Computers and Chemical Engineering 34 (2010) 9–16

Contents lists available at ScienceDirect

Computers and Chemical Engineering

journal homepage: www.elsevier.com/locate/compchemeng

Mathematical modeling of brown stock washing problems and their numerical

solution using MATLAB
Deepak Kumar ∗ , Vivek Kumar, V.P. Singh
Department of Paper Technology, Indian Institute of Technology Roorkee Saharanpur Campus, Saharanpur 247001, Uttar Pradesh, India

a r t i c l e i n f o a b s t r a c t

Article history: The mechanism of the displacement washing of the bed of pulp fibers is mathematically modeled by the
Received 3 September 2008 basic material balance equation. Non-linear Langmuir type adsorption isotherm is used to describe the
Received in revised form 14 August 2009 relationship between the concentration of the solute in the liquor and concentration of the solute on the
Accepted 30 August 2009
fibers. In the present study, the numerical solutions are obtained of the displacement washing model for
Available online 9 September 2009
multistage in counter current manner. For the numerical solution “pdepe” solver in MATLAB is applied
on the axial domain of the system of governing partial differential equations. Numerical solutions thus
Keywords:
obtained are in good agreement with the results of earlier workers. The technique used in the present
MATLAB, “pdepe” solver
Pulp washing model
investigation is simple, elegant and convenient for solving two point boundary value problems with
Peclet number varying range of parameters.
Adsorption isotherm © 2009 Elsevier Ltd. All rights reserved.
Multistage
Counter current

1. Introduction (macroscopic). Lapidus & Amundson (1952) have studied the effect
of longitudinal diffusion in ion-exchange and chromatographic
Modeling of pulp washing is done mainly using three columns and obtained differential equation for the wash liquor.
approaches namely: (a) process modeling, (b) physical modeling Brenner (1962) studied the washing of filter cake by neglecting the
and (c) statistical modeling. In process modeling approach, each accumulation capacity of fibers and assumed that the phenomena
stage in pulp washing operation is treated as black box. Using of longitudinal mixing and obtained model in terms of the differen-
material balances, process models express the efficiency of an indi- tial equation. Sherman (1964) has described the overall movement
vidual washing stage in terms of some performance parameters of solute in the bed of non-porous granular material with the dif-
such as displacement ratio, Norden efficiency factor, and equivalent fusion like differential equation by replacing molecular diffusion
displacement ratio. Although these models are useful for routine coefficient with longitudinal dispersion coefficient as molecular
process design calculations, but provide little information as to diffusion coefficient was found very small as compared to longitu-
how the design or operation of a washer improves its efficiency. A dinal dispersion coefficient. An additional term was used to account
complete review of the various process models used so far describ- for the accumulation (or depletion) capacity of material sorbed by
ing the pulp washing process has been presented by Pekkanen & the solids. Pellett (1966) has studied the longitudinal dispersion of
Norden (1985). solute, intraparticle diffusion of solute and liquid-phase mass trans-
Physical models describe the washing operation in terms of fer for the particles of cylindrical and spherical geometry by using
fundamental fluid flow and mass transfer principles, occurring a modified step function input. Kuo & Barret (1970) neglected the
at microscopic level during displacement washing of a fibrous longitudinal dispersion coefficient to study sodium chloride wash-
bed. These models involve parameters such as longitudinal dis- ing and obtained differential equation for the wash liquor. Grah
persion coefficient and mass transfer coefficients. Physical models (1975) has divided the packed bed of cellulose fibers into three dif-
proposed by various investigators such as Lapidus & Amundson ferent zones namely zone of flowing liquor, stagnant liquor and
(1952), Brenner (1962), Sherman (1964), Pellett (1966), Kuo & fibers. Longitudinal dispersion and mass transfer in the flowing
Barret (1970), Grah (1975) and Perron & Lebeau (1977) has been liquor zone is characterized by the differential equation. Perron &
classified based on mass transfer principles of two types (1) dif- Lebeau (1977) had taken the model equation without considering
ferential contact models (macroscopic) and (2) dispersion models the effect of longitudinal dispersion coefficient.
Most of the researcher described the washing model by cou-
pling the transport equation with various adsorption isotherms
∗ Corresponding author. Tel.: +91 9927090165; fax: +91 1322714011. to describe the relationship between the concentration of the
E-mail address: [email protected] (D. Kumar). solute in the liquor and concentration of the solute on the fibers.

0098-1354/$ – see front matter © 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.compchemeng.2009.08.005
10 D. Kumar et al. / Computers and Chemical Engineering 34 (2010) 9–16

Nomenclature

A surface area of bed (m2 )

c concentration of the solute in the liquor (kg/m3 )
C0 concentration of solute inside the vat (kg/m3 ) Fig. 1. A simple shell balance.
N0 amount of solute accumulated on the fiber surface
at the inlet (kg/m3) In the present study dispersion diffusion based transport equa-
Cs concentration of solute in the wash liquor (kg/m3 ) tions of pulp washing are developed based on the assumptions of
DL longitudinal dispersion coefficient (m2 /s) Sherman (1964) and Pellett (1966) coupled with the equation of
DV molecular diffusion coefficient (m2 /s) mass transfer, i.e. Fick’s second law of diffusion. The equations are
A, B Langmuir constants (m3 /kg) coupled with many other fluid mechanical parameters and finally
L cake thickness (m) two models are obtained, differing in boundary conditions. Non-
n concentration of solute on fibers (kg/m3 ) linear Langmuir type adsorption isotherm given by Fogelberg &
t time (s) Fugleberg (1963) is used for equilibrium between the concentra-
C dimensionless concentration of solute in the liquor tion of the solute in the liquor and concentration of the solute
N dimensionless concentration of solute in the fiber on the fibers. The above mentioned mathematical models of two
Z dimensionless distance simultaneous partial differential equations with various bound-
T dimensionless time ary conditions are extremely intricate in nature and practically
 mean time appear to be unsolvable for multistage washing system even by
u liquor speed in cake pores (m/s) using sophisticated numerical techniques. It is important to men-
z variable cake thickness (m) tion that the problem for single washer with its simplified version
z small increment in cake thickness (m) has been solved analytically by Brenner (1962) and Kukreja (1996)
Cyi inlet vat consistency of the pulp (kg of fiber/kg of using Laplace transform and numerically using orthogonal colloca-
liquor) tion by Grah (1975) and Arora, Dhaliwal, and Kukreja (2006a) and
Cyd discharged consistency of the pulp (kg of fiber/kg of Arora, Dhaliwal, and Kukreja (2006b). Kumar (2002) attempted to
liquor) solve the washing model using Finite difference method. All these
xi dissolves solids inside the vat (%) methods are very complex and time consuming. Similar problem
xs dissolved solids in washing water (%) for a single stage advection-dispersion washing has been solved by
xf dissolved solids in the filtrate (%) Singh, Kumar, and Kumar (2008) by pdepe solver easily by using
xd dissolved solids in discharged pulp (%) much less time and give the similar results as analytical solution
xr dissolved solids in recycle liquor (%) given by Brenner (1962). The present work focuses on the four
Li amount of liquor inside the vat (kg of liquor/kg of stages counter current washing system described by the math-
pulp) ematical model for washing zone of a rotary drum washer and
Ls amount of wash water (kg of water/kg of pulp) washing theory based on Langmuir adsorption isotherm. Pdepe
Ld amount of liquor in discharge pulp (kg of liquor/kg solver in MATLAB source code is used to solve the model of washing
of pulp) zone for multistage counter current cake washing.
Lf amount of filtrate (kg of liquor/kg of pulp)
Lr amount of liquor recycled to previous washer (kg of 2. Description of mathematical models for cake washing
liquor/kg of pulp) zone

The mat of pulp fibers can be assumed to be stationary packed

bed of homogeneous symmetrical cylindrical fibers. Instantaneous
Sherman (1964), Pellett (1966), Neretnieks (1974), Grah (1975)
behavior of any system of this type can only be expressed by an
and Viljakainen (1985) have considered the linear or non-linear
equation involving the variables and their partial derivatives. Using
adsorption isotherm equations along with the dispersion diffusion
simple material balance for setting up a differential equation, con-
based transport equations. Towers & Scallan (1996) used Donnan
sider a thin slice of a filter cake (pulp mat) as shown in Fig. 1, through
equilibrium theory to characterize the distribution; a mathematical
which filtrate or wash water flows.
model has been developed to describe the partitioning of cations
Material balance across the simple shell given in Fig. 1, in the z
between fiber walls and surrounding liquor. Tervola & Rasanen
direction can be written as
(2006) described the cake washing of freely mobile ions by an
advection-dispersion equation combined with Donnan equilibrium rate of mass of solute in
and an overall ion transfer model between the external liquid-
+rate of mass production by chemical reaction = rate of mass out
phase and the fiber phase of the kraft pulp.
A typical washing system in industry consists of three or four + rate of mass accumulation in the liquid phase
washers in counter flow arrangement. Counter flow is an engineer-
ing technique where in two process streams interact as they move +rate of mass accumulation in the solid phase due to
in opposite directions. As applied to pulp washing this means a adsorption-desorption.
series of washers is set up with the final wash being performed with
clean water. The wastewater from the last washer is then used to
wash the stock in the second from the last washer. Water from the If A is the area of the bed, εt the total average porosity (sum of
second to last washer is used to wash stock in a third to last washer porosities in the displaceable liquid εd and in the immobile phase
and so on for the total number of washers used. Tervola (2006) εs ), u the velocity of the liquor in the mat, c the concentration in
developed a Fourier series solution method for solving a multi- the liquid-phase, the equation in one dimension can be written as
stage counter current cake washing problem, so that the solute  
∂
concentration gradient inside a cake between washing stages was (ucεt A )z,t −(ucεt A )z+z,t = {(cεt A z + n(1 − εt )A z} (1)
∂t
preserved. z̄,t
 D. Kumar et al. / Computers and Chemical Engineering 34 (2010) 9–16 11

Table 1
Existing mathematical models for washing zone used in present investigation (dimensionless form).

S. no. Transport equation Adsorption isotherm Boundary conditions


1. 2 2
(∂ C/∂Z ) = Pe(∂C/∂Z + ∂C/∂T +  ∂N/∂T) N = ABC0 C/(1 + BC0 C)N0 (∂C/∂Z) = Pe(C − Cs /C0 ) for (Z = 0, T > 0) and (∂C/∂Z) = 0 at (Z = 1)
2. (∂2 C/∂Z2 ) = Pe(∂C/∂Z + ∂C/∂T +  ∂N/∂T) N = ABC0 C/(1 + BC0 C)N0 (C = Cs /C0 ) for (Z = 0, T > 0) and (∂C/∂Z) = 0 at (Z = 1)

where z < z− < z + z. Taking εt and A as constant and taking the and assumed the liquid solid concentration inside the fibers and
limit as z → 0, one can obtain the following expression: surrounding the fibers to be identical at any time and at any
        position within the bed, implying that diffusion, both within the
∂u ∂c ∂c ∂n fiber and between the fiber and the surrounding fluid is suffi-
−εt c = εt u + εt + (1 − εt ) (2)
∂z ∂z ∂t ∂t ciently rapid which does not affect the rate of the overall transport
process.
The above equation contains principally two accumulation terms,
Perron & Lebeau (1977) used the isotherm equation, i.e.:
one related to dispersion–diffusion and another related to
adsorption–desorption. Other terms are velocity gradient and con- ∂n
vective flow terms. Using Fick’s second law of diffusion, i.e.: = k(c − n) (8)
∂t
   
∂u ∂2 c the diffusion of the solute within the fibers towards the washing
−c = (DL + DV ) (3)
∂z ∂z 2 liquor is described by a partial differential Eq. (8), which is solved
assuming that the mass transfer rate through the stagnant film is
the following equation is obtained: finite.
        Fogelberg & Fugleberg (1963) used non-linear Langmuir type
∂2 c ∂c ∂c 1 − εt ∂n
(DL + DV ) =u + + (4) adsorption isotherm to describe the relationship between the con-
∂z 2 ∂z ∂t εt ∂t
centration of the solute in the liquor and concentration of the solute
According to Sherman (1964) the longitudinal dispersion coeffi- on the fibers as
cient DL is a function of flow pattern within the bed (unless very low ABc
flow rates are used). The molecular diffusion coefficient DV is very n= (9)
1 + Bc
small compared to DL and so may be neglected. Writing (1 − εt )/εt
as  for convenience, Eq. (4) may be written as where A and B are Langmuir constants.
        Arora et al. (2006b) successfully used non-linear Langmuir type
∂2 c ∂c ∂c ∂n adsorption isotherm using one dimensional axial dispersion model
DL =u + + (5)
∂z 2 ∂z ∂t ∂t for single stage washing, by using orthogonal collocation on finite
elements. Thus in the present investigation for sodium species,
This is a non-homogeneous, non-linear, first degree, second order, non-linear Langmuir type adsorption isotherm is used for the solu-
parabolic, partial differential equation. Here u, εt and DL are func- tion of four stage counter current washing problem.
tions of z while c and n are functions of both z and t. As the lumen
of the fiber is porous and the same is true with the wall of the fiber,
2.2. Initial and boundary conditions
the porosity values for these cases are different from the porosity
of the interfiber mass. Therefore three porosity values are required
For the solution of washing models for multistage the initial
to represent the pulp mat system. It is extremely difficult to dis-
condition is c(z, t) = n(z, t) = C0 for 0 < t < L/u, where L/u corresponds
tinguish precisely between the values of porosity at the lumen and
to displacement time. Boundary conditions for Langmuir type
at the wall. Therefore, for practical calculations these are assumed
isotherm are as follows:
to be the same. Hence, to describe the system two porosity values
are assumed, one for the interfibers εd and another for intrafibers ∂c
uc − DL = uCS at z = 0 and t > 0 and
εs, so that εd + εs = εt , the total porosity for the entire system. The ∂z
model Eq. (5) is same as dispersion model for pulp washing given  
∂c
by Sherman (1964) and Pellett (1966). = 0 at z = L and t > 0
∂z
2.1. Adsorption isotherms Perron & Lebeau (1977) give the boundary condition at the inlet of
the bed c = Cs , at z = 0 and t > 0.
The details of the adsorption isotherms which are used by some In the present investigation both the cases of inlet boundary
earlier workers are as follows. conditions are used. Thus two sets of model are obtained for dif-
Lapidus & Amundson (1952) used the adsorption isotherm given ferent boundary conditions. The initial condition is same for both
by the models, i.e. C(Z, 0) = 1 = N(Z, 0) in dimensionless form. Thus pos-
∂n sible models which are differing in boundary conditions only are
= k1 c − k2 n (6) summarized in Table 1 (dimensionless form).
∂t
and assumed that the rate of adsorption is finite and plotted the
effect of longitudinal diffusion for an infinite column in which equi- 2.3. Dimensionless models
librium is established locally. Initial adsorbate concentration was
assumed to be zero. Singh et al. (2008) used this linear isotherm The dimensionless form of the models is obtained by using cer-
successfully and give comparable results. tain dimensionless parameters like Peclet number (or Bodenstein
Sherman (1964) used the adsorption of diacetyl solution by number), dimensionless time, dimensionless thickness and dimen-
porous viscous fibers with simple isotherm equation, i.e.: sionless concentrations given below:

∂n ∂c uL ut z c n N0
n = kc or =k (7) Pe = ,T = ,Z = ,C = ,N = and  =
∂t ∂t DL L L C0 N0 C0
12 D. Kumar et al. / Computers and Chemical Engineering 34 (2010) 9–16

Fig. 2. Flow diagram of a counter current washing system with four drum washers.

Thus the dimensionless form of the transport equation, adsorption ranges (generally 80–100). Poirier et al. (1987) categorically men-
isotherm and boundary conditions are given in Table 1. tioned that displacement efficiency decreases below Pe number 20
and therefore one expects inconsistencies in the profiles. An order
2.4. Steady state modeling for multistage washing process of estimate values of Pe therefore should be predicted. As washing
efficiency has been expressed by Potucek (1997) as a function of
In the present investigation a typical four stages counter current Pe, it becomes more relevant to evaluate for washer performance
washing system is considered. A series of four washers are set up study. The modified Peclet number, Pem and modified dispersion
for pulp washing with the final wash being performed with clean coefficient DLm are then calculated as
water. The steady state material balance equations can be obtained uL
for each washer, with the help of Fig. 2. Pem =
DLm
Steady state mass balance equations for washer 1 are given
below. The algorithmic procedure given by Kumar (2002) is used for
First washer mass balance: estimation of Peclet number. According to Kumar (2002), first we
obtain the C vs. T data from this present investigation for both
Liquor : Lb +Lr1 +Ls1 =Lf 1 +Ld1 , Lf 1 =Lb1 +Le , Lb1 =Lb2 +Lr1 , Li1 =Lr1 +Lb the models by using the Peclet number 71.26 for washer no. 1 is
(10) obtained from the data of Grah (1974), which is used for the sim-
ulation. Then the Peclet number for displacement zone of second
Solids : Lb xb + Lr1 xr1 + Ls1 xs1 = Lf 1 xf 1 + Ld1 xd1 , Lf 1 xf 1 = Lb1 xb1 washer is estimated approximately avoiding iterative calculations
and considering open vessel as used by Han & Edwards (1988),
+ Le xe , Lb1 xb1 = Lb2 xb2 + Lr1 xr1 , Li1 xi1 = Lr1 xr1 + Lb xb (11) Edward, Peyron, and Minton (1986) and Potucek (1997).
First we use the data of Grah (1974) for the kappa no. 49.7 and
Fiber : Lb Cyb = Ld1 Cyd1 (12) for this data we obtain the Peclet number for the first washer of the
series by using the formula, Pe = uL/DL . After that we use the fol-
Water : Lb (1 − Cyb )(1 − Xb ) + Lr1 (1 − Xr1 ) + Ls1 (1 − Xs1 ) lowing algorithm given by Kumar (2002) for calculating the Peclet
number for the subsequent washers.
= Lf 1 (1 − Xf 1 ) + Ld1 (1 − Cyd1 )(1 − Xd1 ) (13)
• obtain C vs. T data for the first washer by solution of the model;
In a similar manner mass balance equations are obtained for • find out the mean time  i = (˙Ti Ci Ti )/(˙Ci Ti );
each washer and then solved, using actual data of a near by pulp • find the spread of the distribution, measured by the
and paper mill. The steady state operational data is given below: variance
 representing  the spread of the distribution
 2 = ( Ti2 Ci Ti )/( Ci Ti ) − i2 ;
(a) Pulp yield = 47%. • fit the dispersion model for the large extent of the dispersion for
(b) Consistency of blown pulp = 13%. open vessel 2 = ( 2 /i2 ) = 10(DL /uL) = (10/Pe).
(c) Solids in blown pulp = 22%.
(d) Liquor in blown pulp = 6.69 kg of liquor/kg of pulp. 3. Result and discussion
(e) Standard consistency = 12%.
The mathematical model of pulp washing given by transport
2.5. Algorithm to calculate approximate Peclet number for the equation (Eq. (5)), combined with the corresponding equation of
displacement zone for second and subsequent washers isotherm and various boundary conditions is given in Table 1.
The model is extremely intricate in nature and practically appears
As varying flow conditions can be assumed in different parts to be unsolvable for multistage washing system, even by using
of the pulp bed, an average value of the dispersion coefficient is sophisticated numerical techniques. As mentioned earlier that the
required to be estimated. This is important as various authors simu- problem for single washer with its simplified version has been
lated with various ranges of Peclet number though the pulp quality solved analytically by Brenner (1962) and Kukreja (1996) using
remains almost the same. As for example Potucek (1997) used very Laplace transform and numerically using orthogonal collocation
low Pe (1.0–11.3) for pulp fibers and high for glass fibers following by Grah (1975) and Arora et al. (2006a). Arora et al. (2006b)
Brenner (1962), whereas Grah (1974), Poirier, Crotogino, Trinh, and first discretized the partial differential equations into differential
Douglas (1987) and Crotogino, Poirier, and Trinh (1987) used high algebraic equations, which are then solved using ODE15s solver
 D. Kumar et al. / Computers and Chemical Engineering 34 (2010) 9–16 13

Table 2
Process data of four stage brown stock washing system.

Input parameters Washer no. 1 Washer no. 2 Washer no. 3 Washer no. 4

Cyi (%) 1.25 1.25 1.25 1.25

Cyd (%) 12.00 12.00 13.00 14.00
xi (%) 15.59 7.29 2.62 0.96
xs (%) 7.00 2.50 0.90 0.00
xf (%) 15.00 7.00 2.50 0.90
xd (%) 10.12 3.77 1.58 0.30
xr (%) 15.00 7.00 2.50 0.90
Li (kg/kg) 79.00 79.00 79.00 79.00
Ls (kg/kg) 10.33 10.33 9.69 9.14
Lf (kg/kg) 82.00 82.00 82.00 82.00
Ld (kg/kg) 7.33 7.33 6.69 6.14
Lr (kg/kg) 72.31 71.67 71.67 72.31

separately. Kumar (2002) attempted to solve the washing model

using finite difference method. While Tervola (2006) developed a
Fourier series solution method for solving a multistage counter cur-
rent cake washing problem. All these methods are very complex
and time consuming. Application of such solution techniques in
control systems is not possible due to the more processing time
and involvement of high mathematical skills at operator level.
For control purpose the transient behavior of the solute concen-
tration in the black liquor is of more interest, rather than solute
concentration in fiber. In the present work value of ∂N/∂T in terms
of ∂C/∂T, C, C0 , N and N0 is obtained by differentiating the isotherm
equation, and then substituted in the transport equation (given in
Table 1).
For the solution of washing zone models for all four washers,
steady state data is observed from a paper mill (given in Table 2) and
values of various process parameters for washing zone of a washer
is taken as given by Grah (1974) (given in Table 3) for sodium ion
species. The washing liquor for the washers’ 1, 2 and 3 contains the
dissolved solid 7.00%, 2.50%, 0.9% and in the last washer, fresh water
is used which also contains 0.5% dissolved solids. The incoming pulp
from the digester has the consistency 12%.
In this investigation dimensionless bed depth as well as dimen- Fig. 3. Solution for washer 1 (Model 1).

sionless time is divided into 21 equal parts and then the influence
of Z and T on C is estimated. The behavior of exit solute concen- been included as dimensionless parameters in the solution of the
tration with respect to time as well as variable cake thickness is models. For the washer no. 1 we used the Peclet number Pe = 71.26
shown by (input curve at Z = 0 to break through curves at Z = 1) 3D based on the simulation data. Since the dimensionless time for the
graphs. The solution obtained for all washers’ are given in Table 4, first and second washers is 10.33/7.33 = 1.41, for the third washer
for both the models, respectively. The variations in the dimension-
less solute concentration with respect to dimensionless time as well
as dimensionless cake thickness are shown in Figs. 3–6 for model
1 and Figs. 7–10 for model 2 for all four washers’, respectively.
In the present investigation Langmuir type non-linear isotherm
is used to describe the relationship between the concentration of
the solute in the liquor and concentration of the solute on the fibers.
Dimensionless initial solute concentrations on fiber for all four
washers’ is obtained are 0.009, 0.002, 0.0002 and 0.00001 (kg/m3 ),
respectively. The solution of the washing zone models of a rotary
vacuum washer is obtained for multistage in counter current man-
ner by pdepe solver in MATLAB source code. Peclet number has

Table 3
Data for simulation for sodium species.

Parameters Values Unit

L 0.105 m
DL 10.8 × 10−7 m2 /s
εt 0.928 –
u 7.33 × 10−4 m/s
C0 0.570 kg/m3
Cs 0.005 kg/m3
A 0.01263 m3 /kg
B 3.955 m3 /kg Fig. 4. Solution for washer 2 (Model 1).
14 D. Kumar et al. / Computers and Chemical Engineering 34 (2010) 9–16

Table 4
Dimensionless exit solute concentration in for washer for dimensionless time T = 0 to T = 1.49.

T Model 1 Model 2

Washer no. 1 Washer no. 2 Washer no. 3 Washer no. 4 Washer no. 1 Washer no. 2 Washer no. 3 Washer no. 4

0.1 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000

0.2 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000
0.3 1.0000 0.9998 0.9997 0.9997 1.0000 0.9998 0.9997 0.9996
0.4 1.0000 0.9964 0.9949 0.9946 1.0000 0.9963 0.9947 0.9944
0.5 0.9991 0.9769 0.9704 0.9695 0.9991 0.9764 0.9696 0.9686
0.6 0.9919 0.9211 0.9067 0.9046 0.9919 0.9297 0.9049 0.9027
0.7 0.9596 0.8195 0.7985 0.7954 0.9594 0.8172 0.7957 0.7925
0.8 0.8708 0.6832 0.6604 0.6568 0.8704 0.6803 0.6569 0.6533
0.9 0.7085 0.5357 0.5153 0.5118 0.7080 0.5327 0.5118 0.5082
1.0 0.4997 0.3990 0.3827 0.3795 0.4993 0.3963 0.3796 0.3762
1.1 0.3039 0.2856 0.2732 0.2702 0.3037 0.2834 0.2706 0.2675
1.2 0.1682 0.1993 0.1892 0.1863 0.1681 0.1976 0.1872 0.1842
1.3 0.0991 0.1377 0.1284 0.1254 0.0990 0.1365 0.1269 0.1239
1.4 – – – 0.0831 – – – 0.0820
1.41 0.0733 0.0925 – – 0.0733 0.0918 – –
1.45 – – 0.0706 – – – 0.0697 –
1.49 – – – 0.0570 – – – 0.0563

Fig. 5. Solution for washer 3 (Model 1). Fig. 7. Solution for washer 1 (Model 2).

Fig. 6. Solution for washer 4 (Model 1). Fig. 8. Solution for washer 2 (Model 2).
 D. Kumar et al. / Computers and Chemical Engineering 34 (2010) 9–16 15

tance which is same as obtained by earlier workers such as Brenner

(1962), Grah (1975), Kumar (2002) and Arora et al. (2006a).

4. Conclusion

The present investigation is for multistage counter current

washing system based on mathematical models derived for wash-
ing zone of a rotary vacuum washer in paper industry. Langmuir
adsorption isotherm is used to describe the relationship between
the concentration of the solute in the liquor and concentration of
the solute on the fibers. The numerical solution are obtained for four
stages in counter current manner by using “pdepe” solver in MAT-
LAB source code and taking Pe = 71.26 for first washer based on the
simulation data. Peclet numbers for second and subsequent wash-
ers are obtained by using the algorithm given by Kumar (2002). The
following conclusions may be drawn from the present study:

1. Boundary conditions do not influence significantly on the wash-

ing results.
2. Pdepe solver can be used successfully for the solution of multi-
stage pulp washing model.
Fig. 9. Solution for washer 3 (Model 2).
3. The pdepe solver used in the present investigation is simple,
elegant and convenient for solving two point boundary value
problems with varying range of parameters and show a compa-
rable performance with QUICK method in terms of CPU time and
average numerical errors.
4. The algorithms in this solver are easy to set up, and so the method
represents an advantage and good alternative to the available
techniques for such type problems.

References

Arora, S., Dhaliwal, S. S., & Kukreja, V. K. (2006a). Simulation of washing of packed
bed of porous particles by orthogonal collocation on finite elements. Computers
and Chemical Engineering, 30, 1054–1060.
Arora, S., Dhaliwal, S. S., & Kukreja, V. K. (2006b). Modelling of the displacement
washing of pulp fiber bed. Indian Journal of Chemical Technology, 13, 1–6.
Brenner, H. (1962). The diffusion model of longitudinal mixing in beds of finite
length. Chemical Engineering Science, 17, 229–243.
Crotogino, R. H., Poirier, N. A., & Trinh, D. T. (1987). The principals of pulp washing.
Tappi Journal, 70(6), 95–103.
Edward, L., Peyron, M., & Minton, M. (1986). Models for cross-flow pulp washing
calculations. Pulp and Paper Canada, 87(1), T17–T21.
Fogelberg, B. C., & Fugleberg, S. (1963). A study of factors influencing the amount of
residual alkali in sulphate pulp. Paprija Puu—Papper och Tra, 45, 675–680.
Grah, L. E. (1974). Washing of cellulose fibers analysis of displacement washing
operations. PhD Thesis, Chalmers University of Technology, Goteborg, Sweden.
Fig. 10. Solution for washer 4 (Model 2). Grah, L. E. (1975). Displacement washing of packed beds of cellulose fibers. Part 1.
Mathematical model. Svensk Paperstidning, 78(12), 446–450.
Han, Y., & Edwards, L. (1988). Optimization of filter washer operation and control.
is 9.69/6.69 = 1.45 and finally for the last washer is 9.14/6.14 = 1.49. Tappi Journal, 71(6), 101–104.
Thus the exit dimensionless solute concentration for washer no. 1 Kuo, M. T., & Barret, E. C. (1970). Continuous filter cake washing performance. AIChE
Journal, 16(4), 633–638.
and 2 is taken at T = 1.41 and similarly for third and fourth washers Kukreja, V. K. (1996). Modeling of washing of brown stock on rotary vacuum washer.
at T = 1.45 and 1.49, respectively. The exit dimensionless solute con- PhD Thesis, University of Roorkee, Roorkee, India.
centrations for all four washers with respect to dimensionless time Kumar, M. (2002). Mathematical modeling of pulp washing systems and solutions.
PhD Thesis, Indian Institute of Technology Roorkee, Roorkee, India.
are shown in Table 4, for both the models, respectively. Exit dimen- Lapidus, L., & Amundson, N. R. (1952). Mathematics of adsorption in beds. Part vi. The
sionless solute concentrations for each washers’ are converted into effect of longitudinal diffusion in ion exchange and chromatographic columns.
absolute concentrations. The exit absolute concentration of previ- Journal of Physical Chemistry, 56, 984–988.
Neretnieks, I. (1974). Mathematical model for continuous counter current adsorp-
ous washer is use to calculate the various parameters like N0 ,  , etc. tion. Svensk Paperstidning, 77(11), 407–411.
for the subsequent washer. Based on C vs. T data of previous washer Pellett, G. L. (1966). Longitudinal dispersion, intraparticle diffusion and liquid-phase
the Peclet number of subsequent washer is obtained by using algo- mass transfer during flow through multiparticle systems. Tappi Journal, 49(2),
75–82.
rithm given by Kumar (2002). The Peclet number obtained by this
Pekkanen, M., & Norden, H. V. (1985). Review of pulp washing models. Paperija Puu,
algorithm for washers nos. 2, 3 and 4 are 21.94, 19.90 and 19.70 for 67(11), 689–696.
model 1 and 21.93, 19.90 and 19.67 for model 2, respectively. Perron, M., & Lebeau, B. (1977). A mathematical model of pulp washing on rotary
drums. Pulp and Paper Canada, 78(3), TR1–TR5.
The washing results for both the models are slightly different at
Poirier, N. A., Crotogino, R. H., Trinh, D. T., & Douglas, W. J. M. (1987). Displace-
three or four decimal places, so it may conclude that the bound- ment washing of wood pulp—A one parameter model for high initial liquor
ary conditions do not influence much on the washing results. It concentration. In Pulp washing symposium, Helsinki, 18–21 May (pp. 49–72).
is clear from figures that the dimensionless concentration of the Potucek, F. (1997). Washing of pulp fiber bed. Collection of Czechoslovak Chemical
Communications, 62, 626–644.
solute in the liquor decreases with the increase of the dimensionless Sherman, W. R. (1964). The movement of a soluble material during the washing of
time, whereas increases with the increase in the dimensionless dis- a bed of packed solids. AIChE Journal, 10(6), 855–860.
16 D. Kumar et al. / Computers and Chemical Engineering 34 (2010) 9–16

Singh, V. P., Kumar, V., & Kumar, D. (2008). Numerical solution of diffusion model Tervola, P. (2006). Fourier series solution for multistage counter current cake wash-
of brown stock washing beds of finite length using MATLAB. In Proceeding of ing and segregated wash effluent circulation. Chemical Engineering Science, 61,
“Second UKSIM European Symposium on Computer Modeling and Simulation”, IEEE 3268–3277.
Digital Library (pp. 295–300). Viljakainen, E. (1985). Mass transfer models and multizone washing efficiency cal-
Towers, M., & Scallan, A. M. (1996). Predicting the ion-exchange of kraft pulps using culations. Pulp and Paper Canada, 86(10), T301–T305.
Donnan theory. Journal of Pulp and Paper Science, 22(9), J332–337.
Tervola, P., & Rasanen, E. (2006). A cake washing model with an overall cation
transfer in kraft pulp washing. Chemical Engineering Science, 61, 3268–3277.