COMBUSTION A N D FLAME 36: 19-26 (1979) 19

Calculation of Turbulent Flame in the One-Dimensional Approach

V. Ya. BASEVICH, V. P. VOLODIN, S. M. KOGARKO, and N. I. PEREGUDOV

Institute of ChemicalPhysics, USSR, Academy of Science, 117334, Moscow B-334

(Vorob/evskoe Shaussee, 2b), USSR

The propagation velocity and reaction-zone structure of turbulent flame are considered. A method is pro-
posed for quantitative calculation of turbulent mixing and combustion processes. The method takes into
account the actual instantaneous parameters (concentrations, temperature, and fluctuation velocities). Dif-
ferential equations are solved for instantaneous parameters; and the averaging procedure is made not before,
but after the solution. A realistic instantaneous and averaged picture of the phenomenon can be obtained in
this way. The suggestion about "one-dimensional" turbulence is a very rough approach, but the solution
that must be considered as a model of the phenomenon can be justified because of its simplicity. The
method can be applied for two- and three-dimensional treatment. The computational results are presented
as an example.

INTRODUCTION rate of chemical reaction is depended on the

inhomogeneity of temperature and composition
The propagation velocity and reaction-zone struc- caused by turbulent fluctuation. Actual param-
ture (concentrations and temperature distribution) eters of the process realized at any time can differ
of flame are greatly dependent on the kinetics of very essentially, by an order of magnitude or more
chemical combustion. from mean values because o f nonlinearity.
The set of the equations for the balance o f The possibility of calculating instantaneous
energy and matter is known and allows description values (concentrations and temperature) during
of the propagation of laminar flames [1]. Thus it the mixing has been demonstrated earlier [7]. The
is possible to get the solution for the complex law o f turbulent mixing in the case of "one-
hydrogen-oxygen flame with a chain reaction [2]. dimensional" turbulence has been obtained [8, 9]
In practice the combustion is almost always without using any phenomenological coefficient
turbulent; and although the models of turbulent of turbulent transport. The present study contains
burning are known [3-6], there is as yet no an attempt o f computational description of the
method to define actual instantaneous parameters turbulent combustion.
of turbulent flame and the mean values therefrom.
First, this relates to the reaction zone and its NUMERICAL METHOD AND COMBUSTION
parameters, such as the thickness and the propaga- SYSTEM
tion velocity. With turbulent environment these
parameters are essentially determined by actual The complete set of differential equations for the
instantaneous velocity fluctuations of gas. Indeed, burning in turbulent environment has to contain
the rate o f reaction strongly depends on relation nonsteady-state equations for balance of heat,
between time t of chemical conversion and fre- matter, and motion in three-dimensional space.
quency v of velocity fluctuations. At tv ~< 1 the To simplify the solution, the mean motion has

Copyright © 1979 by the Combustion Institute

Published by Elsevier North Holland, Inc. 0010-2180/79/070019+8501.75
20 V. Ya. BASEVICH ET AL.

been excluded. The main problem of turbulent of solutions and relating to individual realization
burning, which is retained and can be investigated of the fluctuating velocity field, one can obtain
in this way, is the interference of reaction and the burning velocity in turbulent flow, the sum-
motion fluctuations. mary width of reaction zone usually observed in
Assuming quasisteadiness-stationarity and some experiments, and other values.
other approximations, the equations for heat and The solution of three-dimensional equations
matter balance can be written as (k = 3) is exceedingly complex; the solution of
aT i} aT two-dimensional equations (k = 2) can be obtained
at present; the solutions in this article are obtained
fl k axk for the simplest case-for the one-dimensional
assumption (k = 1). Such approximation for the
-- P ~ ac Tuh
laminar case is common. The suggestion about
ax~
"one-dimensional" turbulence contradicts the
a real physical picture of the phenomenon and
a(ndp) does not correspond to the continuity equation
Po at j Oxk for uncompressible fluid (Eq, 2). But the solution
that must be considered as the phenomenon
a(npk/P)
- - P Z.¢ ' (1) model can be justified in view of its simplicity
k axk and considerable saving of computing time. The
where one-dimensional approach was used for the first
Po, P = initial and current densities; time by Burgers [10].
T= temperature; The combustion process of homogeneous mix-
C = heat capacity; ture in turbulent flow with jet stabilization is
t = time; considered (Fig. 1). A high-temperature jet comes
~= molecular conductivity; out through the central axial channel and interacts
X k == cOordinate;
with the fresh homogeneous mixture that enters
hi, D1= concentration and molecular through the side channels. Burning arises in the
diffusivity of ! component; mixing zone. It spreads through the whole cham-
Wij, htj = rate and heat of elementary step ber. Here the solutions are obtained for chemical
number t for ] component; reaction, the rate of which is assumed as Z Wu =
Uk = projection of instantaneous velocity W = kn, where k = A exp ( - E / R T ) ; and k is the
fluctuatiom onto the xk-eoordinate rate constant of chemical reaction (A is the
axis, connected by continuity equa- preexponential factor, E the activation energy,
tion and R the gas constant). Thus, I¢ is depended on
~)Uk P fluctuating instantaneous concentrations and tem-
k =o. (2) peratures. The set of Eq. 1 for this case consists
of two equations, for T and n, and it was solved
To dose the set, the actual values of the projec- with boundary conditions Ix = x 1, (t =-xa/U), U =
tion of velocity fluctuations must be obtained mean flow velocity] :
either as accidental numbers according to the
space-time correlations observed in turbulent flow
t=O n=no, r=ro
or in some other way. for x = O + b l , b2 +b,
The solution of the equations yields the distri-
bution of the actual values ni and T in space and n = n l, T= TI
time and therefrom the propagation velocity of
for x = b 1 - b2
burning wave, the thickness of the reaction zone,
its structure, and other parameters. By averaging ac ru a(nu/p)
x = O, b p = p a--S--=o, (3)
the actual values nl and T obtained in the series
TURBULENT-FLAME CALCULATIONS 21

where the index 0 denotes side channels, index The sampling procedure is as follows. The
" 1 " the central channel, and b x, b 2, and b the Gauss law
corresponding distances.
There are at least two methods for determina-
tion of u = uk (k = 1) in Eq. 1. The first one is the
,
p(u)=v a exP L 2°2 J
F:(u:o ,l
solution of hydrodynamic turbulence equations.
This method is preferable but too complex. The
is taken for actual velocity fluctuations u. At t = 0
second one used below is the obtaining of u(x, t)
and x = Xl = 0 the value u(0, 0) is obtained in
by the Monte Carlo sampling procedure. Indeed,
accordance with mean velocity a = 0 and the dis-
u(x, t) is the field of random numbers, which obey
persion 0 z = ~z. Then the m random points in the
the statistical correlation. Here not all correlations
of the turbulence are realized (this is impossible at interval (0, b) of the x axis are selected, as con-
the present time), but only two, characterized by sistent with uniform probability distribution.
integral space L and Lagrangian time r scales. Scale The space velocity-correlation coefficient is
L gives the mean statistical size of individual liquid calculated by the approximate expression Rzxx2 =
particle ensemble in flow that moves as a whole; exp ( - A x 2 / L ) for the point x = x 2, which is the
scale T is the time, in the course of which this nearest neighbor to the first one and is separated
individual liquid particle ensemble is maintained. from it by a distance ~ c 2 = x2 - Xl; the actual
Both L and T are independed variables. value o f velocity fluctuation u(x 2, 0) is deter-
It seems to be necessary to make u sampling as mined in terms of normal correlation with the
random numbers at every integration step of mean velocity a = Raxzu(x 1, 0) and the disper-
Eq. 1. Calculation shows that the u spectrum sion o z = ~ ( 1 Rax22) and so on up to the
exhibits in this case very high frequencies, which points x = Xm and x = b.
correspond to the step of calculation but are After a time interval At1 the new m random
absent in the experimental spectrum. The sampling points on the x axis are determined again and the
with steps Ax > L and At > r does not give the actual velocity fluctuations u(x, A t l ) are sampled.
space-time correlations. Therefore, the step of Namely, for the point with the coordinate x =
sampling must be both not too small and not too x 1' = 0 the actual velocity fluctuation is obtained
large. Moreover, it must not be constant during with the time velocity correlation coefficient
the sampling because then false frequencies might RAt1 = exp ( - - A t { r ) , the mean velocity a =
appear, which would correspond to the value of R,atlU(X 1, 0) and the dispersion o 2 = ~2(1 -
this constant step. Therefore, the sampling proce- R ~ t l 2 ). For the next point x = x2', which is at
dures are made with accidental steps, and inter- the distance Ax 2' = x 2' - x 1 ' from the point x =
polation o f the u value is used between the points x l ' , the velocity fluctuation u(x2' , At1) is calcu-
of sampling. This procedure gives a u field similar lated according to the space-time velocity corre-
to the experimental one. lation coefficient RAtlR~x2, , the dispersion
Thus the fluctuating velocity field u(x, t) is a 2 = U2(1 - - R / , t l R / , x 2 , ) , and the mean velocity
obtained separately and used for solution of the
1. t t
equation set in Eq. 1. The actual values of turbu- a -~d[R~tl • u(x2, O)+Rax2'U(X 1 , At1)
lent velocity fluctuations u(x, t) are obtained as
random numbers according to the mean-square + R A t l R A x 2'u(x 1 ', 0)]
velocity fluctuations ~-, space L, and time T scales,
which are assumed the same for the whole field and so on. The values u(x, t) between the points
(x, t). This condition corresponds to assumptions of raffling are obtained by linear interpolation.
on isotropy, homogeneity, quasistationarity of The solution of the series of equations in Eq. 1
turbulent motion, and absence of combustion for k = 1 with the boundary conditions in Eq. 3
influence on velocity fluctuations. Such assump- are obtained by the method of finite differences.
tions are not compulsory and can be different. The steps of integration At are determined in
22 V. Ya. BASEVICH ET AL.

accordance with the permitted extent of increase c** = 10 - a cal/mole • K";

in temperature T and the concentrations n. It has
X = X . - (T/293) °.7,
been assumed that Atrnax <: AtiQ = 1, ""), the
steps of sampling. If the current time step turns X. = 6,25 • 10 - 5 cal/cm • sec • K;
out to be higher than Atl, the actual velocity
p = p . ( 2 9 3 / T ) . (P/100),
calculated at Ati is sent on the preceding layer and
for the new layer the new actual values u(x, t) are p . = 4 • 10 - ~ mole/era ~ ;
raffled and so on up to the terminal time. The
D = D , ( T / 2 9 3 ) 1.75. (100/P),
generator random number [ i i ] has been used in
the course o f raffling. D . = 0.795 cm2/sec,
The solutions were obtained for some set of
h = 57,100 cal/mole.
values n o, T 1 and parameters of turbulence ~, L
and r. One such solution is considered, as the
possible individual realization of actual instanta- From the experimental results [2], A = 2.26-
neous values n and T that correspond to actual l0 s sec- 1 , E = 11,000 cal/mole. T x is assumed to
accidental field of fluctuating velocity u(x, t). be the adiabatic flame temperature, To = 293 K. A
Several N instantaneous sets of values n and T turbulent combustion schematic is presented in
yield mean values Fig. 1.
Figure 2 presents the results of calculation for
= ~., n(x, t)lN, T = ~ T(x, tUN. laminar flame propagation-the distribution of T
N N at times 0, 20, 40, 60, and 80 msec. The flame
burning velocity u , is defined as the motion
velocity of half-temperature point T1/z, and
the reaction zone thickness 8 n as the doubled
RESULTS AND DISCUSSION x projection of the maximum gradient tangent
The results of calculation were compared with to the temperature profile between the points T O
experimental data [12] on combustion of lean and T 1. The appearance of T > T 1 is accounted
hydrogen-air mixtures [H2] o = 8.5-11%, at pres- for by diffusion influence and inequality of the
sures of 14-50 kPa and mean flow velocity U = Lewis-Semenov number to unity.
30.6 m/sec. The scales do not change with pressure The lines in Fig. 3 present the calculated values
increase, and ~ increases ...po.a according to u , and 5 , , the points refer to corresponding
Kramtsov [13]. At P = 50 kPa the parameters of experimental results obtained 1 by the Basevich-
turbulence are taken from Kozlov et al. [14]: Kogarko method [2] (constant volume apparatus,
wire-resistance thermometer techniques; wire
= 190 cm/sec,
parameters: diameter, (2-3).10 - 4 cm; length,
L = 4.2 cm, 0.3-0.5 ram).
r = 3.2 msec, The selected empirical expression for rate of
chemical reaction describes almost correctly the
b = 6 cm,
dependence of laminar flame velocity un on pres-
sure and mixture composition, although the widths
and
of reaction zone are slightly underestimated [case
rn = 20. u(x, t) - 01.
In Fig. 4 the results of calculation are presented
for the turbulent flame; the distribution of actual
The values accepted were:
values T and mean T(N = 9) at time t = 10 msec
C = C, + c**T,
are to the left of the coordinate axis, and mean T

c . = 7 cal/mole • K, 1 In collaboration with Yu. P. Kafyrin.

TURBULENT-FLAME CALCULATIONS 23

x 7"*K.

1 m~

300 _to_L_ _ \
v

,I i

3,O ' ~" 0 ' ' 9,O

' 6,O
cm

Fig. 1. T u r b u l e n t c o m b u s t i o n schematic. Fig. 2. Temperature profile o f laminar flame.

O
g

0
! ,.-...,.. ~ ° .
o,#
0
50 0 0
ttc~A 0°0~ 0
0

oo e ~ l
' 20 ff ~p~ L/O
' ' I

Fig. 3. Propagation velocity u n and reaction-zone width 8n o f laminar flame: (a)

[H2] o = 8.5%; ( b ) P = 25 kPa.

a 6

s/
r2~
0° Crtl
r- i

- %o,¢~

Ii 600
$
5"O
J
.,5," JO0 mot rain
m • • • o

0
' jo ' ~'o ; ;o ,,
& I I I I I

eo ~o ~o
X cm Fig. 5. Propagation velocity UnT, and reaction-zone width
Fig. 4. T e m p e r a t u r e profile o f turbulent flame: L = 4.2 6n7, of turbulent flame: (a) [H2] O = 8.5%; (b) P = 25 kPa.
c m ; r = 3.2 m s e c ; u = 190 cm/sec. L = 4.2 cm, ¢ = 3.2 msec, ~ = 190 (P/50).0.3
24 V. Ya. BASEVICH ET AL.

at times 0, 2, 4, 6, 8, and 10 msec are to the right. 50

The reaction-zone thickness ~,,T Can be obtained
in the previously described way using instanta-
neous values T. Its predicted values are presented
by curves in Fig. 5 and the experimental ones [12]
2E
rntrt
(flow system, wire-resistance thermometer tech-
niques) by points. It can be seen from comparison I I I I
of the calculated curves in Figs. 4 and 5 that
fi~T ~" 8n- Actual instantaneous reaction zone 0 t00 E c /s 200
widths in turbulent flame are nearly equal to those
Fig. 6. Propagation velocity of turbulent flame UnT vs
in laminar flame. This means that under given con- mean square velocity ~: L = 4.2 cm;z = 3.2 msec;P = 50
ditions the solution of the present set of equations kPa, [H2] 0 = 8.5%.
is in favor of the wrinkled model of turbulent
flame [3, 4]. The calculated values of 8nT under There is only one experimental value for the lower
these conditons do not change much and fit satis- boundary UnT at ~ = 190 cm/sec.
factory the experimental values (some disagreement The increase in UnT with pressure drop noted
between experimental values 8nT and 8~-the above does not contradict the known opposite
former is somewhat lower than the latter-has behavior-the decrease-in burning velocity of
probably no significance and is caused by incon- turbulent flame UT.
sistency in the methods used). Here 5nT < L , but One-dimensional calculation gives very roughly
in general it can be not only smaller but larger the turbulent flame velocity UT, usually measured
than L. in experiments. There is a relation between flame
The propagation velocities of the reaction-zone velocity UT and U~T :
surface UnT in normal direction can be obtained
from profiles T and are also presented by the U~TFa
U T --
curves in Fig. 5. No experimental values U~T are F
available. An attempt has been made to determine
the lower boundary of the values UnT presented in where F~ and F are the area of instantaneous
Fig. 5 by points denoted as "min" [12]. These burning surface wrinkled by turbulent fluctuation
lower values are less than the actual ones in the and the area of conventional mean surface of 50%
turbulent:molecular heat conductivities ratio in consumption of burning reactant, respectively. It
the reaction zone. By approximate estimate this could be possible to evaluate F n ~ In 2, l n is the
relation under experimental conditions is several length of line, formed by point T = T1/2 in coordi-
times higher than unity. The calculated propaga- nate plane x - U t (U = mean flow velocity) (Fig. 7).
tion velocities of actual instantaneous reaction When l is the length of line T = 7"1/2, the relation
surface in the normal direction UnT increase with (F~/F) ~-- (ln//) 2 and in the case P = 50 kPa, U =
hydrogen enrichment of the mixture and with 30.6 m/sec, [H2] o = 8.5% is less than 1.1 and,
diminution of pressure and agree with the experi- therefore, u T ~ UnT. That is markedly less than
mental values for its lower boundary. They are the experimental value-for the condition ex-
higher than the laminar flame velocities; the ratio amined, u r ~< 300 cm/sec [12].
of predicted U~T/U~ diminishes with pressure
decrease from 4 to 2. CONCLUSIONS
Figure 6 shows that the increasing of mean-
square-velocity fluctuations ~ results in a higher Comparison of the predicted and experimental
value UnT. From physical considerations such a results shows the possibilities opened up by the
dependence may be inferred as the effect of small- described method of theoretical treatment of
scale turbulence [3, 4]. The value U,T = u~ at ~ = turbulent combustion. Although only the simple
0 is marked on the coordinate axis by circles. Arrhenius reaction has been used in the equations,
TURBULENT-FLAME CALCULATIONS 25

R gas constant, correlation coeffi-

cient
T temperature
3,5 t time
U mean flow velocity
mean-square velocity
u fluctuation velocity components
W chemical reaction rate
x coordinate axis
e0 r cey-- . . . . 6 width o f the flame
molecular conductivity
I I I I I /2 frequency
e q 6 8 tO P density
T time scale
t (Ut)
Fig. 7. Prof'fle "section" 50%-temperature of turbulent Subscripts
flame: ff = 190 em/sec; L = 4.2 em; r = 3.2 msee, P = 50
kPa; [H2] o = 8.5%. 0 initial, side channel
1 initial, central channel
it is not difficult to introduce the quantitative i number of elementary reaction
complex kinetic mechanism. Certainly the one- step
dimensional model of turbulence corresponds to / species
the phenomenon considered very roughly. Descrip- k number of axis
tion o f the large-scale turbulence effect in terms of n laminar or normal to flame
the one-dimensional approach would be worse surface
than that for the small-scale turbulence. T turbulent
Better quantitative results would be expected
with a two-dimensional approach and simulta-
neous use of motion equations. That will be the
purpose o f our studies in future. REFERENCES

1. Kondrat'ev, V. N, Chemical Kinetics o f Gas Reac-

NOMENCLATURE tions, Pergamon Press, New York, 1964, p. 699.
2. Basevich, V. Ya., and Kogarko, S. M., Pyrodynamics
A preexponential factor 4:41 (1966).
b 1 , b 2, b dimensions (Fig. 1) 3. Damkohler, G., Zeitschr. Electrochem. 46:601
(1940).
c heat capacity
4. Shchelkin, K. I.,Zh. Tekhnich. Fis. 13:520 (1943).
D molecular diffusivity 5. Summerfield, M., Reiter, S. H., Kebely, V., and
E activation energy Mascolo, R. W., Jet Propul. 25:377 (1955).
F area 6. Lewis, B., and von Elbe, G., Combustion, Flames and
h reaction heat Explosions o f Gases, Academic Press, New York,
1951, p. 489.
k rate constant
7. Basevich, V. Ya., and Kogarko, S. M, Dokl. Akad.
L space scale Nauk USSR 203:144 (1972).
l length 8. Basevich, V. Ya., Kogarko, S. M., and Dinaburg, E.
m number o f raffling points I., Teoret. Khim. Tekhnol. 10:155 (1976).
N number o f averaging 9. Basevieh, V. Ya., and Kogarko, S. M., Arch. Termo-
dyn. Spalania 7:421 (1976).
n concentration
10. Burgers, I. M., in Advances in Applied Mechanics
P pressure (kPa) (R. Mises and T. Karman, Eds.), Academic Press,
P probability New York, 1948, p. 171.
26 V, Ya. BASEVICH ET AL.

11. Fedorov, R. N., Ed., Biblioteka Programm na FOR- 14. Kozlov, S. N., Basevich, V. Ya., and Kogarko, S. M.,
TRAN'e, ObedinennjiInstitutYadernyeh Issledovanii, Teoret. Khim. Tekhnol. 7:614 (1973).
Dubna, USSR, 1970, Vol. 2, p. 32.
12. Basevieh, V. Ya., and Kogarko, S. M., Arch. Termo-
dyn. Spalania 6:95 (1975).
13. Kramtsov, V. A., Seventh Symposium (International)
on Combustion, Butterworths, London, 1959, p.
609. Received 7April 1978; revised 15 November 1978