A Minlp Model For The Rigorous Design of Shell and Tube

A MINLP MODEL FOR THE RIGOROUS DESIGN
OF SHELL AND TUBE HEAT EXCHANGERS

USING THE TEMA STANDARDS
M. A. S. S. Ravagnani1, and J. A. Caballero2
1
DEQ/PEQ/CTC/UEM, State University of Maringa, Maringa, PR, Brazil.

Chemical Engineering Department, University of Alicante, Alicante, Spain.
Abstract: In this paper a mixed integer non-linear programming (MINLP) model is proposed for
the design of shell and tube heat exchangers. The model rigorously follows the TEMA (Tubular
Exchanger Manufacturers Association) Standards and Bell-Delaware Method is used to the shell
side calculations. Mechanical design features (shell and tube bundle diameters, internal and
external tube diameters, tubes length, pitch and tube arrangement, number of tubes and tube
passes) and thermal-hydraulic variables (heat, area, individual and overall heat transfer coefcients, shell and tube pressure drops and fouling) are variables to be optimized. The equipment
is designed under pressure drop and fouling limits. Three cases from the literature are studied,
with two different objective functions, considering just the heat transfer area minimization or the
annual cost minimization, including area and pumping expenses. More realistic values are
obtained when compared with the literature, considering fouling and pressure drop effects
according to TEMA Standards.
Keywords: TEMA standards; optimization; heat exchanger design; mixed integer non linear
programming; generalized disjunctive programming; mathematical programming.
INTRODUCTION

Correspondence to:
Dr M.A.S.S. Ravagnani,
DEQ/PEQ, CTC/UEMState University of Maringa,
Av. Colombo 5790CEP
87030 900, Maringa,
PR, Brazil.
E-mail: ravag@deq.uem.br
DOI: 10.1205/cherd06231
02638762/07/
$30.00 0.00
Chemical Engineering
Research and Design
Trans IChemE,
Part A, October 2007
# 2007 Institution
of Chemical Engineers
more realistic and accurate results for the

shell side lm heat transfer coefcient and
pressure drop. The ow Bell model considers
ve different streams, not considered in the
method proposed by Kern (1950). Leakages
between tubes and bafes, bypass of the
tube bundle without cross ow, leakages
between shell and bafes, leakages due to
more than one tube passes and the main
stream and tube bundle cross ow.
These streams do not occur in so well
dened regions, but interacts ones to others,
needing a complex mathematical treatment
to represent the real shell side ow.
Both, Kern and Bell-Delaware methods are
used in practice. However, in the majority of
published papers and in industrial applications, heat transfer coefcients are estimated, based, generally on literature tables.
These values have a large degree of uncertainty, and more realistic values can be
obtained if these coefcients are not estimated, but calculated during the design task.
A very few number of papers present shell
and tube heat exchanger design including
overall heat transfer coefcient calculations.
Polley et al. (1990), Polley and Panjeh
Shah (1991), Jegede and Polley (1992) and
Panjeh Shah (1992), using both, the Kern
Shell and tube heat exchangers are the most

used heat transfer equipment in industrial processes due to their resistant manufacturing
features and design exibility. They are also
easy adaptable to operational conditions.
Because of this, the design of shell and tube
heat exchangers is a much known subject.
Nevertheless, some difculties are found,
especially in the shell-side design, because
of the complex characteristics of heat transfer
and pressure drop.
Some methods were proposed to determine
thermal-hydraulic parameters, like heat
exchange area, individual and overall heat
transfer coefcients and pressure drop
(Taborek, 1983). Kern (1950) published the
rst and most known method. Kerns method
was proposed to design heat exchangers, or
to rate existent equipment with respect to
pressure drop and fouling. For the shell side,
correlations were proposed based on the equivalent diameter, overestimating the parameters
design. It is, to this day, the most used method.
Bell-Delaware is a more recent shell and
tube heat exchanger design method based
on mechanical shell side details. According
to Taborek (1983), this method presents
1423
Vol 85 (A10) 1423 1435
1424
RAVAGNANI and CABALLERO
and the Bell-Delaware method, proposed an algorithm for

shell and tube heat exchangers design to be incorporated
into a heat exchanger network synthesis algorithm. The
method consists of solving simultaneously three equations.
In these equations, the authors proposed three constants,
one that depends on the mass owrate and the tube side
uid physical properties. The other two are complex functions
involving shell side uid physical properties, shell geometry
and the Fanning and Colburn ideal factors to the Bell-Delaware method. With the simultaneous solution of these
equations, lm coefcients and pressure drops can be
obtained. The TEMA (1988) Standards were not always
used by the authors, and as result some variables like the
shell diameter and the number of tubes are not always in
accordance with TEMA recommendations.
In Ravagnani (1994), a systematic procedure was proposed for the design of shell and tube heat exchangers
using the Bell-Delaware method. Overall and individual
heat transfer coefcients are calculated based on a TEMA
tube counting table, beginning with the smallest heat exchanger with the biggest number of tube passes, to prot from all
the pressure drop and fouling limits, xed before the design
and that must be respected. If pressure drops or fouling
factor are not satised, a new heat exchanger is tested,
with lower tube passes number or larger shell diameter,
until the pressure drops and fouling are under the xed
limits. This equipment is, inside a trial and error systematic,
the one that presents the minimum heat exchanger area for
xed tube length and bafe cut, for a counting tube TEMA
table including 21 types of shell and tube bundle diameter,
two types of external tube diameter, three types of tube
pitch, two types of tube arrangement and ve types of
number of tube passes. This systematic was also incorporated in a heat exchanger network synthesis method, as it is
shown in Ravagnani et al. (2003).
In a recent paper, Mizutani et al. (2003) presented a mathematical programming model for heat exchangers design.
The optimisation model uses the Bell-Delaware correlations
to calculate the heat transfer coefcient and pressure drop
in the shell-side ow. The model is based on GDP (generalized disjunctive programming) and is optimized with a MINLP
reformulation to determine the heat exchanger design that
minimizes the total annual cost accounting for area and
pumping expenses. The model follows TEMA standards,
but some features, such as the shell diameter and the
number of tubes, that are calculated and optimized, are not
always in accordance with the standards. The model will
nd the best value to optimise the objective function. So,
the number of tubes can be very different from the tube
counting TEMA standards. Also, tube length must be xed
before the design.
More recently in Serna and Jimenez (2005) an analytical
expression that relates the pressure drop, the exchanger
area and the lm heat transfer coefcient for the shell side
of a shell and tube heat exchanger was reported. The
equation was based on the Bell-Delaware method and the
use of the compact formulation within design and optimization algorithms was illustrated.
In the present paper, Bell-Delaware method will be used
to formulate the mathematical model that involves discrete
and continuous variables for the selection of the conguration and operating levels, respectively. A tube counting
table, similar to the one used by Ravagnani et al. (2003)
is proposed. Following the TEMA Standards, it allows to

nd the shell diameter, the tube bundle diameter, the external tube diameter, the tube pitch, the tube arrangement pattern, the number of tube passes and the number of tubes.
By using a GDP formulation, these optimization variables
are easy and rapidly obtained. Other variables are obtained
by using some of the Mizutani et al. (2003) GDP model.
Furthermore, some complementing features are proposed.
Besides the table counting, shell and tube side pressure
drops and fouling factor are calculated and the model has
as constraints operational limits, previously xed, as in
industrial applications. The objective function can be considered as the minimization of area and pumping expenses
or just heat exchange area minimization, depending on data
availability.
MODEL FORMULATION
The problem to be formulated as an optimization problem
is the design of the optimum shell and tube equipment to
exchange heat between a cold and a hot stream. The objective is to nd the heat exchanger that presents the minimum
cost including exchange area cost and/or pumping cost,
rigorously following the Standards of TEMA constrained to
allowable pressure drops and fouling limits. Inlet data for
both uids are: Tin (inlet temperature), Tout (outlet temperature), m (mass owrate), r (density), Cp (heat capacity),
m (viscosity), k (thermal conductivity), allowable DPdesign
(pressure drop), rddesign (fouling factor) and area cost data.
The indexh is used for the hot uid and the indexc for the
cold uid. The mechanical variables to be optimized are
tube inside diameter (din), tube outside diameter (dex), tube
arrangement (arr), tube pitch ( pt), tube length (L), number
of tube passes (Ntp) and number of tubes (Nt), for the tubeside. To the shell-side, the desired variables are the external
diameter (Ds), the tube bundle diameter (Dotl), bafes number
(Nb), number of shells (NS), bafes cut (lc) and bafe spacing
(ls). Finally, thermal-hydraulic variables to be calculated are
heat duty (Q), heat exchange area (A), tube-side and shellside lm coefcients (ht and hs), dirty and clean overall heat
transfer coefcient (Ud and Uc), pressure drops (DPt and
DPs), fouling factor (rd), log mean temperature difference
(LMTD), the correction factor of LMTD (Ft) and the uids
location inside the heat exchanger.
The model is formulated as a GDP Problem and reformulated to a MINLP problem.
Heat Exchanger Fluids Location

The uids allocation in shell and tube heat exchangers
must consider some important issues like the materials of
construction, fouling, operating pressures, pressure drops,
uids viscosity, owrates and temperatures. Therefore,
usually the decision of uids allocation in the tubes side or
in the shell side is taken a priori. However, if the designer
have all this information and is not sure in which side must
be allocated the uids, a mathematical approach can be
used to dene it. Obviously, this decision belongs to the
designer. In the present paper, the formulation proposed by
Mizutani et al. (2003) can be used to dene the hot uids
allocation. There are two possibilities, considering the cold
uid in the shell side or in the tube side. So, two binary variables must be dened, y1f and y2f . If the cold uid is owing in
Trans IChemE, Part A, Chemical Engineering Research and Design, 2007, 85(A10): 14231435

the shell side, or if the hot uid is on the tube side, y1f 1. It
implies that the physical properties and hot uid mass owrate will be in the tube side, and the cold uid physical properties and mass owrate will be directed to the shell side. If
y1f 0, the reverse occurs.
The physical properties (density, viscosity, heat capacity
and thermal conductivity) are assumed to be constant (temperature invariant).
This is formulated as:
2
y1f
y2f
6 mt mh 7 6 mt mc 7
6 s
7 6
7
6 m mc 7 6 ms mh 7
6 t
6
7
7
6 Cp Cph 7 6 Cpt Cpc 7
6 s
7 6 s
7
c
h
6 Cp Cp 7 6 Cp Cp 7
6
7 6
7
6 mt mh
7 _ 6 mt mc
7
6
7 6
7
6 ms mc 7 6 ms mh 7
6
7 6
7
6 kt kh
7 6 kt kc
7
6
7 6
7
6 ks kc
7 6 ks kh
7
6
7 6
7
4 rt rh
5 4 rt rc
5
rs rc
rs rh
y1f y2f 1
h
m mh1 mh2
mc mc1 mc2
mt mh1 mc1
ms mh2 mc2
mh1 mupper y1f
mc1 mupper y2f
mh2 mupper y2f
mc2 mupper y1f
mt y1f mh y2f mc
ms y2f mh y1f mc
Cpt y1f Cph y2f Cpc
Cps y2f Cph y1f Cpc
k t y1f k h y2f k c
k s y2f k h y1f k c
rt y1f rh y2f rc
rs y2f rh y1f rc
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
(10)
(11)
(12)
(13)
(14)
(15)
Table 1. Tube counting table proposed.

Dotl
dex
arr
pt
Ntp
0.20500
0.20500
0.20500
0.20500
0.20500
0.20500
0.20500
0.20500
0.20500
..
.
0.17325
0.17325
0.17325
0.17325
0.17325
0.17325
0.17325
0.17325
0.17325
..
.
0.01905
0.01905
0.01905
0.01905
0.01905
0.01905
0.01905
0.01905
0.01905
..
.
1
1
1
1
1
1
1
1
1
..
.
0.02379
0.02379
0.02379
0.02379
0.02379
0.02540
0.02540
0.02540
0.02540
..
.
1
2
4
6
8
1
2 30
4
6
..
.
24
16
..
.
1.52400
1.52400
1.52400
1.52400
1.52400
1.52400
1.52400
1.47300
1.47300
1.47300
1.47300
1.47300
1.47300
1.47300
0.02540
0.02540
0.02540
0.02540
0.02540
0.02540
0.02540
1
1
2
2
2
2
2
0.03175
0.03175
0.03175
0.03175
0.03175
0.03175
0.03175
6
8
1
2
4
6
8
1761
1726
1639
1615
1587
1553
1522
Ds
Nt
38
32
26
24
18
37
with 565 rows. Obviously, other values of external tube diameters as well as tube arrangement can be easily aggregated
to the table, if necessary. The number of tubes for various
bundle sizes and tube passes is estimated only, using the
correlations presented in Perry and Green (1997). According
to the authors, exact tube counts are part of the design package of most reputable exchanger design software and are
normally used for the nal design. To nd Ds, Dotl, dex, arr,
pt, ntp and Nt, the following equations are used:
2
3
Ds D565
3 2
3
s
Ds D1s
Ds D2s
6
7
6 Dotl D565 7
6
7 6
7
otl
6
7
6 Dotl D1otl 7 6 Dotl D2otl 7
6
7
6
7 6
7
6 dex d 565 7
6 dex d 1 7 6 dex d 2 7
ex
6
7
ex 7 6
ex 7
6
6
7
6
7 6
7
565 7
arr
arr
6 arr arr 1 7 _ 6 arr arr 2 7 _ _ 6
6
7
6
7 6
7
6
7
6 pt pt 1 7 6 pt pt 2 7
6 pt pt 565 7
6
7 6
7
6
7
6
7 6
7
6
7
4 ntp ntp1 5 4 ntp ntp2 5
6 ntp ntp565 7
4
5
Nt Nt 1
Nt Nt 2
565
Nt Nt
2
565
X
Ds
(16)
(17)
1425
dsi ynt(i)
(18)
i1
Dotl
565
X
dotli ynt(i)
(19)
dexi ynt(i)
(20)
arri:ynt(i)
(21)
i1
This formulation [equations (1) (17)] obviously can be

neglected if the designer has information enough to a priori
decide where the uids must be allocated. So, the next
step is the denition of the TEMA parameters.
For the denition of the shell diameter (Ds), tube bundle
diameter (Dotl), tube external diameter (dex), tube arrangement (arr), tube pitch ( pt), number of tube passes (Ntp) and
the number of tubes (Nt), a table containing this values
according to TEMA Standards is constructed, as presented
in Table 1. It contains two types of tube external diameter,
19.05 and 25.4 mm, two types of arrangement, triangular
and square, three types of tube pitch, 23.79, 25.4 and
31.75 mm, ve types of number of tube passes, 1, 2, 4, 6
and 8, and 21 different types of shell and tube bundle
diameter, beginning on 205 mm and 173.25 mm, respectively, and nishing in 1524 mm and 1473 mm, respectively,
dex
565
X
i1
arr
565
X
i1
pt
565
X
pti:ynt(i)
(22)
i1
ntp
565
X
ntpi:ynt(i)
(23)
i1
nt
565
X
nti:ynt(i)
(24)
i1
565
X
ynt(i) 1
i1
(25)
1426
Definition of the Tube Arrangement (arr) and the

Arrangement ( pn and pp) Variables
2
arr tri
arr cua
4 pn 0:5pt 5 _ 4 pn pt 5
pp 0:866pt
pp pt
That can be reformulated as follows:

pn pn1 pn2
1
(27)
pt pt 1 pt 2
(28)
pn1 0:5pt 1
(29)
pn2 pt 2
(30)
pp1 0:866pt 1
(31)
pp pt
1
pt
din (m)
10
11
12
13
14
15
16
17
18
0.0122
0.0129
0.0135
0.0142
0.0148
0.0154
0.0157
0.0161
0.0166
din din1 din2

y1dex
_j y1jbwg
or 1
y1dex
(46)
y1jbwg
1
(47)
y2jbwg 1
(48)
y2dex ) _j y2jbwg or 1 y2dex
X
j
(32)
arr
0:02379ytri
(33)
arr
pt 2 0:02379ycua
(34)
arr
pt 1 0:03175ytri
(35)
arr
pt 0:03175ycua
arr
arr
ytri
ycua
1
(36)
y1dex y2dex 1
(38)
(49)
y2jbwg
(50)
y1jbwg BWG1j
y2dex
or 1
y2jbwg
y2dex
1
Definition of Tube Length (L)

Five kinds of tube length are considered, according with
TEMA (1988):
NL {2:438; 3:658; 4:877; 6:096; 6:706}
According to TEMA (1988) the tube internal diameter

(din) can be found for different values of dex and BWG
(Birmingham Wire Gage), used to dene the tube thickness.
In the present paper, just two tube external diameters will be
considered, which implies just two set of possibilities of din, as
can be seen on Tables 2 and 3. However, other values of
tube external diameter can be aggregated.
BWG determination can be formulated as: for
dex 0.01905 m, BWG f10, 11, 12, . . . , 18g; for
dex 0.0254 m, BWG f8, 9, 10, 11, 12, . . . , 18g
9
X
y1jbwg ) y1dex or 1 y1jbwg y1dex 1
(37)
Definition of Tube Internal Diameter (din)
BWG1
BWG
(26)
pp pp pp
Table 2. Determination of din for dex 0.01905 m.
(39)
L 2:438y1l 3:658y2l 4:877y3l

y1l
6:096y4l 6:706y5l
(51)
y2l
(52)
y3l
y4l
y5l
Definition of Baffle Spacing (ls)

According to TEMA (1988), bafe spacing must be
between Ds and Ds/5. In this case, the following values will
be considered:
j1
9
X
y1jbwg 1
(40)
j1
BWG2
11
X
y2jbwg BWG2j
(53)
(54)
ls Ds
ls Ds =5
(41)
j1
11
X
Table 3. Determination of din for dex 0.0254 m.
y2jbwg 1
(42)
BWG
din (m)
8
9
10
11
12
13
14
15
16
17
18
0.0170
0.0179
0.0186
0.0193
0.0199
0.0206
0.0212
0.0217
0.0221
0.0225
0.0229
j1
BWG BWG1 BWG2
(43)
din can be found by the following equations:

din1
9
X
y1jbwg :din1
(44)
y2jbwg :din2
(45)
j1
din2
11
X
j1

Cross-Flow at or Near Centreline for One
Cross-Flow Section (Sm)
Reynolds number are generally high, laminar ow can be

neglected, and Reynolds number, in this work, will be
considered just for values greater than 100. The following
disjunction can be used:
Sm can be found by the following disjunction:

"
#
arr tri

_
otl dex )
Sm ls: Ds Dotl (ptdex ):(D
pt
"
#
arr sq

otl dex )
Sm ls: Ds Dotl (ptdex ):(D
pn
2
3 2 res
3

y3
y2res
y1res
45
35
4
4
_ Res 10
_ Res 10
Res . 104
Res . 103
Res 102
or
Using a Big M reformulation:

(pt dex ):(Dotl d ex )
Sm ls: Ds Dotl
pt
Res1 104 y1res
(62)
Res2 104 y1res
(63)
103 y2res
103 y2res
102 y3res
(64)
Res2
arr
ytri
)
M(1

(pt dex ):(Dotl dex )
Sm ls: Ds Dotl
pt
(55)
arr
M(1 ytri
)

Sm ls: Ds Dotl
pn
(56)
arr
M(1 ysq
)

Sm ls: Ds Dotl
pn
(57)
Res3
(65)
Res3
X
yrres 1
r
Res
(66)
(67)
Resr
(68)
arr
)
M(1 ysq
(58)
where M is a parameter large enough to be a valid upper

bound in equations (55) (58).
Definition of the Flow Regimen in the

Shell Side
Reynolds number is given by
Res
1427
ms :dex
ms :Sm
(59)
Colburn Factor ( ji) and Fanning Factor (fls)

Determination
Both, Colburn and Fanning factor are functions of
Reynolds number and the tube arrangement, as shown on
Table 4, extracted from Mizutani et al. (2003). Because of
rearr
this it is necessary to introduce a binary set of variables yr,s
,
which represents both the Reynolds number and the tube
arrangement. This set of binary variables is related to the
sets y res and y arr with the following DGP proposition:
rearr
(yrres ^ ysarr ) ) yr;s
This proposition is expressed by the constraints
The shell side uid velocity is given by

ms =rs
vs
(Ds =pt):(pt dex )ls
(60)
According to Smith (2005), the velocity limits must be

vs in m s1
0:5 vs 2;
rearr
1 yrres ysarr yr;s
0
XX
rearr
yr;s 1
(69)
(70)
The Colburn factor ( ji) and the shell-side Fanning factor

( s) are given by the empirical correlations
(61)
The ow regimen is dened as a function of Reynolds

number. Considering that in real heat exchangers the
ji a1 1:064a (Res )a2
(71)
fl s b1 1:064b (Res )b2
(72)
Table 4. Empirical coefcients for equations (69) (80) as function of Reynolds number and tube arrangement.
arr
Res
tri
tri
tri
tri
tri
sq
sq
sq
sq
sq
10 10
104 103
103 102
102 10
,10
105 104
4
10 103
103 102
102 10
,10
a1
a2
a3
a4
b1
b2
b3
b4
0.321
0.321
0.593
1.360
1.400
0.370
0.107
0.408
0.900
0.970
20.388
20.388
20.477
20.657
20.657
20.395
20.266
20.460
20.631
20.667
1.450
1.450
1.450
1.450
1.450
1.187
1.187
1.187
1.187
1.187
0.519
0.519
0.519
0.519
0.519
0.370
0.370
0.370
0.370
0.370
0.372
0.486
4.570
45.100
48.000
0.391
0.082
6.090
32.100
35.000
20.123
20.152
20.476
20.973
21.000
20.148
0.022
20.602
20.963
21.000
7.00
7.00
7.00
7.00
7.00
6.30
6.30
6.30
6.30
6.30
0.500
0.500
0.500
0.500
0.500
0.378
0.378
0.378
0.378
0.378
1428
a
b
a3
Number of Effective Cross-Flow Tube Rows

in Each Windows (Ncw)
(73)
1 0:14(Res )a4
b3
(74)
1 0:14(Res )b4
Ncw
The coefcients a1, a2, a3, a4, b1, b2, b3 and b4 can be
obtained on Table 4 and are related to the Reynolds
number on the shell side and to the tube arrangement. The
coefcients a1, a2, b1 and b2 assume different values
depending on the Reynolds number as well as the tube
arrangement. The coefcients a3, a4, b3 and b4 assume
different values depending on the tube arrangement. This
can be expressed as
XX
r
(75)
rearr a2
yr;s
Ar;s a2
(76)

ls Ds Dotl
Fsbp
(77)
Ssb
yrarr Aar 4
a4
dsb
(78)
XX
r
(79)
earr b2
yr;s
Ar;s b2
(80)
(90)

31 0004(Ds 1000)
1000
Angle values are in radians.
Tube-to-Baffle Leakage Area for One

Bafflese (Stb)
yrarr Abr 3
b3
(81)
Stb 0:0006223dexNt (1 Fc); m2

Ds dsb
2lc
p arccos 1
2
Ds
XX
X
rearr b1
yr;s
Ar;s b1
(89)
where
Sm
Shell-to-Baffle Leakage Area for One

Bafflese (Ssb)
yrarr Aar 3 a3
(88)
Fraction of Cross-Flow Area Available for

Bypass Flow (Fsbp)
XX
X
rearr a1
yr;s
Ar;s a1
0:8lc
pp
yrarr Abr 4
b4
(91)
(82)
Area for Flow Through Window (Sw)

It is given by the difference between the gross window area
(Swg) and the window area occupied by tubes (Swt):
Number of Baffles (Nb)

Nb
L
1
ls
(83)
Sw Swg Swt
(92)
where
Number of Tube Rows Crossed by the Ideal

Cross Flow (Nc)
Nc
Ds 1 2(lc =Ds )
pp
Swg
(Ds )2
24
s 3

l
l
lc 2 5
c
c
1 12
4arccos 12
12
Ds
Ds
Ds
(84)
lc is the bafe cut. The most used value is
(93)
lc 0:25Ds
(85)
and
Swt (Nt =8)(1Fc)p(Ds )2
Fraction of Total Tubes in Cross Flow (Fc)

"
Fc
1
p 2l sinarccos (l ) 2 arccos (l )
p
(86)
Shell-Side Heat Transfer Coefficient for an Ideal

Tube Bank (hoi)
where
Ds 2lc
l
Dotl
(94)
(87)
hoi

2=3
ji Cps ms
ks
Sm
Cps ms
(95)

Correction Factor for Baffle Configuration
Effects (Jc)
Jc Fc 0:54(1 Fc)
0:345
Pressure Drop Across the Shell-Side (DPs)

(96)

Ncw
:Rb
DPs 2:DPbi : 1
Nc
(Nb 1):DPbi :Rb :Rl Nb:DPwi :Rl
Correction Factor for Baffle-Leakage Effects (Jl )

Ssb Stb
Jl a (1 a) exp 2:2
Sm
where

a 0:44 1
Ssb
Ssb Stb
1429
(97)
(106)
This value must respect the pressure drop limit, xed before
the design:
DPs DPs design
(107)

(98)
Tube-Side Reynolds Number (Ret)

Ret
Correction Factor for Bundle-Bypassing

Effects (Jb)
4:mt :Ntp
p:din:mt :Nt
(108)
Friction Factor for the Tube-Side ( flt)
Jb exp (0:3833Fsbp)
(99)
Assuming that very laminar ow is neglected (Res , 100), it

is not necessary to use the correction factor for adverse
temperature gradient build up at low Reynolds number.

1
0:271
0:9
p 4 log
(7=Ret )
dex
flt
(109)
where 1 is the roughness in mm.
Prandtl Number for the Tube-Side (Prt)

Prt
Shell-Side Heat Transfer Coeficient (hs)

hs hoi JcJlJb
(100)
Pressure Drop for an Ideal Cross-Flow

Section (DPbi)
DPbi
2:fls :Nc:(ms )2
rs :Sm2
m t :Cpt
kt
(110)
Nusselt Number for Tube-Side (Nut) Considering

the Turbulent Flow (Ret > 104)
Nut 0:027:(Ret )0;8 :(Prt )1=3
(101)
(111)
Tube-Side Heat Transfer Coeffcient (ht)

ht
Nut :kt din

:
din dex
(112)
Pressure Drop for an Ideal Window Section (DPwi)

DPwi (2 0:6Ncw)
(ms )2
2:Sw:rs :Sm
vt
Correction Factor for the Effect of Baffle Leakage

on Pressure Drop (Rl )
"

Rl exp 1:33 1
Ssb
Ssb Stb
Stb Ssb
Sm
k #
where
Ssb
0:8
Ssb Stb
(104)
Correction Factor for Bundle Bypass (Rb)

Rb exp1:3456Fsbp
Ret :mt
rt :din
(105)
(113)
The velocity limits are

1 vt 3;
(103)

k 0:15 1
Tube-Side Velocity (vt)
(102)
vt in m s1
(114)
Tube-Side Pressure Drop (Including Head

Pressure Drop)
!
2:flt :Ntp :L:(vt )2
2
DPt rt :
1:25:Ntp :(vt ) :NS
din
(115)
This value must respect the pressure drop limit, xed before
the design:
DPt DPt design
(116)
1430

Heat Exchanged
where
The exchange duty can be a data of the problem if all the

temperatures, heat capacities and ow rates are known. If
one of these parameters is unknown, the heat duty is an
optimization variable and can be written as
Q ms :Cps (Tenh Tsaih )s or
Q ms :Cps (Tsaic Tenc )s
Q mt :Cpt (Tenh Tsaih )t or
Q mt :Cpt (Tsaic Tenc )t
The following disjunction, proposed by Mizutani et al.

(2003) is used:
(117:a)
3 2
3 2
3
R 0:99
R 1:01
R 0:99
6
7 6
7 6
7
4 R1 R
5 _ 4 R 1:01 5 _ 4 R2 R
5
Ft f2 (R; S)
Ft f1 (R1 ; S)
Ft f1 (R2 ; S)
(117:b)
That is introduced in the MINLP problem by using the big-M
constraints:
(118)
Area Nt :p:dex :L:NS

LMTD:
t1 Touth Tinc
t2 Tinh Toutc
(119)
(120)
Chen (1987) LMDT approximation is used:

LMTD t1 t2 (t1 t2 )=21=3
(121)
Correction Factor for the LMTD (Ft)

For the Ft determination, the Blackwell and Haydu (1981)
proposition is used:
Tinh Touth
Toutc Tinc
(122)
Toutc Tinc
S
Tinh Tinc
(123)
p!
R2 1
Ft f1 (R; S)
R1
h
i
ln (1 Px1 )=(1 R:Px1 )
"
p #
2=Px1 1 R R2 1
p
ln
2=Px1 1 R R2 1
(124)
where

(127)
Heat exchange area:
Px2 P=(NS NS:S P)
R1 R M(1 yft1 )
(128)
yft1 )
(129)
R1 R M(1
R 0:99 M(1 yft1 )
(130)
Ft f1 (R; S) M(1 yft1 )
(131)
Ft f1 (R; S) M(1 yft1 )
(132)
R 0:99 M(1 yft2 )
(133)
R 1:01 M(1 yft2 )
(134)
Ft f2 (R; S) M(1 yft2 )
(135)
yft2 )
(136)
Ft f2 (R; S) M(1
R 1:01 M(1 yft3 )
(137)
R2 R M(1 yft3 )
(138)
R2 R M(1 yft3 )
(139)
Ft f1 (R; S) M(1 yft3 )
(140)
Ft f1 (R; S) M(1 yft3 )
(141)
yft1 yft2 yft3 1
(142)
According to Kern (1950), practical values of Ft must be

greater than 0.75. This constraint must be aggregated to
the model:
Ft 0:75
1=NS
R:S 1
S1

R:S 1 1=NS
R
S1
(143)
1
Px1
(125)
Dirty Overall Heat Transfer Coeffcient (Ud)

Ud
NS is the number of shells, or, if R 1,
Q
Area:LMTD
Clean Overall Heat Transfer Coeffcient (Uc)
Ft f2 (R; S)

p
R2 1=(1 Px )
p #
2=Px1 1 R R2 1
p
ln
2=Px1 1 R R2 1
"
Px2
(144)
(126)
Uc
1

dex
rin :dex dex : ln (dex =di n)
1
rout
din :ht
din
2:ktube
hs
(145)
1431
Table 5. Example 1 data.

Stream
Tin (K)
Tout (K)
m (kg s21)
m (kg ms21)
r (kg m23)
Cp (J kg21K)
K (W mK21)
rd (W mK21)
Kerosene
Crude oil
371.15
288.15
338.15
298.15
14.9
31.58
.00023
.00100
777
998
2684
4180
0.11
0.60
1.5e-4
1.5e-4
Fouling Factor Calculation (rd)

rd
Uc Ud
Uc :Ud
(146)
This value must respect the fouling heat exchanger limit,

xed before the design:
rd rd design
(147)
For uids with high viscosity, like the petroleum fractions,

the wall viscosity corrections could be included in the
model, both on the tube and the shell sides, for heat transfer
coefcients as well as friction factors and pressure drops calculations, since the viscosity as temperature dependence is
available. If available, the tubes temperature could be calculated and the viscosity estimated in term of these temperatures. For non-viscous uids, however, these correction
factors can be neglected.
The constraints coming from the Bell-Delaware Method
(Details can be seen in Taborek (1983).
COMPUTATIONAL ASPECTS
It is worth mentioning some important computational
aspects related with the solution of the previous MINLP
model. The proposed model is a non-linear non-convex problem. For solving this kind of problems there are available
different algorithms. The most important are decomposition
algorithms (Outer ApproximationDuran and Grossmann,
1986; Generalized Benders DecompositionGeoffrion,
1972) that iterate between a NLP problem, with a xed conguration of integer variables and a Master (MILP) problem
that predicts a new combination of integer variables, and
Branch and Bound (Leyffer, 2001).
Usually the decomposition algorithms performs better in
problems in which the combinatorial part of the model is
important and the branch and bound algorithms tends to be
less sensitive to non-convexities, although, of course these
are general trends and the particular behaviour is problem
dependent.
Another weakness of the decomposition algorithms
appears in MINLP problems in which there are an important
number of infeasible combinations of integer solutions. In
these cases, the Master problem could predict infeasible
combinations of integer variables and then the corresponding NLP problem is infeasible. This problem can be solved
by different strategies: (1) add a binary cut that avoid the
infeasible combination, and continue with the search. However, if the number of potential infeasible solutions is large
the performance use to be very poor with a large number
of iterations. (2) Adding slack variables and a penalty in
the objective function, in order to force feasibility. However,
in our case the performance is again very poor. The
algorithm usually stops in a solution in which the slacks
are activea local optima of the modied problem, but an
infeasible solution of the original one. (3) Solving a

feasibility problem each time a non-feasible solution is
obtained in order to get valid linearizations for the
Master, but once again the performance in the problem is
very poor.
Fortunately, the weakness of the decomposition algorithms
is the strengthens of branch and bound algorithms. The large
number of potential infeasible solutions helps to prune the tree
search and the important combinatorial nature of the problem
is in someway mitigated. Using SBB under GAMS the model
can be solved in some seconds of CPU time. This model is
a clear example of how selecting the correct tool is important
in order to get a reliable solution in an efcient way.
EXAMPLES
Three examples were chosen to apply the developed
model, in designing optimal heat exchangers.
Example 1
The rst example was extracted from Shenoy (1995). In this
case, there is no available area and pumping cost data, and
the objective function will consist in the heat exchange area
minimization. Temperature and ow rate data as well as
Table 6. Results for Example 1.

Present paper
Present paper (xing hot uid
(not xing uids
on the shell
Shenoy (1995)
allocation)
side)
Area (m2)
Q (kW)
Ds (m)
Dotl (m)
Nt
Nb
ls (m)
Ntp
dex (mm)
din (mm)
L (m)
pt (mm)
ht (W m22K)
hs (W m22K)
Ud (W m22K)
Uc (W m22K)
DPt (kPa)
DPs (kPa)
rd (m28C W21)
NS
Ft
DTML (K)
arr
vt (ms21)
vs (ms21)
Hot uid
allocation
28.40
1320
0.549
0.516
368
6
0.192
6
19.10
15.40
1.286
25.40
8649.6
1364.5
776
1000.7
42.00
3.60
4.1e-3
1
0.9
88.60
Square
Shell
28.31
1320
0.438
0.406
194
6
0.105
4
19.05
17.00
2.438
25.40
2759.840
3831.382
779.068
1017.877
26.915
7.00
3.01e-4
1
0.9
88.56
Triangular
1.827
0.935
Tube
38.52
1320
0.533
0.489
264
19
0.122
2
19.05
17.00
2.438
25.40
4087.058
1308.363
572.510
712.422
7.706
7.00
3.43e-4
1
0.9
88.56
Square
1.108
1.162
Shell
1432

Stream
Tin (K)
Tout (K)
m (kg s21)
m (kg ms21)
r (kg m23)
Cp (J kg21K)
k (W mK21)
DP (kPa)
rd (W mK21)
1
2
368.15
298.15
313.75
313.15
27.78
68.88
3.4e-4
8.0e-4
750
995
2840
4200
0.19
0.59
68.95
68.95
1.7e-4
1.7e-4
acos t 123, bcos t 0.59, ccos t 1.31.
uids physical properties and limits for pressure drop and fouling are in Table 5. It is assumed also that the tube thermal conductivity is 50 W mK21 and the roughness factor is 0.0000457.
Pressure drop limits are 42 kPa for the tube-side and 7 kPa for
the shell-side. A dirt resistance factor of 0.00015 m2 K W21
should be provided on each side.
With these uids temperatures the LMTD correction factor
will be greater than 0.75 and one shell is necessary to satisfy
the thermal balance.
Table 6 presents the heat exchanger conguration of
Shenoy (1995) and the designed equipment, by using the
proposed MINLP model. In Shenoy (1995) the author uses
three different methods for the heat exchanger design; the
method of Kern (1950), the method of Bell Delaware
(Taborek, 1983) and the rapid design algorithm developed
in the papers of Polley et al. (1990), Polley and Panjeh
Shah (1991), Jegede and Polley (1992) and Panjeh Shah
(1992) that xes the pressure drop in both, tube-side and
shell-side before the design. The author xed the cold uid
allocation on the tube-side because of its fouling tendency,
greater than the hot uid. Also some mechanical parameters
as the tube outlet and inlet diameters and the tube pitch are
xed. The heat transfer area obtained is 28.4 m2. The other
heat exchanger parameters are presented in Table 6 as
well as the results obtained in present paper with the proposed MINLP model, where two situations were studied,
xing and not xing the uids allocation. It is necessary to
say that Shenoy (1995) does not take in account the standards of TEMA. According to Smith (2005), this type of
approach provides just a preliminary specication for the
equipment. The nal heat exchanger will be constrained to
standard parameters, as tube lengths, tube layouts and
shell size. This preliminary design must be adjusted to
meet the standard specications. For example, the tube
length used is 1.286 m and the minimum tube length recommended by TEMA is 8 ft or 2.438 m. If the TEMA recommended value were used, the heat transfer area would
be at least 53 m2.
If the uids allocation is not previously dened, as commented before, the MINLP formulation will nd an optimum
for the area value in 28.31 m2, with the hot uid in the tube
side and in a triangular arrangement. The shell diameter
would be 0.438 m and the number of tubes 194. Although
with a higher tube length, the heat exchanger would have a
smaller diameter. Fouling and shell side pressure drops are
very close to the xed limits.
If the hot uid is previously allocated on the shell side,
because of the cold uid fouling tendency, the MINLP formulation following the TEMA standards will nd the minimum
area equal to 38.52 m2. It must be taken into account that
when compared with the Shenoy (1995) value that would
be obtained with the same tube length of 2.438 m (approximately 53 m2), the area would be smaller, as well as the
shell diameter and the number of tubes.
Example 2
As previously commented, the objective function in the
model can be the area minimization or a cost function.
Some rigorous parameters (usually constants) can be aggregated to the cost equation, considering mixed materials of
construction, pressure ratings and different types of exchangers, as proposed in Hall et al. (1990).
The second example studied in the present paper was
extracted from Mizutani et al. (2003). In this case, the authors
proposed an objective function composed by the sum of area
and pumping cost. The pumping cost is given by the
equation:
Pcos t

DPt :mt DPs :ms
ccos t :
rt
rs
(148)
The objective function to be minimized is the total annual

cost, given by the equation:
Min total annual cost acos t (Area)bcos t Pcos t
(149)
Mizutani et al.
(2003)
Total annual
cost ($/year)
Area cost ($/
year)
Pumping cost
($/year)
Area (m2)
Q (kW)
Ds (m)
Dotl (m)
Nt
Nb
ls (m)
Ntp
dex (mm)
din (mm)
L (m)
ht (W m228C)
hs (W m228C)
Ud (W m228C)
Uc (W m228C)
DPt (kPa)
DPs (kPa)
rd (m228C W21)
vt (m s21)
vs (m s21)
NS
arr
Hot uid
allocation
Present paper
Present paper (xing hot uid
(not xing uids
on the shell
allocation)
side)
5250.00
5028.29
5191.47
2826.00
3495.36
3663.23
2424.00
1532.93
1528.24
202.00
4339
0.687
0.672
832
8
0.542
2
15.90
12.60
4.88
6480.00
1829.00
860
22.676
7.494
Square
Shell
264.634
4339
1.067
1.022
680
7
0.610
8
25.04
23.00
4.88
1986.49
3240.48
655.298
826.687
23.312
4.431
3.16e-4
1.058
0.500
1
Square
Tube
286.15
4339
0.838
0.796
713
18
0.353
2
19.05
16.00
6.71
4186.21
1516.52
606.019
758.664
13.404
6.445
3.32e-4
1.003
0.500
1
Square
Shell
1433

Stream
Tin (K)
Tout (K)
m (kg s21)
m (kg ms21)
r (kg m23)
Cp (J kg21K)
K (W mK21)
rd (W mK21)
Hot
Cold
483.15
324.81
377.59
355.37
19.15
75.22
0.00012
0.00029
789.72
820.12
2428.34
2135.3
0.106
0.123
3.5e-4
3.5e-4
Table 7 presents costs, temperature and owrate data as

well as uids physical properties. Also known is the tube thermal conductivity, 50 W mK21. As both uids are in the liquid
phase, pressure drop limits are xed to 68.95 kPa, as
suggested by Kern (1950). As in Example 1, a dirt resistance
factor of 0.00015 m2 K W21 should be provided on each side.
Table 8 presents a comparison between the problem
solved with the Mizutani et al. (2003) model and the proposed
model in this paper. Again, two situations were studied, xing
and not xing the uids allocation. In both cases, the annual
cost is smaller than the value obtained in Mizutani et al.
(2003), even with greater heat transfer area. It is because
of the use of non-standard parameters, as the tube external
diameter and number of tubes. If the nal results were
adjusted to the TEMA standards (the number of tubes
would be 902, with dex 19.05 mm and Ntp 2 for square
arrangement) the area should be approximately 264 m2.
However, the pressure drops would increase the annual
cost. Using the MINLP proposed in the present paper, even
xing the hot uid in the shell side, the value of the objective
function is smaller.
Analysing the cost function sensibility for the objective
function studied, two signicant aspects must be considered,
the area cost and the pumping cost. In the case studied the
proposed MINLP model presents an area value greater
(264.15 and 286.15 m2 versus 202.00 m2) but the global
cost is lower than the value obtained by the Mizutani et al.
(2003) model (5250.00 $/year versus 5028.29 $/year and
5191.49 $/year, respectively). It is because of the pumping
Area (m )
Q (kW)
Ds (m)
Dotl (m)
Nt
Nb
ls (m)
Ntp
dex (mm)
din (mm)
L (m)
ht (W m228C)
hs (W m228C)
U (W m228C)
DPt (kPa)
DPs (kPa)
NS
arr
Hot uid
allocation
Serna and
Jimenez
Method of Bell
algorithm
(Serna and
(Serna and
Jimenez, 2004) Jimenez, 2004)
Present paper
167.6
4909.1
0.762
0.715
546
15
0.406
6
19.05
14.83
5.28
1266.22
1260.54
375.3
78.805
83.631
1
Triangular
Tube
148.56
4909.1
0.737
0.659
509
15
0.305
6
19.05
17.00
4.88
1174.36
928.61
425.1
76.738
43.690
1
Triangular
Tube
163.97
4909.1
0.770
0.715
528
18
0.406
6
19.05
14.83
5.28
1270.14
1372.80
385.1
78.805
83.630
1
Triangular
Tube
costs (2424.00 $/year versus 1532.93 $/year and 1528.24

$/year, respectively).
Obviously, if the results obtained by Mizutani et al. (2003)
for the heat exchanger conguration (number of tubes, tube
length, outlet and inlet tube diameters, shell diameter, tube
bundle diameter, number of tube passes, number of shells
and bafe spacing) are xed the model will nd the same
values for the annual cost (area and pumping costs), area,
individual and overall heat transfer coefcients and pressure
drops as the authors found. It means that it represents a local
optimum because of the other better solutions, even when
the uids allocation is previously xed.
Example 3
The third example was extracted from Serna and Jimenez
(2004). The authors presented an efcient and robust algorithm for the rigorous design of shell and tube heat exchangers based on the Bell-Delaware method. The algorithm
makes complete use of the maximum allowable pressure
drops of both streams, relating exchanger area to the lm
coefcients and allowable pressure drops.
In the case studied the objective function will consist in the
heat exchange area minimization. Temperature and ow rate
data as well as uids physical properties and limits for
pressure drop and fouling are in Table 9. It is assumed also
that the tube thermal conductivity is 45 W mK21. Pressure
drop limits are 78.805 kPa for the tube-side and 83.631 kPa
for the shell-side. A dirt resistance factor of
0.00035 m2 K W21 should be provided on each side.
With these uids temperatures the LMTD correction factor
will be 0.9165 (greater than 0.75) and one shell is necessary
to satisfy the thermal balance.
Table 10 shows three heat exchanger congurations. The
rst column presents a pure Bell-Delaware method application with a bafe cut of 22%. The second one presented
the Serna and Jimenez (2004) algorithm, and the third
column present the designed equipment by using the proposed MINLP model. The heat transfer areas obtained are
172.62 m2, 163.97 m2 and 148.46 m2.
The other heat exchanger parameters are also presented
in Table 10. It is interesting to note that the two rst columns
in the table do not take in account the standards of TEMA.
Table 11. Summary of solver results.

Example 1
Equations
Continuous
variables
Discrete
variables
CPU time a
Pentium IV
1 GHz (s)
Example 2
Example 3
166
713
157
706
166
713
53
602
53
0.251
0.561
0.275
1434
As the nal heat exchanger will be restricted to standard parameters, as tube lengths, tube layouts and shell size, this
preliminary design must be adjusted to meet the standard
specications. The tube length used is 5.283 m and it is not
a value recommended by TEMA, according to equations
(51) and (52).
The smaller heat transfer area can be explained by the
use of the different tube length as well as the shell diameter,
smaller than the used by the authors (0.762, 0.770 and
0.737 m, respectively), that affects the number of tubes
(546, 528 and 509, respectively) and in the tube bundle diameters (0.715, 0.725 and 0.659 m, respectively).
The three examples were solved with GAMS, using the
solver SBB, and Table 11 shows a summary of the solver
results. As can be seen, CPU time is not high. As pointed
in the Computational Aspects section, rstly it is necessary
to choose the correct tool to solve the problem. For this
type of problem studied in the present paper, the solver
SBB under GAMS was the better tool to solve the problem.
To set a good starting point it is necessary to give all the
possible exibility in the lower and upper variables limits,
prior to solve the model, i.e., it is important to x very
lower low bounds and very higher upper limits to the
most inuenced variables, as the Reynolds number, for
example.
CONCLUSIONS
In the present paper a model for the optimal design of heat
exchangers based on GDP was proposed. The optimisation
model is a MINLP, following rigorously the Standards of
TEMA. Bell-Delaware method was used to calculate the
shell-side variables. The model was developed for turbulent
ow on the shell side using a bafe cut of 25% but the
model can consider other values of bafe cuts.
The model calculates the best shell and tube heat exchanger to a given set of temperatures, ow rates and uids physical properties. The major contribution of this paper is that all
the calculated heat exchanger variables are in accordance
with TEMA standards, shell diameter, outlet tube bundle
diameter, tube arrangement, tube length, tube pitch, internal
and external tube diameters, number of bafes, bafe spacing, number of tube passes, number of shells and number
of tubes. It avoids heat exchanger parameters adjustment
after the design task. A tube counting table was proposed
and the use of DGP makes the optimisation task not too
hard, avoiding non linearities in the model. The problem
was solved with GAMS, using the solver SBB. During the
solution of the model, the major problems were found in the
variables limits initialisation. Two examples were solved to
test the model applicability. In the rst example the objective
was the heat exchange area minimization and in the second
one the objective function considered area and pumping
expenses in the annual cost minimization.
In the studied examples comparisons were done to
Shenoy (1995), Mizutani et al. (2003) and Serna and Jimenez
(2003). Having a larger eld of TEMA heat exchanger possibilities, the present model achieved more realistic results
than the results obtained in the literature. Besides, the task
of heat exchanger parameters adjustment to the standard
TEMA values is avoided with the proposed MINLP formulation proposition.
NOMENCLATURE
A
acost
arr
a1, a2, a3
and a4
BWG
b1, b2, b3
and b4
ccost
Cp
dex
din
Dotl
Ds
Fc
Fsbp
s
t
Ft
GDP
hoi
hs
ht
Jb
Jc
ji
Jl
L
lc
LMTD
ls
m
Nb
Nc
Ncw
NS
Nt
Ntp
Nu
Pcost
pn
pp
Pr
pt
Q
Re
Rb
rd
Rl
Sm
Ssb
Stb
Sw
Swg
Swt
T
Uc
Ud
vt
y arr
y bwg
y dex
yf
yl
y ls
ynt
heat exchange area

rea cost constant
tube arrangement
empirical coefcients
Birmingham Wire Gauge
empirical coefcients
pumping cost constant
heat capacity
tube outside diameter
tube inside diameter
tube bundle diameter
shell external diameter
fraction of total tubes in cross-ow
fraction of cross-ow area available for bypass ow
shell-side Fanning factor
tube-side Fanning factor
correction factor of LMTD
generalized disjunctive programming
shell-side heat transfer coefcient for an ideal tube
bank
shell-side lm coefcient
tube-side lm coefcient
correction factor for bundle-bypassing effects
correction factor for bafe conguration effects
Colburn factor
correction factor for bafe-leakage effects
tube length
bafes cut
log mean temperature difference
bafe spacing
mass ow rate
number of bafes
number of tube rows crossed in one cross-ow
section
number of tube columns effectively crossed in each
window
number of shells
number of tubes
number of tube passes
number of Nusselt
pumping cost
tube pitch normal to ow
tube pitch parallel to ow
number of Prandtl
tube pitch
heat duty
number of Reynolds
pressure drop correction factor for bundlebypassing effects
fouling factor
pressure drop correction factor for bafe-leakage
effects
reference normal area for shell-side ow
shell-to-bafe leakage area
tube-to-bafe leakage area for one bafe
area ow thought the window
gross window area
window area occupied by tubes
temperature
clean overall heat transfer coefcient
dirty overall heat transfer coefcient
tube-side uid velocity
binary variable which denes tube pattern
arrangement
binary variable which denes internal tube
diameter
binary variable which denes external tube
diameter
binary variable which denes the uid allocation
binary variable which denes the tube length
binary variable which denes the bafe spacing
binary variable which denes the variables of
Table 1

y res
y rearr
e
DP
DPbi
DPwi
k
m
r
binary variable which denes the shell-side

Reynolds number
binary variable which represents y res and y arr
roughness
pressure drop
shell-side pressure drop for ideal cross-ow
pressure drop for the window
thermal conductivity
viscosity
density
Index
h
c
s
t
hot uid
cold uid
shell-side
tube-side
REFERENCES
Blackwell, W.W. and Haydu, L., 1981, Calculating the correct LMDT
in shell-and-tube heat exchangers, Chemical Engineering, October;
101106.
Chen, J.J., 1987, Letter to the editor: Comments on improvement on
a replacement for the logarithmic mean, Chem Eng Sci, 42: 2488
2489.
Duran, M.A. and Grossmann, I.E., 1986, An outer approximation
algorithm for a class of mixed-integer nonlinear programs, Math
Prog, 36: 307 339.
Geoffrion, A.M., 1972, Generalized benders decomposition, J Optim
Theory and Appl, 10(4): 237 260.
Hall, S.G., Ahmad, S. and Smith, R., 1990, Capital cost targets for
heat exchanger networks comprising mixed materials of construction, pressure ratings and exchanger types, Computers and
Chemical Engineering, 14(3): pp. 319 335.
Jegede, F.O. and Polley, G.T., 1992, Optimum heat exchanger
design, Trans Inst Chem Eng, 70(A2): 133141.
Kern, D.Q., 1950, Process Heat Transfer (McGraw Hill, New York,
USA).
Leyffer, S., 2001, Integrating SQP and branch-and-bound for mixed
integer nonlinear programming, Comp Optim Appl, 18: 295309.
Mizutani, F.T., Pessoa, F.L.P., Queiroz, E.M., Hauan, S. and
Grossmann, I.E., 2003, Mathematical programming model for
1435
heat exchanger network synthesis including detailed heat eschanger designs. 1. Shell-and-tube heat exchanger design, Ind Eng
Chem Res, 42: 40094018.
Panjeh Shahi, M.H., 1992, Pressure drop consideration in process
integration, PhD thesis, UMIST, UK.
Perry, R.H. and Green, D.W., 1997, Perrys Chemical Engineering
Handbook, 7th edition (McGraw Hill, USA).
Polley, G.T. and Panjeh Shah, M.H.M., 1991, Interfacing heat
exchanger network synthesis and detailed heat exchanger
design, Trans Inst Chem Eng, 69: 445 447.
Polley, G.T., Panjeh Shah, M.H.M. and Jegede, F.O., 1990, Pressure
drop considerations in the retrot of heat exchanger networks,
Trans Inst Chem Eng, 68: 211 220.
Ravagnani, M.A.S.S., 1994, Projeto e otimizacao de redes de trocadores de calor, PhD thesis, FEQ-UNICAMP, Campinas, Brazil.
Ravagnani, M.A.S.S., Silva, A.P. and Andrade, A.L., 2003, Detailed
equipment design in heat exchanger networks synthesis and
optimization, Applied Thermal Analysis, 23: 141151.
Serna, M. and Jimenez, A., 2004, An efcient method for the design
of shell and tube heat exchangers, Heat Transfer Engineering,
25(2): 5 16.
Serna, M. and Jimenez, A., 2005, A compact formulation of the
Bell-Delaware method for heat exchanger design and optimization,
Chem Eng Res Des, 83 (A5): 539550.
Shenoy, U.V., 1995, Heat Exchanger Network SynthesisProcess
Optimization by Energy and Resource Analysis (Gulf Publishing
Company, USA).
Smith, R., 2005, Chemical Process Design and Integration (Wiley,
UK).
Taborek, J., 1983, Shell-and-tube heat exchangers, Section 3.3, Heat
Exchanger Design Handbook (Hemisphere, USA).
ACKNOWLEDGEMENTS
The authors acknowledge nancial support provided by CNPq
(National Council of Science and Technological Development) and
to the Spanish Ministerio de Educacion y Ciencia under project
CTQ2005-05456.
The manuscript was received 22 November 2006 and accepted for
publication after revision 25 May 2007.

A Minlp Model For The Rigorous Design of Shell and Tube

Загружено:

Сведения о документе

Исходное описание:

Оригинальное название

Авторское право

Доступные форматы

Поделиться этим документом

Поделиться или встроить документ

Параметры публикации

Этот документ был вам полезен?

Это неприемлемый материал?

Авторское право:

Доступные форматы

A Minlp Model For The Rigorous Design of Shell and Tube

Загружено:

Авторское право:

Доступные форматы

A MINLP MODEL FOR THE RIGOROUS DESIGN

OF SHELL AND TUBE HEAT EXCHANGERS

DEQ/PEQ/CTC/UEM, State University of Maringa, Maringa, PR, Brazil.

more realistic and accurate results for the

Shell and tube heat exchangers are the most

Vol 85 (A10) 1423 1435

RAVAGNANI and CABALLERO

and the Bell-Delaware method, proposed an algorithm for

is proposed. Following the TEMA Standards, it allows to

Heat Exchanger Fluids Location