Sie sind auf Seite 1von 9

Dissolved oxygen control through an adaptive

non-linear model approach: a simulation study


J. Ricardo

Pbrez-Correa*,

Francisco

L6pez and Iv&n Solar+

Universidad de Chile, Depto. de lngenieria Quimica, Casilla 2777, Santiago, Chile


Universidad CatHica, Depto. de Ingenierra Quimica, Casilla 6177, Santiago, Chile
(Received 18 October 1990; revised 25 March 1991)

Several exact linearization methods were applied to a simplified non-linear model for the concentration of
dissolved oxygen in a waste water treatment plant, and digital control algorithms were derived based on
these linearizations. A non-linear adaptive control algorithm is proposed and compared with a well tuned

PID, a linear adaptive controller, and a non-adaptive non-linear controller. The proposed algorithm shows
a better performance under a variety of perturbations. However, users must be careful in the choice of the
appropriate model parameters to be estimated.
(Keywords: models; adaptive control; non-linear algorithms; dissolved oxygen; water treatment)

There are many non-linear systems in the process industry with a well known qualitative behaviour. These
systems can be modelled by non-linear dynamic equations obtained from first principles. However, modelling
includes several empirical parameters (kinetic constants,
transport coefficients, etc.) that are difficult to estimate
and vary unpredictably. The usual approach to deal with
this type of system in process control is to use local
linearization and then apply linear control theory to
design the controller. Processes with weak non-linearities
can be properly controlled in this way, if the control
design is robust enough. However, in highly non-linear
processes (pH neutralization, non-isothermal reactors),
control design based on linear theory usually gives an
undesirable performance.
A different approach, which has been shown to be very
successful, consists of incorporating
the non-linear
model of the process directly into the control calculations. This approach is often called model based control
in the process literature. A closely related technique uses
an exact linearization of the non-linear model of the
process to design a control algorithm in the linear transformed space. Several methods have been proposed to
find appropriate transformations;l-9 among these, the
geometric methods are most widely used. Unfortunately,
most of these papers assume perfect model knowledge
and present only simulation results. For an overview of
geometric approaches applied to process control, the
reader is referred to the works of KantorO, Henson and
Seborg, McLellan et al.*, Isidorii3 and Kravaris and
Kantori4.

To whom correspondence should be addressed.

152

J. Proc. Cont. 1991, Vol 1, May

0959-1524~91~030152-09
0 1991 Butterworth-Heinemann

Ltd

Alternatively, some authors have incorporated on line


parameter
estimation
in their model-based
controllers>@, resulting in adaptive non-linear controllers,
Experimental results with this type of algorithm have
been reported*O. The main drawback with these controllers is that exact linearization is not used, and therefore only simple and non-robust control laws can be
applied.
A better solution can be obtained by combining nonlinear transformation with on line parameter estimation.
An algorithm of this kind was experimentally evaluated
by Onderwater et al.z, giving far better results than those
obtained by other methods. However, only process gain
non-linearities were considered, and an ad hoc non-linear
transformation was used. Perez and Solar** have outlined
a simple methodology to design non-linear adaptive controllers based on the process model. This methodology
can be summarized as follows:
i) Linearize the process model using an appropriate
exact linearization technique.
ii) Discretize the obtained transformations.
iii) Estimate on line some chosen model parameters
iv) Design a linear controller based on the linearized
model and the estimated parameters.
v) Transform back to the original variables and
implement the control.
The use of exact linearization improves the robustness of
the design compared with controllers that directly
include the non-linear model in the control formula1sJ6J8.
In addition, adaptation allows the controller to overcome the effects of changes in process behaviour, unmeasured disturbances and unmodelled dynamics. The proposed methodology has been successfully applied to

Dissolved oxygen control: J. Ricardo Pkrez-Correa et al.

control a simulated level tank with varying area** and a


simulated and experimental pH neutralization
process23+24.
A similar approach to the methodology described
above has been proposed and evaluated through simulations and experiments by Bastin and Dochain25. However, in their work a linearizing controller is achieved by
defining a linear model for the tracking error. In a more
theoretical paper, Sastry and Isidori*6 proposed a nonlinear adaptive linearizing control, where its convergence
and stability properties were studied; no simulations or
experiments were reported.
An alternative way of synthesizing adaptive non-linear
controllers can be achieved by a direct linearization of a
discrete non-linear model of the process. Discrete nonlinear transformations27g28 can be used for this purpose;
however, the control algorithm synthesis becomes more
complicated26.
In the present work, the methodology of Ptrez and
Solar is applied to control the dissolved oxygen concentration level in a waste water treatment plant. A simplified mode15,9is used in the simulations and in the design
of the controller. In order to illustrate the difficulties
associated with the linearization step, three simple methods for obtaining non-linear transformations
were
applied. These are: the global linearizing control (GLC)
of Kravaris and Chung, the variable transformation
method of 0gunnaike4 and the generic model control
(GMC) of Lee and Sullivar$. Finally, a simulation study
was carried out in which an adaptive GLC algorithm was
compared with PID, linear adaptive and non-linear nonadaptive control. The study focuses on the implementation issues associated with the adaptive part of the
algorithm.

The dynamics of oxygen dissolution in an activated


sludge process can be represented by the following simplified mode15J9:
=

$2 [G(t)

KdCs

- y(t)] + K,u(t)[Cs(t)

K2UNN

- ~1

The controllability index (r) can be defined as the smallest integer that satisfies:
(dh, a4-(g))
where (dx,y)

# 0

(3)

= Vx.y is called the Lie derivative and:

adz0t_d = y

(44

4 Cv)= kl

WI

adzZ01) = [z,[z,~ll
The Lie bracket is defined by:
[Z,Yl

ay

&Z

a2

ZY

Then, the controllability


since:
(dh, ad? <s)> = Fp

index for Equation

(2) is 1,

= Kl[Cs - y] # 0

and the transformed linear system is:

Pd.vldt+ POY= v

(7)

where PI and POare arbitrary constants.


The transformed controlled variable coincides with the
original measurement, and the transformed control variable can be obtained from Equations (2) and (7):
v = Pov +

Control algorithms

dY) =

Plvw

(8)

&)lu

Discretizing the continuous model gives:


b+,

Toh

Yk[PI

POT01

(9)

Defining p, = 1, and an open loop pole = 0.9 =


BoTo],results in the following transformed input:

- y(t)] (1)

vk

!L!yx+ +[Ci

-yk] - K&b,)

Equation (1) can be written as:

~=fi)+&)u

(24

hb9 =

(2b)

&_v) = QtCi - VI/V

(10)

where To is the sample time.


Now a controller can be designed in the transformed
linear space. In particular, an incremental form of a PI
algorithm is used, where the reference is present only in
the integral part in order to smooth the control action*9:

Global linearizing control7

with

+ Kl[Cs - yk]uk

Different methods can be used to linearize Equation (1)


and some of them are applied next:

[PI

vp = v&l + KCbk_l - yk] + KcT&*

- yJ/Ti

(11)

where y* is the set point.


The following closed loop transfer function can be
obtained from Equations (11) and (9), by applying the z
transform:

J. Proc. Cont. 1991, Vol 1, May

153

Dissolved oxygen control: J. Ricardo Pbrez- Correa et al.

dgldt = Wk_Wgldy) + Cdif,cv4Wdy)

~Y(z) -

(21)

Y*

p.9 -

T&c]z-2

If the following transformations

[Kc/ ToZ]z-
+ [KCT, + (Kc/Ti) -

It is possible to relate the closed-loop


control parameters as:

CS,cV)(dgldY) = a

(22a)

Cf,ot,u)(dgldY)

(22b)

poles with the

Kc = [0.9 - P,P,]/T,

(13)

Ti = KcTo2/[1.9 - (P, + P2) - KcT,1

(14)

Finally, the control signal can be computed from past


and present observations and control actions:

(1%

Yk)]

= bv

the transformed variables are:


g=y

(23)

v = Q(Ci - y)/V + K,u(Cs - y) - K,L(y)

(24)

where the arbitrary constants were set to a = 0, b = 1.


The discrete linear model then is:
Yk+

K(Cs -

are used:

1.9]z- + 1 (12)

Yk

(25)

&Vk

If a digital PI algorithm (Equation (11)) is used to


control the linearized process, the following non-linear
control law results:

Variable transformation4

The previous method is only applicable when the control


appears linearly in the model. 0gunnaike4 has presented
a method which can be used for more general models of
the type:
dy/dt = F(y,u)

(16)

where F is a general non-linear function of u and Y.


However, the method is restricted to first-order processes.
Equation (16) can always be represented in the following form:
dyldt =

Gfd_d + WX_YJ~)

(17)

where C,, C, are constants, and f,, f2 are general functions.


For the model Equation (l), functions S, and f2 can be
chosen as:

(184

GfiW = a
Cf2~,u)

= Q(Ci - y)/ V - a + K,u(Cs - y) - K,L(y)


(18b)

Equation (16) can be transformed


linear model:

into the following

dgldt = a + bv

(19)

where a, b are arbitrary constants, and g, v are the


transformed output and input, respectively. In Equation
(19) g is only a function of Y, so:

Substituting Equation (17) into Equation (20) gives:

154

J. Proc. Cont. 1991, Vol I, May

uk

[ is-_;;

]uk _ , +

W,(Cs-

yk)]

(26)

This is the same result given by the GLC method (Equation (15)), where:
Kc = [0.1/T,, + Kc]

(27a)

Ti = KcT,z/[1.9

(273

- (P, + P2) - KcT,I

If functions f,, f2 (Equations (18a), (18b)) are defined in


another way, a different control algorithm will result.
This method is therefore more general than GLC, when
the original system is first order. An extension of GLC
for more general non-linear processes has been recently
proposed by Henson and Seborg30.
Generic model controP

In the particular case of model (l), GMC reduces to:


fCv,u,d,t) - Kc(y - y*) - Kzj; Q* - y)dt = 0

(28)

where fi,u,d,t)
is a model of the process. If the nonlinear model given by Equation (1) is used, the following
algorithm is obtained:
~[C@)

y(t)1 + K,W[Cdt) - v(t)1 - K&Cv(O)

KcO, - y*) - Kisi (JP - y)dt = 0

Discretizing and transforming


troller yields:

(29)

into an incremental con-

Dissolved oxygen control: J. Ricardo Phrez-Correa et al.


(30)

-KIUk(CS.-Yk)-~CCYk-Yk+I)+~2[LOlk)-~Cyk+,ll=O

7.6~

Rearranging the above equation results in the same controller obtained by the GLC method (Equation (15)).
The same control algorithm can be derived using any
of the three basic methods described above. The GMC is
simpler to use; however, as noted by Henson and
SeborgO, it is not clear which conditions ensure that a
solution to the GMC problem exists.

Estimation
The above control algorithms require the estimation of
the model parameters. This can be done by discretizing
the process model, and then expressing the difference
equation in a form suitable for recursive parameter estimation. Depending on the data used in the estimation, a
positional or an incremental algorithm can be derived.
Positional estimation
The model Equation (1) can be discretized to give:
~k+I=Yk+TO[~k(Cik-Yk)/~+~I~k(C~-Yk)--K21

(31)

and rearranging:

or in an equivalent way:
=

yk+l

Yk

tTo&J'kl/V

&To

@kTo)/u

(33)

+'kek

where:
+'k

[UkTo(cs

wk

[K,,

yk),

ci,

A recursive least squares algorithm is used to estimate


the model parameters K, and Ci. A variable forgetting
factor3 keeps the estimator alert by discarding past data.
Incremental estimation
Using Equation (33) at time k:
Yk

Yk-I

[T,Qk-,Yk-d/V

&To

+'k-lek-1

(34)

and then subtracting Equation (34) from (33):


AYk+,

AYk -

&t&Y/c

Qk-IYk-d/V

@'k

ok

(35)

where:
wk

[To(csAuk

Ah_Yk)),

To@,

Qk-d/q

A recursive estimation algorithm can be used with Equation (35) in the same way as with Equation (33).

Results
Several simulation runs were carried out. The perfor-

4.44
0

20

Time

110 0

80

60

40

(min)

Figure 1 Minimum time non-linear non-adaptive


adaptive GLC discrete control ( x )

control

(0)

vs non-

mance of the adaptive non-linear control algorithm


(closed loop poles = 0.1) was compared with a well
tuned PID (J& = 1.O, Ti = 0.1, Td = O.l), a minimum
time self-tuning regulator and a non-adaptive non-linear
controller. Measured and unmeasured disturbances,
model parameter variations and set-point changes tested
the robustness of the different control algorithms. A
sample time of 0.1 minute was considered in the simulations. Some of the obtained results are now shown.
GLC vs minimum time non-linear control
A sequence of set-point changes (between 7 and 5) is
applied to the system controlled with a non-linear
control algorithm proposed by Goodwin et aLI and with
an algorithm derived using Kravaris GLC technique.
Figure I shows the better performance of the linearizing
controller: less overshoot, fewer oscillations and less settling time. The non-linear control law proposed by Goodwin is a minimum time controller, which is well known
for its lack of robustness. On the other hand, with the
linearizing controller, the closed-loop poles of the transformed linear system can be set by the designer, giving
the desired behaviour.
Control performance under a sequence of set-point
changes
PID control (Figure 2a) presents large overshoots, oscillations and long settling time. The control action (Figure
26) also shows irregular and undesirable behaviour. A
much better control can be achieved with a minimum
time linear self-tuning algorithm (not shown), where the
control gain adapts itself to changes in process dynamicz4. The responses of both linearizing controllers
(adaptive and non-adaptive) are identical. Figure 2
shows the adaptive case; the output overshoot is negligible, the response time is inherent to the process since the
control is saturated (Figure 26), and no oscillations
occur. The estimated model parameters are biased24, but
the control performance is not affected. Then, as in linear
adaptive control, unbiased parameter estimation is not
always necessary for good performance with non-linear
adaptive algorithms.

J. Proc. Cont. 1991, Vol 1, May

155

Dissolved oxygen control: J. Ricardo Phrez-Correa et al.

4.0
0

-L--l-II
8
Time

1
12

16

1
20

l
24

6.95

1
0

12
Time

(min)

Figure 3

Measured

Table 2

load changes.

Inlet oxygen concentration

16

20

24

(min)

PID (0),

adaptive

GLC ( x )

changes

Time (min)

Ci (mg 1-l)

3.75
9.25
15.25
22.96

5.0
0.5
1.5
3.0

0
0

12

b
Figure 2
b, input

Table 1

Time

Set point changes.

PID (0),

16

20

24

[min)

adaptive

GLC ( x ). a, Output,

Inlet flow rate changes


Time (min)

Q (1 min-I)

3.75
9.25
15.25
22.96

0.070
0.020
0.060
0.040

156

J. Proc. Cont. 1991, Vol 1, May

I
4

I
8

Unmeasured

12
Time

Figure 4

Eflect of measured disturbances


The sequence of measured load changes in Table I is
applied.
The PID control (Figure 3) is stable but slow to
respond, and relatively large deviations from the set
point are observed. The self-tuner responds much faster
and with smaller deviations than the PID control24.
Perfect control is achieved with GLC24; the disturbance
(Q) is measured, and the model parameters are known,
hence the algorithm is a non-linear feedforward controller and generates the exact compensation. A more
realistic situation is also shown in Figure 3, where the
model parameters are estimated. In this case, the adaptive GLC control is not perfect, but much better than the
PID one and the linear self-tuner. However, it is worth
mentioning the effect of the disturbance in the estimated

1.51

disturbances.

I
16

I
20

I
24

(min)

PID (0),

adaptive

GLC ( x )

non-linear model parameters. The real parameters are


constant, but the estimated ones change considerably24
after the disturbance enters the process, even though the
control performance is not affected. This behaviour is
typical of linear self-tuners with positional estimatioS2.
E#ect of unmeasured disturbances
The sequence of unmeasured disturbances in Table 2 is
applied.
The PID controller (Figure 4) and the self-tuner (not
shown) presented similar behaviour as in the previous
case24. On the other hand, the GLC24 performed somewhat worse, since in this case the disturbance is not
measured and the algorithm only operates as a feedback
controller, i.e., it works after the disturbances enter the
system. More interesting is the behaviour of the adaptive

Dissolved oxygen control: J. Ricardo PBrez-Correa et al.


0.7

:
i
0.4-i
0

II
4

11
8

11
12

11
16

1
20

I
24

6.8

12

Time (min)
Figure 5

Table 3

16

Time

Real parameter (0), estimated parameter ( x )

Figure 6

Variations

-0.2

III

20

24

(min)

in K2. Linear self-tuner ( x ), adaptive GLC (0)

Variations of the oxygen uptake rate


Time (min)

K2 (mg 1-l min-I)

3.15
9.25
15.25
22.96

3.3
1.4
2.1
1.9

GLC algorithm. This gives the best control in terms of


the output (not shown), but the control action gets extremely oscillatory (Figure 4). The estimated Kl parameter
takes on very small values (Figure 5), then the controller
gain (see Equation (15)) turns very high giving a very
sensitive controller. The estimated inlet concentration
(Ci) is also badly estimated*4. In this case, the bias in the
estimated parameters induced by unmeasured disturbances causes difficulties in the control.

e
0

III
4

12
Time

II
16

I
20

1)

I
24

(min)

Normalized estimated parameters. Linear self-tuner control


gain (O), Z ( x ), Ci (-)
Figure 7

Eflect of model parameter changes

First, the K2 model parameter changes in the form shown


in Table 3. These are like load changes, since in the
process model (Equation (1)) K2 appears as an additive
constant. Consequently, the PID behaves like in the previous two cases, i.e., in a stable but slow manner. The
self-tuner is more affected. The control (Figure 6) is
better than a PID, but the estimation of the controller
gain suffers drastic variations each time the parameter K2
is changed (Figure 7). The non-adaptive GLC algorithm
is very robust to K2 variations24; deviations from the set
point are very small. The adaptive GLC (Figure 6), as in
the previous simulation, is badly affected by the unbiased
estimation resulting from the unmeasured perturbation.
Figure 7 shows the erratic behaviour of the estimated
parameters (Ci, K,).
Secondly, the K, parameter was changed according to
Table 4.

The PID controller presented a stable, but slow, response (not shown). On the other hand, the self-tuner was
faster (Figure 8a), but the control action was more sensitive (Figure 86) and the controller gain suffered some
variations (Figure 9). A good estimation of the mass
transfer coefficient is very important for the GLC algor-

Table 4

Variations of the mass transfer rate


Time (min)

K, (1-I)

3.75
9.25
15.25
22.96

1.30
1.80
0.92
0.60

ithm, since it represents the reciprocal of the controller


gain. In this case the GLC does not work properly*4; the
process gain changes cannot be compensated by the controller, hence output and input responses are very oscillatory. The adaptive version shows better but still unsatisfactory behaviour (Figure 8). The estimated K, tends to
follow the real parameter variations (Figure 9), but the
estimation of Ci is very bad. Among all disturbances,
variations in K, are the ones that mostly affect the performance of the non-linear controller.
Adaptive GLC estimating only K,

The above results show that the GLC algorithm is only


sensitive to variations in K,, and that the adaptive GLC
cannot estimate properly both K, and Ci. Then, another

J. Proc. Cont. 1991, Vol 1, May

157

Dissolved oxygen control: J. Ricardo Phrez-Correa et al.

71
0

I
16

12
Time

I
20

I
24

6.73
0

12

(mini

Time

Figure 10 Variations
in K,. Adaptive
estimated. Output ( x )

Figure 8
a, Output,

Variations
b input

kl

l2 (min) l6

Time

in K,. Linear self-tuner

(0),

20

16

24

[min)

GLC,

only

K, parameter

is

1
P

IL

2o
adaptive

24
GLC ( x ).

12
Time

Figure 11
1 .l

Real parameter

(0),

16

20

24

(mini

estimated

parameter

(x)

I
I

l.Ok

good, but the parameter estimation presents steady-state


bias. Changes in K2 cause larger deviations, but the controller is very fast and no steady-state bias appears in the
estimationz4. The results for variations in K, are shown in
Figure 10; the control response is the best for this type of
disturbance, and the estimation of the mass transfer coefficient is almost perfect (Figure 11).
Adaptive GLC with incremental estimation

II

-0.1

I
4

12
Time

Figure 9
gain (Oh

Normalized estimated
K, (x ), Ci (A)

16

20

24

(min)

parameters.

Linear self-tuner

control

controller was conceived in which only the mass transfer


coefficient is estimated. This controller was subjected to
the same perturbations applied to the previous algorithms. Specifically, under set point changes, the tracking
is excellent and the estimation of K, does not present
steady-state bias, but deviates during transients. Moreover, the response to measured disturbances is perfect. If
the disturbances are not measured, the control is very

15%

J. Proc. Cont. 1991, Vol 1, May

Since incremental estimation gives very good results in


linear adaptive control, a different GLC controller was
designed where the parameters were estimated using
difference variables.
This controller was tested with the same perturbations
as previously. The algorithm responds well to all disturbances except changes in K,. However, the estimated
parameters presented more bias than the positional
algorithm. The incremental controller is unable to cope
well with variations in the mass transfer coefficient.
Figure 12 shows a very oscillatory response after the first
two changes, and then the estimated coefficient remains
constant (Figure 1.3) despite the changes in K,, and the
control gets better.

Dissolved oxygen control: J. Ricardo Pbrez-Correa et al.

6.701

12

Time

Figure 12
tion ( x )

Variations

20

16

24

(mini

in K,. Adaptive

GLC with incremental

estima-

2.0

1.5-

.a,
g
u9

l.O-

Acknowledgement

VI
2
5

meters will be estimated and how they will be estimated.


In this work, three options were tested:
i) positional estimation of Ci and K,
ii) positional estimation of K, only
iii) incremental estimation of Ci and K,.
The second option gave the best results. In this case, the
estimation of the mass transfer rate was almost always
perfect. Steady-state bias only appeared when the inlet
oxygen concentration changed, but the performance of
the controller was not affected. This controller gave the
best response for all the disturbances tested.
The other two options (i, iii) usually gave good
control, although the estimation of the model parameters
presented bias. However, neither was able to cope satisfactorily with changes in K,. Then, it seems that a proper
estimation of model parameters is more an issue in nonlinear than linear control. It can be said that is better to
estimate fewer parameters, and only those which can
cause problems due to a bad estimation.
The methodology proposed by Ptrez and Solar**
appears to be a simple and effective tool to incorporate a
priori process knowledge in order to design better and
more robust controllers; however, the technique requires
a good understanding of the process to be controlled.

The authors wish to thank project FONDECYT 0622/88


for financial support of this work.

0.5-

:
2
0

III
4

Time

Figure 13

Estimated

parameter

12

(-).

16

II
20

I
24

References

(mini

Real parameter

(x)

Conclusions
A non-linear adaptive algorithm was developed to
control the dissolved oxygen concentration in a waste
water treatment plant. A simplified non-linear model was
employed to derive the controller and to test it through
simulations. It was shown that different linearizing techniques can give the same control algorithm; however, the
GMC approach is simpler to use. An adaptive version of
the more general GLC controller was compared with
different standard controllers under a variety of disturbances. In this particular example, the PID gave a reasonable response to all disturbances, but was comparatively slow and oscillatory. A minimum time linear selftuner always showed better behaviour than the PID.
However, model parameter changes caused undesirable
variations in the estimated controller gain, which sometimes took on the opposite sign. The non-adaptive GLC
with nominal parameters can cope efficiently with a wide
range of .disturbances; however, it is very sensitive to
variations in the mass transfer rate. In this situation,
adaptation is called for.
There is no unique way to design an adaptive GLC
controller. It is necessary to specify which model para-

I Su, R. Sysr. Control Left. 1982, 2,48


2 Reboulet, C., Champetier,
C. Inr. J. Control, 1984, 40, 631
3 Boye, A. J., Brogan, W. L. Proc. American Control Conf., Boston,
USA, 1985,948
4 Ogunnaike,
B. A. Ind. Eng. Chem. Process Des. Dev., 1986, 25, 241
5 Hoo, K. A., Kantor, J. C. Chem. Eng. Common., 1986,46, 385
6 Georgakis, C. Chem. Eng. Sci., 1986,41, 1471
7 Kravaris, C., Chung, C. AIChE J. 1987,33, 592
8 Lee, P. L., Sullivan, G. R. Compuf. Chem. Eng., 1988, 12, 573
9 Isidori, A., Ruberti, A. Syst. Control Lert., 1984, 4, 17
10 Kantor, J. C. Shell Process Control Workshop,
D. Prett and M.
Morari. Eds.. Butterworths.
Boston. USA. 1987. 225
11 Henson, M. A., Seborg, D: E. A Unified Differential Geometric
Approach
to Nonlinear
Process Control
presented
at AIChE
Annual Meeting, San Francisco, USA, 1989
12 McLellan, P. J., Harris, T. J. and Bacon, D. W. Chem. Eng. Sci.,
1990,45,3017
Control Systems, Springer-Verlag,
Berlin,
13 Isidori, A. Nonlinear
1989
14 Kravaris, C., Kantor, J. Ind. Eng. Chem. Res., 1990, 29, parts 1 and
2.2295
15 Goodwin, G., McInnis, B., Long, R. Optimal Control Applications
& Methocis, 1982,33,443
16 Alvarez, J., Alvarez, J., Mondit, S. Proc. IFAC Symp.: Adaptive
Systems in Control and Signal Processing,
San Francisco,
USA,
1983, 105
17 Bastin, G., Dochain, D. Installe, M. Proc. IFAC Symp.: Adaptive
Systems in Control and Signal Processing, San Francisco, 1983, 129
18 Dochain, O., Bastin, G. Automatica, 1984, 20,621
19 Ko, K. Y., McInnis, B. C. and Goodwin, G. C. Automatica, 1982,
18, 727
20 Jutila, P., In/. J. Confrol, 1983,38, 639
D., MacGregor,
J. F., Wright, J. D. Can. J. Chem.
21 Onderwater,
Eng., 1988,1X,478
22 P&ez, R., Solar, I. Proc. IFAC Symp.: Low Cost Automation,
Milan, Italy, 1989, I, T-81

J. Proc. Cont. 1991, Vol 1, May

159

Dissolved oxygen control: J. Ricardo Phrez-Correa et al.


23 Lopez, F., Perez, R., Solar, I. Proc. IX Chilean Cong. of Chemical
Engineering (in Spanish), P&n, Chile, 1989,2,507
24 Lopez, F., MSc. thesti, Chemical Engineering Dept., University of
Chile, 1990 (in Spanish)
25 Bastin, G. and Dochain, D. On-line Estimation and Adaptive
Control of Bioreactors, Elsevier, Amsterdam, 1990
26 Sastry, S. S. and Isidori, A IEEE Trans. Aur. Cont., 1989,34, (1 I),
1123
27 Lee, H. and Marcus, S. L. Inr. J. Control, 1986,44 (4), 1103
28 Grizzle. J. W. and Kokotovic. P. V. IEEE Trans. Aur. COW.. 1988.
33 (9) 857
29 Isermann, R., Digital Control Systems, Springer-Vedag, Berlin,
Germany, 1981,85
30 Henson, M. A. and Seborg, D. E. AIChE. J., 1990,36, (II), 1753
31 Fortescue, T. R., Kershenbaum, L. S., Ydstie, B. E. Auromarica,
1981, 17,831
32 Perez-Correa,

J. R., Ph.D. rhesis, Chemical Engineering Dept.


Imperial College, University of London

Nomenclature
;
d

C,, C*
Ci

CS

160

arbitrary constant
arbitrary constant
disturbance
constants
inlet oxygen concentration (mg 1-I)
saturation concentration (mg 1-l)

J. Proc. Cont. 1991, Vol 1, May

%!u,d,r)
f%?
t??

h(z)
k
K,
K2
Kc
L(x)
p,, p*

Q
r
TO
Ti

V
Y
Y
Y
z
2
w. PI
!J
CD

general non-linear function


general non-linear function
general non-linear function
general non-linear function
general non-linear function
general non-linear function
general non-linear function
discrete time
mass transfer coefficient (I I)
oxygen uptake rate (mg I 1min 1)
proportional controller gain
1 forx>O,Oforx<O
closed loop poles
inlet flow rate (1 min)
controllability index
sample time (min)
integral controller time
control variable in the linear space
reactor volume (1)
general vector
dissolved oxygen concentration (mg I - I)
setpoint
z transform variable
general vector
arbitrary constants
estimated parameters vector
positional estimation data vector
incremental estimation data vector
difference operator

Das könnte Ihnen auch gefallen