III. MARCO TEÓRICO
3.1 La infancia en edad escolar
Controller with temperature measured after holding tubes
• • I I I ,. ... I I I
---1-
--�_l---\�
----�-
----1---1-
- - - --- . .... ". I .��.'. I I---�:/ ---�\
..-�
----:?--�
--h-
-J
----[--
---- - --- - .... '! .: .. .' "c ;') • .:� :1'1C • • j,l" .... J \",.: 1 ... � : : I I I I - - - -.� ., - - - r- --
-- - --
- - -- --, - - 10:- : : : .' I I I .' : : .� I I - - - -l-- - - -�- - - -�- - - - .' , .� : .,.4·' : I I ---�� --- - -- -� - -- - -- -- - - --- -� - --- - - � - - - - ,l' : : : :r I I I �- I I I ,. I I - - �- - - - , , ,Figure 7-13 : Disturbance rejection criteria, for the third and first DSI unit controller
configurations.
Cascade controller
We will now consider the third controller configuration in Figure 7-8. This is a cascade controller, which measures the tube surface temperature before the holding tubes and the milk temperature after the holding tubes. The tube surface temperature response has an additional first order dynamic, which we have modelled with a 2 second time constant. The resulting transfer functions are similar to those shown in Figure 7-9. We can now determine the closed loop transfer functions by choosing controller parameters for the cascade controller. Using the
Parameters r = 20s and I
K
c =�
6 for the inner controller and T I =1 50s andKc
=�
for the outer 60controller we can produce the disturbance rejection criteria and this is shown in Figure 7-1 3 . Also shown i s the criteria for the first controller configuration i n Figure 7-8. Clearly the cascade controller provides significantly better disturbance rejection.
Evaporator Controllability Studies
Finally we are interested in the controller input saturation for the DSI unit temperature control loop. The criteria for this was discussed earlier (i.e.,
IG(j.@) I
�IG
d(j.@ � - 1 V @ )
and we can seefrom Figure 7-9 and Figure
7- 10
that it is satisfied. We can also determine the manipulations required for a unit disturbance and the resulting magnitude plots are shown in Figure 7- 14. The magnitude is smaller than one across the entire frequency range, so there are no saturation problems.Temperature measured after DSI unit Temperature measured after holding lubes
6 --- ----�- - -- - -- ---�--- ---�---�--- , , , , , , , , , , , , , , , , , I I I I o - - - -�- - - -�- - - -1- - - -�- - - - I I I I I I I I I I I I I I I I I I I I I I I I , , , -g -5 I I I I - - - -�- - - � - - - -,- - - -,- - - - I I I I I I I I I • I I I I I I • I I I I I I I I • I I ... ... ... _.=��.;tr't.·.:·:'.:i� ... -__ .. _ _ _ � .. ... .. ... ... ... ... ... � ... .. _ _ ... __ .. _ _ .. _
�
_ .. .. ... .. .. .. .. .. .. .. .. .. , , , , ,Figure 7-1 4 : Closed loop manipulations required for a unit disturbance. 7.4.4) MVR section control
Here we are interested in the control of the
MVR
evaporator section evaporation temperature( Tel ) and the product dry matter ( W p5
)
. In the previous section we showed that decentralised controllers could be used if the process cross-over frequencies were above0.01
rad/s. Furthermore, the results of the staticRGA
analysis shows that the correct control loop pairings are Tel/ M
c and W pS/ N
camp . Figure 7-1 5 shows theMVR
evaporator section, with the temperature and product dry matter control loops.MVR
section temperature
The MVR evaporator temperature can be controlled using a cascade controller connected with
the preheat plate heat exchanger. Earlier we calculated the
RGA
by considering the heatexchanger condensate flow
(Mc)
as the manipulated variable. However, the cascade controller inFigure 7- 1 5 provides better disturbance rejection for the plate heat exchanger outlet temperature. The set-point for the outlet temperature controller
(Tmcsp)
is then the manipulated variable for theMVR
evaporator temperature control loop. In Chapter 4 we developed the linear state space representation for theMVR
evaporator section. This includes the transfer function between the preheat condenser inlet temperature(Tmcsp)
and the evaporator temperature(r.J
The resultingBode plot for this transfer function is shown in Figure 7-1 6. It should be remembered that this transfer function neglects the preheat plate heat exchanger dynamics. These have been neglected because of the disturbance rejection capabilities of the plate heat exchanger temperature control loop. MVR FAN
:
e-.---J
1 _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ __ _ _ _ _ _ _ _ _ _ _ _ _ ._. - . . . _ . . _ . _ . . . _ • . . _ . _ • • ' · ' . - . 0 - _ . _.
....
... ... ....
... ... ... ....•.•.. ... _ . . . ....
...Figure 7-1 5 : The MVR evaporator section density (DC) and temperature (TC) control
loops.
There are not many direct disturbances to the
MVR
evaporator section. In fact, the control loops for the preheat plate heat exchanger and DSI unit temperature would suggest that there are no disturbances. Strictly the feed temperature disturbances are removed by the heat exchanger outlet temperature control loop. In addition the potential steam pressure disturbances are removed by the DSI unit temperature control loop. However, there are many indirect disturbances which cause changes in the DSI preheat section flash vessel temperatures. For example the presence of air in the flash vessels was shown in Chapter 6 to cause changes in their temperature. As a result,the preheat section flash vessel temperatures can be a significant disturbance to the
MVR
Evaporator Controllability Studies
We now consider the
MVR
evaporator section in isolation from the DSI preheat section. In Chapter 4 the linear dynamic model was developed by combining the DSI preheat section with theMVR
evaporator section. However, for the consideration of theMVR
evaporator section temperature control, we consider theMVR
evaporator section in isolation. This is done to simplify the analysis and to make the flash vessel temperatures a disturbance to the evaporator. The transfer functions shown in Figure7-1 6
are for theMVR
evaporator section and do not include the feedback through the DSI preheat section.The phase cross-over frequency of the
T.l(S)
transfer function is determined from Figure7-1 6
asTmcsp(s)
0.3 rad/s. Figure
7-1 6
also shows the Bode plot for theT.l(S)
transfer function, from which weTph2(s)
can determine a gain cross-over frequency of 0.0 1 rad/s. This is a very low frequency and more importantly it is considerably lower than the process transfer function phase cross-over frequency. As a result, we do not expect there to be any disturbance rejection problems with this control loop. We can confirm this by calculating the closed loop transfer function. Choosing the
controller parameters Ti = 1 00 s and Kc = 5 , we can determine the disturbance rejection criteria
and this is plotted in Figure
7- 18.
Clearly the gain is low across the frequency range and thisreflects the good disturbance rejection characteristics of the control loop.
�
0 _ _ _ _ _ _ _ _ _ _ _ _ _ _6, -20
Figure 7-16 : Bode plots for the MVR evaporator temperature transfer functions.
The potential problem of controller input saturation can be investigated using the criteria of Section 7.2 (i.e.,
IGCi·w) I
2:IGd (j.w� - 1
at frequencies whereIGd Ci.w� >I ).
Figure7-16
showsthat the process transfer function gain is larger than the disturbance gain at frequencies below 0.25 rad/s. This is consistent with the criteria and therefore we do not expect any controller saturation problems. We can confirm this result by determining the closed loop manipulations
required for a unit disturbance, which are shown in Figure 7-1 9. Clearly the manipulations are small and so no serious saturation problems will occur.
MVR
product dry matter
F or the product dry matter control loop, the disturbance is the feed dry matter
(wjc2)
and themanipulated variable is the MVR compressor speed
(Ncomp).
In Chapter 4 we developed the state space representation of the MVR evaporator. Using the linear dynamic model we can determine the Bode plots between theMVR compressor speed (Nco",p)'
the feed dry matter(wfc2)
and theproduct dry matter
( wps).
The model combines the DSI preheat section with theMVR
evaporator section, but the feedback through the plate heat exchanger is neglected. It should also beremembered that the model considers the entire
MVR
evaporator as a single pass, whereas it actually contains five passes. Later we show that this assumption causes some differences in the process dynamic response. However, the basic conclusion of the controllability analysis is the same when the five passes are modelled as a single pass.Figure 7-17 : Bode plots for the