Experimental Study and

Genetic-Algorithm-Based
Correlation on Shell-Side Heat
Transfer and Flow Performance of
Qiu-wang Wang
e-mail: [email protected] Three Different Types of
Gong-nan Xie
Shell-and-Tube Heat Exchangers
Bo-tao Peng
The heat transfer and pressure drop of three types of shell-and-tube heat exchangers, one
Min Zeng with conventional segmental baffles and the other two with continuous helical baffles,
were experimentally measured with water flowing in the tube side and oil flowing in the
State Key Laboratory of Multiphase Flow in shell side. The genetic algorithm has been used to determine the coefficients of correla-
Power Engineering, tions. It is shown that under the identical mass flow, a heat exchanger with continuous
School of Energy and Power Engineering, helical baffles offers higher heat transfer coefficients and pressure drop than that of a
Xi’an Jiaotong University, heat exchanger with segmental baffles, while the shell structure of the side-in-side-out
Xi’an, 710049, China model offers better performance than that of the middle-in-middle-out model. The pre-
dicted heat transfer rates and friction factors by means of the genetic algorithm provide
a closer fit to experimental data than those determined by regression analysis. The pre-
dicted corrections of heat transfer and flow performance in the shell sides may be used in
engineering applications and comprehensive study. It is recommended that the genetic
algorithm can be used to handle more complicated problems and to obtain the optimal
correlations. 关DOI: 10.1115/1.2739611兴

Keywords: heat transfer, pressure drop, shell-and-tube heat exchanger, experimental

study, genetic algorithm, regression analysis, correlations

Introduction baffles are used in STHXs, higher pump power is needed to sat-
isfy higher pressure drop. Therefore, it is essential to modify the
Heat exchangers are important apparatuses for heat and mass
shape of common baffles.
exchange employed in the fields of refrigeration engineering, oil-
In 1960s, the idea of STHXs with helical baffles was proposed
refinery engineering, chemical engineering, environmental protec-
共Bell 关7兴, Li and Kottke 关8兴, Naim and Bar-Cohen 关9兴兲. In STHX
tion, electric power plants, and so on. Generally speaking, the
with helical baffles, baffles were arranged by approximate heli-
fraction of heat exchanger investment to total investment is about
coids, which makes the flow in the shell side continuously helical,
10% to 20% in oil-refineries, while 30% to 40% in chemical en-
resulting in decrease of pressure drop and in enhanced heat trans-
gineering factories. It is preferable to design and manufacture heat
fer rates in the shell side. Many new productions of this type
exchange apparatuses with high effectiveness, low energy con- appeared in 1990s. Nowadays, these types of STHXs are used in
sumption, and low cost. Shell-and-tube heat exchangers 共STHXs兲 many countries such as Russia and Japan.
are widely used for evaporators and condensers. In STHX, perfor- From 1990s, few studies on STHXs with continuous helical
mance of heat transfer may be improved in the shell side, where baffles are reported in open literatures, while most of them are
baffles are located. Segmental baffles 共supporting tubes and with non-continuous helical baffles. The absence of reliable per-
changing fluid flow direction兲, are commonly used in traditional formance data for STHXs with continuous helical baffles causes
STHXs. Segmental baffles make shell-side fluid to flow in a tor- difficulty for designers and engineers. Thus in this study, an ex-
tuous, zigzag manner across the tube bundles, which enhance heat perimental system has been designed, fabricated, and investigated
transfer rates due to flow normal to the tube banks. In aforemen- experimentally, and performances of heat transfer for STHXs with
tioned scenario, however, there exist many problems 共Kral et al. different baffles have been compared.
关1兴, Peng 关2兴, Reppich and Zagermann 关3兴, Wang 关4兴, Wang et al. In recent years, application of genetic algorithms 共GAs兲 to ther-
关5兴, Xie et al. 关6兴, Bell 关7兴兲: 共1兲 high pressure drop occurs due to mal engineering has received much attention for solving real-word
fluid impact normal to shell wall; 共2兲 heat transfer rates decrease problems 共Sen and Yang 关10兴兲. For example, the fin-tube heat
due to flow stagnate in connection region with baffles and shell exchanger performance was predicted using GA 共Pacheco-Vega et
wall; 共3兲 mass velocity decreases due to bypass and leakage, since al. 关11兴兲. Plate-fin heat exchangers were optimized by means of
accuracy of manufacture and installation is not reached; and 共4兲 GAs 共Ozkol and Komurgoz 关12兴, Xie and Wang 关13兴, Mishra et al.
operation time decreases due to librations from flow normal to 关14兴兲. A new design method was proposed to optimize a STHX
tube banks in high mass velocity. When the traditional segmental from an economic point of view by GA 共Selbas et al. 关15兴兲. Geo-
metrical optimization of cross-wavy and cross-corrugated primary
Contributed by the Heat Transfer Division of ASME for publication in the JOUR-
surface recuperators was studied via GAs 共Liang et al. 关16,17兴兲.
HEAT TRANSFER. Manuscript received July 8, 2006; final manuscript received
NAL OF Heat transfer correlations of compact heat exchangers were ob-
December 3, 2006. Review conducted by Anthony M. Jacobi. tained using GAs, to estimate their performance 共Pacheco-Vega et

Journal of Heat Transfer Copyright © 2007 by ASME SEPTEMBER 2007, Vol. 129 / 1277

Downloaded From: http://heattransfer.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Fig. 1 Experimental loop. Cold-water loop: 1-water box; 2,10-valve; 4,11-
electric value; 3-pump; 5-flowmeter; 6-heater; 7-thermo couples; 8-pressure
transducer; 9-pressure drop transducer; 12-heat exchanger. Hot-water loop:
13-water box; 14-heater; 15,23-valve; 16-pump; 17-electric valve; 18-
flowmeter; 19-heater; 20-thermo couples; 21-pressure transducer; 22-
pressure drop transducer. Oil loop: 24-oil box; 25-heater; 26,34-valve; 27-
pump; 28-electric valve; 29-flowmeter; 30-heater; 31-thermo couples; 32-
pressure transducer; 33-pressure drop transducer; 35-expand box. Cooling
water loop: 36-water box; 37,39-valve; 38-pump; 40-tower; 41-test heat
exchanger.

al. 关18兴, Pacheco-Vega et al. 关19兴兲, other applications in estimating pressure drop between the inlet and outlet of the oil side. The
of the radiative properties and the surface emissivities 关20兴, and in temperatures at the inlet and outlet of the oil side and water side
modeling phonon-phonon normal and umklapp scattering pro- were measured by the copper-constantan thermocouples. The ac-
cesses 关21兴 were presented. These reports suggest that the genetic curacy of the thermocouple reading is estimated with 0.1° C. The
algorithm has a strong ability search and optimize, which can volume flow rates of the oil side and water side were measured by
successfully optimize and predict thermal problems. Therefore, turbine flowmeters, and the oil-side pressure drop was measured
the present study has employed the genetic algorithm to correlate by a differential pressure transducer with an error less than 0.25%.
and predict the heat transfer performance, to correlate the Nusselt All the digital data obtained from thermocouples, flowmeters, and
number versus Reynolds number, and the flow resistance to cor- transducer were collected into the computer, and then translated
relate the friction factor versus Reynolds number, in the shell side into real physical data by a data transition system. Finally, the
for three different types of STHXs. corresponding experimental data are saved and averaged on the
computer for data reduction.
Experimental System In the experiments, once all the temperature data were stabi-
lized within 0.1° C, the system was allowed to approach the
Experimental Loop. The experimental loop, shown in Fig. 1, steady state before any data were recorded and saved. The water
was designed and fabricated at School of Energy and Power En- flow rate was fixed at a specific value, through which the water-
gineering, Xi’an Jiaotong University. There are three sub-loops side Reynolds numbers were around 3000. Thus, Gnielinski’s cor-
共an oil loop, a cold water loop, and a cooling water loop兲 for relation can be validly used for calculation of water-side/tube-side
achieving the heat exchange of the experimental loop in the heat transfer and friction. Experiments were conducted with the
present study. In Fig. 1, 41 is the test heat exchanger. Oil-water inlet oil temperature ranging between 50° C and 60° C, inlet water
共by oil loop and cold water loop兲 or water-water 共by hot water temperature ranging between 15° C and 30° C, and oil flow rate
loop and cold water loop兲 heat exchange is carried out in the ranging between 0.5 and 4.5 kg/ s. The uncertainty of the heat
experimental loop. The cooling water loop is used to cool the balance between the oil side and water side was within 8%, and a
heated water of the cold water loop. A more detailed description 60-set of data per half-minute were averaged and saved into the
of the experimental system and tested heat exchangers can be computer.
found in Ref. 关2兴. Three tested heat exchangers are shown in Fig. 2. Figure 2共a兲 is
Oil loop: The oil loop includes the oil-refined loop and the a heat exchanger with segmental baffles 共hereafter, this type of
experimental loop. In the oil-refined loop, the heated oil flows in heat exchanger is called as HX1兲, the other two are the heat ex-
an expanded box, elicited by a pump, and then flows into the changers with continuous helical baffles, as shown in Figs. 2共b兲
heater after flow in the refiner. In the experimental loop, the and 2共c兲. The only difference between these latter two helical heat
heated oil is elicited by the pump, flows through the turbine flow- exchangers is in the manner of inlet-outlet flow of the shell-side
meter, flows downstream through the test section, and last flows fluid. One is middle-in-middle-out 共HX2, Fig. 2共b兲兲 and the other
into the heater. The loops were switched by a valve. one is side-in-side-out 共HX3, Fig. 2共c兲兲. The cores of HX2 and
Cold water loop: Water flows through turbine flowmeter by a HX3 are the same 共the layouts of tubes and baffles are identical兲,
pump, flows through the test section of tube-side, and then flows the only difference between the HX2 and HX3 is the locations of
into one side of the heat exchanger 12 in the cold water loop, and inlets and outlets of the shell-side flows. The locations of inlet and
last flows into the water box. outlet of HX2 are on the middle of the shell side, which is normal
Cooling water loop: Cooling water flows through the other side to shell, while the positions of inlet and outlet of HX3 are on the
of heat exchanger 12 by the pump, flows into the cooler, and last side of shell side, which is tangential to the shell. It should be
flows into the water box. noted that for helical heat exchangers 共HX2, HX3兲, there is a
In the experimental process, there are seven parameters to be blocked center tube.
measured: inlet and outlet temperature of the oil side 共shell side兲, The heat exchangers are 1-2 type 共one shell pass and two tube
i.e., to1, to2; inlet and outlet temperature of water side 共tube side兲, passes兲, with hot oil flowing in the shell side and cold water
i.e., tw1, tw2; volume flow rate of both sides, i.e., vo , vw; and total flowing in the tube side. The detailed values of the parameters for

1278 / Vol. 129, SEPTEMBER 2007 Transactions of the ASME

Downloaded From: http://heattransfer.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 the three tested heat exchangers are shown in Tables 1 and 2.
Experiments were performed for Reynolds number ranging from
300 to 7000 on the shell side. Heat transfer rates varied from
20 kW to 50 kW. It should be noted that the diameter of center
tube in HX1 is zero because there is no center tube.
Data Reduction. The main purpose of the data reduction is to
determine the shell-side heat transfer and friction characteristic,
i.e., Nu and f, of the heat exchangers from the data, which re-
corded at steady state conditions during the each test run, and to
find out corresponding power-law correlations of Nu versus Re,
and f versus Re for each heat exchanger. The average values of
the inlet and outlet temperatures of the oil side and water side
were used to evaluate the physical properties of oil and water,
respectively.
The oil-side/shell-side heat transfer rate can be given as
Qo = moc p,o共to1 − to2兲 共1兲
where the oil mass flow rate m is obtained from the volume flow
rate multiplying corresponding density, as is the water mass flow
rate. The water-side heat transfer rate can be given as
Qw = mwc p,w共tw1 − tw2兲 共2兲
Thus, the total rate of heat transfer used in the calculation is the
average of the oil side and water side as follows:
Qave = 共Qo + Qw兲/2 共3兲
and the heat transfer balance can be determined by
兩Qo − Qw兩
␧ = 100 % 共4兲
Qave
The oil-side heat transfer coefficient ho is determined from the
overall heat transfer resistance relationship
1 1 do do do 1
Fig. 2 Shell-and-tube heat exchangers = + Ro + ln + + Rw 共5兲
U ho 2kwall di di hw
where do and di are outside and inside diameters of the tube,
Table 1 Geometrical parameters of heat exchangers †2‡ respectively. Because of the new test heat exchanger, the fouling
resistances are neglected, that is, Ro = Rw = 0. The overall heat
Parameters Value transfer coefficient U can be determined from
Qave
Inner diameter of shell, mm 207 U= 共6兲
Outer diameter of tube, mm 10 A⌬tm
Inner diameter of tube, mm 8
Arrangement of tube Triangle where the heat transfer area A is based on the area outside tubes,
Effective length of tube 共l兲, mm 620
Number of tubes 共helical兲 158 A = ␲dolN 共7兲
Number of tubes 共segmental兲 176
Outer diameter of inlet tube in water side, mm 57 where, l and N are the length of tube and the number of tubes,
Inner diameter of inlet tube in water side, mm 50 respectively. The logarithmic-mean temperature difference ⌬tm is
Outer diameter of outlet tube in water side, mm 57
Inner diameter of outlet tube in water side, mm 50 determined by
Outer diameter of inlet tube in oil side, mm 57
Inner diameter of inlet tube in oil side, mm 50 ⌬tm = F⌬tc 共8兲
Outer diameter of outlet tube in oil side, mm 57
Inner diameter of outlet tube in oil side, mm 50 where F is the correction factor and ⌬tm is the counter-flow
Length of heat exchanger 共L兲 mm 670 logarithmic-mean temperature difference. They can be determined

Table 2 Geometrical parameters for test sections †2‡

Spiral pitch, Number of Diameter of center

Test section mm baffles tube, mm

HX with segmental baffles 共HX1兲 70 7 NA

HX with continuous helical baffles, 48 9 48

middle-in-middle-out 共HX2兲

HX with continuous helical baffles, 48 9 48

side-in-side-out 共HX3兲

Journal of Heat Transfer SEPTEMBER 2007, Vol. 129 / 1279

Downloaded From: http://heattransfer.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 by and is now widely used in various research areas, since it has a
powerful ability for global search and optimization of complex
共to1 − tw2兲 − 共to2 − tw1兲

冉 冊
⌬tc = 共9兲 systems. In this paper, the genetic algorithm has been used to
to1 − tw2 search for the numerical values of constants in Eq. 共18兲, for closer
ln
to2 − tw1 fitting to experimental data.

ln 冉 冊
1−P
1 − PR
Genetic Algorithm

冉 冊
F= 共10兲 Simple Descriptions. The genetic algorithm is maintained by a
2 − P共1 + R − 冑R2 + 1兲 population of parent individuals that represent the latent solutions
ln
2 − P共1 + R + 冑R2 + 1兲 of a real-world problem. Figure 3 shows the flowchart of the
genetic algorithm. Each individual is assigned a fitness based on
tw1 − tw2 to1 − to2 how well each individual fits in a given environment, evaluated by
P= R= 共11兲 survival of the fitness. Fit individuals go through the process of
to1 − Tw2 tw1 − tw2
survival selection, crossover, and mutation, creating a next gen-
The water-side/tube-side heat transfer coefficient hw is evalu- eration, called child individuals. A new population is therefore
ated from the semi-empirical correlation of Gnielinski 关22兴: formed by selection of good individuals from parent and child

hw = 冉 冊
kw 共f w/8兲共Re − 1000兲Pr
冑 2/3
di 1 + 12.7 f w/2共Pr − 1兲
1+
d1
l
冋 冉冊册 2/3
共12兲
individuals. After some generations, the algorithm is converged to
the best individual, which possibly represents the best solution of
the given problem. More details about the description of the ge-
where the friction factor is given by netic algorithm can be found in many books 共Goldberg 关23兴;
Michalewicz 关24兴兲.
f w = 共1.82 log10 Re − 1.64兲−2 共13兲
Variables Ranges. A binary string is adopted for encoding the
where the water-side Re= ␳udi / ␮. variables of a given model. Two coefficients of correlations were
Thus, the oil-side heat transfer coefficient ho can be calculated assigned to vary between 关−1 , 1兴. Considering the ability of com-
from Eq. 共4兲, and then the Nusselt number Nu can be calculated puters to handle bit operations and engineering applications, the
by computational precision is set to four decimals 共0.0001兲. Thus, a
do coefficient needs a 15-character binary, and an individual coeffi-
Nu = ho 共14兲 cient of a model needs a 30-character binary.
ko
For simplicity, the flow friction characteristic can be presented Fitness Function. The conventional genetic algorithm deter-
in terms of f-factors as follows mines the global maximum value in search spaces; however, the
objective of the present correlation and prediction is to make the
2⌬P do deviation minimum between the predicted and the experimental
f= 共15兲 results. Thus, when searching the coefficients ␣ and ␤ in Eq. 共18兲,
␳umax
2
L
a fitness function should be given as follows:
where L is the effective length of the heat exchanger, and umax is
CQ

冑
the velocity at the minimum free flow area, and can be determined Fit = 共19兲
by M

兺
1
For HX1: 共Q p − Qe兲2
M i=1
A = Bds共1 − do/S兲 共16兲
For HX2 and HX3 where CQ is a constant and its value is assigned 5000; i.e., about
10% of maximum heat transfer rate 关2兴. M refers to the number of
A = 共1 − B/ds兲Bds共1 − do/S兲 共17兲 experimental data. Qe and Q p refer to experimental and predicted
The uncertainty of measured temperature difference, flow rate, heat transfer rates, respectively. The flowchart of prediction is
heat transfer rate, and heat transfer coefficient are less than 1.8%, shown in Fig. 3. Note that global maximum fitness corresponds
0.14%, 2.5%, and 3.7%, respectively. Typical uncertainty of the minimum deviation leading to the best predicted heat transfer rate.
shell-side heat transfer coefficient is about 15.64 W / m2 k. The Calculation of heat transfer rate can be found in Ref. 关2兴. When
maximum and minimum uncertainties for pressure drop are searching the coefficients ␾ and ␸ in Eq. 共18兲, another fitness
155.9 Pa and 190 Pa, respectively. function should be given as follows:
Cf

冑
Correlations for Friction Factor and Nusselt Number. The Fit = 共20兲
experimental data are often compressed into dimensionless corre- M

兺
1
lations as follows: 共f p − f e兲2
M i=1
Nu Pr−1/3 = ␣ Re␤ f = ␾ Re␸ 共18兲
In this case, C f is assigned 0.05, about 10% of maximum friction
So far, Eq. 共18兲 has been used to correlate the average heat factor 关2兴.
transfer rate and pressure drop 共experimental data兲. Engineers use
Eq. 共18兲, which covers the ranges of the parameters for the same Genetic Operators and Parameters. In the present study,
type of exchangers. The dimensional values for heat transfer co- tournament selection, uniform crossover, and one-point mutation
efficient and friction factor are used in practical design and/or were selected. Niching and elitism were adopted 共Carroll 关25,26兴;
operation. The coefficients ␣ , ␤ , ␾, and ␸ in Eq. 共18兲 can be de- Michalewicz 关27兴兲. The size of the population and maximum evo-
termined by regression analysis or by importing into post-software lution generation were set to 20 and 1000, respectively. The prob-
such as ORIGIN, MATLAB, etc. It is possible that the traditional ability of crossover and mutation were set to 0.5 and 0.005, re-
method, i.e., regression analysis, is not the best method to corre- spectively. The selection of genetic parameters is a trial-and-error
late the experimental data, and thereby the errors may not be process, and with the variation of these parameters, results are not
minimum. Therefore, some other new methods have to be found exactly identical but are very close to one another. On the other
for correlating the data. hand, in this study the selection of genetic parameters and opera-
The genetic algorithm was developed about two decades ago tors is based on the previous recommendations 关25,26兴.

1280 / Vol. 129, SEPTEMBER 2007 Transactions of the ASME

Downloaded From: http://heattransfer.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Fig. 3 Flow chart of genetic algorithm and thermal prediction

Results and Discussion The friction factor number versus Reynolds number in the shell
side of the three heat exchangers is shown in Fig. 6. Again, the
The Nusselt number versus Reynolds number in the shell side
black solid points refer to experimental data while the lines refer
of the three heat exchangers is shown in Fig. 4. The points refer to
to approximate lines fitted by the genetic algorithm. For all three
experimental data, while the lines refer to approximate lines fitted
heat exchangers, it can be seen that the friction factor decrease
by the genetic algorithm. For all three heat exchangers, it can be
with the increase of the Reynolds number. The total pressure drop
seen that the Nusselt number increases with the increase of the
versus the mass flow rate in the shell side of the three heat ex-
Reynolds number. Heat transfer coefficients of three heat ex-
changers is shown in Fig. 7. It can be seen that, under identical
changers versus mass flow rate are shown in Fig. 5. It can be
mass flow rates, pressure drops of both HX2 and HX3 are higher
found that, under identical mass flow rates, heat transfer coeffi-
than that of HX1. The difference is higher at higher mass flow
cients of both HX2 and HX3 are higher than that of HX1. This
rates. The pressure drop of HX2 is slightly higher than that of
difference increases with the increase in mass flow rate. The heat
HX3.
transfer coefficient of HX3 is slightly higher than that of HX2.

Fig. 4 Heat transfer performance for three heat exchangers Fig. 5 Heat transfer coefficient versus flow rate

Journal of Heat Transfer SEPTEMBER 2007, Vol. 129 / 1281

Downloaded From: http://heattransfer.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Fig. 6 Flow friction performance for three heat exchangers

The evolution process for the fitness is shown in Fig. 8. At the

beginning of evolution process 共less than about 20 generations兲,
the differences between every individual are relatively large,
which in turn leads to the individuals with low fitness being elimi-
nated and individuals with higher fitness being saved. After cer-
tain generations 共larger than about 100 generations兲, the variation
of fitness for minimum deviation is small, finally approaching to a
constant value. The predicted Nusselt number versus experimental
Nusselt number of the three heat exchangers is shown in Fig. 9. In
the figure, the middle straight line indicates that the predicted
results agree perfectly with the experimental data. It can be ob-
served that the predicted Nusselt numbers obtained by the genetic
algorithm are quite close to the experimental data. All the devia-
tions are less than 10% and the maximum deviation is 9.6%.
The predicted heat transfer rate versus experimental data of the
three heat exchangers is shown in Fig. 10. In this figure, GA refers Fig. 8 Evolution process of maximizing fitness
to the results obtained by genetic Algorithm, while RA refers to
the results fitted by regression analysis 关2兴. From these figures, it

冉 冊 冉冊
can be seen that the predicted data 共+兲 obtained by genetic algo-
rithms are closer to experimental data than that by regression Qp fp
RiQ = and Rif = 共21兲
analysis 共ⴰ兲, because most of predicted results are much closer to Qe i fe i
the middle straight line, which represents the best agreements with

兺冉 冊
M M
the experimental data. fp
兺
1 1
In order to compare the results obtained by genetic algorithm RQ = RiQ and Rf = 共22兲
M i=1 M i=1 fe
with that by regression analysis 关2兴, two factors, i.e., R and ␴, are
defined as

Fig. 7 Pressure drop versus flow rate Fig. 9 Predicted Nusselt number versus experimental data

1282 / Vol. 129, SEPTEMBER 2007 Transactions of the ASME

Downloaded From: http://heattransfer.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Fig. 10 Predicted heat transfer rates versus experimental data Fig. 11 Predicted friction factors versus experimental data of
of three exchangers three exchangers

␴Q = 冑兺 M

i=1
共RQ − RiQ兲
M
and ␴f = 冑兺 M

i=1
共R f − Rif 兲
M
共23兲
leading to higher consumption of pressure drop than that of HX1.
On the other hand, in HX2 or HX3, there is no sluggish region
and shell fluid does not flow normal to tube bundles. The fluid
The two factors, i.e., R and ␴, are important evaluation factors, flows in a continuous helical manner with uniform velocity, cov-
which quantitatively reflect the order of agreement between the ering and brushing the tubes, resulting in decrease of pressure
predicted results and the experimental results. Sen and Yang 关10兴 drop and in enhancement of heat transfer rates in the shell side.
stated that the R represented the average accuracy of the predicted The shell fluid flows from the side of the shell in HX3, improving
results, while ␴ represented the scatter of the prediction. The val- the manner of helical flow, in turn leading to smaller pressure drop
ues of R and ␴ for three heat exchangers are listed in Table 3, and higher heat transfer rate than that of HX2, in which the shell
where the results obtained by GA and the ones by RA are com- fluid flows through the middle of the shell side.
pared. It can be seen that the values of R by GA are closer to unity As shown above, the heat transfer rate and the friction factor by
than those by RA, and the ␴ by GA is smaller than that by RA. the genetic algorithm are closer to the corresponding real values.
This shows that the predicted results by GA are much better than This indicates that the genetic algorithm can reflect the real per-
those by RA 关2兴. formance better and it has advantages over RA, such as searching
The predicted friction factor versus experimental friction factor and predicting. It is recommended that engineers should use the
of three heat exchangers is shown in Fig. 11. It can be observed genetic algorithm to fit and predict the experimental results, lead-
that, the predicted friction factors obtained by GA are well close ing to further studies or engineering applications.
to the experimental data. Most of all deviations are less than 10%, In engineering practice, the heat transfer rate, Q, is frequently
and only three of the deviations are about 13% 共the number of desired. Application of a computational intelligent method to pre-
datapoints is 39兲. dict performance of thermal devices has received much attention.
The predicted coefficients of correlations, determined by re- For example, in recent years, some researchers use artificial neural
gression analysis 共RA兲 and by the genetic algorithm 共GA兲, are networks to predict the performance of heat exchangers 关28,29兴,
listed in Table 4. Such correlations of heat transfer and flow per- where the aim is to make the predicted results closer to real per-
formance in the shell sides may be used in engineering applica- formance. Thus, the application of the genetic algorithm to more
tions and comprehensive study; for example, engineers may bor- engineering real-world problems can show its superiority with
row them to design STHXs with segmental or continuous helical searching, optimizing, and predicting, leading to further chal-
baffles, or researchers may use them to optimize corresponding lenges.
STHXs under a valid range of parameters.
The cross-sectional area in the shell side of HX2 or HX3 is
smaller than that of HX1, because there is a blocked center tube. Conclusions
In this result, under identical mass flow rate conditions, higher
fluid velocity occurs in the shell side of HX2 or HX3, in turn In this paper the experimental system for testing the shell-side
heat transfer and pressure drop performances of shell-and-tube
heat exchangers is designed and built. Heat transfer and flow per-
Table 3 Quantitative comparison of predicted results formance for three kinds of heat exchanger are investigated and
compared. The genetic algorithm is also applied to correlate the
HX1 HX2 HX3 heat transfer and pressure drop performances in the shell sides.
The main conclusions are summarized as follows.
RQ Obtained by RA 关2兴 0.971386 0.93746 0.988464
Obtained by GA 1.027565 1.045204 1.016262 共1兲 Both the heat transfer coefficient and friction factor of the
␴Q Obtained by RA 关2兴 0.02184 0.01982 0.02452 three heat exchangers increase with the increase of Rey-
Obtained by GA 0.00512 0.00991 0.00868 nolds number. Under identical mass flow rates in the shell
Rf Obtained by RA 关2兴 1.005584 0.997579 0.988199
Obtained by GA 0.999171 0.999466 0.988733
side, heat transfer coefficients and pressure drops in STHXs
␴f Obtained by RA 关2兴 0.01714 0.03832 0.08637 with continuous helical baffles are higher than that of
Obtained by GA 0.01710 0.03566 0.07197 STHXs with segmental baffles. In addition, for heat ex-
changers with continuous helical baffles, the one with

Journal of Heat Transfer SEPTEMBER 2007, Vol. 129 / 1283

Downloaded From: http://heattransfer.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Table 4 Predicted coefficients by RA and GA

Correlations Coefficients Regression Analysis 关2兴 Genetic Algorithm

HX1 HX2 HX3 HX1 HX2 HX3

Nu= ␣ Re␤ Pr1/3 ␣ 0.106 0.0451 0.0598 0.0743 0.0329 0.0533

␤ 0.717 0.699 0.669 0.747 0.742 0.688
f = ␾ Re␸ ␾ 2.02 1.07 0.53 2.05 0.63 1.64

␸ −0.22 −0.40 −0.31 −0.22 −0.33 −0.45

“side-in-side-out” inlet-outlet flow manner has better com- umax ⫽ maximum velocity at minimum free flow area,
prehensive performance than that with “middle-in-middle- m/s
out” flow manner. v ⫽ volume fallow rate, m3 / s
共2兲 The heat transfer rates and friction factors correlated by
means of the genetic algorithm are quite close to the ex- Greek Symbols
perimental data. It is beneficial to use the genetic algorithm ␣,␤ ⫽ constants of heat transfer correlation
to correlate and predict performances for complicated prob- ␾,␸ ⫽ constants of friction factor correlation
lems. The predicted correlations for heat transfer and flow ␳ ⫽ fluid density, kg/ m3
performance in the shell sides may be used in engineering ␮ ⫽ dynamic viscosity, Pa s
application and comprehensive study. ⌬P ⫽ pressure drop in shell side, Pa
⌬tm ⫽ logarithmic-mean temperature difference, K
Acknowledgment Subscripts
1 ⫽ Inlet
This work was supported by the NSFC Fund for Creative Re-
2 ⫽ Outlet
search Groups 共Grant No. 50521604兲 and National Defense Sci-
ence and Technology Key Laboratory Foundations of China o ⫽ Oil side,
共Grant No. 51482100204JW0801兲. The authors also greatly appre- w ⫽ Water side
ciates Professor Hiroyuki Ozoe, a foreign expert at Xi’an Jiaotong wall ⫽ tube wall
University, for his efforts on the improvement of English writing
in this article. References
关1兴 Krai, D., Stehlik, P., Van der Ploeg, H. J., and Master, B. I., 1996, “Helical
Baffles in Shell-and-Tube Heat Exchangers, Part I: Experimental Verification,”
Heat Transfer Eng., 17, pp. 93–101.
Nomenclature 关2兴 Peng, B. T., 2005, “Experimental Study of Heat Transfer and Pressure Drop
A ⫽ area, m2 for Shell-and-Tube Heat Exchangers With Continuous Helical Baffles,” Mas-
B ⫽ pitch of baffles, m ter’s thesis, Department of Energy and Power Engineering, Xi’an Jiaotong
C f , CQ ⫽ constant of fitness University 共in Chinese兲.
关3兴 Reppich, M., and Zagermann, S., 1995, “A New Design Method For Segmen-
cp ⫽ specific heat at constant pressure, J / 共kg K兲 tally Baffled Heat Exchangers,” Comput. Chem. Eng., 19, pp. 37–142.
di ⫽ inside diameter of tube, m 关4兴 Wang, Q. W., 2004, “Current Status and Development of Shell-Side Heat
do ⫽ outside diameter of tube, m Transfer Enhancement of Shell-and-Tube Heat Exchangers With Helical
Baffles,” Journal of Xi’an Jiaotong University, 38, pp. 881–886 共in Chinese兲.
ds ⫽ Diameter of shell, m 关5兴 Wang, Q. W., Luo, L. Q., Zeng, M., and Tao, W. Q., 2005, “Shell-Side Heat
Fit ⫽ fitness Transfer and Pressure Drop of Shell-and-Tube Heat Exchangers With Overlap
fe ⫽ experimental friction factor Helical Baffles,” Journal of Chemical Industry and Engineering 共China兲, 56,
pp. 598–601 共in Chinese兲.
fp ⫽ predicted friction factor by genetic algorithm 关6兴 Xie, G. N., Peng, B. T., Chen, Q. Y., Wang, Q. W., Luo, L. Q., Huang, Y. P.,
共GA兲 and Xiao, Z. J., 2006, “Experimental Study and Prediction of Pressure Drop
k ⫽ thermal conductivity, W / 共m K兲 and Heat Transfer in Shell-Side of Tube-and-Shell Heat Exchangers,” Proc.
m ⫽ mass flow rate, kg/s Chin. Soc. Electr. Eng., 26共21兲, pp. 104–108 共in Chinese兲.
关7兴 Bell, K. J., 2004, “Heat Exchanger Design for the Process Industries,” ASME
h ⫽ heat transfer coefficient of STHX, W / 共m2 K兲 J. Heat Transfer, 126共6兲, pp. 877–885.
l ⫽ length of tube, m 关8兴 Li, H. D., and Kottke, V., 1998, “Effect of the Leakage on Pressure Drop and
L ⫽ effective length of heat exchanger, m Local Heat Transfer in Shell-and Tube Heat Exchangers for Staggered Tube
Arrangement,” Int. J. Heat Mass Transfer, 41, pp. 425–433.
M ⫽ number of experimental datapoints 关9兴 Naim, A., and Bar-Cohen, A., 1996, A New Developments in Heat Exchangers,
N ⫽ number of tubes Gordon and Breach Publishers, Amsterdam, pp. 467–499.
Nue ⫽ experimental Nusselt number 关10兴 Sen, M., and Yang, K. T., 2000, “Applications of Artificial Neural Networks
and Genetic Algorithms in Thermal Engineering,” in: The CRC Handbook of
Nup ⫽ predicted Nusselt number by GA Thermal Engineering, F. Kreith, eds., CRC Press, Boca Raton, FL, pp. 620–
Pr ⫽ Prandtl number 661.
Pop ⫽ size of population 关11兴 Pacheco-Vega, A., Sen, M. Yang, K. T., and McClain, R. L., 1998, “Genetic
Q ⫽ heat transfer rate, W Algorithms-Based Predictions of Fin-Tube Heat Exchanger Performance,”
Proceedings of 11th International Heat Transfer Conference, Kyongju, Korea,
Qe ⫽ experimental heat transfer rate, W August 23–28.
Qp ⫽ predicted heat transfer rate, W 关12兴 Ozkol, I., and Komurgoz, G., 2005, “Determination of the Optimum Geometry
R ⫽ fouling resistance, 共m2 K兲 / W of the Heat Exchanger Body via a Genetic Algorithm,” Numer. Heat Transfer,
Part A, 48, pp. 283–296.
Rf ⫽ predicted factor for friction factors 关13兴 Xie, G. N., and Wang, Q. W., 2006, “Geometrical Optimization of Plate-Fin
RQ ⫽ predicted factor for heat transfer rates Heat Exchanger Using Genetic Algorithms,” Proc. Chin. Soc. Electr. Eng.,
Re ⫽ Reynolds number 26共7兲, pp. 53–57 共in Chinese兲.
S ⫽ distance between two centers of tube 关14兴 Mishra, M., Das, P. K., and Sarangi, S., 2004, “Optimum Design of Crossflow
Plate-Fin Heat Exchangers through Genetic Algorithm,” International Journal
T ⫽ evolution generation of Heat Exchangers, 5, pp. 379–401.
U ⫽ total heat transfer coefficient, W / 共m2 K兲 关15兴 Selbas, R., Kizilkan, O., and Reppich, M., 2006, “A New Design Approach for

1284 / Vol. 129, SEPTEMBER 2007 Transactions of the ASME

Downloaded From: http://heattransfer.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Shell-and-Tube Heat Exchangers Using Genetic Algorithms From Economic fer, 127, pp. 1129–1137.
Point of View,” Chem. Eng. Process., 45, pp. 268–275. 关22兴 Gnielinski, V., 1976, “New Equations for Heat and Mass Transfer in Turbulent
关16兴 Liang, H. X., Xie, G. N., Zeng, M., Wang, Q. W., and Feng, Z. P., 2005, Pipe and Channel Flows,” Indian Chem. Eng., 16, pp. 359–368.
“Application Genetic Algorithm to Optimization Recuperator in Micro- 关23兴 Goldberg, D. E., 1989, Genetic Algorithms in Search, Optimization and Ma-
Turbine,” The 2nd International Symposium on Thermal Science and Engi- chine Learning, Addison-Wesley Publishing Company Inc., Reading, MA.
neering, Beijing, China, October 23–25. 关24兴 Michalewicz, Z., and Fogel, D. B., 2000, How to Solve it: Modern Heuristics,
关17兴 Liang, H. X., Xie, G. N., Zeng, M., Wang, Q. W., and Feng, Z. P., 2006, Springer, New York.
“Genetic Algorithm Optimization for Primary Surfaces Recuperator of Micro- 关25兴 Carroll, D. L., 1996, “Chemical Laser Modeling With Genetic Algorithms,”
turbine,” Proceedings of GT2006, ASME Turbo Expo, Barcelona, Spain, Paper AIAA J., 34, pp. 338–346.
No. GT 2006-90366, May 8–11, 2006.
关26兴 Carroll, D. L., 1996, “Genetic Algorithms and Optimizing Chemical Oxygen-
关18兴 Pacheco-Vega, A., Sen, M., Yang, K. T., and McClain, R. L., 2001, “Correla-
Iodine Lasers,” Developments in Theoretical and Applied Mechanics, H. Wil-
tions of Fin-Tube Heat Exchanger Performance Data Using Genetic Algo-
son et al., eds., School of Engineering, The University of Alabama, Vol. XVIII,
rithms Simulated Annealing and Interval Methods,” Proceedings of ASME the
Heat Transfer Division, New York, November 11–16. pp. 411–424.
关19兴 Pacheco-Vega, A., Sen, M., and Yang, K. T., 2003, “Simultaneous Determina- 关27兴 Michalewicz, Z., Deb, K., Schmidt, M., and Stidsen, T, 2000, “Test-Case
tion of In- and Over-Tube Heat Transfer Correlations in Heat Exchangers by Generator for Constrained Parameter Optimization Techniques,” IEEE Trans.
Global Regression,” Int. J. Heat Mass Transfer, 46, pp. 1029–1040. Evol. Comput., 4, pp. 197–215.
关20兴 Deiveegan, M., Balaji, C., and Venkateshan, S. P., 2006, “Comparison of Vari- 关28兴 Pacheco-Vega, A., Diaz, G, Sen, M., Yang, K. T., and McClain, R. L., 2001,
ous Methods for Simultaneous Retrieval of Surface Emissivities and Gas Prop- “Heat Rate Predictions in Humid Air-Water Heat Exchangers Using Correla-
erties in Gray Participating Media,” ASME J. Heat Transfer, 128, pp. 829– tions and Neural Networks,” ASME J. Heat Transfer, 123, pp. 348–354.
837. 关29兴 Xie, G. N., Wang, Q. W., Zeng, M., and Luo, L. Q., 2007, “Heat Transfer
关21兴 Chen, Y. F., Li, D. Y., Lukes, J. R., and Majumdar, A, 2005, “Monte Carlo Analysis for Shell-and-Tube Heat Exchangers With Experimental Data by Ar-
Simulation of Silicon Nanowire Thermal Conductivity,” ASME J. Heat Trans- tificial Neural Networks Approach,” Appl. Therm. Eng., 27, pp. 1096–1104.

Journal of Heat Transfer SEPTEMBER 2007, Vol. 129 / 1285

Downloaded From: http://heattransfer.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use