EVALUATION OF STEADY FLOW PROFILES IN

RECTANGULAR AND CIRCULAR PACKED BEDS

BY A VARIATIONAL METHOD

D. VORTMEYER andJ. SCHUSTER

Fakultit fiir Maschinenwesen, Technische UniversiGt Miinchen, Arcisstrasse 21, 8000 Miinchen 2, West Germany

(Received 30 _fuune1982: accepted 20 Jmuaty 1983)

AbstracCThe Brinkman-equation is solved by an equivalent variational problem in order to obtain steady flow
profiles which satisfy the conditions w = 0 and l = I at the container wall of a packed bed. The quite large
deviations between calculated and measured profiles are fulIy explained. An analytical expression is developed
with sufficient accuracy. The main body of calculations is performed for isothermal flow, however flows under
non-isothermal conditions also are considered.

1. INTRODUCTION and Pogorski[ 171, Marivoet et al.[ 181 and Szekely and
Recently published research work emphasizes the im- Proveromo[l9]. Figures l(a) and l/b) exhibit typical
portance of porosity changes-Buchlin et of. [ II-and in averaged flow profiles for two different Reynolds-num-
particular of the corresponding velocity changes near the bers.
container wall-Lerou and Froment [Z], Kalthoff and Somewhat different are the experimental findings of
Vortmeyer [3], Vortmeyer and Winter [4l_for the analy- Price[20]. Unlike other authors Price placed a honey-
sis of packed bed chemical reactors. In fact the results of comb structure on top of the bed and suppressed by this
[3,4] show clearly by comparison with experiments that measure the onset of radial flow components’within the
improvements in the theoretical prediction of reactor fluid having left the packed bed. Higher flow maxima
performance are obtained by assuming a non-uniform closer to the wall are found in Fig. 2. In fact Price
flow distribution within the packed bed. reports in his paper that the region of high velocity exists
While the changes in porosity near the wall&111 are within one-half sphere diameter from the wall of the
well known, there is no general consensus on the precise containing vessel.
flow profiles inside the packed bed. Flowmeasurements From our point of view the results of Price are the
inside the bed are very difficult to perform as reported by best approximations of the flow inside the bed because
Cairns and Prausnitz[l2]. Since this type of measure- own measurements by a Laser-Doppler-Anemometer
ments is rather uninformative for various reasons, the clearly indicate a distortion of flow profiles if the
main body of flow profiles was obtained by measure- measurements are performed within a free cross-section
ments a few millimeters above the exit cross-section of plane. It is obvious from Fig. 3 that the flow profiles
the packed bed. This kind of work was performed, e.g. depend strongly on the distance between the exit plane
by Schertz and Bischoff[l3], Dornweiler and Fahien[l4], of the bed and the actual plane where the measurements
Morales et al.[15], Schwartz and Smith[l6], Calderbank are taken. It should be noted that each of the averaged

50

&~I
25

0 10 20 0 10 20
a) wall distance [mm1 b] wall distance [mm]
-
Fig. I. Flow distributions (averaged) above Bxed beds measured by (a) Schwartz and Smith[l6] at Re,, = 118; (b)
Morales et al. [15] at Rq = 30.
1692

&l
ii
tI--
700I\

500

300
‘I
\
\ -----
D. VORTMEYERand J. SCHUSTER

100

0 50 loo 150
wall distance [mm1

Fig. 2. Flow distribution after Price[20~ = 4955: -,

measured; ----, calculated from eqn (14).

curves in Fig. 3 for a rectangular duct is based on about

two hundred measurements. The local measurements are
found in Fig. 4.
In consideration of the obvious distortions connected
with measurements above the bed the question concern-
ing the “true flow profiles” inside the packed bed has to
be raised. We feel that these profiles are only obtainable
from the Brinkman equation which governs the flow Fig. 4. Actual Row distribution above a rectangular packed bed at
inside a porous medium. Rep = 8; dP = 1.25 mm.

2. THEORETICAL FORMULATIONS.
ISOTHERMAL FLOW
2.1 The Brinkman equation for flow within a rectangular duct or
Brinkman was the first[21] to formulate the differen-
tial equation which describes the artificial flow profile
within a porous medium bounded by a rigid wall. Res-
tricting himself to the Darcy flow regimen he extended
Da&s law by a viscosity term in order to include the for tube flow.
viscous forces near the wall. Since the validity of the The factors f, and fz have the meaning
Brinkman equation is restricted to low flow rates, it was
extended in 1251to higher flow rates by incorporating the
f* = 1.75p 9. (3)
Ergun pressure loss relation1221 a

ap
z=- f ,w -f2wZ+71$

80 Equations (1) and (2) which are assumed to be valid also

in the region of porosity changes near the wall, are
I 60 subject to the following boundary conditions
[Cl%1
x=0: w=o
40
x= b/2: E= 0 rectangular duct with width b

and
0 5 IO 15 20
wall distance [mm~l r=R: w=o
Fig. 3. Averaged measurements of flow distributions at various
heights above a rectangular fixed bed at Rep = 7.5 and dr, = 2 mm.
r =0: 2 =0 circular tube.
 Evaluation of steady flow profiles in rectangular and circular packed beds 1693

The porosity E in eqns (1) and (2) varies as a function of

the wall distance with c = 1 at the wall and e. = 0,4
inside the randomly packed bed of spheres. The average
value of L,, = 0,4 is reached by a more or less oscillating
function starting at the wall with E = 1. The evaluation of
this function has been subject of a number of
pubIicationS [5-I I]. For ideal metal spheres Benenati and
Brosilow[S] determined the porosity function of Fig. 5.
In this case about five oscillations are required in order
to reach the average porosity after a distance of five
particle diameters from the wall. Quite different obser-
O0kTT-T-r
wall distanceldp -
vations are made[ 111 for the porosity function of a
Fig. 6. Porosity function for glass particles with slight deviations
packed bed consisting of glass spheres with small devia- from the spherical shape dp = 4 mm: 0, measurements[lll; ----,
tions from the spherical structure. In this case the eqn (61.
porosity function only shows one oscillation (Fig. 6) in
order to approach the average porosity. For later cal-
culations we have approximated the porosity function by integraI[26]. The solution of the problem is based on
the following exponential expressions minimizing an integml in the region of interest rather
than solving the partial differential equation in its original
form.
Tn the case of steady flow through porous media the
governing partial differential equation is usually elliptic.
for a rectangular duct and This elliptic equation leads to eqn (1) in the case of
one-dimensional flow through a rectangular duct. There-
R-r fore, the variational principle can be empIoye$[27] to the
l+Cexp
(
l-27
P
(6) solution of eqn (1) satisfying the boundary conditions
throughout the region.
for a circular tube where C has to be adjusted according In order to express eqn (I) as equivalent to the mini-
to I%. mization of an integral
The problem is to find a suitable solution of the
nonlinear differential equations (1) or (2) with variable b
E = hl H(x, w, w’) dx (7)
porosity functions (5) or (6) valid throughout the region I0
of interest. A finite difference technique is used in the
Iiterature[23-251 to solve this problem only if the with
porosity at the wall was artificially set to a limiting value
b
of e = 0.44 in. [24J and E = OS6in.[25]. This method, VW.1 = 1 w dx = constant. (8)
however, introduces difficulties like stiffness, conver- I0
gence and so on near the waII where l = 1 and w = 0.
Therefore, instead of solving eqns (1) or (2) in their H must be obtained satisfying the Euler equation
original form, the problem is converted to one of opti-
mizing an integral. This method has certain advantages. 0 CJ)
2.2 The oariational method
Many problems involving elliptic partial differential where w ’ = d w/dx, vtjtotslis the volume flow rate, k is the
equations can be related to the optimization of an height, b the width and 1 the length of the duct.
The requirements are fulfilled by the function

. (10)
+’ 1
The integral to be minimized is

E = kl dx

= min (11)

with qti,,,, = const as given by eqn (8).

wall distance /dp - In eqn (10) h is connected with the pressure gradient.
Fig. 5. Porositydistribution for spherical particles near the wall: n , Equation (9) with eqn (11) is equivaIent to eqn (I). If the
Benenati and Brosilow[5] for dp = 4mm brass spheres; ----, width of the channel is divided in intervals of Axi, the
averaged porosity function calculated from eqn (6). volume of this element is k&x,. For these elements the
1694 D. VORTMWEII
and J. SCHLEXER

minimum principle given by eqn (11) is transformed to the calculated profiles is in better agreement with the
experimental results of Price[20] in Fig. 2.
The following Fig. 8 exhibits computed profiles for a
circular packed bed of d, = 8 mm particles. The Re,-
= min number has been varied only between Re, = 1 and Re, =
(12)
100 because beyond Rep = 100 there are no more con-
.Awi = wit, -WI Axi=xi+L-X, siderable changes in the dimensionless flow profiles.
Remembering that the uniform flow profiles correspond
to the line W/I+= 1 we have to conclude from Fig. 8
(D/d, = 5) that the real flow distribution within chemical
wiAxi = constant. (13) reactors (4 < D/d, < 10) differs considerably from the
uniform flow which usually is assumed for theoretical
analysis. The effects of the flow profiles on fixed bed
In a quite analogous way we obtain the minimum prin- chemical reactor theory are outlined in [4]. Figure 9
ciple for the circular tube contains calculated ratios of w,,/w,,,, as a function of
D/d, for circular tubes.
The percentage of bypass-flow near the wall is of
further interest. The bypass-flow is defined as the shaded
area in Fig. 10. For circular tubes we have calculated the
percentage of bypass-flow as a function of Did, and Re,.
It is obvious from Fig. 1 I that the bypass-flow goes
with regard to through a maximum at D/d, = 10. Beyond this ratio it
decreases again with increasing Rep and D/d,.
Comparing the computed velocity distributions inside
3tow,,
= 27r2 wiAr,ri = const. (1% the packed bed with measured profiles like those of Fig.
1 it has to be concluded that the flow profiles are heavily
3. COMPUTED VELOCITY PROFILE3 UNDER distorted after the flow has left the packed bed and has
ISOTRERMAL CONDITIONS entered the empty tube. In fact Schmid[28] was able to
There are no numerical problems concerning the solu- calculate this effect by solving numerically the two-
tion of eqns (12) or (14) for a given total flow rate V,i,, dimensional Navier-Stokes-equations for the developing
which corresponds to an average flow velocity G = flow inside the empty tube with the artificial flow profiles
$,,,/A. Some profiles calculated for flows at 25°C are of the fixed bed as inlet condition. Some of his cal-
plotted in Fig. 7. culations are plotted in Fig. 12 together with flow profiles
The characteristic features of these artificial profiles which were measured 10mm abdve the bed exit. The
are: (i) that the maximal flow rate near the wall is higher computations clearly show a dramatic decrease of the
than expected from the measurements above the bed: maximum velocity and a shift away from the wall.
and (ii) that the flow maxima lies very close to the wall at
4. THEEFFECT OF TEMPERATURE ON
about d,/4. Unlike to other measurements the shape of FIBW PROFILES
The flow distributions shown in the figures of the
proceeding section were always calculated for air passing

dp =8 mm
circular tube

01
0 5
9 1
10 15 20
I I 0 5 10 15 20
wall distance [mm1 wall distance [mm]

Fig. 7. Calculated dimensionless flow profiles for a rectangulm Fig. 8. Calculated dimensionless flow profiles for acircularpacked
packed bed. Air at 25°C: b = 40 mm; Re, = 5; dp = 2,4 and 6 mm. bed. Air at 25°C; D = 40 mm; dp = 8 mm: Ih’d, = 5.
 Evaluation of steady flow profiles in rectangular and circular packed beds 1695

l -a-@B- l-Rep=O,l -0
-a-. -O-*-Rea=l -0

0' 1 1

0 20 40 60 80 100
O/d,, -

Fig. 9. w,~,Jw,.~~~~ as a function of D/d,, for circular tubes. Air at 25°C.

0 5 10 15 20
wall distance [mm] wall distance Ido

Fig. IO. Definition of bypass-flow (shaded area). Fig. 12. Comparison of -, calculated profile inside the packed
bed; 0 l l , measured profile 1Omm above the bed; and -.-,
calculated profile IOmm above the fixed bed by Schmidt27;
Re,, = 7.5.
through the packed bed at 25°C. However, it.is obvious
from the relations (l), (2), (3) and (4) that because of the
temperature dependency of p and IJ the isothermal flow
profiles are influenced by the temperature of the flow.
40-f-l
Figure 13 exhibits this temperature effect for an average
,T : 293 K
flow velocity w = 0.1 ms-’ at temperatures of 293, 473
and 1273 K. An inspection of this figure reveals that the /T = 473K
flow near the wall decreases with higher temperature due
to increasing viscous effects. t 30- ?.- T q 1273 K

I&l

50-
I
s 40-
zo-

W = 10 cm/s
G 30-

2 20-

0 20 40 60 80
Oldp - wall distance [mm1 -
Ag. 11.Percentage of bypass-flow as a function of Rep and of D/d, Fig. 13. Calculated flow profiles corresponding to air temperatures
for circular tubes. Air at 2S”C. of 293,473 and 1273 K; bldV = 12.
16% D. VORTMEYER
and I. SCHUSTBR

5. AN APPROXlMATIVE ANALYTICAL EXPRESSION

FOR FLOW PROFILES AT 293K IN PACKED
BEDS WITH Old, OR b/d,> 5
Tien[29] has derived a general analytical expression
for flow profiles in semi-infinite packed beds which are
bounded at one side by a rigid wall. He found the general
analytical dxpression due to the porosity function near
the wall

Here x is the distance from the wall; * means the

average flow far away from the wall and w..,~ is the
actual calculated flow profile as a function of x. Un-
fortunately, the coefficients ~1 and n cannot be deter-
mined theoretically. We therefore proceeded in calculat-
ing a large number of semi-infinite profiles by the varia-
tional method as a function of x/d, and of Re,. These Ob;’ 2 3
profiles were adjusted to the general eqn (24) and it was wall distance Id,, .-e
found that the coefficient a in this equation also can be
expressed as a function of n only Fig. 14. w,,,,,il+~ and w/W.Calculated profiles for Re,, = IO.

4n
(17)
a =4-n obtained by integration of w,.,,,~ given by eqn (19) for a
particular Reynolds-number
On the other hand n turned out to be solely a function of
Re,. Making use of these empirical findings it is advis-
able to present the functional relations between Re, and
n for three ranges of Re, numbers.
ir,,,, = 2nd,,’ I?
0.1 s Re, < I
n = 112.5 -26.3lRe, + 10.97Re,2-0.1804Re,3
@a)
lSRe,<lOOO
n = -1803 + 201.62(ln Re, + 4) - 3737(ln Re, + 4)“’
<=;lr-. (22)
x .5399(ln Re, + 4)lw (18b) P

Rep > 1WO; n = 27. WC) From eqn (20) it follows that

Equation (16) can also be written for a tubular packed

3,,, = p ir,,,
bed with R +a
where
!$+=l--exp ( fy)(l-ny).
a
(19)

Consecutively, eqns (16) and (19) are corrected in such a so that

way that they are applicable to flows in packed beds of
finite extension. The procedure is illustrated by Fig. 14
with two flow profiles at Re, = 10. The full line
represents the semi-infinite profilq from eqn (19), the a
dotted curve was calculated for tubular flow with D/d, =
--L
6. Both profiles are similar in shape and differ ap- -UP. 1
-e $R*n+l+n;)
proximately by a constant factor b defined as

(20) with R* = R/d, and a = 4n/(4- n).

It is evident from eqn (23) that p depends only on R*
where BG 1. and n which is a function of Re,,. The functional relation
To compute /.S.we need the volume flow q.,,i which is fl =f(R*, Re,) is plotted in Fig. I5 for Re, = IO. For
 Evaluation of steady flow profiles in rectangular and circular packed beds 1697

Fig. 15. Deformation factor fi = f(bld,) and #J= f(ad,); &T, = 10.

large values of R* it is observed that p approaches to 012345

unity as to be expected. wall dlstonce I dp
Finally, Fig. 16 shows a comparison between the exact
Fig. 17. Porosity distributions: -, averaged: ----. real
flow profile computed by the variational method at Re, =
100 inside a tube and a profile evaluated analytically by
the method described above. The agreement is satis-
factory. Similar results are obtained at other Re,-num- the porosity changes, e.g. at a wall distance of d,,/2-
bers. where the porosity function has a minimum-there is
For rectangular packed beds we have obtained also a flow minimum and so forth. Eventually, the
porosity functions die out. This is the point where also a
p =& [&+i[l_e”‘-%‘l constant flow distribution inside the packed bed is
obtained.

+~[l-(l-a$)e”‘b’dp’]]-i (24) 7, FIDW PROmLEs UNDER NON.ISOTHERMAL

CONDITIONS

with &=x/d, and e = @(bld,,Re,). @ is plotted for There are many technical applications with consider-
Re, = loin Fig. 15. able temperature changes across the radius and along the
axis of a fixed bed. We are confronted with this situation,
6. FLOW PROFILJ3-5 WHICH CORRESPONDTO e.g. in chemical reactor theory of wall cooled exothermal
OWIUATING POROSITY FUNCTIONS chemical fixed bed reactors.
The flow distributions so far calculated were always In order to study this effect we have computed the
based on the average porosity functions of eqns (5) and flow profile in a rectangular duct under the assumption
(6). However, if one assumes that the Ergun relation for that the temperature changes linearly across the duct
the pressure loss is locally valid according to the drastic from 29&M K independent of height. Neglecting fur-
changes in porosity (Fig. 17) there is no problem to thermore any free convection effects in the presence of
include this more realistic porosity function in our varia-
tional approach. The results which are plotted in Fig. 18
show that in this case the flow pattern follows exactly

dp-4mm
3 circular tube
dp =4 mm

k Rep =lOO

I 11
w
2
dp

v d
1
-

p’ dP
0 ft=
0 5 10 15 20 012365
wall distance (mm] wall distance I dp

Fig. 16. Comparison of flow profiles evaluated: -, from varia- Fig. 18. Computed flow profiles according to the porosity dis-
tional principle; ----, from approxiinative equations. tributions of Fig. lg.
1698 D. VORTMEYER and J. SCHUSTER

30

I = 293 K

10
.A1
T=443 K T = 358K T - 298 K
20
7.5
.
Did,,= 15
I 5.0
dp =3mm
&Ll

2.5

PfdP
0 -.A 0
0 10 20 30 LO
0 15 30 45
wall distance [mm1
wall distance imm]

Fig. 19. Flow distribution within a rectangular packed bed for Fig. 20. Comparison between -, computed profile inside the
AT = 15OK temperature difference between the walls. bed; and 0 l l , measurements 10 mm above the bed for AT =
60 K.

forced convection the following equations were solved flow profiles is shown as well as the effect of a lateral
analogous to eqns (12) and (13) temperature distribution.
Approximative equations were developed for an
analytical discription of Bow profiles. Their reliability is
only proved for flow at 2S”C. The inclusion of tem-
perature effects-which are not very large-remains to
(25) further work.

and NOTATION

coefficient in eqn (6)

width of rectangular duct
he4 = I i: vpi(T)Axi = const. (26)
i-1 diameter of circular tube
particle diameter
In agreement with many more computations it I”sseen defined by relation (3)
from Fig. 19 that the changes are relatively small in defined by relation (4)
comparison to isothermal flow profiles if the differences height of packed bed
between the maximal flow velocities near the hot and coefficient in eqn (24)
cold wall are considered. pressure
However, flow measurements above the free exit radius
cross-section of the bed no longer are representative at radial coordinate
all as Fig. 20 shows. Here not the least resemblance temperature
between measurements and calculations is observed volume flow
because under non-isothermal conditions the lateral flow flow velocity
components are much higher than those for the isother- average flow rate related to the dqct or tube cross-
mal situation. section
coordinate normal to the flow
8. CONCLUSIONS factor which relates actual flow profiles in ducts or
Flow distributions within packed beds were calculated tubes to flow profiles inside a semi-infinite
by a variational principle which is equivalent to the porous medium
Brinkman equation. porosity
The calculated isothermal profiles exhibit steep max- viscosity
ima near the wall and differ considerably from those flow density
distributions which usually are measured above the constant in eqn (18)
packing. The deviations are fully explained. ll/P
Furthermore, the influence of temperature level on the = #pd,/q Reynolds-number
 Evaluation of steady flow profiles in rectangular and circular packed beds 1699

REFERENCES
[I41 Dornweiler V. P. and Fahien R. W., A.I.Ch.E._J. 1959 5 139.
[l] Buchlin I. M., Lapthom J. C. and Ginoux J. J., vt ver- II51 Morales M., Spinn C. W. and Smith I. M., Jnd. Engng Chem.
fahrenstechnik 1977 11 620. 1951 43 225.
121 Lerou J. J. and Froment G. F.. Chem. Engng Sri. 1977 32 [I61 Schwartz C. E. and Smith J. M., Ind. Engng Chem. 1953 45
853. 1209.
I31 Kalthoff 0. and Vortmeyer D., Chem. Engng Sci. 1980 35 [I71 Calderbank P. H. and Pogorski L. A., Trans. Inst. Chem.
1637. Eners 1957 35 195.
[4] Vortmeyer D. and Winter R., 7th Jnt. Symp. on Chem. I IS] Marlvoet
~ 9 J., Teodoroiu P. and Wajc S. J., Chem. Engng Sci.
Reaction Engng, Boston 1982 ACS Symp. Ser. No. 1% 1981 1974 28 1836.
49. [I91 Szekely J. and Proveromo J. I., A.J.Ch.E.J. 1975 21 769.
[5] Benenati R. F. and Brosilow C. B., A.I.Ch.E.J. 1962 8 359, [ZOIPrice J., Me& Chem. Engng Trans. May 1968 7.
[6] Schuster J., StrBmungs- und Porosit&itsverteilung in Schiit- 1211 Brinkman H. C., Appl. Sci. Res. Sect. Al 1947 27.
tungen, Dissertation, Techn. Universitlt Miinchen 1982. 1221 Ergun S., Chom. Engng Prog. 1952 48 89.
(71 Roblee L. H. S., Baird R. M. and Tiemey J. W., A.J.Ch.EJ. I231 Choudhary M., Szekely J. and Weller S. W., A.1.Ch.E.J.
1958 4 460. 1976 22 1021.
[8] Sonntag G., ,Chem.-lng.-Tech& 1960 32 327. [241 Lederbrink F.-W. and Starnick I., Chemie-Ing.-Technik 1978
191 Speck G., Uber die Randgiingigkeit in gasdurchstrijmten 50 634.
Fii1lkGrpersibde.n. Dissertation, Techn. Hochschule Dresden I251 Chandrasekhara B. C. and Vortmeyer D.. W&me- und
1955. Stofiibertragung 1979 12 105.
[IO1 Korolew V. N., Syromgatnikow N. I. and Tolmachev E. M., [26] Courant R. and Hilbert D., Methods of Mathematical Phy-
Jnrhenero-Firicheskii Zhurnal 1971 21973. sits. Vol. 1. Interscience, New York 1953.
[l I] Schuster J. and Vortmeyer D.. Chem-lng.-Tech& 1980 52 [U] Zienkiewiez 0. C. and Cheung Y. K., Engineer 1963 507.
848. [281 Schmid J., Private communication: present address: Lehrs-
[12] Cairns E. J. and Prausnitz J. M., Ind. Engng Chem. 1959 51 tuhl fiir Strijmungsmechanik TU Minchen, Arcisstr. 21.
1441. 8000 Miinchen 2.
[13] Schertz W. W. and Bischoff K. B., A.1.Ch.E.J. 1969 15 597. [29] Tien C. L., Private communication; Miinchen 1980.