Vous êtes sur la page 1sur 20

ELSEVIER

Modeling of Dynamics, Heat Transfer, and Combustion in Two-Phase Turbulent Flows: 1. Isothermal Flows
L. I. Zaichik V. A. Pershukov M. V. Kozelev A. A. Vinberg Institute for High Temperatures of the Russian Academy of Sciences, Moscow, Russia
This paper presents a review of authors' collective works in the field of two-phase flow modeling done in the past few decades. The paper is aimed at the construction of mathematical models for simulation of particle-laden turbulent flows. A kinetic equation was obtained for the probability density function (PDF) of the particle velocity distribution in turbulent flows. The proposed kinetic equation describes both the interaction of particles with turbulent eddies of the carrier phase and particle-particle collisions. This PDF equation is used for the derivation of different schemes describing turbulent momentum transfer in the dispersed particle phase. The turbulent characteristics of the gaseous phase are calculated on the basis of the k-e turbulence model with a modulation effect of particles on the turbulence. The constructed models have been applied to the calculation of various two-phase gas-particle turbulent flows in jets and channels as well as particle deposition in tubes and separators. For validating the theoretical and numerical results, a wide range of comparisons with experimental data from Russian and foreign sources has been done. Elsevier Science Inc., 1997

Keywords: mathematical model, kinetic equation, turbulence, particle, gas, probability density function, modulation effect, jet, channel, deposition, numerical scheme
INTRODUCTION The known methods for the prediction of two-phase turbulent flows can be subdivided into two groups. The first group is related to the combined Eulerian-Lagrangian formulation of medium motion: the equations of motion and energy of the continuous phase are written and solved with the help of Eulerian variables, whereas the equations of the dispersed phase motion and heat transfer are presented and solved by Lagrangian variables (i.e., these equations are integrated along separate particle trajectories). A consideration of the stochastic particle motion within the framework of this approach [1, 2] considerably increases the computational time because of the necessity of having a representative ensemble of realizations to obtain statistically valid information. With decreasing particle size, the number of such realizations is expected to be increased owing to the enlarged contribution of the particle interaction with small turbulent eddies. Problems with this approach also arise for very small particles because of the decreasing time-step size for solving the equation of motion, which results in very long computational times. Moreover, the coupling between the phases at high loading creates a problem in reaching a converged solution. Therefore, the application of the Lagrangian statistical modeling of single particle dynamics seems to be preferable mainly for the relatively inertial particles (ru//T L > 1, where zu is the particle dynamic relaxation time and T L is the Lagrangian time scale of turbulence). The second group of simulation methods deals with the Eulerian continuous representation of governing equations for both phases [3-10]. The two-fluid modeling method allows use of the same types of equations for both phases and, hence, a uniform numerical algorithm for solving the entire system of equations. In addition, the description of dynamics of very small particles does not present principal difficulties, because the limiting transition to turbulent diffusion problem occurs at 7u/T L ~ O. Moreover, the field of the Eulerian-approach application is extended with an increase in the particle volume con-

Address correspondence to Dr. Leonid Zaichik, Institute for High Temperatures of the Russian Academy of Sciences, Krasnokazarmennaya 17a, Moscow 111250, Russia.

Experimental Thermal and Fluid Science 1997; 15:291-310


Elsevier Science Inc., 1997 655 Avenue of the Americas, New York, NY 10010

0894-1777/97/$17.00 PII S0894-1777(97)00009-5

292 L.I. Zaichik et al. centration when the interparticle collisions can play a significant part. The first part of the paper presents the Eulerian models for the description of the momentum transfer in particleladen turbulent flows The basis of these models is the construction of a kinetic probability density function (PDF) equation for the particle velocity distribution in a turbulent flow. The PDF equations for the transport of particles in a turbulent flow were derived by Derevich and Zaichik [11, 12], Reeks [13, 14] and Zaichik [15], but the particle collisions have not been taken into account in these equations. In two-phase flows with a high concentration of particles, interparticle collisions can play an important role and the PDF equation must become the kinetic Boltzmann equation. The models based on the Boltzmann equation for the description of gas-solid flows with high particle concentrations, when the interaction of particles with the turbulent eddies is negligible, had been proposed by Lun et al. [16], Ding and Gidaspov [17], Andersen [18], and other authors However, these models included only the effects of particle-particle collisions and not those of particle-eddy interactions The kinetic equation presented here takes both effects into consideration The interaction of particles with the random turbulent field of the gaseous phase is described by the generalized Fokker-Planck differential operator, whereas the interparticle interaction of the collisions is determined by the Boltzmann integral operator The proposed PDF equation was used to derive Eulerian models for the description and calculation of particle dynamics The simulation of two-phase turbulent flows with large particles is carried out with the help of the equations for the second moments of particle velocity pulsations (or particle fluctuation energy). For small particles, the complete system of equations obtained for the dispersed phase is reduced to a simple diffusion-inertia model. The constructed models have been applied to the calculation of various two-phase particle-laden turbulent flows in jets, channels, pipes, chambers, and so forth. This paper presents a review of authors' collective works made in the field of two-fluid modeling done in the past few decades. A wide range of comparisons with experimental data from Russian and foreign sources has been done. These comparisons play an important role in ensuring that the right direction is taken in our theoretical and numerical investigations The first part of this paper focuses on isothermal two-phase flow modeling, and the second part of the paper deals with two-phase flows with heat transfer and combustion In the first part of the paper, the dispersed phase is assumed to be monodisperse; in the second part, the combustion of a polydispersed system of particles is considered. EQUATION FOR PROBABILITY DENSITY FUNCTION AND ITS MOMENTS The motion of a single heavy particle in a turbulent gas flow is described as ,dRpi
dVpi
ui --

Upi q - F i q - w i -'l- Wii , (2)

dr

where the particle dynamic relaxation time is defined by the relation

.od~
ru 18pv~(Reo) ' for Rep < 10 3, for Rep > 103. (3) 1 + 0 . 1 5 R e '687 0(Rep) = 0.11Rep/6

Here R_ i and u_i are the position and the velocity of the P . Ij particle, r IS time, u i is the actual velocity of the gaseous phase, pp and p are the densities of the dispersed and gaseous phases, d~ is the particle diameter, and u is the . . . . P klnematm viscosity of the gaseous phase. Equation (2) describes the balance of forces acting on the moving particle. The first term on the right-hand side is the aerodynamic interphase force represented in the form of the Stokes drag law. But relation (3) accounts for the fact that the particle relaxation time is a function of the mean relative velocity of the carrier gas flow with respect to the moving particle So Eq. (2) is not restricted to the Stokes drag law but is also satisfactory at l~ge values of the particle Reynolds number, Rep = I U Vld,/v. This form of representation of Eq. (2) implies a line~rization of the aerodynamic interphase force with respect to velocity fluctuations. The second term on the right-hand side of Eq. (2) defines the external force acting on the particle--for example, gravity. The terms w i and Wi take into account a fluid-dynamic interparticle interaction and a particle-particle interaction through collisions; they are represented as continuous and discrete random processes, respectively Thus the existence of the particle velocity fluctuations is caused by entrainment of particles into the turbulent flow field of the gaseous phase; that is, by the interaction between particles and turbulent eddies, as well as by the interparticle interaction To proceed from the dynamic stochastic Eqs. (1) and (2) to the statistical description, the PDF is introduced:

P(T,~,~)= [ 8 ( ~ - / ~ p ) ~(~' - V'p)].

(4)

With a view to deriving a closed kinetic equation for the PDF, the turbulent velocity field of the carrier gaseous phase is simulated by a Gaussian random process with a known autocorrelation function, and the fluid-dynamic particle interaction is assumed to be a Gaussian 8-correlated in time process Furutsu-Novikov's formula holds true for the Gaussian random functions [19]:

{z(.7, r)R[z]}

= ff[z(

} 6 z ( x lrl) ,r)] r,)] ( t}R[z(~,

dfld %, (5)

dr

Upi'

(1)

where z(, r ) is a random process in the space , R[z] is a functional dependent on the random process, and 8R/Sz is a functional derivative. By means of the functional-derivative method using Eqs. (1), (2), (4), and (5), the PDF equation for a particle

Isothermal Two-Phase Turbulent Flows 293 ensemble can be derived [11, 12, 15]: ~+vk--+-09"1" VgXk 0 Uk - - + F ~
Tu

The total stress tensor, Pik in Eq. (11) consists of the kinetic stress, Z(v~v'k), and one due to interparticle collisions, PiCk. The equation for the second moments of the particle velocity fluctuations has the form Z
' ' O(v~v}) ) Oqijk O(uiu]) "4- V k -Jv - Or Ox k dx k

= fU(u'iU'k) 02P "1" u OUiOUk

(
+ gu(U'iU~) +

02P

)OVn _ _O2P

OxiOv--~k + Ox k c)vic)l.) n

O'~k 0 2 p - + J[P]. q'u ~Vi3Uk

= -Pik b
(a) +
oxk -

o~

The left-hand side of Eq. (6) includes the terms describing evolution in time and convection in the phase space (2, ). The terms on the right-hand side respectively characterize the interaction of particles with turbulent eddies of the carrier phase, fluid-dynamic interparticle interaction, and particle-particle interaction due to collisions. The terms caused by particle-eddy interactions have the form of the generalized Fokker-Planck diffusion operator. These terms are similar to those obtained by Reeks [13], with the exception of the term containing the spatial derivative of the particle averaged velocity. Using direct interaction technique, Reeks [14] performed a more exact analysis of the influence of flow nonuniformity on the form of the PDF equation. The coefficients fu and gu in Eq. (6) characterize an entrainment of particles into the gas velocity fluctuations. With the help of a simple step-by-step approximation of the temporal correlation function of gas velocity fluctuations along the particle trajectories, these coefficients can be presented in the form f , = 1 - e x p ( - T u / r u ) , gu = Tu/'ru - 1 + e x p ( - r u / r u ) . (7) The tensor of the fluid-dynamic interparticle interaction is presented in the form % = ~kk(E " E)(Vj -- Vj)/(Uk - VD(Uk - VD, (8) where, in accordance with Koch [20], O'k~ = dp 2rkp 4~,u (Uk - Vk)(Uk - Vk)"
3

2Z

2Z~j (fu(U'iU}) -- (U~U~)) -t- - -- Qij -- Iij" ru % (12)

The terms in this equation present the next physical processes: variation in time, convection, diffusion by reason of kinetic transfer, and particle collisions qijk = Z( c~V'kl) ~) + qicjk,generation caused by the gradient of the averaged particle velocity, exchange of pulsations between particles and gas, generation caused by the fluid-dynamic particle interaction, dissipation due to energy loss by the inelastic interparticle collisions Qij, and exchange of fluctuations between different components of pulsation energy by the interparticle collisions Iij. It should be noted that Eq. (12) is close to the second-moment equation obtained by Simonin [10]. The equation for the third moments of particle velocity pulsations are derived from the PDF equation, assuming that the fourth moments of fluctuations can be presented as the sum of the products of the second moments. This equation has the form

Or

+g k

OX k

+ qijn ~

+ qikn -3X - n o( c~v'k)

+ qjkm-"7-- +

ax~

+ gu(U'iU'~)

OXn

(9)

The last term in Eq. (6) defining the particle-particle interaction caused by collisions is described by Boltzmann's collision operator in Enskog's form [21]. The PDF equation is used to derive a system of equations for averaged characteristics (moments) of the dispersed phase. The conservation equations of mass and momentum are written as

+ ( ~ + gu(U~U'n))O(v~v'k)- + ( ff-~ + gu(U'kU'n)) dxn


- - + 3x n

o(v;v~)

3(v;v~v'~>
"r u + Mij k
= 0.

(13)

oz ozv~ --+--=0, t~"i" t~Xk


z
(ON i ogi ) o.~ + v k - Ox k

(10) OZ gu(U'iU'k)-Ox k

OPik
Ox~

z(u/- E) + Z F i. ~u

(11)

The system of equations presented describes the momentum transfer at the level of equations of the third moments. To simulate the particle momentum transfer at the level of equations of the second moments, it is necessary to derive the algebraic relations for the third moments of pulsations. These algebraic approximations can be obtained from Eq. (13) for the third moments when a contribution of some convection and generation terms is neglected. If the particle collisions also are not taken into consideration in Eq. (13), the algebraic approximation for

294

L.I. Zaichik et al.

the third moment is

,,,

ru[

(v~v)vD = - T ((v~v'.)

+ gu(U'iU'.))--

e( v; v'~>
OX n

N[P]

=--+vk--+

8P dr

OP ~x k

( U k - Vk ) 3P +F k % Ov k

2 2guk ( a2P
-

a(viv'D +((vSv'> + gu<u}u'.)) ax.


]

+ 7--(ko -fok) Ov2 Jru

d2P

aV.
+ -

32P)
. (20)

3
(14)

aXk,9 Vk

aX k a Vk d Vn

-[-((UIkU'n)

-+-

gu(dkU'n))

~x n

I l

Further simplification of the calculation scheme is possible by employing a single differential equation of particle fluctuation energy, kp = (v'kv~)/2 instead of the system of equations for the second moments. This equation is obtained from Eq. (12): 3kp
z

Here, operator T[P] describes the interaction of particles with turbulent gas eddies, and N[P] is the convection operator. The collision operator is presented by the Enskog's approach, J[ P, P] = YJo + J1 + "", where Jo[ P, P] is the Boltzmann collision integral. The ratio between the frequencies of particle collisions in the dense and dilute particle-laden flows, Y, is determined with the help of the correlation by Ogawa et al. [22]: r= [1 - (./**)1/3]-1, (21)

Or

+ vk-;:-!

Okp ~ u.% ]

+ --

3qk c~xk

Pi~

3 Vi 3x k
ru

2Z ru

(Lk

~)
(15)

+ -

ZO'~k

- Q.

In this case, the particle turbulent stress and diffusion flux of fluctuation energy are defined by the following expressions in Boussinesq's gradient form:
2

where the limiting concentration of the dispersed phase is * * = 0.6. A solution of Eq. (18) is obtained with the help of the perturbation technique; that is, it is presented in the form P = P0 + P1 + "", where functions P0 and Pj are determined from the following equations: T[Po] + YJoIPo, Po] = 0, T [ P 1] + Y(Jo[Po, P1] + Jo[PJ,Po])
= N[Po] -

eg.
,,~ ax i
'

ev~
-OX k

)
6ij ,
(16)

(22)

(v;v~) = g k p ~ i j ' '

3X i

Jl[Po, Po]

= L[Po]-

(23)

(vi VkVk) 2

AK 3kp ,
3X i

A solution of Eq. (22) is the well-known Maxwellian velocity distribution: P0 = Z exp 4kp (24)

where the kinetic coefficients of particle viscosity and diffusion of fluctuations are determined as
PK

fuTcl Pt PT Vj + - - , AK Tu -[- %1 /iT + Vj

ATA J

A T + Aj

(17)

Here, PT, AT and vj, A s are the transfer coefficients caused by entrainment of particles into the turbulent gas flow field and interparticle collisions, respectively. For closure of the system of equations presented, it is necessary to determine the terms Piej, qicjk, Qij, Iij, and Mij k caused by interparticle collisions or particle transfer coefficients v K and A K. These terms as well as transfer coefficients may be defined from a solution of the kinetic equation for the PDF of the particle velocity distribution.
SOLUTION OF THE EQUATION FOR PDF

With the help of the Maxwcllian distribution, the particle collision stress, Pi~, J and the diffusion flux of pulsation energy, q/~, can be defined. In the same way, the dissipation due to energy loss by inelastic interparticle collisions Q [16, 17] can be determined as well. These quantities have the form

Pi~ = ~(1 + e ) Z * Y
/ 2kp

(v~v 5) + kp ~ij

~1/2[ ~Vii

+- 3xj

q'-

Ox k ~ij

'

(25)

The kinetic equation is solved by assuming that the fields of gas turbulence and fluid-dynamic interparticle interaction are isotropic; that is, (u'iu'i) = 2k6iJ3, trij = tr~k6ij/3. This equation is presented in an operational form:

q~k = -~(qi~6jk + q~6ik + q~6i),


[ 2kp ]1/2 3kp ] q/C = (1 + e ) Z * Y

-(v~v'kv'

(26)

T[P] + J [ P , P ] = N[P],
where

(18)

Qij = ~Q6ij,

( 2fuk Okk ) 32P 1 3(vk -- Vk) P + , (19) T[P] = 3% + 3% 3v, r. Ov k

(alp4 2kp ) 1/2 c~Vk Q= 2 ( 1 - e 2)Z*gkp[[ .-2-.~ ---

(27)

Isothermal Two-Phase Turbulent Flows The second term of perturbation expansion, P1, is expressed by combination of the particular solutions of Eq. (23) that respectively describe the interaction of particles with the turbulent eddies, PIT, and the particle-particle interaction because of collisions, Pu. These particular solutions are found from the following equations: T[P~T] = L[P0], (28) J0[P0, P1] + Jo[Pl, P0] = L[Po]/Y. (29) Then, if the solutions of Eqs. (28) and (29) have the form P I T = Ax~(vi) and PIJ = Ajq~(v), where A T and Aj are constants and (v,) is a function of vi, the solution of Eq. (23) may be written as

295

The total particle stress and diffusion flux of fluctuation energy are defined as

e,, =

pp

zt-g ) 8,, z.. ox, + oxG oxG8"


Okp Oxi,
(37)

qi =

-ZAp

where Pp, and %, Ap are written in the form 2 Pp = =Zkp[1 + 2(1 + 5

P1

ATAj - A T + Aj ~(vi)

P1TP1J PIT + PIJ "

(30)

e)Y],

(38)

Equation (28) can be solved exactly, but the solution of Eq. (29) is obtained approximately and corresponds to Grad's approach for inelastic collisions of the solid spheres [23]. These solutions are PIJ J re, ) 4kp t (1

4
=-~(1 +

e)rbYdp(2kpl '/2
37r ] '

(39)

Vp= 1 + ~(1 + e ) r P Y

[ 4

] v K+ ~ 3, ]A K + -~. 3

(40)
(41)

X U;U;- T S i j

OX'~j --

9%2/4

Ap = 1 + ~(1 +

[ 6

e)~Y

x~

,o[ 1 + g(1 3

- e)2(2e - 1 ) * Y

]
(31)

The particle transfer coefficients vp and Ap tend to corresponding limiting relations for dilute particle-laden turbulent flows as the particle concentration decreases Tukp 10( guk ) vp0 =fuVt + ---~--,Ap0 = ~-~ 1 + kp rukp" (42) The particle transfer coefficients for dilute particleladen flows are formed because of entrainment of particles into the gas turbulent flow field and are governed by the particle dynamic relaxation time and fluctuation energy. The particle transfer coefficients for particle-laden flows with high particle concentrations also depend on the particle volume concentration, ~, and the parameter A = dp/ru kl/2. The influence of interparticle collisions may be neglected when << A, but the collisions are a dominating phenomenon when qb >> A. Therefore, for the massive (large) particle (when A << 1), the interparticle collisions can play a significant role in both dense and dilute particle-laden flows. The particle shear viscosity coefficient is shown as a function of volume concentration in Fig. 1. As can be observed, for particle-laden turbulent flows with small particle concentrations, the particle viscosity coefficient can both decrease and increase with increasing ~; that is, with increasing interparticle collisions. This effect depends on the value of parameter A. For particle-laden flows with high particle concentrations, the particle transfer coefficient grows with an increase in value of qb for all values of parameter A. EQUATIONS FOR GASEOUS PHASE

x 1+ ~

~V k - - .

The typical times between collisions have the form %1 =

5dp(2r/3kp) 1/2 8(1 + e)(3 - e)~Y'

5dp(2"n'/3kp) 1/2
rc2 = (1 + e)(49 - 33e)~Y"

(32)
Orad's approach allows the derivation of the following simple expressions for Iij and Mijk:

Mijk =

Iiy = -rd - (C;V;) -- -~kpaij , 1 3z [(~;v;v;) -~((v;v'v')ajk


rc2
t t t i i t [

(33)

-[-(V)VnVn)aik -[- (CkVnVn)aij) . (34) J From Eq. (31), the following expressions for the kinetic transfer coefficients caused by entrainment of particles into a turbulent flow field and by interparticle collisions are obtained:
vj

,
%1

1 + ~(1 + e ) ( 3 e -

1)~Y -~-, (35)

{A~} = {10%/271[ J 1+ 3(1 +e)2(2e 1+

_kp.

(36)

The conservation equations of mass and momentum of the continuous gaseous phase may be written for isother-

296

L.I. Zaichik et al.


' ?" I" ' II l|

~o

___::'0: l'

!:' l' l; i,I,/,'i, :'I' I'


o.,, o.,o o.,,

Equations (45) and (46) include the source terms accounting for the turbulence modulation effect caused by the influence of the dispersed phase [25]. The term A k has a similar form as that in the papers of Simonin [10] and He and Simonin [26]. However, Simonin and He applied a differential equation for determination of the gas-particle velocity correlation (u),v~); in Eq. (47), the local relation (u'kv'k ) = 2 f u k is used instead. Coefficients f~, g~ characterize the interaction of particles with the micropulsating turbulent motion of the gaseous phase and are defined as
L=l-e~(-L/%),g~= L/%-L,

(48)

Figure 1. Influence of volume concentration on particle shear viscosity coefficient. mal flows and large Reynolds number in the form a(1 - ~ )
Or

where T, = ( 1 5 v / 8 ) 1/2 is the time microscale of turbulence. The turbulent stresses are given by
Og 2 OUk _ 3

a(1 - ~,)U~
3x k

<u'iu}> = g k a i j - ,,, axj

3xi

(49)

O,

(43)
where the turbulent viscosity coefficient is Ct~
1)t
= 1 -~-

a(1 - ~ ) u ,
3r 3
- -

a O - )u,.uk
3x k
[(1 ~)(u'~u),)]

k2
Ak)/6C I e

(G - e-

(5o)

3X k

(1 - )
p

aP~
3x k

z(u~ - v~)
pro

gu(U'~U'k) o z
p Ox k

+ ( 1 - ~)F~.

(44)

The turbulent characteristics of the continuous phase are described by means of the k - ~ turbulence model based on the transport equations for the turbulence kinetic energy and its rate of dissipation [24]:
-+ uk---

Relation (50) follows as an algebraic approximation from the second-moment equation and accounts for the effects of both turbulence nonequilibrium--that is, G ~ e [27], and particle presence. The time of interaction of particles with the high-energy gas fluctuations of velocity is approximated by

3c.k [
T~ = 2Sct---~11 +

31o:
2k ]

,j2
(51)

Or

Ox~

(1 - ) Ox~
+G e -- A k ,

(1

)--

~k

ovi G = - (u;u'k) axk


~8 OT 08 3X k

(45)

The turbulent constants are of the following values: C~ = 0.09, C 1 = 2.0, ~k = 1.0, tr~ = 1.3, C~1 = 1.44, and C~2 = 1.92. BOUNDARY C O N D I T I O N S Boundary conditions for the dispersed phase velocities are defined for the elastic interaction of particles and a wall in the normal direction, with allowance both for possible particle deposition and for a momentum loss in the longitudinal direction. Therefore, in this case, the probability density relating the particle velocity transition from a situation before a collision with the surface (index 1) to a situation after a collision with the surface (index 2) is determined as

1 (1 - )

O [(1 6~'9 ] ex~ ) v ~ 7-~x~ ]


8

7Pt

+ ( C , 1 G - Ce2s) ~ - A , ,

(46)

Ak}

~u

(1 - L ) ~

gu ( U'i U tk > dX k

try(2/1)
"

= X6(Ux2 -

(~vUxl)~(Uy 2 + Uyl) ,

(52)

aZ

(47)

~g~-aTx,

where X, the reflection coefficient, is equal to the probability of particle rebound from the wall and return into the flow after impact and ~b v is the coefficient of restitution of

Isothermal Two-Phase Turbulent Flows the longitudinal velocity component by particle-wall collision. For an ideally smooth wall, the restitution coefficient is equal to unity; for a very rough wall, we may assume thv = 5/7. The distribution of the normal velocity component in the near-wall space is assumed to be quasinormal:

297

presence of the dispersed phase can be determined with fair accuracy from the same relation as that for singlephase turbulent flows [28]:

k = u 2* C p~ -l/z

(58)

where u . is the friction velocity. In the logarithmic layer, the dissipation rate is determined by the expression

,~ = C3/4k3/2/~ey, P(Vy > O) = xP(Vy


< 0). (53)

(59)

With the help of Eqs. (52) and (53), we can find the following boundary conditions for the particle velocities:

I -- X ( ,2(U'y2) - - ) 1/2
Vy I+X ~"

(54)

where, for single-phase flows, ~e = 0.4. If we assume that the length scale of turbulence is not much affected by particles, ~e may be taken to be equal to the PrandtlKarman constant in two-phase flows, too. The friction velocity can be found from the universal logarithmic velocity profile in the near-wall region:

u+= In y+/~e + A,

(60)

l+x4,v
--

a+x
UxUy) + .

Vx
(55)

where u + = Ux/U, and y + = y u , / v. The applicability of the wall law [Eq. (60)] with the same values of the constant ~e and A as in single-phase flows was experimentally confirmed for gas-particle flows by Kulick et al. [29] and other authors. DIFFUSION-INERTIA MODEL For the small particles that have a parameter of inertia, zu/TL, below unity, a diffusion-inertia model obtained from a complete system of equations is proposed. In this approximation, the transport of particles is described by a single equation of concentration:

Expressions (54) and (55) determine the boundary conditions for conservation equations of mass and momentum. The boundary conditions for equations of the particle fluctuating velocity second moments are proposed with an assumption that the diffusion fluxes of these moments through the wall are absent. In this case, it can be written
as

O(v~v~) Oy

Okp Oy

O.

(56)

--+-Z Ui + "r u Fi 0"1" OXi


-

OT OqDik ) . Ox k

Uk OXk
(61)

According to Eq. (56), the pulsation energy of the dispersed phase may be nonzero on the wall in spite of the zero value of the turbulent energy of the gaseous phase. This effect is conditioned by the transport of pulsations from the turbulent region of the flow because of particle inertia and can play a significant role for massive particles. As boundary conditions for the equations of average motion of the gaseous phase [Eqs. (43) and (44)], no-slip conditions are prescribed on the walls: U/= 0 for y = 0. (57)

0( OZ - OXi Dig Ox k

Here Dij = Tu(u'iu ~) is the diffusion tensor, and q = Zufu/T u is the parameter of turbulent migration of particles. For the very small particles (zu -~ 0), the concentration equation [Eq. (61)] becomes a usual diffusion equation:
- -

0"I"

OXi

- -

OXi

Oik

(62)

For determining the fluctuation characteristics of the gas flow near the wall, it is worth using the wall function technique that has gained wide acceptance in single-phase flow calculation [24]. Introducing wall functions considerably reduces the number of computational grid points in the direction normal to the wall. In this case, the boundary conditions for Eqs. (45) and (46) are prescribed not on the wall itself but at some distance from it, outside the viscous sublayer in the so-called logarithmic layer. The basis of the wall function method requires the assumption of constant turbulent shear stress (U'xU',) and turbulent energy k, as well as universality of tl~e profile of the streamwise velocity Ux in the near-wall region, where the first calculation grid point for Eqs. (45) and (46) is positioned. Analysis of the system equations for the second moments of the gaseous phase velocity pulsations shows that the turbulence energy in the logarithmic layer in the

In comparison with a usual diffusion model, the diffusion-inertia model allows us to take into account a number of effects caused by the particle inertia: interphase velocity slip stimulated by the gravity force or other external force, turbulent migration due to gradient of turbulent energy, and inertial transfer by reason of deviation of the particle movement from the gas streamlines. The boundary condition for Eq. (54) has the form

----g-OqDyy + Uy + 7uFy]Z. ay
(63)

As a simplification of the calculation scheme, the boundary condition for the particle concentration equa-

298 L.I. Zaichik et al. tion [Eq. (61)] in the form of wall functions also is proposed. This approximation connects the particle deposition flux, Gyw = VewZw with the particle concentration in the near-wall region Z n and is presented on the fully absorbing wall ( X = 0) as rain(2.5 X 10-%2'5; 0.25)u, Gyw = Zn max[0.61; rain(1.32 - 0.271n ~-+; 1)] Chen and Wood [4], Mostafa and Mongia [5], Risk and Elghobashi [6], Derevich et al. [7], Shraiber et al. [9], and other authors for the prediction of the dispersed phase stress tensor may lead to considerable errors for sufficiently inertial particles Consider a solution of Eq. (12) for the particle normal pulsation intensity <Uy2>in the near-wall turbulent flow. As an approximation of the fluctuation structure of the carrier flow, the simplest two-layer model consisting of a viscous sublayer with zero pulsation intensity and a turbulent region with constant pulsation intensity is taken; that is, (u~) = 0 for 0 < y < 6,, <u~2) = u2, for y > 6,. (66) Figure 2 presents the distribution of the particle pulsation intensity in the near-wall region The dependence of the pulsation intensity at the wall on the parameter T. = r u U . / 6 1 characterizing the particle inertia is given in Fig. 3. It is important to emphasize that the pulsation intensity of the fine particles in the viscous sublayer is zero, whereas the velocity fluctuations of the inertial particles in the viscous sublayer and at the wall itself are nonzero The presence of velocity fluctuations of the dispersed phase in the viscous sublayer could be obtained only on the basis of the nonlocal theory and is explained by the diffusion mechanism of transport of pulsations from the turbulent region of the flow to the wall, owing to particle inertia. The decrease in the pulsation intensity at the wall with increasing ~'. for the large particles is connected to a reduction in the response of the particles to gas velocity fluctuations. Straight Round Jet In free jet flows, the turbulent characteristics are essentially closer to an isotropic state than in near-wall flows Therefore, turbulent stresses in the dispersed phase can be simulated with the help of the particle turbulent energy equation [Eq. (15)] instead of the system of Eqs. (12) for the second moments With the use of Eqs. (10), (11), (15), and (43)-(46), the calculations of two-phase turbulent jets for a wide range of particle diameter d_ and mass loading ratio m 0 have been carried out. The predicted results have been compared with experimental data of Modarress et al. [31] and Tsuji et al. [32].
P

-G-ruG},
M E T H O D OF NUMERICAL SIMULATION

(64)

where T+= "ruU2,/P is the dimensionless particle relaxation time. Formula (64) is valid for the relatively small particles (for T+ < 100) when choosing concentration Z n in the range of distance variance from the wall 30 < y+ < 100, where Z varies slightly with the variation of y.

The finite volume method based on integral conservation laws is used for a numerical solution of the equation system. The pressure field is determined by means of the SIMPLE procedure consisting of a global iteration process for solving the pressure correction equation [30]. Each equation can be written as a generalized differential elliptical one:
-+ -Fik + S,

(65)

Or

Ox i

Ox i

where q~ is a generalized dependent variable, W/ is the velocity of a gaseous or particle phase, Fzk is a generalized diffusivity, and S is a source term. To obtain a discrete analog of Eq. (65), the flow field is covered by a Cartesian grid, and Eq. (65) is integrated over each control volume. Discretization of the diffusion fluxes in Eq. (65) is performed with the help of the central-differences scheme. The convective fluxes are approximated by using a hybrid scheme that combines central and upwind-differences schemes. The pressure correction equation has the form of a Poisson equation. The number of grid nodes is 50 50 (or, in some cases 100 100) for two-dimensional calculations and 40 40 60 for three-dimensional calculations. P R E D I C T I O N EXAMPLES The proposed models have been used for predictions of various two-phase particle-laden turbulent flows in jets, pipes, channels, and combustion chambers. Near-Wall Region with Viscous Sublayer Within the framework of the Eulerian two-fluid approach, an accurate simulation of the dispersed phase characteristics in the near-wall region, including the viscous sublayer, could be conducted only on the basis of the secondmoment models. Such models allow a description of the near-wall turbulent flow that takes into account the nonlocal effects of momentum and heat transfer by inertial particles. The application of algebraic (local equilibrium) Eulerian models used by Pourahmadi and Humphrey [3],

0.4

?
I

~'.=0.5

0.2

10.

YI61

Figure 2. Distributions of the particle pulsation intensity in

the near-wall region

Isothermal Two-Phase Turbulent Flows 299


I I

ult , v,

_
t~.'~.

,,2
0.05

Im~._~. ~a

~.a

A,o,

o,-----

tJ I'-- "-',~o.

-I

,0LS
o_ r
5 lO ~',,
L 0 0.05 ~0.10 ~o..

]
1
rht

Figure 3. Particle pulsation intensity at the wall. For boundary conditions, on the jet axis, we use the flow symmetry requirement and, at the outer edge, the equality of the averaged characteristics of the carrier and dispersed phases to the parameters of the external flow. The inlet conditions are determined either on the basis of experimental data or by using the numerical solution for the mean and fluctuating characteristics of the carrier phase in the fully developed pipe flow and the assumption that the distribution of the dispersed phase characteristics is uniform. In Fig. 4, the calculations are compared with the experimental data of Tsuji et al. [32] on the decay in the axial velocities of the gas and particles. It is seen that the presence of the dispersed phase leads to an increase in the range of the jet in comparison with a single-phase jet. This effect is explained by the increase in the main stream total momentum and flow laminarization because of the dispersed phase presence. But, in the large-particle case dp = 500 /xm, the influence of the dispersed phase on the parameters of the gaseous phase is not very pronounced. It should be noted that, in the coarse dispersed phase case, with increasing particle size, there is also a decrease in the reaction of particles on the mean and fluctuating characteristics of the carrier phase. Because of the particle inertia, the decay in the longitudinal velocities of the dispersed phase proceeds more slowly than does the decay in the axial gas velocities. The calculated and measured results of radial distributions of the axial gaseous and dispersed phases velocities are illustrated in Fig. 5. The most important effect is a decrease in the jet width associated with the increase in the range that was found both in predictions and in

Figure 5. Comparison of radial distributions of (1) axial gas and (2) particle velocities in a straight jet with the experimental data of Modarress et al. [31] at x / D o = 20. (a) rn 0 = 0.32, (b) m 0 = 0.86, (c) m 0 = 0; (a,b) dp = 46 p.m. experiments. The predictions of longitudinal fluctuations of the continuous and dispersed phases are demonstrated in Fig. 6. A remarkable effect of the dispersed phase on flow pulsation structure can be seen. With increasing particle mass loading, the fluctuations of both gaseous and dispersed phases are suppressed. This phenomenon is in agreement with experimental data by Modarress et al. [31] and other authors. Figure 7 presents the variations in concentration along the jet axis for relatively small particles that were used in experiments by Laats and Frishman [33, 34]. The simulation was carried out on the basis of the diffusion-inertia model [Eq. (61)]. It is easy to see from Fig. 7 that a abnormal rise of particle concentration along the jet axis can take place. The implemented calculations are suggesting that this abnormal rise effect is connected to a phenomenon of turbulent migration of particles in a nonuniform fluctuation flow field--namely, caused by transport of particles from a high-level turbulence energy region into a low-level turbulence energy region. Swirling Round Jet A theoretical investigation flows was conducted under in experiments by Bulzan swirling two-phase jet was of particle-laden swirling jet conditions that were realized et al. [35]. The axisymmetric injected vertically downward

Uo ' Uo
|,.~d" " "-o'" ~ "o~o---. "o--.

iii 'v" I oe ~ ' I C I I

10

2o

X/D o

o.oJ

o.1

r~

Figure 4. Comparison of axial distributions of (1) axial particle and (2) gas velocities in a straight jet with the experimental data of Tsuji et al. [32]. (a,b) m 0 = 0.86, (c) m 0 = 0; (a) dp = 170 p.m, (b) dp = 500 p.m.

Figure 6. Comparison of radial distributions of (1) particle and (2) gas fluctuation intensities in a straight jet with the experimental data of Modarress et al. [31] at x / D o = 20. (a) m 0 = 0.32, (b) m 0 = 0.86, (c) m 0 = 0; (a, b) dp = 46 p.m.

300 L.I. Zaichik et al.


(~m-e' ! i
e ! B-I A-2

Oxs.,um

o.gF ~ ~ ~ , . , . .

"

1.0 i o-3

"I
I

0.,s

1o

2o

X/Do

Figure 7. Comparison of axial distributions of particle concentration in a straight jet with the experimental date of Laats and Frishman [33, 34]. (1, 2) m 0 = 0.22, (3) m 0 = 0; (1) dp = 7/zm, (2) dp = 17/zm.

0.I

0.2

r/x

from a nozzle of diameter D o = 3 . 8 . 1 0 2 m, with the swirl number S ranging from 0 to 0.33 and a loading ratio of m 0 = 0.2 with particles of 39-/xm diameter. The calculation model is the same as that used in the preceding section. On the jet axis, symmetry conditions were imposed; whereas, on the outer boundary, the characteristics of the carrier and dispersed phases were assumed to be equal to those in the oncoming flow. As inlet conditions, the experimental data or uniform distributions of the characteristics of the carrier and dispersed phases were used. The initial circumferential velocity was assumed to be described by rigid-body rotation: In Fig. 8, the axial velocity of the carrier phase is seen to decay more quickly for greater values of the initial swirl. Moreover, in Fig. 8, the decay of the axial gas velocity along the jet axis by the addition of particles is slower than that in a single-phase swirling jet. This result is the same as that in a straight nonswirling two-phase jet and is likewise connected to the increase in the stream m o m e n t u m and the effect of turbulence attenuation caused by particles. Figures 9 and 10 show the predicted and measured profiles of the axial particle flux, G x = V ~ Z , normalized by the maximum value in the same cross section for x / D o = 5. The maximum of the particle flux was found to shift

Figure 9. Comparison of radial distributions of axial particle mass flux in a swirling jet with the experimental data of Bulzan et al. [35] at x / D o = 5. (1) Sg = 0 (Sp = 0), (2) 0.16 (0.08), (3) 0.3 (0.12). outward from the center of the jet with an increase in the swirl number. This shift is caused by the angular and radial inertia of particles. Figure 10 presents the calculated profiles of the axial particle flux obtained by Bulzan et al. [35] with the help of three widely disseminated models: Lagrangian deterministic separated flow (DSF) analysis, Eulerian locally homogeneous flow (LHF) analysis (i.e., a usual diffusion model), and Lagrangian stochastic separated flow (SSF) analysis. As can be observed from Fig. 10, the proposed model is in better agreement with experimental data than are the other models. Figure 11 shows the radial distributions of the particle fluctuation energy. Of significance is the predicted decrease in the particle fluctuation energy with increasing swirl number, corresponding to experimental data. Flow in Pipe In this section, the behavior of rather massive particles in an upward, fully developed pipe flow is considered. As suggested in a set of previous studies [29, 36, 37], the profiles of the axial velocity and fluctuation intensities of
OI

Oxn~
~-1 ,l,. - 2 v-3

1.0

o-4
e-5

'~/

/ /

\ \

0.,S -

/1
2

0.5

1o

20

x/~

o.1

o~

Figure 8. Comparison of axial distributions of axial gas velocities in a swirling jet with the experimental data of Bulzan et al. [351. (1) S e = 0 (Sp = 0), (2) 0.16 (0.08), (3) 0.3 (0.12), (4) = Sg0" 0.19, (5) 0.3; (1-3) dp 39 ~m, m o 0.2; (4, 5)
m o

Figure 10. Comparison between calculated and measured radial distributions of axial mass flux in a swirling jet for S~ = 0.3 (S o = 0.12) at x / D o - 5. (1) LHF model, (2) DSF niode|, (3) SSF model, (4) authors' model. Symbols are from the experiment of Bulzan et al. [35].

Isothermal Two-Phase Turbulent Flows 301


V|

0.7J : . _ _ _

z.._ % 7
0 0.l 0.2 r/x

Figure 11. Comparison of radial distributions of particle fluctuation energy in a swirling jet with the experimental data of Bulzan et al. [35] at x/D o=5. (1) S g = 0 ( S p = 0 ) , (2) 0.16 (0.08), (3) 0.3 (0.12).

0.2~

03

0.7~

['/R

the large particles in pipes and channels are becoming relatively flat owing to intensive transverse mixing (i.e., these distributions in the pipe cross section for massive particles are virtually uniform). Taking account of this fact, we can carry out a theoretical analysis for large particles on the basis of an asymptotic solution of conservation equations for the dispersed phase. In this approach, we may use second-moment Eqs. (12) in so-called algebraic form when the diffusion terms are neglected. It should be mentioned that the influence of parameter e on the particle characteristics under the considered conditions is weak, and the profiles of particle velocity and pulsation energy, even when e = 0 and when e = 1 are close. On the contrary, as can be seen from Figs. 12-14, the effect of p a r a m e t e r ~b v is considerable. Moreover, it should be emphasized that the uniform profiles of the particle velocity and pulsation energy do not testify to any deficiency or any feature of the proposed model but are the result of the asymptotic approach for large particles. The results of predictions of particle velocity are compared with the experimental data of Laats and Mulgi [38], Lee and Durst [36], and Tsuji et al. [39] in Figs. 12, 13, and 14, respectively, for two values of the restitution coefficient for particle-wall collisions ~v = 5 / 7 and ~bv = 1. In Figs. 15 and 16, the prediction results of the axial and radial components of the particle pulsation energy for the conditions realized in the experimental investigation of Lee and Durst [36] are presented. As can be observed from Figs. 15 and 16, the most important effects in this case are the considerable difference between gas and particle velocities due to gravita-

Figure 13. Comparison of axial particlevelocity profilesin a vertical pipe with the experimental data of Lee and Durst [36]. (1) dp = 100 /zm, (2) dp = 200 /xm, (3) dp = 400 /zm, (4) dp = 800 /zm; (dashed line) ~b v = 1, (solid line) ~b v = 5/7.
tional force and to loss of particle m o m e n t u m by collisions with the wall (at ~b v = 5 / 7 ) and the quite high values of the particle pulsation energy that are induced by the large difference in interphase velocities. It is interesting to note (see Fig. 16) that the dependence of the pulsation energy on particle size is not a monotonic function. The diminution of pulsation energy is caused by the decrease in particle entrainment into turbulent gas pulsations with the increase in particle size. The increase in the particle pulsation energy is induced by the increase in the interphase velocity difference. For very massive particles (Rep > 103), when e = 1, the pulsation energy does not depend on particle size and can be determined by the correlation
..pk1/2 = 0 . 3 ( P l O p )
1/3

[ Ux -- Vx[.

(67)

Particle Deposition in Pipe The simulation of particle (or droplet) deposition from a turbulent stream in round pipes was performed with the help of Eqs. (12) for the second moments of velocity
Vx Uxsuz

~i~i -

0.7.~

Vz
Uzmex

0.75

03
l 2

LX

~ - - / ~ -

O- 1

2"/0 0
I

o21

-"~"- ~ -zxI 0.J I 0.75

0
I

0"/0
t

0
I

O..S

.,.

025

03

0.75

dR

I 0.25

r/R

Figure 12. Comparison of axial particle velocity profiles in a horizontal pipe with the experimental data of Laats and Mulgi [38]. (1) d p = 4 4 / ~ m , ( 2 0 d p = 8 8 p . m ; (dashed line) ~b v = 1, (solid line) thv = 5/7.

Figure 14. Comparison of axial particle velocity profiles in a vertical pipe with the experimental data of Tsuji et al. [39]. (1) dp = 200 /zm, (2) d o = 500 /zm, (3) dp = 3000 /~m; (dashed line) ~b v = 1, (solid line) ~bv = 5/7.

302 L.I. Zaichik et al.

<vz> t/z
Ux~
D

J+
I ~

00

0.~

000

~ 1 0

e-1 o-4 x-2 v-5 Q.3 A-6

"

'

02J

OJ

0.75

fiR

10-2 I0"1 Figure 17. (1) Forney Ganic and (5) Lee et (70).

100

.to

Figure 15. Predicted and measured profiles of particle axial fluctuation intensity for d_ = 800 /xm Line, calculation; U symbols, experiment of Lee and Durst [36]. pulsations The two-phase flows in the case of low concentrations of the dispersed phase when the effect of particles on the gaseous phase characteristics may be neglected is considered. Therefore, the turbulent characteristics of gas are determined by knowing analytical approximations for a single-phase flow [40]. The particles are assumed to absorb on the pipe walls after collisions. In accordance with this assumption, the coefficient of particle reflection X in boundary conditions (54) and (55) is taken to be zero The results of analytical and numerical solutions are presented in the form of widespread dependence of the deposition coefficient j + = Gyw/U. Z m on the dimensionless particle relaxation time ~-+. The deposition process of the fine particles is governed by Brownian and turbulent diffusion, and the deposition rate is determined in this case as j + = 0.ll5SCB 3/4 = Ol15B-3/4'r+ 3/8, (68) where Sc n = u/D n is the Brownian Schmidt number, D B is the Brownian diffusivity, and B = SCB~'T1/2 is the Brownian diffusion parameter With an increase in particle size, the effect of the turbulent migration of particles in a inhomogeneous pulsation field on the deposition also increases The prediction results in the range 1 < ~'+ < 10 can be approximated by the formula j + = 2.5..r~--4 25 Iv ~'+'. (69) For the relatively large particles (~-+>> 1), the following analytical relation is obtained: 0.25 J+ 1 + 144~-01/2" (70) Here ~'0 = 2~'uu./D is the dimensionless particle relaxation time Figure 17 compares expression (70) with experimental data obtained for the deposition of the large particles.
I

Deposition rate of large particles. Experiments: and Spielman [41], (2) Liu and Agarwal [42], (3) Mastanaiah [43], (4) El-Kassaby and Ganic [44], al. [45], (6) Binder and Hanratty [46]. Line, Eq.

On the basis of Eqs. (68)-(70), a generalized approximate formula for the coefficient of deposition on the fully absorbing wall is proposed: 0115B J+= 1 +
. c,-3 Iv 3 / 4 , / . + 3 / 8 ~_ 25 ~-+'

25
10

113""~-4T+'25

+ 1.44

3'r+3R+l/2'

(71)

where R+= R u , / ~ , is the dimensionless pipe radius. A comparison of the results of calculations made with the help of approximation (71) with the experimental data collected by McCoy and Hanratty [47] is presented in Fig. 18. The initial decrease in j+ with the increase in r+ is explained by the diminution of the Brownian diffusion coefficient with increasing particle inertia according to Eq. (68). The growth of the deposition rate is induced by the turbulent migration mechanism of the particle transport from the core flow to the wall. The subsequent reduction in the deposition rate for the large particles is connected to the decrease in response of the particles to velocity turbulent fluctuations of the carrier flow.

J+

g,

.6

10.2

o *
o 0 o

<v~z> t/z O~

<,,;2> xa
I0"a
I I I

i0

102

104 ~+

3.SO

600

g.sO dp,/~m

Figure 16. Effect of particle size on particle axial and radial fluctuation intensities.

Figure 18. Comparison of formula (71) with experimental data collected by M c C o y and Hanratty [47].(I) B = 105, (2) B = 3.105, (3) B = 10 6, (4) R+= 10 3, (5) R + = 10 4, (6) R+= 105.

Isothermal Two-Phase Turbulent Flows 303 Figures 17 and 18 show, rather significantly, deviations of the experimental data from Eqs. (70) and (71). This fact testifies that the deposition rate is not dependent only on r or %. As suggested by Johansen [48], one of the reasons for this phenomenon can be connected to the effect of transversal lift force acting on the moving particle in a shear flow.
V+~
09
m

0.7

Channel with a Backward-Facing Step Computations of a particle-laden flow in a channel with a backward-facing step were conducted under the same conditions as those realized in an experimental investigation by Ruck and Makiola [49]. Four different sizes of particles (dp = 1, 15, 30, and 70/zm) are chosen for these computations. Calculations are carried out on the basis of Eqs. (10), (11), and (15) for the dispersed phase and Eqs. (43)-(46) for the continuous phase. The wall boundary conditions for the dispersed phase equations are specified according to Eqs. (54)-(56), with X = 1 and ~b v = 5/7; for the gaseous phase, Eqs. (57)-(60) are used. Uniform profiles of all computational variables are given at the inlet cross section. The particle velocity profiles for three particle sizes are depicted in Fig. 19, in which it is easy to see that the entrainment of particles into the recirculating motion of the gaseous phase diminishes with increasing particles size. The particles of dp = 70 /xm are not entrained in this motion, and the profiles of velocity become smooth faster. On the whole, the general flow pattern corresponds to the pattern that was observed in the experiments by Ruck and Makiola [49]. For a quantitative comparison of prediction results with the experimental data, we analyzed the variation of the maximum positive streamwise velocity of particles along the channel and the maximum negative velocity of particles in the recirculating zone. These results are presented in Figs. 20 and 21, respectively, where Xstep is the step length and x r is the reattachment length ( x r -xstep = 200 mm). It can be seen in Fig. 21 that the
~-3 0
I

0.5

X-Xl~ Xr-X~

Figure 20. Maximum positive axial velocity of particles in a channel with a backward-facing step. Lines, calculation; symbols, experiment by Ruck and Makiola [49]. (1) dp = 1 /zm, (2) dp = 15 /zm, (3) d p = 30 p~m.

position of the maximum negative velocity is shifted toward the step (upstream) compared with the experimental data. Straight Flow in a C h a m b e r with Expansion In this section, we examine the straight flow in a cylindrical chamber with a sudden expansion. Computations are performed for the same conditions as those in the experiments of Hahn and Sohn [50]. The primary gas-particle flow is injected through the central nozzle into the nearaxis region of the chamber and mixed with the secondary air flow injected through the annular split. Numerical results are obtained on the basis of the model described in the preceding subsection. Uniform profiles of all flow characteristics of the primary and secondary streams at the inlet sections have been imposed. Figure 22 presents the streamlines and the velocity profiles of the gas and dispersed phases. The expansion of the chamber section is the reason for the recirculating zone generation. The small particles (dp < 10 /~m) are

U0 250

I1-1 0-2
~-3

!o

m
o/

//~-~
~\

25" 0 0.1

250

0.5

xr-x xr-x~p

z~o

4~o

6/~o

8~o

x, mm

Figure 19. Profiles of particle axial velocity in a channel with a backward-facing step. (1) dp = 1 p~m, (2) d p = 30 /xm, (3) d o = 70 p~m.

Figure 21. Maximum negative axial velocity of particles in a channel with a backward-facing step. Lines, calculation; symbols, experiment by Ruck and Makiola [49]. (1) dp = ] /.~m, (2) d p = 15 /zm, (3) dp = 30 /zm.

304

L.I. Zaichik et al.

otr, mm ~. . . . . . . . . . . . . . .
- - - - - . . . . . . . .

o.,I I I I I I I ~ I I I I I I I I I I I I I I I / I I I / ~ I I I A ~ I I I I I I . . . . . . . . . . . . I ~

I
. o

25m/~

r~ mlal

v/llllllll~lrlflflll

100 5O o
5O-

L
................ ;...
2o0

Figure 22. Streamlines and axial velocities for straight flow in a chamber with expansion: (a) gas, (b) particles of dp = 40 /zm, (c) particles of dp = 100 /zm.

-100

,
4oo

,
6oo

L
$oo 1 -

XI mill

lnGC~
0-2 A-3 -2

turbulence attenutation, the mixing process between the primary and secondary flows is reduced. This phenomenon is illustrated in Fig. 23, in which the premixing function is determined with the help of the expression
f = (G x Ge)/(G

G2) ,

(72)

-d

2.0

3.s

~a(2x/D)

Figure 23. Variation of premixing function along the chamber axis. Lines, calculation; symbols, experiment by Hahn and Sohn [50]. (1) one-phase flow; (2, 3) two-phase flow for m 0 = 0.2, (2) d p = 19/zm, (3) d o = 40 /zm.

where G t and G 2 are the primary and secondary mass particle fluxes. As in Fig. 23, an increase in particle size reduces the mixing. A similar result concerning the decrease in the mixing process with increasing particle size was obtained in the experiments of Hahn and Sohn [50]. Swirling Flow in a C h a m b e r with Expansion

virtually completely entrained into the gas motion, so the regions of the negative velocities of the gas and particles coincide. With increasing particle size, the entrainment of the dispersed phase into the recirculating zone decreases and the size of the negative velocity of particles diminishes. As mentioned earlier, the dispersed phase promotes the laminarization of turbulent flows. As a consequence of

A swirling of the primary or secondary flow in combustion chambers is often used to stabilize the burning process due to the formation of a central recirculating zone. The simulation of a two-phase turbulent swirling flow in a chamber with a sudden expansion was carried out under the same conditions as those of the experimental investigation by Sommerfeld and Qiu [51]. The primary nonswirling gas-particle flow enters the chamber through a

~p

mill
50

-511

Figure 24. Streamlines for swirling flow in a chamber with expansion.

ao

1so

24o

32o

x, mm

Isothermal Two-Phase Turbulent Flows 305


0
I

20
I

0
t

15
I

0
I

5
I

0
I

50
I

5
I

0
I

~d'

"

"

ry
315
4
I

o -50~ 0
0
I

Y~ 15
I

112
5
I

155
4
I

195
4 0
I I

X, Into

Figure 25. Comparison of distributions of (a) gas and (b) particle axial velocities (m/s) with the experimental data of Sommerfeld and Qiu [51] for d o = 30 /zm.

0
1

4
I

Y,

$0

.;... a
|

-50

52

112

155

195

315

X~ t a m

Figure 26. Comparison of distributions of (a) gas and (b) particle tangential velocities (m/s) with the experimental data of Sommerfeld and Qiu [15] f o r d p = 30 /zm.

I ~J

o4,

o4,

o, I

mitt

O
50

I/52 112 155 195 315 X,

-50

mm

Figure 27. Comparison of distributions of (a) gas and (b) particle root-mean-square fluctuating velocity (m/s) with the experimental data of Sommerfeld and Qiu [51] for d p = 30 /xm.

4> ,I,,

/
0.5
I I i I

0.4

o.s

12

x/D

Figure 28. Distributions of particle concentration along the flow axis. (1) d p = 30 /xm, (2) dp = 45 /xm, (3) d o = 60 /xm.

central tube. The particle mass loading ratio in the primary flow is rn 0 = 0.034, so the influence of the dispersed phase on the gas flow is virtually absent. The secondary swirling airstream enters through a coaxial annulus. The swirl number based on the total inflow is 0.47. Computations were carried out on the basis of a mathematical model similar to that given in the preceding subsection. The tangential velocity of the secondary flow in the inlet section is assumed to correspond to the rotation law of a solid body. Flow streamlines a r e presented in Fig. 24. The calculated size of the central recirculating zone deviates noticeably from experimental data. The zone of the particle reverse flow differs negligibly from the central recirculating zone of the gaseous phase (Fig. 25). A comparison of the calculated and experimental data on the averaged gas and particle axial velocities is presented in Fig. 25. The same comparison for the tangential velocities is depicted in Fig. 26. The profiles of the gas and particle root-meansquare fluctuating velocity are shown in Fig. 27. Figure 28

306

L.I. Zaichik et al.

IOO
50

O'

-50 '

Figure 29. Isolines of droplet concentration (~/qb 0) in separator with conical central body. (a) d p = 1 /zm, (b) d p = 3 /xm.

-100'

200

400

600

800

X, mm

deviation of the inlet conditions in our calculations from those realized in the experiments.
0.4

0.3

Separation of Droplets

0.2

0.1

4 dp, ptm

Figure 30. Effect of droplet size on separation efficiency. (1) diffusion-inertia model, (2) Lagrangian model, (3) experimental data of Johansen and Anderson [53].

presents the variation in the particle concentration related to the inlet along the chamber axis. It can be seen that the motion of the particles is characterized by the increase in their concentration in the central recirculating zone. However, with increasing size, the particle entrainment into the reverse flow decreases. The implemented comparison shows some differences between calculation and measurement. A reason for these discrepancies is the necessity for modification of the k - e turbulence model for predicting vortex swirling flows. A possible way of improving the calculation results is the use of the second-moment model for the description of the gaseous and dispersed phases instead of the models based on the pulsation energy. On the other hand, the results obtained by Sommerfeld et al. [52] on the basis of the k - e turbulence model are in somewhat better agreement with the experimental data considered. So, apparently, these discrepancies may also be connected to some

One of the biggest problems in modeling the water-steam separation process is the description of the motion of fine aerosols and their deposition on the surface of separators. Here, the diffusion-inertia model [Eq. (61)] is used for the calculation of motion and deposition of aerosols in the vortex separators. The separator is presented as a cylindrical pipe with a conical central body. The angle of the central conical body is determined from the condition in which a separated flow is absent in the core part of the flow. In the pipe inlet, a swirling flow is formed. The results of the calculations of droplet concentration normalized by the inlet droplet concentration are presented in Fig. 29, in which the deposition rate (or separation effect) can be s e e n t o grow with increased size of the droplets. Figure 30 compares the calculated and experimental data of Johansen and Anderson [53] on the separation efficiency for the annular centrifugal separator. Also represented is the predicted dependence derived on the basis of a deterministic Lagrangian model by Johansen and Anderson [53]. In this case, the difference between various models is not considerable. The main advantage of the proposed diffusion-inertia model is the use of a uniform calculation algorithm for the simulation of both gaseous and dispersed phases.

Concentrator of Abrasive Particles

The same model as that used for a channel with a backward-facing step was used for the simulation of the abrasive particle motion in an apparatus for cutting the fine

Figure 31. Scheme of the cutting apparatus.

I:

3.5 mm

~.l.t

4.0 ram

,I

308 L.I. Zaichik et al. kg/(m s 3) Qij dissipation tensor due to inelastic collisions, kg/(m s3) R functional Rpi particle position vector, m R+ dimensionless pipe radius (= D u , / 2 v ) Rep Reynolds number (= [L 7 - V[dp/v), dimensionless r radial coordinate, m S swirl number, dimensionless ScB Brownian Schmidt number, dimensionless Sc t turbulent Schmidt number, dimensionless T operator of interaction of particles with turbulent eddies T L Lagrangian time scale of turbulence, s ~u time of interaction of particles with high-energy turbulent eddies, s L time microscale of turbulence [= (15v/e)~/2], dimensionless Ui, u i, uri averaged, actual, and pulsation velocity of the gaseous phase, m / s u, friction velocity, m / s u+ dimensionless velocity ( = UJu , ) <utiu ~ ) gas turbulent stress, m2/s 2 U~o velocity of the primary flow in the inlet section, m / s re, vi, ~; averaged, actual, and pulsation velocity of the dispersed phase, m/s" Upi velocity of particle, m / s w~ force acceleration due to particle-particle collisions, m / s 2 wi force acceleration due to fluid-dynamic interaction, m / s 2 X distance in the longitudinal direction on the wall, m X i coordinates, m X r reattachment length, m Xstep step length, m Y ratio between frequencies of particle collisions in dense and dilute particleladen flows, dimensionless y distance in the normal direction to the wall, m y+ dimensionless distance in the normal direction to the wall ( = y u , / v ) Z mass concentration of the dispersed phase (= pp~), k g / m 3 Z n mass particle concentration in the nearwall region, k g / m 3 Z m mean concentration of the dispersed phase in the pipe section, k g / m 3 Greek Symbols A dimensionless particle diameter,
(-.~ dp/,.ruklp/2)
8

Aj, AK, Ap, AT Apo


V

PJ' VK, Pp, PT

Vp0
Pt

rate of turbulence dissipation, m2/s 3 Prandtl-Karman constant, dimensionless coefficients of diffusion of pulsations in the dispersed phase, mZ/s coefficients of diffusion in dilute particle-laden flow, m2/s kinematic viscosity of the gaseous phase, mZ/s coefficients of shear viscosity in the dispersed phase, mZ/s coefficients of shear viscosity in dilute particle-laden flow, m2/s coefficient of gas turbulent viscosity,
m2/s

Wv
P, Pp

~rij ~rk, ~r~ r rc, %l, rc2 %


TO
T+ T,

qb,

6v

separation efficiency, dimensionless probability density relating particle velocity by particle-wall interaction density of gas and particles, k g / m 3 tensor of fluid-dyn_amic particle interaction, mZ/s z) turbulence constants, dimensionless time, s typical times between collisions, s particle dynamic and thermal relaxation time, s particle relaxation time (= % u , / D = r +/R +), dimensionless particle relaxation time (= %u 2 / v ) , dimensionless particle relaxation time (= %u,/61), dimensionless volume particle concentration, dimensionless limiting volume particle concentration, dimensionless coefficient of restitution of longitudinal velocity component by particle-wall collision, dimensionless function of particle Reynolds number, dimensionless Superscripts collision pulsating

ax

B g i,j,k,n J K max P
S

T
t W

6 Dirac function 6ii Kroneker symbol, dimensionless 61 thickness of viscous sublayer, m

x, y 0

Subscripts axis Brownian gas coordinates particle-particle interaction kinetic transfer maximum value particle single-phase flow particle-eddy interaction turbulent wall parallel and normal to the wall value in the inlet section

Isothermal Two-Phase Turbulent Flows 1 2 + b e f o r e p a r t i c l e - w a l l collision o r the p r i m a r y flow after p a r t i c l e - w a l l collision v a l u e b a s e d o n friction velocity and gas k i n e m a t i c viscosity, d i m e n s i o n l e s s REFERENCES 1. Berlemont, A., Desjonqueres, P., and Gouesbet, G., Particle Lagrangian Simulation in Turbulent Flows. Int. J. Multiphase Flow 16(1), 19-34, 1990. 2. Sommerfeld, M., Modelling of Particle-Wall Collisions in Confined Gas-Particle Flows. Int. J. Multiphase Flow 18(6), 905-926, 1992. 3. Pourahmadi, F., and Humphrey, J. A. C., Modeling Solid-Fluid Turbulent Flows with Application to Predicting Erosive Wear. Phys. Chem. Hydrodyn. 4(3), 191-219, 1983. 4. Chen, C. P., and Wood, P. E., Turbulent Closure Modeling of the Dilute Gas-Particle Axisymmetric Jet. AIChE J. 32(1), 163-166, 1986. 5. Mostafa, A. A., and Mongia, H. C., On the Modeling of Turbulent Evaporating Sprays: Eulerian and Lagrangian Approach. Int. J. Heat Mass Transfer 30(12), 2583-2593, 1987. 6. Rizk, M. A., and Elghobashi, S. E., A Two-Equation Turbulence Model for Dispersed Dilute Confined Two-Phase Flows. Int. J. Multiphase Flow 15(1), 119-133, 1989. 7. Derevich, I. V., Yeroshenko, V. M., and Zaichik, L. I., Hydrodynamics and Heat Transfer of Turbulent Gas Suspension Flows in Tubes 1: Hydrodynamics. Int. J. Heat Mass Transfer 32(12), 2329-2339, 1989. 8. Hong, T., and Zhou, L., Numerical Simulation of Three-Dimensional Turbulent Gas-Particle Flows in Boiler Furnaces by a Continuum Model of Particle Phase. First Asian-Pacific Int. Syrup. on Combustion and Energy Utilization, Beijing, pp. 184-189, October 1990. 9. Shraiber, A. A., Gavin, L. B., Naumov, V. A., and Yatsenko, V. P., Turbulent Flows in Gas Suspensions. Hemisphere, New York, 1990. 10. Simonin, O., Second-Moment Prediction of Dispersed Phase Turbulence in Particle-Laden Flows. Eighth Symp. on Turbulent Shear Flows, Munich, pp. 7.4.1-7.4.6, August, 1991. 11. Dercvich, I. V., and Zaichik, L. I., Particle Deposition from a Turbulent Flow. Fluid Dyn. 23(5), 722-729, 1988. 12. Derevich, I. V., and Zaichik, L. I., An Equation of Probability Density of Velocity and Temperature of Particles in a Turbulent Flow Modeled by a Random Gaussian Field. Appl. Math. Mech. 54, 722-729, 1990. 13. Reeks, M. W., On a Kinetic Equation for the Transport of Particles in Turbulent Flows. Phys. Fluids A 3(3), 446-456, 1991. 14. Reeks, M. W., On the Continuum Equations for Dispersed Particles in Nonuniform Flows. Phys. Fluids A 4, 1290-1303, 1992. 15. Zaichik, L. I., Modelling of Turbulent Transport and Heat Transfer in the Dispersed Phase Using Equations for the Second and Third Moments of Particle Velocity and Temperature. Inzh.-Fiz. Z. 63(4), 404-413, 1992. 16. Lun, C. K. K., Savage, S. B., Jeffrey, D. J., and Chepurniy, N., Kinetic Theories for Granular Flow: Inelastic Particles in Couette Flow and Slightly Inelastic Particles in a General Flow Field. J. Fluid Mech. 140(4), 223-256, 1984. 17. Ding, J., and Gidaspov, D., A Bubbling Fluidization Model Using Kinetic Theory of Granular Flow. AIChE J. 36(4), 523-538, 1990. 18. Andersen, E., Statistic Approach to Continuum Models for Turbulent Gas-Particle Flows. Ph.D. Thesis, Technical University of Denmark, Lyngby, 1990. 19. Klyatskin, V. I., Stochastic Equations and Waves in Random Inhomogeneous Media. Nauka, Moscow, 1980.

309

20. Koch, D. L., Kinetic Theory for a Monodispersed Gas-Solid Suspension. Phys. Fluids A 2(10), 1711-1723, 1990. 21. Hirschfelder, J. O., Curtiss, Ch. F., and Bird, R. B., Molecular Theory of Gases and Liquids, John Wiley, New York, 1954. 22. Ogawa, S., Umemura, A., and Oshima, N., On the Equations of Fully Fluidized Granular Materials. J. Appl. Math. Phys. 31, 483-493, 1980. 23. Grad, H., On the Kinetic Theory of Rarefied Gases. Commun. Pure Appl. Math. 2(4), 331-407, 1949. 24. Launder, B. E., and Spalding, D. B., The Numerical Computation of Turbulent Flow. Comput. Math. Appl. Mech. Eng. 2, 269-289, 1974. 25. Vinberg, A. A., Zaichik, L. I., and Pershukov, V. A., Calculation of the Momentum and Heat Transfer in Turbulent Gas-Particle Jet Flows. Fluid Dyn. 27(3), 353-362, 1992. 26. He, J., and Simonin, O., Non-Equilibrium Prediction of the Particle-Phase Stress Tensor in Vertical Pneumatic Conveying. Fifth Symp. on Gas-Solid Flows. ASME FED 166, 253-263, 1993. 27. Rodi, W., Turbulent Models for Environment, in Prediction Methods for Turbulent Flows, W. Kollman, Ed., Hemisphere, New York, 1980. 28. Volkov, E. P., Zaichik, L. I., and Pershukov, V. A., Modelling of Solid Fuel Combustion. Nauka, Moscow, 1994. 29. Kulick, J. D., Fessler, J. R., and Eaton, J. K., Particle Response and Turbulence Modification in Fully Developed Channel Flow. J. Fluid Mech. 277, 109-134, 1994. 30. Patankar, S., Numerical Heat Transfer and Fluid Flow. Hemisphere, Washington, DC, 1980. 31. Modarress, D., Tan, H., and Elghobashi, S., Two-Component LDA Measurement in a Two-Phase Turbulent Jet. A/AA J. 22(5), 624-630, 1984. 32. Tsuji, Y., Morikava, Y., Tanaka, T., Karimine, K., and Nishida, S, Measurement of an Axisymmetric Jet Laden with Coarse Particles. Int. J. Multiphase Flow 14(5), 565-574, 1988. 33. Laats, M. K., and Frishman, F. A., Assumptions Used in Calculating the Two-Phase Jet. Fluid Dyn. 5(2), 333-338, 1970. 34. Laats, M. K., and Frishman, F. A., Development of Technique and Study of the Intensity of Turbulence on the Axis Two-Phase Turbulent Jet. Fluid Dyn. 8(2), 304-307, 1973. 35. Bulzan, D. L., Shuen, J.-S., and Faeth, G. M., Particle-Laden Swirling Free Jets: Measurements and Predictions. AIAA Paper No. 303, pp. 1-13, 1987. 36. Lee, S. L., and Durst, F., On the Motion of Particle in Turbulent Duct Flows. Int. J. Multiphase Flow 8(2), 125-146, 1982. 37. Zaichik, L. I., and Pershukov, V. A., Modelling of Particle Motion in a Turbulent Flow with Allowance for Collisions. Fluid Dyn. 30(1), 49-63, 1995. 38. Laats, M. K., and Mulgi, F. S., Experimental Investigation of Kinematic Picture of Small-Dispersed Pipe Flow, in Turbulent Two-Phase Flows, M. K. Laats, Ed., pp. 32-46, Estonian Academy of Sciences, Tallin, 1979. 39. Tsuji, Y., Morikava, Y., and Shiomi, H., LDV Measurements of an Air-Solid Two-Phase Flow in a Vertical Pipe. J. Fluid Mech. 139, 417-434, 1984. 40. Gusev, I. N., Guseva, E. I., and Zaichik, L. I., Model of Particle Deposition from Turbulent Gas-Solid Flow in Channel with Absorbent Walls. Fluid Dyn. 27(1), 43-48, 1992. 41. Forney, L. J., and Spielman, L. A., Deposition of Coarse Aerosols from Turbulent Flow. J. Aerosol Sci. 5(2), 257-271, 1974. 42. Liu, B. Y. H., and Agarwal, J. K., Experimental Observation of Aerosol Deposition in Turbulent Flow. J. Aerosol Sci. 5(2), 145-155, 1974. 43. Ganic, E. N., and Mastanaiah, K., Investigation of Droplet Deposition from a Turbulent Gas Stream. Int. J. Multiphase Flow 7, 401-422, 1981. 44. E1-Kassaby, M. M., and Ganic, E. N., Droplet Deposition in Two-Phase Turbulent Flow. Int. J. Heat Mass Transfer 29(8), 1149-1158, 1986.

310

L . I . Zaichik et al. of Particles in a Turbulent Axisymmetric Gas Jet Injected into a Flash Furnace Shaft. MetaU. Trans. 19(12), 871-884, 1988. 51. Sommerfeld, M., and Qiu, H.-H., Detailed Measurements in a Swirling Particulate Two-Phase Flow by a Phase Doppler Anemometer. Int. J. Heat Fluid Flow 12(1), 20-28, 1991. 52. Sommerfeld, M., Ando, A., and Wennerberg, D., Swirling, Particle-Laden Flows Through a Pipe Expansion. Trans. A S M E 114, 648-656, 1992. 53. Johansen, S. T., and Anderson, N. M., A Model for Predicting Effects of Thermophoresis on Collection Efficiency in Swirling Flow Precipitators. IChemE Syrup. Sen 99, 73-88, 1987.

45. Lee, M. M., Hanratty, T. J., and Adrian, R. J., The Interpretation of Droplet Deposition Measurements with Diffusion Model. Int. J. Multiphase Flow 15(3), 459-469, 1989. 46. Binder, J. L., and Hanratty, T. J., A Diffusion Model for Droplet Deposition in Gas/Liquid Annular Flow. Int. J. Multiphase Flow 17(1), 1-11, 1991. 47. McCoy, D. D., and Hanratty, T. J., Rate of Deposition of Droplets in Annular Two-Phase Flow. Int. J. Multiphase Flow 3, 319-331, 1977. 48. Johansen, S. T., The Deposition of Particles on Vertical Walls. Int. J. Multiphase Flow 17(3), 355-376, 1991. 49. Ruck, B., and Makiola, B., Particle Dispersion in a Single-Sided Backward-Facing Step Flow. Int. J. Multiphase Flow 14(6), 787-800, 1988. 50. Hahn, Y. B., and Sohn, H. Y., The Trajectories and Distribution

Received March 28, 1995; revised November 15, 1996

Vous aimerez peut-être aussi