You are on page 1of 9

Design optimization of shell-and-tube heat exchanger using particle swarm

optimization technique
V.K. Patel, R.V. Rao
*
Sardar Vallabhbhai National Institute of Technology, Surat-395 007, Gujarat, India
a r t i c l e i n f o
Article history:
Received 19 February 2009
Accepted 2 March 2010
Available online 6 March 2010
Keywords:
Heat exchanger design
Shell-and-tube heat exchanger (STHE)
Particle swarm optimization (PSO)
Genetic algorithm (GA)
a b s t r a c t
Shell-and-tube heat exchangers (STHEs) are the most common type of heat exchangers that nd
widespread use in numerous industrial applications. Cost minimization of these heat exchangers is a key
objective for both designer and users. Heat exchanger design involves complex processes, including
selection of geometrical parameters and operating parameters. The traditional design approach for shell-
and-tube heat exchangers involves rating a large number of different exchanger geometries to identify
those that satisfy a given heat duty and a set of geometric and operational constraints. However, this
approach is time-consuming and does not assure an optimal solution. Hence the present study explores
the use of a non-traditional optimization technique; called particle swarm optimization (PSO), for design
optimization of shell-and-tube heat exchangers from economic view point. Minimization of total annual
cost is considered as an objective function. Three design variables such as shell internal diameter, outer
tube diameter and bafe spacing are considered for optimization. Two tube layouts viz. triangle and
square are also considered for optimization. Four different case studies are presented to demonstrate the
effectiveness and accuracy of the proposed algorithm. The results of optimization using PSO technique
are compared with those obtained by using genetic algorithm (GA).
2010 Elsevier Ltd. All rights reserved.
1. Introduction
Heat exchangers are used in industrial process to recover heat
between two process uids. Shell-and-tube heat exchangers
(STHEs) are the most widely used heat exchangers in process
industries because of their relatively simple manufacturing and
their adaptability to different operating conditions. The design of
STHEs, including thermodynamic and uid dynamic design, cost
estimation and optimization, represents a complex process con-
taining an integrated whole of design rules and empirical knowl-
edge of various elds [1,2].
The design of STHEs involves a large number of geometric and
operating variables as a part of the search for an exchanger
geometry that meets the heat duty requirement and a given set of
design constrains. Usually a reference geometric conguration of
the equipment is chosen at rst and an allowable pressure drop
value is xed. Then, the values of the design variables are dened
based on the design specications and the assumption of several
mechanical and thermodynamic parameters in order to have
a satisfactory heat transfer coefcient leading to a suitable
utilization of the heat exchange surface. The designer's choices are
then veried based on iterative procedures involving many trials
until a reasonable design is obtained which meets design speci-
cations with a satisfying compromise between pressure drops and
thermal exchange performances [3e6].
There are many previous studies on the optimization of heat
exchanger. Several investigators had used different optimization
techniques considering different objective functions to optimize
heat exchanger design. Chauduri et al. [7] used simulated annealing
approach for the optimal design of heat exchanger and developed
a command procedure to link HTRI (Heat Transfer Research Inc.)
design program to the annealing algorithm. The authors had
analyzed the problemconsidering two different objective functions
namely, total heat transfer area and a linearized purchased cost
index. Selbas et al. [2] used genetic algorithm (GA) for optimal
design of STHEs, in which pressure drop was applied as a constraint
for achieving optimal design parameters. The authors had consid-
ered minimization of total heat exchanger cost as an objective
function. A case study was also considered to demonstrate the
application of algorithm.
Ozcelik [8] considered mixed integer nonlinear programming
problem of STHE taking into account the sizing and exergy cost of
the STHE. The authors had considered minimization of sum of
annual capital cost and exergetic cost of STHEs as an objective
* Corresponding author. Tel.: 91 261 2201661; fax: 91 261 2201571.
E-mail address: ravipudirao@gmail.com (R.V. Rao).
Contents lists available at ScienceDirect
Applied Thermal Engineering
j ournal homepage: www. el sevi er. com/ l ocat e/ apt hermeng
1359-4311/$ e see front matter 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.applthermaleng.2010.03.001
Applied Thermal Engineering 30 (2010) 1417e1425
function and applied a genetic algorithm for optimization. Caputo
et al. [9] carried out heat exchanger design based on economic
optimization using GA. The authors had considered minimization
of total annual cost as an objective function to nd out the optimal
heat exchanger architecture. Fesanghary et al. [1] used global
sensitivity analysis to identify the most inuential geometrical
parameters that affect total cost of STHE in order to reduce the size
of optimization problem and carried out the optimization of
inuential parameter of STHE from economic point of view by
applying harmony search algorithm. The authors had considered
minimization of capital cost and operating cost as an objective
function. Several other investigators also used strategies based on
GA and traditional mathematical optimization algorithms [10e26]
for various objectives like minimumentropy generation [18,19] and
minimum cost of STHEs [12e17,22e26] to optimize heat exchanger
design. Some of these studies spotlight on a single geometric
parameter like optimal bafe spacing [15,16,21] and some others
tried to optimize a variety of geometrical and operational param-
eter of the STHEs. However, there is a need to investigate the
potential of application of non-traditional optimization techniques.
Particle swarm optimization (PSO) is one such technique and the
same is investigated in the present work for its effectiveness.
The main objectives of this study are (i) to optimize the inu-
ential parameters of STHEs from economic point of view and (ii)
to demonstrate the effectiveness of particle swarm optimization
(PSO) technique in the design optimization of STHEs from
economic point of view. Ability of the PSO technique is demon-
strated using different case studies. Parametric analysis is also
carried out to evaluate and compare the sensitivity of the relevant
cost parameters and change in objective function for both the
approaches.
2. Particle swarm optimization (PSO) technique
Particle swarm optimization (PSO) is an evolutionary compu-
tation technique developed by Kennedy and Eberhart [27]. It
exhibits common evolutionary computation attributes including
initialization with a population of random solutions and searching
for optima by updating generations. Potential solutions, called
birds or particles, are then own through the problem space by
following the current optimum particles. The particle swarm
concept was originated as a simulation of a simplied social system.
The original intent was to graphically simulate the graceful but
unpredictable choreography of a bird ock. Each particle keeps
track of its coordinates in the problem space, which are associated
with the best solution (tness) it has achieved so far. This value is
called pBest. Another best value that is tracked by the global
version of the particle swarm optimization is the overall best value
and its location obtained so far by any particle in the population.
This location is called gBest.
The particle swarm optimization concept consists of, at each
step, changing the velocity (i.e. accelerating) of each particle
Nomenclature
a
1
numerical constant (V)
a
2
numerical constant (V/m
2
)
a
3
numerical constant
a
s
shell side pass area (m
2
)
A heat exchanger surface area (m
2
)
B bafes spacing (m)
C numerical constant
C
e
energy cost (V/kW h)
C
i
capital investment (V)
cl clearance (m)
C
o
annual operating cost (V/yr)
C
od
total discounted operating cost (V)
C
p
specic heat (J/kg K)
C
tot
total annual cost (V)
d
e
equivalent shell diameter (m)
d
i
tube inside diameter (m)
d
o
tube outside diameter (m)
D
s
shell inside diameter (m)
F temperature difference correction factor
f
s
shell side friction coefcient
f
t
tube side friction coefcient
H annual operating time (h/yr)
h
s
shell side convective coefcient (W/m
2
K)
h
t
tube side convective coefcient (W/m
2
K)
i annual discount rate (%)
k thermal conductivity (W/m K)
L tubes length (m)
LMTD logarithmic mean temperature difference (K)
m
s
shell side mass ow rate (kg/s)
m
t
tube side mass ow rate (kg/s)
n number of tube passes
n
1
numerical constant
ny equipment life (yr)
N
t
number of tubes
P pumping power (W)
Pr
s
shell side Prandtl number
Pr
t
tube side Prandtl number
Q heat duty (W)
Re
s
shell side Reynolds number
Re
t
tube side Reynolds number
R
fs
shell side fouling resistance (m
2
K/W)
R
ft
tube side fouling resistance (m
2
K/W)
S
t
tube pitch (m)
T
ci
cold uid inlet temperature (K)
T
co
cold uid outlet temperature (K)
T
hi
hot uid inlet temperature (K)
T
ho
hot uid outlet temperature (K)
U overall heat transfer coefcient (W/m
2
K)
v
s
shell side uid velocity (m/s)
v
t
tube side uid velocity (m/s)
Dh heat transfer difference (W/m
2
K)
DP pressure drop (Pa)
DP
tube elbow
tube elbow pressure drop
DP
tube length
tube length pressure drop (Pa)
Greek letters
m dynamic viscosity (Pa s)
r density (kg/m
3
)
h overall pumping efciency
Subscripts
c cold stream
e equivalent
h hot stream
i inlet
o outlet
s shell side
t tube side
wt wall
V.K. Patel, R.V. Rao / Applied Thermal Engineering 30 (2010) 1417e1425 1418
toward its pBest and gBest locations (global version of PSO).
Acceleration is weighted by a random term with separate random
numbers being generated for acceleration toward pBest and
gBest locations. The updates of the particles are accomplished as
per the following equations.
V
i1
wV
i
c
1
r
1
pBest
i
X
i
c
2
r
2
gBest
i
X
i
(1)
X
i1
X
i
V
i1
(2)
Equation (1) calculates a new velocity (V
i1
) for each particle
(potential solution) based on its previous velocity, the best location
it has achieved (pBest) so far, and the global best location (gBest),
the population has achieved. Equation (2) updates individual par-
ticle's position (X
i
) in solution hyperspace. The two random
numbers r
1
and r
2
in equation (1) are independently generated in
the range [0, 1].
The acceleration constants c
l
and c
2
in equation (1) represent
the weighting of the stochastic acceleration terms that pull each
particle toward pBest and gBest positions. c
1
represents the
condence the particle has in itself (cognitive parameter) and c
2

represents the condence the particle has in swarm (social


parameter). Thus, adjustment of these constants changes the
amount of tension in the system. Low values of the constants allow
particles to roam far from target regions before being tugged back,
while high values result in abrupt movement toward, or past
through target regions [28].
Particle's velocities on each dimension are conned to
a maximum velocity parameter V
max
, specied by the user. If the
sum of accelerations would cause the velocity on that dimension to
exceed V
max
, then the velocity on that dimension is limited to V
max
.
Also if X
i1
is greater than maximum value for the corresponding
design variable then it is set to the maximum value for that design
variable and if X
i1
is less than the corresponding minimum design
variable then it is set to the minimumvalue for that design variable.
The inertia weight w plays an important role in the PSO
convergence behavior since it is employed to control the explora-
tion abilities of the swarm. The large inertia weights allow wide
velocity updates allowing to globally explore the design space
while small inertia weights concentrate the velocity updates to
nearby regions of the design space. The optimum use of the inertia
weight w provides improved performance in a number of appli-
cations. The effect of w, c
1
, and c
2
on convergence for standard
numerical benchmark functions was provided by Bergh and
Engelbrecht [29].
The next section presents the mathematical models for heat
transfer and pressure drop in a heat exchanger and the formulation
of objective function for optimization.
3. Mathematical models
3.1. Heat transfer
According to ow regime, the tube side heat transfer coefcient
(h
t
) is computed from following correlation,
h
t

k
t
d
i
_
3:657
0:0677Re
t
Pr
t
d
i
=L
1:33
1=3
1 0:1Pr
t
Re
t
d
i
=L
0:3
_
(3)
(if Re
t
< 2300 [9])
h
t

k
t
d
i
_
f
t
=8Re
t
1000Pr
t
1 12:7f
t
=8
1=2
_
Pr
2=3
t
1
_
_
1
d
i
L
_
0:67
_
(4)
(if 2300 < Re
t
< 10,000 [9])
h
t
0:027
k
t
d
0
Re
0:8
t
Pr
1=3
t
_
m
t
m
wt
_
0:14
(5)
(for Re
t
> 10,000 [9])
where, f
t
is the Darcy friction factor [5] given as,
f
t

_
1:82 log 10
Ret
1:64
_
2
(6)
Re
t
is the tube side Reynolds number and given by,
Re
t

r
t
v
t
d
i
m
t
(7)
Flow velocity for tube side is found by,
v
t

m
t
p=4d
2
t
r
t
_
n
N
t
_
(8)
N
t
is the number of tubes and n is the number of tube passes which
canbe foundapproximatelyfromthe followingequation[4,6,30,31],
N
t
C
_
D
s
d
0
_
n1
(9)
C and n
1
are coefcients that are taking values according to ow
arrangement and number of passes. These coefcients are shown in
Table 1 for different ow arrangements.
Pr
t
is the tube side Prandtl number and given by,
Pr
t

m
t
C
pt
k
t
(10)
also, d
i
0.8d
o
Kern's formulation for segmental bafe shell-and-tube
exchanger is used for computing shell side heat transfer coefcient
h
s
[3],
h
s
0:36
k
t
d
e
Re
0:55
s
Pr
1=3
s
_
m
s
m
wts
_
0:14
(11)
where, d
e
is the shell hydraulic diameter and computed as [3,30],
d
e

4
_
S
2
t

_
pd
2
o
=4
__
pd
o
(12)
(for square pitch)
d
e

4
_
0:43S
2
t

_
0:5pd
2
o
=4
__
0:5pd
o
(13)
(for triangular pitch)
Cross section area normal to owdirection is determined by [2],
A
s
D
s
B
_
1
d
o
S
t
_
(14)
Flow velocity for the shell side can obtain from [2],
Table 1
Values of C and n
1
coefcient.
No. of passes Triangle tube pitch
S
t
1.25d
o
Square tube pitch
S
t
1.25d
o
C n
1
C n
1
1 0.319 2.142 0.215 2.207
2 0.249 2.207 0.156 2.291
4 0.175 2.285 0.158 2.263
6 0.0743 2.499 0.0402 2.617
8 0.0365 2.675 0.0331 2.643
V.K. Patel, R.V. Rao / Applied Thermal Engineering 30 (2010) 1417e1425 1419
v
s

m
s
r
s
A
s
(15)
Reynolds number for shell side follows,
Re
s

m
s
d
e
A
s
m
s
(16)
Prandtl number for shell side follows,
Pr
s

m
s
C
ps
k
s
(17)
The overall heat transfer coefcient (U) depends on both the
tube side and shell side heat transfer coefcients and fouling
resistances are given by [2],
U
1
1=h
s
R
fs
d
o
=d
i

_
R
ft
1=h
t

_ (18)
Considering the cross ow between adjacent bafe, the loga-
rithmic mean temperature difference (LMTD) is determined by,
LMTD
T
hi
T
co
T
ho
T
ci

lnT
hi
T
co
=T
ho
T
ci

(19)
The correction factor F for the ow conguration involved is
found as a function of dimensionless temperature ratio for most
ow conguration of interest [32,33].
F

R
2
1
R1

ln1P=1PR
ln
__
2PR1

R
2
1
_
___
2PR1

R
2
1
_
__
(20)
where R is the correction coefcient given by,
R
T
hi
T
ho

T
co
T
ci

(21)
P is the efciency given by,
P
T
co
T
ci

T
hi
T
ci

(22)
Considering overall heat transfer coefcient, the heat exchanger
surface area (A) is computed by,
A
Q
UFLMTD
(23)
For sensible heat transfer, the heat transfer rate is given by,
Q m
h
C
ph
T
hi
T
ho
m
c
C
pc
T
co
T
ci
(24)
Based on total heat exchanger surface area (A), the necessary
tube length (L) is,
L
A
pd
o
N
t
(25)
3.2. Pressure drop
The pressure drop allowance in heat exchanger is the static uid
pressure which may be expended to drive the uid through the
exchanger. In all heat exchanger there is close physical and
economical afnity between heat transfer and pressure drop. For
a constant heat capacity in the heat exchanger that is to be
designed, increasing the ow velocity will cause a rise of heat
transfer coefcient which results in compact exchanger design and
lower investment cost. However increase of owvelocity will cause
more pressure drop in heat exchanger which results in additional
running cost. For this reason when designing a heat exchanger
pressure drop must be considered with heat transfer and best
solution for the system must be found.
Tube side pressure drop includes distributed pressure drop
along the tube length and concentrated pressure losses in elbows
and in the inlet and outlet nozzle [3].
DP
t
DP
tube length
DP
tube elbow
(26)
DP
t

r
t
v
2
t
2
_
L
d
i
f
t
p
_
n (27)
Different values of constant p are considered by different
authors. Kern [3] assumed p 4, while Sinnot et al. [30] assumed
p 2.5.
The shell side pressure drop is,
DP
s
f
s
_
r
s
v
2
s
2
__
L
B
__
D
s
d
e
_
(28)
where,
f
s
2b
o
Re
0:15
s
(29)
and b
o
0.72 [34] valid for Re
s
< 40,000.
Considering pumping efciency h, pumping power computed
by,
P
1
h
_
m
t
s
t
DP
t

m
s
s
s
DP
s
_
(30)
3.3. Objective function
Total cost C
tot
is taken as the objective function, which includes
capital investment (C
i
), energy cost (C
e
), annual operating cost (C
o
)
and total discounted operating cost (C
od
) [9].
C
tot
C
i
C
od
(31)
Adopting Hall's correlation [35], the capital investment C
i
is
computed as a function of the exchanger surface area.
C
i
a
1
a
2
A
a=3
(32)
where, a
1
8000, a
2
259.2 and a
3
0.93 for exchanger made
with stainless steel for both shell-and-tubes [35].
The total discounted operating cost related to pumping power to
overcome friction losses is computed from the following equation,
C
o
PC
e
H (33)
C
od

ny
x 1
C
o
1 i
x
(34)
Based on all above calculations, total cost is computed from
equation (31). The procedure is repeated computing new value of
exchanger area (A), exchanger length (L), total cost (C
tot
) and
a corresponding exchanger architecture meeting the specications.
Each time the optimization algorithm changes the values of the
design variables d
o
, D
s
and B in an attempt to minimize the
objective function.
Present approach uses the following steps for optimal heat
exchanger design:
V.K. Patel, R.V. Rao / Applied Thermal Engineering 30 (2010) 1417e1425 1420
Step 1: Assuming values of a set of design variables and esti-
mating heat transfer area of the heat exchanger based on the
required heat duty and other design specication.
Step 2: Evaluation of the capital investment, operating cost and
formulation of the objective function.
Step 3: Utilization of the PSO algorithm to select a new set of
values for the design variables.
Step 4: Iteration of the previous steps until a minimum of the
objective function is found.
4. Case studies
The effectiveness of the present approach using PSO is assessed
by analyzing four case studies.
Case 1: 4.34(MW) duty, methanolebrackish water exchanger
[30].
Case 2: 1.44(MW) duty, keroseneecrude oil exchanger [3].
Case 3: 0.46(MW) duty, distilled watereraw water exchanger
[30].
Case 4: 2.09(MW) duty, waterewater exchanger.
The rst three case studies were analyzed by Caputo et al. [9]
using GA approach and taken from literature [3,30]. The forth
case study was analyzed by Selbas et al. [2] using GA approach.
The original design specications, shown in Table 2, are
supplied as inputs to the described PSO algorithm for each of the
rst three cases. The resulting optimal exchanger architectures
obtained by PSO are compared with the results obtained by
Caputo et al. [9] using GA approach and with original design
solution given by Sinnot et al. [30] and Kern [3] (shown in Tables
3a and 3b). In order to allow a consistent comparison, cost
functions of all three approaches are computed as described in
Section 3.3. Also for the same reason, all the values related to cost
are taken from the work of Caputo et al. [9] who tried all the case
studies by GA approach.
In the PSO approach following upper and lower bounds for the
optimization variables are imposed: shell inside diameter (D
s
)
ranging between 0.1 m and 1.5 m; tubes outer diameter (d
o
)
ranging between 0.015 m and 0.051 m; bafe spacing (B) ranging
from 0.05 m to 0.5 m. All the values of discounted operating costs
are computed with ny 10 yr, annual discount rate (i) 10%,
energy cost (C
e
) 0.12 V/kW h and an annual amount of work
hours H 7000 yr/h.
Table 2
The process input and physical properties for different case studies [9].
Mass ow (kg/s) T
input
(

C) T
output
(

C) r (kg/m
3
) C
p
(kJ/kg K) m (Pa s) k (W/m K) R
fouling
(m
2
K/W)
Case 1:
Shell side: methanol 27.80 95.00 40.00 750.00 2.84 0.00034 0.19 0.00033
Tube side: sea water 68.90 25.00 40.00 995.00 4.20 0.0008 0.59 0.0002
Case 2:
Shell side: kerosene 5.52 199.00 93.30 850.00 2.47 0.0004 0.13 0.00061
Tube side: crude oil 18.80 37.80 76.70 995.00 2.05 0.00358 0.13 0.00061
Case 3:
Shell side: distilled water 22.07 33.90 29.40 995.00 4.18 0.0008 0.62 0.00017
Tube side: raw water 35.31 23.90 26.70 999.00 4.18 0.00092 0.62 0.00017
Table 3a
Optimal heat exchanger geometry using different optimization methods.
Case study 1 Case study 2 Case study 3
Literature GA PSO Literature GA PSO Literature GA PSO
L (m) 4.83 3.379 3.115 4.88 2.153 1.56 4.88 1.548 1.45
d
o
(m) 0.02 0.016 0.015 0.025 0.02 0.015 0.013 0.016 0.0145
B (m) 0.356 0.5 0.424 0.127 0.12 0.1112 0.305 0.44 0.423
D
s
(m) 0.894 0.83 0.81 0.539 0.63 0.59 0.387 0.62 0.59
S
t
(m) 0.025 0.02 0.0187 0.031 0.025 0.0187 0.023 0.02 0.0181
cl (m) 0.005 0.004 0.0037 0.006 0.005 0.0037 0.004 0.004 0.0036
N
t
918 1567 1658 158 391 646 160 803 894
v
t
(m/s) 0.75 0.69 0.67 1.44 0.87 0.93 1.76 0.68 0.74
Re
t
14,925 10,936 10,503 8227 4068 3283 36,400 9487 9424
Pr
t
5.7 5.7 5.7 55.2 55.2 55.2 6.2 6.2 6.2
h
t
(W/m
2
K) 3812 3762 3721 619 1168 1205 6558 6043 5618
f
t
0.028 0.031 0.0311 0.033 0.041 0.044 0.023 0.031 0.0314
DP
t
(Pa) 6251 4298 4171 49,245 14,009 16,926 62,812 3673 4474
a
s
(m
2
) 0.032 0.0831 0.0687 0.0137 0.0148 0.0131 0.0236 0.0541 0.059
d
e
(m) 0.014 0.011 0.0107 0.025 0.019 0.0149 0.013 0.011 0.0103
v
s
(m/s) 0.58 0.44 0.53 0.47 0.43 0.495 0.94 0.41 0.375
Re
s
18,381 11,075 12,678 25,281 18,327 15,844 16,200 8039 4814
Pr
s
5.1 5.1 5.1 7.5 7.5 7.5 5.4 5.4 5.4
h
s
(W/m
2
K) 1573 1740 1950.8 920 1034 1288 5735 3476 4088.3
f
s
0.33 0.357 0.349 0.315 0.331 0.337 0.337 0.374 0.403
DP
s
(Pa) 35,789 13,267 20,551 24,909 15,717 21,745 67,684 4365 4721
U (W/m
2
K) 615 660 713.9 317 376 409.3 1471 1121 1177
A (m
2
) 278.6 262.8 243.2 61.5 52.9 47.5 46.6 62.5 59.15
C
i
(V) 51,507 49,259 46,453 19,007 17,599 16,707 16,549 19,163 18,614
C
o
(V/yr) 2111 947 1038.7 1304 440 523.3 4466 272 276
C
od
(V) 12,973 5818 6778.2 8012 2704 3215.6 27,440 1671 1696
C
tot
(V) 64,480 55,077 53,231.1 27,020 20,303 19,922.6 43,989 20,834 20,310
V.K. Patel, R.V. Rao / Applied Thermal Engineering 30 (2010) 1417e1425 1421
To obtain the optimum solution using PSO, experiments are
conducted to decide the minimumnumber of particles and number
of generations required to obtain the optimum solution using PSO.
After obtaining the number of particles and number of generations,
PSO algorithm is run for 50 times by considering the following
parameters.
Number of particles (population) 50
Maximum number of generations (iterations) 100
Variation of inertial weight (w) 0.9e0.4
Cognitive parameter (c
1
) 2
Social parameter (c
2
) 2
Maximum velocity (V
max
) 1
Table 3a shows the optimized parameters of the case studies
obtained using PSOand comparisonwith the optimized parameters
obtained by Caputo et al. [9] using GA approach, while Fig. 5 shows
the cost comparison of both the approaches for all the three case
studies. Similarly, Table 3b shows the comparison of the optimized
parameters obtained by using present approach (PSO) with the
earlier optimized parameters obtained by Selbas et al. [2] using GA
approach.
4.1. Case 1: 4.34(MW) duty, methanolebrackish water exchanger
The original design as well as the design proposed by Caputo
et al. [9] using GA approach assumed an exchanger with two tube
side passages and one shell side passage. The same conguration is
retained in the present approach.
Result shows that a signicant increase in the number of tubes
reduces the tube side ow velocity which consecutively reduces
the tube side heat transfer coefcient by 1.08%. The reduction in
shell diameter increases the shell side ow velocity which
consecutively increases the shell side heat transfer coefcient by
12.1%. The overall effect of these higher shell side heat transfer
coefcient result in 8.2% increase in overall heat transfer coefcient
Table 3b
Optimal heat exchanger geometry for case study 4.
GA PSO
L (m) 0.74 0.61
d
o
(m) 0.016 0.01
B (m) 0.45 0.45
D
s
(m) 1.1 0.88
S
t
(m) 0.02 0.0125
v
t
(m/s) 0.625 0.7
Re
t
18,383 12,868
Pr
t
2.87 2.87
h
t
(W/m
2
K) 4814 5790
f
t
0.0267 0.0293
DP
t
(Pa) 3163.8 5950
a
s
(m
2
) 0.099 0.0792
d
e
(m) 0.158 0.009
v
s
(m/s) 0.34 0.42
Re
s
6277 4903
Pr
s
5.8 5.8
h
s
(W/m
2
K) 3217 4494
f
s
0.3879 0.4025
DP
s
(Pa) 1976.2 3279
U (W/m
2
K) 676 752
A (m
2
) 94.1 84.6
C
i
(V) 1372.4 1281.3
C
o
(V/yr) 58.8 95.99
C
tot
(V/yr) 1431.2 1377.2
Fig. 1. Convergence of PSO for case study 1.
Fig. 2. Convergence of PSO for case study 2.
Fig. 3. Convergence of PSO for case study 3.
V.K. Patel, R.V. Rao / Applied Thermal Engineering 30 (2010) 1417e1425 1422
which in turn leads to 7.45% reduction in heat exchanger area and
7.81% reduction in exchanger length in the present approach as
compared to GA approach considered by Caputo et al. [9]. The
capital investment is also decreased corresponding to 5.7% because
of reduction in heat exchanger area The reduction in tube side ow
velocity reduces the tube side pressure drop while the high shell
side ow velocity increases the shell side pressure drop. Therefore,
14.1% increment in the annual pumping cost is observed in the
present case. The combined effect of capital investment and oper-
ating costs led to a reduction of the total cost of about 4.1% with
respect to GA approach considered by Caputo et al. [9].
Fig. 1 shows the convergence of the objective function using
PSO. It can be observed that the objective function converges
within about 10 generations for this case.
4.2. Case 2: 1.44(MW) duty, keroseneecrude oil exchanger
The original designas well as the designproposedbyCaputoet al.
[9] using GA approach assumed an exchanger with four tube side
passages (with square pitch pattern) and one shell side passage. The
same conguration is retained in the present approach. It is
observed that in this case higher tube side ow velocity increases
the tube side heat transfer coefcient by 3.2%. Similarly, high shell
side owvelocity increases the shell side heat transfer coefcient by
24.5%. An 8.85% increment in overall heat transfer coefcient is
observed in the present case because of the combined increment in
tube side and shell side heat transfer coefcient. As a result of high
overall heat transfer coefcient, a reduction of 10.2% in heat
exchanger area and reduction of 27.5% in heat exchanger length is
observed compared to GA approach considered by Caputo et al. [9].
The capital investment is decreased by 5.1%. The higher tube side
and shell side ow velocity increases the tube side and shell side
pressure drop. Therefore 15.9% increment in the annual pumping
cost is observed in the present case. Overall 3.35% reduction in total
investment is observed using PSO approach as compared to the GA
approach considered by Caputo et al. [9].
As shown in Fig. 2, in this case also PSO converges to optimum
solution quite rapidly (within 22 generations).
4.3. Case 3: 0.46(MW) duty, distilled watereraw water exchanger
This case study is taken from Kern [3] and the same case study
was considered by Caputo et al. [9]. Two tube side passages
(triangle pitch pattern) and one shell side passage are assumed for
this exchanger.
Fig. 4. Convergence of PSO for case study 4.
49259
46453
17599
16707
19163
18614
5818
6778
2704
3215
1671
1696
0
10000
20000
30000
40000
50000
60000
GA Approch
Caputo et al.[9]
Present work
(PSO)
GA Approch
Caputo et al.[9]
Present work
(PSO)
GA Approch
Caputo et al.[9]
Present work
(PSO)
)
(
t
s
o
C
Initial Cost
Operating Cost
Case 1
Case 2
Case 3
TC=55077
TC=53231
TC=20303
TC=19922
TC=20834
TC=20310
TC = Total Cost
Fig. 5. Overall cost comparison.
V.K. Patel, R.V. Rao / Applied Thermal Engineering 30 (2010) 1417e1425 1423
In this case also a reduction in the heat exchanger area is
observed (about 5.36%) because of higher overall heat transfer
coefcient. The capital investment is decreased by 2.86%
compared to GA approach considered by Caputo et al. [9].
However, the increment in total pressure losses (about 14.4%)
results in 1.5% increment in annual operating expense. Therefore,
a cumulative effect of reduction in capital investment and
increment in operating expense led to a reduction of the total cost
of about 2.51% compared to GA approach considered by Caputo
et al. [9].
Fig. 3 shows the convergence of the objective function using PSO
algorithm. The convergence is obtained within 7 generations.
To demonstrate the effectiveness of PSO technique, one more
case study is considered which was earlier analyzed by Selbas et al.
[2] using GA approach. For comparison of the results, the design
methodology, cost function and all the values related to cost are
taken from Selbas et al. [2].
4.4. Case 4: 2.09(MW) duty, waterewater exchanger
The design proposed by Selbas et al. [2] using GA approach
assumed an exchanger with two tube side passages (with square
pitch pattern) and one shell side passage. The same conguration is
retained in the present approach. Results show that higher tube
side and shell side ow velocity increases the tube side and shell
side heat transfer coefcient (20.2% on tube side and 39.6% on shell
side) which in turn results in 11.24% increment in overall heat
transfer coefcient in the present approach. The higher overall heat
transfer coefcient results in 10.1% reduction in heat exchanger
area and 17.5% reduction in heat exchanger length in the present
approach compared to GA approach considered by Selbas et al. [2].
The capital investment is decreased by 6.6%. Though the higher
tube side and shell side ow velocity along with the smaller tube
and shell size increases the tube side and shell side pressure drop
(46.8% on tube side and 65% on shell side). The annual operating
cost is increased by 63.2%. In general, the combined effect of capital
investment and operating costs results in 3.77% reduction in the
total cost/year in the present approach compared to GA approach
considered by Selbas et al. [2].
Fig. 4 shows the convergence of the objective function which is
obtained within about 15 generations using PSO.
Table 4
Comparison of sensitivity to electricity price for case study 1.
C
e
0.06 V/kW h C
e
0.12 V/kW h C
e
0.18 V/kW h
GA PSO GA PSO GA PSO
L (m) 3.98 3.59 3.379 3.115 3.021 2.95
d
o
(m) 0.016 0.015 0.016 0.015 0.016 0.015
B (m) 0.46 0.3734 0.5 0.424 0.5 0.5
D
s
(m) 0.75 0.73 0.830 0.81 0.9 0.86
S
t
(m) 0.02 0.0187 0.020 0.0187 0.02 0.0187
cl (m) 0.004 0.0037 0.004 0.0037 0.004 0.0037
N
t
1227 1837 1567 1658 1824 1892
v
t
(m/s) 0.88 0.843 0.69 0.67 0.59 0.56
Re
t
13,963 13,213 10,936 10,503 9394 9202
Pr
t
5.7 5.7 5.7 5.7 5.7 5.7
h
t
(W/m
2
K) 4574 4471.2 3762 3721 3331 3249.1
f
t
0.028 0.0291 0.031 0.0311 0.032 0.0322
DP
t
(Pa) 8741 6967.3 4298 4171 3483 3167
a
s
(m
2
) 0.0696 0.0545 0.0831 0.0687 0.0901 0.0860
d
e
(m) 0.011 0.0107 0.011 0.0107 0.011 0.0107
v
s
(m/s) 0.53 0.67 0.44 0.53 0.41 0.4310
Re
s
13,378 15,974 11,075 12,678 10,347 10,126
Pr
s
5.1 5.1 5.1 5.1 5.1 5.1
h
s
(W/m
2
K) 1906 2215.1 1740 1950.8 1655 1723
f
s
0.347 0.337 0.357 0.349 0.361 0.361
DP
s
(Pa) 20,920 38,593 13,267 20,551 10,936 11,986
U (W/m
2
K) 707 777.73 660 713.9 626 660.25
A (m
2
) 245.5 223.2 262.8 243.2 277 263
C
i
(V) 46,776 43,573 49,259 46,453 51,280 49,290
C
o
(V/yr) 828 1004.3 947 1038.7 1163 1045.1
C
od
(V) 5088 6171 5818 6778.2 7148 6422
C
tot
(V) 51,864 49,744 55,077 53,231.1 58,427 55,711
Table 5
Sensitivity to objective function for case study 1.
Objective function: Dh h
t
h
s
Literature GA PSO
L (m) 4.83 1.489 1.7471
d
o
(m) 0.02 0.016 0.015
B (m) 0.356 0.39 0.3425
D
s
(m) 0.894 1.4 1.1944
S
t
(m) 0.025 0.02 0.0187
cl (m) 0.005 0.004 0.0037
N
t
918 4935 3907
v
t
(m/s) 0.75 0.22 0.3136
Re
t
14,925 3458 4680
Pr
t
5.7 5.7 5.7
h
t
(W/m
2
K) 3812 1491.19 1770.04
f
t
0.028 0.042 0.0394
DP
t
(Pa) 6251 302 805.4
a
s
(m
2
) 0.032 0.1096 0.0818
d
e
(m) 0.014 0.011 0.0107
v
s
(m/s) 0.58 0.33 0.4530
Re
s
18,381 8429 10,670
Pr
s
5.1 5.1 5.1
h
s
(W/m
2
K) 1573 1491 1769.9
f
s
0.33 0.372 0.3582
DP
s
(Pa) 35,789 7529 15,733
U (W/m
2
K) 615 474 540.2
A (m
2
) 278.6 366.2 321.4
C
i
(V) 51,507 63,792 57,563
C
o
(V/yr) 2111 360 670.8
C
od
(V) 12,973 2210 4122
C
tot
(V) 64,480 66,002 61,685
Dh (W/m
2
K) 0.19 0.14
V.K. Patel, R.V. Rao / Applied Thermal Engineering 30 (2010) 1417e1425 1424
To assess sensitivity of PSO, trials are made in the present work
by parametrically changing electricity cost in the total cost function.
The effect of 50% variation of electricity price with respect to
nominal value is examined for case study 1, the results of both the
approaches are shown in Table 4. It is observed that PSO responded
correctly by reducing the operating cost when electricity price
increased at the expense of increased exchanger surface area, and
making opposite when electricity cost is reduced.
Finally, sensitivity of both the approaches is compared for some
other objective function instead of total cost minimization. Mini-
mization of difference between tube side heat transfer coefcient
and shell side heat transfer coefcient is considered as an objective
function.
Dh Dh
t
Dh
s
(35)
The results of the present approach using PSO are shown in
Table 5. Quite high shell side and tube side heat transfer coefcients
(1770 W/m
2
K) are observed in the present approach.
5. Conclusion
Heat exchangers are an integral component of all thermal
systems. Their designs should be adapted well to the applications in
which they are used; otherwise their performances will be
deceiving and their costs excessive. Heat exchanger design can be
a complex task, and advanced optimization tools are useful to
identify the best andcheapest heat exchanger for a specic duty. The
present study has demonstrated successful application of PSO
technique for the optimal design of a shell-and-tube heat exchanger
fromeconomic viewpoint. The presentedPSOtechnique is simple in
concept, few in parameters and easy for implementation. These
features boost the applicability of the PSO particularly in thermal
system design, where the problems are usually complex and have
a large number of variables and discontinuity in the objective
function. The presented PSO technique's ability is demonstrated
using different literature case studies and the performance results
are compared with those obtained by the previous researchers. PSO
converges to optimum value of the objective function within quite
fewgenerations and this feature signies the importance of PSO for
heat exchanger optimization. The PSO technique can be easily
modied to suit optimization of various thermal systems.
References
[1] M. Fesanghary, E. Damangir, I. Soleimani, Design optimization of shell-and-
tube heat exchangers using global sensitivity analysis and harmony search
algorithm. Applied Thermal Engineering 29 (2009) 1026e1031.
[2] R. Selbas, O. Kizilkan, M. Reppich, A new design approach for shell-and-tube
heat exchangers using genetic algorithms from economic point of view.
Chemical Engineering and Processing 45 (2006) 268e275.
[3] D.Q. Kern, Process Heat Transfer. McGraw-Hill, New York, 1950.
[4] W.M. Rohsenow, J.P. Hartnett, Handbook of Heat Transfer. McGraw-Hill, New
York, 1973.
[5] G.F. Hewitt, Heat Exchanger Design Handbook. Begell House, New York, 1998.
[6] R.K. Shah, K.J. Bell, CRC Handbook of Thermal Engineering. CRC Press, Florida,
2000.
[7] P.D. Chauduri, U. Diwekar, J. Logsdon, An automated approach for the optimal
design of heat exchangers. Industrial & Engineering Chemistry Research 36
(1997) 3685e3693.
[8] Y. Ozcelik, Exergetic optimization of shell-and-tube heat exchanger using
a genetic based algorithm. Applied Thermal Engineering 27 (2007)
1849e1856.
[9] A.C. Caputo, P.M. Pelagagge, P. Salini, Heat exchanger design based on
economic optimization. Applied Thermal Engineering 28 (2008)
1151e1159.
[10] I. Ozkol, G. Komurgoz, Determination of the optimum geometry of the heat
exchanger body via a genetic algorithm. International Journal of Heat and
Mass Transfer 48 (2005) 283e296.
[11] R. Hilbert, G. Janiga, R. Baron, D. Thevenin, Multi objective shape optimization
of a heat exchanger using parallel genetic algorithm. International Journal of
Heat and Mass Transfer 49 (2006) 2567e2577.
[12] G.N. Xie, B. Sunden, Q.W. Wang, Optimization of compact heat exchangers by
a genetic algorithm. Applied Thermal Engineering 28 (2008) 895e906.
[13] J.M. Ponce-Ortega, M. Serna-Gonzalez, A. Jimenez-Gutierrez, Use of genetic
algorithms for the optimal design of shell-and-tube heat exchangers. Applied
Thermal Engineering 29 (2009) 203e209.
[14] J.M. Ponce-Ortega, M. Serna-Gonzalez, A. Jimenez-Gutierrez, Design and
optimization of multipass heat exchangers. Chemical Engineering and Pro-
cessing 47 (2008) 906e913.
[15] M. Saffar-Avval, E. Damangir, A general correlation for determining optimum
bafe spacing for all types of shell-and-tube exchangers. International Journal
of Heat and Mass Transfer 38 (1995) 2501e2506.
[16] B.K. Soltan, M. Saffar-Avval, E. Damangir, Minimizing capital and operating
costs of shell-and-tube condensers using optimum bafe spacing. Applied
Thermal Engineering 24 (2004) 2801e2810.
[17] M. Reppich, J. Kohoutek, Optimal design of shell-and-tube heat exchangers.
Computer & Chemical Engineer 18 (1994) S295eS299.
[18] S. Sun, Y. Lu, C. Yan, Optimization in calculation of shell-and-tube heat
exchanger. International Communications in Heat and Mass Transfer 20 (1993)
675e685.
[19] P.P.P.M. Lerou, T.T. Veenstra, J.F. Burger, H.J.M. Brake, H. Rogalla, Optimization
of counter ow heat exchanger geometry through minimization of entropy
generation. Cryogenics 45 (2005) 659e669.
[20] A.L.H. Costa, E.M. Queiroz, Design optimization of shell-and-tube heat
exchanger. Applied Thermal Engineering 28 (2008) 1798e1805.
[21] H. Li, V. Kottke, Visualization and determination of local heat transfer coef-
cients in shell-and-tube heat exchangers for staggered tube arrangement by
mass transfer measurements. Experimental Thermal and Fluid Science 17
(1998) 210e216.
[22] P. Wildi-Tremblay, L. Gosselin, Minimizing shell-and-tube heat exchanger cost
with genetic algorithms and considering maintenance. International Journal
of Energy Research 31 (9) (2007) 867e885.
[23] B. Allen, L. Gosselin, Optimal geometry and ow arrangement for minimizing
the cost of shell-and-tube condensers. International Journal of Energy
Research 32 (10) (2008) 958e969.
[24] B.V. Babu, S.A. Munawar, Differential evolution strategies for optimal design of
shell-and-tube heat exchangers. Chemical Engineering Science 62 (14) (2007)
3720e3739.
[25] L. Valdevit, A. Pantano, H.A. Stone, A.G. Evans, Optimal active cooling perfor-
mance of metallic sandwich panels with prismatic cores. International Journal
of Heat and Mass Transfer 49 (21e22) (2006) 3819e3830.
[26] F. Pettersson, J. Soderman, Design of robust heat recovery systems in paper
machines. Chemical Engineering and Processing 46 (10) (2007) 910e917.
[27] J. Kennedy, R. Eberhart, Particle swarm optimization, in: Proceedings of the
IEEE International Conference on Neural Networks, Perth, Australia, 1995,
pp.1942e1948.
[28] Y. Dong, J. Tang, B. Xu, D. Wang, An application of swarm optimization to
nonlinear programming. Computer and Mathematics with Applications 49
(2005) 1655e1668.
[29] F. Bergh, A.P. Engelbrecht, A study of particle swarm optimization particle
trajectories. Information Sciences 176 (2006) 937e971.
[30] R.K. Sinnot, J.M. Coulson, J.F. Richardson, Chemical Engineering Design, vol. 6,
Butterworth-Heinemann, Boston MA, 1996.
[31] F.P. Incropera, D.P. DeWitt, Fundamentals of Heat and Mass Transfer. John
Wiley, New York, 1996.
[32] A.P. Fraas, Heat Exchanger Design, second ed. John Wiley, New York, 1989.
[33] M.M. Ohadi, The Engineering Handbook. CRC Press LLC, Florida, 2000.
[34] M.S. Peters, K.D. Timmerhaus, Plant Design and Economics for Chemical
Engineers. McGraw-Hill, New York, 1991.
[35] M. Taal, I. Bulatov, J. Klemes, P. Stehlik, Cost estimation and energy price
forecast for economic evaluation of retrot projects. Applied Thermal Engi-
neering 23 (2003) 1819e1835.
V.K. Patel, R.V. Rao / Applied Thermal Engineering 30 (2010) 1417e1425 1425

You might also like