Sie sind auf Seite 1von 13

Nahrain University, College of Engineering Journal (NUCEJ) Vol.13 No.2, 2010 pp.

129-140

Dynamics and Control of Double-Pipe Heat Exchanger

Dr.Khalid.M.Mousa
Chemical engineering Department
Nahrain University

Abstract
2
Experimental investigation was conducted to
study the dynamics of double-pipe heat exchanger
using step change in flow rate of hot water and
[Burn ,1981] used lead / lag approximation in
observe its effect on the temperature of cold water
modeling a double pipe heat exchanger system.
and it was found that the transfer function can be
The analysis involved extraction of successive
represented by first order system plus dead time
real poles and zeros of the system transfer
(FOPDT).
functions for temperature or flow forcing. [
Shinsky,1967] formulated a model describing
dimensionless temperature as function of flow
rates.[Stephanopouls,1967] shown that the heat
exchanger can be represented by a partial
MATLAB 7.4 simulation and programming was
differential equation i. e distributed parameter
used to show the response of close-loop system
system.[Timothy,2004] studied a double pipe
using P, PI, PID and fuzzy logic control.
helical heat exchanger numerically and
The Smith predictor was used to remove the
experimentally for both heat transfer and
effect of transport delay on the process and it was
hydrodynamic characteristics. Simulation were
found that this method greatly improve the
performed using various flow rates, his study
response of the system. The stability of the overall
involved the determination of uniformity of the
closed loop system was proved by using Bod plot
residence time and temperature distribution as
diagram.
well as the heating-cooling uniformity.[Sudhir et
Introduction al.,2006] used neural network for controlling
temperature of double pipe heat exchanger .The
Simple double pipe heat exchanger consist of results indicated that the feedback control system
concentric pipes. It is commonly used in is suitable when the process detail are known
applications involving relatively low flow rates completely. whereas neural network control is
and high temperature or pressure, for which they useful for the process with nonlinear and less
are well suited. other advantage include low knowledge about the process.[ Anton et al.,2009]
installation cost, ease of maintenance, and indicated that some parameters ,for example fluid
flexibility [Robert ,2007 ] density and heat capacity are not constant in
[Gerard ,1974] formulated a model for a counter operating temperature rang and this introduced
flow heat exchanger, which has taken into anon-negligible nonlinearities in heat exchangers
account variation of the heat transfer coefficient modeling.
with respect to fluid flow rates and temperature. The aim of this work is to find the transfer
The approach used in this study to solve four function of the double pipe heat exchanger
nonlinear partial differential equations by experimentally and then using a MATLAB
linearization method . Experimental investigation simulink to study the closed loop system via
using various test such as step function and sine applying both conventional and advanced control
wave was conducted. The results showed that the methods.
transfer function with variation of temperature in
both stream could be approximated by the
following forms:
Process Identification
Experimental identification of process
dynamics has been an active area of search for
many years by workers in several area of
1 engineering.
A time domain fitting of step test data is the most
direct way to obtaining an empirical linear
dynamics model of process in order to find the
parameter (dead time, time constant and process

NUCEJ Vol.13 No.2 Mousa 129


gain) that fit the experimentally obtained step phase angle is 180. At this frequency look at the
response data. The process is being identified is magnitude plot to determine the gain margin
usually open loop, but experimental testing of which is the gain required to raise the magnitude
close loop system is also possible. A step curve to 0 db. A positive gain margin mean that
disturbance in hot water flow rate was introduced the system is stable, and negative gain margin
and record the temperature of could water . mean that the system is unstable. [Luyben, 1997]
probably 80% of open loop process can be The phase margin is found by using the
modeled by gain, a dead time, and one lag. magnitude curve to find the frequency when the
gain is 0 db. On the phase curve at that frequency,
the phase margin is the difference between the
phase value and 180.
3
Mathematical model of a heat
exchanger
The steady state process gain (k) is equal to the
ratio of the final steady state change in the output For double pipe heat exchanger, a liquid flows
over the size of input, while the time constant through the inner tube and it is being heated by
estimated from the response by using several another liquid that flows counter currently around
methods such as time takes (63.2) percent of the the tube. The temperature of the inner liquid not
final steady state change and initial slop methods only changes with time but also changes along the
[Luyben,1997]. Bode plots is used to examine the axial direction z, the energy balance equation is:-
stability of the system and evaluate the gain and
phase margin. The gain margin is found by using
the phase plot to find the frequency, where the

Dividing by Z t and taking the limit of Z and


t go to zero yields
+ 7

= DQ 5
It is very robust and can be used when we do not
know the plant model very accurately. The design
Where Q=U( -T), substitute in Eq. (5) , yields: process involves finding values of Kc, i and D
such that the design specifications are satisfied.
The PID gain search process is sometimes
6 referred to as tuning. For certain plant type, The
= DU( -T) Ziegler-Nicolas method provides a structure
tuning procedure, however not all plants are
amenable to that type of tuning. One method
Where relies on interactive computer simulation and
Q=amount of heat transferred from hot fluid to analysis. Initial nominal gain values are choose ,
other per unit of time and unit of heat transfer and then tune the controller parameters by
area. simulating the response and verifying that the
A=cross-sectional area of the inner tube. performance is acceptable. If the performance is
V=average velocity of the inner liquid . not acceptable, update the gains and iterate the
D=external diameter of the inner tube process. Naturally MATLAB is very helpful in
U=overall heat transfer coefficient. this process. [Bishop,1977]
Smith predictor
PID controller A Smith predictor was used to design a
The PID controller is widely used in industry. controller to remove the effect of transport delay .
The transfer function is expressed as follows:- The first step to design the controller Gc(s) is to
design a suitable controller cs (s) when the
system has no dead time. J.M. Smith designed an

NUCEJ Vol.13, No.2 Control of Double-Pipe 130


innovative controller scheme, to control the plant
with dead time. The Smith predictor consist of
Gc(s), dead time e-DS, and the plant transfer 9
function G(s).
Figure (1) shows the Smith predictor as a minor
loop in the block diagram. The transfer function
between v(s) and E(s) is

10
8

Hence the block diagram of figure (1) can be


modified to that of figure (2). The close-loop
Then the close-loop transfer function C(s)/R(s)
response of the system with dead time (e-DS) is the
can be given by
same as the response of the system without dead
time (e-DS), [Ogata]

R + + V G(s) e-DS C
c
- -

-e-DS G(s)

Figure1. Minor loop of Smith predictor [Ogata,2002]

+
c G(s) e-DS
R - C

Figure 2. modified block diagram of Smith predictor[ Ogata,2002 ]

3.10 Fuzzy Logic Control (FLC) Fuzzy mathematics is attractive not only
because it is based on the very intuitive idea of
Fuzzy logic is a super set of traditional logic, Fuzzy sets, but because it is capable of generating
according to [Zadeh,1989], who invented this many structures that provide today's scientists and
concept in 1965. He said it is the logic of engineers with new insights into interesting,
approximate reasoning and it is unlike other significant and often-debated problems in both
branches of artificial intelligence (AI) that use science and engineering. [ Zadeh L.A , 1989]
conventional logic. Since then, the theory of The theory of Fuzzy sets has one of its aims,
mathematics has gained more recognition from the development of a methodology for the
many researches in a wide range of scientific formulation and solution of problems that are too
fields. complex or too ill-defined to be analyzed by
conventional techniques.
NUCEJ Vol.13 No.2 Mousa 131
Hence the theory of Fuzzy sets is likely to be configuration single pass. The specification of
recognized as a natural development in the double pipe are 37 mm outer diameter, 12 mm
evaluation of scientific thinking inner diameter and 1 m length was used.
[McCusker,1990]. Hot and cold water were employed as circulating
The Fuzzy logic can be considered as a tool of fluid. steady state values of cold and hot water
promise for control. It can AL's sector, which flow rates are 300 and 600 l/min respectively.
ensures to enhance tomorrow's industrial control After reach the desired temperature of hot water
system [ Zadeh ,1989]. tank (341K), and steady-state of cold water
stream temperature (103K),a step change in hot
Experimental work:- water flow rate was introduced and record the
outlet temperature of exit cold water each two
The process involved a pilot plant of double pipe second until a steady stat was reached and noting
heat exchanger consisted mainly of two the apparent dead time.
concentric pipes assembled in counter-current

Temp.
Glass wool (insulation) recorder

heat exchanger Hot


Temp. flow meter
record water
er tank

valve
flow meter

pum Heating
coil

Cold
water
tank

Valv

pum

Figure 3. Schematic diagram of experimental set-up

shape of the response (using deviation variable), it


Results and Discussion is clearly that the system can be represented by
first order system plus dead time (FOPDT)
- System identification
The actual response of the outlet could water
temperature using step change in hot water flow
rate is shown in figure 4. A process reaction curve
was used to find the process parameters. From the
NUCEJ Vol.13, No.2 Control of Double-Pipe 132
A theoretical value of the sped of response
( = m Cp /h A ) is 25s. The deviation
Gp(s)= 11 between the theoretical and experimental
value is due to a neglected the variation of
heat transfer coefficient with respect to fluid
flow rates and temperature. The transport
Where: delay involved the heat exchanger and all the
K is a process gain = 0.6 k/(kg/s) accessories between the valve of hot water
is the time constant =20 s and temperature recorder.
D is the transport delay = 4 s

1.2

0.8
Temp. ( K )

0.6

0.4

0.2

0
0 10 20 30 40 50 60
Time (sec )

Figure 4. Temperature response for step change in hot water flow rate

If the system is first order, then the output (Ziegler-Nicolas) ,and attest for closed loop
response curve has anon-zero slope at the initial control system and fine-tuning was conducted
time. The response of the system with a relative using a MATLAB simulik . Figures (5,6,7,8 and
order greater than one has zero slope at initial 9) show the block diagram of P, PI , PID, fuzzy
time. [wayne Bequette ,1998] logic control and Smith predictor, while figures
(10,11,12,13 and 14) show the response of these
Closed loop block diagram respectively. The results indicate
For simplicity we assume that the transfer that the smith predictor is greatly improved the
function of close-loop system involved only the response or controlled variable. A PI and PID
transfer function of process and controller due to removed the offset but the settling time response
small time constant of valve and transmitter is large compared with that in using smith
compared with process and controller. predictor which only ten second. From figures (11
A servo problem was considered using and 12) one can see that the time required for
MATLAB simulink via unit step change in set response curve to reach and stay within rang
point of controlled variable which is the outlet about the final value of specified by absolute
percentage of the final value usually 2% or 5% ( a
temperature of cold water stream .The initial settling time) equal 50 second when using PI
tuning constants(proportional gain (Kc), time controller and 40 second when using PID
integral ( I) and time derivative ( D ))for P, PI and controller while only 5 second in using Smith
PID controller were found using empirical model predictor .The controlled variable in P ,PI ,and
NUCEJ Vol.13 No.2 Mousa 133
PID control methods overshoots its set point approximation via Bode diagram. Figure (15)
while dos not in using Smith predictor and fuzzy indicate that the system is stable because gain and
logic control phase margin are positive.
A stability of open loop and closed loop using
smith predictor was stated using first order pade

.6
7.5
20 s+1
Step Gain Transfer Fcn Transport Scope
Delay

Figure 5. Block diagram of closed-loop system using P controller

12 .5s+1 .6
-K -
12 .5s 20 s+1
Step Gain Transfer Fcn 1 Transfer Fcn Transport Scope
Delay

Figure 6. Block diagram of closed-loop system using PI controller

14 s2 +9.3s+1 .6
-K - s2 +7.5s 20 s+1
Gain Transfer Fcn 1 Transfer Fcn Transport
Step Scope
Delay

Figure 7. Block diagram of closed-loop system using a PID controller

NUCEJ Vol.13, No.2 Control of Double-Pipe 134


.6
95 20 s+1
Transfer Fcn Transport Scope
Step Gain
Delay

Figure8. Block diagram of closed-loop system using smith predictor

Figure9. Block diagram of closed-loop system using fuzzy logic control

1.4

1.2

0.8
Temp. (K)

0.6

0.4

0.2

0
0 10 20 30 40 50 60 70 80 90 100
Time (sec)

Figure. 10 response of closed loop using P controller

NUCEJ Vol.13 No.2 Mousa 135


1.5

1
temp. (K)

0.5

0
0 10 20 30 40 50 60 70 80 90 100
time (sec)

Figure. 11 response of closed loop using PI controller

1.8

1.6

1.4

1.2
Temp. (K)

0.8

0.6

0.4

0.2

0
0 10 20 30 40 50 60 70 80 90 100
time (sec)

Figure. 12 response of closed loop using PID controller

NUCEJ Vol.13, No.2 Control of Double-Pipe 136


1.5

1
Temp. (K)

0.5

0
0 10 20 30 40 50 60 70 80 90 100
Time (sec)

Figure 13. Response of closed loop using smith predictor


Temp.(K)

Figure14. Fuzzy logic control

NUCEJ Vol.13 No.2 Mousa 137


Bode Diagram

10

-5
M agnitude (dB)

-10

-15

-20

-25

-30

-35

-40
360

270
Phas e (deg)

180

90
-2 -1 0 1 2
10 10 Frequency10(rad/sec) 10 10

Figure 15. Bode diagram for open loop system without time delay

Bode Diagram

-10
Magnitude (dB)

-20

-30

-40
360
Phase (deg)

270

180

90
-2 -1 0 1 2
10 10 Frequency10(rad/sec) 10 10

Figure 16. Bode diagram for close loop system using smith predictor

NUCEJ Vol.13, No.2 Control of Double-Pipe 138


Conclusion : 3. Coughanowr and Koppel Process
systems and control , Mc Grow-hill,
There are a few points which must be New York,p 94, 1981.
emphasized: 4. Georg Stephanopoulos, chemical process
control An introduction to theory and
1) The system Can be represented by FOPDT practice, prentic-hall, p 69-70, 1984.
with process gain (Kp=0.6) and time constant of
20s 5. Gerard, G.J. dynamic system,
2) Smith predictor greatly improved the response measurement and control ASME, 9
of closed-loop, a settling time only 5 second in (1974).
using Smith predictor
3) The open and closed-loop systems are stable 6. Katrsuhiko Ogata, modern control
due to positive gain margin engineering, prentic-hall, P 411-413 ,
4) Fuzzy logic controller is recommended due to 2002.
low cost ,easy to implement and it does not
required an accurate mathematical model. 7. Larsen P.M., Industrial application of
Abbreviatin Fuzzy Logic Control, academic press,
inc., may, 1979.
p proportional control 8. Mamdine E.H, Application of Fuzzy
PI proportional integral control Logic to Approximate Reasoning using
PID proportional integral derivative Languistic Synthesis, IEEE Trans. On
Notation Computers, December 1977.
9. McCusker T., Neural network and Fuzzy
A Surface area of heat transfer, (m2) logic, Tools of Promise Control, Control
Eng. Magazine, 84-85, May 1990.
C Control variable (temperature), (k) 10. Michael L. Lyben and William L. lyben,
Essentials of process control, Mc Grow-
Cp Heat capacity of water, kjkg-1.k-1 hill, Inc New York, p 548-549,1997.
11. Norman S. N control system
D Transport delay, (s) engineering, johan wiley and sons. Inc
2000.
G Transfer function, (k.s.kg-1)
12. Robert H. Bishop modern control
Gc Transfer function of controller system analysis and design using
MATLAB and simulink. 1977 .
h Heat transfer coefficient (KJ.m-2.s-1) 13. Robert w .Serth ,Process heat transfer
principles and applications ,Elsevier
Kp Process gain, (k.s.kg-1) science and technology book,2007
Kc Controller gain 14. Shinsky process control system, Mc
Grow-hill, New York,p236-238, 1967.
R Input variable 15. Sudhir ,etl , Automation of heat
exchanger using neural network,
m mass of water in heat exchanger, (kg) PWASET ,15, 2006
Time constant, (s) 16. Tiomthy J. Rennie Ph.D a thesis,
University McGill 2004.
i Integral time of controller
17. Wayne Bequette Process dynamics
d derivative time of controller modeling, Analysis ,and Simulation,
prentic-hall RTR ,p 219, 1998.
Referances
18. Zadeh L.A., Fuzzy relation Equations
th and Applications to Knowledge
1. Anton Sodja1 et al., proceeding 7
Modelica conference , Como, Engineering, Kluwer Academic
Italy,Sep.20-22,2009 Publishers, Holland, 1989
2. Burne, A. Journal of dynamics system
ASME, 3 (1981).

NUCEJ Vol.13 No.2 Mousa 139


MATLAB

Smith predictor

NUCEJ Vol.13, No.2 Control of Double-Pipe 140


This document was created with Win2PDF available at http://www.daneprairie.com.
The unregistered version of Win2PDF is for evaluation or non-commercial use only.