You are on page 1of 8

# International Journal of Heat and Mass Transfer 50 (2007) 3925–3932

www.elsevier.com/locate/ijhmt

## A fractal resistance model for ﬂow through porous media

Jinsui Wu a,b, Boming Yu a,*
a
Department of Physics, Huazhong University of Science and Technology, 1037 Luoyu Road, Wuhan 430074, PR China
b
Department of Physics, North China Institute of Science and Technology, Yanjiao Town, Sanhe City, Hebei 06520, PR China

## Received 19 September 2006; received in revised form 25 January 2007

Available online 5 April 2007

Abstract

A fractal model for resistance of ﬂow through porous media is developed based on the fractal characters of porous media and on the
pore–throat model for capillary. The proposed model is expressed as a function of the pore–throat ratio, porosity, property of ﬂuid,
pore/capillary and particle sizes, ﬂuid velocity (or Reynolds number) and fractal dimensions of porous media. There is no empirical con-
stant and every parameter has clear physical meaning in the proposed model. The model predictions are compared with experiment data,
and good agreement is found between them.

## 1. Introduction ﬁed Reynolds number (Rep = (Dpqvs/l)(1  e)1) is less

than 10 [2]. If the ﬁrst term on the right side of2 Eq. (1) is
qvs
The widely employed resistance model for ﬂow through neglected, Eq. (1) is reduced to DPL0
¼ 1:75 1e
e3 Dp
, which is
porous media was proposed by Ergun [1] in 1952. This called Burke-Plummer equation. Burke-Plummer equation
model is called Ergun equation: denotes the kinetic energy loss primarily in turbulent ﬂow
2 and the kinetic/local energy loss dominates the pressure
DP 150lð1  eÞ vs 1  e qv2s drop when the modiﬁed Reynolds number Rep is higher
¼ þ 1:75 ð1Þ
L0 D2p e3 e3 D p than 100 [2]. The interior mechanism for the kinetic energy
loss is not well understood. Ergun equation indicates that
where DP is the pressure drop, L0 is the length along the the pressure drop across the packing length is dependent
macroscopic pressure gradient in porous media, vs is the upon the ﬂow rate, the viscosity and density of ﬂuid, and
superﬁcial velocity (deﬁned by vs = Q/A, where Q is the to- the size, shape and surface of packing materials [1]. It has
tal ﬂow rate through a cross section of area A), l is the been shown that the pressure loss as indicated by Eq. (1)
absolute viscosity of ﬂuid, e is the porosity, and Dp is the is obtained by adding the viscous and kinetic energy losses.
appropriate characteristic length for a medium or the equiv- Ergun equation has been hotly debated in the past decades.
alent mean diameter of particles, q is the density of ﬂuid. Many investigators [3–10] discussed its applicability and
Eq. (1) is based on the average hydraulic radius [2]. The diﬀerent empirical constants under diﬀerent porosities
ﬁrst term on the right side 2of Eq. (1) is called Blake-Kozeny and particles.
equation, i.e. DP
L0
¼ 150lð1eÞ
D2p e3
vs
, which represents the viscous It has been shown that the fractal geometry theory [11]
energy loss primarily in laminar ﬂow, and pressure drop has been used as a tool in many disciplines to characterize
for ﬂow at low speed (or low Reynolds number) is mainly irregular or disordered objects [11–13] such as coast lines,
determined by the viscous energy loss, i.e. when the modi- clouds and islands, roughness of surfaces [14–16], sand-
stone pores [17,18], fracture surfaces of metal [19], and
*
Corresponding author. Tel.: +86 27 87542153; fax: +86 27 87543882. granular materials [20], etc. The pores and their distribu-
E-mail address: yuboming2003@yahoo.com.cn (B. Yu). tions in porous media are analogous to islands or lakes

doi:10.1016/j.ijheatmasstransfer.2007.02.009
3926 J. Wu, B. Yu / International Journal of Heat and Mass Transfer 50 (2007) 3925–3932

## on earth and to contact spots on engineering surfaces. dN

 ¼ Df kDmin
f
kðDf þ1Þ dk ¼ f ðkÞ dk ð5Þ
Therefore, it is possible to model the transport properties Nt
such as ﬂow resistance and permeability for ﬂow in porous
where f ðkÞ ¼ Df kDmin
f
kðDf þ1Þ is the probability density func-
media by fractal geometry theory. In the light of this point,
tion. The probability density function f(k) should satisfy
Yu et al. [21,22] proposed a fractal geometry model for per-
the following normalization relation:
meability of porous media and their model has been shown
Z kmax  D
to be suitable not only for particle porous media [21] but kmin f
also for porous fabrics [22]. Their model is analytically f ðkÞ dk ¼ 1  ¼1 ð6Þ
kmin kmax
expressed as a function of fractal dimensions (for pore
spaces and for tortuous capillaries/streamlines) and micro- As a result, Eq. (6) holds if and only if [27]
structural parameters of the media. Karacan and Halleck ðkmin =kmax Þ
Df
¼0 ð7Þ
[23] extended Yu and Cheng’s [21] model to the prediction
2
of the permeability for grain fragments. Recently, Shi et al. In general, kmin/kmax 6 10 in porous media, and Eq. (7)
[24,25] extended Yu et al.’s fractal permeability model holds approximately, thus the fractal geometry theory
[21,22] to modeling the permeability for the gas diﬀusion and technique can be used to analyze properties of porous
layer (GDL) of PEM fuel cells, whose pore size is in the media. In above equations, fractal dimension Df is given by
order of 105–108 m. Meng et al. [26] also applied the [27]
fractal geometry theory to model the permeation of mem- ln e
brane fouling in membrane bioreactor. The cake layer Df ¼ d E  ð8Þ
lnðkmin =kmax Þ
formed on membrane surface presents a major challenge
to membrane permeation, and it can be considered as a where dE is the Euclidean dimension, and dE = 2 (3) in two
porous media. The cake layer permeability was derived (three) dimensions.
and found to be a function of the pore-area fractal dimen- If one is interested in fractal particles, the above param-
sion and microstructural parameters. eters and equations are immediately applicable as long as
From the above brief review, it is seen that the wide appropriate changes are made, for example, changing the
applications of the fractal geometry theory in many ﬁelds porosity e into the particle volume fraction, and pore diam-
have been found. It, therefore, may be possible to develop eter k into the particle diameter, etc.
the analytical model for resistance of ﬂow in porous media The tortuous capillaries have also been shown to follow
based on the fractal geometry theory. In this paper, we the fractal scaling law given by [21]
derive a fractal model for resistance of ﬂow through porous
Lt ðkÞ ¼ k1DT LD0 T ð9Þ
media with particles of diﬀerent shapes based on the fractal
characters of the media and on the pore–throat model for where DT is the fractal dimension for tortuous capillaries
capillary. In the following section, the fractal characters of with 1 < DT < 2 in two dimensions, representing the convo-
porous media are addressed ﬁrst. luted extent of capillary pathways for ﬂuid ﬂow through a
porous medium, and Lt(k) is the tortuous/real length. Due
2. Fractal characters of porous media to the tortuous nature of the pore channel, Lt(k) P L0,
where L0 is the length along the macroscopic pressure gra-
It has been shown that the cumulative size distribution dient in the medium. Note that DT = 1 represents a straight
of pores in porous media follows the fractal scaling law capillary path, and a higher value of DT corresponds to a
[21,22]: highly tortuous capillary.
Df
Based on the above fractal characters of pores and tor-
N ðL P kÞ ¼ ðkmax =kÞ ð2Þ tuous capillaries in porous media, a fractal model for resis-
tance of ﬂow in porous media is derived in the following
where k is the diameter of pores, kmax is the maximum
section.
diameter of pores, N is the cumulative population of pores
whose sizes are greater than or equal to k, and Df is the
fractal dimension for pores, with 1 < Df < 2 in two dimen- 3. Fractal model for ﬂow resistances
sions and 2 < Df < 3 in three dimensions.
It is evident that the total number of pores, from the 3.1. The viscous energy loss along the ﬂow path at low
smallest diameter to the largest diameter, can be obtained Reynolds numbers
from Eq. (2) as
In this model, we assume that a porous medium is com-
Df
N t ðL P kmin Þ ¼ ðkmax =kmin Þ ð3Þ prised of a bundle of tortuous capillaries. The ﬂow rate
through a tortuous capillary is given by modifying the well
Diﬀerentiating on both sides of Eq. (2) results in
known Hagen–Poiseulle equation as [28]
dN ¼ Df kDmax
f
kðDf þ1Þ dk ð4Þ p DP 1 k4
qðkÞ ¼ ð10Þ
Dividing Eq. (4) by (3) yields 128 Lt l
J. Wu, B. Yu / International Journal of Heat and Mass Transfer 50 (2007) 3925–3932 3927

## The total ﬂow rate Q through a unit cell can be obtained by

integrating the individual ﬂow rate, q(k), over the entire
range of pores/capillaries from the minimum pore size kmin
to the maximum pore size kmax. According to Eqs. (4), (9)
and (10), we have [21]
Z kmax
Q¼ qðkÞ dN ðkÞ
kmin

p Dp1 Df
¼ L1DT k3þD T
ð11Þ
128l L0 3 þ DT  Df 0 max

## The total pore area Ap is related to the total cross area A by

A = Ap/e. The total pore area Ap can be obtained by
Z kmax 2 "  2Df #
pk pDf 2 kmin
Ap ¼  dN ¼ k 1
kmin 4 4ð2  Df Þ max kmax
Fig. 1. An idealized pore–throat model.
ð12Þ

Thus, the total cross area A is DT) aﬀecting the ﬂow resistance are not revealed. Eq. (17)
"  2Df # also indicates that the maximum pose size signiﬁcantly
Ap pDf 2 kmin
A¼ ¼ k 1 ð13Þ inﬂuences the pressure drop. However, Blake-Kozeny
e 4eð2  Df Þ max kmax equation does not show this dependence. The following is
devoted to deriving a simple model for the maximum pore
Due to [27] size.
 2Df Based on the idealized pore–throat model as shown in
kmin
e¼ ð14Þ Fig. 1 for ﬂow through porous media, the area of the unit
kmax
cell is
So, the total cross sectional area A of a unit cell perpendic-
pD2p
ular to the ﬂow direction is S¼ ð18Þ
4ð1  eÞ
p Df 1  e 2
A¼ kmax ð15Þ The area of the pore is
4 2  Df e
 
Thus, the superﬁcial velocity is pD2p pD2p 1
S p; max ¼S ¼ 1 ð19Þ
4 4 ð1  eÞ
1 Dp1 2  Df e
vs ¼ Q=A ¼ L1DT k1þD T
ð16Þ
32l L0 3 þ DT  Df 1  e 0 max The irregular geometry of the macro-pore area is approxi-
mated as a circular pore as
According to Eq. (16), we can get the pressure drop across
p
the length L0 along the macroscopic pressure gradient as S p; max ¼ k2max ð20Þ
4
Dp1 32lvs 3 þ DT  Df 1  e 1
¼ 1DT ð17Þ From Eqs. (19) and (20), kmax can be obtained as
L0 L0 2  Df e k1þD
max
T
rﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ
e
Eq. (17) depicts the pressure drop caused by the viscous en- kmax ¼ Dp ð21Þ
1e
ergy loss along the ﬂow path. This pressure drop is linearly
proportional to the superﬁcial velocity. Eq. (17) is similar Eq. (21) is a simple model for the maximum pore size,
to Blake-Kozeny equation, which represents the viscous which is related to porosity and the equivalent mean diam-
energy loss. Note that the empirical constant 150 in eter Dp. It is evident that if porosity e = 0, kmax = 0; if
Blake-Kozeny equation has no physical meaning and is porosity e = 1, kmax is inﬁnity. This is consistent with the
independent of porosity. Whereas every parameter in Eq. practical situations. Eq. (21) also indicates that the maxi-
(17) has clear physical meaning, and the viscous energy loss mum pore size kmax is proportional to the mean particle
expressed by Eq. (17) depends on ﬂuid viscosity, pore size, diameter Dp. This is also consistent with the physical situ-
ﬂuid velocity, porosity, pore area dimension Df and the tor- ation. In Eq. (17), kmax is determined by Eq. (21).
tuosity fractal dimension DT. However, Blake-Kozeny Eq. (17) can also be written as the dimensionless form:
equation only depends on ﬂuid viscosity, particle size, ﬂuid 3
Dp1 qDp e3 32 3 þ DT  Df Dp
2

## velocity and porosity, and some other mechanisms (such as G1 ¼ 3

¼ 1DT 1þD
Rep ð22Þ
L0 l2 ð1  eÞ L0 2  Df kmax T
pore size distribution described by the fractal dimension Df
and capillary tortuosity described by the fractal dimension where Rep = (Dpqvs/l)(1  e)1 is the Reynolds number.
3928 J. Wu, B. Yu / International Journal of Heat and Mass Transfer 50 (2007) 3925–3932
 
3.2. The kinetic energy loss at high Reynolds numbers 3 1 5 qv2t
DP 2 ¼ qghf ¼ þ 4 2 ð33Þ
2 b 2b 2
In Fig. 1, the ratio of pore diameter (BF + Dp = b) to
BF þD The average velocity in a straight pipe/channel is
throat diameter (BF) is deﬁned by b, i.e. b ¼ BF p , and l
is the length of a pore–throat unit. a1 and a2 represent dL0
the cross sectional areas. v0 ¼ ð34Þ
dt
The length of a pore–throat unit is assumed to be l = b,
and based on Eq. (18), this yields Due to Eq. (9), we can get the relation between average
rﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ velocity vt in a tortuous capillary and average velocity v0
pﬃﬃﬃ p
l ¼ b ¼ BF þ Dp ¼ S ¼ Dp ð23Þ in a straight capillary as follows
4ð1  eÞ
vt ¼ dLt =dt ¼ DT L0ðDT 1Þ kð1DT Þv0 ð35Þ
Thus, we obtain
rﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ  Due to overlapping of particles randomly distributed in a
p
BF ¼ Dp 1 ð24Þ porous medium, the ﬂuid passes not only through the max-
4ð1  eÞ imum macro-pores but also through the narrow channels
BF þ Dp 1 with the width of BF, and these maximum pores and the
b¼ ¼ qﬃﬃﬃﬃﬃﬃﬃﬃ ð25Þ
BF narrow channels may be in the same ﬂow direction. Thus,
1  2 ð1eÞ p the average pore size k in Eq. (35) may be taken as
Now, we consider the kinetic/local energy loss. The loss k ¼ ðkmax þ BF Þ=2 ð36Þ
of head is expressed by [29]
X v2 Substituting Eqs. (21) and (24) into Eq. (36) yields
hf ¼ n t ð26Þ rﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ rﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ 
2g k ¼ ðkmax þ BF Þ ¼ Dp e
þ
p
1 ð37Þ
2 2 1e 4ð1  eÞ
where vt is the real average velocity in a throat of cross sec-
tion area a1 (see Fig. 1), and n represents the coeﬃcient of Inserting Eq. (35) into Eq. (33) results in
kinetic/local energy loss. The coeﬃcient for a sudden  
3 1 5 q ðD 1Þ
½DT L0 T kð1DT Þ  v20
expanding pipe/channel is [29] 2
Dp2 ¼ qghf ¼ þ  ð38Þ
 2 b4 2b2 2
a1 2
ne ¼ 1  ð27Þ
a Due to the relation between the superﬁcial velocity vs
where a is the cross sectional area of the pore. For a sudden and the average velocity v0 , v0 ¼ vs =e, Eq. (38) is rewritten
contracting pipe/channel of cross section area a2, the coef- as
 
ﬁcient is [29] 3 1 5 qv2s 2 ð2DT 2Þ ð22DT Þ
 Dp2 ¼ qghf ¼ þ 4 2 DT L0 k ð39Þ
a2  2 b 2b 2e2
nc ¼ 0:5 1  ð28Þ
a
Eq. (39) represents the pressure drop along a pore–throat,
So, the local loss of head caused by the sudden expanding and the pressure drop per unit length is then
pipe/channel is  
 2 2 Dp2 3 1 5 qv2s 1 2 ð2DT 2Þ ð22DT Þ
v2e 1 ve ¼ þ 4 2 DT L0 k ð40Þ
hfe ¼ ne ¼ 1 2 ð29Þ l 2 b 2b 2e2 l
2g b 2g
Eq. (40) denotes that the kinetic/local energy loss is linearly
and the local loss of head caused by the sudden contracted proportional to v2s , which (vs) as related to Eq. (16). Eq.
pipe/channel is (40) represents the kinetic energy loss and is similar to
  Burke-Plummer equation. However, Burke-Plummer equa-
1 v2 v2
hfc ¼ 0:5 1  2 c ¼ nc c ð30Þ tion contains an empirical constant 1.75, whereas there is
b 2g 2g
no empirical constant in Eq. (40), in which every parameter
Due to continuity of ﬂow rate, we have has clear physical meaning. Eq. (40) is expressed as a func-
q ¼ a1ve ¼ a2vc ð31Þ tion of the pore–throat ratio, porosity, ﬂuid property, pore
size, velocity, and fractal dimensions Df and DT. While
Since a1 = a2, ve ¼ vc ¼ vt . Adding Eqs. (29) and (30) re- Burke-Plummer equation does not reveal the dependences
sults in the total loss of head of the pore–throat ratio, pore size, fractal dimensions Df
  and DT on the pressure drop. Therefore, the physical prin-
v2 3 1 5 v2
hf ¼ hfe þ hfc ¼ ðn1 þ n2 Þ t ¼ þ 4 2 t ð32Þ ciples of the kinetic energy loss are clearly revealed in the
2g 2 b 2b 2g
proposed model Eq. (40).
The pressure drop across the pore–throat unit with length l Eq. (40) can also be expressed in terms of the dimension-
then is less form:
J. Wu, B. Yu / International Journal of Heat and Mass Transfer 50 (2007) 3925–3932 3929

2 r
ﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ
2 3
Dp qD3p e3
G2 ¼ 2 pﬃﬃﬃﬃﬃ
1 ﬃ
 1 þ 47 1
l l2 ð1  eÞ3 16 1 pﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ pﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ 1e
s¼ 6 1 þ 1  e þ 1  e p ﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ 7 ð46Þ
   2DT 2 24 2 1 1e 5
3 1 5 Dp e 2 L0
¼ þ  D Re2p ð41Þ
2 b4 2b2 2l ð1  eÞ T 
k
Interested readers may consult reference [30] for the gen-
eral behavior of the fractal dimension DT, and for the sen-
3.3. Total pressure drop sitivity analyses of DT on the ﬂow resistance/permeability,
The total pressure drop per unit length is the sum of the Eq. (43) is now rewritten as the dimensionless form,
viscous energy loss and kinetic/local energy loss along the G ¼ G1 þ G2
ﬂow path, i.e. 2
1 32 3 þ DT  Df Dp
Dp1 Dp2 32lvs 3 þ DT  Df 1  e 1 ¼ Rep
þ ¼ 1DT /s L1D
0
T 2  Df k1þD
max
T

L0 l L0 2  Df e k1þD
max
T
   2DT 2
   2DT 2 1 3 1 5 Dp e 2 L0
3 1 5 qv2s 1 2 L0 þ 2 þ  DT  Re2p
þ þ 4 2 D ð42Þ /s 2 b4 2b2 2l ð1  eÞ k
2 b 2b 2e2 l
T 
k
ð47Þ
Eq. (42) is similar to Ergun equation (1), which has two
and Ergun equation is also be rewritten as the dimension-
empirical constants, 150 and 1.75. Whereas there is no
less form, i.e.
empirical constant in Eq. (42) and every parameter in Eq.
3
(42) has the clear physical meaning, and more physical DP qDp e3
principles are revealed in Eq. (42). ¼ 150Rep þ 1:75Re2p ð48Þ
L0 l2 ð1  eÞ3
However, in reality, not all the pores/capillaries are
spherical, so Eq. (42) is modiﬁed by introducing /s, called Fig. 2 compares the dimensionless ﬂow resistances pre-
sphericity of particles [7], then the modiﬁed equation is dicted respectively by Eqs. (22) and (47) versus Reynolds
expressed as numbers when Rep < 10 for the spherical particles. The
results show that when Reynolds numbers Rep < 2.75, the
Dp1 Dp2 1 32lvs 3 þ DT  Df 1  e 1
þ ¼ relative error between the predicted values by the two equa-
L0 l /s L1D
0
T 2  Df e k1þD
max
T
tions is less than 10%, This means that the pressure drop is
  2  2DT 2
1 3 1 5 qvs 1 2 L0 mainly determined by the viscous energy loss Eq. (22) at
þ 2 þ 4 2 DT  low Reynolds numbers, and Eq. (22) can be a good approx-
/s 2 b 2b 2e2 l k
imation to the ﬂow resistance at low Reynolds numbers.
ð43Þ However, at higher Reynolds numbers, Eq. (47) should
1=3 be used.
where /s ¼ ð36pV 2P =S 3P Þ [8], here VP and SP are volume
Fig. 3 compares the dimensionless ﬂow resistances pre-
and surface area of particles, with /s = 1 for spherical par-
dicted respectively by Eqs. (41) and (47) versus the modi-
ticles and /s < 1 for other shaped particles.
ﬁed Reynolds numbers as Rep > 220. The results show
that when the modiﬁed Reynolds numbers Rep > 220, the
4. Results and discussion

## It was generally recognized that the minimum pore size

is smaller than the maximum pore size in porous media by 800
0.15
Relative error
at least two orders of magnitude. In this work we thus
Relative error

0.10

## dimension D for particles can be found by modifying Eq.

0.05
(8) as
G1, G

400
lnð1  eÞ 0.00
0 1 2 3 4
D ¼ dE  ; ð44Þ Rep
lnðd min =d max Þ
200
and the fractal dimension DT for tortuous capillaries can be Eq. (47)
obtained from Eq. (9) as Eq. (22)

ln s 0
DT ¼ 1 þ ð45Þ 0 1 2 3 4 5 6 7 8 9 10
lnðL0 =kÞ Rep

where 
k is determined by Eq. (37), and the tortuosity s is Fig. 2. A comparison on the dimensionless ﬂow resistance versus Rep at
deﬁned by s = Lt/L0 and is expressed as [30] Rep < 10, e = 0.40, Dp = 10 mm.
3930 J. Wu, B. Yu / International Journal of Heat and Mass Transfer 50 (2007) 3925–3932

0.20 5
6
2.4x10
2.5x10 Present model Eq. (47)
0.15
5
Relative error
Relative error
2.0x10 Ergun equation (48)
6 0.10 Experiment data
2.0x10
5
0.05 1.6x10
6
G2, G

1.5x10 0.00 5
200 400 600 800 1000 1.2x10

G
Rep
6
1.0x10 4
8.0x10
5
5.0x10 Eq. (47) 4
4.0x10
Eq. (41)
0.0 0.0
200 400 600 800 1000 1200 0 40 80 120 160 200
Rep Re p

Fig. 3. A comparison on the dimensionless ﬂow resistance versus Rep at Fig. 5. A comparison among the present model Eq. (47), Ergun equation
Rep > 100, e = 0.40, Dp = 10 mm. (48) and experiment data for Trilobes particles [8] at e = 0.511, /s = 0.63
and Dp = 1.41 mm.

## relative error between the predicted values by the two equa-

tions is less than 10%, This means that at high Reynolds
numbers the ﬂow resistance is dominated by the kinetic/
local energy loss Eq. (41). 1000
It should be noted that Eq. (47) is the general form for Present Eq. (42)
resistance of ﬂow through porous media. From Eq. (47) it Ergun Eq. (1)
Pressure drop (pa/m)
800 Experiment data
can be seen that the pressure drop for ﬂow at low speed (or
at low Reynolds number) is mainly determined by the vis- 600
cous energy loss, which is represented by the ﬁrst term on
the right-hand side of Eq. (47). In this case, the irregularity 400
of porous shape can be ignored. At high speed (or at high
Reynolds number) the irregularity of porous shapes has the 200
signiﬁcant inﬂuence on ﬂow resistance, and the kinetic/
local energy loss dominates the pressure drop, which is 0
mainly determined by the second term on the right-hand 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
side of Eq. (47). Eq. (47) is similar to Ergun equation, vs (m/s)
which has two empirical constants 150 and 1.75. Whereas
Fig. 6. Pressure drop when air ﬂows through a bed packed with glass balls
there is no empirical constant and every parameter has [31].
the clear physical meaning in Eq. (47).
Fig. 4 compares the present model predictions by Eq.
(47) with those by Ergun equation (48) and the experiment
data for beds packed with Quadralobes particles [8] of
/s = 0.593, Dp = 1.26 mm at e = 0.502. Fig. 5 compares
the present model predictions by Eq. (47) with those by
2.0x10
5 Ergun equation (48) and the experiment data [8] for beds
Present model Eq. (47) packed with Trilobes particles of diﬀerent shapes of
5
Ergun equation (48)
1.6x10 Experiment data /s = 0.63, Dp = 1.41 mm at e = 0.511. Fig. 6 compares
the present model predictions (at e = 0.433) by Eq. (42)
5
1.2x10 with those by Ergun equation (1) and the experiment data
[31], which were measured for beds packed with spherical
G

4
8.0x10 particles of Dp = 10 mm and L = 0.5 m [31], L is the real
length of a porous sample. From Figs. 4–6, it can be seen
4
4.0x10 that the model predictions are in good agreement with
the experiment data with non-sphericity particles. How-
0.0 ever, Ergun equation underestimates the resistance, and
0 20 40 60 80 100 120 140 160 180
this may be explained that Ergun equation may only be
Re p suitable for spherical particles and for his experimental
Fig. 4. A comparison among the present model Eq. (47), Ergun equation data.
(48) and experiment data for Quadralobes particles [8] at e = 0.471, If Eq. (42) is written as the other dimensionless form, we
/s = 0.593 and Dp = 1.26 mm. obtain
J. Wu, B. Yu / International Journal of Heat and Mass Transfer 50 (2007) 3925–3932 3931
 
8000 Dp1 Dp2 Dp e3
Fk ¼ þ 2
7000
L0 l qv ð1  eÞ
   2DT 2
6000 1 3 1 5 Dp e L0
¼ 2 þ 4 2 D2T 
5000 /s 2 b 2b 2l ð1  eÞ k
2
1 32 3 þ DT  Df Dp
fv

4000 1
þ 1DT ð1þD Þ
ð51Þ
3000 /s L0 2  Df kmax Rep
T

2000 Present model Eq. (49) Fig. 8 compares the present model predictions by Eq. (51)
Ergun equation (50)
1000
Experiment data
with the experiment data [6] for cylindrical particles. From
0 Fig. 8 it can be found that the model predictions also pres-
1000 1500 2000 2500 3000 3500 4000 ent good agreement with the experimental data.
Re p
5. Conclusions
Fig. 7. A comparison among the present model Eq. (49), Ergun equation
(50) and experiment data for ball particles [32] at e = 0.364, /s = 1, and
Dp = 12 mm. We have shown an analytical model for resistance of
ﬂuid ﬂow in porous media based on the capillary model
  and pore–throat model as well as the fractal characters of
Dp1 Dp2 D2p e3 pores and capillaries. The proposed model is expressed as
fv ¼ fv1 þ fv2 ¼ þ
L0 l lvs ð1  eÞ2 a function of the pore–throat ratio, porosity, ﬂuid prop-
2   erty, pore/capillary and particle size, ﬂuid velocity (or Rey-
32 3 þ DT  Df Dp 3 1 5 Dp
¼ 1DT þ þ  nolds number) and fractal characters (Df and DT) of pores
L0 2  Df ð1þDT Þ
kmax 2 b4 2b2 2l and capillaries in porous media. There is no empirical con-
 2DT 2 stant and every parameter has clear physical meaning in the
e L0
 D2T  Rep ð49Þ proposed model. The physical principles of the viscous
ð1  eÞ k
energy loss along the ﬂow path and the kinetic energy loss
Similarly, if Ergun equation (1) is written as the other are clearly revealed. The model predictions are compared
dimensionless form, we have with experiment data, and good agreement is found
between them. The validity of the propose model is thus
DP D2p e3 veriﬁed.
¼ 150 þ 1:75Rep ð50Þ
L0 lvs ð1  eÞ2 It should be pointed out that the accuracy of the present
fractal model may crucially depends on the correct determi-
Fig. 7 compares the present model predictions by Eq. (49) nation of sphericity /s of particles, the maximum pore or
with those by Ergun equation (50) and the experiment data particle size, the ratio of kmin/kmax, and the ratio of pore
[32] for beds packed with spherical particles of to throat. Therefore, the correct determination of these
Dp = 12 mm, e = 0.364. From Fig. 7, it can be found that parameters is critical for successfully predicting the resis-
the model predictions also present much better agreement tance of ﬂow in porous media.
with the experimental data than those by Ergun equation.
If Eq. (43) is written as Acknowledgement

10
4 This work was supported by the National Natural Sci-
Eq. (51), ε=0.41 Dp=2.82mm ence Foundation of China through Grant Number
Eq. (51), ε=0.414 Dp=2.24mm 10572052.
3
10 Experiment, ε=0.41 Dp=2.82mm
Experiment, ε=0.414 Dp=2.24mm References
log10Fk

2
10 [1] S. Ergun, Fluid ﬂow through packed columns, Chem. Eng. Prog. 48
(1952) 89–94.
[2] R.B. Bird, W.E. Stewart, E.N. Lightfoot, Transport Phenomena,
10
1 Wiley, New York, 1960 (Chapter 6).
[3] R.E. Hicks, Pressure drop in packed beds of spheres, Ind. Eng. Chem.
Fund. 9 (1970) 500–502.
[4] R.D. Bradshaw, J.E. Myers, Heat and mass transfer in ﬁxed and
-1 0 1 2 3
ﬂuidized beds of large particles, AIChE. J. 9 (1963) 590–598.
10 10 10 10 10 [5] D. Handly, P.J. Heggs, Momentum and heat transfer mechanisms in
log10Rep regular shaped packings, Trans. Inst. Chem. Eng. 46 (1968) 251–259.
[6] F. MacDonald, M.S. Ei-Sayed, K. Mow, F.A.L. Dullien, Flow
Fig. 8. A comparison of the model predictions with the experimental data through porous media—the Ergun equation revised, Ind. Eng. Chem.
by Pahl (1975) [6] at e = 0.41, /s = 0.862 and e = 0.414, /s = 0.872. Fundam. 18 (1979) 199–208.
3932 J. Wu, B. Yu / International Journal of Heat and Mass Transfer 50 (2007) 3925–3932

[7] D. Metha, M.C. Hawley, Wall eﬀect in packed columns, Ind. Eng. [21] B.M. Yu, P. Cheng, A fractal permeability model for bi-dispersed
Chem., Proc. Des. Dev. 8 (1969) 280–286. porous media, Int. J. Heat Mass Transfer 45 (2002) 2983–2993.
[8] D. Nemec, J. Levec, Flow through packed bed reactors: 1. Single- [22] B.M. Yu, L.J. Lee, H.Q. Cao, A fractal in-plane permeability model
phase ﬂow, Chem. Eng. Sci. 60 (2005) 6947–6957. for fabrics, Polym. Compos. 23 (2002) 201–221.
[9] Y. Endo, D-R. Chen, D.Y.H. Pui, Theoretical consideration of [23] C.O. Karacan, P.M. Halleck, A fractal model for predicting perme-
permeation resistance of ﬂuid through a particle packed layer, Powder ability around perforation tunnels using size distribution of frag-
Technol. 124 (2002) 119–126. mented grains, J. Petroleum Sci. Eng. 40 (2003) 159–176.
[10] R.J. Hill, D.L. Koch, A.J.C. Ladd, Moderate-Reynolds-number ﬂows [24] Y. Shi, J.S. Xiao, M. Pan, Z.R. Yuan, A fractal permeability model
in ordered and random arrays of spheres, J. Fluid Mech. 448 (2001) for the gas diﬀusion layer of PEM fuel cells, J. Power Sources 160
243–278. (2006) 277–283.
[11] B.B. Mandelbrot, The Fractal Geometry of Nature, W.H. Freeman, [25] Y. Shi, J.S. Xiao, M. Pan, Z.R. Yuan, Addendum to ‘‘A fractal
New York, 1982. permeability model for gas diﬀusion layer of PEM fuel cells”, J.
[12] J. Feder, A. Aharony, Fractals in Physics, North-Holland, 1989. Power Sources 165 (2007) 299.
[13] J. Feder, Fractals, Plenum Press, New York, 1988. [26] F.G. Meng, H.M. Zhang, Y.S. Li, X.W. Zhang, F.G. Yang,
[14] T.L. Warren, D. Krajcinovic, Random Cantor set models for the Application of fractal permeation model to investigate membrane
elastic–perfectly plastic contact of roughness surfaces, Wear 196 fouling in membrane bioreactor, J. Membr. Sci. 262 (2005) 107–
(1996) 1–15. 116.
[15] F.M. Borodich, A.B. Mosolov, Fractal roughness in contact prob- [27] B.M. Yu, J.H. Li, Some fractal characters of porous media, Fractals 9
lems, J. Appl. Math. 56 (1992) 681–690. (2001) 365–372.
[16] A. Majumdar, B. Bhushan, Role of fractal geometry in roughness [28] M.M. Denn, Process Fluid Mechanics, Prentice Hall, NJ, 1980.
characterization and contact mechanics of surfaces, J. Tribol. 112 [29] Y.Y. Zhang, Fluid Mechanics, Higher Education Press, 1999 (Chap-
(1990) 205–216. ters 3 and 6).
[17] A.J. Katz, A.H. Thompson, Fractal sandstone pores: implications for [30] B.M. Yu, Fractal character for tortuous streamtubes in porous
conductivity and pore formation, Phys. Rev. Lett. 54 (1985) 1325– media, Chin. Phys. Lett. 22 (2005) 158–161.
1328. [31] Z.L. Wang, Y.L. Ding, M. Ghadiri, Flow of a gas–solid two-phase
[18] C.E. Krohn, A.H. Thompson, Fractal sandstone pores: automated mixture through a packed bed, Chem. Eng. Sci. 59 (2004) 3071–3079.
measurements using scanning-electron-microscope images, Phys. Rev. [32] J. Yu, M.C. Zhang, W.D. Fan, Y.G. Zhou, G.F. Zhao, Study on
B 33 (1986) 6366–6374. performance of the ball packed-bed regenerator: experiments and
[19] B.B. Mandelbrot, D.E. Passoja, A.J. Paullay, Fractal character of simulation, Appl. Therm. Eng. 22 (2002) 641–651.
fracture surfaces of metals, Nature 308 (1984) 721–722.
[20] H. Xie, R. Bhaskar, J. Li, Generation of fractal models for
characterization of pulverized materials, Miner. Metall. Process.
(February) (1993) 36–42.