PFC Heat Exchanger 1 2001

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

Chemical Engineering and Processing 40 (2001) 449– 457

www.elsevier.com/locate/cep

Predictive functional control of a counter current heat exchanger


using convexity property
M.A. Abdelghani-Idrissi a,*, M.A. Arbaoui a, L. Estel a, J. Richalet b
a
L.R.C.P., INSA de Rouen B.P. 08, Place Emile Blondel, 76131 Mont Saint Aignan Cedex, France
b
ADERSA, 10 Rue Croix Martre, 91873 Palaiseau Cédex, France

Received 19 May 2000; received in revised form 31 August 2000; accepted 1 September 2000

Abstract

This paper deals with the global modelling and the predictive functional control (PFC) of a tubular counter current heat
exchanger. The hot product flowing through the inner tube is cooled and its outlet temperature is monitored under varying the
flow rate of cold fluid circulating in the annular duct. A global model representing the response to inlet temperature variations
is used to implement the PFC algorithm. The control law takes into account the convexity property of the heat exchanger, which
distinguishes the linear effects of state perturbations from the non-linear effects of structure disturbances. The control equation
corresponds to a generic algebraic solver, which enables to assess the inlet temperature or the flow rate of cold fluid. In this study,
the manipulated variable used to control the heat exchanger is the flow rate of cold fluid corresponding to a parameter of the
dynamic model while the inlet temperature, which is the principal input, is kept constant. The PFC algorithm is then ‘parametric’
and the manipulated parameter is derived from the control equation. The robustness of this controller has also been studied when
inlet temperatures and flow rate of product are subjected to sudden fluctuations. © 2001 Elsevier Science B.V. All rights reserved.

Keywords: Heat exchange; Counter current; Convexity

1. Introduction real time by the actual measurements of the flows and


temperatures, a permanent natural adaptation of the
Heat exchangers are extensively used in all kinds of controller to the environment is feasible.
industries to control temperature of liquids [1,2]. A A black box model in which the parameters have no
simple PI controller, manually tuned, is most of the physical significance cannot be used. Complex, finite
time sufficient to control the product outlet tempera- elements, distributed parameter models with iterative
ture. But even in that case if the pipes are long enough, solvers cannot be implemented on line. An intermediate
the tubular counter current heat exchanger exhibits a approach is necessary: simplified models based on en-
time delay and PID controllers lose their efficiency. The thalpy and mass balances in a limited number of com-
parameters of the model linking the different tempera- partments are used, provided they have been identified
tures vary with the flow rates of product and fluid and on actual data in order to have a proven predictability.
the controller tuning should vary with the flows, which They are based on a set of physical equations with a
might be a tedious work. The PID controllers are thus clear physical significance and thus, they can be used
no more up to what an industrial user would expect for on-line adaptation. Their limited accuracy is com-
from an efficient modern control system. Model based pensated by their restricted use in a feed-back loop that
control uses a model of the processes in real time. If desensitises their uncertainty. The Convexity Theorem
this model is based on First Principles and if it is fed in (see Appendix A) is a convenient way to represent heat
transfer, and clearly separates the linear effect of tem-
perature changes (state variables) from the non-linear
* Corresponding author. Tel: +33-235-140051; fax: +33-235-
146371.
effect of flow variations (parametric changes). Richalet
E-mail address: ahmed.abdelghani@univ-rouen.fr (M.A. Abdel- et al. [3–5] developed a predictive functional controller
ghani-Idrissi). (PFC), in which the effort was focused on the easiness

0255-2701/01/$ - see front matter © 2001 Elsevier Science B.V. All rights reserved.
PII: S 0 2 5 5 - 2 7 0 1 ( 0 0 ) 0 0 1 4 3 - 4
450 M.A. Abdelghani-Idrissi et al. / Chemical Engineering and Processing 40 (2001) 449–457

to tune and to implement in a simple controller such as exchanger. Two valves make it possible to monitor the
PLC’s. It has been used in fast processes (e.g. mechani- flow rate of the cold fluid and the product via this card.
cal systems) and slow processes (e.g. furnaces). The Physical and geometrical parameters of different ele-
stability of model predictive control has been the sub- ments of the heat exchanger are reported in Table 1.
ject of several studies [6,7]. This paper presents the
application of PFC to a counter current heat ex-
changer. Based on a First Principles model, the generic 2.2. Assumptions and modelling
control strategy can use as a manipulated variable
either the inlet temperature of the fluid or its flow rate. The model used in this study represents the response
In this study, the outlet temperature of the product to inlet temperature variations. The assumptions con-
flowing in the inner tube is adjusted under varying the sidered here to derive the differential equations are:
flow rate of the cold fluid circulating through the outer “ fluids are in turbulent flow;
duct, while the temperatures at the inlet are kept con- “ fluids are incompressible and single phased;
stants. In this case, the process is monitored by the “ thermophysical properties of the fluids are assumed
parameter while keeping the principal inputs constant. to be constant;
The controller is then a parametric one and the ‘manip- “ heat capacity of the tube wall is neglected;
ulated parameter’ is calculated from the control law. “ axial heat conduction in the tube wall is neglected.
Considering energy balance overall physical system, a
global modelling of the system is used. The transient
behaviour of the heat exchanger to inlet temperatures
2. Description and formulation steps can be described by a first order response with a
delay time [8–10]. In this case, ordinary differential
2.1. Experimental de6ice and description equations are obtained to characterise the global tran-
sient response of the system. This method is usually
The process studied in this work corresponds to a used when investigations focus only on the system
tubular counter current heat exchanger as described in outlets without interesting on its spatial profiles. In
Fig. 1. The inner tube is copper-made and the outer one addition, the steady state of this intermediate model is
is in stainless steel. The experiments are performed to compatible with the criteria of systems controllability
test the PFC using a water– water counter flow in the [11].
heat exchanger. Temperature probes are placed at the In order to derive the governing differential equa-
inlet and the outlet of the tubular insulated heat ex- tions, the heat exchanger is considered globally as
changer for both fluids. Their diameter is equal to 0.001 indicated in Fig. 2. In this case, the average tempera-
m and their rising time is less than 0.5 s. The sensors tures of product and cold fluid may be used to charac-
are linked to a data acquisition card inserted into a terise energy content of both sides. While flowing
computer, which enables to acquire and display the through the occupied volume, hot fluid transfers heat to
temperatures at the inlets and the outlets of the heat the wall by convection. Therefore, its outlet enthalpy

Fig. 1. Experimental set-up with its electronic cards and system acquisition.
M.A. Abdelghani-Idrissi et al. / Chemical Engineering and Processing 40 (2001) 449–457 451

Table 1
Physical and geometrical properties of the heat exchanger

k (W K−1 m−1) Cp (J kg−1 K−1) z (kg m−3) D (m) a (m) L (m)

Inner tube 384 394 8900 0.02 0.001 4.5


Outer tube 45 490 7850 0.04 0.003 4.5

and internal stored thermal energy are reduced. Energy case, the product outlet temperature can be written in
balance applied to the occupied volume of hot fluid the following form:
leads to the first equation of system (Eq. (1)) after
T op(p) = Fp1(p) e − tdpp
T ip(p)+ Fp2(p)T ic(p) (5)
simplification and rearrangement. Similar energy bal-
ances on cold fluid give the second equation of the The same analysis is also valid for outlet temperature of
system (Eq. (1)). The heat exchange between cold fluid cold fluid versus inlet temperatures.
and product considered global includes convection and Note that in the steady state, the outlet temperatures
conduction through the separating wall. The average can be expressed as function of inlet temperatures in
overall heat transfer coefficient is used in system (Eq. the following form (see Convex Theorem of Appendix
(1)). A):

Á
à z p C p
p V p
d T op
= m; pC pp(T ip −T op) + UA(T oc −T op) ! T op = kpT ip + (1− kp)T ic
(6)
Í
d t
(1) T oc = kcT ic + (1− kc)T op
o
Ãz C c V d T c = m; C c (T i −T o) +UA(T o −T o)
Ä c p c
d t
c p c c p c
where kp and kc are given as:

This same global approach is also given in references Á vp


[12,13]. In order to derive the simple model to be Ãkp = [1− (1− v ) (1− v )]
Í p c
(7)
implemented with the PFC algorithm, the system (Eq.
Ãk = vc
(1)) is rearranged in the following way: Ä c [1− (1− vp) (1− vc)]
Á d T op kp and (1– kp) represent the static gains while the inputs
Ã~p d t + T p = vpT p +(1 −vp)T c
o i o

Í (2) correspond to inlet temperatures. The same analysis is


o
à ~ d T c + T o =v T i +(1 −v )T o also valid for cold fluid. This steady state model (Eq.
Ä c
d t c c c c p
(6)) expresses the Convexity Property, which distin-
guishes the linear influence of inlet temperatures repre-
The time constants are expressed as: senting the principal inputs from the non-linear effect
zpCppVp zcCpcVc of the flow rates, which are the parameters of the
~p = and ~c = (3) system. This important Convexity Property allows to
m; pCpp + UA m; cCpc +UA
derive in the PFC algorithm a generic law permitting to
The parameters vp and vc are given as: control the outlet temperature of product by manipu-
lating either the inlet temperature or the flow rate of
Cppm; p Cpcm; c
vp = and vc = (4) cold fluid with the same approach. The aim of this
Cppm; p +UA Cpcm; c +UA work is the control of the product outlet temperature
System (Eq. (2)) representing the model allows to
obtain the transfer functions Fp1(p) between product
outlet temperature and inlet temperature and Fp2(p)
when the input corresponds to the cold fluid inlet
temperature. These transfer functions do not take into
account the fluids dwell times which are linked to the
transport of fluids from the inlet to the outlet. A delay
time has to be added in order to express the relation-
ship between outlet temperature and inlet temperatures.
When the heat exchanger is in the counter flow configu-
ration, the effect of product inlet temperature on
product outlet temperature appears after a delay time
corresponding to the dwell time tdp. But the effect of Fig. 2. Global energy balance taking into account convective heat
cold fluid inlet temperature is instantaneous. In this transfer and counter current flow.
452 M.A. Abdelghani-Idrissi et al. / Chemical Engineering and Processing 40 (2001) 449–457

under varying the flow rate. Therefore, the analysis is


focused on the parameter kp, which is given as a
function of the number of transfer unit:

(1 +NTUc)
kp = (8)
(1+ NTUp +NTUc)

where NTUp and NTUc are defined as:

UA UA
NTUp = = Gpgp and NTUc = =Gcgc
m; pCpp m; cCpc
(9)

gp and Gp are defined as follow:

1 UA Fig. 3. Predictive control principle.


gp = and Gp = (10)
m; p Cpp
(C(n)− ypr(n)) (1− u H)= i1yB1(H)+ h Hypr(n)−ym(n)
Note that the time constants can be obtained from the (11)
parameters kp and kc.
The proposed System (Eq. (2)) represents a simple where
lumped model where each stream is considered as a DT

mixed tank. A first order system characterised by a gain u= e (TRBF/3) (12)
and a time constant is also discussed by Gögüs et al.
[14] and Ataer et al. [15,16]. Simple models may differ yB1 is given by the following expression:
slightly from the experimental results but their errors
are generally compensated by the use of models in a yB1 = K(1− h H) (13)
feed back loop that desensitises their accuracy provided
they are based on physical representation. where

h=e − DT/~ (14)

3. Predictive functional control and K is the static gain of the system.


It is worth noting that the control Eq. (Eq. (11))
The model presented in the section above represents describes a universal approach to control heat exchang-
an approximating first order response to change of inlet ers. Indeed, the outlet temperature is a function of inlet
temperatures with a delay time. It is known that while temperatures and flow rates. The theorem of convexity
the delay time is important compared to time constant, distinguishes the state perturbations from the structure
the PID controller looses its performance and effi- perturbations. The latter have non-linear effects on the
ciency. This is the case of a long tubular heat ex- product outlet temperature. Relation (Eq. (11)) repre-
changer. In addition, if perturbations affect the inlet sents a simple algebraic solver, which enables to assess
temperatures and flow rate of the product, the PID the inlet temperature or the flow rate of the cold fluid.
controller cannot reject correctly these perturbations In addition, this control law is naturally adaptive when
due to delay time and non-linearity. Therefore, the changes occur on inlet temperatures and product flow
control of this kind of system requires the use of rate. In this work, the heat exchanger is monitored
advanced control based on dynamic internal models. under varying the flow rate but not under varying the
As discussed before, the predictive control has been inlet temperature. Therefore, instead of calculating the
applied successfully in many industrial applications [4]. principal input corresponding to the inlet temperature,
The principle of this method is shown in Fig. 3 and is the flow rate is then assessed. The algorithm used in
based on the elements described in Appendix B. this study corresponds to a Parametric Predictive
Some industrial applications are constrained in order Controller.
to ensure the safety of the process. Moreover, different
elements of the process could be physically limited in
their functioning. The aspect of these constraints can be 4. Results and discussion
added easily to the PFC algorithm [5,7] by bounding
the manipulated variable. In order to identify the model, experimental results
The command law is recalled from Reference [5] by are achieved in the steady state. The parameter kp can
the following expression: be extracted from the experimental steady state as:
M.A. Abdelghani-Idrissi et al. / Chemical Engineering and Processing 40 (2001) 449–457 453

Fig. 4. Parameter kp as function of cold fluid flow rate for different


values of product flow rate.

T op −T ic
kp = (15)
T ip − T ic
Fig. 4 shows the experimental kp as a function of cold
fluid flow rate for different values of product flow rate.
In the same figure, the theoretical values of kp are
reported. An average error of 8% is observed for 0.03
kg/s, 9% for 0.05 kg/s and 5% for 0.06 kg/s. This
difference is quite reasonable for this identification, Fig. 6. Closed loop response to steps of set value with constraint on
which will be used in the algorithm to compute the cold fluid flow rate. (A) Temperature. (B) Error. (C) Cold fluid flow
manipulated variable. rate.
Before implementing the algorithm, a simulation has
been achieved to test the performance of the PFC rejects this variation by maintaining the outlet tempera-
controller. The transfer function Fp1(p) is approximated ture at the set values. The variation of product inlet
by the well known Broı̈da transfer function: temperature has also been tested. The controller main-
tains the outlet temperature at the set values.
kpe − ~dp
Fp1(p) = (16) The algorithm has been implemented on a computer
1+ ~p to control the outlet temperature by the valve aperture
Note that the delay time indicated in expression (Eq. as indicated previously in the experimental device. Fig.
(16)) has no physical signification but it is used to 6 shows the experimental response results to different
compensate the order of the function F1p(p). steps applied on the set value. The temperature evolu-
This result is shown in Fig. 5, which represents the tion is plotted in graph A, the error in graph B and the
temporal evolution of output temperature of the flow rate of cold fluid in graph C. This example is
product in many configurations. In this simulation, the achieved without any perturbations. Indeed, the inlet
flow rate of product, which is considered as a perturba- temperatures and the product flow rate are kept con-
tion has been submitted to a sudden change. The PFC stant during this experiment. The product flow rate is
0.06 kg/s, and the inlet temperature of cold fluid is
about 17°C. At the beginning, the outlet temperature
surpasses the set value due to the initialisation of the
flow rate of cold fluid. This observation is also visible
on the error. Afterward, a negative step is applied on
the set value. At once, the variable of control reacts in
order to respect the set value. The manipulated variable
being constrained in the PFC algorithm (m̄c Bm̄csat =
0.35 kg/s), the outlet temperature is slightly shifted
from the lowest set value. This constraint is clearly
indicated in the graph C of Fig. 6. The positive step
does not induce the saturated flow rate. In this case, the
controlled variable reaches the set values without sur-
pass. Fig. 7, obtained in the same previous control
conditions with an increased constraint value, illustrates
Fig. 5. Simulation of predictive functional control of a counter the product outlet temperature, the error evolution and
current heat exchanger in a different desired configuration. the cold fluid flow rate. This figure shows the good
454 M.A. Abdelghani-Idrissi et al. / Chemical Engineering and Processing 40 (2001) 449–457

experimental results. Indeed, the outlet temperature


does not exceed the set value and has not shifted from
the desired value. This result is obtained with a
smoothed manipulated variable as indicated in graph C
of Fig. 7. In the experiments of Fig. 6 and Fig. 7, the
desired closed loop response time TRBF is adjusted to
45 s and H to 1 s after the delay time. When changing
the control parameters such as TRBF or H in order to
increase the rapidity of the closed loop, oscillatory
behaviour of the manipulated variable occurs. The re-
sults of Fig. 6 and Fig. 7, take into account the
compromise between the rapidity of closed loop re-
sponse and the stability of the manipulated variable.
In order to test this technique when the product inlet
temperature is subjected to a sudden change, Fig. 8
shows the response to step of desired value and to
negative change of T ip. Graph A corresponds to the
temperature, B is relative to the error and C presents
the flow rate of cold fluid. The inlet temperature varia-
tion occurs at 372 s. The controller reacts correctly,
rejecting this perturbation and tracking the set value
Fig. 8. Closed loop response to perturbations appearing on product
profile. The controller reduces the difference between inlet temperature. (A) Temperature. (B) Error. (C) Cold fluid flow
the set value and the outlet temperature and the error rate.
converges to zero. Graph C of Fig. 8 shows the change
in the manipulated variable to compensate the pertur- C the flow rate of cold fluid. The sudden change
bations corresponding to the product inlet temperature appears on the product flow rate at 370 s as indicated
variation. in graph B. In spite of this sudden variation, the
The product flow rate could also be submitted to an controller reacts and adjusts the outlet temperature to
unexpected variation, which can modify the functioning the set value in real time. This reaction is shown in
of the controller. This is illustrated in Fig. 9 where the graph C, which indicates a rapid evolution of the flow
graph A showing the response of the closed loop sys- rate of cold fluid. Indeed, the time response is shorter in
tem, the graph B the flow rate of product and the graph this example and the manipulated variable overshoots.

Fig. 7. Closed loop response to steps of set value with increased value Fig. 9. Closed loop response to perturbations appearing on product
of constraint. (A) Temperature. (B) Error. (C) Cold fluid flow rate. flow rate. (A) Temperature. (B) Error. (C) Flow rate of both fluids.
M.A. Abdelghani-Idrissi et al. / Chemical Engineering and Processing 40 (2001) 449–457 455

5. Conclusion where
m; 1
In this paper, the PFC is applied to a counter current k= (A3)
heat exchanger. This device is used as a cooler of m; 1 + m; 2
product flowing through the inner tube. The outlet k depends on the flow rates through a non-linear (hy-
temperature of the product is adjusted by the flow rate perbolic) function and T depends linearly on tempera-
of the cold fluid circulating through the outer duct tures T1 and T2.
while the temperatures at the inlet are kept constant. Let us now consider the heat transfer from a fluid
The PFC algorithm integrates a First Principles model with flow rate m; to a plenum of volume V and charac-
of the heat exchanger. This model representing the teristics z and Cp through a surface A with a global
response of the system to variations of inlet tempera- heat transfer coefficient U where T o is the fluid output
tures takes into account an average global heat transfer temperature and T i the input fluid temperature. A
coefficient. The convexity property of the heat ex- classical ‘one compartment’ representation [12,13] gives:
changer permits to distinguish the linear influence of
dT o
the state disturbance from the non-linear effect of the zCpV =UA(Tm − T o)+ m; Cp(T i − T o) (A4)
structure perturbations. Taking into account this prop- dt
erty, the control law corresponds to an algebraic solver, dTm
zCpVm = UA(T o − Tm) (A5)
which enables to assess the inlet temperature or the dt
flow rate of cold fluid in order to control the product
where Tm represents the uniform temperature of outer
outlet temperature. In this study, the process is moni-
volume Vm. Eqs. (Eq. (A4)) and (Eq. (A5)) can be
tored by the parameter corresponding to the flow rate
rearranged in the following form:
while the principal inputs corresponding to inlet tem-
peratures are kept constant. Therefore, the controller is dT o
~ + T o = kTm + (1− k)T i (A6)
parametric and from the generic command equation, dt
the parameter is assessed instead of the principal input.
with
This technique provides good results for different de-
sired configurations. Robustness is also investigated zCpV
~= (A7)
while the perturbations affect the product inlet temper- m; Cp + UA
ature and the product flow rate. The model used takes
UA
into account these perturbations. The controller has a k= (A8)
permanent physical adaptation and rejects these pertur- m; Cp + UA
bations by tracking the set-point profile and keeping The steady state is given as follow:
the error at zero.
T o = kTm + (1− k)T i (A9)
We find a similar structure in the modelling of heat
exchangers. Assuming stirred tank heat exchanger, the
Appendix A. Convexity theorem
steady state is given as follow:
Transfer of energy follows the basic principles of T op = kpT ip + (1− kp)T ic (A10)
mass and enthalpy conservation. Let us consider for
with
instance the mixing of two identical fluids with different

 
flow rates and temperatures as depicted in the following UA + m; cCpc
kp = (A11)
figure: UA
UA +m; cCpc 1+
m; pCpp
More complete steady state model [13,17,18] could be
extracted from the well known effectiveness of heat
exchanger:
m; pCpp (T ip − T ip) m; cCpc (T oc − T ic)
m= = (A12)
(m; Cp)min (T ip − T ic) (m; Cp)min (T ip − T ic)
If the mixing is assumed without any mechanical For counter flow heat exchangers, the effectiveness is
effect, the following relations are obtained: expressed as follow:
1− exp[−NTU(1 −C*)]
m; = m; 1 + m; 2 (A1) m= (A13)
1 −C* exp[− NTU(1 −C*)]
T=kT1 + (1− k)T2 with 0Bk B1 (A2) where
456 M.A. Abdelghani-Idrissi et al. / Chemical Engineering and Processing 40 (2001) 449–457

UA (m; Cp)min Appendix B. Predictive Functional Control


NTU = and C* =
(m; Cp)min (m; Cp)max
The PFC is based on the following elements:
From these definitions, we obtain the following “ A dynamic model ym(n) to predict the future system

  n
expressions: behaviour.
UA m; pCpp “ A reference trajectory yr(n), which consists in bring-
1 − exp − 1− ing more slowly the controlled variables up to the
T ip −T op m; pCpp m; cCpc
= (A14) future set-point C(n) within a prediction horizon Hp,
T ip −T ic m; pCpp
1− in order to avoid any overshoot.

  n
m; cCpc
“ A quadratic performance objective with a criterion
UA m; pCpp J(u,e) as a function of the future controller error
1−exp − 1−
T ic − T oc m; pCpp m; cCpc e(n) between reference trajectory and predicted out-
= (A15)
i
T p −T c i
m; cCpc put over a coincidence horizon [H1, Hc].
1− “ A compensation for modelling error.
m; pCpp
Note that the first element of this technique is an
The product outlet temperature extracted from the important point, which can modify the controller per-
steady state (Eq. (A14)) and (Eq. (A15)) should be formance when the model is not very representative of
written in the same way as expression (Eq. (A10)): the process. A great effort has to be undertaken in this
T op =k%pT ip +(1−k%p)T ic (A16) area before implementing the PFC. But a simple model
is desirable according to the system controllability
with the following k%p:
    n
criteria.
The reference trajectory can also be considered as the
m; pCpp UA m; pCpp
1− exp − 1− desired response of the closed loop. Particularly, PFC

  n
m; cCpc m; pCpp m; cCpc
k%p = (A17) considers a reference trajectory of first order. For a
m; Cp UA m; Cp given set-point profile, the objective criterion minimisa-
1− p p exp − 1− p p
m; cCpc m; pCpp m; cCpc tion leads to optimal values of the manipulated variable
PFC algorithm computes a ‘functional’ manipulated in the future for the coincidence horizon. That gives the
variable Teq, corresponding to the second member of nearing predicted target variables by the dynamic
equation (A6), defined as an equivalent temperature: model to the reference trajectory values. In PFC al-
gorithm, the manipulated variable is constituted of a
dT o superposition of a set of basis functions. The basis
~ + T o = Teq =kTm +(1 −k)T i (A18)
dt functions are generally taken polynomial [8]:

Taking the case of stirred tank heat exchanger, the u(n+ i )= % [ikuBk](i ) (B1)
‘physical’ manipulated variable will depend on the na- k
ture of the actuators. If the control of product outlet
uBK(i )= i k (B2)
temperature is achieved by the inlet temperature of cold
fluid, the Manipulated Variable is defined as: Note that there is no limit to select them; they are
taken according to the form of the desired controlled
Teq(n)−kpT ip
T ic(n)= (A19) variable.
1− kp The increment between the actual model output and
When the control is performed with the flow rate of the reference trajectory at the point of coincidence in
cold fluid, the manipulated variable is obtainable from the future is:
kp: DH= yr(n+H)− ypr(n) (B3)
i
Teq(n)− T with
kp(n) =
c
(A20)
T ip − T ic
C(n+ i )− yr(n+i )= u i[C(n)−ypr(n)] (B4)
UA
(1−kp) Taking into account the coincidence of the model
 
Cpc
m; c(n)= (A21) output and the reference trajectory, DH can be written
UA
kp 1+ −1 in the following form:
m; pCpp
DH=[C(n)− ypr(n)](1− u H) (B5)
As shown in expressions (Eq. (A19)) and (Eq. (A21)),
the stirred tank heat exchanger model presents the Considering the forced response of the model and its
advantage to assess easily the physical manipulated homogeneous solution, the increment can be trans-
variable. formed into the following expression:
M.A. Abdelghani-Idrissi et al. / Chemical Engineering and Processing 40 (2001) 449–457 457

DH = ypr,x(n+ H) −ypr(n) + % [ikuBk ](i ) (B6) pr process


k
r reference
Note that to each input function uBk corresponds an sat saturation
output function yBk. If the set-point profile is character-
ised by different steps, then only the first function uB1 Superscript
has to be considered. i inlet
In this case, the law command is given by the follow- o outlet
ing expression: p product
final steady state
(C(n)−ypr(n)) (1− u H) =i1yB1(H) +h Hypr(n) − ym(n)
(B7)

References
Appendix C. Nomenclature
[1] K. Preuß, M.-V. Le Lann, J. Richalet, M. Cabassud, G. Casa-
matta, Thermal control of chemical batch reactors with predictive
functional control, Journal A 39 (4) (1998) 13 – 20.
a thickness (m) [2] H.A.B. Te Braake, E.J.L. Van Can, J.M.A. Scherpen, H.B.
A heat transfer area (m2) Verbruggen, Control of nonlinear chemical processes using neural
models and feedback linearization, Comput. Chem. Engng. 22
Bi Biot number (1998) 1113 – 1127.
C set-point [3] J. Richalet, A. Rault, J.L. Testud, J. Papon, Model heuristic
Cp specific heat (J · kg−1 · K−1) control-application to industrial processes, Automatica 14 (1978)
D diameter (m) 413 – 428.
[4] J. Richalet, Industrial applications of model based predictive
e error (K)
control, Automatica 29 (5) (1993) 1251 – 1274.
h convective heat transfer coefficient (W K−1 [5] J. Richalet, Pratique de la commande predictive, Hermes, Paris,
m−2) 1993.
H horizon [6] A. Zheng, Stability of model predictive control with time-varying
k thermal conductivity (W K−1 m−1) weights, Comput. Chem. Engng. 21 (12) (1997) 1389 – 1393.
[7] T.T.J. Van den boom, R.A.J. De Vries, S.B. Bouwmeester, H.B.
K static gain
Verbruggen, Constrained predictive control with guaranteed sta-
L length bility and convex optimisation, Journal A. 35 (3) (1994) 27 – 34.
m; flow rate (kg s−1) [8] P. Pierson, J. Padet, Etude théorique et expérimentale des
NTU number of transfer échangeurs en régime thermique instationnaire. Simulation d’une
n unit time sample phase de relaxation, Int. J. Heat Mass Transfer 31 (8) (1988)
1577 – 1586.
T temperature (K)
[9] M. Hadidi, M. Guellal, M. Lachi, J. Padet, Loi de réponse d’un
t time échangeur thermique soumis à des échelons de températures aux
TRBF closed loop (s) response time entrées, Int. Comm. Heat Mass Transfer 22 (1) (1995) 145 –154.
U global average coefficient heat transfer (W [10] M. Lachi, N. El Wakil, J. Padet, The time constant of double pipe
K−1 m−2) and one pass shell-and-tube heat exchangers in the case of varying
flow rates, Int. J. Heat Mass Transfer. 40 (9) (1997) 2067 –2079.
V occupied volume (m3)
[11] K.W. Mathisen, M. Morari, S. Skogestad, Dynamic models for
y target variable heat exchangers and heat exchanger networks, Comput. Chem.
Engng. 18 (1994) 459 – 463.
Greek symbols [12] J.F. Sacadura, Initiation aux Transferts Thermiques, techniques
h parameter et documentation, 1982.
i parameter [13] S. Kakaç, H. Liu, Heat Exchangers: Selection, Rating, and
u parameter Thermal design, CRC Press LLC, 1998.
k gain [14] Y.A. Gögüs, O8 .E. Ataer, Transient behaviour of heat exchangers,
in: J. Padet, F. Arinç (Eds.), Transient convective heat transfer,
v parameter Begel House, New York, 1997, pp. 289 – 303.
z density (kg m−3) [15] O8 .E. Ataer, A. Ileri, Y.A. Gögüs, Transient behaviour of finned-
~ time constant (s) tube cross-flow heat exchangers, Int. J. Refrig. 18 (3) (1995)
DT time sampling (s) 153 – 160.
[16] O8 .E. Ataer, Y.A. Gögüs, An approximate approach for transient
Subscript behaviour of finned-tube-crossflow heat exchanger, in: J. Padet,
B basic functions F. Arinç (Eds.), Transient convective heat transfer, Begel House,
New York, 1997, pp. 533 – 542.
c cold fluid
[17] W. Roetzel, Y. Xuan, Dynamic Behaviour of Heat Exchangers,
m model Wit Press, Boston, 1999.
x forced response [18] W.M. Rohsenow, J.P. Hartnett, E.N. Ganic, Handbook of heat
p product transfer applications, McGraw Hill, New York, 1985.

You might also like