You are on page 1of 9

Modeling and Simulation of an Industrial Process of

Oxygen Delignification and ECF Bleaching of


Eucalyptus Pulp
Carlos Yamamoto: Professor at Chemical Engineering Department, UFPR, Brazil,
ciyama@ufpr.br
Osvaldo Vieira: Research and Development Coordinator, Klabin S.A, Brazil,
osvaldov@klabin.com.br
Viridiane Vianna: Chemical Engineer, UFPR, Brazil, viridianevianna@uol.com.br

Abstract
The Brazilian productive sector of pulp and paper represents a relevant contribution for
the development of Brazil. To increase the competitiveness of Brazilian companies at the
international level, products must have high standards of quality and high added value. Thus,
the mathematical modeling and simulation of industrial processes ensures the stability of
production. Mathematical models that represent the Oxygen Delignification and ECF Bleaching
process of eucalyptus pulp were incorporated in the commercial simulator CADSIM Plus to
provide an optimization tool to the pulp industries. The simulator predicts the kappa number
after the oxygen delignification process, the kappa number after the OP stage of the ECF
bleaching sequence and the brightness of the bleached pulp. The results of the models for the
Oxygen Delignification process indicate that possibly there is excess oxygen in the reaction
medium, and the simulation results showed an absolute error of ± 2 kappa number. The ECF
bleaching represented in the CADSIM Plus simulator had an absolute error of ± 1 kappa
number, and ± 2 units of brightness. The results on the ECF bleaching indicate that probably
there is excess of chlorine dioxide used in the DHT stage.

Key words: Oxygen delignification, ECF bleaching sequence and CADSIM Plus simulator.

Introduction
The use of process simulators helps the mill to optimize the pulp production assuring
quality specifications that fit the clients demand with a safety and environmentally friend
process. However, this comes with a high degree of complexity, grate number of parameters
and requests tuning and validation efforts.
This study simulation was done with mathematical models that represent the oxygen
delignification and the ECF bleaching processes of a kraft Eucalyptus pulp from Klabin Papéis
Monte Alegre pulp and paper industry. The commercial simulator CADSIM Plus was used to
represent these processes by the prediction of the bleach plant quality parameters, such as:
kappa number and brightness. These processes involve complex physical-chemical phenomena
and have a high number of variables that influence direct or indirectly the reaction progress,
being important the research of better process control and optimization tools.
By the 70’s and the 80’s, the computational system started to be trustable enough to
realize process simulations. The Fortran program created the opportunity to realize fast
calculations of complex math, even with an archaic graphic interface. The solution was found, in
the stationary state, by pressure and mass flows, valve opening, pumps velocity and tanks level,
after the quality variables were calculated, like concentration, reaction rate and temperature [9].
In this same period, the Honeywell acquired the SCADA (Supervisory Control and Data
Acquisition), a process simulator that had: solutions to differential algebraic equations, flow and
pressure calculations, and the interaction with control systems. This resulted in tools like the
MASSBAL to engineer and operational training [7].
In the beginning of the 90’s, it was clear the need of the dynamic simulation to allow the
user to evaluate process variables by the time. So, CADSIM was created, which combined the
drawing interface with a new and original tool to do the process dynamic simulation. Because of
this need, the other softwares were also improved to supply this demand.

_________________________________________________________________________________
5th International Colloquium on Eucalyptus Pulp, May 9-12, 2011. Porto Seguro, Bahia, Brazil.
Nowadays, the most used pulp and paper commercial simulators are: WinGEMS, from
Metso Automation, IDEAS, from Andritz and CADSIM Plus from Aurel Systems Inc.
The WinGEMS allows the user to create dynamic or stationary simulations, by simple
graphic tools for blocks and streams that help to construct the models. The simulation could be
integrated with a financial Microsoft Excel sheet to evaluate business strategies and market
plans. Another characteristic of this simulator is the tool that creates the communication with
other programs, such as MS Excel, MS Access and PI, allowing the automatic data interchange
between applications, for in or out data [10].
The IDEAS simulator realizes mass, moment and energy balances, besides flow and
pressure calculations for complex net pipes. It can be used approximately 60 components on
the solid, liquid and gas phases. This program is also used as a platform to process modeling,
operational training, advanced control and process optimization. This simulator can model any
equipment and it is capable to communicate with DCS systems [2]. One example of the use of
this software the Aracruz Cellulose pulp industry expansion. During the implementation of this
project, the process and the DCS system were analyzed together. By that, more than 1800
errors of the DCS system were corrected and the mill reached the startup 17% above the goal,
generating millions in additional yield [4].
The CADSIM was the original draw interface to SACDA Inc.'s, the pioneer of the
stationary process simulator MASSBAL. The CADSIM Plus simulator allowed a CAD draw to
provide a data source to process simulation. By that, it was possible to draw a process
flowsheet more quickly in the CADSIM than in AutoCAD and, in the same time, a process
simulation was created [3].
The dynamic simulation can be used to process optimization, operation plan, test of the
DCS (Distributed Control System) control system and development of control strategies.
The CADSIM Plus software allows the user to create particular modules of simulation,
called DLM (Dynamically Linked Modules). Trough a pre-defined program language, ‘C++’, the
programmer can incorporate his own calculations to the DLM. One example of this usage is the
work done by Chbel e Lapierrière (2009) that utilized the software to develop one Jacobian
simulation module that generates automatic variations on manipulated variables to calculate the
process gain in different operational conditions. These gains can be used to do the parameters
tuning of the process controllers [6].

Materials and Methods


To the model development, first the flowsheets to the oxygen delignification and ECF
bleach plant were created in the CADSIM Plus simulator. In the beginning, process lines were
added, followed by process controllers. In the process lines, two patterns were considered: the
first is characterized by the presence of water, chemicals, inorganic dissolved solids, organic
dissolve solids, fiber and lignin. The second characterizes the steam line used in the process.
The pulp that goes to the oxygen delignification first passes trough a washing filter, them
goes to the delignification reactor and, to the washing stage, which is composed by a
pressurized diffuser washer in series with a washing filter. These equipments were represented
in the CADSIM Plus flowsheet with filter and pulp tanks of the process.
To specify the pulp kappa number it was necessary the creation of a derivate variable,
given by the equation below [5].

Kappa  %lignin * 6,67 (1)

Where %lignin is the lignin percent of the pulp, what is given by the ratio between the
quantity of lignin and the quantity of total fibers of the pulp.
In the main stream of the process it was specified consistency, mass pulp flow and
temperature. Regarding the inorganic dissolved solids, organic dissolved solids and lignin
content, it was defined a percent relation to the total quantity of total fiber as 0,1 %, 0,12 % e 2,7
%, respectively [5].
In the simulator library, there is a washing paste, where several types of industrial

_________________________________________________________________________________
5th International Colloquium on Eucalyptus Pulp, May 9-12, 2011. Porto Seguro, Bahia, Brazil.
washers are represented. For the washing filter, the program asks to specify the filtrate fibers
content, so it was adopted to use a relation between the filtrated fibers content and the feed
fibers content, given by Equation 2 [5].

FeedFibers
FiltrateFibers  FiltrateTotalFibers  (2)
TotalFeedFibers

Where FiltrateFibers is the mass flow of the fibers in the washing filter filtrate,
FiltrateTotalFibers is the total mass flow of the filtrate fibers, FeedFibers is the mass flow of the
fibers in the feed pulp flow and TotalFeedFibers is the total mass flow of the fibers in the feed
pulp flow.
The other characteristics of the washing filter, like filtrate’s consistency, washed pulp
consistency, dilution factor, displacement ratio are already pre-defined with typical values for this
equipment. However, it is possible to change these values to adapt the model to the existent mill
site.
To represent the oxygen delignification reactor, the lignin was set as the limit reactant.
The reaction that represents de alkali and oxygen consumption by the lignin was defined by a
global reaction that has as end products organic and inorganic dissolved solids. As the main
goal of this work is to represent the industrial process by kinetic models, it is not necessary to
represent the several reactions, with the intermediates species that occurs in the delignification,
in the mathematical model. This practical approach avoids the model to become complex and
with a slow simulation time. The global reaction is given by:

(a)lignin  (b) NaOH  (c)O2  (d )SolInDis  (e)SolOrgDis (3)

Where a, b, c, d and e are the stoichiometric coefficients of lignin, alkali, oxygen,


inorganic dissolves solids and organic dissolves solids respectively.
The stoichiometric coefficients were simulated to show representative values of the
stoichiometric coefficients of the oxygen delignification reaction of the Klabin Papéis Monte
Alegre industry.
To the kinetic equation rate, first it was adopted the model similar to the one proposed by
Violette (2003), with the parameters fitted to the process of this study [11]. Then it was used the
model proposed by Agarwal et al (1999), having the partial oxygen pressure represented by the
concentration trough the creation of a derived variable in the software [1].
In the simulation, DDE communication protocols were created to connect CADSIM Plus
and Microsoft Excel. The CADSIM Plus acts such as client or server.
For the client case, the simulator imports the data that will be used as inputs to the main
stream of the simulation, in this case, the pulp stream that goes to the wash filter before the
reactor. These data are the data acquired from the process trough PI software and Microsoft
Excel sheets. For this study, the data collected were pulp consistency, calculated production,
reactor temperature and kappa number before the reactor.
When CADSIM Plus acts as server, it sports the data of kappa number calculated after
the reactor to Microsoft Excel sheets. In these sheets, the relative error between the simulated
and real process values is calculated. Visual inspection tests are also made in these sheets to
evaluate the simulation accuracy.
Regarding the process controllers represented in the simulation, it was used one steam
flow controller to the pulp temperature before the reactor, having the temperature set point given
by the temperature imported by the DDE protocol from real process data. In the tanks, level
control was used. In the pulp stream before the pressurized diffuser washer, a consistency
control was used by the addition washing liquor from the own washer. At last, but not least, in
the filter washer after the pressurized diffuser washer, a flow control for the washing filtrate
related to the dilution factor was used.
The oxygen and alkali flows are given by a control that calculates the ratio between the

_________________________________________________________________________________
5th International Colloquium on Eucalyptus Pulp, May 9-12, 2011. Porto Seguro, Bahia, Brazil.
chemical mass flow and the pulp flow, this is also known as specific consumption.
In the other oxygen delignification line, only the wash equipments after the reactor were
changed, as this line has two washing presses. The filtrate fiber content was also given by
Equation 2. Process controls were also used for consistency and wash flow.
The ECF bleach line is composed by DHT, OP e D1 stages with washing press between
the reaction stages. The filtrate fiber content was also given by Equation 2.
Before each bleaching stage there is a mixer for the chemical reactants. For the DHT and
D1 stages there is the pulp mixture with chlorine dioxide. The simulation module considers
perfect mixture into the mixer volume.
In the simulation it was also necessary to create a derived variable to represent the kappa
number, given by Equation 1.
In this step of the process the models obtained are empirical, so math modules were
created to represent these models. Two modules were created, one for DHT and OP stages and
one for D1 stage. The output signal from the fist modules represents the kappa number after the
reactor and the output signal of the second modules represents the brightness of the bleached
pulp.
DDE communication protocols were also created to do the connection between CADSIM
Plus and Microsoft Excel. The imported data from CADSIM Plus were consistency, calculated
production rate, DHT reactor temperature, kappa numbers after digester and before DHT
reactor, chorine dioxide concentration, volumetric flows of chlorine dioxide, alkali, hydrogen
peroxide and brightness before D1 reactor. The data exported by CADSIM Plus were kappa
number after OP stage and brightness of the bleached pulp.
Level control were used in the filtrate tanks, consistency control were used in the pulp
lines before the reactors, a filtrate wash flow control related to the dilution factor was used in the
washing press after D1 stage and a steam flow to the DHT temperature control was used,
having as temperature set point the data imported by DDE protocol.

Results and Discussion


The oxygen delignification process of one line was simulated with kinetic models similar
to Violette (2003) and Agarwal et al (1999). Only the results given by the model similar to
Agarwal et al (1999) will be shown because it presented better results besides having a smaller
simulation time.
The fist simulation, constructed with the flowsheet and DDE protocol, where CADSIM
Plus acted as client, gave an absolute average error of 1,30 for the simulated kappa number
after the reactor. To the second oxygen delignification line, the absolute average error was 1,20
to the simulated kappa number.
Figure 1 shows the tendency chart of the simulated and real kappa number in function of
the sample numbers. In this figure, the values were divided by a constant to assure the
industrial data confidentiality.

_________________________________________________________________________________
5th International Colloquium on Eucalyptus Pulp, May 9-12, 2011. Porto Seguro, Bahia, Brazil.
0,5

Kappa_Cadsim
Kappa_PI
0,45

0,4

0,35
Kappa

0,3

0,25

0,2

0,15
1 50 99 148 197 246 295 344 393 442
Amostras

Figure 1 – Kappa number comparison between simulated and real values for the model similar
to Agarwal et al (1999) in CADSIM Plus.

It can be noticed that the model follows the process variations excepted for some points
that probably have influence of a transient period of the operation or represent conditions that
were not considered in the data statistical treatment. This is a new process in the mill operation
what results in process variations and non stationary operation. Even with cautiously analyses
to assure the stationary state, some data still can have errors from instrument calibration, data
acquisition, or signal processing for PI software.
With the CADSIM Plus simulator it was possible to evaluate de chemical consumption
into the oxygen delignification reactor. Figure 2 shows the values of oxygen mass flow that
enters in the reactor and the residual oxygen (the mass flow that leaves the reactor with the
pulp stream). Figure 3 represents the same mass balance but with the alkali volumetric flow
used in the process. The values presented in the graphics were also divided by a constant to
assure the industrial data confidentiality.

10

O2_in
9
O2_res

6
O2 (kg/h)

0
1 50 99 148 197 246 295 344 393 442
Amostras

Figure 2 – Oxygen mass flow for the in and out stream of the oxygen delignification reactor.

_________________________________________________________________________________
5th International Colloquium on Eucalyptus Pulp, May 9-12, 2011. Porto Seguro, Bahia, Brazil.
0,5
NaOH_in
NaOH_res
0,45

0,4

0,35

0,3
NaOH (L/min)

0,25

0,2

0,15

0,1

0,05

0
1 50 99 148 197 246 295 344 393 442
Amostras
Figure 3 – Alkali mass flow for the in and out stream of the oxygen delignification reactor.

It can be noticed from Figure 2 that the process operates with oxygen in excess and by
Figure 3 that all the alkali is consumed in the reaction.
The ECF bleach process was simulated with the empirical models. The DHT and OP
stages were represented as one and the prediction model parameter was the kappa number
after the OP reactor. For the D1 stage, the simulated parameter was the brightness of the pulp
after the reactor.
For the first simulation it was used DDE protocol to interchange data between the
CADSIM Plus software and Microsoft Excel with the purpose of evaluate the model prediction
accuracy.
The model for the DHT and OP stages presents an absolute average error of 0,54 for the
simulated kappa number after OP stage. Figure 4 shows the tendency charts the simulated and
real kappa number of the process in function of the sample numbers.

_________________________________________________________________________________
5th International Colloquium on Eucalyptus Pulp, May 9-12, 2011. Porto Seguro, Bahia, Brazil.
0,3
KappaOP_PI
KappaOP_CADSIM Plus

0,25

0,2
Kappa

0,15

0,1

0,05

0
1 50 99 148 197 246 295 344
Amostras
Figure 4 – Kappa number comparison between simulated and real values for the model of DHT
and OP stages in CADSIM Plus.

By the figure above, it is clear that the model follows the process, having some outliers
points for high kappa numbers, probably caused by some transient period or conditions that
were not considered in the statistical treatment data.
The brightness presented an absolute average error of 0,97. Figure 5 shows the
tendency chart of the simulated and real brightness of the process in function of the sample
numbers. The values presented in the graphics were also divided by a constant to assure the
industrial data confidentiality.

100

Alvura_PI
98 Alvura_CADSIM Plus

96

94

92
Alvura

90

88

86

84

82

80
1 50 99 148 197 246 295 344
Amostras

Figure 5 – Brightness comparison between simulated and real values in CADSIM Plus.

_________________________________________________________________________________
5th International Colloquium on Eucalyptus Pulp, May 9-12, 2011. Porto Seguro, Bahia, Brazil.
The models simulated by CADSIM Plus were efficient regarding the predictions of the
process quality parameters. For all cases, the simulated values followed the real values pattern.
The models were not capable of predicting high kappa numbers and low brightness values
compared to the process average. As they sporadically happen, usually caused by production
rate changes or row material variations, this does not affect the model efficiency.
The results of the second simulation done in the CADSIM Plus simulator, having the
entering data being the average usual values of the process allows qualitative analyzes of
kappa number and brightness variations.
When the chlorine dioxide amount applied to the DHT reactor is changed, the kappa
number does not showed a significant variation, probably because there is excess of this
reactant in the reaction media. The model could be improved by doing a laboratorial analyzes of
the reaction, having operational conditions assuring that the chlorine dioxide is not in excess or
by the acquisition of a new sample data when the process operates with lack of this reactant.
To evaluate de quality of the CADSIM Plus simulation, some tendency charts were done
related to the kappa number and brightness related to the chemical charge applied in the
reactors.
As the brightness also depends on the kappa number, the higher the kappa number
entering into DHT stage, higher the kappa number after the OP stage will be, what decreases
the final brightness.
Other variations caused by chorine dioxide changes into D1 stage were given by the
software.
The brightness gain starts in the delignification stage with chlorine dioxide (DHT) and the
brightness gain is higher as the chlorine dioxide substitution increases. The kappa number after
the extraction stage (OP) is linearly correlated with the chlorine dioxide charge required by the
next bleach stages to reach the final desired brightness [8].
As more chemical reactants are applied to the reaction media, they become less effective
and the reaction efficiency decreases. This is because the reaction reaches a limit beyond
which there is not any more delignification and brightness gain. The reaction limit for kappa
number and brightness is certainly connected to the lignin fraction known as non reactive lignin.
The increase in the pulp temperature before D1 stage resulted in a brightness gain, as it
was expected. However, the brightness gain is low when compared to the necessary
temperature increase, what is limited by economical factors related to the steam use. This
indicates that the process temperature is close to the optimal temperature operational point. It
must be noticed that there is chlorine dioxide excess in the reaction media, what means that the
reaction probably is stabilized, and so, the temperature variations has no influenced on the
brightness gain.

Conclusions
The main objective of this study was to create a representative simulator of the oxygen
delignification and ECF bleach industrial process using kinetic models, providing a useful toll to
realize daily simulations or to predict the results when operational changes are need.
The representative process models for the delignification reactor can be applied to
predict normalized kappa number between 1,2 to 1,6 units of kappa number. To the ECF
bleaching process, the absolute average error was ± 1 kappa number after DHT and OP stages
and ± 2 brightness for the bleached pulp.
The results of the ECF bleaching plant indicate that probably there is chlorine dioxide in
excess charged to the DHT stage.
The simulator was efficient to reproduce interactions between the process variables,
mainly related to the deviations caused by changes in the chemical flow. The simulator can also
be used to tune and analyze control strategies of the process. With this tool it was possible to
analyze if there were chemical excess in the process and to evaluate the quality of the reaction.

_________________________________________________________________________________
5th International Colloquium on Eucalyptus Pulp, May 9-12, 2011. Porto Seguro, Bahia, Brazil.
References
1. AGARWAL, S.B; GENCO, J.M.; COLE, B.J.M.; MILLER, W. Kinetics of oxygen
delignification. Journal of Pulp and Paper Science, v. 25, n.10. out. 1999.
2. Andritz Automation, Available at: <http://www.ideas-simulation.com>. Accessed on:
January/2009.
3. Aurel Systems Inc. Chemical Process Simulation Software Tools & Services. Available at:
<http://www.aurelsystems.com>. Accessed on: July/2008.
4. Bogo, A. L., Herschmiller, D. W., Hidalgo, G., Nunes. P. Utilização de simulação dinâmica
para verificação de sistema de controle da Fábrica C da Aracruz Celulose MONTE ALEGRE
e treinamento de operadores. 35° Congresso e Exposição Anual de Celulose e Papel,
ABTCP. São Paulo, Brasil, 2002.
5. CADSIM Plus, Software library. 2008.
6. Chbel, M., Laperrière, L. Development of a Jacobian module for advanced pulp and paper
simulation and control. Tappi Journal, 2009.
7. Dahlquist, E. Process Simulation for Pulp and Paper Industries: Current Practice and Future
Trend. Chemical Product and Process Modeling, Vol. 3, article 18, University of Malardalen,
Sweden, 2008.
8. Dence, C. W., Reeve, D. W. Pulp Bleaching Principles and Practices. Tappi Press, Atlanta,
GA, 1996.
9. HONEYWELL. Available at: <http://www.honeywell.com>. Accessed on: January/2009.
10. Metso’s Automation Available at: <http://www.metso.com/automation>. Accessed on:
January/2009.
11. Violette, S. M. Oxygen delignification kinetics and selectivity improvement. Theses (Chemical
Engineer Doctor Degree). University of Maine. Orono, 2003.

_________________________________________________________________________________
5th International Colloquium on Eucalyptus Pulp, May 9-12, 2011. Porto Seguro, Bahia, Brazil.

You might also like