ARTICLE

Identification of design criteria for

district cooling distribution network
Gerardo L. Augusto1,*, Alvin B. Culaba1, Raymond R. Tan 2
1
Department of Mechanical Engineering, De La Salle University, Manila, Philippines
2
Department of Chemical Engineering, De La Salle University, Manila, Philippines

T
he hydraulic calculation and system simulation of solved using multivariable Newton-Raphson method. The
two (2) district cooling distribution network linearized equations revealed that coefficient matrices formed
models with and without secondary lines are between the two networks were different from each other which
presented in which the theoretical system pressure suggested that different decomposition algorithms must be used
drop, system flow rate and flow rate requirements to ensure that solution vectors are properly determined.
in each energy transfer station were determined considering a Distribution networks with and without secondary lines showed
temperature difference set-point of 9°C. Variable primary flow that Jacobian matrix can be solved using singular-value
of chilled water pumping arrangement was used to improve decomposition and LU decomposition methods, respectively.
energy usage and to eliminate the need for a separate distribution The results of system simulation defined the coordinates of
pump in the network. system characteristic curve which indicated the best efficiency
points during selected part load and full load conditions.
Pipe sizing and friction factor identification were necessary
to determine the frictional coefficients of distribution network An optimization technique such as exhaustive search
components. Implicit Colebrook-White equation was used to method was used to determine the piping network design criteria
determine the pipe friction factor. The method of least squares that could give minimum overall costs of construction and
and Cholesky decomposition method were also adopted to maintenance of piping system. Numerical results show that
derive new set of pump characteristic curve considering that distribution network with secondary lines yields minimum
pumps are modulated at its best efficiency points. The governing overall costs as compared with piping network without
equations consisted of mass conservation and energy equations secondary line considering that nominated loads, pipe lengths
in the form of pump characteristic curve and distribution and fittings, normalized annual demand factor and costs
network characteristics. The system of nonlinear equations was parameters associated with components of objective function are
held constant.

*Corresponding author KEYWORDS

Email Address: gerardo_augusto@dlsu.ph
Submitted: March 24, 2013 District Cooling System, Distribution Network, Variable Primary
Revised: July 28, 2013 Flow Pumping System, Multivariable Newton-Raphson Method,
Accepted: August 26, 2013 Singular-Value Decomposition Method, LU Decomposition
Published:October 31, 2013 Method, Cholesky Decomposition Method
Editor-in-charge: Leonardo Q. Liongson
Reviewers:
Menandro S. Berana
INTRODUCTION
Marrick C. Neri
District cooling system (DCS) uses thermal energy in the

182 Philippine Science Letters Vol.6 | No.2 | 2013

form of chilled water from a central plant to multiple buildings increase performance during piping system design (Hodgson and
through a distribution network of underground pipes for use in Walters 2002; Walters 2002). To implement optimization
space cooling. The cooling process takes place in the central methods in pumping system design, Hodgson et al. introduced
plant which eliminates the use of separate conventional cooling hydraulic solver and optimizer for flow analysis and
system from each building. optimization technique, respectively. The hydraulic solver
operated as a subroutine which is called repeatedly by the
The district cooling system primarily consists of three (3) optimizer to evaluate a series of designs. However, the details of
components, namely; the central cooling plant, the distribution hydraulic calculation were not specified.
system and the customer’s energy transfer station (ETS). The
cooling equipment, cooling towers, power generation, and Recent study of distribution network used numerical
thermal storage if any, are the main components of a central optimization to determine the optimal pipeline size and pump
cooling plant. The distribution system is a piping network that combination for a 90.4 mile long, treated-seawater transfer line
transfers the cooling medium or chilled water from the central in the Kingdom of Saudi Arabia. The design study described the
cooling plant to different energy transfer stations of individual best configuration of pipeline material and sizes including
building at controlled rates. The customer’s energy transfer pumps to meet both seawater requirements and the flexibility to
station consists of plate type heat exchangers (PHE), secondary accommodate an increased demand in the future. The system
pumping system, and chilled water piping for the building. constraints were maximum pipe operating pressure and
minimum thickness for a given pipe diameter (Thorp and Olsen
A large-scale district cooling plant could save energy 2008).
consumption costs from 25% to 40% as compared with the sum
of conventional centralized air-conditioning system of each In district cooling with closed-loop system, the appropriate
building (Bin Shafar 2011). However, as the distribution network pipe size of distribution network shall be determined by the
is often the most expensive portion and requires large initial economic standpoint of life-cycle cost for construction,
investment cost of the district cooling system, careful design is operation and maintenance. However, this study is seldom done
needed to optimize its use. because the calculation requires tedious works. Instead, the
criteria that have evolved from practice are usually used for
Over the years, the codes, standards and regulations design. These design criteria are maximum flow velocity and
essential for the piping industry are well established (ASHRAE pressure drop limits (ASHRAE 2008; ASHRAE 2013).
2008; ASME 1997; ASME 2009). Studies on pipeline design,
installation, and operation are widely discussed by several This study aimed to determine the piping network design
investigators in which application of appropriate codes and criteria that would minimize the sum of initial investment,
system hardware components to meet safety standards are maintenance, and operational costs. The criteria that have
specified (Antaki 2003; McAllister 2005). Antaki (2003) evolved from practice were still used in the piping design. The
discussed the fundamental principles in materials, design, design constraints were pressure drop limit, maximum flow
fabrication, inspection, testing, operation, maintenance and velocities at primary line, secondary line, tertiary line if any, and
integrity of plant piping. However, even prior to construction, a at plot take-off (PTO). The calculation method was divided into
thorough investigation using system simulation and optimization three (3) modules namely; friction factor identification, chilled
is needed to imitate the configuration of real system and to water pumping system simulation, and optimization. The system
understand the performance of distribution network at different pressure drop, system flow rate and flow rate requirements in
operating conditions with minimum cost. each energy transfer station (ETS) were determined to properly
evaluate the performance of complex piping system before any
Stoecker (1989) used system simulation of a simple water hardware is procured. Optimization technique such as exhaustive
pumping system which transfers fluid from one reservoir to search method was used to solve the objective function and to
another. The flow rates and pressure drop were obtained from a identify the piping network design criteria.
determined system of equations using multivariable Newton-
Raphson method. In a separate study, he also introduced MATERIALS AND METHODS
optimization procedures of water chilling unit (e.g., refrigeration
system with cooling tower) being optimized for minimum first District Cooling System and Distribution Network Models
cost.
Figure 1 shows an overall view of district cooling system
For several decades, numerical optimization techniques with temperature limits at full load condition. Variable primary
have been widely used on structural applications but their use in flow pumping scheme was used to reduce operating and
distribution network is quite new (Schmit 1960; Vanderplaats investment costs since it requires smaller space due to fewer
1999). Combining both system flow analysis and optimization plant components. The pumping arrangement is shown in Fig. 2.
techniques yields a powerful approach to minimize costs and to The distribution network with temperature difference set-point

Vol.6 | No.2 | 2013 Philippine Science Letters 183

of 9°C was used as the reference for plant design and operation components. Knowing these coefficients, the energy equations
for chilled-water systems. To compare the piping network design due to pipe friction can be derived. Combining mass
criteria of two distribution network models, the schematic conservation, pump performance curve and energy equations
diagrams DNModel-01_12 and DNModel-03_12 were used as yielded a system of nonlinear equations. The results of
shown in Fig. 3 and Fig. 4, respectively. Figure 4 shows the calculation defined the coordinates of system characteristic
presence of secondary lines in the network. These network curve at full load and part load conditions. The coordinates of
models were used as benchmarks for fluid flow analysis. In both system characteristic curve were sets of system pressure drop,
cases, twelve (12) energy transfer stations (ETS) in each network system flow rate, and individual flow rate in each energy transfer
were used with total pipe length of approximately 5.2 km. The station at selected operating conditions.
nominated cooling loads are listed in Tab. 1. Table 2 shows the
breakdown of pipe lengths installed in two (2) distribution Mass Conservation. Mass conservation was introduced at
network models. To properly accommodate the cooling loads, district cooling plant main entry point and at the nodes where the
three (3) plate type heat exchangers were placed in each intersection between primary and secondary lines existed. The
building. continuity equation included the time rate of change of mass
within the control volume and mass flux as shown below.
Identification of Piping Network Design Criteria

Methods of hydraulic calculation have been developed

primarily for domestic water distribution system (Jeppson 1977;
Stephenson 1981). However, these may apply to chilled water
distribution network of district cooling system with appropriate
modifications in which the algebraic sum of head losses around
any closed loop is neglected. Prior to head loss calculation, the
friction factor is determined using an implicit Colebrook-White
method (GL Augusto, unpublished observations). Pertinent
variables such as system pressure drop and volume flow rates of In an incompressible fluid, the chilled water density is
chilled water in a thermal system operating at steady-state must constant. If a steady-state condition exists then the continuity
be calculated to determine the required pumping performance equation at any arbitrary volume can be reduced to Eq. 1 which
characteristics. Calculating flow rates and system pressure in a means that the divergence of velocity field q is zero everywhere.
piping network with branches, fittings, pumps and plate type
heat exchangers could be difficult without the aid of a computer. Eq. 1
Computer tools needed for the source codes are written in
Fortran programming and Basic programming using macro The simplified mass balance equations are shown below
platform of MS Excel. The hydraulic calculation and chilled
water pumping system simulation were carried-out in an Eq. 1a
iterative fashion considering that chilled water supply and return
temperatures throughout the network were maintained. Initial
estimates of system pressure and system flow rate were
determined. Refined values of these parameters including flow
rate in each energy transfer station were calculated, and adjusted Eq. 1b
to ensure that the cooling loads were met. The system simulation
was treated as continuous and deterministic. This research work
is based on the following fluid mechanics assumptions namely; Pump Performance Curve. The pump performance curve
one-dimensional flow, incompressible flow, steady-state was taken from the technical data sheet of pump model that
condition, and no chemical reactions. An optimization approach satisfies the chilled water cooling flow rate requirements of
commonly known as exhaustive search method was used to water-cooled chiller with temperature difference of 9°C. The
determine the piping network design criteria that provided the pump flow curve can be generally expressed as a function of
minimum overall cost. Figure 5 shows the logical structure of volume flow rate as shown in Eq. 2.
identifying the piping network design criteria using the
combined fluids engineering and optimization approach.
Eq. 2
Identification of Governing Equations

Prior to chilled water pumping system simulation, pipe Energy Equations due to Pipe Friction. The energy
sizing and friction factor identification were necessary to equations due to pipe friction were derived based on the
determine the frictional coefficients of distribution network

184 Philippine Science Letters Vol.6 | No.2 | 2013

 Figure 1. Overview of district cooling system with temperature limits

Figure 2. Variable primary flow of chilled water pumping arrangement

Vol.6 | No.2 | 2013 Philippine Science Letters 185

 Figure 3. Schematic diagram of distribution network DNModel-01_12

Figure 4. Schematic diagram of distribution network DNModel-03_12

186 Philippine Science Letters Vol.6 | No.2 | 2013

 Figure 5. Logical structure of optimum pumping system for distribution network

configuration of distribution network from central cooling plant Enthalpy Balance at Return Line. The enthalpy balance
to respective energy transfer station of each building which can was neglected because chilled-water supply temperature (Ts) and
be generally expressed as temperature difference set point (δT) were held constant
throughout the network at 4.5 C and 9°C, respectively.
Eq. 3 Evaluating the enthalpy balance of return line at the junction
before entering the district cooling plant reduced the equation to
mass conservation.
where, i indicates the number of pipe components and j the
number of ETS in the distribution network. The notations ci and
ωi denote the frictional coefficients of each pipe components and Eq. 4
flow rate along the distribution network, respectively. The
friction factor denoted by f is a function of Reynolds number and
pipe relative roughness. The frictional coefficient can be
expressed in terms of friction factor and a ratio of pipe length
and pipe diameter. These parameters are given by
The above equations require that the following conditions
Eq. 3a are satisfied.

Eq. 3b

Vol.6 | No.2 | 2013 Philippine Science Letters 187

 Subsequently, based on the above equations, the Affinity
Laws for geometrically similar pumps such as head, torque and
volume flow rate were derived as shown in Eq. 9a, Eq. 9b and
Eq. 9c, respectively.
Eq.4a

The fluid thermal properties such as mass density, absolute Eq. 9a

viscosity and specific heat at constant pressure are functions of
chilled water supply temperature.
Eq. 9b
Eq. 5a

Eq. 5b Eq. 9c

Eq. 5c
Numerical Methods

Cholesky Decomposition Method. In a large-scale district

Dimensional Analysis
cooling plant, the chilled-water pumps are connected in parallel.
To derive the pump performance curve that would represent a set
Buckingham Π Theorem. As variable primary flow
of similar pumps operating at full load, the method of least
pumping system was used, the pumps were modulated at its best
squares was used. But, if the total cooling load downstream of
efficiency point at different operating conditions. Dimensional
the network decreases then the pumps operate at part load
analysis with a generalized approach commonly known as
condition. In this case, the pressure rise developed by the pumps
Buckingham Π Theorem was adopted to derive pertinent
is a function of both flow rate and number of operating pumps.
parameters to predict the performance of pump model under
As system equilibrium occurs at pump’s best efficiency point,
selected conditions of operation. When the length scale was
the pump rotational speed can be identified using Affinity Laws.
neglected the following variables were used:
The method of least squares can be expressed as
Eq. 6

The dimensionless parameters can be expressed as

Eq. 10
Eq. 7

Simplifying the above parameters, the new functional where, pk(Q) is a theoretical pump curve which is a
relations yield polynomial function of flow rate with an exponential index,
m=3.
Eq. 11
Eq. 8a

Eq. 8b

Eq. 8c

188 Philippine Science Letters Vol.6 | No.2 | 2013

 As the pump curve should be fitted through new sets of Simplifying the governing equations could lead to system
coordinates, a necessary condition for u to be minimum is to of nonlinear equations in which the variables s be an entire
satisfy the following: vector denoted as sjjand F be the entire vector of functions Fii. It
can be solved using multivariable Newton-Raphson (MNR)
method where the functions can be expanded in Taylor series of
Eq. 12 the form

Eq. 17

Neglecting higher order terms O(δs2)and by considering the

left-hand side equation equal to zero, a set of equations for the
corrections δs that move each function closer to zero is
In matrix form, the normal equations are shown below.
simultaneously obtained. Hence, Eq. 17 can be reduced to an
equation with Jacobian matrix J as shown below.

Eq. 13
Eq. 18

In vector form, it can be rewritten as

If matrix Eq. 18 is a determined system of equations, then
LU decomposition method can be used. Otherwise, it would be
more appropriate to use singular-value decomposition method
where the corrections δsll are then added to the solution vector
Eq. 14 and the process is iterated to convergence. The convergence
criteria are function convergence and root convergence with
tolerance values of 1.0x10-8.
As the elements of coefficient matrix are symmetric and
positive-definite, then it has positive real eigenvalues which
suggests that the inverse of coefficient matrix is equal to its Eq. 19
transpose. Hence, Cholesky decomposition method can be used
in solving the unknown vector parameter denoted as a.
Optimization Method
Multivariable Newton-Raphson Method. The continuity
equation, pump performance curve and energy equations due to In optimization using exhaustive search method, the values
pipe friction can be expressed as functional relations to be of objective function were determined and conclusions were
zeroed as shown in Eq. 15. drawn from the results of the calculation at various combinations
of independent variables. The objective function denoted as
is a cost function derived from engineering economics.
Eq. 15 Comparative cost analysis such as present worth pattern method
was used. To further verify the results of computation, the annual
cost pattern method was also adopted for comparison. The cost
function variables included the (a) the initial investment costs for
The independent variables were system pressure drop, pumps, plate type heat exchangers (PHE), and distribution
system flow rate and cooling flow rate requirements in each network, (b) installation costs for item (a), (c) excavation cost
ETS. The above equation can be rewritten as for distribution network including backfill and surface
restoration, (d) operation and maintenance costs, (e) payroll
taxes including taxes for property and insurance, and (f)
depreciation cost. The optimization problem can be written as
Eq. 16

Eq. 20

Vol.6 | No.2 | 2013 Philippine Science Letters 189

subject to the following design constraints, DNModel-03_12, respectively.

The results of system simulation indicate the best efficiency

points during selected part load and full load conditions for
Eq. 21
distribution network models DNModel-01_12 and DNModel-
03_12 as shown in Fig. 6 and Fig. 7, respectively. Table 4
Eq. 22 summarizes the description of best efficiency points at different
operating conditions. PC_1 indicates the general pump curve of
all pumps operating at rated speed. Pump curves denoted as
PC_2 and PC_3 represent the upper and lower limits with six (6)
The equality and inequality design constraints were piping and five (5) number of pumps operating at rated speed.
network design criteria such as pressure drop limit and velocity PC_Predicted illustrates the pump curve when operating at full
limits, respectively. The pressure drop limit was set at 100 Pa/m
(ASHRAE 2008). The velocity limits for primary line, secondary load with a diversity factor of 80%. PC _Mod shows the pump
line and plot take-off should not exceed 3.0 m/s with a search curve when operating at part load with nominated loads reduced
interval of 0.05 m/s. Table 3 shows the normalized annual power to 72% and diversity factor of 80%. At full load, distribution
load demand factor. network DNModel-01_12 requires higher pump head i.e., 28.3
m as compared with DNModel-03_12 which is 21.7 m.
RESULTS AND DISCUSSIONS
Optimization Method
Hydraulic Calculation
Optimization technique commonly known as exhaustive
A determined system of nonlinear equations was formed search method was used in identifying the piping network design
when consolidating all the governing equations using criteria that would give minimum overall costs of construction
distribution network DNModel-01_12. An overdetermined and maintenance of two (2) distribution network models. Table 5
system of nonlinear equations was derived when evaluating shows pertinent cost parameters that are associated with
distribution network DNModel-03_12. In system simulation, components of objective function. The optimization results for
multivariable Newton-Raphson method with LU decomposition distribution network models are listed in Tab. 6. Pipe sizes for
and singular-value decomposition methods were used in two (2) distribution network models were obtained when the
calculating the independent variables for DNModel-01_12 and network design criteria became known. The results are listed in

Figure 6. Best efficiency points at part load and full load conditions for DNModel-01_12

190 Philippine Science Letters Vol.6 | No.2 | 2013

 Figure 7. Best efficiency points at part load and full load conditions for DNModel-03_12

Tab. 7. 1. Prior to chilled water pumping system simulation, pipe

sizing and friction factor identification were necessary
Figure 8 illustrates the profile of overall costs for to properly determine the frictional coefficients of
distribution network DNModel-01_12. This is based on present distribution network components. The pipe friction
worth pattern method as a function of limiting velocity at factor was calculated using an implicit Colebrook-
primary line considering the velocity limit of 1.75 m/s at plot White equation.
take-off is held constant. The annual cost pattern method was
also adopted for comparison as shown in Fig. 9. The results 2. A determined system of equations was formed when
suggest that minimum overall cost occurred when the velocity evaluating all the governing equations derived from a
limit at primary line was 3.0 m/s. distribution network without secondary line.
Multivariable Newton-Raphson and LU decomposition
The overall costs profiles for distribution network methods were used to determine the independent
DNModel-03_12 based on present worth and annual cost pattern variables.
methods are shown in Figures 10 and 11. The results indicate
that minimum overall cost occurred when the velocity limit at 3. An overdetermined system of nonlinear equations was
primary line was 2.7 m/s considering that the velocity limits at created when consolidating all the governing equations
secondary line and plot take-off were held constant at 3.0 m/s derived from a distribution network with secondary
and 1.9 m/s, respectively. lines. The independent variables were identified using
multivariable Newton-Raphson and singular-value
CONCLUSIONS decomposition methods.

The method of identifying the piping design criteria of 4. The results of system simulation defined the coordinates
district cooling distribution network was developed. To of system characteristic curve which indicated the best
determine the piping design criteria involved three (3) research efficiency points during selected part load and full load
areas namely; friction factor identification, chilled water conditions.
pumping system simulation, and optimization. In this study, two
(2) distribution networks models with and without secondary 5. At full load, distribution network DNModel-01_12
lines and total pipe length of approximately 5.2 km. were used as required higher pump head of 28.3 m as compared with
benchmarks for fluid flow analysis. Both networks had twelve distribution network DNModel-03_12 which was only
(12) energy transfer stations with total nominated load of 21.7 m.
24,000-ton of refrigeration and diversity factor of 80%. Main
results are summarized as follows: 6. Optimization results using exhaustive search method

Vol.6 | No.2 | 2013 Philippine Science Letters 191

 Figure 8. Optimization based on present worth pattern method for DNModel-01_12

Figure 9. Optimization based on annual cost pattern method for DNModel-01_12

192 Philippine Science Letters Vol.6 | No.2 | 2013

 Figure 10. Optimization based on present worth pattern method for DNModel-03_12

Figure 11. Optimization based on annual cost pattern method for DNModel-03_12

suggest that piping network design criteria for components of objective function as listed in Tab 1, Tab. 2, Tab.
distribution network DNModel-01_12 without 3 and Tab. 5, respectively will be used.
secondary line have velocity limits at primary line of
3.0 m/s and at plot take-off of 1.75 m/s considering a NOMENCLATURE
pressure drop limit of 100 Pa/m.
µ absolute viscosity
7. Furthermore, the piping network design criteria for Ts chilled-water supply temperature
distribution network DNModel-03_12 with secondary
lines had velocity limits at primary line of 2.7 m/s, at δs correction vector
secondary line of 3.0 m/s and at plot take-off of 1.9 m/s
considering a pressure drop limit of 100 Pa/m. dm elemental mass

dv elemental volume
The piping network design criteria discussed in item nos. 6
and 7 can only be achieved provided that input parameters such ds elemental surface
as nominated cooling loads, pipe lengths, normalized annual
power load demand factor and interest rates associated with c frictional coefficient

Vol.6 | No.2 | 2013 Philippine Science Letters 193

 Table 1. Nominated cooling loads

Table 2. Pipe lengths for two (2) distribution network models

Table 3. Normalized annual power load demand factor

194 Philippine Science Letters Vol.6 | No.2 | 2013

 Table 4. Description of best efficiency points at part load and full load conditions

f friction factor t time

g gravity torque
J Jacobian matrix
unit vector
ρ mass density
q velocity field
LC nominated cooling load
vu.l. velocity upper limit
D pipe diameter
ω volume flow rate
L pipe length
Subscript
ε /D pipe relative roughness
cv control volume
PHE plate-type heat exchanger
i indicates the number of pipe components
PTO plot take-off
j refers to the number of ETS in the distribution network
δp1 pressure drop due to pump curve
k number of data points
δpj pressure drop due to piping configuration
l number of pump curve coefficients
δpl.v. pressure drop limiting value
ii number of functional relations to be zeroed in MNR
η pump efficiency
jj number of independent variables in MNR
pump head
kk number of velocity limits

φ pump impeller diameter ll number of iteration in MNR

ε/φ pump relative roughness s number of secondary lines

n pump rotative speed

ACKNOWLEDGEMENT
Q pump volume flow rate
The authors wish to thank the FLUIDNOVATION
R Reynolds number Research, Co. for providing financial support.
S solution vector of MNR
CONFLICTS OF INTERESTS
Cp specific heat at constant pressure
The authors certify that there is no conflict of interest in the
δT temperature difference set point conduct, preparation and submission of this manuscript.

Vol.6 | No.2 | 2013 Philippine Science Letters 195

 Table 5. Cost parameters associated with components of objective function

Table 6. Piping network design criteria for distribution network models

Table 7. Pipe sizes for two (2) distribution network models

196 Philippine Science Letters Vol.6 | No.2 | 2013

CONTRIBUTIONS OF INDIVIDUAL AUTHORS Engineers, 1997.
ASME. Power Piping: ASME Code for Pressure Piping, B31.1. New
Gerardo L. Augusto The main author and researcher of York: The American Society of Mechanical Engineers, 2009.
Bin Shafar A. Exclusive Interview of Empower CEO, Ahmad Bin
this study. This research work is part of his doctoral dissertation
Shafar by Michael Palmer. Dubai UAE, June 16, 2011. Available
at De La Salle University – Manila. at http://www.utilities-me.com/article-1383-exclusive-interview-
empower-ceo-ahmad-bin-shafar/3/#.UUHqU46UAy4
Alvin B. Culaba He guided the main author in Hodgson J, Walters T. Optimizing pumping systems to minimize first or
conceptualizing design optimization in district cooling system. life-cycle cost. Proceedings of the 19 th International Pump Users
He is the adviser of the main author in his PhD work at De La Symposium. Turbomachinery Laboratory, Texas A&M
Salle University – Manila. University 2002; 1-8.
Jeppson RW. Analysis of Flow in Pipe Networks. Michigan: Ann Arbor
Raymond R. Tan Provided lectures on engineering Science, 1977.
optimization and discussed some key points of heuristic and McAllister EW. Pipeline Rules of Thumb Handbook. 5 th ed., Oxford:
Elsevier, Inc., 2005.
metaheuristic algorithms with the main author.
Schmit LA. Structural design by systematic synthesis. ASCE
Proceedings of the 2nd Conference on Electronic Computation
REFERENCES 1960; 105-122.
Stephenson D. Pipeline Design for water Engineers. New York: Elsevier
Antaki GA. Piping and Pipeline Engineering: Design, Construction, Scientific, 1981.
Maintenance, Integrity and Repair. New York: Marcel Dekker, Stoecker WF. Design of Thermal Systems, 3 rd ed., New York: McGraw
Inc. 2003. Hill, 1989.
ASHRAE. 2008 ASHRAE Handbook HVAC Systems and Equipment. Thorp J, Olsen J. Application of numerical optimization of sea water
Atlanta: American Society of Heating, Refrigerating and Air-
pumping systems. Proceedings of the 24 th International Pump
Conditioning Engineers, 2008.
Users Symposium. Turbomachinery Laboratory, Texas A&M
ASHRAE. 2013 ASHRAE Handbook Fundamentals. Atlanta: American
University 2008; 19-28.
Society of Heating, Refrigerating and Air-Conditioning
Vanderplaats GN. Structural design optimization – status and directions.
Engineers, 2013.
AIAA Journal Aircraft 1999; 11-20.
ASME. Building Services Piping: ASME Code for Pressure Piping,
Walters T. Cutting costs in pump and pipe sizing. Chemical Processing
B31.9. New York: The American Society of Mechanical
2002; 35-38.

Vol.6 | No.2 | 2013 Philippine Science Letters 197