You are on page 1of 12

02638762/05/$30.00+0.

00
# 2005 Institution of Chemical Engineers
www.icheme.org/journals Trans IChemE, Part A, May 2005
doi: 10.1205/cherd.03192 Chemical Engineering Research and Design, 83(A5): 539550

A COMPACT FORMULATION OF THE BELL DELAWARE


METHOD FOR HEAT EXCHANGER DESIGN
AND OPTIMIZATION
M. SERNA1 and A. JIMENEZ2
1
Facultad de Ingeniera Qumica, Universidad Michoacana de San Nicolas de Hidalgo, Morelia, Mich. Mexico
2
Departamento de Ingeniera Qumica, Instituto Tecnologico de Celaya, Celaya, Gto. Mexico

A
n analytical expression that relates the pressure drop, the exchanger area and the film
heat transfer coefficient for the shell side of a shell and tube heat exchanger is
reported. The equation has been developed based on the Bell Delaware method,
and may aid significantly in tasks such as heat exchanger design and optimization procedures.
The mathematical approach used for this formulation can be extended to other heat exchanger
systems to provide suitable compact pressure drop relationships. The use of the compact
formulation within design and optimization algorithms is illustrated.

Keywords: heat exchangers; heat exchanger design; Bell Delaware method; Kern method.

INTRODUCTION and hS is the shell-side film coefficient. The constants


KPT and KPS depend on the physical properties and mass
In this paper we develop compact analytical relationships flowrates of the streams, and on geometric data of the
for the pressure drop, heat transfer coefficient and exchan- equipment.
ger area for the shell side of a heat exchanger, taking into Equation (2) was based on the Kern method (Kern,
account the effects of baffle leakage and bypass. Such 1950), which does not take into account the effect of the
effects have been neglected in previous compact relation- leakage and bypass streams of the shell-side flow model
ships. Two examples are given of the application of the proposed by Tinker (1951, 1958). This flow pattern has
new formulations to the design and optimization of heat been shown to give a better representation of the actual
exchangers. performance of a segmentally baffled heat exchanger
Several expressions have been published to relate the (Coulson et al., 1989; Saunders, 1988), and provided the
pressure drop of a stream with its film heat transfer coeffi- basis for the development of the Bell Delaware method
cient and the equipment area for shell and tube heat (Bell, 1963, 1981; Taborek, 1983). Figure 1 shows how
exchangers. For instance, the following relationship applies the model incorporates the different flowstreams within
to turbulent flow through smooth tubes (Cichelli and Brinn, the flow pattern of the shell side of the exchanger.
1956; Polley et al., 1990; Peters and Timmerhaus, 1991; Assuming window flow areas equal to cross-flow areas
Jegede and Polley, 1992) and same baffles spacing in the end zones as in the internal
sections, Polley et al. (1991) developed another relation-
DPT KPT Ao h3:5
T (1) ship for the shell-side pressure drop:

while for the shell side the following relationship has been DPS K1 K2 Ao h2S for Res  100 (3)
proposed (Polley et al., 1990; Jegede and Polley, 1992)
where the constants K1 and K2 depend on the correction
DPS KPS Ao h5:1
S (2) factors and the geometric characteristics of the equipment
(according to the Bell Delaware method), and on the phys-
where Ao is the outside heat transfer surface area, DPT is ical properties and mass flowrate of the stream.
the pressure drop on the tube side, DPS is the pressure Although more accurate, equation (3) is more complex
drop on the shell side, hT is the tube-side film coefficient, than equation (2). Through simulation studies, Polley
et al. (1991) developed the following simpler expression

Correspondence to: Professor A. Jimenez, Departamento de Ingeniera
Qumica, Instituto Technologico de Celaya, 38010, Celaya, Gto. Mexico.
E-mail: arturo@iqcelaya.itc.mx DPS KS Ao (hS )4:412 (4)

539
540 SERNA and JIMENEZ

Figure 1. Flow model for the shell side of the heat exchanger.

In addition to limiting this equation by their geometric The ideal heat transfer coefficient can be calculated from
assumption of equal flow areas, Polley et al. do not
report the dependency of the constant KS. In principle, by fs ks jsi Res (Prs )1=3
rating exchangers with geometric similarity, one could hsi (6)
Dt (103 )
use the above equation form and fit the exponent for
specific heat exchanger geometries; a general relationship, where ks is the thermal conductivity of the shell-side fluid,
however, remains to be explored. Dt is the tube outlet diameter, jsi is the heat transfer factor
We show how an analytical relationship for the shell side for an ideal tube bank, Res and Prs are the shell-side
of an exchanger based on the full Bell Delaware method Reynolds number and Prandtl number, and fs is the
can be developed. The formulation also provides a viscosity correction factor.
generalization of the simple forms reported by equations
(2) and (4).
Pressure Drop
The shell-side pressure drop, DPS , is the addition of the
BELL DELAWARE METHOD pressure drops for the internal cross-flow sections, DPc , the
The basic principles of the Bell Delaware method are window sections, DPw , and the inlet and outlet sections, DPe ,
summarized in this section.
DPS DPc DPw DPe (7)

These terms are individually calculated for an ideal tube


Heat Transfer bundle, and then corrected for leakage and bypass streams.
The basic relationship for the effective average film heat The pressure drop for a cross-flow section of ideal tube
transfer coefficient is given by bundles, DPbi , is calculated from
  
hs hsi Jc Jl Jb Jr Js hsi Jtot (5) 2Ds rs Bc
DPbi 12 fsi vs 2 (8)
fs Lpp 100
where hsi is the heat transfer coefficient for an ideal tube
bundle, and Jc, Jl , Jb, Jr and Js are correction factors for where Ds is the inside shell diameter, Bc is the baffle cut as
the baffle cut, baffle leakage effects, bundle bypass flow, percent of Ds, Lpp is the tube layout pitch parallel to the
laminar flow and unequal baffle spacing in the inlet and flow direction, and fsi is the ideal tube bundle friction
outlet sections, respectively. factor.

Trans IChemE, Part A, Chemical Engineering Research and Design, 2005, 83(A5): 539550
BELL DELAWARE METHOD FOR HEAT EXCHANGER DESIGN 541

For turbulent flow, the pressure drop in an equivalent Heat Transfer Factor and Friction Factor for an Ideal
ideal tube bundle for the baffle window section, DPwi , Tube Bank
depends on the fluid cross-flow velocity, vs, according to
Both heat transfer and friction factors can be expressed
    as a function of the Reynolds number, and therefore of
Sm the shell-side fluid velocity. The heat transfer factor is
DPwi 1 0:3Ntcw r vs 2
Sw s given by
for Res . 100 (9)  rh
ms
jsi ch Res rh ch vs rh (15)
while for laminar flow D t rs
  where rs and ms are the density and the viscosity of the
3 (vs )ms Ntcw Lbc
DPwi 26  10 p fluid, Dt is the tube outlet diameter, and the constants rh
Sw =Sm Ltp  Dt (Dw )2
  and ch are given by
Sm
r (vs )2 for Res , 100 (10)
Sw s rh a2 (16)
 
1:33 a
where Ltp is the tube layout pitch, Ntcw is the number of c h a1 (17)
effective tube rows crossed in one baffle window, Sm is Ltp =Dt
the minimum cross-flow area in the shell-side flow direc-
tion, Sw is the net cross-flow area through one baffle where
window, and Res is the shell-side Reynolds number. a3
For all interior cross-flow sections, the pressure drop is a (18)
calculated as follows: 1 0:14Res a4

The coefficients a1, a2, a3 and a4 depend on the tube layout


DPc DPbi Nb  1Rb Rl (11) angle and on the Reynolds number [see Table 3.3.7.1 from
Taborek (1983)].
where Nb is the number of baffles. The friction factor is calculated from
The combined pressure drop of all the window sections
can be calculated from  rp
ms
fsi cp Res rp cp vs rp (19)
D t rs
DPw Nb DPwi Rl (12)
where the constants rp and cp are given by
The pressure drop in the inlet and outlet sections is
given by
rp b2 (20)
 
DPe DPbi (1 Ntcw =Ntcc )Rb Rs (13) 1:33 b
c p b1 (21)
Ltp =Dt
where Ntcc is the number of effective tube rows crossed in
one crossflow section. where
If one combines equations (7), (11), (12) and (13), the
following relationship is obtained, b3
b (22)
1 0:14Res b4
DPS Nb  1Rb Rl 1 Ntcw =Ntcc Rb Rs DPbi
Nb DPwi Rl (14) Values of b1 to b4 are also given in Taborek (1983).
To arrive at heat transfer coefficient and pressure drop
In these equations, Rl, Rb and Rs are pressure drop cor- equations that depend directly on the velocity, the Reynolds
rection factors due to baffle leakage effects, bypass flow, number is evaluated for a base fluid velocity, vs .
and the entrance and exit sections having a different
baffle spacing than the internal sections.
Heat Transfer Coefficient as a Function of Velocity
By combining equations (6) and (15), one can express
the heat transfer coefficient for an ideal tube bundle as
DEVELOPMENT OF THE COMPACT RELATIONSHIP
 
We take the equations of the Bell Delaware method, fs ch ks Prs 1=3 Dt rs 1rh
hsi vs 1rh (23)
and develop a compact formulation that relates the shell- Dt 103 ms
side pressure drop with the heat exchanger area and the
heat transfer coefficient. We first develop expressions for where Prs (msCps/ks).
the shell-side film coefficient and for pressure drop as a Equations (5) and (23) are combined to obtain the heat
function of the fluid cross-flow velocity. transfer coefficient as a function of the fluid cross-flow

Trans IChemE, Part A, Chemical Engineering Research and Design, 2005, 83(A5): 539550
542 SERNA and JIMENEZ

velocity, The effective length for heat transfer can be related to the
total number of tubes, Ntt, and the outside area through
hs KS1 vs 1rh (24)  
106 Ao
where the constant KS1 is given by Lta (35)
pDt Ntt
 
fs ch ks Prs 1=3 Dt rs 1rh Taborek (1983) reports a useful empirical equation to
KS1 Jtot (25)
103 Dt ms calculate the number of tubes
 
Pressure Drop as a Function of Area and Velocity p Dctl 2  
Ntt 1  cn (36)
4C1 Ltp
Substitution of the terms for DPbi, DPwi and fsi, in
equation (14) provides the following expression for the
where C1 is 0.866 for a triangular tube layout and 1.0 for a
total shell-side pressure drop for turbulent flow:
square tube layout, Dctl is the diameter of the circle through
the centers of the outermost tubes of a bundle, cn is a tube
DPS bKS2 KS3 vs rp cDs Nb 1vs 2 (26) count correction factor for multiple tube pass layouts with
values between 0.013 and 0.234.
where the constants KS2 and KS3 are given by The average cross-flow velocity of the fluid in the shell
  side is calculated from
Sm 1 0:3Ntcw Rl Nb rs
KS2 (27)
Sw Nb 1Ds Qs
vs  
(37)
  106 Lbc Lbb Dctl =Ltpe Ltp  Dt
Nb  1 Ntcc Ntcw
KS3 Rl Rs
Nb 1 Ntcc Nb 1 where Qs is the volumetric throughput, Lbb is the inside
     shell-to-bundle bypass clearance; Ltpe 0.707Ltp for
Bc 2cp Rb rs ms rp
 12 (28) tube layouts of 458 and Ltpe Ltp for tube layouts of 308
100 fs Lpp Dt rs
and 908.
Equations (34) to (37) are combined to give,
To obtain the desired compact relationship for pressure
drop, we first reduce equation (26) to:
Ds Nb 1 KS5 Ao vs (38)
2r0
DPS KS4 Ds Nb 1vs p (29) where the constant KS5 is defined as
To evaluate the constants rp0 and KS4, let us call DP0S to the  
value given by one of the two equations, say equation (26). 4C1 Ltp 2 Ds Nb 1Lbc
KS5  
Then, for a fluid base velocity vs , Dt Qs 1  cn pDctl Nb  1Lbc Lbi Lbo 
  
    Ltp  Dt
DP0S vs DPS vs (30)  Lbb Dctl (39)
  Ltp,eff
@DP0S  @DPS 
(31)
@ vs vs @ vs vs After substitution of equation (38) into equation (29), the
general pressure drop relationship as a function of the
The application of these two conditions provides the fol- cross-flow velocity of the fluid and the outside heat transfer
lowing expressions for the constants rp0 and KS4 (Serna- surface area can be determined as
Gonzalez, 1999)
3  rp0
DPS KS4 KS5 Ao vs (40)
rp
rp0  rp (32)
KS2 =KS3 vs 1
  rp0  r0 r
KS4 KS2 vs KS3 vs p p (33)
Compact Formulation
The product Ds Nb 1 in the right hand side of The combination of equations (24) and (40) provides the
equation (29) must be expressed as a function of the outside desired compact relationship for the shell-side pressure
heat transfer area, Ao, and the cross-flow velocity of the drop for turbulent flow,
fluid.
For straight tube bundles, the effective tube length for
DPS KS Ao hS m (41)
heat transfer is given by

Lta Nb  1Lbc Lbi Lbo (34) where

where Lbc is the central baffle spacing, Lbi is the inlet baffle KS4 KS5
KS (42)
spacing, and Lbo is the outlet baffle spacing. KS1 m

Trans IChemE, Part A, Chemical Engineering Research and Design, 2005, 83(A5): 539550
BELL DELAWARE METHOD FOR HEAT EXCHANGER DESIGN 543

and Kindiv 0 and n 3.5, so that equation (45) simplifies to

3  rp0 1=2
 11=6  
m (43) f4:5
t Dti =1000 mt =1000 Dti
1  rh KT 2:5 7=3 7=6
(53)
2
0:023 gc Qt rt kt Cpt Dt
Since no restriction has been taken to develop equation
With these results we recover equation (1) that was
(41), this compact formulation has the same range of appli-
derived earlier by several authors (Cichelli and Brinn,
cation as the original Bell Delaware method. Thus, the
1956; Polley et al., 1990; Peters and Timmerhaus, 1991;
pressure drop relationship can be used for shell and tube
Jegede and Polley, 1992), which constitutes a limiting
exchangers with different geometries. This compact
case of the general pressure drop relationship presented in
expression can also provide a convenient basis for suitable
this work.
algorithms for the design and optimization of shell and tube
For situations requiring a more accurate prediction of the
heat exchangers.
tubeside heat transfer coefficient than that given by the
Colburn equation, Petukhov equation is recommended
(Ozisik, 1985). To implement this modification, it is
necessary to replace equation (51) by
A COMPACT EXPRESSION FOR THE TUBE SIDE
  
The procedure used in this work can be generalized, as 0:023ft 103 kt Dti rt 0:8
KT1 Prt (54)
shown in Appendix A. Through the application of such a X Dti mt
procedure, for instance, a compact relationship can be
obtained for the case of turbulent flow through the tube- where
side of a shell-and-tube exchanger. Using ch 0.023, 
rh 0.2, p 1/3, cp 0.046, rp 0.2, and the inside 0:5
X 1:07 12:7 Prt2=3  1 0:023=Re0:2
t (55)
tube diameter as the characteristic length (McAdams,
1954), one can obtain the following compact relationship
(Serna-Gonzalez, 1999),
APPLICATIONS
DPT KT Ao hT n (44) Two applications of the compact formulations derived
above are shown. The first one deals with their imple-
with the constants KT and n given by mentation within an efficient design algorithm, while the
 n second one shows how optimization of heat exchangers
KT5 KT4 1 based on economic criteria can be carried out aided by
KT (45)
2gc KT1 the compact formulations. An earlier work by Jegede and
3  r0 Polley (1992) provides an interesting background, since
n (46) they proposed rapid algorithms for the design and optimi-
0:8 zation of shell and tube heat exchangers based on the
where Kern method for the shell-side flow pattern. Their algor-
ithm uses equations (1) and (2) along with the heat exchan-
Dti rt ger design equation:
KT5 (47)  
4Qt Dt
 r 0  r0 0:2 Q 1 Dt
KT4 KT2 vt KT3 vt (48) Ao Rdw (56)
FT DTLM hS Dti hT
  
2Lts mt 0:2
KT3 0:184ft 1 (49) where Ao is the external surface area required for the speci-
Lta Dti rt fied heat duty Q, FT is the temperature correction factor for
Dti multipass heat exchangers, DTLM is the log-mean tempera-
KT2 Kindiv (50)
Lta ture difference, Dt is the outside tube diameter, and Dti is
  the inside tube diameter; Rdw is the combined resistance
0:023kt Prt 1=3 Dti rt 0:8
KT1 (51) of tube wall and fouling factors,
103 Dti mt
 3   
0:2 10 Dt Dt Dt
r0  0:2 (52) Rdw Rds ln Rdt (57)
KT2 =KT3 vt 1 2kw Dti Dti

where vt is a base fluid velocity, Qt is the tube-side volu- where kw is the thermal conductivity for the tube wall
metric throughput, Ltt is the total tube length, Lts is the material, Rds is the shell-side fouling factor, and Rdt is the
tubesheet thickness, and Kindiv is the loss coefficient due tube-side fouling factor.
to contraction at the tube inlets, expansions at the exits, In addition to the limitation of the use of the Kern
and flow reversals in the headers. A recommended value method in the work by Jegede and Polley (1992), the
for Kindiv is 2.5 (Coulson et al., 1989). return losses for the tube-side fluid were neglected in
A special case of equation (44) is when no indivi- their algorithm. The application of the compact formulation
dual pressure losses are taken into account. In this case of the Bell Delaware method provides a more efficient

Trans IChemE, Part A, Chemical Engineering Research and Design, 2005, 83(A5): 539550
544 SERNA and JIMENEZ

algorithm; also, tube-side return losses are incorporated in A detailed description of the algorithm is given in
the algorithm here presented. Appendix B. The following design problem shows the
use of the algorithm.

Application Within a Design Algorithm


In the design procedure one requires the following vari- Example 1
ables for each stream: inlet and outlet temperatures, allow- This case problem was taken from Thomas (1993), who
able pressure drop, mass flowrate, fouling resistance, and also reports a solution based on the conventional design
physical properties such as density, heat capacity, viscosity method. We highlight here only the most relevant charac-
and thermal conductivity. The designer also sets the follow- teristics. A shell and tube exchanger must be designed to
ing geometric parameters of the heat exchanger: outside cool down an oil, 308 API, using cooling water. The oil
and internal tube diameter, tube pitch, tube layout, is the shell-side fluid. Table 1 shows the design data.
number of tube passes, number of sealing trips, tube lane The allowable pressure drops are 10.13 kPa for the tube
partition bypass width, tubesheet thickness, baffle cut, and side and 11.346 kPa for the shell side; these values
clearances. The application of the design algorithm come from the performance evaluation of the original
provides the following items: total flow length of tubes, exchanger.
effective tube length for heat transfer area, number of Two designs were obtained with the algorithm, as
tubes, number of baffles, central baffle spacing, inlet and reported in Table 2, where the solution given by Thomas
outlet baffle spacing, shell diameter, shell and tube side is also shown. Each solution took six iterations to converge.
heat transfer coefficients, and heat transfer area. The performance calculations show that the three designs
As indicated by Jegede and Polley (1992), it is con- make a full use of the specified allowable pressure drops.
venient to formulate a design algorithm that ensures that For this problem, there is a significant difference in the
the pressure drops within the exchanger are equal to their properties of the fluids. The oil viscosity is 3.6 times
allowable (specified) values for both streams, since this higher than that of water, its fouling factor 4.67 higher
condition provides the smallest exchanger for a given and its thermal conductivity almost five times lower. As a
heat duty. The algorithm we have developed provides an consequence, the fouling heat transfer coefficient for the
efficient use of pressure drops for both sides of the shell side is relatively small. The three designs in Table 2
exchanger. reflect the higher resistance to heat transfer of the shell-
The algorithm uses a nested calculation method. In the side fluid.
inner loop, the parameters for the compact pressure drop Design 1 was based on a ratio Sm/Sw 1 and provides a
formulation, or compact parameters, are treated as con- baffle cut of 22.73% and an installed area of 253.544 m2. It
stants. Thus, the combination of equations (41), (44) and is very common that if the design is based on equal flow
(56) leads to areas, Sm Sw, a solution with a fractional number of baf-
" #1=n fles is obtained, 10.2945 in this case. The solution must be
DPT FT DTML =KT Q revised for this aspect. Thus, Design 2 was developed to
hT   1=m 0 obtain 10 baffles, with a baffle cut of 20.8% and an installed
KS DPT =KT DPS hnT Rdw Dt =Dti hT area of 251.717 m2. For such a design the required ratio of
(58) Sm/Sw was 1.1528. It should be noted that in the series of
 
KT DPS hnT 1=m
hS (59)
KS DPT
Table 1. Design data for Example 1.

where hT and hS need to be determined to provide the heat Shellside Tubeside


exchanger configuration.
Flowrate (kg s) 43.6 45.377
The inner loop problem is identical to the problem of Density (kg m3) 820 993
shell and tube heat exchanger design based on the Kern Heat capacity (J kg . K) 2170 4170
method (Jegede and Polley, 1992). The solution to this Viscosity (ctp) 2.45 0.682
problem is quite simple. First, equation (58) is solved Thermal conductivity (W m . K) 0.128 0.63
Inlet temperature (8C) 114 26
numerically for hT. Then, hS and Ao are calculated using Outlet temperature (8C) 66 50
equations (59) and (56), respectively; next, the correspond- Allowable DP (kPa) 11.346 10.13
ing exchanger details can be determined using some of the Fouling factor (K m2 W) 0.0007 0.00015
preceding equations. However, the compact parameters are Tube wall thermal 0
related to the unknown heat exchanger variables, which in conductivity (W K m2)
Heat duty (kW) 4541.4
turn depend on those parameters. Thus, an outer loop is
needed in which the values of the dependent variables of Geometry
the heat exchanger are used to compute new approxi- Outside tube diameter (mm) 19.1
Inside tube diameter (mm) 16.6
mations to the compact parameters; these new approxi- Tube layout (degrees) 90
mations are used in the inner loop of the following Tube pitch (mm) 25.4
iteration to compute another heat exchanger configuration Number of tube passes 4
as a new approximation to the actual solution. The Shell/baffle clearance (mm) 5.72
Tube/baffle clearance (mm) 0.794
algorithm converges when two successive sets of com- Shell/tube bundle clearance (mm) 12.7
pact parameters differ by a sufficiently small amount.

Trans IChemE, Part A, Chemical Engineering Research and Design, 2005, 83(A5): 539550
BELL DELAWARE METHOD FOR HEAT EXCHANGER DESIGN 545

Table 2. Designs obtained for Example 1.

Design 1 Design 2 Thomas (1993)

Geometry
Shell diameter (mm) 1015.71 1014.75 1070
Total flow length of the tubes (mm) 3872.97 3852.14 4880
Baffle cut (%) 22.73 20.8 20
Central baffle spacing (mm) 342.9 349.94 375
Inlet baffle spacing (mm) 342.9 349.94 375
Outlet baffle spacing (mm) 342.9 349.94 375
Number of baffles 10.2945 10 12
Number of tubes 1091 1089 1195
Number of tube passes 4 4 4
Number of shells in series 1 1 1
Installed area (m2) 253.544 251.717 349.922
Performance
Required area (m2) 253.352 251.496 262.815
Shell side Re 36608.44 33768.21 33544.94
Shell side DP (kPa) 11.346 11.346 11.346
Tube side DP (kPa)
DP for straight section (kPa) 7.151 7.136 7.65
DP for ends (kPa) 2.978 2.989 2.483
Total tube side DP (kPa) 10.129 10.128 10.13
hS (W m2 K) 692.091 701.05 657.84
hT (W m2 K) 3775.964 3782.149 3510.477
U (W m2 K) 381.29 384.13 367.58
Sm Sw 1.0004 1.1528 1.22819

problems we have solved, a proper value for the ratio as (Jegede and Polley, 1992):
Sm/Sw seems to be between 0.8 and 1.2.
The solution reported by Thomas (1993) is overdesigned. CTOT KF N(c1 c2 Aco3 ) c4 c5 (NQT DPT )c6
The ratio of installed area to required area is 1.33. Design 2,
on the other hand, shows an efficient design for which the c7 c8 (NQS DPS )c9  (Cpow =h)HY NQT DPT
installed area is practically equal to the required area. (Cpow =h)HY NQS DPS (60)
Two of the major factors for this difference between the
two designs are the tube length and the baffle spacing: where KF is the annualization factor for capital cost, HY is
4880 mm and 375 mm for the design by Thomas, against the plant operation time per annum, c1 to c9 are coefficients
3852.14 mm and 350 mm for Design 2. Therefore, the in capital cost laws for heat exchanger and pumping
pressure drop for the shell-side fluid per unit length of devices, Cpow is the cost per unit of power, h is the
Design 2 is 26.7% higher, which also increases the value efficiency of pumping devices, QT is the volumetric flow
of the shell-side heat transfer coefficient (701.05 W m2 K rate of the tube side stream, QS is the volumetric flow
for Design 2, against 657.84 W m2 K for the design by rate of the shell-side stream and N is the number of
Thomas). The overall heat transfer coefficient is therefore shells in series required for multipass flow heat exchan-
better for Design 2 (384.13 W m2 K as compared to gers. In this case, it is considered that each shell will
367.58 W m2 K for the solution by Thomas). The result is have the same heat transfer area and geometric
that the installed area of the solution by Thomas is 39% arrangement; thus, stream pressure drops are the same
higher than the value given by Design 2. for each shell.
The results show a significant advantage of the design If the density of the fluid is assumed to be constant, the
algorithm here presented over the conventional design power required for each stream equals the product of volu-
method. Through a better use of the allowable pressure metric flow rate and the pressure drop (QT . DPT for the
drops, higher heat transfer coefficients are obtained which tube side and QS . DPS for the shell side).
in turn provide a lower exchanger area. Substitution of the values for DPT and DPS from
equations (41) and (44) into equation (60) gives a new
expression for the total annual cost
Application Within an Optimization Algorithm
CTOT KF bN(c1 c2 Aco3 ) c4 c5 (NKT QT hnT Ao )c6
We now show how the design algorithm can be used
c7 c8 (NKS QS hm c9
S Ao ) c (CT =hT )fU NKT
within a formal optimization method. The problem is find-
ing the design that minimizes the yearly cost of the heat  QT hnT Ao (CS =hS )fU NKS QS hm
S Ao (61)
exchanger (investment plus operating costs). The capital
investment includes the cost of the exchanger and the two The solution to this problem must satisfy the exchanger
pumping devices. The operating cost is related to the cost basic design equation given by equation (56), which relates
of pumping the streams through the exchanger. If one the three independent variables hT, hS and Ao of the objec-
uses power-law cost expressions for heat exchangers and tive function. Incorporating equation (56) into equation
pumping devices, the objective function can be expressed (61), one obtains a suitable form of the objective function

Trans IChemE, Part A, Chemical Engineering Research and Design, 2005, 83(A5): 539550
546 SERNA and JIMENEZ

with only two independent variables (hT and hS), which Table 3. Economic data for Example 2.
aids the optimization procedure. Heat exchanger cost ($) c1 c2(Ao)c3 30000 750(Ao)0.81
   c3 Tubeside pump capital c4 c5(QTDPT)c6 2000 5(QTDPT)0.68
Q cost ($)
CTOT NKF c1 c2 Shellside pump capital c7 c8(QSDPS)c9 2000 5(QSDPS)0.68
NFT DTML cost ($)
   1.2501  1028
1 D t c3 Cost of power ($/J)
 Rdw Plant operation (h/year) 8000
hS Dti hT Pump efficiency 70%
   Annualization factor (year21)
KT QT Q c6 0.322
KF c4 c5
FT DTML
  
1 D t c6 n c6
 Rdw (hT ) verified through a rating procedure with the shell and
hS Dti hT
   tube heat exchanger model based on the Bell Delaware
KS QS Q c9 method (Taborek, 1983). The rating results confirm that
KF c7 c8
FT DTML the optimum heat exchanger satisfies the heat duty require-
   ment with pressure drops in both streams equal to the opti-
1 D t c9 m c 9
 Rdw (hS ) mum ones. The heat exchanger needs 21 baffles (an integer
hS Dti hT
   number) for a baffle cut of 16.75%; thus, the design meets
HY CT KT QT Q 1 Dt important practical considerations.
Rdw hn
hT FT DTML hS Dti hT T For a formal optimality test, the matrix of second partial
   derivatives of the objective function is evaluated at the
HY CS KS QS Q 1 Dt
Rdw hm stationary points hT 5567.9 and hS 1009.17,
hS FT DTML hS Dti hT S
(62)  2 
 @ CTOT @2 CTOT 
 
 @h2 @hT @hS 
To solve the problem using differential calculus, the total  T
 2 
annual cost equation and its first and second partial deriva-  @ CTOT @2 CTOT 
 
tives are treated as continuous functions. Again, a nested  @hS @hT @h2S 
solution procedure is used. In the inner loop, the compact  
 0:222  103 0:2498  103 
parameters are regarded as independent of the heat exchan- 
 
ger configuration. Then, two nonlinear algebraic equations  0:2498  103 0:3088  101 
are obtained after applying the necessary conditions for the
optimality criteria to the nonlinear objective function, 0:6793  105 . 0 (65)

@CTOT
0 (63)
@hT Table 4. Optimal design for Example 2.
@CTOT
0 (64) Geometry
@hS Shell diameter (mm) 898.93
Total flow length of the tubes (mm) 3976.5
A set of local optimum heat transfer coefficients for each Baffle cut (%) 16.75
iteration can be obtained by solving equations (63) and (64) Central baffle spacing (mm) 180.75
simultaneously (e.g., with the Newton Raphson method) Inlet baffle spacing (mm) 180.75
and, if one wishes to conduct a more formal optimality Outlet baffle spacing (mm) 180.75
Number of baffles 21
test, by analysing the second order derivatives of the Number of tubes 838
objective function. Once the values for hT and hS have Number of tube passes 4
been obtained, it is straightforward to calculate the heat Number of shells in series 1
exchanger details and new values of the compact para- Installed area (m2) 200.02
meters in the outer loop; the procedure is applied until con- Performance
vergence is obtained. Notice that for each iteration, the Required area (m2) 200.05
compact parameters must be updated. The iterative Shell side Re 8256.1
Shell side DP (kPa) 61.354
method, outlined in Appendix B, has proven to be quite Tube side DP (kPa)
robust in the applications we have carried out. DP for straight section (kPa) 11.789
DP for ends (kPa) 4.033
Total tube side DP (kPa) 15.822
Example 2 hS (W m2 K) 1009.17
The case study of Example 1 was taken and reformulated hT (W m2 K) 5567.91
as an optimization problem. The process stream properties U (W m2 K) 483.06
Sm Sw 0.8892
and geometric data of Table 1 are supplemented with the
economic parameters given in Table 3. Cost
Table 4 shows the results of the optimization procedure. Annual pump cost ($/year) 1824.22
Annual heat exchanger cost ($/year) 27311.7
This test problem took eight iterations to reach the opti- Power cost ($/year) 2050.68
mum solution. The optimum value of the objective function Total annual cost ($/year) 31,186.6
(minimum cost) was 31 187 $/year. The design was

Trans IChemE, Part A, Chemical Engineering Research and Design, 2005, 83(A5): 539550
BELL DELAWARE METHOD FOR HEAT EXCHANGER DESIGN 547

As the first derivatives of the total annual cost are zero at c1 to c9 coefficients in capital cost laws for equipment
C1 tube count constant for multiple tube pass layouts
optimum shell-side and tube-side heat transfer coefficients
[equation (36)]
given by the algorithm, and the Hessian matrix is positive ch coefficient in heat transfer factor relationship
definite, then the solution satisfies the necessary and suffi- [equation (17)]
cient conditions for a minimum of the objective function. cp coefficient in friction factor relationship
The use of the compact formulation provides an efficient [equation (21)]
Cp fluid specific heat at constant pressure and average
numerical algorithm, with which a detailed configuration of temperature, J (kg8C)
the optimal design of a heat exchanger is obtained. The Cpow cost per unit of power, $/J
effect of various economic, process and geometric CTOT total annual cost, $/year
parameters on the optimum design can be readily analysed Dctl diameter of the circle through the centres of the
with the proposed algorithm. outermost tubes of a bundle, mm.
Ds inside shell diameter, mm
Dti tube inside diameter, mm
Dt tube outside diameter, mm
CONCLUSIONS FT correction factor to logarithmic mean temperature
difference for non-countercurrent flow
A compact formulation to relate the shell-side pressure fsi friction factor for an ideal tube bank [equation (19)]
drop with the exchanger area and the film coefficient g exponent for heat transfer coefficient in general
based on the full Bell Delaware method has been pre- pressure drop relationship [equation (A10)]
sented. This compact formulation retains the simplicity h clean heat transfer coefficient, W (m28C)
and robustness of the type of relationship based on the hsi clean shell-side heat transfer coefficient for an ideal
tube bundle, W (m28C)
Kern method, but provides a more accurate description of HY annual plant operation time, h/year
the flow pattern within the shell of the exchanger. It can jsi heat transfer factor [equation (15)]
be applied to any of the shell side geometries typically Jb bundle bypass correction factor for heat transfer
used in industry. In addition to the derivation of the shell Jc segmental baffle window correction factor for heat
transfer
side compact expression, we have developed a compact Jl baffle leakage correction factor for heat transfer
pressure drop equation for the tube-side stream, which Jr laminar flow heat transfer correction factor
accounts for both straight pressure drops and return Js heat transfer correction factor for unequal baffle
losses. The procedure can be extended to derive analytical, spacing
compact pressure drop equations for other types of heat ks, kt, kw thermal conductivity of shell-side fluid, tube-side
fluid, and tube wall J (m s8C).
exchanger. K constant for general pressure drop relationship
The mathematical approach shown in this work to [equation (A10)]
develop compact expressions provides a more convenient K1 shell-side constant for pressure drop relationship
tool than the procedure of rating and correlating data for [equation (3)]
K2 shell-side constant for pressure drop relationship
exchangers of similar geometry, such as the one used by [equation (3)]
Polley et al. (1991), since such a procedure requires exten- KF annualization factor for capital cost
sive computational efforts to find individual expressions for Kh constant for general heat transfer relationship
the constant KS and a proper value for the exponent of the [equation (A3)]
film coefficient. Kp1 constant for general pressure drop relationship
[equation (A5)]
We have shown how the compact formulations can be Kp2 constant for general pressure drop relationship
used within an efficient design algorithm. The compact par- [equation (A5)]
ameters are the core of the design algorithm. For given Kp3 constant for general pressure drop relationship
values of the compact parameters, simple algebraic [equation (A6)]
Kp4 fluid constant in equation (A9)
equations are easily solved in an inner loop for the shell- KPS shell-side constant for pressure drop relationship
side and tube-side heat transfer coefficients, and the heat [equation (2)]
transfer area. With the newly obtained heat exchanger KPT tube-side constant for pressure drop relationship
configuration, the compact parameters are corrected in the [equation (1)]
outer loop. The inclusion of the compact formulations KS shell-side constant for pressure drop relationship
[equation (41)]
within an optimization algorithm has also been presented. KS1 shell-side fluid constant as defined by equation (25)
We have found a satisfactory performance of the proposed KS2 shell-side fluid constant as defined by equation (27)
algorithms over the entire geometry range of single phase, KS3 shell-side fluid constant as defined by equation (28)
shell and tube heat exchangers. KS4 shell-side fluid constant as defined by equation (33)
KS5 shell-side fluid constant as defined by equation (39)
Given its simple form and general use, the compact for- KT tube-side constant for pressure drop relationship
mulation may find a suitable application in the develop- KT1 tube-side fluid constant as defined by equation (47)
ment of more detailed algorithms for the synthesis of heat KT2 tube-side fluid constant as defined by equation (48)
exchanger networks. KT3 tube-side fluid constant as defined by equation (49)
KT4 tube-side fluid constant as defined by equation (50)
KT5 tube-side fluid constant as defined by equation (51)
NOMENCLATURE L characteristic length
Lbb inside shell-to-bundle bypass clearance (diametral),
a, a1, a2, a3 correlational coefficients for the estimation of the mm
heat transfer factor, jsi Lbc central baffle spacing, mm
A heat transfer surface area, m2 Lbi inlet baffle spacing, mm
Ao heat transfer surface based on outside tube area, m2 Lbo outlet baffle spacing, mm
b, b1, b2, b3 correlational coefficients for the estimation of the Lpp tube layout pitch parallel to fluid flow, mm
friction factor, fsi Lta effective tube length for heat transfer area
BC baffle cut as percent of inside shell diameter calculations, mm

Trans IChemE, Part A, Chemical Engineering Research and Design, 2005, 83(A5): 539550
548 SERNA and JIMENEZ

Ltp tube layout pitch, mm Coulson, J.M., Richardson, J.F. and Sinnott, R.K., 1989, Chemical Engin-
Lts tubesheet thickness, mm eering. An Introduction to Chemical Engineering Design, Chapter 12
Ltt total tube length, mm (Pergamon Press, Singapore).
m exponent for shell-side heat transfer coefficient in Jegede, F.O. and Polley, G.T., 1992, Optimum heat exhanger design, Trans
pressure drop relationship [equation (41)] IChemE Part A, 70: 133.
m shell-side fluid mass velocity, kg (m2 s) Kern, D.Q., 1950, Process Heat Transfer (McGraw-Hill, New York,
n exponent for tube-side heat transfer coefficient in USA).
pressure drop relationship [equation (44)] McAdams, W.H., 1954, Heat Transmission, 3rd edition (McGraw-Hill,
N number of shells New York, USA).
Nb number of baffles Ozisik, M.N., 1985, Heat Transfer. A Basic Approach (McGraw Hill
Ntcc number of tube rows crossed between baffle tips of International Editions, Singapore).
one baffle compartment Peters, M.S. and Timmerhaus, K.D., 1991, Plant Design and Economics
Ntcw number of tube rows crossed in one baffle window for Chemical Engineers, 4th edition (McGraw Hill International Edi-
Ntt total number of tubes tions, Singapore).
Ntw number of tubes in baffle window Polley, G.T., Panjeh Shahi, M.H. and Jegede, F.O., 1990, Pressure drop
Nu Nusselt number for fluid stream considerations in the retrofit of heat exchanger networks, Trans
p exponent for Prandtl number in heat transfer IChemE, Part A, Chem Eng Res Des, 68: 211.
correlation [equation (A2)] Polley, G.T., Panjeh Shahi, M.H. and Picon Nunez, M., 1991, Rapid design
Pr Prandtl number for fluid stream algorithm for shell-and-tube and compact heat exchangers, Trans
Q heat duty, W IChemE, Part A, Chem Eng Res Des, 69: 435.
Q volumetric throughput for fluid stream, m3 s Saunders, E.A.D., 1988, Heat Exchangers: Selection, Design and
rh exponent for Reynolds number in heat transfer Construction (Longman Scientific & Technical, New York, USA).
factor relationship [equation (15)] Serna-Gonzalez, M., 1999, Desarrollo de algoritmos rigurosos para la
rp exponent for Reynolds number in friction factor integracion termica de procesos, PhD thesis, Instituto Tecnologico de
relationship [equation (A4)] Celaya, Celaya, Gto. Mexico.
rp0 exponent for velocity in pressure drop relationship Soumerai, H., 1987, Practical Thermodynamic Tools for Heat Exchanger
[equation (29)] Design Engineers (John Wiley & Sons, New York, USA).
Rb bundle bypass correction factor for pressure drop Taborek, J., 1983, Shell-and-tube exchangers: single-phase flow, in
Rds shell-side fouling factor, K m2 W Schlunder, E.U. (ed). Heat Exchangers Design Handbook, Vol. 3,
Rdw combined resistance of tube wall and fouling factors, Section 3.3 (Hemisphere Publishing Corp., Washington, DC, USA).
K m2 W Thomas, L.C., 1993, Heat TransferProfessional Version (Prentice Hall,
Rdt tube-side fouling factor, K m2 W New Jersey, USA).
Rl baffle leakage correction factor for pressure drop Tinker, T., 1951, Shell side characteristics of segmentally baffled shell-
Rs baffle end zones correction factor for pressure drop and-tube heat exchangers, Parts I, II, IIIGeneral discussion on heat
Re Reynolds number for fluid stream transfer, Proc Inst Mech Engrs (Lond.), 1: 89.
Sm cross-flow area near shell centreline, mm2 Tinker, T., 1958, Shell side characteristics of shell-and-tube heat exchan-
St Stanton number for fluid stream gersa simplified rating system for commercial heat exchangers, ASME
Sw net cross-flow area through one baffle window, mm2 Trans, 80: 36.
v velocity for fluid stream, m s
v base velocity for fluid stream, m s

Greek letters ACKNOWLEDGEMENTS


m viscosity of fluid stream, cP Financial support from CONACyT, Mexico (grant 25970-A) for the
f viscosity correction factor, (m/mw)0.14 development of this project is acknowledged.
h efficiency of pump
r density of fluid stream, kg m3
DP pressure drop for fluid stream, Pa The manuscript was received 14 May 2003 and accepted for publication
DPbi cross-flow pressure drop in an equivalent ideal tube after revision 11 February 2005.
bundle, Pa
DPc total pressure drop in cross flow between baffle APPENDIX A. DEVELOPMENT OF COMPACT
tips, Pa FORMULATIONS
DPe pressure drop in the end zones, Pa
DPw pressure drop in window of a segmentally baffled The derivation shown for the shell side of an exchanger
exchanger, Pa can be extended to other exchangers or geometries to obtain
DPwi baffle window pressure drop in an equivalent ideal
tube bundle, Pa simple pressure drop relationships.
DTLM log-mean temperature difference, 8C The heat transfer factor, j, is expressed in a general
form as
Subscripts
s, S shell-side j ch Rerh (A1)
t, T tube-side
w at wall temperature with constants rh and ch.
The Stanton number is used for the film heat transfer
coefficient,
REFERENCES
Nu
Bell, K.J., 1963, Final report of the cooperative research program on shell j StPr 1p Pr 1p (A2)
and tube heat exchangers, University of Delaware Engineering Exper- Re Pr
imental Station Bulletin No. 5, Newark, Delaware.
Bell, K.J., 1981, Delaware method for shell side design, in S. Kakac, where Nu is the Nusselt number, Pr is the Prandtl number,
A.E., Bergles, F. and Mayinger, F. (eds). Heat Exchangers: Thermal- and p is Prandtl number exponent in heat transfer
Hydraulic Fundamentals and Design (Hemisphere/McGraw-Hill,
Washington, DC).
correlation.
Cichelli, M.T. and Brinn, M.S., 1956, How to design the optimum heat From equations (A1) and (A2), after some rearrange-
exchanger, Chem Engng, 4: 196. ment, the film heat transfer coefficient can be directly

Trans IChemE, Part A, Chemical Engineering Research and Design, 2005, 83(A5): 539550
BELL DELAWARE METHOD FOR HEAT EXCHANGER DESIGN 549

related to the velocity of the fluid: on the shellside. The heat duty Q, the log-mean temperature
difference DTML, and the correction factor FT are calcu-
h Kh v1rh (A3) lated from the specifications of (1).
Step 2. Guess initial values for KT, KS, n and m. A good
where the constant Kh depends on physical properties, a set of initial values can be provided by the Kern method
single characteristic dimension, and ch. (i.e., KS and KT can be estimated with equations (B1) and
The friction factor is related to the Reynolds number, (53), while m 5.109 and n 3.5).
f cp Rerp (A4)    
67:062C1 Ltp  Dt Ltp D1:109
e m1:297
s
KS
where the constants rp and cp are defined on the basis of the 10003:406 gc Dt Qs r2s ks3:406 Cp1:703
s
original friction factor correlation. (B1)
For incompressible fluids, the total fluid pressure drop
from heat exchanger inlet to outlet is the sum of wall fric- Step 3. Solve equation (58) numerically to obtain hT.
tion loss and individual losses (Soumerai, 1987). The wall
friction loss is directly proportional to the friction factor, Step 4. Determine hS and Ao sequentially from equations
whereas the individual losses are expressed in terms of (59) and (56), respectively.
losses of velocity heads. After the use of the friction Step 5. Obtain geometric parameters for the exchanger,
factor given by equation (A4), one can express the total once hT, hS and Ao have been calculated:
fluid pressure drop as
. The shell-side fluid velocity is calculated from hS. Kern
DP0 bKp1 Kp2 vrp cLv2 (A5) method is used for the first iteration to obtain
!1=0:55
where Kp1 and Kp2 are functions of fluid properties and of m1:3=6 D0:45
s e hs
the heat exchanger geometry, and L is a characteristic vs (B2a)
length. The constant Kp1 involves the sum of individual 36ks Cp1=3
2=3
s rs
0:55

loss coefficients Kindiv associated with velocity heads


rv 2/2. while for other iterations the following relationship,
The simple pressure drop relationship can be obtained if based on the Bell Delaware method, applies
equation (A5) is written as
 1=1rh
hs
DP Kp3 Lv 2rp0
(A6) vs (B2b)
KS1

To determine the unknowns rp0 and Kp3, the following . With the values of hT, the tube-side fluid physical prop-
conditions are used erties and the inside tube diameter, the tube-side fluid
velocity can be calculated:
DP0 v DPv (A7)
  !1=0:8
@DP0  @DP  D1=5 7=15
(A8) ti mt hT
@ v  v @ v  v vt
2:3kt2=3 r4=5 1=3
(B3a)
t Cpt

The characteristic length L can be related to the fluid


velocity and exchanger surface area through This equation is based on Colburn equation (McAdams,
1954) for the determination of the film heat transfer
L Kp4 Av (A9) coefficient for turbulent flow inside circular tubes. If
Petukhov equation (Ozisik, 1985) is used instead of
The combination of equations (A4), (A6) and (A9) yields Colburn equation, the previous formula applies only
the desired compact relationship for the pressure drop, for the first iteration; after the first iteration, the tubeside
stream velocity based on Petukhov equation is given by
DP KA(h)g (A10)  1=0:8
hT
vt (B3b)
KT1
APPENDIX B. DESIGN AND OPTIMIZATION . The total number of tubes is obtained from vt
ALGORITHMS
 6
The design algorithm consists of the following steps. 10 Ntp Qt
Ntt  2  (B4)
Step 1. Specify design data. These data include: (1) mass pDti =4 vt
flowrates, physical properties, inlet and outlet temperatures,
fouling factors and allowable pressure drops for each . The effective tube length for heat transfer is calculated
stream; (2) inside and outside tube diameters, tube pitch, from equation (35).
tube layout, and number of tube passes; and (3) the . The shell diameter is obtained from the total number of
number of sealing strips, the clearances and the baffle cut tubes, the specified inside shell diameter-to-tube bundle

Trans IChemE, Part A, Chemical Engineering Research and Design, 2005, 83(A5): 539550
550 SERNA and JIMENEZ

bypass clearance, and the tube outside diameter. First, functions are smaller than some given tolerance, 1:
Dctl is estimated from equation (36). Then, the shell  
diameter is calculated as g1 KT  KTcalc  (B8a)
 
g2 KS  KScalc  (B8b)
Ds Dctl Lbb Dt (B5)  
g3 m  mcalc  (B8c)
 
g4 n  ncalc  (B8d)
. The central baffle spacing is calculated from the diameter
of the circle through the centres of the outermost tubes If convergence is not achieved, calculate the values of the
and the shell-side fluid velocity. tear variables for the next iteration and go back to Step
3. Although other substitution methods can be used, the fol-
  lowing direct substitution procedure provides satisfactory
106 Qs convergence properties:
Lbc  
(B6)
vs Lbb Dctl =Ltpe Ltp  Dt
KT KTcalc , KS KScalc , m mcalc and n ncalc
(B9)
. The number of baffles is calculated from the central
baffle spacing and the effective tube length for heat trans- The optimization algorithm is closely related to the
fer. design algorithm, since they have the same outer loops.
Hence, these algorithms differ only in steps 1, 3, and
4. Then, for the optimization algorithm the following modi-
Lta  Li Lo Lbc
Nb 1 (B7) fied steps are employed:
Lbc Step 1. The initial data also include the economic and
technical parameters for heat exchangers needed for
It should be noticed that in order to allow for larger inlet equation (62), that is: the annualization factor for capital
and outlet baffle spacings, a set of variables Lbi and Lbo cost, the plant operation time per annum, the coefficients
can be written as LiLbc and LoLbc, respectively, where in capital cost laws for heat exchanger and pumps, the
Li is the ratio of Lbi to Lbc and Lo is the ratio of Lbo to cost per unit of power, and the efficiency of pumping
Lbc. The uniform baffle spacing solution is obtained by devices.
setting Li Lo 1.
Step 3. Use equations (63) and (64) to solve for shell side
. Given the baffle cut, shell diameter, baffle spacings,
and heat transfer coefficients with any proper technique,
shell-side velocity and number of tubes, one can
such as the Newton Raphson method.
use the Bell Delaware correlations (Taborek, 1983) to
These equations are valid if and only if the compact
calculate the constants a2 and b2, the ideal heat transfer
equation parameters are not functions of the unknown
factor ( jsi), the ideal friction factor ( fsi), as well as the
heat exchanger variables. These approximate analytical
leakage and bypass areas and correction factors for the
derivatives have been used, however, with reasonable suc-
shellside (Jc, Jl, Jb, Jr, Js, Rl, Rb, Rs).
cess. The constants in the derivatives are updated with each
new base point.
Step 6. Calculate new values for the tear variables, KScalc ,
Step 4. Compute Ao using equation (56).
mcalc , KTcalc and ncalc , using the equations (42), (43), (45)
The authors have made available both the design and the
and (46), respectively.
optimization algorithms described in this work to interested
Step 7. Check for convergence of the main loop. The users. The codes can be downloaded from the Website:
algorithm converges when all of the following deviation http://posgrado.fiq.umich.mx/division/pag6.html.

Trans IChemE, Part A, Chemical Engineering Research and Design, 2005, 83(A5): 539550

You might also like