Accepted Manuscript

Title: Optimal Design and Analysis of Mechanical Draft Cooling Tower Using
Improved Jaya Algorithm

Author: R.V. Rao, K.C. More

PII: S0140-7007(17)30258-X
DOI: http://dx.doi.org/doi: 10.1016/j.ijrefrig.2017.06.024
Reference: JIJR 3689

To appear in: International Journal of Refrigeration

Received date: 13-2-2017

Revised date: 10-4-2017
Accepted date: 19-6-2017

Please cite this article as: R.V. Rao, K.C. More, Optimal Design and Analysis of Mechanical
Draft Cooling Tower Using Improved Jaya Algorithm, International Journal of Refrigeration
(2017), http://dx.doi.org/doi: 10.1016/j.ijrefrig.2017.06.024.

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service
to our customers we are providing this early version of the manuscript. The manuscript will
undergo copyediting, typesetting, and review of the resulting proof before it is published in its
final form. Please note that during the production process errors may be discovered which could
affect the content, and all legal disclaimers that apply to the journal pertain.
 Optimal Design and Analysis of Mechanical Draft Cooling Tower Using

Improved Jaya Algorithm

R. V. Rao*, K. C. More
Dept. of Mech.Engg., S.V. National Institute of Technology, Surat-395007, India
(* Corresponding author: - ravipudirao@gmail.com, Tel- +912612201982)

Highlights

 Improved Jaya algorithm is proposed for design of mechanical draft cooling tower.

 Results of the proposed algorithm are comparatively much better.

 The algorithm’s performance is superior in terms of time and convergence.

Abstract:

A Cooling tower is an imperative component of industrial plants. The minimization of energy

related expenses is critical for conservation of resources and energy savings. Hence, the present

study explores the use of an improved Jaya algorithm called self-adaptive Jaya algorithm for

optimal design of cooling tower from economic facets. In this work, six different examples are

considered in the design optimization of mechanical draft cooling tower. Various researchers

have attempted the same mathematical models by using different methods like Merkel method,

Poppe method and artificial bee colony (ABC) algorithm. The results achieved by using the

proposed self-adaptive Jaya algorithm are compared with the results achieved by using the

Merkel method, Poppe method, ABC algorithm and basic Jaya algorithm. The proposed self-

adaptive Jaya algorithm determines the population size automatically and the user need not tune

the population size. The proposed self-adaptive Jaya algorithm is proved better as compared to

the other optimization methods with respect to achieving the optimal value of the objective

function at less computational effort.

Page 1 of 35
Keywords: Cooling tower, Cost optimization, Merkel method, Poppe method, Artificial bee

colony algorithm, Jaya algorithm, self-adaptive Jaya algorithm.

Nomenclature
a correlation coefficient for the estimation of enthalpy of saturated air, hsa
afi surface area per unit volume (m−1)
Afr cross-sectional area of the cooling tower (m2)
b correlation coefficient for the estimation of enthalpy of saturated air, hsa
c correlation coefficient for the estimation of enthalpy of saturated air, hsa
Ccap,CT installed capital cost of the cooling tower (US$)
CCTF fixed cooling tower cost (US$)
CCTMA incremental cooling tower cost based on air mass flow rate (US$ s/kg)
CCTV incremental cooling tower cost based on tower fill volume (US$/m3)
CkCT,V disaggregated variables for the capital cost coefficients of cooling towers
Cop annual operating cost (US$/yr)
Ce unit operating cost of electricity (US$/J)
cj variables for the Merkel number calculation
cp specific heat at constant pressure (J/kg K)
cw unit operating cost of make-up water (US$/kg)
d correlation coefficient for the estimation of enthalpy of saturated air, hsa
dl variables used in the calculation of the loss coefficient
ek coefficient cost for different fill types
G mass velocity (kg/m2 s)
ha enthalpy of air (J/kg dry air)
hsa enthalpy of saturated air at the local bulk water temperature (J/kg dry air)
hd mass transfer coefficient (kg/m2 s)
HY yearly operating time (s/yr)
K loss coefficient
Kf annualized factor of capital cost (yr−1)
L fill height (m)

Page 2 of 35
M mass flow rate (kg/s)
mav air-vapor flow rate (kg/s)
Me Merkel number
Ncycles number of cycles of concentration
P electric power (W)
Ptot total pressure (Pa)
PV vapor pressure of water (Pa)
Q heat load (W)
T temperature (K or ◦C)
TA dry-bulb temperature of the air (◦C)
TAC total annual cost (US$/yr)
TCH Chebyshev constant
TMIP minimum process inlet temperature (◦C)
TMOP minimum process outlet temperature (◦C)
TW water temperature (◦C)
TWB wet-bulb temperature of the air (◦C)
v velocity (m/s)
w mass-fraction humidity of moist air (kg of water/kg of dry air)
yk binary variable for the fill type
ymfd binary variable for the selection of mechanical forced draft
ymid binary variable for the selection of mechanical induced draft
Greek symbols

Δh local enthalpy difference (J/kg dry air)

ΔP pressure drop (Pa)

ΔTmin minimum allowable temperature difference (◦ C)

ɳf fan efficiency

ρ density of air–water vapor mixtures (kg/m3)

Subscripts

a dry air

av air-vapor

bw blow-down water

Page 3 of 35
fi fill

fin fan inlet

i index to denote the temperature interval, i=1,. . ., 4

in at tower inlet

j index used for the constants in the fill type, j=1,. . ., 5

k superscript to denote the fill type, k=1, 2, 3
l index used for the constants in the fill type, l=1,. . ., 6
m mean

n index used in Eq. (40), n =−1, ..., 3

mcl miscellaneous component

min minimum

misc miscellaneous

mw water make-up

out at tower outlet

t total

vp velocity pressure

w water

wb wet-bulb temperature

wbin wet-bulb temperature of the inlet air

wd water drift

wev water evaporated

1. Introduction

Cooling tower is used to release waste heat into the environment and wide spread

utilization of cooling tower can be found power plants, cooling systems, chemical and

Page 4 of 35
petrochemical industries. A cooling tower is equipment used to reduce the temperature of a water

stream by extracting heat from water and emitting it to the atmosphere. Cooling towers make use

of evaporation whereby some of the water is evaporated into a moving air stream and

subsequently discharged into the atmosphere. As a result, the remainder of the water is cooled

down significantly. Cooling towers are able to lower the water temperatures more than devices

that use only air to reject heat, like the radiator in a car, and are therefore more cost-effective and

energy efficient. The COP and energy efficiency ratio (EER) of air cooled chillers is 50% less

than the water cooled chillers [1]. They represent a relatively inexpensive and dependable means

of removing low grade heat from cooling water [2].

Common applications for cooling towers include providing cooled water for air-

conditioning, manufacturing and electric power iteration. The smallest cooling towers are

designed to handle water streams of only a few gallons of water per minute supplied in small

pipes like those might see in a residence, while the largest cool hundreds of thousands of gallons

per minute supplied in pipes as much as 15 feet (about 5 meters) in diameter on a large power

plant.

In last past few decades the interest of researchers is growing in the field of thermo

economic design optimization of cooling towers. Finite set of discrete variables method was

applied to a natural draft cooling tower by Reinschmidp and Narayanan [3]. By using sequential

quadratic programming (SQP) method Conradie et al. [4] obtained an optimal design cost of dry-

cooling system for the thermal plant application. Ecker and Wiebking [5] investigated the

optimal design of dry-type natural-draft cooling towers using geometric programming (GP)

method.

Page 5 of 35
 The condenser water system was optimized by Howell [6] to provide maximum plant

efficiency during part-load conditions. An exact finite element analysis was done by Pieczara [7]

for the buckling behavior of tower in which a basic genetic algorithm (GA) was applied to

minimize the weight of cooling tower. `Soylemez [8] developed an iteration optimization

technique to study the optimal heat and mass transfer area to minimize the cost of forced draft

cooling tower. Kloppers and Kröger [9] optimized a counter flow natural draft wet-cooling tower

to achieve the optimal cost of the cooling tower. Pornce et al. [10] developed an optimization

model for the design of recirculating cooling water system to minimize the total annual cost.

Ataei et al. [11] described the optimization aspects of a mechanical draft cooling tower and

variety of packing material were used for optimizing the heat transfer.

An objective function to minimize the total annual cost of the cooling tower was

formulated by Serna et al. [12] to develop a model of mechanical draft cooling tower by using

Mixed Integer Non-Linear Programming (MINLP) technique, while a cooling tower assisted

vapor compression refrigeration system was optimized by Sayyaadi and Nejatolahi [13] using

Non-dominated sorting genetic algorithm (NSGA-II) by considering the total exergy and the

overall cost of the system as objectives. Rao and Patel [14] optimized a mechanical draft

counter-flow cooling tower by using an optimization technique called artificial bee colony

(ABC) to minimize the overall cost for heavy duty requirement. Rubio-Castro et al. [14]

minimized the overall cost of cooling systems by using MINLP technique, while Eusiel et al.

[16] used Poppe method for design optimization of mechanical draft counter flow cooling tower

in which six examples of cooling tower are presented for minimize the total cost. Zou et al. [17]

developed an optimal design for solar enhanced natural draft cooling tower, in which a complete

cost model was offered including capital cost, labor cost, operation cost and maintenance cost.

Page 6 of 35
Cui et al. [18] studied the performances of parallel and serial hybrid ground-source heat pump

(GSHP) systems. Hongfang et al. [19] simulated the heat transfer performance of a 1000 MW

indirect air cooling tower in big power stations with four different wind-break structures. Singh

and Das [20] investigated a mechanical draft cooling tower using various fills for maximizing the

performance of cooling tower. Mondal et al. [22] investigated the thermal performance of a

crosswind-influenced natural draft counter flow wet cooling tower. Singla et al. [22]

investigated the performance of a forced draft cooling tower with expanded wire mesh packing.

The experimental analysis was conducted for controlling the parameters such as mass flow rates

of water and air. Singh and Das [23] proposed a feedback model to predict the parameters for

controlling the performance of a mechanical draft cooling tower. The performance of the cooling

tower was optimized with an inverse feedback model in combination with genetic algorithm.

Singh and Das [24] used an improved constrained inverse optimization method for optimization

of mechanical draft cooling towers. Omid et al. [25] optimized exergy efficiency and total annual

cost of cooling tower using genetic algorithm. Bornman et al. [26] used generalized reduced

gradient optimization method for design optimization of forced draft, direct-contact bulk air

cooling tower. Table 1 shows the summary of literature on the optimization of cooling tower.

From the literature survey, it is observed that various researchers had used different

optimization techniques like MNILP, SQP, GP, Merkel method, Poppe method, GA, NSGA-II

and ABC algorithm for optimization of mechanical draft cooling tower. It is also observed that

all the evolutionary algorithms are probabilistic and it requires common controlling parameters

like; population size and the number of iterations. Beside the common control parameters,

different algorithms require their own algorithm-specific control parameters. The improper

Page 7 of 35
tuning of algorithm-specific parameters either increases the computational effort or yields the

local optimal solution.

The proposed Jaya and self-adaptive Jaya algorithms do not have any algorithm-specific

parameters. Thus, the Jaya and self-adaptive Jaya algorithms can be called as an algorithm

specific-parameter-less algorithm. In the present research, the cooling tower problems are solved

by using Jaya and self-adaptive Jaya algorithms to perceive whether further improvement in

terms of the reduction in the total annual cost of the cooling tower is possible.

In this paper, an improved Jaya algorithm named as “self-adaptive Jaya algorithm” is

proposed and the key feature is that the self-adaptive Jaya algorithm determines the population

size automatically without user intervention. Hence, the user is not required to concentrate on

choosing the population size. The improved Jaya algorithm (i.e. self-adaptive Jaya algorithm) is

selected for this study due to its simplicity, robustness, algorithm-specific parameter free nature

and its ability to get optimal solutions with less number of function evaluations and less memory

requirement. The structure of the paper is as follows: Section 1 presents the introduction to the

cooling tower and the literature review, section 2 describes the proposed optimization

algorithms, section 3 describes the case study of a mechanical draft cooling tower, section 4

presents the results and discussions and section 5 presents the concluding statements.

The main objectives of this work are (i) to optimize the influential parameters of cooling

tower from economic point of view and (ii) to demonstrate the effectiveness of Jaya and self-

adaptive Jaya algorithms in the design optimization of cooling tower. The ability of the Jaya and

self-adaptive Jaya algorithms is demonstrated using different application examples. The results

obtained by using Jaya and self-adaptive Jaya algorithms are compared with those obtained by

Serna-González et al. [12], Eusiel et al., [16] and Rao and Patel [14], respectively using MINLP

Page 8 of 35
(Merkel method), Poppe method and ABC algorithm respectively for the same application

example.

The next section presents the proposed optimization algorithms.

2 Proposed optimization algorithms

2.1 Jaya Algorithm

Jaya algorithm is a recently developed advanced optimization algorithm [27]. The description of

Jaya algorithm is given below:

Let the objective function is Z(x) which is to be maximized or minimized and at any

iteration i, let ‘l’ is number of design variables and ‘n’ is the population size. Let the best and

worst values of the objective function during an iteration are denoted by Z(x)best,i and Z(x)worst,i

respectively. Also, the Xl,best,i and Xl,worst,i are the best and worst values respectively of variable

‘l’ during iteration ‘i’corresponding to Z(x)best,i and Z(x)worst,i in the population. Then the updated

value is computed as per equation (1) [27, 28].

X'l,n,i= Xl,n,i+ r1,l,i (Xl,best,i- │Xl,n,i│) - r2,l,i (Xl,worst,i- │Xl,n,i│) (1)

where, r1,l,i and r2,l,i are two random numbers between [0, 1]. The term “r1,l,i (Xl,best,i-

│Xl,n,i│)” shows the ability of solution to go nearer to the best solution and the term “r2,l,i

(Xl,worst,i- │Xl,n,i│)” shows the ability of the solution to shun the worst solution. X'l,n,i is accepted if

it gives superior function value. For ensuring better exploration of the search space two random

numbers r1,l,i and r2,l,i are used for Jaya algorithm. The absolute value of the candidate solution

(│Xl,n,i│) used in equation (1) helps the algorithm to further increase the exploration ability.

These features make the algorithm to converge towards the global solution rather than towards

the local optimal solution. At the end of an iteration all accepted function values are considered
9

Page 9 of 35
as input to the next iteration. The flowchart of the basic Jaya algorithm is shown in Fig. 1. The

Jaya algorithm is based on the concept that the solution obtained for a given problem should

move towards the best solution and must avoid the worst solution. The algorithm always tries to

get closer to success (i.e. reaching the best solution) and tries to avoid failure (i.e. moving away

from the worst solution) and hence it is named as Jaya (a Sanskrit word meaning victory).

2.2 Self-adaptive Jaya Algorithm

In this paper, an improved version of Jaya algorithm is proposed. The basic Jaya algorithm

requires the common control parameters of population size and no. of iterations. The choice of a

particular population size for appropriate case studies is a difficult task. However, not much

work has been conducted yet on self-adaptive population sizes; hence this aspect is taken up in

this work and the proposed algorithm is named as self-adaptive Jaya algorithm. The key feature

is that the self-adaptive Jaya algorithm determines the population size automatically. Hence, the

user need not concentrate on choosing the population size. Let the random initial population is

(10 * l), where l is the number of design variables then the new population is formulated as [29];

P new = round (P old+ R *P old). (2)

R is a random value between [−0.5, 0.5]; it acts as a comparative population development

rate. The population size may decrease or increase due to negative or positive random value of R.

The flowchart of the proposed self-adaptive Jaya algorithm is presented in Fig. 2. Elitism is

implemented when the population size of the next iterations is larger than the population size of

the present iterations (Pnew>Pold). Then all of the existing population will go into the next

iterations and the optimal solutions in the current population are assigned to the remaining

Pnew−Pold solutions. When the population size of the next iterations is less than the population

10

Page 10 of 35
size of the current iterations (Pnew< Pold), then, only the best population of the current iteration

will go to the next iteration. No change takes place if the population sizes of the next and current

iterations are equal (Pnew = Pold). If the number of population decreases and less than the number

of design variables (DV), then the population size is considered equal to the number of design

variables (if Pnew < DV, then Pnew=DV). Thus, the solutions will not be stuck-up in local optima.

The next section presents a case study of the design optimization of a mechanical draft
cooling tower.

3 Case study

This problem was attempted by Serna-Gonzalez et al. [12], Eusiel et al., [16] and Rao and Patel

[14], in which the objective function total annual cost was considered for a mechanical draft

cooling tower. Three design variables such as ratio of mass of water to mass of air, ratio of mass

of water to cross sectional area of cooling tower and ratio of mass of air to cross sectional area of

cooling tower are considered for design optimization of cooling tower. Serna-Gonzalez et al.

[12] developed a mixed integer nonlinear programming (MINLP) formulation for optimizing

mechanical draft cooling tower which looks in to consideration a group of independent variable

together with a set of constraints. Merkel’s method [12] was used for estimating the size and

performance of the tower. Merkel developed the theory for the thermal evaluation of cooling

towers in 1925 [30]. This work was largely neglected until 1941 when the paper was translated

into English. Since then, the model has been widely applied. The Merkel theory relies on several

critical assumptions to reduce the solution to a simple hand calculation. Because of these

11

Page 11 of 35
assumptions, however, the Merkel method does not accurately represent the physics of heat and

mass transfer process in the cooling tower fill. Merkel’s method is used for estimating the size

and performance of the tower. This method has been widely applied to design these units and is

recommended by international standards. The Merkel’s method [30] consists of an energy

balance, and it describes simultaneously the mass and heat transfer processes coupled through

the Lewis relationship; however, these relationships oversimplify the process because they do

not account for the water lost by evaporation and the humidity of the air that exits the cooling

tower. The critical simplifying assumptions of the Merkel theory are:

• The Lewis factor, Lef , relating heat and mass transfer is equal to 1;

• The air exiting the tower is saturated with water vapor and it is characterized only by its

enthalpy;

• The reduction of water flow rate by evaporation is neglected in the energy balance. The

objective was used to minimize the total annual cost of the tower.

Eusiel et al., [16] attempted the same problem using Poppe method. Poppe and Rögener

[31] developed a more complete and accurate model of a cooling tower, which is commonly

known as the Poppe method. This method considers the effects of Lewis factor and water

evaporation on the air process states along the vertical length of the tower. The Poppe method

also accounts for the possibility of supersaturation of the moist air during the heat and mass

transfer processes. In this method, the Merkel number for the cooling tower is obtained by

solving numerically three simultaneous differential equations governing heat and mass transfer

and air flow in the tower fill. The Poppe’s method [28] avoids the simplifying assumptions made

by Merkel, and consists of differential equations that evaluate the air outlet conditions in terms of

enthalpy and humidity, taking into account the water lost by evaporation and the NTU.

12

Page 12 of 35
 And Rao and Patel [14] used artificial bee colony (ABC) algorithm. ABC algorithm

proposed by Karaboga [32] and ABC algorithm is a nature inspired algorithm based on the

intelligent foraging behavior of honey bee swarm. The algorithm mimics insect’s food searching

ability. The ABC algorithm describes the foraging behavior, learning, memorizing and

information sharing characteristics of honeybees. The honeybee swarms consists of two essential

components (i.e. food sources and foragers) and define two leading modes of the behavior (i.e.

recruitment to a nectar source and abandonment of a source). The colony of artificial bees

consists of three groups of bees: employed bees, onlookers and scouts. The colony of the

artificial bees is divided into two groups, first half of the colony consists of the employed

artificial bees and the second half includes the onlooker bees. Scout bees are the employed bees

whose food source has been abandoned. In ABC algorithm, the position of a food source

represents a possible solution to the optimization problem (value of design variables) and the

nectar amount of a food source corresponds to the quality of the associated solution (fitness

value). More details of Merkel’s method, poppe method and ABC algorithm are available in

Serna-González et al. [12], Eusiel et al., [16] and Rao and Patel [14], respectively.

Serna-Gonzalez et al. [12] presents six examples of mechanical draft counter flow

cooling tower. The design specifications and process constraints for the six examples were

presented with example 1 taken as a base case. Table 2 shows the input conditions for all six

examples. In remaining examples (i.e. examples 2-6) one input variable was altered keeping

other variables at their base value. Forced draft was assumed as draft type and film packing for

filling material. The algorithm was coded in MATLAB 2009a and executed on a PC with a 2.83-

GHz Intel Core 2 Quad CPU and 4 GB of Ram memory.

13

Page 13 of 35
 The objective function is to minimize total annual cost (TAC).

Minimize (3)

The annual operating cost is given by,

(4)

The capital cost of cooling towers is given by [12]:

(5)

The values of cw is US$ 5.283*10-4/kg, ce, is US$ 0.085/kWh, Cctf, is US$ 31,185, CCTMA

is US$ 1097.5 (kg of dry air/s), ɳf is 0.75 and ɳcycle is 4. The ambient pressure is considered

101,325 pa, the specific heat of water Cpw is 4.187 kJkg-1 oc and Hy parameter = 2.934*107

s/year.

Subjected to,

(6)

(7)

(8)

Tain 50 oC (9)

(10)

(11)

(12)

The ranges of design variables are given by,

(13)

(14)

(15)

14

Page 14 of 35
 Fig. 3 shows the schematic diagram of a cooling tower.

4 Results and discussion

To demonstrate and validate the proposed self-adaptive Jaya algorithm, the same case study of

cooling tower (containing six examples) is considered. Table 3 shows the comparison of the

optimal results of cooling tower obtained by Jaya and self-adaptive Jaya algorithms. From Table

3 it is observed that the total annual cost is reduced by 10.19%, 5.3% and 0.411% as compared to

Merkel method, Poppe method and ABC algorithm respectively. Reduction in mass flow rate of

air reduces the mass of evaporated water and mass of blowdown water which in turn reduces the

mass of make-up water. The reduction in the mass of make-up water reduces the make-up water

cost. The longer fill height (Lfi) with the higher loss coefficient per meter depth of fill (Kfi) in the

present approach increases the pressure drop through the fill matrix but at the same time the fill

matrix pressure drop is reduces. So, the combined effect of both the parameters result in

reduction in fill matrix pressure drop which in turn reduces the total pressure drop of air stream

about. The reduction in total pressure drop of air stream reduces the power cost about 40.36%,

50.08% and 5.71% than Merkel’s method, Poppe method and ABC algorithm respectively.

Therefore, 0.4 -10% reduction in operating cost is observed in the present approach. The

reduction in total cost is observed due to reduction in operating and electricity cost. For a given

heat load and inlet air conditions, the model yields the optimal ratio of mass of water to mass of

air, ratio of mass of water to cross sectional area of cooling tower and ratio of mass of air to

cross sectional area of cooling tower, water consumption, power consumption. It is observed that

reducing pressure losses reduces electricity cost. Therefore, the proposed formulation by using

15

Page 15 of 35
Jaya and self-adaptive Jaya algorithm could be placed in the context of integrated cooling water

systems to develop a more reliable optimization framework for cooling systems.

Fig. 4 shows the effect of number of iterations on convergence rate of example 1 using

Jaya and self-adaptive Jaya algorithms. The convergence takes place after 8 th and 7th iterations by

using Jaya and self-adaptive Jaya algorithms respectively. The convergence takes place after 18

iterations in the case of ABC algorithm [14]. For Jaya algorithm the population size is taken as

50 and the required computational time is 18.22s, but the self-adaptive Jaya algorithm required

small average population size of 27 and less computational time of 8.6s for example 1.

In example 2, Tain is decreased from 22 oC to 17 oC. Table 4 shows the comparison of the

optimal results of cooling tower obtained by Jaya and self-adaptive Jaya algorithms for examples

2-6. It can be seen from Table 4 that with the reduction in dry bulb temperature (about 5 oC) of

air, the total annual cost is reduced by 10.35%, 19.75% and 0.73% as compared to Merkel

method, Poppe method and ABC algorithm respectively. Moreover, because of reduction in dry

bulb temperature of air the total cost in the present case is 2.38% lower than the base case.

Example 3 explains the effect of variation of wet-bulb temperature of air i.e. it reduced from 12
o
C to 7 oC. From Table 4 it is observed that the total annual cost is reduced by 14.7%, 1.94% and

6.32% as compared to Merkel method, Poppe method and ABC algorithm respectively.

Moreover, compared to base case, 16.55% reduction in total cost is observed due to increase in

tower approach.

In example 4, minimum process inlet temperature is reduced by 10 oC from base case.

The decrease in minimum process inlet temperature increases the air mass flow rate which in

16

Page 16 of 35
turn increases the total annual cost of the tower. Table 4 shows the comparison of the optimal

results of cooling tower obtained by Jaya and self-adaptive Jaya algorithms. From Table 4 it is

observed that the total annual cost is reduced by 11.97%, 0.90% and 1.64% as compared to

Merkel method, Poppe method and ABC algorithm respectively. So, compared to the base case,

4.64% increment in total cost is observed because of reduction in tower range.

Example 5 explains the simultaneous effect of reduction in tower approach and increment

in tower range with reducing minimum process outlet temperature by 5 oC from base case. Due

to the decreases in minimum process outlet temperature, Merkel number increases which in turn

which increases the total cost of the tower. Table 4 shows the comparison of the optimal results

of cooling tower obtained by Jaya and self-adaptive Jaya algorithms. From Table 4 it is observed

that the total annual cost is reduced by 3.57%, 24.88% and 2.33%as compared to Merkel method,

Poppe method and ABC algorithm respectively. So 90.59% increment in total cost is observed

compared to base case.

In example 6, minimum allowable temperature difference is reduced by 5 oC from base

case. Compared to the base case, in the present example the reduction in tower range is obtained

by increasing the outlet water temperature with a corresponding increase in cooling driving

forces (i.e. tower approach). Thus, as the cooling driving force is increased the cooling range is

decreased. Table 4 shows the comparison of the optimal results of cooling tower obtained by

Jaya and self-adaptive Jaya algorithms. From Table 4 it is observed that the total annual cost is

reduced by 13.2%, 1.21% and 2.32% as compared to Merkel method, Poppe method and ABC

algorithm respectively. Therefore, compared to base case, 20.36% reduction in total cost is

observed in the example 6.

17

Page 17 of 35
 Fig. 5 shows the effect of no. of iterations on convergence rate of example 2 using Jaya

and self-adaptive Jaya algorithms. The convergence takes place after 11th and 9th iterations by

using Jaya and self-adaptive Jaya algorithms respectively. The convergence takes place after 20

iterations in the case of ABC algorithm [14]. The population size is taken as 50 in the case of

Jaya algorithm and the required computational time is 21.3s. The self-adaptive Jaya algorithm

required less average population size of 26 and less computational time is 8.33s for example 2.

Fig. 6 shows the effect of no. of iterations on convergence rate of example 3 using Jaya

and self-adaptive Jaya algorithms. The convergence takes place after 13 iterations by using Jaya

as well as self-adaptive Jaya algorithms. The convergence takes place after 18 iterations in the

case of ABC algorithm [14]. The population size is taken as 50 in the case of Jaya algorithm and

the required computational time is 15.6s. The self-adaptive Jaya algorithm required less average

population size of 25 and less computational time is 8s for example 3.

Fig.7 shows the effect of no. of iterations on convergence rate of example 4 using Jaya

and self-adaptive Jaya algorithms. The convergence takes place after 10 and 8 iterations by using

Jaya and self-adaptive Jaya algorithms respectively. The convergence takes place after 17

iterations in the case of ABC algorithm [14]. The population size is taken as 50 in the case of

Jaya algorithm and the required computational time is 18.34s. The self-adaptive Jaya algorithm

required less average population size of 27 and less computational time is 8.24s for example 4.

Fig. 8 shows the effect of no. of iterations on convergence rate of example 1 using Jaya

and self-adaptive Jaya algorithms. The convergence takes place after 13 and 10 iterations by

18

Page 18 of 35
using Jaya and self-adaptive Jaya algorithms respectively. The convergence takes place after 26

iterations in the case of ABC algorithm [14]. The population size is taken as 50 in the case of

Jaya algorithm and the required computational time is 12.4s. The self-adaptive Jaya algorithm

required less average population size of 28 and less computational time is 8.4s for example 5.

Fig. 9 shows the Effect of no. of iterations on convergence rate of example 1 using Jaya

and self-adaptive Jaya algorithms. The convergence takes place after 11 and 9 iterations by using

Jaya and self-adaptive Jaya algorithms respectively. The convergence takes place after 14

iterations in the case of ABC algorithm [14]. The population size is taken as 50 in the case of

Jaya algorithm and the required computational time is 11.6s. The self-adaptive Jaya algorithm

required less average population size of 20 and less computational time is 8.1s for example 6.

In all the examples it observed that the Jaya and self-adaptive Jaya algorithms give better

results than Merkel method [12], Poppe method [16], and ABC algorithm [14]. The self-adaptive

Jaya algorithm required about 50% less computational time and comparatively small population

size as compared to other optimization algorithms. The results for Jaya and self-adaptive Jaya

algorithms are exactly equal probably because of the reason that the achieved solution is global

optimum. However, in Jaya algorithm we need to perform experiments by changing the

population size to achieve the optimal solution but in self-adaptive Jaya algorithm it is not

necessary to change the population size manually, thus reducing the number of experiments and

computational time. The above discussion is now added in the revised manuscript.

Conclusions

19

Page 19 of 35
A cooling tower is a heat removal device; it is commonly used for circulation of water in power

plants, oil refineries, cooling systems, chemical and petrochemical industries. In the present

work, the optimization aspects of a mechanical draft cooling tower are considered with six

different examples. Design optimization of the considered case study is performed to find the

best possible design from economic point of view. A newly evolved advanced optimization

algorithm called Jaya algorithm is adopted and its improved version of self-adaptive Jaya

algorithm is proposed for the optimal design of a mechanical draft cooling tower. The proposed

self-adaptive Jaya algorithm determines the population size automatically and the user need not

tune the population size.

The same mechanical draft cooling tower model was optimized previously by Serna-

González et al. using Merkel method on the mixed integer non Linear programming (MINLP)

formulation, Eusiel et al. using Poppe method, and Rao and Patel using ABC algorithm. The

results achieved by using Jaya and self-adaptive Jaya algorithms are compared with those

obtained by using the Merkel method, Poppe method, and ABC algorithm. The ability of Jaya

and self-adaptive Jaya algorithms is demonstrated and the convergence behavior of the self-

adaptive Jaya algorithm is found to be more efficient than the convergence behavior of ABC and

Jaya algorithms. The Jaya and self-adaptive Jaya algorithms are proved superior to other

optimization algorithms in terms of results. The proposed self-adaptive Jaya algorithm is proved

superior to the other optimization algorithms in terms of optimal results, convergence and

computational time.

20

Page 20 of 35
6 References

1. Chang, C.K; Fui, Song K. F.; Bin, N. W.; Azizan, M. H., (2013). Comparison of air-

cooled chiller, water-cooled chiller and hybrid evaporative air-cooled chiller in Malaysia

– a review, International Journal of Arts & Sciences; 6 (1): 337-355.

2. Kröger D. G., (1998). Air-Cooled Heat Exchanger and Cooling tower, Penn Well, Tulsa,

OK, USA.

3. Reinschmidp K., Narayanan R., (1975). The optimum shape of cooling towers,

Computers and Structures, 5: 321-325.

4. Conradie A. E., Buyst J. D., Kroger D, G., (1998). Performance optimization of dry-

cooling systems for power plants through SQP methods, Applied Thermal Engineering 18

(1-2): 25-45.

5. Ecker J. G., Wiebking R. D., (1978). Optimal design of a dry-type natural-draft cooling

tower by geometric programming, Journal of Optimization Theory and Applications, 26

(2): 310-320.

6. Howell R., (1998). Chilled water plant optimization based on part-load cooling tower

performance, ASHRAE Fellow University of South Florida Tampa, FL.

7. Pieczara J., (2000). Optimization of cooling tower shells using a simple genetic

algorithm, Structural and Multidisciplinary Optimization, 19: 311–316.

8. Soylemez M.S., (2001). On the optimum sizing of cooling towers, Energy Conversion

and Management, 42: 783-789.

9. Kloppers J. C., Kröger D. G., (2004). Cost optimization of cooling tower geometry,

Engineering Optimization, 36(5): 575–584.

21

Page 21 of 35
10. Ponce-Ortega J. M., Serna-Gonzáleza M., Jiménez-Gutiérrezb A., (2010). Optimization

model for re-circulating cooling water systems, Computers and Chemical Engineering,

34: 177–195.

11. Atai A., Panjeshahi M.H. Gharaie M., (2009). A new algorithm for design of mechanical

draft wet cooling tower, Journal of Applied Science, 9 (3): 561-566.

12. Serna-Gonzáleza M., Ponce-Ortegaa J. M., Jiménez-Gutiérrez B.A., (2010). MINLP

optimization of mechanical draft counter flow wet-cooling towers, Chemical Engineering

Research and Design, 88: 614–625.

13. Sayyaadi H., Nejatolahi M., (2011). Multi-objective optimization of a cooling tower

assisted vapor compression refrigeration system, International Journal of Refrigeration,

34(1): 243-256, 2011.

14. Rao R.V., Patel V.K., (2011). Optimization of mechanical draft counter flow wet cooling

tower using artificial bee colony algorithm, Energy Conversion and Management, 52:

2611-2622.

15. Rubio-Castro E., Serna-González M., Ponce-Ortega J.M., (2011) Optimal design of

effluent-cooling systems using a mathematical programming model, Applied Thermal

Engineering, 30: 2116-2126.

16. Eusiel R-C, Medardo S-G, José M P-O, Miguel A M-C., (2011). Optimization of

mechanical draft counter flow wet-cooling towers using a rigorous model, Applied

Thermal Engineering, 31: 3615-3628.

17. Zou Z., Guan Z., Gurgenci H., (2013(). Optimization design of solar enhanced natural

draft dry cooling tower. Energy Conversion and Management, 76: 945-955.

22

Page 22 of 35
18. Cui W., Zhou S., Liu X., (2015). Optimization of design and operation parameters for

hybrid ground-source heat pump assisted with cooling tower, Energy and Buildings, 99:

253-262.

19. Hongfang G, , Haijun W, Yuqian G, Jianan Y., (2016). A numerical study on the

mechanism and optimization of wind-break structures for indirect air-cooling towers,

Energy Conversion and Management, 108: 43-49.

20. Singh K., Das R. (2016). An experimental and multi-objective optimization study of a

forced draft cooling tower with different fills, Energy Conversion and Management, 111:

417-430.

21. Mondal P. K., Mukherjee S., Kundu B., Wongwises S., (2015) Investigation of the

crosswind-influenced thermal performance of a natural draft counter flow cooling tower.

International Journal of Heat and Mass Transfer.85: 1049–1057.

22. Singla R. K., Singh K., Das R. (2016). Tower characteristics correlation and parameter

retrieval in wet cooling tower with expanded wire mesh packing. Applied Thermal

Engineering, 96: 240–249.

23. Singh K., Das R. (2016). Feedback model to predict parameters for controlling the

performance of a mechanical draft cooling tower. Applied Thermal Engineering, 105:

519–530.

24. Bornman, W., Dirker, J., Arndt, D. C., Meyer, J. P., (2016) Operational energy

minimisation for forced draft, direct-contact bulk air cooling tower through a

combination of forward and first-principle modelling, coupled with an optimisation

platform. Energy, 114: 995-1006

23

Page 23 of 35
25. Singh, K., Das, R., (2017). An improved constrained inverse optimization method for

mechanical draft cooling towers. Applied Thermal Engineering, 114:573-582.

26. Omid, K., S., Shahram, K., Iraj, M., (2017). Energic, Exergic, Exergo-economic

investigation and optimization of Auxiliary Cooling System (ACS) equipped with

Compression Refrigerating System (CRS). Case Studies in Thermal Engineering,

http://dx.doi.org/10.1016/j.csite.2016.12.004

27. Rao R.V., (2016). Jaya: A simple and new optimization algorithm for solving constrained

and unconstrained optimization problems, International Journal of Industrial

Engineering Computations, 7(1):19-34.

28. Rao R.V., More K.C., Taler J., Oclon P., (2016). Dimensional optimization of a micro-

channel heat sink using Jaya algorithm, Applied Thermal Engineering, 103: 572–582.

29. Teo, J., (2006). Exploring dynamic self-adaptive populations in differential evolution.

Soft Comput, 10: 673–86.

30. Merkel, F., (1925). Verdunstungsku¨hlung. VDI-Zeitchrift, 70: 123– 128.

31. Poppe, M., Rögener, H., (1991). Berechnung von Rückkühlwerken. VDI-Wärmeatlas, 1-

15.

32. Karaboga, D., Basturk, B., (2003). A powerful and efficient algorithm for numerical

function optimization: artificial bee colony (ABC) algorithm. Journal of Global

Optimization, 39:459–71.

24

Page 24 of 35
Fig. 1. Flowchart of Jaya algorithm [28]

25

Page 25 of 35
Fig. 2 Flowchart of self-adaptive Jaya algorithm

26

Page 26 of 35
Fig 3 Schematic diagram of a cooling tower model [12]

60000

59900
Total annual cost (US $)

ABC [14]
59800

59700 Jaya

59600
self-
59500 adaptive
Jaya
59400

59300
0 10 20 30 40
No. of iterations

Fig. 4 Convergence of ABC, Jaya and self-adaptive Jaya algorithms for example 1.

27

Page 27 of 35
 60300

Total annual cost (US $)

59800
ABC [14]

59300
Jaya
58800
self-
58300 adaptive
Jaya

57800
0 10 20 30 40
No. of iterations

Fig. 5. Convergence of ABC, Jaya and self-adaptive algorithms for example 2.

56000
55000
Total annual cost (US $)

54000 ABC [14]

53000
52000 Jaya
51000
50000
49000
0 10 20 30 40
No. of iterations

Fig. 6 Convergence of ABC, Jaya and self-adaptive Jaya algorithms for example 3.

28

Page 28 of 35
 65500
65000

Total annual cost (US $)

64500 ABC
[14]
64000
63500
Jaya
63000
62500
62000
61500
0 10 20 30 40
No. of iterations

Fig. 7 Convergence of ABC, Jaya and self-adaptive Jaya algorithms for example 4.

117500
Total annual cost (US $)

ABC [14]
116500

115500 Jaya
114500
self-
113500
adaptive
112500 Jaya
0 10 20 30 40
No. of iterations

Fig. 8 Convergence of ABC, Jaya and self-adaptive Jaya algorithms for example 5.

29

Page 29 of 35
 50000
ABC

Total annual cost (US $)

49500 [14]

49000 Jaya

48500
self-
48000 adaptiv
e Jaya
47500

47000
0 10 20 30 40
No. of iterations

Fig. 9 Convergence of ABC, Jaya and self-adaptive Jaya algorithms for example 6.

30

Page 30 of 35
 Table 1 Summary of literature on the optimization of cooling tower.

References Cooling tower type Optimizing parameter Optimization method

Reinschmidp and Natural draft Shape Finites set of discrete
Narayanan [3] variable method
Conradie et al. [4] Dry cooling system Cost Sequential Quadratic
Programming
Ecker and Wiebking Dry type natural draft Total annual cost Geometric
[5] Programming
Howell [6] Wet cooling tower Plant efficiency Simple control
algorithm
Pieczara [7] Natural draft Volume of cooling Genetic algorithm
tower
Soylemez [8] Forced Draft Cost Iteration method
Kloppers and Kroger Natural draft wet- Cost Leap-frog
[9] cooling tower optimization
programme with
constraints
Pornce et al. [10] Forced draft wet Total annual cost Mixed integer non-
cooling tower linear programming
Ataei et al. [11] Mechanical draft wet Heat transfer Computational
cooling tower method using Visual
Studia.Net 2003
(C++)
Serna et al. [12] Mechanical draft Total annual cost Merkel’s method with
counter flow wet Mixed integer non-
cooling tower linear programming
Sayyaadi and Natural draft cooling Thermodynamic and Non-dominated
Nejatolahi [13] tower economic sorting genetic
algorithm
Rao and Patel [14] Natural draft cooling Total annual cost Artificial bee colony
tower algorithm
Rubio-Castro et al. Mechanical draft Total annual cost Mixed integer non-
[15] counter-flow wet linear programming
cooling tower
Eusiel et al. [16] Mechanical draft Total annual cost Poppe method
counter-flow wet
cooling tower
Zou et al. [17] Natural draft dry Total annual cost Global Search and

31

Page 31 of 35
 cooling tower Multi Start solver
Cui et al. [18] Natural draft dry Cost TRNSYS 16. A
cooling tower Transient System
Simulation Program
Hongfang et al. [19] Indirect-air cooling Efficiency of power Computational fluid
tower plant in a windy area. dynamics (CFD)
Singh and Das [20] Forced draft cooling Effectiveness and Elitist Non-
tower Evaporation rate Dominated Sorting
Genetic Algorithm
(NSGAII)
Mondal et al. [21] Natural draft counter Thermal performance Experimental analysis
flow cooling tower
Singla et al., [22] counter flow forced controlling parameters Differential evolution
draft cooling tower mass flow rates of
water and air satisfy a
given value of the
Merkel
number
Singh and Das [23] Mechanical draft Mass flow rate and Inverse feedback
cooling tower power consumption model and Genetic
algorithm
Singh and Das [24] Mechanical draft power consumption improved constrained
cooling tower inverse optimization
Omid et al. [25] Dry cooling tower Exergy efficiency and Genetic Algorithm
Total annual cost
Bornman et al. [26] Forced draft, direct- Power consumption Generalized reduced
contact bulk gradient optimization
air cooling tower method

32

Page 32 of 35
 Table 2 Input conditions for case study 1 [12]:

Example Example 2 Example 3 Example 4 Example 5 Example 6

1
Q (kW) 3400 3400 3400 3400 3400 3400
TAin (oc) 22 17 22 22 22 22
TWBin (oc) 12 12 7 12 12 12
TMPI (oc) 65 65 65 55 65 65
TMPO (oc) 30 30 30 30 25 30
(oc) 10 10 10 10 10 5
win (kg 0.0047 0.0047 0.0047 0.0047 0.0047 0.0047
water/kg dry
air)
ρin (kgm-3) 1.1925 1.1925 1.1925 1.1925 1.1925 1.1925

33

Page 33 of 35
Table 3 Optimal cooling tower geometry using different optimization methods for example 1.

Merkel method Poppe method ABC [14] Jaya self-adaptive

[12] [16] Jaya
mw (kgs-1) 25.720 29.984 25.720 25.594 25.594
ma (kgs-1) 31.014 43.237 26.070 28.06 28.06
mw / ma 0.829 0.682 0.987 0.912 0.912
mmw(kgs-1) 1.541 1.123 1.521 1.508 1.508
mwev(kgs-1) 1.156 0.842 1.078 1.056 1.056
mbw(kgs-1) 0.334 - 0.327 0.300 0.300
mwd(kgs-1) 0.051 - 0.051 0.051 0.051
TWin(oc) 50.000 38.886 50.000 50.000 50.000
TWout(oc) 20.000 20.000 20.000 20.000 20.000
TAout(oc) 37.077 28.387 39.000 40.100 40.100
Range (oc) 30.000 18.886 30.000 30.000 30.000
Approach 8.000 8.000 8.000 8.000 8.000
(oc)
Lfi(m) 2.239 1.273 2.418 2.450 2.450
Afr (m2) 8.869 10.173 8.869 8.750 8.750
Kfi 21.946 - 22.070 24.450 24.450
Δpfi(pa) 280.331 - 211.350 206.418 206.418
ΔPmics (pa) 36.189 - 25.750 23.560 23.560
Δpt(pa) 527.640 - 395.260 390.077 390.077
P (hp) 24.637 29.733 15.500 14.550 14.550
Me 3.083 2.367 3.720 3.312 3.312
Cmw (US$) 23885.190 17412.400 23449.110 23374.000 23374.000
Cpower 12737.540 15373.000 8013.020 7555.100 7555.100
Cop (US$)
(US$) 36622.703 32785.400 31462.130 30930.400 30930.400
K Cap,ct 29442.436 29866.700 28110.040 28398.900 28398.900
TAC (US$) 66065.139 62652.100 59572.170 59329.300 59329.300

34

Page 34 of 35
 Table 4 Optimal cost of cooling tower geometry using different optimization methods for
examples 2-6.
Merkel Poppe ABC [14] Jaya self-
method method adaptive
Example TAC (US$) [12]
64604.640 [16]
72,171.700 58343.930 57916.000 57916.000 Jaya

2 Computational NA NA NA 21.3 8.33

time
Convergence NA NA 20 11 9

Example TAC (US$) 58716.800 50448.600 52848.170 49508.000 49508.000

3
Computational NA NA NA 15.6 8
time
Convergence NA NA 18 13 13

Example TAC (US$) 70533.909 62652.100 63123.900 62087.000 62087.000

4
Computational NA NA NA 18.34 8.24
time
Convergence NA NA 17 10 8

Example TAC (US$) 117278.680 150539.600 115786.700 113080.000 113080.000

5
Computational NA NA NA 12.4 8.4
time
Convergence NA NA 26 13 10

Example TAC (US$) 54456.026 47826.500 48370.880 47246.000 47246.000

6
Computational NA NA NA 11.6 8.9
time
Convergence NA NA 14 11 9

 Bold values indicates the best results

35

Page 35 of 35