ANALYSIS OF A SEMI-BATCH REACTOR FOR CONTROL PURPOSES

František Gazdoš and Lubomír Macků

Faculty of Applied Informatics
Tomas Bata University in Zlín
Nad stráněmi 4511, 760 05, Zlín, Czech Republic
E-mail: [email protected]

KEYWORDS view by simulation means mainly to obtain useful

Exothermic Semi-batch Reactor, Chromium Waste, information for subsequent optimal control design.
Modelling, Analysis, Simulation. The contribution is structured as follows: after the
introduction, a detailed description of the reactor
ABSTRACT follows. In the next sections, a mathematical model of
The paper presents a control system design oriented the system is derived and all variables are defined
analysis of a semi-batch reactor used for tanning waste together with their physical values and limits. Further,
recovery. The analysis is performed with the help of steady-state and dynamical behaviour is studied by
modelling and simulation means and provides useful simulation means. A detailed analysis of the reactor
information for optimal control design of the process. from the systems theory point of view follows in next
sections, giving useful information for optimal control
INTRODUCTION design. Possible control strategies are discussed at the
end of the contribution together with areas for possible
The tannery is an essential industry process today. Its future research.
product is a natural hide. Some of the leather properties
(such as softness, plasticity, stability, absorption) cannot A SEMI-BATCH REACTOR
be replaced by any artificial material.
The chromium sludge is processed in a chemical reactor
There are a lot of technological operations during the
sketched in Fig.1. by an exothermic chemical reaction
leather-to-hide conversion, starting with washing,
with chrome sulphate acid. During this reaction a
continuing by tanning etc. These are usually connected
considerable quantity of heat is developing so that
with huge water, energy and chemicals consumption
control of the reaction is necessary.
resulting in negative impact on the environment.
Recently, one of the very fundamental operations is
chromium salt tanning. It is possible to obtain only 200
[kg] of hide from 1000 [kg] of leather, but over 600
[kg] of solid waste originates from this process; the rest
is drained away in the form of liquid waste during the
chromium salt tanning process. The USA produces
almost 60 000 tons of this solid waste and the
worldwide production is approximately 10 times bigger
(Cabeza et al. 1998). Currently, majority of this solid
waste is land filled. It can result in leakage of Cr3+ into
groundwater. After oxidation, for example in sewage
treatment plants during purification from unhealthy
bacteria, the Cr3+ to Cr6+ conversion can occur. The
Cr6+ compounds belong to cancerous substances so
they are dangerous to health. As a consequence, at
present, alternative methods of dealing with the
chromium waste are sought and investigated (Aloy and
Vulliermet 1998; Tiravanti et al. 1996, 1997). The
enzymatic hydrolysis is one of the considered
alternatives (Kolomazník et al. 1996). This technique
separates the chrome from protein in the form of the
chromium filter cake. All products of this process are
usable – it is a waste-free technology.
This paper deals with an analysis of a chemical reactor
for chromium sludge (chromium filter cake) recovery.
The reactor is used for the enzymatic hydrolysis and the
analysis is performed from the control theory point of
Figure 1: Chemical Reactor

Proceedings 22nd European Conference on Modelling and

Simulation ©ECMS Loucas S. Louca, Yiorgos Chrysanthou,
Zuzana Oplatková, Khalid Al-Begain (Editors)
ISBN: 978-0-9553018-5-8 / ISBN: 978-0-9553018-6-5 (CD)
System description
⎡⎣ m ( t ) aFK ( t ) ⎤⎦′ [kg.s-1], where aFK ( t ) [ − ] denotes the
In order to investigate main properties of the real
mass concentration of the chromium sludge in the
process, a mathematical model of the chemical reactor
reactor and m ( t ) [ kg ] describes weight of the reaction
was derived based on Fig.2.
components in the system. The expression
k m ( t ) aFK ( t ) [kg.s-1] defines the chromium sludge
extinction by the chemical reaction. Here, k[s-1] is the
reaction rate constant expressed by the Arrhenius
equation (2) where A [s-1], E [J.mol-1] and R [J.mol-
1
.K-1] are pre-exponential factor, activation energy and
gas constant respectively.
E
−
RT ( t )
k = Ae (2)

The third equation describes the enthalpy balance. The

input heat entering the reactor in the form of the
Figure 2: Chemical Reactor Scheme
chromium sludge is expressed by the term mFK cFK TFK ,
The scheme above shows a chemical semi-batch reactor the heat arising from the chemical reaction is given by
with initial filling mP[kg] given by the solution of the expression ΔH r k m ( t ) aFK ( t ) and the heat
chemicals without the chromium sludge (filter cake). transmission through the reactor wall is expressed by
This is fed into the reactor by mFK [kg/s] to control the
the formula K S ⎡⎣T ( t ) − Tv ( t ) ⎤⎦ . The individual symbols
developing heat since the temperature has to stay under
used above mean: cFK [J.kg-1.K-1] – chromium sludge
a certain critical level ( T ( t ) ≺ 100o C ), otherwise the
specific heat capacity, cR [J.kg-1.K-1] – specific heat
reactor could be destroyed. On the other hand it is capacity of the reactor content, TFK [K] – chromium
desirable to utilise the maximum capacity of the reactor sludge temperature, ΔH r [J.kg-1] – reaction heat, K
to process the maximum amount of waste in the shortest
[J.m-2. K-1.s-1] – conduction coefficient, S [m2] – heat
possible time (higher temperature is desirable).
Therefore an optimal control strategy has to find a transfer surface, T ( t ) [K] - temperature of reaction
trade-off between these opposite requirements. components in the reactor, Tv ( t ) [K] – temperature of
a coolant in the reactor double wall.
Mathematical model The last equation describes coolant heat balance. The
Under usual simplifications, based on the mass and heat input heat is given by mv cvTvp , the heat entering the
balance, the following 4 nonlinear ordinary differential coolant by the reactor wall is expressed by
equations can be derived (Macků 2005, 2004): K S ⎡⎣T ( t ) − Tv ( t ) ⎤⎦ , the heat going out with the coolant
d is described as mv cvTv ( t ) and the heat accumulated in
mFK = m (t ) (1)
dt the double wall describes the last term mvR cvTV′ ( t ) . The
d symbols mean: mv [kg.s-1] – coolant mass flow, cv
mFK = k m ( t ) aFK ( t ) + ⎡⎣ m ( t ) aFK ( t ) ⎤⎦
dt [J.kg-1.K-1] – coolant specific heat capacity, Tvp [K] –
mFK cFK TFK + ΔH r k m ( t ) aFK ( t ) =
input coolant temperature, mvR [kg] – coolant mass
d
= K S ⎡⎣T ( t ) − Tv ( t ) ⎤⎦ + ⎡⎣ m ( t ) cR T ( t ) ⎤⎦ weight in the reactor double wall.
dt Variables mFK , mv , TFK , Tvp are manipulated signals,
d
mv cvTvp + K S ⎡⎣T ( t ) − Tv ( t ) ⎤⎦ = mv cvTv ( t ) + mvR cv Tv ( t ) however, from practical point of view, only mFK and
dt
mv are usable. The temperature change of TFK or Tvp
The first equation expresses the total mass balance of is inconvenient due to the economic reasons (great
the chemical solution in the reactor. The symbol energy demands).
mFK [kg.s-1] expresses the mass flow of the entering
chromium sludge and m′ ( t ) [kg.s-1] describes the For simulation purposes, the following values of the
constants used in the mathematical model (1) were
accumulation of the in-reactor content. adopted (obtained analytically, experimentally,
The second equation represents the chromium sludge estimated or taken from the literature, for details see
mass balance. The input is mFK [kg.s-1] again, the Macků 2004): A = 219.6 s-1, E = 29968 J.mol-1, R =
accumulation is given by the last term 8.314 J.mol-1.K-1, cFK = 4400 J.kg-1.K-1, cv = 4118
J.kg-1.K-1, cR = 4500 J.kg-1.K-1, mvR = 220 kg, ΔH r =
1392350 J.kg-1, K = 200 J.m-2.K-1.s-1, S = 7.36 m2,
mv = 1 kg.s-1, TFK = 293.15K = 20°C, Tvp = 288.15K
= 15°C. The mFK feeding varied from 0 to 3kg.s-1.
Initial conditions were defined as:
m ( 0 ) = m p = 1810 kg (initial reactor filling),
aFK ( 0 ) = 0 (initial mass concentration of the
chromium sludge in the reactor), T ( 0 ) = 323.15 K =
50°C (initial temperature of the reactor filling) and
initial coolant temperature Tv ( 0 ) = 293.15 K = 20°C.

Technological limits and variables saturation Figure 3: Total Mass Response

Maximum filling of the reactor is limited by its volume The figure above shows increase of the total mass in the
to the value of m < 2450 kg approximately. Then the reactor for various input flow rates of the chromium
sludge. The simulation reveals integrating, astatic
process of feeding by the chromium sludge mFK has to
behaviour and it already respects the limitations
be stopped. The feeding can be practically realized in introduced previously (max. mass in the reactor
the range mFK ∈ 0;3 kg.s-1. As stated in the system mmax = 2450 [ kg ] ).
description, the temperature cannot exceed the limit
T ( t ) ≺ 100o C ; this holds also for the coolant (water)
but it is not so critical in this case as shown by the
further experiments.

SIMULATION ANALYSIS
In order to prepare a convenient control strategy for the
process described above, first a steady-state and
dynamic analyses were performed with the help of
simulation means.

Steady-state analysis
The reactor described above, i.e. a semi-batch reactor,
has only one steady state given by: Figure 4: Chromium Sludge Mass Concentration
Response
mFK = 0; aFK = 0; T = Tv = Tvp (3)
The response presented in Fig. 4 reveals derivative
It means that the steady state can occur only at the behaviour of the variable aFK ( t ) for various values of
beginning of the process (before feeding the reactor) or mFK .
at the end of the reaction running inside.

Dynamics analysis
Dynamical behaviour analysis of a system is also an
important part in the control design process. It gives
insight into stability and other important properties of
the system, such as time-constants, gains and non-
minimum-phase behaviour. Using simulation means
(MATLAB/Simulink), the following graphs were
recorded as a step-responses of the input variable mFK
(other variables were kept at their operating values
defined previously). The input variable was chosen in
the range: mFK = [ 0.05 0.1 0.5 1 3] ⎡⎣ kg ⋅ s −1 ⎤⎦ .
Figure 5: Temperature-in-the-Reactor Response
The response of Fig. 5 shows temperature increase Linear model
inside the reactor – it can be seen how the temperature
Having generally a nonlinear model defined by a system
rises as a result of the chemical reaction. The faster
of formulas
input flow rate of the chromium sludge, the faster
reaction and temperature increase. Then, the next
increase is limited by the restriction on the maximum x ′ ( t ) = f ⎡⎣t , x ( t ) , u ( t ) ⎤⎦ , (4)
possible mass in the reactor followed by gradual
temperature fall. From the graph it is also clear that for where x (t ) defines a vector of state-variables
the simulated range of mFK the temperature goes
⎡⎣ x1 ( t ) x2 ( t ) xn ( t ) ⎤⎦ , u (t ) vector of input
beyond the allowed limit T ( t ) ≺ 100 [°C] for higher
mFK -values, therefore the process needs to be variables ⎡⎣u1 ( t ) u2 ( t ) um ( t ) ⎤⎦ and f is a
controlled properly. nonlinear vector function [ f1 f2 f n ] , then the
linear model in a given operating (steady-state) point
( u , y ) can be generally obtained using formulae:
s s

x ′ ( t ) = Ax ( t ) + Bu ( t ) (5)

with constant matrices A , B calculated as:

⎡ ∂f1 ∂f1 ∂f1 ⎤

⎢ ∂x ∂x2 ∂xn ⎥
⎢ 1 ⎥
⎢ ∂f 2 ∂f 2 ∂f 2 ⎥
⎢ ⎥
A = ⎢ ∂x1 ∂x2 ∂xn ⎥ (6)
⎢ ⎥
⎢ ⎥
Figure 6: Coolant Temperature Response ⎢ ∂f n ∂f n ∂f n ⎥
⎢ ∂x ∂x2 ∂xn ⎥⎦ x = x s
⎣ 1
A record of the coolant temperature is presented in Fig.
6. As can be seen from the graph, the temperature of the
media for the whole range of mFK is not critical since ⎡ ∂f1 ∂f1 ∂f1 ⎤
⎢ ∂u ∂u2 ∂um ⎥
water is used for the cooling (provided the defined flow ⎢ 1 ⎥
rate mv = 1 [kg/s] is ensured). ⎢ ∂f 2 ∂f 2 ∂f 2 ⎥
⎢ ⎥
B = ⎢ ∂u1 ∂u2 ∂um ⎥
CONTROL THEORY POINT OF VIEW ⎢ ⎥
⎢ ⎥
From the systems theory point of view the reactor has ⎢ ∂f n ∂f n ∂f n ⎥
⎢ ∂u ∂um ⎥⎦ u = us
four input signals mFK ( t ) , mv ( t ) , TFK ( t ) and Tvp ( t ) , ⎣ 1 ∂u2

four state variables m ( t ) , aFK ( t ) , T ( t ) , Tv ( t ) and

As the reactor embodies astatic behaviour, it is not
one output signal to be controlled given by the possible to compute the matrices in a chosen (steady-
temperature inside the reactor T ( t ) . Hence, it can be state) operating point. However, the linearization can be
generally seen as a Multi Input – Multi Output (MIMO) performed generally, resulting in a time-variant system:
system of 4th order. In addition it possesses strongly
nonlinear behaviour. Practically, the only manipulated x′ (t ) = A (t ) x (t ) + B (t ) u (t ) (7)
variables are input flow rates of the chromium sludge
mFK ( t ) and of the coolant mv ( t ) . Therefore, input where the matrices A ( t ) , B ( t ) are no longer constant
temperatures of the filter cake TFK ( t ) and of the but time-dependent. Using the formulas above, the
coolant Tvp ( t ) can be alternatively seen as originally nonlinear model of the reactor has been
transformed into a linear time-variant model with the
disturbances. For further analysis, the reactor model
matrices A ( t ) , B ( t ) of the form (8),
described by the system of differential equations (1) is
transformed into a linear time-variant (LTV) system.
 ⎡ 0 0 0 0 ⎤ Since the system generally has one output T ( t ) and 4
⎢a t a t a t 0 ⎥⎥
21 ( ) 22 ( ) 23 ( ) inputs mFK ( t ) , mv ( t ) , TFK ( t ) , Tvp ( t ) , the resultant
A (t ) = ⎢ (8)
⎢ a31 ( t ) a32 ( t ) a33 ( t ) a34 ( t ) ⎥ t.f. is a vector of the 1-by-4 size:
⎢ ⎥
⎣⎢ 0 0 a43 a44 ( t ) ⎦⎥
G = ⎡GT GT GT GT ⎤. (13)
⎣ mFK mv TFK Tvp ⎦
⎡ 1 0 0 0 ⎤
⎢b t 0 ⎥⎥
21 ( ) 0 0
B (t ) = ⎢ The first term GT mFK describes the relation between the
⎢b31 ( t ) 0 b33 ( t ) 0 ⎥
⎢ ⎥ temperature inside the reactor T ( t ) and the input flow
⎢⎣ 0 b42 ( t ) 0 b44 ( t ) ⎥⎦
rate of the chromium sludge mFK ( t ) . The other terms
where the particular ralations describing the coefficients describes the relations between the temperature and the
(except for a43 all coefficients are time-dependent) are variables mv ( t ) , TFK ( t ) , Tvp ( t ) respectively. As
presented in the appendix. Generally, output from a stated earlier, the only practically manipulated variables
linear system is defined as: are mFK ( t ) and mv ( t ) . Therefore, next analysis is
y ( t ) = Cx ( t ) + Du ( t ) (9) focused on the relation between these variables and the
inner reactor temperature. The “pseudo” transfer
function for these two input variables have these
with matrices C , D defined in our case as:
general forms (using (8)-(10) and (12)):
C = [ 0 0 1 0] , D = [ 0 0 0 0] (10)
B3 ( t ) s 3 + B2 ( t ) s 2 + B1 ( t ) s + B0 ( t )
GT ( s, t ) = , (14)
(output for control purposes is only the temperature
mFK
(
s s 3 + A2 ( t ) s 2 + A1 ( t ) s + A0 ( t ) )
inside the reactor and the system is strictly proper).
B1′ ( t ) s + B0′ ( t )
Transfer function GT mv ( s, t ) = , (15)
s + A2 ( t ) s 2 + A1 ( t ) s + A0 ( t )
3
If we have the matrices A , B , C , D of a linear (or
linearized) model then it is possible to compute a
transfer function (matrix in a MIMO case) of a system, where relations for the time-dependent
which is a useful relation for control system analysis coefficients Ai ( t ) , Bi ( t ) are given in the appendix (due
and design. It gives insight into input-output behaviour to the limited space, for information only formulas
of the system variables and provides useful information, concerning the most important t.f. GT mFK are
such as stability, non-minimum-phase behaviour, time-
constants, gain etc. For constant matrices mentioned presented).
above, it can be computed simply as: From the equations presented above, it can be seen that
the relation between T ( t ) and mFK ( t ) is generally
−1
G ( s ) = C ( sI n − A ) B , (11) integrative whereas between T ( t ) and mv ( t ) only
proportional. At present, the only practically
where s is the complex Laplace variable and I n is the manipulated variable is mFK ( t ) , therefore, the further
n -by- n identity matrix. As the reactor analyzed in this investigation is focused on the analysis of the transfer
contribution is astatatic, the linearized model is time- function GT mFK . Although it would be also useful to
dependent and the “proper” transfer function introduced
above cannot be computed. However, in a given time- investigate also GT mv , due to the limited space of this
moment t , the time-dependent matrices A ( t ) , B ( t ) paper, this analysis is not provided.
can be considered constant, i.e. it is possible to compute
a “pseudo” transfer function (t.f.) describing Transfer function coefficients range
instantaneous relation between the output and input
In order to determine the range of GT mFK coefficients, a
signals. Then this “pseudo” transfer function is also
time-dependent, but for a given time-moment it yields series of simulation experiments were performed in the
instantaneous transfer function of the system: MATLAB/Simulink environment. The variable
mFK ( t ) was changing in the admissible range from 0 to
G ( s, t ) = C ( sI n − A ( t ) ) B ( t ) .
−1
(12) 3 [kg/s] with the step 0.1 and both maximum and
minimum values of the GT mFK coefficients were
Using this idea and the relations (8)-(10) it is possible to recorded. Summary of the achieved results is presented
compute a “pseudo” transfer function of the reactor. in Table 1.
 Table 1: Range of GT mFK Coefficients values of complex parts of the poles are relatively small
which shows that this effect is not so significant. From
the results, it can be also deduced that the system
Coefficient Minimum value Maximum value
possesses non-minimum phase (NMP) behaviour –
B3 -0.0426 -8.632x10-4 some of the zeros may become positive (unstable).
B2 9.944 x10-5 0.0016 Generally, NMP-systems are more difficult to control.
B1 5.221x10 -7
1.022x10-5 Having the range of poles, it would be also possible to
compute the range of time-constants and of the gain for
B0 -1.220x10-8 5.810x10-11 the system. Although this information would be also
A2 0.0051 0.0301 interesting from the controller design point of view, due
-6 to the limited space it is not provided.
A1 -8.777x10 1.770 x10-4
A0 -1.336x10-8 1.787x10-7 Discussion
The computed range of GT mFK coefficients from Table
The table shows that some of the coefficients are very
small and consequently they could be possibly 1 shows that for the control design, some of the
neglected for the control system design. coefficients could be neglected (e.g. B0 , possibly also
B1 , A0 ). Consequently, the linear time-variant model
Poles and zeros
will be of a simpler form, which is useful.
Poles pi of a system are defined as roots of the transfer
Suggested control strategies
function denominator whereas zeros zi are given by the
numerator roots. Given the range of coefficients from Havig the approximate uncertainty intervals of the
the Table 1, it is possible to compute also the range of GT mFK coefficients, it suggests using the robust control
poles and zeros of GT mFK . Results are summarized in approach (e.g. Morari and Zafirou 1989). As parameters
Table 2 where the roots are divided into their real and of the linearized model change, an alternative idea
imaginary parts for the case they are complex. could also be the usage of adaptive control strategies
(e.g. Åström and Wittenmark 1989). Another possible
Table 2: Range of GT mFK Poles pi and Zeros zi approach which proved to be successful is the
predictive control (Srinivasan and Bonvin 2007, Sámek
and Macků 2007). At present, the system is controlled
zi / pi Imag. Imag.
Real min. Real max. only as a Single Input – Single Output (the temperature
min. max.
inside the reactor is controlled only by means of the
z1 0.0185 0.1601 0 0 chromium sludge input flow rate). It would be useful to
z2 -0.0062 -0.0062 0 0 include also the coolant flow rate mv ( t ) as a
z3 -1.973x10-5 0.0016 0 0 manipulated variable and use the MIMO control design
p1 0 0 0 0 strategies (e.g. Skogestad and Postlethwaite 1996).

p2 5.590 CONCLUSION
-0.0226 -0.0059 0
x10-4
-5.590 9.108 This contribution was focused on the analysis of a
p3 -0.0067 0.0021 reactor used for the tanning waste recovery. The
x10-4 x10-4
-9.108 analysis was made by both theoretical and simulation
p4 -0.0018 8.464x10-4 0 means in order to provide useful information for
x10-4
optimal controller design. The nonlinear model of the
Zeros at (or very close to) the origin indicate derivative system was derived and transformed into a linear time-
behaviour whereas poles at the same position signalize variant system which was used for further investigation
integrative properties. The table shows that one pole of the process properties. The range of transfer function
( p1 ) is directly at the origin resulting in integrative coefficients was determined and possible control
strategies were outlined. In this work, the main focus
behaviour of the temperature T ( t ) with respect to was on the relationship between the controlled variable
mFK ( t ) . If the poles are located in the left part of the – temperature inside the reactor and the manipulated
complex plane (their real parts are negative), the system variable given by the chromium sludge input flow rate.
is stable. From this point of view the table shows that From the controller design point of view, it would be
generally the system embodies also instability. In useful to investigate also the relationship between the
addition, when the poles are complex (they also have second practically possible manipulated variable – the
imaginary parts), it indicates oscillatory behaviour. As coolant mass flow rate and the temperature.
revealed by the table, in some conditions the system
may embody oscillatory behaviour, however absolute
ACKNOWLEDGEMENTS Society of Leather Technologies & Chemists 82, No.4 (Jul-
Aug), 140-142.
The work was supported by the Grant Agency of the Åström, K.J. and B. Wittenmark. 1989. Adaptive Control.
Czech Republic under the grant no. 102/07/P148 and by Addison-Wesley, Reading, MA.
the Ministry of Education of the Czech Republic under Cabeza, L.F., Taylor, M.M., Dimaio, G.L., Brown, E.M.,
the grant no. MSM 7088352102. Authors are also Mermer, W.N., Carrio, R., Celma, P.J. and J. Cot. 1998.
grateful to Dr. Chalupa for the refinement of the “Processing of leather waste: pilot scale studies on chrome
presented mathematical model. shavings. Isolation of potentially valuable protein products
and chromium.” Waste Management 18, No.3, 211-218.
APENDIX Kolomazník, K., Mládek, M., Langmaier, F., Taylor, M.,
Diefendorf, E.J., Marmer, W.N., Brown, E. M. and L.
Formulas for coefficients of the linear time-variant Tribula. 1996. CR Patent 280 655, Czech Republic (Feb).
model Macků, L. 2004. “Control design for the preparation of
regenerate for tanning”. Ph.D. thesis. Tomas Bata
E
mFK ( t ) mFK ( t ) −
RT ( t )
University in Zlín, Czech Republic (in Czech).
a21 ( t ) = − (1 − aFK ( t ) ) , a22 ( t ) = − m (t )
− Ae Macků, L. 2005. “Modeling of tanning salts regeneration
m2 (t ) process.” In Proceedings of the 15th Int. Conf. Process
E Control 2005 (High Tatras, Slovakia, Jun.7-10).
−
RT ( t )
⎛ E ⎞
a23 ( t ) = − Ae ⎜ ⎟ a (t ) Bratislava: Slovak University of Technology, 127/1-
⎜ RT 2 ( t ) ⎟ FK 127/4.
⎝ ⎠
Morari, M. and E. Zafirou. 1989. Robust Process Control.
1
a31 ( t ) = − ⎡ mFK ( t ) ( cFK TFK ( t ) − cR T ( t ) ) + KS (Tv ( t ) − T ( t ) ) ⎦⎤ Prentice Hall, Englewood Cliffs, New Jersey.
m 2 ( t ) cR ⎣
Sámek, D. and L. Macků. 2007. “Simulation of model
E predictive control of semi-batch reactor.” In Proc. Int.
−
RT ( t )
Ae ΔH r KS Symp. on Systems Theory SINTES 13 (Craiova, Romania,
a32 ( t ) = , a34 ( t ) = Oct.18-20). Craiova: University of Craiova, 180-185.
cR m ( t ) cR
Skogestad, S. and I. Postlethwaite. 1996. Multivariable
E Feedback Control: Analysis and Design. John Wiley &
AΔH r EaFK ( t ) −
RT ( t ) KS m (t )
a33 ( t ) = e − − FK Sons, Chichester, England.
RT 2
( t ) cR m ( t ) cR m (t ) Srinivasan, B. and D. Bonvin. 2007. “Controllability and
stability of repetitive batch processes.” Journal of Process
KS 1 ⎛ KS ⎞ Control 17, 285-295.
a43 = , a44 ( t ) = − ⎜ mv ( t ) + ⎟
mvR cv mvR cv ⎠ Tiravanti, G., D. Petruzzelli and R. Passino. 1996. “Low and
⎝
non waste technologies for metals recovery by reactive
1 − aFK ( t ) 1 ⎛ cFK TFK ( t ) ⎞ polymers.” Waste Management 16, No.7, 597-605.
b21 ( t ) = , b31 ( t ) = ⎜ − T (t ) ⎟
m (t ) m (t ) ⎝ cR ⎠
Tiravanti, G., D. Petruzzelli and R. Passino. 1997.
“Pretreatment of tannery wastewaters by an ion exchange
mFK ( t ) cFK 1
b33 ( t ) =
m ( t ) cR
, b42 ( t ) =
mvR
(
Tvp ( t ) − Tv ( t ) ) process for Cr(III) removal and recovery.” Water Science
and Technology 36, No.2-3, 197-207.

mv ( t ) AUTHOR BIOGRAPHIES
b44 ( t ) =
mvR FRANTIŠEK GAZDOŠ was born in Zlín,
Czech Republic, and graduated from the
Formulas for transfer function coefficients Technical University of Brno in 1999 with
MSc. degree in Automation. He then
B3 ( t ) = b31 ( t ) followed studies of Technical cybernetics
B2 ( t ) = a32 ( t ) b21 ( t ) − b31 ( t ) a22 ( t ) + a31 ( t ) − a44 ( t ) b31 ( t ) at Tomas Bata University in Zlín,
obtaining Ph.D. degree in 2004. Now he works as a
B1 ( t ) = a21 ( t ) a32 ( t ) − a31 ( t ) a22 ( t ) −
senior lecturer in the Department of Process Control,
− a44 ( t ) ⎡⎣ a32 ( t ) b21 ( t ) − b31 ( t ) a22 ( t ) + a31 ( t ) ⎤⎦ Faculty of Applied Informatics of the same University.
E-mail: [email protected].
B0 ( t ) = −a44 ( t ) ⎡⎣ a21 ( t ) a32 ( t ) − a31 ( t ) a22 ( t ) ⎦⎤
LUBOMÍR MACKŮ was born in
A2 ( t ) = − a44 ( t ) − a22 ( t ) − a33 ( t )
Uherské Hradiště, Czech Republic and
A1 ( t ) = a22 ( t ) a44 ( t ) − a34 ( t ) a43 − a32 ( t ) a23 ( t ) + graduated from the Technical University of
+ a33 ( t ) a44 ( t ) + a22 ( t ) a33 ( t ) Brno in 1996 with MSc. degree in Process
Control Engineering. He followed studies
A0 = − a22 ( t ) a33 ( t ) a44 ( t ) + a22 ( t ) a34 ( t ) a43 + of Technical cybernetics at Tomas Bata
+ a32 ( t ) a23 ( t ) a44 ( t ) University in Zlín, obtaining Ph.D. degree in 2004.
Now he woks as a senior lecturer in the Department of
REFERENCES Electrotechnics and Measurements, Faculty of Applied
Informatics of the same University.
Aloy, M. and B. Vulliermet. 1998. “Membrane technologies E-mail:[email protected].
for the treatment of tannery residual floats.” Journal of the