a radial segment at z = 6 mm for both the 1- and 0-g flames discussed in Fig. 3.2. The temperature profiles reach peak values downstream of the inner premixed reaction zone and thereafter exhibit a gradual decrease. The shift in the maximum temperature locations for the 0-g flame is indicative of the increase in the flame height. Without radiation, the peak temperatures are essentially identical for the 1- and 0-g flames (2028 K vs. 2020 K). However, when radiation is included, the difference between these two temperatures is significant (1973 K vs. 1760 K, respectively). The decrease in the peak temperature caused by radiation is 260 K for the 0-g flame compared to 55 K for the 1-g flame. Clearly, radiation effects on PPFs are significantly enhanced in p-g.


- [(a) r = 0 mm] -

-

Figure 3.2.3: Temperature profiles along the centerline and along a radial cut at z = 6 mm for both the 1- and O-g flames discussed in Fig. 5.

Figure 3.2.4 presents heat release rate profiles for these flames. Their peak along the centerline (Fig. 3.2.4(a)) occurs at the inner premixed flame tip, while the two peaks in the radial profiles (cf. Fig. 3.2.4 (b)) occur in the inner premixed and outer nonpremixed reaction zones, respectively. In the absence of radiation the peak heat release rate for the O-g flame is only one half of that for the 1-g flame. This difference can be attributed to two factors. First, the O-g flame is spatially broader and longer than the 1-g flame. Second, the reactivity at the premixed flame tip is weaker for the O-g flame due to a decrease in oxidizer advection in the absence of buoyancy. This figure also corroborates the results discussed in the context of Figs. 3.2.2 and 3.2.3. For the 0-g flame, the peak heat release rate when radiation is addressed is less than half of that without it, indicating that the reactivity at the premixed flame tip is significant] y weakened by radiation.

Figure 3.2.4: Heat release rate profiles along the centerline and along a radial cut at z = 6 mm for both the 1- and O-g flames discussed in Fig. 5.

80

In order to assess the validity of the optically-thin gas assumption, the maximum optical thickness is calculated along the axial (z) direction, i.e., [48]

q ( r ) = G K P ( t - ' (1)

and along the radial (r) direction by

K~ (z) = I: Kpdr . (2) Figure 3.2.5 presents the variation of K, and K~ for the flames corresponding to Figs. 3.2.2(b) and (d). The value of K, is much larger than of K, since the Planck mean coefficients are functions of the local species concentrations and


l I _ _ _ - - O-g no Rad. - O-g with Rad. .

I I(a) r = 0 mml -

- - I

' I

1-g no Rad. -

I I - - - - - l -g with Rad. -

temperatures, and because the flames are optically thicker in the axial direction than in the radial direction. The fact that K, is larger than K, is also partly due to the high concentration of C&, H20, CO, and C02 along the axial segment that is considered. The optical thickness is significantly larger for the 0-g flame than for the 1-g flame. The peak values of K, occur near the centerline and are 0.132 and 0.162 for the 1- and 0-g flames, respectively. Although these values are not insignificant, they are nevertheless much smaller than unity, and thus justify the optically thin gas assumption.

t A [(a) r = o mm J

Figure 3.2.5: Heat release rate profiles along the centerline and along a radial cut at z = 6 mm for both the 1- and 0-g flames

150

100 CI

b 5. u

50

The global effect of radiation on 1- and 0-g PPFs is summarized in Table 1, which presents the simulated values of the maximum flame temperature T,,, maximum heat release rate q-, total heat of combustion Q, total thermal radiation heat loss Qrd, radiation fraction xnd, maximum optical thickness K,,,, and inner flame height H (which is defined by the location of peak heat release rate along the centerline) for five flamer;. The simulated radiation fraction values for 1-g flames are in good agreement with those reported in literature for laminar methane- air nonpremixed flames [53]. It is logical that corresponding values for PPFs should be of the: same order. A comparison of xrd for 1- and 0-g flames again demonstrates that the effect of radiation on PPFs is significantly enhanced in the absence of gravity for which the xrad value can be as high as 50.5%., although it drops to roughly 22.9% in the presence of a coflow at 0.3 m il. The optically-thin gas model is known to overpredict the radiation heat loss in flames. Nevertheless, it provides a limiting value for the radiation loss, and can be used to compare the effects of radiation in 1- and 0-1: flames.

Table 1 Properties of double flames under different gravity level. For all the cases, 4, = 2.5, F,, = 0.3 m s-', and = 0.

. . I , .-,r , - I , , , ,

8 10

8 * 8 - l m q ~ ~ l r ~ ~ 8 ~ q v - - ~ m q L T

1 -g no Rad. - - - - - - 1 -g with Rad. - - - - - - 0-9 no Rad.

- 0-9 with Rad. -

vout Gravity Tmac %ax Q Qrad Xmd Kz,mm H (ms-') (8) (K) (W cm") (w) (W) (%I (mm)

0 1 1974 632 32.78 3 .03 9.2 0.132 10.7 0 0 1760 268 32.02 16.17 50.5 0.162 14.4

0.3 1 1989 1008 32.90 2.69 8.2 0.130 8.2 0.3 0 1954 890 32.78 7.5 1 22.9 0.149 8.6 0.6 0 1990 1264 32.66 4.02 12.3 0.135 9.7 1 .O 0 2008 1454 33.31 2.36 7.1 0.128 7.3

- I

Differences between the 1- and 0-g flames are enhanced when radiation is addressed. Figure 3.2.6 presents velocity vectors, and temperature and heat release rate contours for the 1- and 0-g flames discussed in Fig. 3.2.2. The base of the 0-g flame moves further away from the centerline and stabilizes below the burner rim. The height of the inner premixed reaction zone is significantly longer than for the 1-g flame (14.4 vs. 10.7 mm). The flame base displacement occurs due to the entrainment of air. In 1-g, entrainment produces a flow that pushes the flame closer


I I I

I I

[ ( b ) z - 6 j

-

to the centerline and also pulls the flame base radially inward toward the burner wall. In 0-g, the flame base remains away from the centerline due to the absence of the buoyant flow. The maximum flame temperature and heat release rate are significantly lower in the absence of gravity (Table 1). The flame was assumed to occupy the volume described by temperatures T I 1000 K and the volume of the 0-g flame was 3.9 times larger than of the 1-g flame (9.298 vs. 2.381 cm3).

Table 2 lists the properties of the double flame at different levels of the gravitational acceleration, i.e. 0, 0.1, 1, 2, 3, 5, and l&g. Above I-g, the flame is unsteady. Therefore, average values of T-, q,,, Qrd, xrdr and H are provided in italics. The maximum flame temperatures and local heat release rates increase as the gravitational acceleration increases, while the radiation fractions and inner flame heights decrease. The flickering frequency also increases from 14.7 Hz at 1-g to 41.4 Hz at 10-g.

Fig. 3.2.6: Velocity vector fields, temperature and heat Figure 3.2.7 presents the relationship release rate contours for the 1- and O-g flames discussed in between the inverse Froude numbers ;and the Strouhal the context of Fig. 3.2.2. numbers (St = $!INin) of the flickering flames under

various gravitational accelerations. The results of premixed flames of Durex et al. [54] and nonpremixed flames of Arai et al. [55] are also plotted for comparison. The solid line represents the correlation St oc Fr-'." provided by Hamins et al. [56] through a large compilation of normal gravity data. Our results are in good agreement with those of Durox et al. and closely follow the correlation. The nonpremixed flame results of Arai et al. also fall along the slope. This implies that flickering at enhanced gravity is influenced by the same physical factors as under normal gravity conditions [56].

Table 2 Properties of double flames at different gravity levels. For all the cases, b, = 2.5, fin = 0.3 m il, and A", = 0, Vout = 0.

Besides the radiation fraction, the radiation intensity can also be expressed by the radiation (fourth) Damkijhler number, which compares heat generated to the heat lost by radiation, i.e., Dad = QgeJ(~l,ss - z&Irc. AS the Dad-+l, the radiation heat loss becomes significant. .r;: is related to the flow time of the gas through the radiating volume. It is possible to show that w - (XfQICpT - I)-' =AXf, 2'). Therefore, D h - T ~ / T , ~ - U(g) AXf,T), where U(g) is effective velocity as a function of gravity and T is a function of Xf. The radiative heat release increases with a decrease in U(g) or increa%e of T. The Froude number is significant for the radiation importance related by the Damkijhler number. It can be shown that Dard - (1 + ~ r - ' ) AXf, T). This shows that the Damkijhler number is iversely proportional to the Froude number. As Fr increases (during the transition from 1-g to kg) , Dad decreases (i.e., the relative significance of radiation increases).


3.3 Coflow Effects The effect of coflow on the structures of 1- and 0-g PPFs with coflow velocities V,,, = 0.3 m/s, and 0.6

m/s have ben simulated. There is little difference between the flames with the 0.3 and 0.6 d s coflow velocities, and both flames are shorter and more compact. Figure 3.3.1 presents velocity vectors along with temperature and heat release rate contours for the flame of V,,, = 0.3 m/s. Other conditions pertaining to this flame are fj,, = 2.5, Vi, = 0.3 m/s, and #,, = 0. Although, the presence of a coflow makes both the 1- and 0-g flames shorter and more compact, its influence is more pronounced on the 0-g flame. The flame base is pushed closer to the centerline by the presence of the coflow and the difference in the structures of 1- and 0-g flames is less significant as the coflow velocity is increased. Table 1 presents some global properties of these flames. Both the maximum flame temperature and heat release rate become larger in the presence of a coflow, since it increases the advective oxidizer flux, thus enhancing the global reaction rate. Takahashi and Katta have reported a similar "blowing effect" for laminar methane jet nonpremixed flames [57]. The radiation fraction also decreases with increasing coflow, and the differences between the maximum flame temperatures and heat release rates for the I- and 0-g flames become less pronounced.

Fig. 3.3.1: Effect of coflow on the structures of 1- and 0-g double flames in terms of velocity vectors fields, temperature and heat release rate contours for the condition of 6 = 2.5, Vi, = 0.3 m s-', h,, = 0, and VOut = 0.3 m il. (The figure on the left has a zero coflow.)

Figure 3.3.2 illustrates the effect of varying inner jet velocities on the 1- and 0-g PPFs. The heat release rate contours, velocity vectors and isotherms are presented for three cases: (a) Vh = 0.1 mls; (b) 0.4 d s ; and (c) 0.8 mls. Other conditions for all three flames are q&, = 2.5, #out = 0, and Vou, = 0. The values of the Froude number of the 1-g flames for these three cases are 0.23,3.63, and 14.51. For all flames, the heights of the inner and outer reaction zones heights increase as the jet velocity is increased, since the fuel residence time, which is determined by the reaction time, remains constant. The difference between the 1- and 0-g flame heights is more sigrdficant at higher jet velocities. The dependence of the inner flame height on the reactant velocity is attributed to the: residence time, since the chemical reaction time essentially depends only on the equivalence ratio. The increase in the height of the outer flame is attributed to the strong synergistic interaction between the two flames and to the enhanced advection fluxes of CO and H2 from the mner reaction zone. At 0-g, the weaker air entrainment and the enhancement of radiation heat loss due to the absence of buoyancy acceleration causes the flame heights to increase further for larger inner flow rates.

Table 3 presents the properties of similar triple flames under different gravitational accelerations. Average values of T-, q-, Qd, xrd, and H are provided in italics. Similar to the double flames, as the gravitational accelerations increases, the maximum flame temperature and local heat release rate increase, and the radiation fraction and inner flame height tiecrease. The flickering frequency of the triple flames is larger than for the double flames under the same gravitat~onal acceleration. This is probably due to the fact that triple flames are more premixed-like in their structure and have a broader shear layer [58] The relationship between the inverse Froude number and the Strouhal number for these flame is also shown in Fig. 10. The data lie above the Hamins et al.


correlation but still follow the power law relationship showing that the flickering mechanisms for both triple and double flames are similar.

30 30

20 20 E E

E E - - N N

10 10

0 0

15 10 5 0 5 10 15 15 10 5 0 5 10 15 15 10 5 0 5 10 15 r (mm) r (mm) r (mm) r (mm) r (mm) r (nm)

Figure 3.3.2: Effect of inner jet velocities on the 1- and O-g PPFs. The heat release rate contours, velocity vectors and isotherms are presented for three cases (from left): (a) Vh = 0.1 mls; (b) 0.4 d s ; and (c) 0.8 d s . Other conditions are & = 2.5, hut = 0, and V,,, = 0.

Table 3 Properties of triple flames at different gravity levels. For all the cases, +&, = 2.5, Vh = 0.6 m s-', and hut = 0.35, Vout = 0.5 m s-'.

3.4 Flame Liftoff and Dilution Flame stabilization and liftoff are complex processes involving transport, partial premixing, ignition and

extinction. Peters and Williams (591 and Pitts [60]0 have provided reviews on theories of turbulent diffusion flame stabilization. Lifted flames in Iminar nonpremixed jets have been extensively investigated in order to gain a fundamental understanding of the liftoff and stabilization phenomena [61,62,63,64,65,66]. These investigations have observed that lifted flames often have a triple or tribrachial flame structure at their base. Chung and Lee [61,62] reported that for nonpremixed laminar jets, propane and n-butane flames can become lifted while methane and ethane flames blowout directly from a burner-stabilized mode. Their analysis showed that the Schmidt number Sc plays an important role in flame liftoff. Stable lifted flames are possible only for fuels for which Sc > 1 or less than 0.5. Kioni et al. [63], and Plessing et al. [67] established lifted triple flames using nitrogen diluted methane fuel and investigated the effect of strain rate. Ghosal and Vervisch [65] demonstrated analytically that a lifted laminar flame is possible for a fuel for which Sc is greater than a critical value S c , where Sc, can be less than unity. For values of Sc < Scc, they showed that a lifted flame is subcritical and can only survive in a narrow parametric region.

The lifted flames in these investigations were stabilized in the near field of a jet, with the liftoff heights being of the order of 1 cm. This allowed additional fuel-air mixing in the upstream region and thereby the formation of a triple flame structure at the flame base. These lifted flames are characterized by a flame propagation speed such that a flame is stabilized where the flame speed equals the local flow velocity along the stoichiometric mixture fraction line. The flame propagation speed however differs from the corresponding unstretched laminar speed due to the effects of curvature and flow divergence upstream of the flame base or the triple point. Echekki and Chen [68]


concluded From direct numerical simulations that both curvature and diffusion effects augment radical production that enhances the flame propagation speed. These investigations have identified the minimum flow speed located at the triple point as the premixed laminar flame speed, and distinguished it from the triple flame propagation speed, which is considered further upstream of the triple point, and is generally larger than the former.

Figure 3.4.1: Five images each of reaction rate intensity contours, and predicted heat release rate contours and velocity vector plots for 1- and p-g l i d PPFs at 4=2.00,2.25,2.50,2.75,3.00, Vh=VOot=50 cmls, and 25% N2

The behavior of lifted flames that are stabilized near the burner exit (i-e., with liftoff heights that are typically smaller than 1 cm) can be expected to be different from those established in the far field. Takahashi et al. [69] investigated the stabilization of nonpremixed flames and instead of a triple flame structure they found the existence of a reaction kernel of high reactivity. This kernel provided radicals and served as a flame stabilization point in a small premixing zone Kim et al. [70] examined liftoff characteristics with respect to fuel concentration gradients and noted that the flame liftoff height and propagation velocity can be controlled by varying the mixture concentration gradient. They found that as the concentration gradient was increased, the liftoff heights of both methane and propane triple flames first decreased and then increased, showing a minimum value at a critical concentration gradient corresponding to the maximum propagation velocity. They suggested that this critical concentration gradient represented a criterion for transition from a premixed flame to a triple flame.

Lifted PPFs show a strong effect of equivalence ratio on the flame liftoff height. Figure 3.4.1 presents the measured reaction rate intensities of N2-diluted 1- and p-g lifted PPFs for several equivalence ratios. Figure 3.4.1 also depicts the simulated counterparts of these flames in terms of the heat release rate contours, velocity vectors and stoichiometric mixture fraction contours. The maximum reaction zone intensities and heat release rates for these flames occur at the flame base. As the level of partial premixing is decreased (i.e., as 4 is increased), the flame liftoff height decreases, whereas the flame length increases. At higher 4 values, however, the liftoff height becomes nearly independent of 4. The reaction zones of both 1- and y g flames move away from the centerline as 4 increases. Both 1- and p-g lifted PPFs show that increasing equivalence ratio increases the flame width, as observed in Figure 3.4.1. The reduction in g reduces the liftoff height and pushes the flame away from the centerline for Nz-diluted flames. Both of these effects can be attributed to the absence of buoyant acceleration.

Nitrogen has a noticeable effect on the flame liftoff height. It is a desirable suppressant because it is stable, non-toxic, and easily introduced into terrestrial applications. The results show that N2 is less effective as a suppressant in p-g relative to 1-g. The influence of argon is far less substantial. There is a negligible difference in the liftoff height as a function of gravity for Ar-diluted flames. This implies that the p-g effectiveness of the suppressant can be approximated by its 1-g effectiveness. Additionally Ar has a relatively low heat capacity, which makes it a poor supressant.


Carbon Dioxide is used as a fire suppressant on the International Space Station [71]. Using the same levels as for N2 (which produce moderate liftoff) a flame is blown out when C02 is used. Figure 3.4.2 shlows that for the same base fuel and air flowrates, 15% dilution with C02 is nearly as effective as 25% dilution with nitrogen. The difference between the 1-g and p-g liftoff heights of a COz-diluted flame is much smaller than those observed with N2 dilution and are on the order of a 10% difference. Another difference observed between N2 and C02 dilution is that while the N2-diluted flames appear to decrease in height as gravity is reduced, the C02 flame height seems to increase under the same circumstances. The implication is that C02 may be more effective in p-g than at 1-g. Carbon dioxide is generally considered to be a thermally acting suppressant relying on its large heat capacity to absorb heat and inhibit the chemical reaction rates. However, C02 may act as a chemical agent to a small extent when the equivalence ratio is high enough, approaching the nonpremixed case [27].

Figure 3.4.2: Diluted lifted flames Vflout=50cmls a) +=2.5,25 % Nz, b) +=SO, 15 % COz, c) +=2.5,25% Ar.

Figure 3.4.2 also illustrates the differences between the three diluents. The cases were chosen so that they show approximately the same liftoff height with each diluent. The C02-diluted flame is lifted the farthest with the least diluent. The N2-diluted flame shows a very pronounced difference between the 1-g and p-g cases where the others show a negligible change All three diluents are thermally acting suppressants for which C102 has both the largest density and largest heat capacity indicating that it would be the most effective fire suppressant, since N2 and Ar have lower densities and heat capacities.

Figure 3.4.3: bh=2.25, Vflout=50cmls, COz diiuent a) 5 %(in), b) 10%(in), c) 15 %(out), d) 25 %(out).

For a nonpremixed cup burner flame, the blow off point is approximately 19% COZ by volume [71]. At equivalence ratios of 2.25 and 2.75 we have observed that it is not possible to stabilize a flame in 1-g above approximately 20% dilution of the fuel. A somewhat larger dilution was allowable when this diluent was introduced with the coflowing air jet. Figure 3.4.3 a) and b) show the effect of increasing the inner jet diluent concentration on both the 1-g and y g flames. The introduction of the diluent causes the flame to lift off from the burner. Very small quantities of C02 facilitate flame lifting, e.g., 5% as seen in this figure. As the quantity of diluent is increased, the


flame moves farther away from the burner. Carbon dioxide acts as a thermal sink for the energy being released from the flame, which inhibits the triple point flame speed [14] and allows the flame to stabilize farther up in the divergent jet flow.

Diluent introduction through the inner jet, as discussed above, may not mimic the actual process utilized in fire suppression. More likely, the suppressant will be introduced from the periphery of the flame. This situation is mimicked by introducing C02 through the outer coflowing jet. Figure 3.4.3 c) and d) show such an effect. The dilution rates are 15% and 25%, respectively, of the outer jet volume. The outer jet is nearly 3 times larger than the inner jet and this translates to a larger net volumetric flowrate of C02 relative to the former case. The flame moves farther downstream in p-g. The difference in liftoff height between the 1- and p-g conditions is now approximately 50% of the 1-g liftoff height. One explanation for the increased liftoff height in p-g is the effect of radiation. The excess C02 present in the flame produces relatively more radiative heat loss from the flame, which is sufficient to inhibit the flame speed enough to counteract the reduced flow velocity in p-g.

Fig. 3.4.4: 4i,=2.25, Vi,,/V,,,=50 cmls, inner and outer dilution, a) 15% N2, b) 10% C02.

A condition that has greater physical meaning is the inclusion of diluent into both the inner and outer jets. This case is shown in Figure 3.4.4. When a flame becomes lifted through dilution or excessive jet velocity it is reasonable to assume that a fire suppressant would be mixed equally with the fuel jet and its surroundings. In Figure 3.4.4 a) the flame is diluted with N2. In Figure 3.4.4 b) the flame is diluted with C02. Again, we see a similar trend to that observed for the flames that were only diluted through the inner jet. The C02-diluted flame is lifted higher than the N2-diluted flame with the smaller dilution. In both cases the p-g flame is stabilized closer to the burner than the 1-g flame.

IV. FLAME PROPAGATION Since the first publicat~on by Phillips [72] that described a triple flame in a mixing la:yer, triple flame

behavior has been extensively investigated because of the important role of these flames during flame stabilization, flame spread, and re-ignition in turbulent combustion [64]. Kioni et al. [63], and Plessing et al. [67] investigated the formation of triple flames in a nitrogen-diluted methane stream. They concluded that the existence of a fuel mass fraction gradient normal to the flow direction, along which the equivalence ratio varies from fuel rich to fuel lean, is an essential condition for the formation of a triple flame. The ability to propagate in a nonuniform mixture is an important characteristic of a triple flame. The flame can propagate with either a positive or a negative flame speed, depending upon various system parameters, such as the Lewis number, flame stretch, and heat loss [73].

Hartley and Dold [lo] and Ruetsch et al. [ l l ] have reported analytical and numerical investigations of triple flame propagation in nonuniform mixtures. They found that the flame propagation speed decreases as the mixture fraction gradient is increased. However, their results were obtained using a single-step global reaction model with idealized assumptions that ignored the effects of unequal species diffusivities. Furthermore, realistic


flames are stretched, and experience time-varying mixture fraction gradients and changes in their shape and curvature, factors that have not previously been addressed in detail. Echekki and Chen [74] numerically investigated the influence of gravity in the upward and downward propagation and structure of triple flames. However, they too considered single step chemistry and diffusionally neutral mixtures. Their investigation considered freely propagating triple flames and, hence, those results are inapplicable for flame positions in proximity to a burner.

Kioni et al. [63] observed that with increasing strain rate the complex structure of a triple flame transforms into that of a conventional nonpremixed flame. Using the concept of a transport layer, Wichman [75] demonstrated that near a burner wall the premixed flame wing of a triple flame becomes a "flame nub" due to bleat losses to the wall. Fernandez et al. [76] analyzed the structure of nonpremixed flames near the rim of an injector <and showed that the Karlovitz number plays an important role in characterizing the flame shape, and its attachment and liftoff. Im and Chen [77] investigated the effect of flow strain on triple flame propagation and showed that the precise identification of the triple point of a propagating triple flame that is subject to a large strain, field becomes ambiguous. They demonstrated that large discrepancies arise in the flame displacement speeds determined at the triple point that is variously defined as the location of the maximum curvature or by the maximum heat release.

Watson et al. [78] and Su et al. [79] have reported experimental investigations of turbulent lifted flames. They observed that lean premixed and nonpremixed reaction zones existed, but did not find a rich premixed reaction zone in their turbulent lifted flames. They hypothesized that the rich premixed flame branch was folded into the nonpremixed flame tail. Takahashi et al. [80] investigated the stabilization of nonpremixed laminar flames, and also did not report a triple flame structure at the base of these lifted flames. Instead, they proposed that ac reaction kernel of high reactivity provides the radical flux and serves as a stabilization point that is located in a small premixing zone. KO and Chung [17] investigated the propagation of laminar triple flames in nonpremixed methane jets, and examined the correlations for flame propagation speed with respect to flame curvature, stretch, and fuel mass fraction gradient. However, they focused on the far field of various axisymmetric jets and did not address the transition of a t r i~ l e flame into a nub. which is ex~ected to occur near a burner rim.

(a) Case A (b) Case B Qin et al. investigated the transient processes associated with ignition and flame propagation in methane-air mixtures [81]. It is generally assumed that the triple point or the "leading edge" of triple flames propagates along the stoichiometric contour. The authors point out that previous studies did not distinguish between the stoichiometric line defined using the equivalence ratio and an alternative definition based on the mixture fraction. While the local equi.valence ratio is determined in the context of the available fuel and oxygen, the mixture fraction is a conserved scalar that represents the fluid originating from the fuel (or air) stream. The stoichiometric equivalence ratio contour and mixture fraction contours do not coincide in the region beyond the triple point due to product formation. KO et ol. [82] experimentally r (mm) demonstrated that the equivaleince ratio is stoichiometric at the leading edge of a triple flame in

Fig. 4.1: Triple flame structure near the triple point in laminar jets. However, it can be seen from Figure 4.1 terms of heat release contours. The solid line represents that the stoichiometric equivalence ratio contour and the stoichiometric mixture fraction and the dashed line the stoichiometric mixture fraction contour do not the stoichiometric equivalence ratio. coincide in the vicinity of the triple point, although they are nearly identical upstream of the triple point. However, the simulated values of the flame speeds at these two locations are quite different, since the local equivalence ratio changes significantly in this region.

Figure 4.2 presents the variation of the local equivalence ratio at the triple points of propagating flames for the two cases. The local equivalence ratio at the triple points is smaller than unity for both cases, implying that the local effective Lewis number for these methane-air mixtures is also lower than unity. When a flame is positively stretched, as in these two cases, for Le c 1 the focusing of mass transport dominates the defocusing of heat transport. Consequently, the positive stretch rate increases the flame propagation speed by enhancing the local reaction and heat generation rates [83]. We observe from Figure 4.2 that as the flame reaches the burner rim, the local equivalence ratio decreases to an even lower value due to oxidizer entrainment.


o Case A Case 6

Qin et al. argue that it is more logical to define the triple point by the stoichiometric mixture fraction contour 5,, rather than by the unity equivalence ratio curve. The e,, contour clearly separates the flame into two separate parts and its intersection with the heat release rate contours also identifies the location where the three reaction zones merge. Moreover, the triple flame propagates along the <,, contour rather than the i$=1 contour. Therefore, as can be seen from Figure 4.1, the triple points (marked as a) are located in a fuel lean region of the methandair mixture for both cases A and B. The difference in the triple point locations influences the propagation of triple flames. In addition, as shown by Im and Chen [77], the

- . .

0 5 10 15 20 25 "leading edge" of a propagating triple flame that is (mm) subject to a large strain field is ambiguous. They

demonstrated that large discrepancies could arise in Fig. 4.2: Local equivalence ratio at the triple points of the flame displacement speeds determined at the propagating flames from the simulations. triple point that is variously defined as the location of

the maximum curvature or by the maximum heat release. The analysis by Qin et al. found that the flame leading edge or triple point, which was defined by the intersection of the stoichiometric mixture fraction line and a specific OH isocontour, propagates along the stoichiometric mixture fraction line [81].

The local equivalence ratio at the triple points is smaller than unity for both the partially premixed and nonpremixed jet modes. This has important implications in determining the flame propagation speed near the triple point, where both the equivalence ratio and Lewis number are lower than unity. When the flame approaches the burner wall, the mixture fraction gradient increases and the two premixed wings of the triple flame diminish in size and finally merge with the nonpremixed wing. Qin et al. also found that differences in the locations of the peak reaction rates can be used to illl~strate the transition from a triple flame to a burner-stabilized nonpremixed flame [81]. The spatial profile of key reactions changes from one containing two peaks to a single peak and the flame thickness simultaneously decreases, since the mixing layer thickness diminishes as the flame approaches the burner.

precise identification of the triple point or the

V. CONCLUDING REMARKS In this paper, we have presented a review of partially premixed flames (PPFs) and their relevance to fire

safety in space. Experimental and numerical results concerning the burner-stabilized and lifted PPlD under lg and pg conditions have been discussed. The pg experiments have been conducted in the 2.2-second drop tower at the NASA Glenn Research Center in a self-contained rig. The effects of partial premixing and diluents (e.g., N2, COz. and Ar) on the liftoff and blowout characteristics of these flames have also been presented. Important observations are as follows: 1 For both the burner-stabilized and lifted PPFs, the absence of gravity increases the spatial separation between

the reaction zones, since diffusive transport is enhanced relative to advection as buoyant entrainment of the oxidizer is eliminated. For PPFs containing two reaction zones, as the oxidizer flux into the outer nonpremixed reaction zone diminishes in pg, this zone moves laterally outward to meet its oxidizer demand. Similarly in a triple flame, both the nonpremixed and the outer lean premixed zone move laterally outward in pg due to the reduction in oxidizer flux. In addition, the heights of these two reaction zones increase in pg. These effects increase the effective flame volume in pg PPFs compared to their lg counterpart.

2 Both experiments and simulations indicate that during transition fiom lg to pg, the flame structure experiences a rapid change and another relatively slower transient change. The faster transient stabilizes within 200ms and is due to changes in advection caused by reduced gravity. The slower transient does not filly stabilize in the available pg time. It is attributed to changes in the diffusive fluxes in response to the change in advection.

3 The thermal radiation fiom PPFs increases significantly in pg. This is due to the increased flame volume and the reduced advective heat flux in pg. A scaling analysis is developed that explains these results at various levels of gravity. Due to radiation effects, the heights of both the inner premixed and outer nonpremixed reaction zones in the Og double flame increase, and the heat release rate intensity near the premixed reaction


zone tip decreases. When radiation effects are not included in the simulations, the peak temperatures are nearly the same for the lg and pg flames. However, with radiation the difference in these temperatures is significant. The decrease in the peak temperature due to radiation for the Og flame can be as large as five times compared to that for the 1-g flame. The simulations of lg to pg PPFs also indicate that the radiation-chemistry interactions are significantly enhanced in the absence of gravity, implying that the microgravity conditions; are well suited for a hndamental investigation of such interactions.

4 While the pg PPFs flames are steady, the corresponding lg flames are unsteady, exhibiting well-organized oscillations due to an absolute instability induced by buoyant acceleration. The flickering frequency increases with increasing gravitational acceleration. The correlation between the Strouhal number and the Froude number for both the double and triple flames follow the relation St cc Fr-0.57, which is in good ag;reement with a previous compilation of normal gravity data from a variety of flames.

5 Both the simulations and measurements indicate that with nitrogen dilution, a lifted pg PPF is ;stabilized closer to the burner compared to the lg flame, and the difference in liftoff heights becomes more pronounced as the level of partial premixing is increased (i.e., at lower 4 values). At lg, the local flow velocity increases due to buoyant acceleration, and the lg flame is stabilized at a higher axial location to achieve a balance between flame propagation speed and local flow velocity. The higher liftoff height and entrainment caused by buoyant acceleration lead to enhanced mixing for lg flames. As a result, l g lifted flames exhibit a triple: flame structure at the base, while pg flames show a transition 6om triple flame to double flame structure. In addition, the liftoff heights of both Ig and pg flames decrease with increasing f, and approach their respective nonpremixed flame limits.

In future investigations. we plan to use lifted PPFs to study different flame suppressio:n scenarios, for example, the evaluation of different fuel fire suppressants (such as COJ in extinguishing flames at different gravity levels.

ACKNOWLEDGMENTS This research was supported by the NASA Microgravity Research Division through Grant No. NCC3-688 for which Dr. Uday Hegde serves as the technical monitor.

REFERENCES

Z. Shu, S. K. Aggarwal, V. R. Katta, I. K. Puri, Combust. Flame 11 1 (1997) 276-295. R. Azzoni, S. Ratti, I. K. Puri, S. K. Aggarwal, Phys. Fluids 11 (1999) 3449-40. K. R. Sacksteder, Proc. Combust. Inst. 23 (1990) 1589. Law, C. K., and Faeth, G. M., Prog. Energy Combust. Sci. 20: 65-1 13 (1 994). Beyler, C. L., Combust. Sci. Tech. 49:287 (1984). Peters, N., Proc. Combust. Inst. 20:353-360 (1984). Rogg, B., Behrendt, F., and Warnatz, J., Proc. Combust. Inst. 2 1: 1533 (1 986). Shu, Z., Krass, B. J., Choi, C. W., Aggarwal, S. K., Katta, V. R., and Puri, I. K., Proc. Combust. Inst. 27:625- 632 (1998). Dold, J. W., Combust. Flame 76:71 (1989). Hartley, L. J., and Dold, J. W., Combust. Sci. Tech. 80,23 (1991). Ruetsch, G. R., Vervisch, L., and Lifihn, A., Phys. Fluids 7:1447 (1995). Choi, C. W., and Puri, I. K., Flame stretch effects on partially premixed flames, Combust. Flame 123:119-139 (2000). Chung, S. H., and Lee, B. J , Combust. Flame 86:62 (1991). Qin, X., Puri, I. K., and Aggarwal, S. K., Proc. Combust. Inst. 29: 1565-1572 (2002). Muniz, L., and Mungal, M. G., Combust. Flame 1 1 I : 16 (1 997). X. Qin , I. K. Puri, S. K. Aggarwal, V. R. Katta, Physics ofFluids, 16:2963-74 (2004). KO, Y. S., and Chung S. H., Combust. Flame 118:151-163 (1999). Upatnieks, A., Driscoll, J., Ceccio, S., Proc. Comb. Inst. 29:1897 (2002). Buckmaster, J., Prog in Energy Combus. Sci., 28:435-475 (2002). Azzoni, R., Ratti, S., Aggarwal, S. K., and Puri, I. K., Combust. Flame 119:23-40 (1999). Aggarwal, S. K., and Puri, I. K., MAA J. 36: 1 190 (1 998). Xue, H., and Aggarwal, S. K., AIAA J. , 40:2289-2297 (2002).


23. Xue, H., Aggarwal, S. K., Osborne, R. J., Brown, T. M., and Pitz, R. W., AIAA J. 40: 1236-1239 (2002). 24. X.L. Zhu, T. Takeno and J.P. Gore, Combust. Flame 135: 351-355 (2003). 25. Barlow, R. S., Karpetis, A. N., Frank, J.H., and Chen, J.Y., Combust. Flame 127: 2102-21 18 (2001). 26. Blevins, L. G., and Gore, J. P., Combust. Flame, 116: 546 (1999). 27. Li, S. C., and Williams, F. A., Combust. Flame 11 8: 399 (1999). 28. Xue, H., and Aggarwal, S. K., Combust. Flame 132,723-74 1 (2003). 29. Arana C., Choinski M., Sen S., and Puri, I.K., Combust. Flame 138:362-372 (2004). 30. Z. Shu, C. W. Choi, S. K. Aggarwal, V. R. Katta, and I.K. Puri, Combust. Flame 118:91-107 (1999). 3 1. J. Lock, R. Ganguly, I. K. Puri, S. K. Aggarwal, U. Hedge, Proc. Combust. Inst, 30 To appear (2004). 32. Yamaoka, I., and Tsuji, H., Combust. Inst, 5:737 (1974). 33. Yamaoka, I., and Tsuji, H. Combust. Inst, 6: 1 145 (1976). 34. Yamaoka, I., and Tsuji, H., Combust. Inst, 7:843 (1978). 35. Rogg, B., Behrendt, F., and Warnatz, J., Combust. Inst, 21:1533 (1986). 36. Hamins, A., Thridandam, )I., and Seshadari, K., Chem. Eng. Sci., 40:2027-2038 (1985). 37. Smooke,M.D.,Seshadri,K.,andPuri,I.K.,Combust. Inst,22:15550(1988). 38. Law, C. K., Li, T. X., Chung, S. H., Kim, J. S., and Zhu, D. L., Combust. Sci Technol., 64:199 (1989). 39. Tanoff, M.A., Smooke, M.D., Osborne, R.J., Brown, T.M., and Pitz, R.W., Proc. Combust. Inst. 26 (1996)

1121 - 1128. 40. Li, S. C. and Williams, F. A., Proc. Combust. Inst., 28: 103 1 : 1038 (2000). 41. Barlow, R.S., Karpetis, A. N., Frank, J.H., and Chen, J.Y., Combust. Flame 127: 2 102-2 1 18 (2.00 1). 42. Berta, P., Puri, I. K., and Aggarwal, S. K., Structure of Partially Premixed n-HeptaneIAir Counterflow Flames,

Proc. Combust. Inst. 30 (2004), to appear. 43. H.S. Hura, I. Glassman, Proc. Combust. Inst. 22:371 (1988). 44. Law, C. K., and Faeth, G. M., Prog. Energy Combust. Sci. 20: 65-1 13 (1994). 45. Kono, M., Ito, K., Niioka, T., Kadota, T., and Sato, J., Proc. Combust. Inst. 26: 1189-1 199 (1996). 46. Romey, P. D., Proc. Combust. Inst. 27: 2485-2506 (1998). 47. B. A. V. Bennett, C. S., McEnally, L. D. Pfefferle, M.D. Smooke, Combust. Flame 123522-546 (2000). 48. Siegel, R., and Howell, J. R., Thermal Radiation Heat Transfer, Hemisphere Publishing Corporation, New

York, 1981. 49. Smith, N., Gore, J., Kim, J., and Tang, Q., Radiation Models, International Workshop on Measurement and

Computation of Turbuleni Nonpremixed Flames, h~://www.ca.sandia.&td~orkshop/Sul, 2001.

50. Bahadori, M. Y., Edelman, R. B., Stocker, D. P., and Olson, S. L., AIAA Journal 28: 236-244 (1990). 51. Urban, D. L., Yuan, Z. -G , Sunderland, P. B., Linteris, G. T., Voss, J. E., Lin, K.-C., Dai, Z., Sun, K., and

Faeth, G. M., AIAA Journat'36: 13461360 (1998). 52. Lin, K.-C., Faeth, G. M., Sunderland, P. B., Urban, D. L., and Yuan, Z.-G., Combust. Flame 1 16: 415-43 1

(1 999). 53. Markstein, G. H., Proc. Combust. Inst. 20: 1055-1061 (1984). 54. Durox, D., Yuan, T., Baillot, F., and Most, J. M., Combust. Flame 102: 501-51 1 (1995). 55. Arai, M., Sato, H., and Amitgai, K., Combust. Flame 118: 293-300 (1999). 56. Hamins, A., Yang, J. C., and Kashiwagi, T., Proc. Combust. Inst. 24:1695-1702 (1992). 57. Takahashi, F., and Katta, V R., Proc. Combust. Inst. 29:2509-2518 (2002). 58. Yuan,T.,Durox,D.,andVdlermaux,E.,Exp. Fluia!~ 17:337-349(1994). 59. Peters, N., Williams, F. A., AIAA J. 21:423-429 (1983). 60. Pitts, W. M., Proc. Combust. Inst. 222309-816 (1988). 61. Chung, S. H., Lee, B. J., Combust. Flame 86 (1991) 62-72. 62. Lee, B. J., Chung, S. H., Combust. Flame 109 (1997) 163-172. 63. Kioni, P. N., Rogg, B., Bra); K. N. C., and Liiih, A., Combust. Flame 95 (1993) 276. 64. Vervisch, L., Proc. Combut. Inst. 28: 1 1-24 (2000). 65. Ghosal, S., Vervisch, L., Combust. Flame 123 (2001) 646-655. 66. Lifkin, A., in Combustion in High Speed Flows (J. Buckmaster, T. L. Jackson, and A. Kumar, eds.), Kluwer

Academic, Boston, 1994, p.46 1. 67. Plessing, T., Terhoeven, P., Peters, N., Mansour, M. S., Combust. Flame 115 (1998) 335-353.


Echekki, T., Chen, J. H., Combust. Flame 114 (1998) 231. Takahashi, F., Schmoll, W. J., and Katta, V. R., Proc. Combust. Inst. 27 (1998) 675-684. Kim, N. I., Seo, J. I., Oh, K. C., Shin, H. D., Proc. Combust. Inst. 30, to appear. Katta, V.R., Takahashi, F., Linteris, G.T., Combust. Flame 137506-522 (2004). Phillips, H., Proc. Combusr. Inst. 10:1277 (1965). Thatcher, R. W. and Dold, J. W., Combust. Theory Modell. 4:435 (2000). Echekki, T. and Chen, J.-Y., Combust. Theory Modell. 5:499 (2001). Wichman, I. S., Combust. Flame 117:384 (1999). Fernandez, E., Kurdyumov, V., and Liflhn, A., Proc. Combust. Inst. 28:2125 (2000). Im, H. G., and Chen, J. H., Combust, Flame 126:1384 (2001). Watson, K. A., Lyons, K. M., Donbar, J. M., and Carter, C. D., Combust. Flame 117:199 (1999). Su, L. K., Han, D., and Mungal, M. G., Proc. Combust. Inst. 28:327 (2000). Takahashi, F., Schmoll, W. J., and Katta, V. R., Proc. Combust. Inst. 27:675 (1998). Qin, X., Choi, C. W., Mukhopadhyay A., Puri, I. K., Aggarwal, S. K., and Katta, V. R., Combust. Theory Modell. 8: 1-22 (2004). KO, Y. S., Chung, S. H., Kim, G. S., and Kim, S. W., Combust. Flame 123:430 (2000).

83. Law, C. K., and Sung, C. J., Prog. Energy Combust. Sci. 26:459 (2000).


[american institute of aeronautics and astronautics 43rd aiaa aerospace sciences meeting and exhibit...

Documents