PROCESS CONTROL
Model predictive control
of the chip level in a continuous
pulp digester, a case study
By T. Lindgren, T. Gustafsson, H. Forsgren, D. Johansson and J. Östensson
Abstract: This paper describes the digester chip level control problem and shows a MPC system
implementation for the level control.
The objective with implemented chip level control system is to stabilize the transport conditions
in the bottom of the digester by using the chip feed and not the blow flow as the primary manipulated variable. The results indicate a smoother digester operation and a reduction of the kappa
number variations compared with the original control system.
he continuous digester (The Kamyr
digester) is a complex tubular reactor
where the delignification of the wood
chips takes place through combined
chemical treatment and thermal
effects. The digester consists of three important
zones: the impregnation zone, the cooking zone
(delignification reaction) and the washing zone
where the counter flow of the bottom dilution
flow slows done the reaction and removes residual chemicals and lignin. The key parameter for
monitoring the digester system performance is
the kappa number, which represents the residual
lignin in the pulp.
The kappa number is influenced by the following variables, Svanholm [5], Lundqvist [10]
• Chip quality (humidity, mixture, species, size,
etc.)
• The quantity of Effective Alkali (EA) in the
chips
• The temperature that chips are exposed to
• The time the chips are exposed to EA and temperature (the time in the delignification process)
The statements above are valid when looking
at the delignification process at the chip scale,
but it is very important to also consider how the
variables above are behaving when the chips are
transported through the digester process (the
digester scale). It is important to understand the
general sequence of transport and reaction processes that govern the overall digester operation.
Stable transport conditions at every subpart in the
digester system are essential for having good conditions for the chemical and thermal treatment of
the chips. The kappa number variability is related
to variations in chip retention time in the
digester, caused by fluctuations in chip packing,
blow line consistency, etc. These variations can
be, to some extent, detected by measurements of
chip level, blow line differential pressure, and
bottom scraper amperage.
In this paper a number of reports on digester
chip level control have been analysed regarding the
control strategy and controller structure. A chip level control strategy (two-vessel Kamyr digester) is
proposed, aiming to stabilize digester bottom conditions. The strategy is implemented as a MPC controller scheme and as a PID-solution in DCS.
T
T. LINDGREN
Luleå University of
Technology, Luleå, Sweden,
Network for Process Intelligence (NPI), Ö-vik, Sweden
Eurocon AB, Örnsköldsvik,
Sweden
[email protected]
T. GUSTAFSSON
Luleå University of
Technology
Luleå, Sweden
H. FORSGREN
Eurocon AB
Örnsköldsvik, Sweden
D. JOHANSSON
M-real, Husum Mill
Husum, Sweden
J. ÖSTENSSON
M-real, Husum Mill
Husum, Sweden
46 ❘❘❘ 106:11 (2005)
The MPC scheme is straightforward regarding
the handling of constraint variables in the control
problem. By combining the controlled and constraint variables, it is possible to set-up and tune
the MPC controller both to control the level to a
set point and to have upper and lower limits to
ensure the production level, i.e. to keep the chip
level at an average value (set point) but at the
same time allow the level to fluctuate within certain limits (as a buffering functionality). The
MPC controller is implemented on a continuous
digester at the M-real Husum mill, Sweden and
has been running since the end of 2003. The
digester operators state that digester system has
operated more smoothly during this winter period with the updated level control system, compared to last year’s operation.
Another purpose of this work is to implement
a multivariable model-based controller, such as a
MPC controller, on a selected pulp and paper
process. The selected process (the digester system) and the MPC controller will be used as a test
system for the development of a monitoring and
diagnosis system for multivariable and multi-loop
control systems.
THE CHIP LEVEL CONTROL PROBLEM
The primary objective of using a digester supervisory control system is to minimize the variation
in the kappa number and to improve (stabilize)
digester operation. The chip level control has an
important effect on that objective. It is important to understand how the variations in the
chip level itself are influencing the chip cooking
process, and also how the manipulated variables,
used for controlling the level, are influencing
the digester process.
A number of papers have been written regarding the control of the Kamyr digester chip level.
The following section is an attempt to review the
control strategies, controller structures, Manipulated Variable (MV) and Controlled Variable (CV)
selection, etc., described in those papers.
Fiber discharge rate as MV
• Sastry [12] designed a Self Tuning Regulator
(STR) which used the blow flow as the primary
MV. The chip flow and bottom scraper speed
T 239 Pulp & Paper Canada
PROCESS CONTROL
TABLE I: Summary of reviewed papers. FF = Feed-Forward
Signal.
FIG. 1. Digester system at M-real mill, Sweden.
were used to deal with extreme level conditions. No primary control of the blow
line consistency was presented. The
results showed a decrease in chip and
black liquor flow variations to the digester
by manipulating blow flow compared to
the original control system. According to
Belanger et al. [9] this control system was
abandoned because of bad reliability of
the STR, but Belanger mentioned nothing about any problems with the selected
control strategy.
• Belanger et. al. [9] made a comparison
between using the blow flow and the
digester bottom scraper as the primary
manipulated variable for the chip level
control (using a STR control structure).
The two alternatives showed similar
results regarding top separator amperage
(level measurement) variations but the
digester bottom scraper (outlet device)
was selected because of different practical
reasons. The operators also preferred the
bottom scraper as the primary MV
because of the more steady blow flow.
Controlling dilution flow, where the set
point was ratioed to the scraper set point,
aimed to stabilize the blow flow consistency. The dilution controller output was
slowly changing the blow flow set point.
The STR controller increased the stability
of the production.
• Allison et al. [7] designed an adaptive
controller based on a generalized predictive control structure (GPC). The blow
flow was used as the primary manipulated
variable. No clear motive was presented
regarding why the blow flow was selected.
The blow flow set point is set by the GPC
together with the production feed forward
signal and a proportional feedback signal
from the digester top separator current.
No primary control of the blow line consistency was presented. The adaptive controller showed improvement of the chip
level control by means of a closer mean
value to level target and a reduction of the
standard deviation of the chip level signal.
Pulp & Paper Canada T 240
Chip feed rate and fibre discharge rate as
MV
• Fuchs et al. [13] used the chip metering
speed as the primary manipulated variable for the chip level control. They pointed out that it is important to have a stable
fibre discharge rate in order to stabilize
the residence time in the cooking zones
of the digester. The manipulations of the
blow flow were minimized, but a very slow
feed-forward signal was added to the fibre
discharge target (blow flow target) based
on the difference between the nominal
production and the actual production i.e.
the chip meter speed. Manipulating the
digester bottom scraper, with some correction of the bottom dilution flow when
the scraper speed was at its limits, primarily controlled the blow line consistency.
No specific results were shown regarding
the level control performance.
• Al-Shaikh et al. [11] presented a control
strategy where the chip meter speed was
used as the primary manipulated variable
for short-term level variations, and the
blow flow ratio, for long term variations.
The chip meter speed sets the production
target, with a limited bias added for the
level control. When the accumulated
error for the chip meter speed (from the
nominal production speed) exceeds a
certain limit the blow flow ratio target is
adjusted to maintain target production.
The blow line constistency was not directly controlled.
• Allison et al. [6] used both the chip
meter speed and the blow flow as manipulated variables. The chip meter and the
blow flow simultaneously reacted to level
disturbances, but with the chip meter
slowly returning to steady state production rate and with blow flow to compensate for the persistent load disturbance.
The GPC was set to constrain the chip
meter speed about the production target
speed. The motive for the use of the chip
meter was to reduce blow flow manipulations. No specifications about blow line
consistency control were described.
Discussions with operating personnel
indicate that large blow flow manipulations could disturb the chip column
movement and change the mass and energy balances. The results showed a
decrease of 50 % in blow flow manipulations, but the expected improvements in
pulp quality were not achieved (a slight
increase in kappa variations). The MIMO
GPC controller was disabled after six
month of operation. According to Allison
et al. [6] the bad chip level measurement
(strain gauge measurement) and the fact
that the mill had been experimenting
with a number of new grades were contributing factors to why the GPC was
switched off.
• Lundqvist [10] presents a chip level
control strategy that utilizes both the chip
meter speed and the blow flow as manipulated variables depending on the size of
the chip density variations. At small chip
density and wood property variations, the
chip meter is used as primary MV and the
blow flow is kept constant. If the chip density varies considerably, the blow flow is
used as MV and the chip meter is kept
constant. The tuning of the mix of the two
MV’s should be done regarding the chip
level response characteristics and with the
kappa number chosen as the control performance criterion.
• Amirthalingam et al. [2,3] discussed
how chip level control relates to pulp
quality parameters such as the kappa
number. They proposed that the manipulated variable for the chip level control
should be selected according to the origin
of the disturbance, e.g. if a change in the
digester chip level is caused by a change
in chip bulk density, the right action
would be to change the chip metering
speed. However if a disturbance in the
chip level were caused by a change in the
fibre discharge rate (change in blow line
pulp consistency), the right action would
be to change the blow flow (or digester
106:11 (2005) ❘❘❘
47
PROCESS CONTROL
bottom scraper). They demonstrated the
proposed digester control structure
through a simulation model (fundamental model). They also showed that a very
tight control of chip level (via a well tuned
PI-controller) actually disturbed the inferential kappa number control.
Table I is a summary of the reviewed
papers, showing the selection of manipulated variables and the controller structure for the digester level control.
PROCESS ANALYSIS AND
PROPOSED CONTROL
STRATEGY
Process description of the M-real digester
system
The continuous digester at M-real Husum
mill is a conventional Kamyr digester system, Fig.1, which consists of an impregnation vessel and a steam/liquor phase
digester. The digester has four liquor circulation flows (C5, C6, C7, and C8) where
the C7 is not used and the C6 trim circulation is only used for alkali measurement.
The alkali is added in the impregnation
vessel and in the C5 and C8 circulation.
The production rate is about 1000
tons/day (softwood) and the kappa number target is 30. The digester supervisory
control system at M-real Husum mill of
consists of following modules
1) Production rate change control
2) Pre-steaming
3) Chip level control of digester and
impregnation vessel (Blow flow, gamma
source as level ind.)
4) Wood/liquor ratio and circulation
flow control
5) Pulp consistency and sluice flow control
6) White liquor addition (concentration)
control
7) H-factor control
8) Kappa number prediction and feedback control
All control modules are implemented
in the Siemens DCS system (mainly with
PID controllers). The actual production
load on the digester is about 50 % higher
than design figures, production per bottom square meter [ton/m2*day]. During
the cold wintertime, snow, ice and frozen
chips make it more difficult to get a
smooth digester operation compared to
summertime.
Process analysis
The chip level gives an indication of the
chip column movement and the residence time (the extent of the chemical
and thermal treatment) for the wood
chips in the digester.
The downward movement of the chip
column is due to the gravitational force,
which mainly depends on the height of
the chip pile (bulk density, etc) that is
above the liquor level and the density difference between the chips and the liquid.
48 ❘❘❘ 106:11 (2005)
The fluctuation of the chip pile may cause
changes in the driving force and furthermore variations in the chip column movement and compaction. The chip column
is, however, elastic and compressible and
the elasticity of the column may be
reversible or irreversible [8] and the compaction degree of the chip column may
fluctuate.
If the chip column is moving quite
freely through the digester, the residence
time in the impregnation, cooking and
washing zones depends mainly on the
fiber discharge rate [13]. It is then important that the fibre discharge rate is stabilized. Amirthalingam et al [2] pointed out
that a good chip level controller structure
should react to the root cause of the level
disturbance.
The choice of control philosophy (selection of MV’s and CV’s, controller structure
etc.) for the chip level control may be more
important than actually achieving a very
tight control of the level [2].
Proposed control strategy
The proposed control strategy for the chip
level control focuses on stable conditions
in the digester bottom. A conclusion
regarding the described strategies above is
that excessive blow flow manipulation will
disrupt the chip column movement and
the digester operation. Blow flow manipulations will affect the washing liquor up
flow (dilution factor control), which will
disturb the conditions in the washing zone
and change the position where the cooking process (delignification) is disrupted.
The following proposal for the chip
level control strategy is based on
[10,11,13], discussion with process engineers and digester operators at m-real
Husum mill, and experiences from other
digester chip level control system installations made by Eurocon AB. This control
strategy also includes the level control of
the impregnation vessel.
The digester chip level is controlled by:
• Manipulating the speed of the chip
metering wheel. The control signal is
added as a production bias to the nominal
production with upper and lower limits
(+/- 2 t/h).
• Manipulating the digester bottom
scraper controls the pulp blow consistency. The pulp consistency measurement is
primarily based on the differential pressure between digester bottom and blow
line. The in-line pulp consistency measurement device (yield stress) and the
load on the bottom scraper may be used
to adjust the pulp consistency controller.
• The blow flow is manipulated slowly to
move the actual production back to target
production.
• Depending on the performance of the
pulp consistency controller the dilution
flow could be used to improve consistency control.
The impregnation vessel chip level is controlled
by:
• Manipulating the impregnation vessel
bottom scraper and the bottom dilution
flow.
• The dilution flow (sluice) is used as the
primary manipulating variable to maintain the level close to target.
• The scraper speed is used to maintain
the level within an upper and lower limit.
When the level is within these limits no
action is made by the scraper. (MPC constraint functionality, see below)
• The scraper speed is slowly moved to a
minimum speed target to move the average control action to the bottom dilution
flow and to prevent too high scraper speed.
CASE STUDY —
M-REAL HUSUM MILL
The proposed control strategy above is
mainly implemented as a Model Predictive
Control scheme on a stand alone PC communicating via an OPC link to the
Siemens DCS system. A watchdog function
for the OPC link checks that the communication is established and running. If the
communication is disrupted, a back-up
control function is implemented in the
DCS system with mainly the same control
strategy as the MPC implementation.
Implemented control strategy
A system structure of the implemented
control strategy is shown in Figs. 2 and 3.
The digester chip level is controlled by
adding a production bias (+/- 2 t/h) to
the nominal production target, which is
controlled by manipulating chip-metering
speed. The digester chip level is also set
up as a constraint variable (CT) with an
upper and lower limit. If a violation of
these limits is detected, the MPC algorithm will make a powerful adjustment of
the chip meter speed to bring the level
within the limits again. When the level is
within the limits, the control actions made
by the level constraint variable are disabled.
The actual production is used as a
feed-forward signal for the impregnation
vessel level control to improve the chip
level control at production rate changes
and to shorten the time delay for the chipmetering wheel.
The blow flow to production ratio is
very slowly adjusted to keep the production bias within +/- 1 t/h. The implementation of the blow flow control function is
still a matter of discussion and is, at the
moment, not implemented in the MPC
controller. The differential pressure
between the digester bottom and the blow
line is controlled by the bottom scraper
speed to maintain a constant pulp consistency and to stabilize the fibre discharge
rate.
The impregnation vessel chip level
(controlled variable) is controlled to the
level target by adjusting the set point for
the bottom dilution flow (FC5.2 in Fig. 2).
T 241 Pulp & Paper Canada
PROCESS CONTROL
FIG. 2. Digester chip level control system.
FIG. 3. MPC controller structure.
FIG. 4. Step response for the differential pressure.
FIG. 5. Step response for digester level.
The impregnation vessel chip level is also
set up as a constraint variable with an
upper and lower limit. If a violation of
these limits is detected, the bottom scraper will react powerfully, together with the
dilution flow, to bring the level within the
limits again. When the level is within the
limits, the scraper speed is slowly brought
back to a target speed to move more of
the average control action onto the dilution flow (FC5.2). The impregnation vessel scraper speed is both a manipulated
variable and a control variable in the MPC
controller, Table II.
The constraint variable can be seen, in
this case, as a safety function to prevent
plugging of the impregnation vessel. That
allows a smoother control action for the
set-point control. It is not very critical to
have a tight level control of the impregnation vessel, and that also gives the possibility of using the impregnation vessel as
a buffer for chip feed variations into the
digester system.
Step response modeling and MPC controller design
A key element in the design of the MPC
controller is the dynamic process model,
Pulp & Paper Canada T 242
which is used for the output predictions
and to relate the MV’s to the CV’s. The
data used for the modelling and design
has been generated by step response
experiments. The possibilities of using
more sophisticated data generation tools
like PRBS signals were limited. Figures 4
through 6 show the step responses for the
chip levels and the differential pressure.
The step response figures show that
the digester and impregnation vessel level
process can be characterized as an integrator process with a time delay at the
input. The time delay is, however, longer
than physically expected. That could be
explained by the time it takes to accelerate the chip column in the impregnation
vessel and the compaction of the chip column in the digester due to chip column
elasticity [8]. The chip column elasticity
can possibly be modelled as a second
order system. An integrating process with
dead time was selected as the model structure for the models needed in the MPC
design, except for the differential pressure control, where a first order model
structure was used.
Table II shows the controller design
matrix for MPC controller where each cell
in the matrix contains the dynamic model
that relates the MV’s with CV’s and CT’s.
The prediction horizon for the MPC controller was set to 50 minutes (100 control
intervals). The prediction horizon for the
level constraint variables was set to 30 and
20 minutes to make that control function
less restrictive. The MPC algorithms are
described in Maciejowski [1].
RESULTS AND DISCUSSION
The MPC controller has been running
since start-up in mid December 2003. The
back up level control system in the
Siemens DCS system was installed in
August 2003 and was running until the
MPC controller was installed. The original
digester level control system, with the blow
flow as the primary manipulated variable,
is still available for the operators but has
not been in operation since the start-up of
the new level control strategy.
Table III shows some figures regarding
the standard deviation for the chip meter
speed [rpm], blow flow [m3/h], kappa
number, digester chip level [m] and pulp
consistency [%]. The standard deviation
calculations are made from periods of
equal length and similar production rates
106:11 (2005) ❘❘❘
49
PROCESS CONTROL
FIG. 6. Step response for impregnation vessel leve.
TABLE II: MPC controller design matrix.
(40 to 46 t/h) during January and February in 2003 and 2004
(sample time 20 minutes).
The results show a clear reduction of blow flow manipulations and pulp consistency variations. The reasons for the reduction in kappa number variations may need a closer investigation
of the overall production conditions. The digester operators’
confidence for the new control system is high and they state that
the digester operation has been smoother this winter period
than previous ones.
Figure 7 shows a trend plot of the digester level control. The
production bias (chip meter speed) is manipulated to control the
digester level. The blow flow set point is adjusted at time 16:48
when the production bias is outside the +/- 1 t/h control limits.
Figure 8 shows how the blow line pulp consistency is controlled
by manipulating the digester bottom scraper speed. The goal is to
keep the pulp consistency within +/- 0.25 %. The digester bottom
dilution flow may need to be involved in the pulp consistency control function to improve control performance [13].
The impregnation vessel chip level is primary controlled by
the bottom dilution flow (FC5.2 in Fig. 2) but with the bottom
scraper acting on the upper and lower limit. In the beginning of
the trend plot, Fig. 9, the scraper speed increases to correct a
detected violation of the upper constraint limit. When the level
is back within the limits the scraper speed slowly returns to target speed, which is set to 3 rpm.
CONCLUSIONS AND FUTURE WORK
This paper describes the implementation of a new control strategy for the digester chip level control at the M-real Husum mill.
The objective of the new control strategy was to stabilize the
transport process in the bottom of the digester. The chip feed
50 ❘❘❘ 106:11 (2005)
FIG. 7. Digester level control. The production bias is changing between — 2 t/h to + 2 t/h.
TABLE III: Comparison between old level control system
(2003) and new level control system (2004).
was used as the primary manipulated variable for the level control, instead of the blow flow, which was used in the original level control system. A comparison of production data for January
to February 2004 and the same period in 2003 indicates that the
new control strategy decreases the kappa number variations. A
smoother digester operation (stated by the operators) together
with an increased focus, from the operators, on the complete
digester mass balance control was probably two contributing factors for the reduction of the kappa number variations. Even if
the chip level variations did not decrease significantly, the reduction of the blow flow manipulations and pulp consistency variations probably stabilized the chip column movement and the
conditions in the digester bottom.
The MPC controller implementation has improved the control of the impregnation vessel chip level. With the constraint
functionality in the MPC scheme it has been possible to combine
disturbances reduction with safety function control in the same
control algorithm.
The future work regarding the level control system will be to
• Modify and improve the control function for the blow flow
and implement that function into the MPC controller
• Improve the pulp consistency control by utilizing dilution flow
and bottom scraper motor amperage.
• Utilize gain scheduling for improved tuning at low production
rates
• Start-up of the digester and MPC controller performance
monitoring project.
ACKNOWLEDGEMENT
The authors would like to thank the digester operators and process engineers at M-real Husum Mill for their assistance, and the
Network for Process Intelligence (NPI) at Mid Sweden University, Örnsköldsvik, Sweden for their financial support. A special
thanks goes to Capstone Technology, Seattle USA, for the use of
their MPC software.
LITERATURE
1. MACIEJOWSKI, J.M. Predictive Control with Constraints. Harlow, UK. Prentice Hall,
331p. (2002).
2. AMIRTHALINGAM, R., LEE, J. H. Inferential Control of A Continuous Pulp
Digester In the Presence of Chip Level Variations. Proceedings Control Conference (2000).
3. AMIRTHALINGAM, R., LEE, J. H. Subspace Identification Based Inferential
T 243 Pulp & Paper Canada
PROCESS CONTROL
FIG. 8. Blow line pulp consistency control.
Control Applied to a Continuous Pulp Digester. J. of
Process Control 9:397-406 (1999).
4. PELLETIER, S., EVANS, R., WESSLEN, T. New
Digester Controls at Donohue Boost Kraft Pulp Quality and Production , www.paperloop.com (1998).
5. SVANHOLM, C. Supervisory control of continuous
digesters. Certification of digester operators,
Markaryd Education, internal document Eurocon AB,
(1997).
6. ALLISON, B. J., DUMONT, G. A., NOVAK, L. H.
Multi-Input Adaptive-Predictive Control of Kamyr
Digester Chip Level. The Canadian J. of Chem. Eng.
69:111-119 (1991)
7. ALLISON, B. J., DUMONT, G. A., NOVAK, L. H.,
CHEETHAM, W. J. Adaptive-Predictive Control of
Kamyr Digester Chip Level. AIChE Journal 36(7):10751085 (1990).
8. HÄRKÖNEN, E. A mathematical model for twophase flow in a continuous digester. Tappi Journal
70(12):112-126(1987).
9. BELANGER, P. R., ROCHON, L., DUMONT, G. A.,
GENDRON, S. Self-Tuning Control of Chip Level in a
Kamyr Digester. AIChE Journal 32(1):65-74 (1986).
10. LUNDQVIST, S. O. State of Art in Continuous
digester control. Proc. EUCEPA Symp. Control Sys. in
Pulp and Paper industry, Stockholm (1982).
Pulp & Paper Canada T 244
FIG. 9. Impregnation vessel level control.
11. AL-SHAIKH, A., TU, F. Multilevel control of continuous digesters: Applications and results. Proc. IFAC
7th World Cong., Helsinki, 239 (1978).
12. SASTRY, V. A. Self-Tuning Control of Kamyr
Digester Chip Level. Pulp & Paper Canada 79(5):T160T163 (1978).
13. FUCHS, R. E., SINGLE, G. J. Advances in Computer Control of Kamyr Continuous Digesters. Tappi
Journal 59(2):96-99 (1976).
Résumé: La présente communication décrit le problème du contrôle du niveau des copeaux
dans le lessiveur et illustre la mise en oeuvre d’un système MPC pour le contrôle du niveau. Le
but de ce système est de stabiliser les conditions du transport au fond du lessiveur à l’aide de l’alimentation des copeaux et non par le soufflage comme variable de commande principale. Les
résultats font état d’un meilleur fonctionnement du lessiveur et d’une réduction des variations de
l’indice Kappa par rapport au système de contrôle initial.
Reference: RLINDGREN, T., GUSTAFSSON, T., FORSGREN, H., JOHANSSON, D., ÖSTENSSON, J. Model predictive control of the chip level in a continuous pulp digester, a case study. Pulp
& Paper Canada 106(11):T239-244 (November 2005). Paper presented at the Control Systems
2004 in Quebec, Quebec, June 14-17, 2004. Not to be reproduced without permission of PAPTAC.
Manuscript received April 01, 2004. Revised manuscript approved for publication by the Review
Panel on April 6, 2005.
Keywords: PROCESS CONTROL, CONTROL SYSTEM, CHIPS, CONTINUOUS PROCESS,
CHEMICAL PULPING, FEEDERS, DIGESTERS.
106:11 (2005) ❘❘❘
51