Академический Документы
Профессиональный Документы
Культура Документы
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
1424
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.
Trans IChemE, Part A, Chemical Engineering Research and Design, 2007, 85(A10): 14231435
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)
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
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
Trans IChemE, Part A, Chemical Engineering Research and Design, 2007, 85(A10): 14231435
(25)
1426
arr tri
arr cua
4 pn 0:5pt 5 _ 4 pn pt 5
pp 0:866pt
pp pt
(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
_j y1jbwg
or 1
y1dex
(46)
y1jbwg
1
(47)
y2jbwg 1
(48)
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
(37)
BWG1
BWG
(26)
pp pp pp
(39)
6:096y4l 6:706y5l
(51)
y2l
(52)
y3l
y4l
y5l
j1
9
X
y1jbwg 1
(40)
j1
BWG2
11
X
y2jbwg BWG2j
(53)
(54)
ls Ds
ls Ds =5
(41)
j1
11
X
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
(43)
9
X
y1jbwg :din1
(44)
y2jbwg :din2
(45)
j1
din2
11
X
j1
Trans IChemE, Part A, Chemical Engineering Research and Design, 2007, 85(A10): 14231435
2
3 2 res
3
y3
y2res
y1res
45
35
4
4
_ Res 10
_ Res 10
Res . 104
Res . 103
Res 102
or
(62)
(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
)
(pt dex ):(Dotl dex )
Sm ls: Ds Dotl
pn
(56)
arr
M(1 ysq
)
(pt dex ):(Dotl dex )
Sm ls: Ds Dotl
pn
(57)
Res3
(65)
Res3
X
yrres 1
r
Res
(66)
(67)
Resr
(68)
arr
)
M(1 ysq
(58)
1427
ms :dex
ms :Sm
(59)
(60)
0:5 vs 2;
rearr
1 yrres ysarr yr;s
0
XX
rearr
yr;s 1
(69)
(70)
(71)
(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
Trans IChemE, Part A, Chemical Engineering Research and Design, 2007, 85(A10): 14231435
1428
a
b
a3
(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
yrarr Abr 3
b3
(81)
Ds dsb
2lc
p arccos 1
2
Ds
XX
X
rearr b1
yr;s
Ar;s b1
(89)
where
Sm
yrarr Aar 3 a3
(88)
XX
X
rearr a1
yr;s
Ar;s a1
0:8lc
pp
yrarr Abr 4
b4
(91)
(82)
L
1
ls
(83)
Sw Swg Swt
(92)
where
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)
(93)
lc 0:25Ds
(85)
and
Swt (Nt =8)(1Fc)p(Ds )2
Fc
1
p 2l sinarccos (l ) 2 arccos (l )
p
(86)
where
Ds 2lc
l
Dotl
(94)
(87)
hoi
2=3
ji Cps ms
ks
Sm
Cps ms
(95)
Trans IChemE, Part A, Chemical Engineering Research and Design, 2007, 85(A10): 14231435
0:345
(96)
Ncw
:Rb
DPs 2:DPbi : 1
Nc
(Nb 1):DPbi :Rb :Rl Nb:DPwi :Rl
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)
4:mt :Ntp
p:din:mt :Nt
(108)
Jb exp (0:3833Fsbp)
(99)
1
0:271
0:9
p 4 log
(7=Ret )
dex
flt
(109)
(100)
2:fls :Nc:(ms )2
rs :Sm2
m t :Cpt
kt
(110)
(101)
(111)
(112)
(ms )2
2:Sw:rs :Sm
vt
Ssb
Ssb Stb
Stb Ssb
Sm
k #
where
Ssb
0:8
Ssb Stb
(104)
Ret :mt
rt :din
(105)
(113)
(103)
k 0:15 1
(102)
vt in m s1
(114)
!
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
Trans IChemE, Part A, Chemical Engineering Research and Design, 2007, 85(A10): 14231435
(116)
1430
where
(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)
(119)
(120)
(121)
(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)
R1 R M(1 yft1 )
(128)
yft1 )
(129)
R1 R M(1
(130)
(131)
(132)
(133)
(134)
(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)
(140)
(141)
(142)
1=NS
R:S 1
S1
R:S 1 1=NS
R
S1
(143)
1
Px1
(125)
Q
Area:LMTD
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)
Trans IChemE, Part A, Chemical Engineering Research and Design, 2007, 85(A10): 14231435
1431
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
Uc Ud
Uc :Ud
(146)
(147)
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
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
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
Trans IChemE, Part A, Chemical Engineering Research and Design, 2007, 85(A10): 14231435
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
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)
(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
Trans IChemE, Part A, Chemical Engineering Research and Design, 2007, 85(A10): 14231435
1433
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
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
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.
Example 2
Example 3
166
713
157
706
166
713
53
602
53
0.251
Trans IChemE, Part A, Chemical Engineering Research and Design, 2007, 85(A10): 14231435
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
Trans IChemE, Part A, Chemical Engineering Research and Design, 2007, 85(A10): 14231435
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.
Trans IChemE, Part A, Chemical Engineering Research and Design, 2007, 85(A10): 14231435