Modelling and Predictive Control of An Olive Oil Mill

Download as pdf or txt
Download as pdf or txt
You are on page 1of 6

&VSPQFBO$POUSPM$POGFSFODF

&$$

1PSUP 1PSUVHBM 4FQUFNCFS 

  
     
 
          
  

" $ & ( * & , - / 0 " 2 " 4 5 7 9

Escuela Superior de Ingenieros. Universidad de Sevilla


Camino de los Descubrimientos s/n, 41092-Sevilla (Spain)
Phone: +34 954487348, Fax: +34 954487340
e-mail: {bordons,cueli}@cartuja.us.es

Process Control, Predictive Control, Industrial Olives


: < = ? A B D F H

Processes, Factory Modelling and Simulation, Agricultural Addition water


Processes Heating water

I K - L ( M N L

Decanter (two phases) Addition water


x x

x x

This paper describes the modelling and predictive control of


y x

x x

x x

the extraction process in an olive oil mill. The work is focused


Water
y x

on the thermal part of the process, where the raw material is


q r s t u w s r

x x

prepared for the mechanical separation. The paper shows the y x


g h j k l

Oleaginous
development of a model based upon rst principles combined O
x x

m o p m

phase
with experimental results and validated with real data. Differ-
Termo-mixer Filter Centrifugal
ent control strategies have been tested under simulation, show- Mill Alpeorujo
ing that good performance can be obtained by the use of a pre- pump
(to dryer)
dictive controller that takes into account the measurable distur-
bances that appear in the process. Constraints in actuators are
also included in the control strategy. Figure 1: Process
, L ( & * 4 N L 9 & ,

P R

The automatic control of the extraction of oil out of olives is


still an open eld, since many installations are usually operated O
Therefore the predictive controller is used in this part of the
in manual mode. As olive oil mills are becoming bigger the process.
chances for automation are increasing, therefore it is important The discontinuous way of feeding the paste is the main dif-
to acquire the necessary knowledge of the process behaviour in O culty that appears when trying to maintain the optimal op-
order to design the appropriate control strategies. erating conditions in the . These changes intro-
V W Y [ ] ^ ] _ ` Y [

The process is composed of several operations: reception of duce continuous variations in the level and therefore changes
raw material (olives), washing, preparation, extraction, and in temperature since the quantity of product inside the machine
storage of the produced oil [3]. Figure 1 shows the most im- varies. As level can be easily measured, it can be considered as
portant phases of the process: preparation and extraction. a measurable disturbance and hence can be taken into account
by the predictive algorithm as a action.
Y Y ^ [ | c [

z a z a

The preparation phase is crucial for the whole process it con- T

sists of two subprocesses. The rst one is olive crushing by an O


The control strategy will also consider the existence of con-
especial mill, whose objective is to destroy the olive cells where straints. There exist physical limitations to the heating power
oil is stored. The second one aims at homogenizing the paste by and also operating limitations since temperature must be kept
revolving it while its temperature is kept constant at a speci ed O
inside a range, out of which the quality of the product is drasti-
value (around 35 C). This is performed in a machine called U
cally reduced.
, which homogenizes the three phases of the paste
V W Y [ ] ^ ] _ ` Y [

The paper is organized as follows. In section 2 a description of


(oil, water and by-product) while exchanges energy with sur- the process is presented, whose model is obtained in section 3
rounding pipes of hot water. This is done in order to facilitate using nonlinear differential equations. This model is validated
oil extraction in the following process: mechanical separation with real data obtained from the process. The control strategy
in the . This paper is focused on control
Y b c e V Y [ V W Y [ ] ^ ] _ ` Y [

that is used is described in section 4. The simulated results


since homogenization is really important in the whole process, obtained when applying the predictive controller are described
because bad operation conditions in the can dra-
V W Y [ ] ^ ] _ ` Y [

in section 5 and nally the major conclusions to be drawn are O

matically reduce the quality and quantity of the nal product. O

given.

Proceedings of the European Control Conference 2001 2291

*4#/

Authorized licensed use limited to: Bursa Uludag Universitesi. Downloaded on December 12,2021 at 14:58:42 UTC from IEEE Xplore. Restrictions apply.
          

Measured levels in the three bodies


100
Third body
95
The system considered corresponds to a , whose     " % &  
Second body
First body
90
main objective is to homogenize the three phases of the paste 85
(oil, water and by-product) and keep it at a certain tempera- 80

ture in order to facilitate oil extraction. Heating of the paste 75

is achieved by means of hot water circulating through a jacket. 70


0 1 2 3 4 5

The machine is divided into different (usually three or four) Measured inlet water temperature
80
tanks or , each one with revolving blades to facilitate
) " + %  -

75
homogenizing. The bodies are composed of semi-cylinders 70

about 3 metres long with a diameter of 1 metre. Paste is 65

dropped over one side of the rst body and pushed by the re- .
60

volving blades, which make the paste fall down to the second 55

50
body through the over ow and so on. The existence of sev- /
0 1 2
Time(hours)
3 4 5

eral bodies allows a gradual temperature increment along the


   , since abrupt changes in paste temperature would
" % &  

affect the quality of the end product.


Figure 3: Disturbances: levels and water temperature
1 3 4 5 6

7 8 : ;

< 6 4 = >

8 ;

6 @

? 8 ;

ter to several processes in the factory, so it is affected by load


Q changes.
Therefore, the outlet temperature presents oscillations at the
frequency of level variations with changes produced by heating
water variation. The controller must be able to reduce the effect
of these disturbances as much as possible. Level and water
Q
temperature can easily be measured and level evolution can be
predicted as shown later. Figure 3 shows level evolutions in the
three bodies (the solid one is the last body) and also the random
variation of temperature.

Q A        E   

To decanter
G H I K L M O P M Q R S U P V V Q S Q M Y P R O Q Z [ R Y P L M ^

The plant can be modelled as a thermodynamic process where


Figure 2: Thermomixer bodies. both mass and energy balances can be used for modelling. The
model is obtained by applying the following balance equations
The paste is heated in order to facilitate mixing since the paste [10] to each body (energy balance):
turns more uenty when temperature rises. However, there ex- /

ists an upper temperature limit behind which olive oil loses _ ` a c d e f d h c i

f f

(1)
d e d h d e d h

quality ( avour, fragrance, etc.) due to the oxidation process


/
_ j k m n n p m r

and the loss of volatile components. s t t v w s t z

Another important fact to be considered is the mixing time (res- _ ` a | d e


f }

d h | i

| d e
f }
d ` h
}
h | i

t t
v 

idence time), whose optimal value is around one and a half _ j


k m n p p p

hours. A shorter time drives to incomplete mixing and a longer t


d ‚ d ` h „ h i

one can give rise to emulsions, which interfere with the extrac-
k  p

d ` h h ‡ i

 

t v w … v w …

tion process. k † p

The feeding of the machine with the paste coming from the
Inlet ow in each body is not known, it must be calculated from
crushing mill is done by an on-off level controller that turns the
/

the available measures: levels and outlet paste ow, using the /

feeding pump on when the level is low and turns it off when
following equation :
it reaches a maximum. Therefore the evolution of the level re-
sembles a kind of saw-teeth wave, which has a great in uence / _ ` ‚ ˆ d ‰ d ‹ i

on temperature, constituting an important disturbance. Another _ j k m n p m r (2)


disturbance that appears in this process is the temperature of
the heating water. It comes from a boiler that supplies hot wa-
where:

Proceedings of the European Control Conference 2001 2292

Authorized licensed use limited to: Bursa Uludag Universitesi. Downloaded on December 12,2021 at 14:58:42 UTC from IEEE Xplore. Restrictions apply.

 


Mass of paste and water in the heating jacket 45
R e a l vs s i m u l a t e d o u tp u t


Speci c heat of paste and water


Paste and water temperature


 

 


Inlet paste and water temperature

3º body temperature (ºC)


    

40

Inlet and outlet paste ow


   

Environment temperature
 

35
Water ow


    
Exchanged heat
 
Loss heat 30

Friction heat
Heat transfer coef cient


25

   

68 69 70 71 72 73 74

" #
Loss factor T im e ( h o u r s )
Cross section of the body
Measured level
$

Paste density Figure 5: Measured and simulated temperature (I)


The model takes the form:

Outlet paste flow be different depending on external factors and values that can
Third body level Outlet paste change as inlet paste density (which is not homogeneous) or
Non linear temperature
Heating water temperature heat transfer coef cient.
model

R e a l v s s im u la t e d o u tp u t
Inlet paste temperature 45

Heating water flow

40
3º body temperature (ºC)

Physical parameters
35

Figure 4: Simpli ed model of the thermo-mixer

30

This model includes a loss factor that describes the heat ex-
25
changed with the environment and a constant factor to model 86 87 88 89 90 91 92

T im e ( h o u r s )
the heat generated by friction.
The complete model is obtained assembling the body models,
taking into account that: Figure 6: Measured and simulated temperature (II)

- Paste inlet temperature and ow of body i equal outlet 


& ( & > 7 = 0 5 @ A - / 0 2

temperature and ow of body i-1, except the rst body, 

whose temperature is measured and whose ow is esti- 

A linear model has to be developed in order to design the pre-


mated from level measures. dictive controller. It is obtained from step tests on the plant. In
fact, it is dif cult to do step tests in the real plant, since it is not
- Flow and temperature of the heating water is the same for

possible to maintain some variables in steady state while per-


the three bodies, since they are in parallel. The outlet tem-
forming step tests in other. For instance, since feeding is done
perature is the arithmetic mean of the three bodies.
in an on-off way, level cannot be kept constant while a step in
& ( ) + - / 0 2 4 5 2 7 / 5 ; 7 - =
inlet temperature is performed.
Therefore, the linear model is obtained from simulation using
This model has been validated using real data obtained from the nonlinear model. All manipulated variables can be changed
an olive-oil mill located in Málaga (Spain). Data was obtained independently to see their in uence on temperature behaviour. 

from a series of tests performed in the plant during this year’s With the results obtained from simulations is it possible to nd

campaign. This data was used to estimate many of the para- linear models using simple identi cation techniques [9]. The

meters that appear in the model which are not perfectly known, models needed for control give the nal paste temperature (the

since they depend on several circumstances: kind and moisture paste that leaves the last body of the thermomixer) as function
of olives, dirt in the heating circuit, etc. of ow and temperature of the heating water and level.


Figures 5 and 6 show a comparison of real (the bold one) and After several simulations, the following models were obtained,
simulated output obtained with real input data. The error can in the form of a CARIMA description:

Proceedings of the European Control Conference 2001 2293

Authorized licensed use limited to: Bursa Uludag Universitesi. Downloaded on December 12,2021 at 14:58:42 UTC from IEEE Xplore. Restrictions apply.
of the last body as constant as possible in spite of disturbances
               
 

(level and variations in hot water). The manipulated variable is




(3) the hot water ow.


The process is characterized by the big deadtimes in tempera-
Numerical values for the model that gives temperature as a ture dynamics. What is more, the effect of disturbances on the
function of ow: controlled variable shows faster dynamics (mainly at high pro-
duction rates) that the manipulated variable itself, which makes
disturbance rejection more dif cult by the controller. The con- K

   # % & ( *   +  # % ( / 0   1 3 # % ( * #   7 3 # % # # *   9

(4) trol scheme is shown in gure 8. K

3 # % # ( 0 *   : 3 # % # # ;   3 # % # & >   @ & #  A

 & 3 ( % / 0 


 & % ; # 
 @

3 # % F & 
 G
Temperature

Water
Set point flow
Temperature with respect to level: Controller System

  3 J % # # (   A  J % # & 0   +

Measurable disturbances:
(5) Noise
 & 3 # % 0 ; ; 

level and water temperature

And with respect to water temperature:


Figure 8: Control scheme

  # % # # > &   A
Predictive Control can be an interesting candidate to control
(6) this system. There are many applications of several predictive
 & 3 # % 0 ; *  

controllers in industry [8] and the one chosen for this applica-
tion is Dynamic Matrix Control, DMC [4]. This controller, as
shown in [2], can easily deal with measurable disturbances.
-3
R e sp o n s e to s te p in flo w o f h e a tin g w a te r
x 10
2

1.5
As was said before, the effect of the manipulated variable in the
1 process output is slower than the effect of disturbances. This
0.5

0
fact makes it interesting to include a prediction of the distur-
2
R es p o n s e to s te p in le ve l bances to improve the results. This is a slight change with
0 respect to the standard DMC algorithm, which considers that
-2

-4
disturbances are kept constant (and equal to their current value)
-6 in the future. The information that provides this future evolu-
R e s p o n s e to s te p in w a ter te m pe ra tu re
0.6
tion is very important in this case, allowing the controller to
0.4
anticipate its in uence on the process output.
0.2

0
0 1 2 3 4 5 6
In this application, as the main disturbance acting on the out-
T im e (h o u rs) put (level) exhibits a predictable behaviour, the control law is
calculated considering an Auto-Regressive second order model
[7] of disturbance.
Figure 7: Step responses
The control law that minimizes the general cost function:
The sampling time was chosen as 100 seconds according to
the dynamics of the process. Step responses of the model are 

shown in gure 7. K

Z
 [ \
_ ` a

^     d 3 f    d g
@ h

 i k
_ ` a

     d g
@

It can be observed that temperature response with respect to a b : b

step in level shows an initial inverse behaviour. This is due to (7)


the fact that a sudden cold paste inlet increment reduces the
temperature in the thermomixer until it recovers once the mix- is given by:
ing process has taken place.    p r p  i t  p r  [  3 f g

(8)
b u u w

L N P Q R S P T U R S V R W X Y

where
The control objective is to maintain the operating conditions [ y z

in the thermomixer, that is equivalent to keep the temperature - is the expectation operator

Proceedings of the European Control Conference 2001 2294

Authorized licensed use limited to: Bursa Uludag Universitesi. Downloaded on December 12,2021 at 14:58:42 UTC from IEEE Xplore. Restrictions apply.
 

- is the vector of future control action increments The expression for assuming a truncated step response


model is (see for instance [2]):


- is the calculated free response without disturbances
 

-  is the expected value of the free response due to




measurable disturbances 



K

(13)
 J

 H  H  W

-
is the reference trajectory L N

G P Q S T P

   

where
is the samples of the truncated step response and
is the increment of the disturbance signal in the instant
P
Q

The objective function can be expressed as


 H

H
T

. P

The expression (13) can be separated in two terms, the rst one -

containing past values of the measurable disturbance and the


# % '

(9)
 

    
  

second one containing future values.


Note that the operator only applies to the rst term of the  ) + , -

functional. Remember that the predicted value for the output F G

is the sum of two terms, the rst one due to the control law -

  % Y


K

[
% Y %
K

J
% Y

and the other one due to measurable disturbances. The future  H

L N

G P
Q S

T P
 H  W

L N

P
Q S

T P
 H  W

values of the measurable disturbances are stochastic. The rest [

(14)
of variables are deterministic.
% 

Lets substitute the prediction of the output by in  .

Applying the expectation operator to the equation (14), we ob-


equation (9). The free response has two terms, due respectively tain the desired expression:
to the past control law and to the measurable disturbances. So   2 % 


it can be expressed as . The result is:




F G

@   % Y B @ % Y B %

 [

  H   H  W

%  %  # % '

(10)
  

L N

Q S

G P T P

   . 
 . 

%  2 %  %  2 %  % '

  #


  . 
 . 

% % Y

 H  W (15)
L N

Q
P S T P

Applying linearity to (10) and rearranging terms leads to (11) [

The expected value for the measurable disturbance can be eas-


 
% ' 9 % : 

2
% 


% ily calculated most cases with an optimal prediction. This can
%
 .


2
.

%   


2


%  


#

.

be done if the disturbance is a stationary or quasi-stationary


   
 
(11) signal, so it can be modelled as an AR process (as it has been
done in this work). For the trivial case, in which the polyno-
mial AR is equal to 1, we obtain the tipical assumption that


And solving we can obtain the control law:


 ?

considers future values of disturbance as constant.


=

^ ` a b c d e b

: % ' 9 : @ @    B

 . . . 
 

This section presents the results obtained when applying the


control strategy previously de ned to the simulated model of -


% ' 9 F G @ 
2

  B

the thermomixer. Notice that olive oil production is a highly


seasonal process. Model validation was done with real data
 . . .  
 

% ' 9 F G @    B

(12)


 . . .
   

available this year and the controller will be tested in the real
plant during next year’s campaign.
That indicates that the best control law should include the best
The rst graph shows the effect of disturbance estimation on
prediction for the future values of disturbances. The standard
-

the outlet temperature. Figure 9 shows the clear effect of in-


DMC algorithm makes the simplest assumption that distur-
cluding the AR model of level in the control algorithm. No-
bances will keep constant along the horizon. In this work, dis-
tice the improvement of the output response when considering
turbance estimation is included.
measurable disturbances. The dotted line is the simplest DMC
The expected value for the part of the free response due to algorithm, without considering measurable disturbances. The
measurable disturbances can be calculated as follows in equa- next approach (thin solid line) includes explicitly constant dis-
tion (15). turbances along the horizon and the next strategy (bold line)

Proceedings of the European Control Conference 2001 2295

Authorized licensed use limited to: Bursa Uludag Universitesi. Downloaded on December 12,2021 at 14:58:42 UTC from IEEE Xplore. Restrictions apply.
   
     

uses a second order AR model to estimate future evolution of


disturbances. Notice that the control action in this case al-
ways goes before that calculated without considering measur- This work has presented the modelling and predictive control
able disturbances. Although the output has been simulated with of an olive oil mill. The method was developed as a result of
the linear model of the plant, the tests have been done using studies for the control of a real plant located in Spain, where
values of level obtained from the real plant. real data has been taken for validating the model. The control
strategy proposed will be tested there during the next campaign.
DMC applied to lineal simulator Simulation results have shown that a DMC considering esti-
Third body temperature (ºC)

36
mation of level variations and constraints in the manipulated
34
variable can be a good solution for the problem that exists in
32 No measurable disturbance the real plant.
Constant disturbance estimation
AR model disturbance estimation
30 Set point 


           ! 

25 26 27 28 29
4
x 10
3
The authors would like to acknowledge E.F. Camacho, M.R.
Heating water flow(l/h)

2 Arahal, D. Limón and D. Rodríguez for collaborating in the


1 project and people from Sociedad Cooperativa Ntra. Sra. de los
0 Remedios for their help in carrying out the tests. This work has
-1
been supported by EC under FEDER programme under grant
-2
25 26 27 28 29
1FD97-0836.
Time (hours)
#  $  &  
 

Figure 9: Results from linear simulator [1] K.J. Astrom and B. Wittenmark. "Computer Controlled
Systems", (1984) ' ( * , . 0 1 * 3 5 7 8 8

When testing the controllers on the nonlinear model, the be-


haviour deteriorates. Figure 10 shows the controllers in a non- [2] Camacho, E.F. and C.Bordons. "Model Predictive Con-
linear simulation. This is due to the fact that the manipulated trol”, (1999) 9 : ( 0 , = * ( 3 ? * ( 8 7 = @ A C , E C ,

variable saturates during the experiment, reaching its physical [3] Civantos, L. “Obtención del aceite de oliva virgen”, F E H

limits. Comparing the responses, it is clear that the proposed I

(1999) = ( J 1 C 8 7 F L : 7 N C 8 7 @ 9 H
I

@ P Q , 9 : 7 , 0 L U V

strategy behaves better than the other, although the difference


is not as clear as in the linear case. For instance, values of IEA [4] Cutler, C.R. and B.C. Ramaker. “Dynamic Matrix Con- I W

(Integral of Absolute Error) for the three cases are 0.64, 0.52 trol - A Computer Control Algorithm”, Q , . C X 7 . 0 1 Z C , 3

and 0.35. This gure also shows level evolution, which apart (1980) . ( C 8 Z C , ] * ( * , 1 * @ 9 7 , _ ( 7 , 1 0 L 1 C H

from being oscillatory, present sudden changes. These changes


are due to stops in the feeding, which cause temperature to in- [5] Ljung L. “ System Identi cation - Theory for the User”,
crease since there is no paste inside the machine.
' (1987) ( * , . 0 1 * 5 7 8 8 @ F , = 8 * ` C C E Z 8 0 ] ] L @ b H c H

[6] Ljung L. and T. Glad. "Modeling of Dynamic Systems",


38
Heating water flow(l/h) Temperature (ºC)

No measurable disturbances ' (1994) ( * , . 0 1 * 5 7 8 8 @ F , = 8 * ` C C E Z 8 0 ] ] L @ b H c H

36 Constant disturbance estimation


AR model for disturbances
34 [7] Oppenheim J. and A.S. Willsky. “Signals and Systems”,
32 ' (1985) ( * , . 0 1 * 5 7 8 8 @ F , = 8 * ` C C E Z 8 0 ] ] L @ b H c H

15 16 17 18 19 20 21
6000 Time (hours)
[8] Qin, S.J. and T.A. Badgwell. "An overview of Industrial
4000 Model Predictive Control Technology." , Q , Z U * X 0 1 7 8

2000 ' ( C 1 * L L Z C , . ( C 8 e L L * L L X * , . 7 , E b * ` g 0 ( * 1 . C , L ] C ( h * 3

I W

0 L * 7 ( 1 U H Q , Q Z U F 9 l X : C L 0 X 9 * ( 0 * L o p q @ r o H c * ] ] ( * l Z H

15 16 17 18 19 20 21 s

1.5 Time (hours) 7 , . C ( @ Z 7 ( 8 C L F H t 7 ( 1 J 7 7 , E w ( 0 1 * Z 7 ( , 7 U 7 , F E L H ' : e

(1997)
y y y z
Level (m)

1
o 3 q H

0.5 [9] Söderström T. and P. Stoica. “System Identi cation”,


0 ' (1989) ( * , . 0 1 * 5 7 8 8 Q , . * ( , 7 . 0 C , 7 8 @ A C , E C , H

15 16 17 18 19 20 21
Time (hours)
[10] Thompson E.V. and W.H. Ceckler. “Introduction to
Chemical Engineering”, (1979) { 1 H t ( 7 ` 3 5 0 8 8 H

Figure 10: Results from non linear simulator

Proceedings of the European Control Conference 2001 2296

Authorized licensed use limited to: Bursa Uludag Universitesi. Downloaded on December 12,2021 at 14:58:42 UTC from IEEE Xplore. Restrictions apply.

You might also like