- Email: [email protected]
Extinguishment of hydrogen laminar diffusion flames by water vapor in a cup burner apparatus
Journal of Loss Prevention in the Process Industries 38 (2015) 260e267
Contents lists available at ScienceDirect
Journal of Loss Prevention in the Process Industries journal homepage: www.elsevier.com/locate/jlp
Extinguishment of hydrogen laminar diffusion flames by water vapor in a cup burner apparatus Ming-Hui Feng, Jun Qin* State Key Laboratory of Fire Science, University of Science and Technology of China, Hefei 230026, Anhui, China
a r t i c l e i n f o
a b s t r a c t
Article history: Received 10 November 2014 Received in revised form 20 July 2015 Accepted 11 October 2015 Available online 23 October 2015
Transient computations with full hydrogen chemistry were performed to reveal the flame structure and extinguishment process of co-flow, hydrogen diffusion flame suppressed by water vapor. As the concentration of water vapor was increased, the flame detached away from the burner brim and formed an edge flame at the flame base. Water vapor showed larger chemical inhibition effect than nitrogen when extinguishing hydrogen flame, which was attributed to its enhanced third body effect in the reaction H þ O2 þ M ¼ HO2 þ M. The minimum extinguishing concentration (MEC) of water vapor and nitrogen was predicted by Senecal formula and perfectly stirred reactor (PSR) model respectively. The MECs predicted by PSR model agree with the MECs calculated by Fluent, which shows that 1) the flame extinction is controlled by the flame base, and 2) radiation absorption is negligible. The measured MECs are in a reasonable agreement with the values calculated by Fluent, which demonstrates the accuracy of the CFD model. A simple model was used to investigate the relative importance of extinguishing mechanisms of water vapor. The results show that in a co-flow configuration the thermal cooling and chemical inhibition effect are the main extinguishing mechanisms in suppressing hydrogen diffusion cup burner flame. © 2015 Published by Elsevier Ltd.
Keywords: Hydrogen safety Diffusion flame Water vapor Nitrogen Flame suppression
1. Introduction Hydrogen is viewed as a promising clean energy at present, but it is dangerous for its wide explosion range and low-energy ignition. The hazards of hydrogen are a concern in many industries, not only where hydrogen is served as a reactant or product in the chemical industry, but also where hydrogen is an unwanted product in nuclear power plants due to the reaction between zirconium and water (Ingram and et al., 2012; Dobashi, 2014). Many researchers have investigated on the extinguishment/inhibition of hydrogen flame/explosion by some inert gases and ultra-fine water mist (Battersby and et al., 2012; Papas et al., 1994; Ng and Lee, 2008), but few efforts have been paid on the effects of water steam. The lack of interest may be attributed to the fact that the application of water vapor is restricted by its low saturated vapor pressure at room temperature. However, in the chemical or nuclear industry where water steam is used as a primary heat transfer medium, water steam is a potential alternative for ordinary
* Corresponding author. E-mail address: [email protected] (J. Qin). http://dx.doi.org/10.1016/j.jlp.2015.10.004 0950-4230/© 2015 Published by Elsevier Ltd.
suppressants. The scope of the present study is limited to the interaction of water steam and hydrogen diffusion flames in a cupburner apparatus at normal atmospheric environment (298 K and 1 atm). This arrangement is based on the following reasons besides of the hydrogen safety. The cup-burner method (Association, 2004) is a widely accepted way to measure gas agents’ minimum extinguishing concentration (MEC), which serves as a reference for the design of fire protection systems. Besides, the cup-burner flames are stable and simple enough so that the extinguishing mechanisms can be readily studied. Some simplified models have been proposed to predict the MECs of gas agents in order to harmonize data used in fire protection standards. Senecal (Senecal, 2005) proposed a phenomenological approach to predict the MECs of inert gases in terms of heat capacity and fuel properties. Zhang et al. (Zhang and Soteriou, 2011) and Liu et al. (Liu and et al., 2008) predicted the MECs of inert and chemical gas agents by using a perfectly-stirred reactor (PSR) model with full chemistry. Linteris et al. (Linteris and et al., 2012) extended the scope of the PSR model to understand the unwanted combustion enhancement by potential halon replacements. Although reasonable agreements are achieved between experiments and calculations, the validity of those models on the
M.-H. Feng, J. Qin / Journal of Loss Prevention in the Process Industries 38 (2015) 260e267
hydrogen cup-burner flame has not been testified. For laminar hydrogen premixed flame, many experimental and numerical studies have been performed to investigate the extinguishing efficiency and mechanisms of various suppressants. Qiao et al. (Qiao and et al., 2010) simulated the detailed flame structure of premixed hydrogen flame at different suppressant concentrations and found that the third-body reaction H þ O2 þ M ¼ HO2 þ M was an important chain-terminating reactions. Seiser et al. (Seiser and Seshadri, 2005) investigated the influences of water on the extinction and ignition of hydrogen/methane premixed flames and found that the addition of water vapor made the hydrogen flames easier to extinguish due to the enhanced chaperon efficiencies of water in reactions H þ O2 þ M ¼ HO2 þ M, H þ OH ¼ M ¼ H2O, and H þ H þ M ¼ H2 þ M. The extinguishing efficiency and mechanisms of various suppressants change with the flame types due to different flame structures and flame speeds. For diffusion flames, Takahashi et al. (Takahashi et al., 2007; Takahashi and Katta, 2009) studied the interaction process of co-flow diffusion flames and fire suppressants in a cup burner. Their results showed that the peak reactivity spot formed at the flame base was crucial for flame stabilization and the extinguishment occurred via a blowoff process rather than the global chemical extinction. Cong et al. simulated the interaction of water steam and methane cup burner flame with one-step chemistry (Cong and Liao, 2008). Their results supported the extinguishing process proposed by Takahashi and further demonstrated that the flame-base oscillation was the key step for flame extinction. It should be noted that hydrogen is free of carbon, and the fact that its combustion product is water indicates water steam's potential capability of curbing specific chemical reactions. Few data have been reported on the MEC of water vapor (MECH2O) and the relative importance of the extinguishing mechanisms when water steam interacts with hydrogen diffusion flame. In this study, the hydrogen co-flow diffusion flame was simulated with full chemistry by Fluent and its interaction with water steam was investigated. A PSR model, as well as the Senecal's formula, was used to predict the MECs for hydrogen cup-burner flame. Their results were compared with the Fluent's predicted values and the measured values to confirm the validity of the simplified models and to interpret the influences of different extinguishing mechanisms on MECs. Finally, the relative importance of the extinguishing mechanisms was investigated with an energy balance equation. 2. Experiment and CFD modeling 2.1. Physical model setup The sketch of the apparatus is shown in Fig. 1(a). The apparatus used here is similar to that in (Cong and Liao, 2008). The cup burner has a cylindrical stainless steel cup and a cylindrical quartz chimney. The bottle of the burner is connected to the TSI atomizer (Model 9306A) through a diffuser. The fuel and air flow rate are measured by a calibrated mass flow meter and vortex flow meter respectively. Their uncertainty is 1% of the indicated flow. For the case of water vapor, the air is preheated before entering the mist chamber. As the air enters the mist chamber, it mixes with ultrafine water mist and passes through an electrically heated fine screen upstream the exit of the cup. The temperature of the fine screen is carefully adjusted to ensure a total evaporation of ultrafine water mist and maintain the temperature of the inlet oxidizer at z373 K. The mass flow rate of water vapor equals to that of water mist during the experiments so its value is determined by collecting the water mist at the oxidizer inlet during a given period. The numerical parameters were set according to the experimental conditions. The stainless steel cup burner had an outer
261
diameter of 31 mm and length of 25 mm while the quartz chimney had an inner diameter of 90 mm and length of 200 mm. The temperature of the quartz wall and burner wall was set at 298 K except the 600 K burner wall within 1 mm away from the top burner brim (Ananth and Mowrey, 2008). No slip boundary condition was applied to the wall. The normal thermal emittance of the hot burner wall and cool wall was 0.24 and 0.22 respectively (Bergman et al., 2011) and the hemisphere integrated radiant emittance was 0.93 for the chimney wall. The initial fuel temperature was 298 K and the inlet velocity of the fuel was fixed at 6.5 cm/s to achieve a laminar flame with a visible flame height of about 75 mme85 mm (Association, 2004). Three types of oxidizer stream were considered in the simulations, they are: dry air, dry air with nitrogen, dry air with water vapor. The temperature of the inlet oxidizer stream was 298 K and its velocity was fixed at 13.6 cm/s at 1 atm. The mass fraction of the extinguishers in the oxidizer stream was gradually increased to its MEC where the flame sustained combustion.
2.2. The numerical method The Da number, chemical reaction time tr and flow time tf (Quintiere, 2006) were estimated from the following equations:
Da ¼
tf ; tr
(1)
tr ¼ rCp T∞ ½E=RT∞ Dhc AeE=ðRT∞ Þ ;
(2)
tf ¼ d=v:
(3)
In these equations, r and Cp are the gas density and specific heat capacity near the flame sheet, T∞ and v is the temperature and velocity of the inlet air, E and A is the activation energy and preexponential factor for the hydrogen respectively, Dhc is the idea heat of combustion and d is the equivalent distance for the air transporting to the flame sheet. The Da number is nearly 1 at the flame base under the present situations. This means that for the hydrogen cup-burner flame the chemical dynamic effect is at least equivalent to the transport ef Conaire and fect. So a detailed reaction mechanism of hydrogen (O et al., 2004) was incorporated into the laminar finite rate model. The gas phase combustion was solved by the laminar, transient, Navier-Sokes and energy equations with a time step of 0.1 s. The mass diffusion coefficient and thermal diffusion coefficient were calculated by the Chapman-Enskog and kinetic theory respectively. Fluent's DO model was used to account for the radiation loss of the flame. Radiation of the gaseous specie H2O was incorporated into the model by a polynomial fits and its absorption coefficient was taken from (Peters and Rogg, 1993). The radiation equation was calculated after solving the energy equation for 10 times at each time step. Fluent solves the conversation equations in finite volume forms. The gradients were computed using Least Squares Cell-Based Gradient method. The diffusion terms in governing equations were discretized by the secondeorder accurate central-differenced method. The convection terms in momentum equations were discretized using the Fluent's QUICK Scheme and those of the species and energy equations were discretized using a second-order upwind scheme. The coupled pressureevelocity equations were solved using Fluent's PISO method before solving energy and species conservation equations. The residuals for the gas phase energy equation were reduced to 104% for convergence while those for other equations were reduced to 102% for convergence.
262
M.-H. Feng, J. Qin / Journal of Loss Prevention in the Process Industries 38 (2015) 260e267
Fig. 1. Models for the interaction of water steam and cup-burner flame: (a) a sketch of cup-burner apparatus; (b) a sketch of the PSR model for the prediction of MEC; (c) a simplified model for the interpretation of extinguishing mechanisms.
2.3. The mesh grid size and sensitivity analysis Fine mesh size helps to accurately capture the flame dynamics but is time-consuming for computation. Since the flame thickness turned wider when extinguishers were added, the sensitivity analysis was restricted to the case where the oxidizer stream was dry air. The unburned fuel-air mixture at the uninhibited cupburner flame base was nearly stoichiometric (Takahashi et al., 2011), so the equation for calculating the thickness of laminar premixed flame was used:
Tb Tig dR z Tig Tu
!
a : Su
(4)
In this equation, dR is the premixed flame thickness, Tb is the adiabatic flame temperature, Tig is the auto-ignition temperature, Tu is the unburned gas temperature, a is the thermal diffusion coefficient and Su is the laminar burning velocity of the flat flame. The flame thickness was z 0.1 mm at the flame base. Consequently, the sensitivity analysis were performed on three structured meshes with minimum grid size of 0.125 mm by 0.125 mm, 0.0625 mm by 0.0625 mm, and 0.03125 mm by 0.03125 mm respectively. The maximum grid size of the three meshes was 0.25 mm by 0.25 mm. The preliminary simulations on the flow field of the non-reacting flow showed that the boundary layer was insensitive to the mesh grid when the first row height of the meshes near the walls was reduced to less than 0.3 mm. So the maximum grid size in the simulations would not influence the
results near the walls. The mesh independent issues were analyzed by examining the time-average temperatures distribution along the axis and the transient species mole fraction distribution along a horizontal line 3 mm away from the burner brim. The calculated temperature distribution and species mole fraction distribution of the cases are shown in Fig. 2. The results of the case with a minimum grid size of 0.03125 mm by 0.03125 mm are not shown in Fig. 2 because its results are nearly identical to the results of the case with a minimum grid size of 0.0625 mm by 0.0625 mm. The temperature distribution in Fig. 2 shows that the time-average flame height is nearly 75 mm, which meets the requirement for the cup burner test (Association, 2004). The temperature and specie distribution indicate that the minimum grid of 0.0625 mm by 0.0625 mm at the flame base is appropriate for capturing the flame dynamic. So this grid size was used in the rest simulations.
3. Simplified models for MEC prediction PSR is an ideal reactor where perfect mixing is achieved inside the control volume. It is extremely useful to study flame stabilization. Fig. 1 (b) shows the schematic diagram of the adiabatic PSR model. The governing equations of mass, species, and energy are showed below:
m_ out m_ in ¼ 0;
(5)
M.-H. Feng, J. Qin / Journal of Loss Prevention in the Process Industries 38 (2015) 260e267
263
Fig. 2. Temperature distribution along the axis and mole fraction distribution along a horizontal line 3 mm away from the burner brim.
$V; m_ in $ Yi;out Yi;in ¼ m;;; i
(6)
X Yi;out $hi;out Yi;in $hi;in ¼ 0:
(7)
i
In these equations, m_ out and m_ in are the inlet and outlet mass flow rates, Yi,out and Yi,in are the outlet and inlet mass fractions of ith specie, m;;; is the production rates of ith specie, V is reactor volume, i hi,out and hi,in are the outlet and inlet enthalpies of ith specie. m;;; could be solved from Eq. (8)e(10): i
m;;; ¼ w_ i $MWi ; i w_ i ¼
L X
(8)
qj $ v;;ij v;ij ;
(9)
flame base was comparable to the stoichiometric laminar flame Conaire and et al., 2004). Soot and radiation are not speed (O considered in this model because they are negligible for the nearlimit hydrogen flame. We then assume that the extinguishing mechanisms are similar for nitrogen and water stream, so the critical residence time of N2, at which point the flame is unable to sustain, is used as the critical residence time for water steam. The PSR model can be solved by CHEMKIN with the kinetic mechanism Conaire (O Conaire and et al., 2004) and thermochemical of O properties of Kee et al. (Kee and et al., 1996). The MEC of water vapor could be achieved by adjusting its concentration in the oxidizer stream to a critical value where the temperature suddenly drops. Senecal's approach is based on the assumption that heat capacity of the inert gas agent is the primary extinguishing mechanism. The detailed method is available in (Senecal, 2005).
j¼1
4. Results and discussion
qj ¼ kfj $
N Y
v;ij
½xi krj $
i¼1
N Y
½xi
v;;ij
(10)
In these equations, MWi is the molar weight of ith specie, v;ij and v;;ij are forward and backward stoichiometric coefficients of ith specie in jth reaction, kfj and krj are forward and backward rate constants of jth reaction, [xi] is the concentration of ith specie. m_ out can be expressed as Eq. (11),
m_ out ¼ r$V=t;
4.1. Extinguishment limits
i¼1
(11)
where r is the average density of the mixture and t is the residence time. hi,out and hi,in can be correlated with the temperature in the reactor. For a given residence time and an appropriate inlet condition, the temperature in the reactor can be solved. To apply the PSR for MEC prediction, we assume that the fuel and air are stoichiometrically premixed and reacts instantaneously at the flame base. This assumption is plausible for the cup-burner apparatus since researches showed that the unburned mixture velocity at the
Different extinguishing mechanisms of suppressants have a direct impact on MECs. Nitrogen and water vapor are selected as the extinguishers in this study. The extinguishing mechanisms in real applications and those adopted in the three models are summarized in Table 1 based on the previous studies (Senecal, 2005; Conaire and et al., 2004; Ni and Takahashi and Katta, 2009; O Chow, 2011). The blow-off conditions for various hydrogen/air/N2/ H2O mixtures were first determined by Fluent. The time-dependent calculation was run for separate cases in which the volume fraction of the extinguishers in the oxidizer stream was stepwise increased at 1% interval. As the volume fraction of the extinguishers reached to the critical value (defined as MEC), the flame detached away from the burner, drifted downstream and exited the computational domain. The calculated MECs of nitrogen (MECN2) and water vapor (MECH2O) are shown in Table 2. When the temperature of the inlet oxidizer stream (Toxidzer) was fixed at 298 K, the calculated MECH2O could not be reached in real life. In the current study, this effect was
264
M.-H. Feng, J. Qin / Journal of Loss Prevention in the Process Industries 38 (2015) 260e267
Table 1 Extinguishing mechanisms of N2 and water vapor. Extinguishment mechanism
H2O (vapor)
N2
Oxygen dilution Increase of sensible enthalpy Chemical kinetic effects Radiation attenuation by gas species
Real
Senecal
PSR
Fluent
Real
√ √
√
√ √
√ √
√ √ √ √
Table 2 A comparison of MECN2 and MECH2O. Model
Oxidizer temperature
N2
H2O
Fluent Senecal formula PSR model Fluent Experiment
298 298 298 373 373
36 e e 38 37.7
32 31.0 32.1 37 35.9
K K K K K
not considered so that the extinguishing efficiency of N2 and water vapor could be compared and the different mechanisms could be interpreted without the influence of oxidizer temperature. Table 2 also shows the measured MECs by experiments and the calculated MECs by the simplified models. It should be noted that the MECH2O was measured when Toxidzer was maintained 373 K in the experiments. So additional cases where Toxidzer was fixed at 373 K were calculated by Fluent to testify the accuracy of the model. The MECH2O predicted by the Senecal formula and the adiabatic PSR model was calculated based on the MECN2 calculated by Fluent in order to have some basic knowledge about the influences of different mechanisms on the MEC. When Toxidzer is 373 K, the calculated MECN2 by Fluent agrees well with its experimental counterpart. This indicates that the CFD model can accurately predict the MEC for the hydrogen cup burner flame. Besides, the increase of Toxidzer from 298 K to 373 K slightly increases the calculated MECN2 from 36% to 38%. The extinguishing mechanisms of N2 include thermal cooling and dilution. The increase of Toxidzer brings additional enthalpies to the mixture and compromises the cooling effect of N2. The predicted MECH2O by Fluent and PSR model are in reasonable agreement if taking account for the 1% increment in Fluent's calculation. This agreement supports the assumption in the PSR model: 1) the flame base is the key factor for the flame stabilization and 2) the radiation attenuation of water vapor is negligible for hydrogen flame. Table 2 also shows that the predicted MECH2O by Senecal's formula is 1.1% lower than the PSR model. This fact indirectly demonstrates that chemical kinetic effect may be important for the suppression of hydrogen cup-burner flame by water vapor since only increase of sensible enthalpy effect is included in the Senecal's formula. The increase of Toxidzer increases the calculated MECH2O by Fluent from 32% to 37%. The reason for this increase is the same with that for N2. For the case with Toxidzer¼373 K, the measured MECH2O is z 3% lower than the calculated value by Fluent. It is assumed that the deviation mainly comes from two aspects. Firstly, a portion of ultrafine water mist in the mixing chamber is heated by the preheated air and changes into water vapor. This portion may not be collected at the oxidizer inlet. Secondly, the edge flame at the flame base is susceptible to the turbulence at the oxidizer inlet. The introduction of water mist upstream the oxidizer inlet increases the turbulence at the oxidizer inlet and subsequently reduces the measured MECH2O. 4.2. Flame structure Fig. 3 shows the temperature contours and streamlines of the flame with saturated water vapor (298 K) at two instants separated
Senecal
PSR
Fluent
√
√ √ √
√ √ √ √
by 0.05 s. The flame was attached to the burner brim and inclined inwardly due to the large velocity gap between the fuel and oxidizer stream as showed in the temperature contour. The flow in the tube was affected by the flame. The generation and shredding of the vortex near the flame tip caused the cutting off of the flame tip as showed in the streamlines in Fig. 3. Two vortices existed at the bulk air side at the left half of Fig. 3. They merged into one and a new vortex was formed at the right half part. The process repeated itself at a fixed frequency corresponding to the flickering frequency of the flame. Five temperature monitors were set to measure the instantaneous temperature of the simulated flame along the axis in the continuous flame zone and fire plume. We transformed the temperature-time sequence into the magnitude-frequency sequence by a Fast Fourier Transformation. The obtained flickering frequency was nearly 10.94 Hz, which is slightly higher than 9 Hz estimated from Pagni’ s empirical correlation for pool fire (Chen and et al., 2008). Fig. 4 shows the calculated structure of the near-limit hydrogen flame suppressed by water vapor. The right part of Fig. 4 shows that the oxidizer was entrained into the flame base and its velocity was substantially increased due to the buoyancy-driven flow. As the concentration of H2O in the oxidizer side increased, the flame base gradually detached away from the burner brim. As showed in the left part of Fig. 4, the gap between the flame base and burner brim left a path for the oxidizer to penetrate the flame. This allows the fuel and oxidizer to be partially premixed before entering the flame base. The concentration gradient of the partially premixed mixture resulted in the formation of edge flame at the flame base. However, the edge flame did not develop into edge flame because the fuelsided and oxidizer-sided branch merged to the trailing diffusion flame. This can be explained by the finding that the formation of triple flame is fuel-dependent: acetylene and ethylene form the triple flame structure, but alkanes fails it (Takahashi and Katta, 2005). For the chemical reactions, Seiser et al. (Seiser and Seshadri, 2005) found that for hydrogen flames, the enhanced chaperon efficiencies of water in reactions H þ O2 þ M ¼ HO2 þ M, H þ OH þ M ¼ H2O þ M, and H þ H þ M ¼ H2 þ M made flames easier to extinguish. In Fig. 5, we tracked the reaction rates of the above three reactions as well as two chain-branching reactions at different mass fraction of water vapor. The reaction kernel was roughly determined by the position where the chain-branching reactions reached its maximum values. The chain-branching reactions were significantly inhibited as the mass load of water vapor was increased and the flame was nearly extinguished when the chain-branching reaction rates were below 1 Kg-mol/m3sec. In the current reaction mechanism, the chaperon efficiencies of water vapor are 12, 12 and 14 respectively, which are much more than other species. As the concentration of water vapor increased, the reaction H þ H þ M ¼ H2 þ M was enhanced while the other two were nearly insensitive to the mass fraction of water vapor. This meant that for this configuration the reaction H þ H þ M ¼ H2 þ M was crucial for flame suppression. When water vapor reached its critical value, the reactions were completely suppressed and the flame was eventually blown off.
M.-H. Feng, J. Qin / Journal of Loss Prevention in the Process Industries 38 (2015) 260e267
265
Fig. 3. Streamlines and temperature contours at two instants for the cases with saturated water vapor at the oxidizer inlet: left part at 20.40 s, right part at 20.45 s.
Fig. 4. Calculated structure of the near-limit hydrogen cup-burner flame for the cases with the inlet oxidizer diluted by 31% H2O. Left half: contours of mole fraction for H2O and O2. Right half: contours of temperature and mole fraction for H.
266
M.-H. Feng, J. Qin / Journal of Loss Prevention in the Process Industries 38 (2015) 260e267
strain rate of the edge flame and bring more heat away from the flame base so that the temperature at flame base was easier to reach the critical temperature. Besides, the calculated MECN2 and MECH2O slightly increased as the oxidizer temperature was increased from 298 K to 373 K because of the additional heat brought by the oxidizer. 4.4. Flame extinguishing mechanisms
Fig. 5. The variation of the reaction rate inside the reaction kernel with the mass fraction of water vapor at the inlet.
4.3. Influences of oxidizer temperature and velocity on flame extinction In real applications, the mass fraction of water vapor is restricted by its saturated vapor pressure at a given temperature. Besides, the release of gas agent will affect the local strain rate, which is crucial for flame extinction. Fig. 6 shows the calculated MEC at different oxidizer temperatures while the oxidizer velocity was maintained at 13.6 cm/s, or at different oxidizer velocities while the oxidizer temperature was maintained at 298 K. The results show that the calculated MECs were nearly insensitive to the changes of oxidizer velocity and temperature over the investigated range. The insensitivity of MEC to the oxidizer velocity is similar to the plateau region observed by Takahashi et al. (Takahashi and Katta, 2009). We assumed that this insensitivity could be ascribed to the flow conditions at the flame base. The stream-tube shrinkage and boundary separation at the burner brim made the gas temperature and velocity at the flame base less sensitive to the corresponding changes at the oxidizer inlet. The calculated MECN2 and MECH2O decreased mildly as oxidizer velocity was increased from 5.6 cm/s to 17.6 cm/s. It was assumed that the increase of velocity might increase the local
Fig. 6. The variation of MECN2 and MECH2O with the temperatures and velocities of the inlet oxidizer.
In order to further figure out the relative importance of different mechanisms, we proposed a simplified model based on the cup burner configuration as showed in Fig. 1 (c). The cup burner flame is simplified into a cylindrical flame, with fuel inlet at the bottom, oxidizer entrained from the side and products emitted at the top. Heat is transferred to the ambient from the cylinder side surface by convection, conduction and radiation. The flame is considered as a fast reactor where the reaction is assumed to occur at infinite rate compared with the supply rate of fuel and oxidizer. The supply rates of fuel and air is assumed in stoichiometric proportion. This assumption is appropriate unless the suppressant concentration reaches its extinguishing limit. The equation is showed below:
m_ f DHf ¼ m_ f Cpf Th Tf þ m_ e Cpair ðTh Tox Þð1 Xs Þ þ m_ e Xs Cps ðTh Tox Þ þ hc hbðTh Tox Þ;
(12)
where m_ f and m_ e are the mass flow rates of fuel and entrained oxidizer respectively, DHf is the heat of combustion per unit mass of fuel, Cpf, Cpair and Cps are the specific heats of fuel, air and suppressant. Th, Tf and Tox are the temperatures of products, inlet fuel and oxidizer, Xs represents the mass fractions of suppressant in the oxidizer side, hc is the constant overall heat transfer coefficient, h is the flame height and b is the circumference of the cylinder. The flame height h was determined by the place where the maximum flame temperature was along the axis for each case. The results are showed in Fig. 7. As the suppressant was added, the flame height increased first to entrain more oxidizer to obtain a stoichiometric mixture and decreased sharply as the suppressant concentration was close to the MEC. The coefficient hc was calculated from the base case where no suppressant was applied, and was assumed constant throughout the extinguishing process. The specific heat was taken from the polynomial fittings of NIST chemistry book. For each case, Th was calculated from the above equation by the trial-and-error method. The relative flame
Fig. 7. The variation of the time-average flame height over the mass fraction of suppressants at the inlet.
M.-H. Feng, J. Qin / Journal of Loss Prevention in the Process Industries 38 (2015) 260e267
temperature drop between the suppressed flame and base flame was defined as the indicator for extinguishing effect. Fig. 8 shows the extinguishing effect of N2 and water vapor. Only dilution effect and increase of sensible heat effect are included in the simplified model. Since the specific heat of water vapor is twice than that of N2, the difference between the N2 theoretical data and H2O theoretical data can be ascribed to the thermal cooling effect. Furthermore, from the difference of the two theoretical data line, it is assumed that the thermal cooling effect is more important than the oxygen dilution effect in terms of flame suppression for the current configuration. The theoretical results of N2 are in reasonable agreement with the simulated one while substantial difference exists for H2O. Since chemical kinetics effect was excluded from the simplified model, we deduced this difference was due to the influence of H2O on the chemical reactions. However, the marginal effect of this chemical effect gradually reduced as the mass fraction of H2O was increased. As stated previously, higher temperature drop represented higher efficiency for an inert gas. But for H2O, its participation in flame reactions enhanced the suppression efficiency (reduced extinguishment limit) but lowered the temperature drop. Consequently, the temperature drop could not be used as an indicator of suppression efficiency in terms of the interaction of water vapor and hydrogen cup burner flame. 5. Conclusion Computational simulations were performed to show the interaction of “cold water steam” and hydrogen flame in a cup burner apparatus. The extinguishment limits of N2 and H2O were calculated by three models and were measured by experiments. The results demonstrate the accuracy of the CFD model and show that the flame extinction is controlled by the flame base. Besides, thermal cooling is a determinant on MEC and radiation is negligible in the interaction of water vapor and hydrogen cup burner flame. The computations shows the detailed flame structure and flame dynamics. A reaction kernel exists at the flame base and controls the flame extinction. An edge diffusion flame can be formed as the flame base of the near-limit flame detaches away from the burner brim. The chain-branching reactions at the reaction kernel can be significantly inhibited by the presence of water vapor and the hydrogen flame sustains combustion when the rates of these reactions are reduced below 1 kg mol/m3s. The reaction H þ O2 þ M ¼ HO2 þ M is crucial for the flame extinction in terms of chemical inhibition since the enhanced third body effect of water
Fig. 8. The variation of the relative temperature drops with the mass fractions of N2 and water vapor at the inlet.
267
vapor increases its reaction rate. The calculated MECs are insensitive to the temperature and velocity changes due to stream-tube shrinkage and boundary layer separation near burner brim. As the mass fraction of H2O gradually increases, the difference of DT/T0 between the results of Fluent and the simplified equation becomes larger but the marginal effect of chemical inhibition is reduced. The suppression effect of water vapor mainly includes dilution, thermal cooling and chemical kinetic effect but the latter two are the dominant extinguishing mechanisms. Acknowledgments This work is supported by the National Natural Science Foundation of China (Grant No. 51276176). The numerical calculations in this paper have been done on the supercomputing system in the Supercomputing Center of University of Science and Technology of China. References Ananth, R., Mowrey, R.C., 2008. Ultra-fine water mist extinction dynamics of a coflow diffusion flame. Combust. Sci. Technol. 180 (9), 1659e1692. Association, N.F.P., 2004. Standard on Clean Agent Fire Extinguishing Systems. National Fire Protection Association. Battersby, P., et al., 2012. Suppression of hydrogeneoxygenenitrogen explosions by fine water mist: Part 2. Mitigation of vented deflagrations. Int. J. Hydrogen Energy 37 (24), 19258e19267. Bergman, T.L., Incropera, F.P., Lavine, A.S., 2011. Fundamentals of Heat and Mass Transfer. John Wiley & Sons. Chen, Z., et al., 2008. Flame oscillation frequency based on image correlation. J. Combust. Sci. Technol. 4, 014. Cong, B., Liao, G., 2008. Mechanisms of suppressing cup-burner flame with water vapor. Sci. China Ser. E Technol. Sci. 51 (8), 1222e1231. Dobashi, R., 2014. Fire and explosion disasters occurred due to the great east Japan earthquake (March 11, 2011). J. Loss Prev. Process Ind. 31, 121e126. Ingram, J., et al., 2012. Suppression of hydrogeneoxygenenitrogen explosions by fine water mist: Part 1. Burning velocity. Int. J. Hydrogen Energy 37 (24), 19250e19257. Kee, R.J., et al., 1996. A fortran chemical kinetics package for the analysis of gas phase chemical and plasma kinetics. Sandia Natl. Lab. Linteris, G.T., et al., 2012. Stirred reactor calculations to understand unwanted combustion enhancement by potential halon replacements. Combust. Flame 159 (3), 1016e1025. Liu, S., et al., 2008. Determination of cup-burner extinguishing concentration using the perfectly stirred reactor model. Fire Saf. J. 43 (8), 589e597. Ng, H.D., Lee, J.H., 2008. Comments on explosion problems for hydrogen safety. J. Loss Prev. Process Ind. 21 (2), 136e146. Ni, X., Chow, W., 2011. Performance evaluation of water mist with bromofluoropropene in suppressing gasoline pool fires. Appl. Therm. Eng. 31 (17), 3864e3870. Conaire, M., et al., 2004. A comprehensive modeling study of hydrogen oxidation. O Int. J. Chem. Kinet. 36 (11), 603e622. Papas, P., Glassman, I., Law, C., 1994. Effects of pressure and dilution on the extinction of counterflow nonpremixed hydrogen-air flames. In: Symposium (International) on Combustion. Elsevier. Peters, N., Rogg, B., 1993. Reduced Kinetic Mechanisms for Applications in Combustion Systems. Springer Science & Business Media. Qiao, L., et al., 2010. Extinction of premixed methane/air flames in microgravity by diluents: effects of radiation and lewis number. Combust. Flame 157 (8), 1446e1455. Quintiere, J., 2006. Fundamentals of Fire Phenomena. Wiley. Seiser, R., Seshadri, K., 2005. The influence of water on extinction and ignition of hydrogen and methane flames. Proc. Combust. Inst. 30 (1), 407e414. Senecal, J.A., 2005. flame extinguishing in the cup-burner by inert gases. Fire Saf. J. 40 (6), 13. Takahashi, F., Katta, V.R., 2005. Structure of propagating edge diffusion flames in hydrocarbon fuel jets. Proc. Combust. Inst. 30 (1), 375e382. Takahashi, F., Katta, V.R., 2009. Extinguishment of diffusion flames around a cylinder in a coaxial air stream with dilution or water mist. Proc. Combust. Inst. 32 (2), 2615e2623. Takahashi, F., Linteris, G.T., Katta, V.R., 2007. Extinguishment mechanisms of coflow diffusion flames in a cup-burner apparatus. Proc. Combust. Inst. 31 (2), 2721e2729. Takahashi, F., Linteris, G.T., Katta, V.R., 2011. Extinguishment of methane diffusion flames by inert gases in coflow air and oxygen-enriched microgravity environments. Proc. Combust. Inst. 33 (2), 2531e2538. Zhang, S., Soteriou, M.C., 2011. An analytical model for the determination of the cup-burner minimum extinguishing concentration of inert fire suppression agents. Proc. Combust. Inst. 33 (2), 2505e2513.
Contents lists available at ScienceDirect
Journal of Loss Prevention in the Process Industries journal homepage: www.elsevier.com/locate/jlp
Extinguishment of hydrogen laminar diffusion flames by water vapor in a cup burner apparatus Ming-Hui Feng, Jun Qin* State Key Laboratory of Fire Science, University of Science and Technology of China, Hefei 230026, Anhui, China
a r t i c l e i n f o
a b s t r a c t
Article history: Received 10 November 2014 Received in revised form 20 July 2015 Accepted 11 October 2015 Available online 23 October 2015
Transient computations with full hydrogen chemistry were performed to reveal the flame structure and extinguishment process of co-flow, hydrogen diffusion flame suppressed by water vapor. As the concentration of water vapor was increased, the flame detached away from the burner brim and formed an edge flame at the flame base. Water vapor showed larger chemical inhibition effect than nitrogen when extinguishing hydrogen flame, which was attributed to its enhanced third body effect in the reaction H þ O2 þ M ¼ HO2 þ M. The minimum extinguishing concentration (MEC) of water vapor and nitrogen was predicted by Senecal formula and perfectly stirred reactor (PSR) model respectively. The MECs predicted by PSR model agree with the MECs calculated by Fluent, which shows that 1) the flame extinction is controlled by the flame base, and 2) radiation absorption is negligible. The measured MECs are in a reasonable agreement with the values calculated by Fluent, which demonstrates the accuracy of the CFD model. A simple model was used to investigate the relative importance of extinguishing mechanisms of water vapor. The results show that in a co-flow configuration the thermal cooling and chemical inhibition effect are the main extinguishing mechanisms in suppressing hydrogen diffusion cup burner flame. © 2015 Published by Elsevier Ltd.
Keywords: Hydrogen safety Diffusion flame Water vapor Nitrogen Flame suppression
1. Introduction Hydrogen is viewed as a promising clean energy at present, but it is dangerous for its wide explosion range and low-energy ignition. The hazards of hydrogen are a concern in many industries, not only where hydrogen is served as a reactant or product in the chemical industry, but also where hydrogen is an unwanted product in nuclear power plants due to the reaction between zirconium and water (Ingram and et al., 2012; Dobashi, 2014). Many researchers have investigated on the extinguishment/inhibition of hydrogen flame/explosion by some inert gases and ultra-fine water mist (Battersby and et al., 2012; Papas et al., 1994; Ng and Lee, 2008), but few efforts have been paid on the effects of water steam. The lack of interest may be attributed to the fact that the application of water vapor is restricted by its low saturated vapor pressure at room temperature. However, in the chemical or nuclear industry where water steam is used as a primary heat transfer medium, water steam is a potential alternative for ordinary
* Corresponding author. E-mail address: [email protected] (J. Qin). http://dx.doi.org/10.1016/j.jlp.2015.10.004 0950-4230/© 2015 Published by Elsevier Ltd.
suppressants. The scope of the present study is limited to the interaction of water steam and hydrogen diffusion flames in a cupburner apparatus at normal atmospheric environment (298 K and 1 atm). This arrangement is based on the following reasons besides of the hydrogen safety. The cup-burner method (Association, 2004) is a widely accepted way to measure gas agents’ minimum extinguishing concentration (MEC), which serves as a reference for the design of fire protection systems. Besides, the cup-burner flames are stable and simple enough so that the extinguishing mechanisms can be readily studied. Some simplified models have been proposed to predict the MECs of gas agents in order to harmonize data used in fire protection standards. Senecal (Senecal, 2005) proposed a phenomenological approach to predict the MECs of inert gases in terms of heat capacity and fuel properties. Zhang et al. (Zhang and Soteriou, 2011) and Liu et al. (Liu and et al., 2008) predicted the MECs of inert and chemical gas agents by using a perfectly-stirred reactor (PSR) model with full chemistry. Linteris et al. (Linteris and et al., 2012) extended the scope of the PSR model to understand the unwanted combustion enhancement by potential halon replacements. Although reasonable agreements are achieved between experiments and calculations, the validity of those models on the
M.-H. Feng, J. Qin / Journal of Loss Prevention in the Process Industries 38 (2015) 260e267
hydrogen cup-burner flame has not been testified. For laminar hydrogen premixed flame, many experimental and numerical studies have been performed to investigate the extinguishing efficiency and mechanisms of various suppressants. Qiao et al. (Qiao and et al., 2010) simulated the detailed flame structure of premixed hydrogen flame at different suppressant concentrations and found that the third-body reaction H þ O2 þ M ¼ HO2 þ M was an important chain-terminating reactions. Seiser et al. (Seiser and Seshadri, 2005) investigated the influences of water on the extinction and ignition of hydrogen/methane premixed flames and found that the addition of water vapor made the hydrogen flames easier to extinguish due to the enhanced chaperon efficiencies of water in reactions H þ O2 þ M ¼ HO2 þ M, H þ OH ¼ M ¼ H2O, and H þ H þ M ¼ H2 þ M. The extinguishing efficiency and mechanisms of various suppressants change with the flame types due to different flame structures and flame speeds. For diffusion flames, Takahashi et al. (Takahashi et al., 2007; Takahashi and Katta, 2009) studied the interaction process of co-flow diffusion flames and fire suppressants in a cup burner. Their results showed that the peak reactivity spot formed at the flame base was crucial for flame stabilization and the extinguishment occurred via a blowoff process rather than the global chemical extinction. Cong et al. simulated the interaction of water steam and methane cup burner flame with one-step chemistry (Cong and Liao, 2008). Their results supported the extinguishing process proposed by Takahashi and further demonstrated that the flame-base oscillation was the key step for flame extinction. It should be noted that hydrogen is free of carbon, and the fact that its combustion product is water indicates water steam's potential capability of curbing specific chemical reactions. Few data have been reported on the MEC of water vapor (MECH2O) and the relative importance of the extinguishing mechanisms when water steam interacts with hydrogen diffusion flame. In this study, the hydrogen co-flow diffusion flame was simulated with full chemistry by Fluent and its interaction with water steam was investigated. A PSR model, as well as the Senecal's formula, was used to predict the MECs for hydrogen cup-burner flame. Their results were compared with the Fluent's predicted values and the measured values to confirm the validity of the simplified models and to interpret the influences of different extinguishing mechanisms on MECs. Finally, the relative importance of the extinguishing mechanisms was investigated with an energy balance equation. 2. Experiment and CFD modeling 2.1. Physical model setup The sketch of the apparatus is shown in Fig. 1(a). The apparatus used here is similar to that in (Cong and Liao, 2008). The cup burner has a cylindrical stainless steel cup and a cylindrical quartz chimney. The bottle of the burner is connected to the TSI atomizer (Model 9306A) through a diffuser. The fuel and air flow rate are measured by a calibrated mass flow meter and vortex flow meter respectively. Their uncertainty is 1% of the indicated flow. For the case of water vapor, the air is preheated before entering the mist chamber. As the air enters the mist chamber, it mixes with ultrafine water mist and passes through an electrically heated fine screen upstream the exit of the cup. The temperature of the fine screen is carefully adjusted to ensure a total evaporation of ultrafine water mist and maintain the temperature of the inlet oxidizer at z373 K. The mass flow rate of water vapor equals to that of water mist during the experiments so its value is determined by collecting the water mist at the oxidizer inlet during a given period. The numerical parameters were set according to the experimental conditions. The stainless steel cup burner had an outer
261
diameter of 31 mm and length of 25 mm while the quartz chimney had an inner diameter of 90 mm and length of 200 mm. The temperature of the quartz wall and burner wall was set at 298 K except the 600 K burner wall within 1 mm away from the top burner brim (Ananth and Mowrey, 2008). No slip boundary condition was applied to the wall. The normal thermal emittance of the hot burner wall and cool wall was 0.24 and 0.22 respectively (Bergman et al., 2011) and the hemisphere integrated radiant emittance was 0.93 for the chimney wall. The initial fuel temperature was 298 K and the inlet velocity of the fuel was fixed at 6.5 cm/s to achieve a laminar flame with a visible flame height of about 75 mme85 mm (Association, 2004). Three types of oxidizer stream were considered in the simulations, they are: dry air, dry air with nitrogen, dry air with water vapor. The temperature of the inlet oxidizer stream was 298 K and its velocity was fixed at 13.6 cm/s at 1 atm. The mass fraction of the extinguishers in the oxidizer stream was gradually increased to its MEC where the flame sustained combustion.
2.2. The numerical method The Da number, chemical reaction time tr and flow time tf (Quintiere, 2006) were estimated from the following equations:
Da ¼
tf ; tr
(1)
tr ¼ rCp T∞ ½E=RT∞ Dhc AeE=ðRT∞ Þ ;
(2)
tf ¼ d=v:
(3)
In these equations, r and Cp are the gas density and specific heat capacity near the flame sheet, T∞ and v is the temperature and velocity of the inlet air, E and A is the activation energy and preexponential factor for the hydrogen respectively, Dhc is the idea heat of combustion and d is the equivalent distance for the air transporting to the flame sheet. The Da number is nearly 1 at the flame base under the present situations. This means that for the hydrogen cup-burner flame the chemical dynamic effect is at least equivalent to the transport ef Conaire and fect. So a detailed reaction mechanism of hydrogen (O et al., 2004) was incorporated into the laminar finite rate model. The gas phase combustion was solved by the laminar, transient, Navier-Sokes and energy equations with a time step of 0.1 s. The mass diffusion coefficient and thermal diffusion coefficient were calculated by the Chapman-Enskog and kinetic theory respectively. Fluent's DO model was used to account for the radiation loss of the flame. Radiation of the gaseous specie H2O was incorporated into the model by a polynomial fits and its absorption coefficient was taken from (Peters and Rogg, 1993). The radiation equation was calculated after solving the energy equation for 10 times at each time step. Fluent solves the conversation equations in finite volume forms. The gradients were computed using Least Squares Cell-Based Gradient method. The diffusion terms in governing equations were discretized by the secondeorder accurate central-differenced method. The convection terms in momentum equations were discretized using the Fluent's QUICK Scheme and those of the species and energy equations were discretized using a second-order upwind scheme. The coupled pressureevelocity equations were solved using Fluent's PISO method before solving energy and species conservation equations. The residuals for the gas phase energy equation were reduced to 104% for convergence while those for other equations were reduced to 102% for convergence.
262
M.-H. Feng, J. Qin / Journal of Loss Prevention in the Process Industries 38 (2015) 260e267
Fig. 1. Models for the interaction of water steam and cup-burner flame: (a) a sketch of cup-burner apparatus; (b) a sketch of the PSR model for the prediction of MEC; (c) a simplified model for the interpretation of extinguishing mechanisms.
2.3. The mesh grid size and sensitivity analysis Fine mesh size helps to accurately capture the flame dynamics but is time-consuming for computation. Since the flame thickness turned wider when extinguishers were added, the sensitivity analysis was restricted to the case where the oxidizer stream was dry air. The unburned fuel-air mixture at the uninhibited cupburner flame base was nearly stoichiometric (Takahashi et al., 2011), so the equation for calculating the thickness of laminar premixed flame was used:
Tb Tig dR z Tig Tu
!
a : Su
(4)
In this equation, dR is the premixed flame thickness, Tb is the adiabatic flame temperature, Tig is the auto-ignition temperature, Tu is the unburned gas temperature, a is the thermal diffusion coefficient and Su is the laminar burning velocity of the flat flame. The flame thickness was z 0.1 mm at the flame base. Consequently, the sensitivity analysis were performed on three structured meshes with minimum grid size of 0.125 mm by 0.125 mm, 0.0625 mm by 0.0625 mm, and 0.03125 mm by 0.03125 mm respectively. The maximum grid size of the three meshes was 0.25 mm by 0.25 mm. The preliminary simulations on the flow field of the non-reacting flow showed that the boundary layer was insensitive to the mesh grid when the first row height of the meshes near the walls was reduced to less than 0.3 mm. So the maximum grid size in the simulations would not influence the
results near the walls. The mesh independent issues were analyzed by examining the time-average temperatures distribution along the axis and the transient species mole fraction distribution along a horizontal line 3 mm away from the burner brim. The calculated temperature distribution and species mole fraction distribution of the cases are shown in Fig. 2. The results of the case with a minimum grid size of 0.03125 mm by 0.03125 mm are not shown in Fig. 2 because its results are nearly identical to the results of the case with a minimum grid size of 0.0625 mm by 0.0625 mm. The temperature distribution in Fig. 2 shows that the time-average flame height is nearly 75 mm, which meets the requirement for the cup burner test (Association, 2004). The temperature and specie distribution indicate that the minimum grid of 0.0625 mm by 0.0625 mm at the flame base is appropriate for capturing the flame dynamic. So this grid size was used in the rest simulations.
3. Simplified models for MEC prediction PSR is an ideal reactor where perfect mixing is achieved inside the control volume. It is extremely useful to study flame stabilization. Fig. 1 (b) shows the schematic diagram of the adiabatic PSR model. The governing equations of mass, species, and energy are showed below:
m_ out m_ in ¼ 0;
(5)
M.-H. Feng, J. Qin / Journal of Loss Prevention in the Process Industries 38 (2015) 260e267
263
Fig. 2. Temperature distribution along the axis and mole fraction distribution along a horizontal line 3 mm away from the burner brim.
$V; m_ in $ Yi;out Yi;in ¼ m;;; i
(6)
X Yi;out $hi;out Yi;in $hi;in ¼ 0:
(7)
i
In these equations, m_ out and m_ in are the inlet and outlet mass flow rates, Yi,out and Yi,in are the outlet and inlet mass fractions of ith specie, m;;; is the production rates of ith specie, V is reactor volume, i hi,out and hi,in are the outlet and inlet enthalpies of ith specie. m;;; could be solved from Eq. (8)e(10): i
m;;; ¼ w_ i $MWi ; i w_ i ¼
L X
(8)
qj $ v;;ij v;ij ;
(9)
flame base was comparable to the stoichiometric laminar flame Conaire and et al., 2004). Soot and radiation are not speed (O considered in this model because they are negligible for the nearlimit hydrogen flame. We then assume that the extinguishing mechanisms are similar for nitrogen and water stream, so the critical residence time of N2, at which point the flame is unable to sustain, is used as the critical residence time for water steam. The PSR model can be solved by CHEMKIN with the kinetic mechanism Conaire (O Conaire and et al., 2004) and thermochemical of O properties of Kee et al. (Kee and et al., 1996). The MEC of water vapor could be achieved by adjusting its concentration in the oxidizer stream to a critical value where the temperature suddenly drops. Senecal's approach is based on the assumption that heat capacity of the inert gas agent is the primary extinguishing mechanism. The detailed method is available in (Senecal, 2005).
j¼1
4. Results and discussion
qj ¼ kfj $
N Y
v;ij
½xi krj $
i¼1
N Y
½xi
v;;ij
(10)
In these equations, MWi is the molar weight of ith specie, v;ij and v;;ij are forward and backward stoichiometric coefficients of ith specie in jth reaction, kfj and krj are forward and backward rate constants of jth reaction, [xi] is the concentration of ith specie. m_ out can be expressed as Eq. (11),
m_ out ¼ r$V=t;
4.1. Extinguishment limits
i¼1
(11)
where r is the average density of the mixture and t is the residence time. hi,out and hi,in can be correlated with the temperature in the reactor. For a given residence time and an appropriate inlet condition, the temperature in the reactor can be solved. To apply the PSR for MEC prediction, we assume that the fuel and air are stoichiometrically premixed and reacts instantaneously at the flame base. This assumption is plausible for the cup-burner apparatus since researches showed that the unburned mixture velocity at the
Different extinguishing mechanisms of suppressants have a direct impact on MECs. Nitrogen and water vapor are selected as the extinguishers in this study. The extinguishing mechanisms in real applications and those adopted in the three models are summarized in Table 1 based on the previous studies (Senecal, 2005; Conaire and et al., 2004; Ni and Takahashi and Katta, 2009; O Chow, 2011). The blow-off conditions for various hydrogen/air/N2/ H2O mixtures were first determined by Fluent. The time-dependent calculation was run for separate cases in which the volume fraction of the extinguishers in the oxidizer stream was stepwise increased at 1% interval. As the volume fraction of the extinguishers reached to the critical value (defined as MEC), the flame detached away from the burner, drifted downstream and exited the computational domain. The calculated MECs of nitrogen (MECN2) and water vapor (MECH2O) are shown in Table 2. When the temperature of the inlet oxidizer stream (Toxidzer) was fixed at 298 K, the calculated MECH2O could not be reached in real life. In the current study, this effect was
264
M.-H. Feng, J. Qin / Journal of Loss Prevention in the Process Industries 38 (2015) 260e267
Table 1 Extinguishing mechanisms of N2 and water vapor. Extinguishment mechanism
H2O (vapor)
N2
Oxygen dilution Increase of sensible enthalpy Chemical kinetic effects Radiation attenuation by gas species
Real
Senecal
PSR
Fluent
Real
√ √
√
√ √
√ √
√ √ √ √
Table 2 A comparison of MECN2 and MECH2O. Model
Oxidizer temperature
N2
H2O
Fluent Senecal formula PSR model Fluent Experiment
298 298 298 373 373
36 e e 38 37.7
32 31.0 32.1 37 35.9
K K K K K
not considered so that the extinguishing efficiency of N2 and water vapor could be compared and the different mechanisms could be interpreted without the influence of oxidizer temperature. Table 2 also shows the measured MECs by experiments and the calculated MECs by the simplified models. It should be noted that the MECH2O was measured when Toxidzer was maintained 373 K in the experiments. So additional cases where Toxidzer was fixed at 373 K were calculated by Fluent to testify the accuracy of the model. The MECH2O predicted by the Senecal formula and the adiabatic PSR model was calculated based on the MECN2 calculated by Fluent in order to have some basic knowledge about the influences of different mechanisms on the MEC. When Toxidzer is 373 K, the calculated MECN2 by Fluent agrees well with its experimental counterpart. This indicates that the CFD model can accurately predict the MEC for the hydrogen cup burner flame. Besides, the increase of Toxidzer from 298 K to 373 K slightly increases the calculated MECN2 from 36% to 38%. The extinguishing mechanisms of N2 include thermal cooling and dilution. The increase of Toxidzer brings additional enthalpies to the mixture and compromises the cooling effect of N2. The predicted MECH2O by Fluent and PSR model are in reasonable agreement if taking account for the 1% increment in Fluent's calculation. This agreement supports the assumption in the PSR model: 1) the flame base is the key factor for the flame stabilization and 2) the radiation attenuation of water vapor is negligible for hydrogen flame. Table 2 also shows that the predicted MECH2O by Senecal's formula is 1.1% lower than the PSR model. This fact indirectly demonstrates that chemical kinetic effect may be important for the suppression of hydrogen cup-burner flame by water vapor since only increase of sensible enthalpy effect is included in the Senecal's formula. The increase of Toxidzer increases the calculated MECH2O by Fluent from 32% to 37%. The reason for this increase is the same with that for N2. For the case with Toxidzer¼373 K, the measured MECH2O is z 3% lower than the calculated value by Fluent. It is assumed that the deviation mainly comes from two aspects. Firstly, a portion of ultrafine water mist in the mixing chamber is heated by the preheated air and changes into water vapor. This portion may not be collected at the oxidizer inlet. Secondly, the edge flame at the flame base is susceptible to the turbulence at the oxidizer inlet. The introduction of water mist upstream the oxidizer inlet increases the turbulence at the oxidizer inlet and subsequently reduces the measured MECH2O. 4.2. Flame structure Fig. 3 shows the temperature contours and streamlines of the flame with saturated water vapor (298 K) at two instants separated
Senecal
PSR
Fluent
√
√ √ √
√ √ √ √
by 0.05 s. The flame was attached to the burner brim and inclined inwardly due to the large velocity gap between the fuel and oxidizer stream as showed in the temperature contour. The flow in the tube was affected by the flame. The generation and shredding of the vortex near the flame tip caused the cutting off of the flame tip as showed in the streamlines in Fig. 3. Two vortices existed at the bulk air side at the left half of Fig. 3. They merged into one and a new vortex was formed at the right half part. The process repeated itself at a fixed frequency corresponding to the flickering frequency of the flame. Five temperature monitors were set to measure the instantaneous temperature of the simulated flame along the axis in the continuous flame zone and fire plume. We transformed the temperature-time sequence into the magnitude-frequency sequence by a Fast Fourier Transformation. The obtained flickering frequency was nearly 10.94 Hz, which is slightly higher than 9 Hz estimated from Pagni’ s empirical correlation for pool fire (Chen and et al., 2008). Fig. 4 shows the calculated structure of the near-limit hydrogen flame suppressed by water vapor. The right part of Fig. 4 shows that the oxidizer was entrained into the flame base and its velocity was substantially increased due to the buoyancy-driven flow. As the concentration of H2O in the oxidizer side increased, the flame base gradually detached away from the burner brim. As showed in the left part of Fig. 4, the gap between the flame base and burner brim left a path for the oxidizer to penetrate the flame. This allows the fuel and oxidizer to be partially premixed before entering the flame base. The concentration gradient of the partially premixed mixture resulted in the formation of edge flame at the flame base. However, the edge flame did not develop into edge flame because the fuelsided and oxidizer-sided branch merged to the trailing diffusion flame. This can be explained by the finding that the formation of triple flame is fuel-dependent: acetylene and ethylene form the triple flame structure, but alkanes fails it (Takahashi and Katta, 2005). For the chemical reactions, Seiser et al. (Seiser and Seshadri, 2005) found that for hydrogen flames, the enhanced chaperon efficiencies of water in reactions H þ O2 þ M ¼ HO2 þ M, H þ OH þ M ¼ H2O þ M, and H þ H þ M ¼ H2 þ M made flames easier to extinguish. In Fig. 5, we tracked the reaction rates of the above three reactions as well as two chain-branching reactions at different mass fraction of water vapor. The reaction kernel was roughly determined by the position where the chain-branching reactions reached its maximum values. The chain-branching reactions were significantly inhibited as the mass load of water vapor was increased and the flame was nearly extinguished when the chain-branching reaction rates were below 1 Kg-mol/m3sec. In the current reaction mechanism, the chaperon efficiencies of water vapor are 12, 12 and 14 respectively, which are much more than other species. As the concentration of water vapor increased, the reaction H þ H þ M ¼ H2 þ M was enhanced while the other two were nearly insensitive to the mass fraction of water vapor. This meant that for this configuration the reaction H þ H þ M ¼ H2 þ M was crucial for flame suppression. When water vapor reached its critical value, the reactions were completely suppressed and the flame was eventually blown off.
M.-H. Feng, J. Qin / Journal of Loss Prevention in the Process Industries 38 (2015) 260e267
265
Fig. 3. Streamlines and temperature contours at two instants for the cases with saturated water vapor at the oxidizer inlet: left part at 20.40 s, right part at 20.45 s.
Fig. 4. Calculated structure of the near-limit hydrogen cup-burner flame for the cases with the inlet oxidizer diluted by 31% H2O. Left half: contours of mole fraction for H2O and O2. Right half: contours of temperature and mole fraction for H.
266
M.-H. Feng, J. Qin / Journal of Loss Prevention in the Process Industries 38 (2015) 260e267
strain rate of the edge flame and bring more heat away from the flame base so that the temperature at flame base was easier to reach the critical temperature. Besides, the calculated MECN2 and MECH2O slightly increased as the oxidizer temperature was increased from 298 K to 373 K because of the additional heat brought by the oxidizer. 4.4. Flame extinguishing mechanisms
Fig. 5. The variation of the reaction rate inside the reaction kernel with the mass fraction of water vapor at the inlet.
4.3. Influences of oxidizer temperature and velocity on flame extinction In real applications, the mass fraction of water vapor is restricted by its saturated vapor pressure at a given temperature. Besides, the release of gas agent will affect the local strain rate, which is crucial for flame extinction. Fig. 6 shows the calculated MEC at different oxidizer temperatures while the oxidizer velocity was maintained at 13.6 cm/s, or at different oxidizer velocities while the oxidizer temperature was maintained at 298 K. The results show that the calculated MECs were nearly insensitive to the changes of oxidizer velocity and temperature over the investigated range. The insensitivity of MEC to the oxidizer velocity is similar to the plateau region observed by Takahashi et al. (Takahashi and Katta, 2009). We assumed that this insensitivity could be ascribed to the flow conditions at the flame base. The stream-tube shrinkage and boundary separation at the burner brim made the gas temperature and velocity at the flame base less sensitive to the corresponding changes at the oxidizer inlet. The calculated MECN2 and MECH2O decreased mildly as oxidizer velocity was increased from 5.6 cm/s to 17.6 cm/s. It was assumed that the increase of velocity might increase the local
Fig. 6. The variation of MECN2 and MECH2O with the temperatures and velocities of the inlet oxidizer.
In order to further figure out the relative importance of different mechanisms, we proposed a simplified model based on the cup burner configuration as showed in Fig. 1 (c). The cup burner flame is simplified into a cylindrical flame, with fuel inlet at the bottom, oxidizer entrained from the side and products emitted at the top. Heat is transferred to the ambient from the cylinder side surface by convection, conduction and radiation. The flame is considered as a fast reactor where the reaction is assumed to occur at infinite rate compared with the supply rate of fuel and oxidizer. The supply rates of fuel and air is assumed in stoichiometric proportion. This assumption is appropriate unless the suppressant concentration reaches its extinguishing limit. The equation is showed below:
m_ f DHf ¼ m_ f Cpf Th Tf þ m_ e Cpair ðTh Tox Þð1 Xs Þ þ m_ e Xs Cps ðTh Tox Þ þ hc hbðTh Tox Þ;
(12)
where m_ f and m_ e are the mass flow rates of fuel and entrained oxidizer respectively, DHf is the heat of combustion per unit mass of fuel, Cpf, Cpair and Cps are the specific heats of fuel, air and suppressant. Th, Tf and Tox are the temperatures of products, inlet fuel and oxidizer, Xs represents the mass fractions of suppressant in the oxidizer side, hc is the constant overall heat transfer coefficient, h is the flame height and b is the circumference of the cylinder. The flame height h was determined by the place where the maximum flame temperature was along the axis for each case. The results are showed in Fig. 7. As the suppressant was added, the flame height increased first to entrain more oxidizer to obtain a stoichiometric mixture and decreased sharply as the suppressant concentration was close to the MEC. The coefficient hc was calculated from the base case where no suppressant was applied, and was assumed constant throughout the extinguishing process. The specific heat was taken from the polynomial fittings of NIST chemistry book. For each case, Th was calculated from the above equation by the trial-and-error method. The relative flame
Fig. 7. The variation of the time-average flame height over the mass fraction of suppressants at the inlet.
M.-H. Feng, J. Qin / Journal of Loss Prevention in the Process Industries 38 (2015) 260e267
temperature drop between the suppressed flame and base flame was defined as the indicator for extinguishing effect. Fig. 8 shows the extinguishing effect of N2 and water vapor. Only dilution effect and increase of sensible heat effect are included in the simplified model. Since the specific heat of water vapor is twice than that of N2, the difference between the N2 theoretical data and H2O theoretical data can be ascribed to the thermal cooling effect. Furthermore, from the difference of the two theoretical data line, it is assumed that the thermal cooling effect is more important than the oxygen dilution effect in terms of flame suppression for the current configuration. The theoretical results of N2 are in reasonable agreement with the simulated one while substantial difference exists for H2O. Since chemical kinetics effect was excluded from the simplified model, we deduced this difference was due to the influence of H2O on the chemical reactions. However, the marginal effect of this chemical effect gradually reduced as the mass fraction of H2O was increased. As stated previously, higher temperature drop represented higher efficiency for an inert gas. But for H2O, its participation in flame reactions enhanced the suppression efficiency (reduced extinguishment limit) but lowered the temperature drop. Consequently, the temperature drop could not be used as an indicator of suppression efficiency in terms of the interaction of water vapor and hydrogen cup burner flame. 5. Conclusion Computational simulations were performed to show the interaction of “cold water steam” and hydrogen flame in a cup burner apparatus. The extinguishment limits of N2 and H2O were calculated by three models and were measured by experiments. The results demonstrate the accuracy of the CFD model and show that the flame extinction is controlled by the flame base. Besides, thermal cooling is a determinant on MEC and radiation is negligible in the interaction of water vapor and hydrogen cup burner flame. The computations shows the detailed flame structure and flame dynamics. A reaction kernel exists at the flame base and controls the flame extinction. An edge diffusion flame can be formed as the flame base of the near-limit flame detaches away from the burner brim. The chain-branching reactions at the reaction kernel can be significantly inhibited by the presence of water vapor and the hydrogen flame sustains combustion when the rates of these reactions are reduced below 1 kg mol/m3s. The reaction H þ O2 þ M ¼ HO2 þ M is crucial for the flame extinction in terms of chemical inhibition since the enhanced third body effect of water
Fig. 8. The variation of the relative temperature drops with the mass fractions of N2 and water vapor at the inlet.
267
vapor increases its reaction rate. The calculated MECs are insensitive to the temperature and velocity changes due to stream-tube shrinkage and boundary layer separation near burner brim. As the mass fraction of H2O gradually increases, the difference of DT/T0 between the results of Fluent and the simplified equation becomes larger but the marginal effect of chemical inhibition is reduced. The suppression effect of water vapor mainly includes dilution, thermal cooling and chemical kinetic effect but the latter two are the dominant extinguishing mechanisms. Acknowledgments This work is supported by the National Natural Science Foundation of China (Grant No. 51276176). The numerical calculations in this paper have been done on the supercomputing system in the Supercomputing Center of University of Science and Technology of China. References Ananth, R., Mowrey, R.C., 2008. Ultra-fine water mist extinction dynamics of a coflow diffusion flame. Combust. Sci. Technol. 180 (9), 1659e1692. Association, N.F.P., 2004. Standard on Clean Agent Fire Extinguishing Systems. National Fire Protection Association. Battersby, P., et al., 2012. Suppression of hydrogeneoxygenenitrogen explosions by fine water mist: Part 2. Mitigation of vented deflagrations. Int. J. Hydrogen Energy 37 (24), 19258e19267. Bergman, T.L., Incropera, F.P., Lavine, A.S., 2011. Fundamentals of Heat and Mass Transfer. John Wiley & Sons. Chen, Z., et al., 2008. Flame oscillation frequency based on image correlation. J. Combust. Sci. Technol. 4, 014. Cong, B., Liao, G., 2008. Mechanisms of suppressing cup-burner flame with water vapor. Sci. China Ser. E Technol. Sci. 51 (8), 1222e1231. Dobashi, R., 2014. Fire and explosion disasters occurred due to the great east Japan earthquake (March 11, 2011). J. Loss Prev. Process Ind. 31, 121e126. Ingram, J., et al., 2012. Suppression of hydrogeneoxygenenitrogen explosions by fine water mist: Part 1. Burning velocity. Int. J. Hydrogen Energy 37 (24), 19250e19257. Kee, R.J., et al., 1996. A fortran chemical kinetics package for the analysis of gas phase chemical and plasma kinetics. Sandia Natl. Lab. Linteris, G.T., et al., 2012. Stirred reactor calculations to understand unwanted combustion enhancement by potential halon replacements. Combust. Flame 159 (3), 1016e1025. Liu, S., et al., 2008. Determination of cup-burner extinguishing concentration using the perfectly stirred reactor model. Fire Saf. J. 43 (8), 589e597. Ng, H.D., Lee, J.H., 2008. Comments on explosion problems for hydrogen safety. J. Loss Prev. Process Ind. 21 (2), 136e146. Ni, X., Chow, W., 2011. Performance evaluation of water mist with bromofluoropropene in suppressing gasoline pool fires. Appl. Therm. Eng. 31 (17), 3864e3870. Conaire, M., et al., 2004. A comprehensive modeling study of hydrogen oxidation. O Int. J. Chem. Kinet. 36 (11), 603e622. Papas, P., Glassman, I., Law, C., 1994. Effects of pressure and dilution on the extinction of counterflow nonpremixed hydrogen-air flames. In: Symposium (International) on Combustion. Elsevier. Peters, N., Rogg, B., 1993. Reduced Kinetic Mechanisms for Applications in Combustion Systems. Springer Science & Business Media. Qiao, L., et al., 2010. Extinction of premixed methane/air flames in microgravity by diluents: effects of radiation and lewis number. Combust. Flame 157 (8), 1446e1455. Quintiere, J., 2006. Fundamentals of Fire Phenomena. Wiley. Seiser, R., Seshadri, K., 2005. The influence of water on extinction and ignition of hydrogen and methane flames. Proc. Combust. Inst. 30 (1), 407e414. Senecal, J.A., 2005. flame extinguishing in the cup-burner by inert gases. Fire Saf. J. 40 (6), 13. Takahashi, F., Katta, V.R., 2005. Structure of propagating edge diffusion flames in hydrocarbon fuel jets. Proc. Combust. Inst. 30 (1), 375e382. Takahashi, F., Katta, V.R., 2009. Extinguishment of diffusion flames around a cylinder in a coaxial air stream with dilution or water mist. Proc. Combust. Inst. 32 (2), 2615e2623. Takahashi, F., Linteris, G.T., Katta, V.R., 2007. Extinguishment mechanisms of coflow diffusion flames in a cup-burner apparatus. Proc. Combust. Inst. 31 (2), 2721e2729. Takahashi, F., Linteris, G.T., Katta, V.R., 2011. Extinguishment of methane diffusion flames by inert gases in coflow air and oxygen-enriched microgravity environments. Proc. Combust. Inst. 33 (2), 2531e2538. Zhang, S., Soteriou, M.C., 2011. An analytical model for the determination of the cup-burner minimum extinguishing concentration of inert fire suppression agents. Proc. Combust. Inst. 33 (2), 2505e2513.