Iranian Journal of Management Studies (IJMS) http://ijms.ut.ac.

ir/
Vol. 8, No. 4, October 2015 Print ISSN: 2008-7055
pp: 653-685 Online ISSN: 2345-3745

A multi-objective resource-constrained optimization of

time-cost trade-off problems in scheduling project
Online ISSN 2345-3745
Mostafa Zareei, Hossein Ali Hassan-Pour
Faculty of Industrial Engineering, Imam Hossein (AS) University, Tehran, Iran

(Received: 8 August 2014; Revised: 22 December, 2014; Accepted: 24 December, 2014)

Abstract
This paper presents a multi-objective resource-constrained project scheduling
problem with positive and negative cash flows. The net present value (NPV)
maximization and making span minimization are this study objectives. And since
this problem is considered as complex optimization in NP-Hard context, we present
a mathematical model for the given problem and solve three evolutionary
algorithms; NSGA-II, MOSA and MOPSO are applied to find the set of Pareto
solutions for this multi-objective scheduling problem. In order to show performance
of the algorithms, different metrics are applied and comparisons between the two
algorithms are also considered. The computational results for a set of test problems
taken from the project scheduling problem Bandar Abbas Gas condensate Refinery
project and library are presented and discussed. Finally, the computational results
illustrate the superior performance of the NSGA-II, MOSA and MOPSO algorithm
with regard to the proposed metrics. In order to solve proposed method from NSGA-
II algorithm, the results are compared with GAMS software in some problems. The
proposed method is a Converge to the optimum and efficient solution algorithm.

Keywords
Comparative indicators of evolutionary algorithms, MOSA and MOPSO algorithm,
NSGA-II, Payment patterns, Project scheduling, Resource constraints.

 Corresponding Author, Email: [email protected]

654 (IJMS) Vol. 8, No. 4, October 2015

Introduction
Resource-constraint project scheduling problem (RCPSP) is a class of
project scheduling that one’s activities should be scheduled subject to
precedence and resource constraints and it is proven to be NP-hard
(Aboutalebi1 et al., 2012). Minimization of project duration is often
used as an objective of a general project scheduling problem while
other objectives such as Maximizing of net present value of cash
flows, and leveling of resource usage are also considered. Resources
involved in a project can be single or multiple varieties, and can be
renewable or nonrenewable (Ritwik & Paul, 2013). The time–cost
tradeoff problem in project management originates when activity time
can be reduced with some extra direct cost (Jongyul et al., 2012).
Time Cost Trade off analysis is the compression of the project
schedule to achieve a more favorable outcome in terms of project
duration, cost, and projected revenues. The objectives of the Time
Cost Trade off analysis are compressing the project to the optimum
duration which minimizes the total project cost (Rifat & Önder Halis,
2012).
Another important type of objective emerges if cash flows occur while
the project is carried out. Cash outflows are induced by the execution
of activities and the usage of resources. On the other hand, cash
inflows result from payments due to the completion of specified parts
of the project. Typically, discount rates are also included. Note that
cash flows related to activity j might occur at several points in time
during execution of j. However, they can easily be compounded to a
single cash flow at the beginning or the end of j. These considerations
result in problems with the objective to maximize the net present
value (NPV) of the project which subject to the standard RCPSP
constraints (Shu-Shun & Chang-Jung, 2008). Project payment
scheduling problem involves how to schedule progress payments
effectively including the amount, time or spots (i.e. the key activities
or events associated with payments), and so on of payments in the
project so as to maximize the profits of the contractor or/and the
client.
A multi-objective resource-constrained optimization of time-cost trade-off … 655

In real life situations, there are at least two parties involved in the
project: the client, who is the owner of the project, and the contractor,
whose job is to execute the project. They have to agree with the
method of payment transferring from the client to the contractor for
the execution of the project. The ideal situation for the client would be
a single payment at the end of the project. The contractor, on the other
hand, would like to receive the whole payment at the beginning of the
project (Zhengwen & Yu, 2008). Time-Cost optimization (TCO)
problem has been extensively examined by a number of research
studies. Various approaches have been proposed for optimizing
construction time and cost including (1) heuristic methods (Moselhi,
1993; Siemens, 1971); (2) mathematical programming (Liu et al.,
1995; Moussourakis & Haksever, 2004); and (3) meta-heuristic
methods. Mathematical programming such as linear programming is
suitable for problems with linear time-cost relationships, but they
often fail to solve the problem with discrete time-cost relationships
(Feng et al., 1997). Moreover, it requires a lot of computational efforts
to solve a large scale project network. Heuristic methods are able to
overcome such limitation of a large scale problem, but fail to
guarantee optimal solutions. Therefore, many research studies have
focused on utilizing meta-heuristic methods in time-cost tradeoff
analysis to overcome the limitation of heuristic methods and
mathematical programming.
Liu et al. (1995) have developed optimization model using a hybrid
method that integrates linear and integer programming. Linear
programming was used to find lower bounds of the solutions, and then
integer programming was used to obtain the exact solution. The
integer programming was then used to minimize total project cost with
the constraints of activity precedence and the selection of a single
resource utilization option for each activity. The hybrid model was
developed using Microsoft Excel to provide a construction planner
with an efficient means of analyzing time-cost trade-off decisions.
Zheng et al. (2005) have developed GA-based multi-objective
optimization model that simultaneously minimizes time and cost. In
order to overcome the problem of genetic drift, the model utilized
656 (IJMS) Vol. 8, No. 4, October 2015

Pareto ranking, niche formation, and adaptive mutation rate. The

genetic drift occurs when GA converge to a single peak due to the
stochastic errors during processing. The model incorporated which
modified adaptive weight approach (MAWA) to exert a search
pressure in GA. Pareto ranking mechanism overcomes the limitation
of traditional proportional selections. Niche formulation is useful for
stabilizing multiple subpopulations that arise along the Pareto-optimal
front, and thereby it maintains diversity of population.
Xiong and Kuang (2008) have used ant colony optimization (ACO)
algorithm to solve time-cost tradeoff problem. The modified adaptive
weight approach (MAWA) proposed by Zheng et al. (2004) was
incorporated in the model to generate optimal time-cost tradeoff
curve. The performance of developed model was compared with GA
by two test examples. The results showed that ACO-based multi-
objective approach provides an attractive alternative to solving
construction time-cost optimization. El-Rayes and Kandil (2005) have
presented the multi-objective optimization model for time-cost-quality
tradeoff analysis. Multi-objective genetic algorithm was used in the
model to generate the optimal solutions for those conflicting
objectives. The multi-objective genetic algorithm was implemented in
the model to search for optimal resource utilization options for each
activity that provides minimum project cost and time while
maximizing quality performance. The model also provided
visualization of tradeoffs among time, cost, and quality for the
analyzed example.
Najafi and Niaki (2006) consider a cash flow related to a subset of
activities, that is, a cash flow is initiated when the last activity of the
subset is finished. Furthermore, the cash flow associated with activity
j might depend on the mode chosen for j as lined out in Icmeli and
Erenguc (1996). Smith-Daniels et al. (1996) propose the objective to
maximize the discounted amount of cash available in each period.
This amount is influenced by cash flows associated with each activity.
Ulusoy and Cebelli (2000) investigate the negotiation process to find
the timing of payments and the amount of each specific payment
between a client and a contractor. Obviously, the client seeks to
A multi-objective resource-constrained optimization of time-cost trade-off … 657

minimize the NPV while the contractor aims at maximizing it. The
objective in Ulusoy and Cebelli (2000) is to find the payment structure
which minimizes each party’s loss in comparison to the respective
ideal payment structure. Dayanand and Padman (2001) treat a similar
problem, but restrict themselves to the client’s point of view. The
client might associate a specific value with each event (starting or
completion of a job). Cash outflows can be assigned to each event
having a positive value. The problem is to find a project schedule and
decide cash outflows to happen at a given number of events.
The total outflow might exceed the total value of finished activities
at no point of time. The objective is to find a solution such that
discounted cash inflow (associated with finishing the project) minus
total discounted cash outflow will maximize. Dorner et al. (2008)
employ three objectives within a variant of the time-cost tradeoff
problem. The first objective is a function of the project make span,
while the second and the third are functions of the monetary and non-
monetary costs for crashing the activities, respectively. Afshar et al.
(2009) proposed Non-dominated Archiving ACO (NA-ACO)
algorithm in which all ant colonies are initiated by the same number of
ants, and arbitrary order is given to the colonies. Ants in a certain
colony simultaneously explore a solution according to the objective
assigned to that colony. If there is an improvement, the optimal path is
updated. The progress payment model corresponds to what Dayanand
and Padman (1998) refer to as the periodic payment model. Dayanand
and Padman provide mixed integer linear programming formulations
for the so-called basic client, equal time intervals and periodic
payment models. They provide insights about the characteristics of
optimal payment schedules obtained with each model. Kim et al.
(2012) present an improved genetic algorithm to solve a multi-mode
resource-constrained discrete time–cost tradeoff problem; which is
applicable to special knowledge intensive projects, and does not
consider project activities' quality.
Zhou (2011) has used ant colony algorithm to trade-off costs and
time. The problem is considered as a multimode discrete and the
objective function as the sum of the direct and indirect costs which
658 (IJMS) Vol. 8, No. 4, October 2015

indirect cost is defined as a linear function of project time (Xu, 2011).

Aladini et al. (2011) have proposed a multi-objective ant colony
optimization model for minimizing the project direct cost by
calculating discounted cash flow to solve costs and time tradeoff
problem considering value of money over time (considering the
discount rate), and importance of discounted cash flow analysis for
owners and contractors. Vanhoucke (2010) has presented a scatter
search algorithm for project scheduling problem with constrained
resource and with discounted cash flows. He has assumed payment in
connection with the implementation of the project activities as a
constant parameter and has developed a heuristic optimization method
for maximizing the project net present value of the objective function
with respect to priority and renewable resources constraints
(Khalilzadeh et al., 2011).
Table 1. Literature of multi-objective optimization of time-cost trade-off problems in project scheduling
NO. Name Years Using of Way Type
Have developed optimization model using a hybrid method that
integrates linear and integer programming to minimize total
1 Liu et al. 1995
project cost with the constraints of activity precedence and the
selection of a single resource utilization option for each activity.
Provide mixed integer linear programming formulations for the
Dayanand
2 1998 so-called basic client, equal time intervals and periodic payment
and Padman
models.
El-Rayes Have presented the Multi-objective genetic algorithm for time-
3 2005
and Kandil cost-quality tradeoff analysis.
Have developed GA-based multi-objective optimization model
that simultaneously minimizes time and cost. The model
4 Zheng et al. 2005
incorporated modified adaptive weight approach (MAWA) to
exert a search pressure in GA.
consider a cash flow related to a subset of activities, that is, a
Najafi and
5 2006 cash flow is initiated when the last activity of the subset is
Niaki
finished
Xiong and have used ant colony optimization (ACO) algorithm to solve
6 2008
Kuang time-cost tradeoff problem
Proposed Non-dominated Archiving ACO (NA-ACO) algorithm
7 Afshar et al. 2009 in which all ant colonies are initiated by the same number of ants
and arbitrary order is given to the colonies.
Has presented a scatter search algorithm for project scheduling
8 Vanhoucke 2010 Problem with constrained resource and with discounted cash
flows.
9 Zhou 2011 Has used ant colony algorithm to trade-off costs and time.
Present an improved genetic algorithm to solve a multi-mode
10 Kim et al. 2012
resource-constrained discrete time–cost tradeoff problem
A multi-objective resource-constrained optimization of time-cost trade-off … 659

Problem description and mathematical formulation model

Our proposed model is categorized in resource-constrained project
scheduling problem with discounted cash flows (RCPSPDCF) that can
be defined as follows. A project consisting of n activities is
represented by an activity-on-node network, G= (J,E), |J|= n, where
nodes and arcs correspond to activities and precedence constraints
between activities, respectively. Nodes in graph are topologic and
numerically numbered, that is an activity has always a higher number
than all its predecessors. No activity may be started before all its
predecessors are finished. The duration of activity j= (1,2,…,n)
executed is dj. There are R renewable resources. The number of
available units of renewable resource k= (k=1,2,…,R) is Rk. Each
activity j is executed requiring rjk units of renewable resource k=
(k=1,2,…,R) for its processing. A negative cash flow CF-j is
associated with the execution of activity j. For each completed activity
occurs a negative cash flow until the completion time of a project.
Finally, the contractor receives amount of cash flows CF+j for each
activity that has completed successfully. The value of an amount of
money is a function of the time of receipt or disbursement of cash.
Money received today is more valuable than money to be received
some time in the future, since today’s money can be invested
immediately. In order to calculate the value of NPV, a discount rate i
α has to be chosen, which represents the return following from
investing in the project. The objective is to find an assignment of
modes to activities as well as precedence and resource-feasible
starting times for all activities such that the net present value of the
project is maximized.
All the parameters are used in the proposed RCPSPDCF model are
summarized below:
N: Number of activities
G: Acyclic digraph representing the project
dj: Duration of activity j executed
CF-j: Negative cash flow associated with activity j executed
CF+j: Positive cash flow associated with activity j executed
660 (IJMS) Vol. 8, No. 4, October 2015

STj: Starting time of activity j

FTj: Finishing time of activity j
EFj: Earliest finishing time of activity j
LFj: Latest finishing time of activity j
Pj: Setting of all predecessors of activity j
R: Number of renewable resources
Rk: Number of available units of renewable resource k, k=1,2,…R
rjk: Number of units of renewable resource k required by activity j
executed
α: Discount rate
Cmax: The maximum time for completion
T: Horizon of Project Scheduling
NPV: Net Present Value of the project
Pk: Paid the amount of k
K: The number of continuous payments
U: The total amount of payments

By using the above notations, the proposed model can be

formulated as the following mathematical programming problem:
(1)
maxNPV
LFj
min c max  max(  tX
t  EF j
jt ) (2)

st :
LFj LFj

 tx jt   (t  d j ) x jt  j ,w  p j
t  EF j t  EF j
(3)

LFj
Cj   tX
t  EF j
jt  j  1,2,..., n (4)

C max  C j  j  1,2,..., n (5)

A multi-objective resource-constrained optimization of time-cost trade-off … 661

n
T   max( d j ) (6)
j 1

C max  T (7)
n t  d j 1

 rjk
j 1
X
b t
jb  Rk ,k ,t (8)

ES1  0 (9)

EFi  ES i  d i , i  1,2,..., n (10)

ESj  maxEFi i  p j , j  1,2,..., n (11)
d0  0, d n1  0 (12)
n

y
j 1
jk  1, k  1,2,..., k  1 (13)
k

y
k 1
jk  1, j  1,2,..., n (14)
k

P
k 1
k  u , k  1,2,..., k (15)
Pk  0, k  1,2,..., k (16)

Equation (1) represents the objective function which is to maximize

the net present value of the project, and the contractor is calculated
according to the method of payment. Equation (2) represents the
objective function, the maximum completion time of activity n+1 that
should be minimized. The constraint set (3) makes sure that all
precedence relations are satisfied. The Constraints set (4) shows the
completion time of project activities. Constraint (5) calculates
maximum project completion. Constraint (6) calculates project
planning horizon which is equal to all project activities. Constraint (7)
ensures that the project is completed before project planning horizon.
Constraints (8) are for applying renewable resource constraints, and in
each period, summation of consumption of all activities from each
resource in each time unit cannot exceed from maximum amount of
that resource (Rk) in its relevant time unit. Constraints (9) expresses
project starts time. Constraints (10) are related to transposition
relations (without delay) between project activities. In such a way that
662 (IJMS) Vol. 8, No. 4, October 2015

no activity can start before the end of all its prerequisite activities, and
on the other hand, projects activities are continuous. Constraints (11)
show that j the activity start time is equal or larger than its prerequisite
activities end time. Constraints (12) show that 0 and n+1 activity are
virtual activities. Constraints (13) shows number of payments K for
certain event m. constraint (14) ensures that one payment is allocated
at the end of event. Constraint (15) ensures that summations of all
payments are equal to project contractor price. Constraint (16) also
shows that payments values always are positive.
In real life situations, there are at least two parties involved in the
project: the client, that is, the owner of the project and the contractor
who undertakes the execution of the project. The legal basis of the
execution of a project is provided by a contract organizing aspects of
the interactions between the stakeholders. There are a large number of
contract types with considerable amount of detail involved. A treatise
of different contract types is given by Herroelen et al. (1997). For the
purposes of this paper, we are interested in the basic payment
structures specified in the contracts. Four types of payment scheduling
models are of particular interest in practice: Lump-sum payment,
payment at event occurrences, payment at equal time intervals, and
progress payment.
Lump-sum payment (LSP) is one of the more commonly used
payment structures in the literature. Here, the whole payment is paid
by the client to the contractor upon successful termination of the
project (Seifi & Tavakkoli-Moghaddam, 2008). The LSP model
represents the ideal situation for the client––he makes a single
payment to the contractor only at the end of the project. However, in
general, this shifts the entire financial burden on the contractor, which
may not be acceptable in some project environments (Marek et al.,
2005).
LFj Mj
n CF jm
maxz  CFlsp (1   )  FTn
    X jmt
i 1 t  EFi m 1 (1  ) t
(17)
n
CFlsp   CF 
j
j1
A multi-objective resource-constrained optimization of time-cost trade-off … 663

In the payments at event occurrences (PEO) model, payments are

made at predetermined set of event nodes. The problem is to
determine the amount and timing of these payments (Seifi &
Tavakkoli-Moghaddam, 2008). PEO is a very reasonable model,
where the contractor gets his payments for successful completion of
each activity (Marek et al., 2005).
LFj M j
n n CF jm
maxz   CF j (1   )   (1   ) (18)
 FT j
t
 X jmt
j 1 j 1 t  EFi m1

In the equal time intervals (ETI) model, the client makes H

payments for the project. Of these payments, the first (H-1) are
scheduled at equal time intervals over the duration of the project, and
the final payment is scheduled on project completion (Seifi &
Tavakkoli-Moghaddam, 2008). In the ETI model the client and the
contractor agree about the number of payments over the course of the
project. The payments are then made at equal time intervals (Marek et
al., 2005).
LFj M j
H n CF jm
maxz   Pp (1   ) TP    (1   ) t
 X jmt (19)
P 1 j 1 t  EFi m1

In the progress payment (PP) model, the contractor receives the

project payments from the client at regular time intervals until the
project is completed. For example, the contractor might receive at the
end of each month a payment for the work accomplished during that
month multiplied by a profit rate agreed upon by both the client and
the contractor (Seifi & Tavakkoli-Moghaddam, 2008). A similar
situation concerns the PP model, where the payments are also made at
regular time intervals, but in this case the two parties agree about the
length of this interval, not the number of payments (Marek et al.,
2005).
The difference between the ETI and PP models is that in the latter
case the number of payments is not known in advance.
664 (IJMS) Vol. 8, No. 4, October 2015

H 1 n LFj M j
CF jm
maxz  ( Pp (1   i )  PT  PH (1   i ) _ FTn )      X jmt (20)
P 1 j 1 t  EFi m1 (1   ) t

Schedule Generation Schemes

Schedule generation schemes (SGS) are the core of most heuristic
solution procedures for the RCPSP. SGS start from scratch and build a
feasible schedule by stepwise extension of a partial schedule. A partial
schedule is a schedule where only a subset of the n+2 activities have
been scheduled. The so-called serial SGS performs activity
incrimination and the so-called parallel SGS performs time-
incrimination.
Serial schedule generation scheme
We begin with a description of the serial SGS. It consists of g = 1… n
stages, in each of which one activity is selected and scheduled at the
earliest precedence and resource feasible completion time. Associated
with each stage, g is two disjoint activity sets. The scheduled set S g
comprises the activities which have been already scheduled; the
eligible set Dg comprises all activities which are eligible for
scheduling. Note that the conjunction of Sg and Dg does not give the
set of all activities J because, generally, there are so-called ineligible
activities, that is activities which have not been scheduled and cannot
be scheduled at stage g because not all of their predecessors have been
scheduled. Consider Rk(t)=Rk-jεA(t)rj,k is the remaining capacity of
resource type k at time instant t and is the set of all
finish times, and further is the set of eligible
activities. The initialization assigns the dummy source activity j = 0 a
completion time of 0 and puts it into the partial schedule. At the
beginning of each step g, the decision set Dg, the set of finish times Fg,
and the remaining capacities Rk(t) at the finish time’s t are
calculated. Afterwards, one activity j is selected from the decision set.
The finish time of j is calculated by first determining the earliest
precedence feasible finish time EFj and then calculating the earliest
(precedence-and) resource-feasible finish time Fj within [EFj, LFj].
LFj denotes the latest finish time as calculated by backward recursion
A multi-objective resource-constrained optimization of time-cost trade-off … 665

(cf. Elmaghraby, 1977) from an upper bound of the project's finish time T
(Möhring & Stork, 2000).

NSGA-II methodology
The NSGA-II algorithm is the first and one of the commonly used
evolutionary multi-objective optimization (EMO) algorithms which
search for solution space to find Pareto-optimal solutions in a multi
objective optimization problem. NSGA-II uses the elitist principle and
an explicit diversity preserving mechanism. In addition, it emphasizes
non-dominated solutions, and forms the Pareto front as Pareto-optimal
solutions. The NSGA-II algorithm uses two effective strategies
including an elite-preserving and an explicit diversity-preserving.
NSGA-II uses an explicit diversity-preservation or niching strategy to
assign a diversity rank to all the individuals that are in the same non-
dominated front and thus have the same non-dominated rank in the
population. The members within each non-dominated front that are in
the least crowded region in that front are assigned a higher rank. For
calculating the density of solutions surrounding a particular solution in
the population, a crowding distance metric is used that is achieved
from the average distance of the two solutions on either side of the
solution along each of the objectives. Respecting that this particular
niching strategy does not require any external parameters; therefore, it
was chosen for NSGA II. Details can be found elsewhere. Because of
the nature of the models of the multi-objective optimization problems,
non-dominated sorting genetic algorithm (NSGA) can be used to find
the non-dominant optimal solutions. In the absence of any additional
information about multi-objective optimization problem, one of these
Pareto-optimal solutions cannot be considered as better solution than
the others. Superiority and Suitability of one solution over the others
depends on several factors including user’s choice and problem
environment. Therefore, the NSGA II determines a set of dominant
solution and as a result Pareto front is obtained (Salimi et al., 2013).
Initial population for NSGA-II algorithm
Initial individuals are obtained by fixing the activities modes
666 (IJMS) Vol. 8, No. 4, October 2015

randomly from the remaining set of modes after the preprocessing

procedure. Then, the activity list is constructed with respect to the
resulting mode assignment. The ensuing schedule is precedence and
nonrenewable resource feasible.
The initial generation is obtained by repeating the next steps POP
times. Firstly, a mode assignment is generated by randomly selecting
for activities j=1,…J. Secondly, the resulting mode
assignment is checked for nonrenewable resource feasibility. In
presence of violation, a local search procedure is applied trying to
improve the current mode assignment. The process consists of
randomly selecting a job j which has more than one mode alternative,
and its current mode is changed by randomly selecting
. The result is a new mode assignment . If there is
improvement, is replaced by . This process is repeated until J
consecutive unsuccessful trials to improve the mode assignment have
been made, or in the best case, until the individuals become
nonrenewable resource feasible. Thirdly, we adopt the mode
assignment and construct a precedence feasible schedule by randomly
sampling. The procedure consists of J stages; at each stage, a decision
(say also eligible) set is determined; this set includes all unscheduled
activities which every predecessor has scheduled yet. An activity is
chosen randomly and scheduled at its earliest precedence and
renewable resource feasible start time. These steps are repeated until
the J steps are fulfilled. Finally, activities finish times are computed
and the makespan of the project is derived (Elloumi & Fortemps, 2010).

Updated the population

Crossover operator
pc Parameter is considered as Crossover probability and for selecting
parent’s chromosome in Crossover, we repeat the following process
( pop  size  pc ) times. For i=1, 2… pop-size, we use three Crossover
types as one point and two point unified Crossover. This process is
described as follow. First, we must select a stochastic number in one
point Crossover in [0, N-1] and then we break both parents in this
A multi-objective resource-constrained optimization of time-cost trade-off … 667

point and by moving their sequence, we produce two new child. Then
in two point Crossover we select two different random number in [0,
N-1] interval and we break both generator in these two points and by
moving points between two parts of both generators, we produce two
child and then in unified intersection we produce two random numbers
like V in [0, 1] interval and if V≤pc (in proposed algorithm is equal to
0.9), xi chromosome is selected as a parent in Crossover operation.
Then, we reach the number of (pop-size) pc parents for Crossover
operation. We number them again from the start and specify them by
prime sign as ( ). In the next phase, if we want to have an
Crossover between two parents like ,
, we must first produce a random number in
[0, 1] interval and then do the intersection operation by using the
following equation which are new chromosome and named as child
chromosome and are signed by ". If both Childs are feasible, then we
replace parents with them. If one of the parents is possible then we
keep that and repeat Crossover operations to reach another possible
child. If both of them are not possible, we repeat the operation to two
possible childe.
Mutation operator
pm Parameter is considered as probability of mutation. Parent
chromosome is selected by the same method which was mentioned in
intersection operations. Parent chromosome is selected which is
almost as many as . Then mutation operation is
applied as the following method. In this research, gussing method is
used for producing mutants that for X variables which is and
, new variable must have normal distribution with zero mean and
variance. That and . This means that a
standard value is produced and multiplied by and summed by X
value and is equal to . Which is equal to
mutation steps. Therefore, mu% (mutation ratio) is selected randomly
and to have an integer value for mutants and at least one case may be
found, value is rounded up.
668 (IJMS) Vol. 8, No. 4, October 2015

The Multi-Objective Particle Swarm Optimization (MOPSO)

PSO simulates a social behavior such as bird flocking to a promising
position or region for food or other objectives in an area or space. Like
evolutionary algorithm, PSO conducts search using a population,
which is called swarm, of individuals, which are called particles. Each
particle represents a candidate position or solution to the problem at
hand to represent a potential solution. During searching for optima
each PSO particle adjusts its trajectory towards its own previous best
position, and towards the best previous position attained by any
member of its neighborhood (i.e., the whole swarm). Thus, global
sharing of experience or information takes place and particles profit
from the discoveries of themselves (i.e., local best), and previous
experience of all other companions (i.e., global best) during search
process. PSO is initialized with a population of M random particles
and then searches for best position (solution or optimum) by updating
generations until getting a relatively steady position or exceeding the
limit of iteration number (i.e., T). In every iteration or generation, the
local bests and global bests are determined through evaluating the
performances (Azimi et al., 2011; Bashiri et al., 2011). A swarm
consists of a set of particles and each particle represents a potential
solution. is the position of each particle that is defined by adding
a velocity to a current position:
+

That the velocity vector is defined as fallow:

where is position of the best particle member of the

neighborhood of the given particle, is the best position of
the best particle member of the entire swarm (leader),w is inertia
weight, is the cognitive learning factor and is the social learning
factor (usually defined as constants), and are random
values (Ritwik & Ginu, 2013; Bashiri et al., 2011).
A multi-objective resource-constrained optimization of time-cost trade-off … 669

Main algorithm of MOPSO

In case of the relative simplicity of PSO, multi objective particle
swarm optimization allows PSO algorithm to solve multi objective
problems. This algorithm is based on Pareto dominance and it
considers every non-dominated solution as new leader. This approach
also uses a crowding factor to filter out the list of available leaders.
This algorithm works thus. First, a swarm is initialized and a set of
leaders is also initialized with the non-dominated particles from the
swarm. This set is usually stored in an external archive. Then, some
sort of quality measure is calculated for all the leaders in order to
select one leader for each particle of the swarm. At each generation
for each particle, a leader is selected and a flight is performed. Then,
the particle is evaluated and its corresponding is updated. A
new particle replaces its particle usually when this particle is
dominated or if both are non-dominated with respect to each other.
After all the particles have been updated, the set of leader is updated,
too. Finally, the quality measure of the set of leaders is re-calculated
and this procedure is repeated for a certain number of criterions.
External repository: The main objective of the external repository is to
keep a record of non-dominated vectors found during the search
process. The external repository consists of two components; the
archive controller and the grid, which are discussed in more details.
The function of archive controller is to decide whether a certain
solution should be added to archive or not. The mechanism of the grid
is to produce well-distributed Pareto fronts (Aboutalebi et al., 2013).
Multi-Objective Simulated Annealing
Simulated annealing (SA) is one of the stochastic search algorithms
which have been originally motivated by thermodynamic process of
annealing in physics. Though SA was designed originally to use only
one search agent, and therefore it does not need large memory to keep
the population, there are some techniques for empowering SA to find
the set of estimated Pareto-optimal solutions for multi-objective
problems. Considering a minimization problem, SA allows “uphill”
moves so as to avoid getting stuck at a local optimum by accepting
670 (IJMS) Vol. 8, No. 4, October 2015

worse solutions with a probability defined by “acceptance probability

function.” We find that the performance of the MOSA depends on the
type of acceptance probability function applied and the rule of
replacing current solution by candidate solution (Varadharajan &
Rajendran, 2005). The SA algorithm starts with an initial solution for
the given problem and repeats an iterative neighbor generation
procedure that improves the objective function. During searching for
the solution space and in order to escape from local minima, the SA
algorithm offers the possibility to accept the worse neighbor solutions
in a controlled manner. A neighboring solution (S') of the current
solution (S) is generated in each iteration of the inner loop. If the
objective function value of S' is better than S, then the generated
solution replaces with the current one; otherwise, the solution can be
also accepted with a probability . where T is the value of
current temperature (i.e., higher values of T give a higher acceptance
probability) and Δ = f(S) − f (S′). The acceptance probability is
compared to a number y [0, 1] generated randomly, and S' is
accepted whenever p > y (Chen et al., 2010).
A target-vector approach to solve a bi-objective optimization
problem has been used. Ulungu et al (1999) have proposed a complete
MOSA algorithm which they had tested on a multiobjective
combinatorial optimization problem. A weighted aggregating function
to evaluate the fitness of solutions has been used. The algorithm
worked with only one current solution but maintained a population
with the non-dominated solutions found during the search. The
algorithm uses only one solution and the annealing process adjusts
each temperature independently according to the performance of the
solution in each criterion during the search. An archive set stores all
the non-dominated solutions between each of the multiple objectives.
A new acceptance probability formulation based on an annealing
schedule with multiple temperatures (one for each objective) has also
been proposed. The acceptance probability of a new solution depends
on whether or not it is added to the set of potentially Pareto-optimal
solutions. If it is added to this set, it is accepted to be the current
A multi-objective resource-constrained optimization of time-cost trade-off … 671

solution with probability equal to one. Otherwise, a multiobjective

acceptance rule is used. However, the acceptance decision of new
solution has to take into consideration the improvement (or
deterioration) of k objectives, simultaneously. In case of two
optimization objectives (i.e. k = 2), the comparison of the actual and
the new solution (i.e. decision vector) results depicted in three cases.
Case (a): The move from Xact to Xnew is improving with respect to
all k objectives. This means that
(Supposing a minimization MOP) for
.
Case (b): An improvement and deterioration can be simultaneously
observed on different objectives. This means, there exist a and a
with < 0 and > 0. This is the case where the new solution is
indifferent to the actual one. Thus, a strategy has to be defined to
decide if the new solution should be accepted as current solution for
the next iteration.
Case (c): All objectives are deteriorated, with ≥ 0 for all
and such that > 0. In this case, an acceptance
probability to accept Xnew has to be calculated.

Parameter settings for the NSGA-II, MOPSO and MOSA algorithms

Tables 2, 3, 4 present the control parameters for the NSGA-II,
MOPSO and MOSA algorithms.

Table 2. Control parameters for the NSGA-II algorithms

Parameter The parameters amount
Maximum Number of Iterations 200
Population Size 100
Crossover Percent 0.9
Number of Crossover 2*round(pCrossover*nPop/2)
Mutation Percent 1/ String length of chromosome
Number of Mutation round(pMutation*nPop)
Mutation rate 0.01
Mutation step size 0.2*(VarMax-VarMin)
Parent selection method Binary tournament
Stopping rule Production of 200 generations
672 (IJMS) Vol. 8, No. 4, October 2015

Table 3. Control parameters for the MOPSO algorithms

Parameter The parameters amount
Maximum Number of Iterations 200
Population Size 100
Repository Size 100
Inertia Weight 0.5
Inertia Weight Damping Rate 0.99
Personal Learning Coefficient(c1) 1
Global Learning Coefficient(c2) 2
Number of Grids per Dimension 5
Inflation Rate 0.1
Leader Selection Pressure 2
Deletion Selection Pressure 2
Mutation Rate 0.1
Stopping rule Production of 200 generations

Table 4. Control parameters for the MOSA algorithms

Parameter The parameters amount
Maximum Number of Iterations 1000
Population Size 1
Number Variables 35
Number move 5
Initial Temperature Pareto. Cost/10000

Computational Results
Proposed method presented in this research is coded by using multi
objective genetic algorithm which proposed in MATLAB software. In
this part, input parameters which consider general and control
variables are presented and results of proposed algorithm solving are
discussed and the proposed multi objective genetic algorithm is
validated by GAMS. In Table 5, required information for Bandar
Abbas Gas Condensate Refinery Construction Project including
activities time, prerequisite relations, required resources for activities,
and positive and negative financial flows for activities. In this project,
it is assumed that there is no limit in non-renewable resources and
maximum values of these resources are as follow: human resources
equal to 150 (R1=150) and 100 unite machinery (R2=100).
A multi-objective resource-constrained optimization of time-cost trade-off … 673

Table 5. information about the activities of the installation of steel structures

Prerequisite Resource
Activities duration
activities requirements
1 0 - - - 0 0
2 16 1 20 23 24800 55800
3 45 1 41 30 104400 234900
4 30 2 27 23 50700 114075
5 15 2 33 29 31650 71212.5
6 60 2 38 34 147600 332100
7 25 2 32 22 43500 97875
8 16 2 23 15 19360 43560
9 31 2 34 18 48980 110205
10 30 2 36 19 50100 112725
11 61 4 59 47 215330 484492.5
12 45 4,5 44 38 125100 281475
13 25 5,12 39 18 42000 94500
14 45 5 37 26 91800 206550
15 15 5 21 16 18300 41175
16 16 5 29 25 29280 65880
17 45 6 39 40 125100 281475
18 25 2,3,9 55 34 70000 157500
19 439 2,7,8 97 78 2569600 5781600
20 15 13,14,15 34 14 20700 46575
21 16 11,20 12 16 17680 39780
22 45 13,20 23 45 121950 274387.5
23 46 16 19 21 67210 151222.5
24 30 16 34 23 54900 123525
25 45 16,24 23 13 49950 112387.5
26 25 16,25 27 23 49010 110272.5
27 61 25 43 34 158720 357120
28 45 24 23 13 54390 122377.5
29 25 20,22 18 12 22080 49680
30 504 23,24,25,26 120 79 3321050 7472362.5
31 16 27,30 23 12 18020 40545
32 25 26,28,30 36 22 45500 102375
33 30 23,30 19 21 42900 96525
34 30 17,19 21 28 54600 122850
35 45 3,19 65 45 166850 375412.5
36 241 30 95 96 1654900 3723525
37 0 21,28,32,33,34,35,36 - - 0 0
674 (IJMS) Vol. 8, No. 4, October 2015

With the algorithm implementation Surface of the front Pareto of

generated solutions by the algorithm is based on payment type.
According to these results, the contractor is faced with a set of
answers which could be due to the importance of each objective
function (Maximum completion time and maximizing the NPV of the
project) one alternative that represents a method to select the mode is
operational activity.
Therefore to prove efficiency of the proposed algorithms, several
sample problems in small scale including sub sets of Reference are
listed in addition Bandar Abbas Gas condensate Refinery project
with14 (Pan et al., 2008), 18 (Rifat & Önder, 2012), 20 (Luong &
Ario, 2008), and 25 (Kwan et al., 2003) activity is solved by the
proposed algorithms based on the four types of payments.
Indicators Performance
Because of the multi objective nature of the problem, the numerical
results obtained by each algorithm were evaluated in terms of quality
of the produced set of non-dominated solutions and of the associated
approximation of the Pareto front. This quality of the Pareto front
commonly includes not only the number of non-dominated solutions,
but also convergence and distribution concepts. In our experiments,
we evaluated the results (i.e., the pareto fronts) obtained by NSGA-II
and MOPSO algorithms for each problem according to different
quality metrics usually adopted in the literature of MOO. The adopted
metrics were: Spacing, Maximum Spread, and Spread. Each metric
considers a different aspect of the pareto front.
Spacing (S): this metric was proposed by Schott. S estimates the
diversity of the achieved Pareto Front. S is derived by computing the
relative distance between adjacent solutions of the Pareto Front as
follows:
N 1

 (d i  d mean )
(21)
S i 1
( N  1)d mean

where n is the number of non-dominated solutions, di is the distance

between adjacent solutions to the solution VI and is the average
A multi-objective resource-constrained optimization of time-cost trade-off … 675

distance between the adjacent solutions. S=0 means that all solutions
of the Pareto Front are equally spaced. Hence, values of S near zero
are preferred (Kashif Gill et al., 2006).
Maximum Spread (MS): it was proposed by Zitzler et al. (2000)
and evaluates the maximum extension covered by the non-dominated
solutions in the Pareto Front. MS is computed by using as follows.
m
D  (max f
j 1
i
j
 min f i j )2 (22)

where n is number of solutions in the Pareto front, k is the number of

objectives. This measure can be used to compare the techniques and
thus define which of them covers a bigger extension of the search
space. Hence, large values of this metric are preferred.
Criterion of Pareto solutions: Criterion of Pareto solutions numbers
represents Pareto optimum solutions which can be found in every
algorithms and the more number of Pareto front lead to a better
situation (Zitzler et al., 2000).
Diversity: The scale measures disturbance among a set of non-
dominated solutions. The related formula is as follows:
m
d  max( x
i 1
j
i  y ji (23)

In place where, is direct distance between the xi non-

dominated solutions and yi non-dominated solutions. The more the
criterion, the less similar solutions and more diversity among the
responses will be which covers a larger space; this shows the extent of
the Pareto Front.
The algorithm execution time criterion: the algorithm execution
time is considered as quality assessment criteria.
Comparative results: The performance of proposed NSGA-II
algorithms and MOPSO are analyzed to solve condensate Abbas the
algorithm with 35 activities and show examples of problems with
number 14 (Pan et al., 2008), 18 (Rifat & Önder, 2012), 20 (Luong &
Ario, 2008), and 25 (Kwan et al., 2003) activities based on the four
676 (IJMS) Vol. 8, No. 4, October 2015

types of payments. The algorithms results are shown in Tables 6, 7, 8,

and 9.
Table 6. Comparative results based on the LSP
Comparative
Indicators Number
Maximum Diversification
Spacing of Pareto Time
Number of Spread Metric
solution
activities
NSGA-II 0.5 259.9 35 361.67 258.3
J=14 MOPSO 1 265.03 16 243.89 157.29
MOSA 0.54 258.09 28 323.04 39.6
NSGA-II 1 7330.63 16 1055.24 282.55
J=18 MOPSO 0.52 6873.13 14 993.45 151.51
MOSA 0.37 7010.87 15 1026.42 55.62
NSGA-II 0.18 438.16 20 652.58 265.8
J=20 MOPSO 0.36 436.13 19 635.95 160.49
MOSA 0.28 436.08 20 652.74 50.94
NSGA-II 0.17 431.13 22 638.47 303.05
J=25 MOPSO 0.24 366 19 593.89 204.91
MOSA 0.26 429.01 22 636.54 87.43
NSGA-II 0.66 775236.35 73 6265.14 756.42
J=35 MOPSO 0.78 752524.75 35 4936.62 997.06
MOSA 0.9 677365.5 37 4568.86 286.92

Table 7. Comparative results based on the PEO

Comparative
Indicators Number
Maximum Diversification
Spacing of Pareto Time
Number of Spread Metric
solution
activities
NSGA-II 0.51 54.82 13 492.01 252.48
J=14 MOPSO 0.28 54.68 9 230.92 161.54
MOSA 0.25 55.04 10 202.43 41.74
NSGA-II 0.74 808.15 9 745.42 291.15
J=18 MOPSO 1.06 956.76 7 681.1 155.14
MOSA 59 958.78 5 681.11 46
NSGA-II 0.34 146.94 15 579.18 276.39
J=20 MOPSO 1.18 149.67 15 579.23 174.35
MOSA 0.51 151.54 14 559.54 54.47
NSGA-II 0.55 95.14 16 557.09 327.13
J=25 MOPSO 1.22 94.03 13 502.2 177.17
MOSA 1.49 89.61 11 462.01 87.57
NSGA-II 0.79 365502.22 33 13116.45 776.43
J=35 MOPSO 1.13 318008.8 12 7946.42 814.57
MOSA 0.76 306008.4 15 9218.79 286.3
A multi-objective resource-constrained optimization of time-cost trade-off … 677

Table 8. Comparative results based on the ETI

Comparative
Indicators Number of
Maximum Diversification
Spacing Pareto Time
Number of Spread Metric
solution
activities
NSGA-II 0.48 63.63 9 138.56 262.98
J=14 MOPSO 0.71 249.49 8 182.12 137.92
MOSA 0.41 99.3 6 133.69 41.78
NSGA-II 0.44 1209.15 6 607.69 296.56
J=18 MOPSO 1.33 2471.8 4 545.39 153.37
MOSA 0.99 6373.86 4 610.39 55
NSGA-II 0.47 154.49 10 468.39 280.54
J=20 MOPSO 1.56 324.81 6 357.88 120.77
MOSA 0.49 394.57 6 358.11 54.32
NSGA-II 0.31 111.29 8 384.28 329.09
J=25 MOPSO 1.55 98.54 6 303.81 180.73
MOSA 0.54 64.73 5 303.81 90.31
NSGA-II 0.9 516742.1 10 5565.03 707.15
J=35 MOPSO 0.97 498873.4 8 5843.67 786.45
MOSA 0.91 334016.3 9 6134.72 299.64

Table 9. Comparative results based on the PP

Comparative
Indicators Number
Maximum Diversification
Spacing of Pareto Time
Number of Spread Metric
solution
activities
NSGA-II 1 1272.56 12 124.51 253.8
J=14 MOPSO 1.61 1080.91 7 118.45 155.67
MOSA 1.65 1080.7 8 134.78 42.14
NSGA-II 1.2 6919.69 10 780.44 288.01
J=18 MOPSO 1.29 5981.53 6 680.54 161.06
MOSA 1.27 6527.37 8 876.52 60.17
NSGA-II 0.25 3252.61 11 306.01 274.51
J=20 MOPSO 0.3 2612.6 7 269.89 169.67
MOSA 0.93 2106.09 7 291.64 54.44
NSGA-II 0.96 1333.59 10 251.27 320.54
J=25 MOPSO 0.69 2522.06 8 146.44 211.9
MOSA 0.58 2127.25 5 179.26 89.23
NSGA-II 0.89 1717730.93 13 4251.28 727.16
J=35 MOPSO 0.58 907479.38 7 3318.16 521.72
MOSA 0.57 159564.1 10 3619.34 281.79

The main problem that tables and the selected sample problems
with 10, 18, 20 and 25 activity tables represent are the following
topics:
NSGA-II algorithms have a greater ability to achieving higher
number of answers in Pareto front, that from this perspective,
678 (IJMS) Vol. 8, No. 4, October 2015

contractor is faced with more number of options to choose

administrative methods. NSGA-II algorithms have more extensive
solutions in many cases. NSGA-II algorithm have more Pareto front
extensive compared to other algorithms and show more diversity
among the responses. NSGA-II algorithm has more uniformity in
Pareto solutions than other algorithms. MOSA algorithm has less
solving time compared to the NSGA-II and MOPSO algorithm in four
payment and all studied problems.
Validation of the proposed algorithm
For proving efficiency of the Meta heuristic algorithm, solution of the
algorithm is compared with solution of GAMS software. Therefore, to
prove efficiency of the proposed method, several sample problems in
small scale including sub sets of real problem (with 10, 14, 18 and 20
activity) are solved by the proposed NSGA-II algorithm and GAMS
software. Results and duration of executing NSGA-II algorithm and
GAMS software are shown and analyzed which compared in Table 10
to 12. To compute means of different percentage of the results of
GAMS and NSGA-II, we use the following formulation.
Average difference percentage= ((NSGA-II result – GAMS result) /
GAMS result) * 100

Table 10. Results from GAMS software and NSGA-II

Type of payment LSP PEO ETI PP

Problem NPV NPV NPV NPV

NSGA-II 15 4297.59 15 4335.43 14 3840.09 16 1773.26
J=10
GAMS 14 4304.09 14 4436.2 14 3840.34 14 1948.15
NSGA-II 85 3959.68 85 4214.3 85 3954.71 87 2772.83
J=14
GAMS 83 3969.90 83 4214.58 83 3962 83 2780.51
NSGA-II 190 73489.38 188 93336.19 196 77852.39 188 78445.25
J=18
GAMS 187 73816.44 187 93427.17 187 78435.38 187 78611.21
NSGA-II 64 21516.57 64 22502.47 65 22079.6 66 12407.68
J=20
GAMS 63 21595.11 63 22519.73 63 22100.15 63 13612.09
J=25 NSGA-II 67 18784.05 66 19516.25 68 18533.39 65 12968.16
GAMS 65 18834.45 64 19529.84 65 18555.29 64 13412.05
A multi-objective resource-constrained optimization of time-cost trade-off … 679

Table 11. Percentage Average difference between Results from GAMS software and NSGA-II
Type of payment LSP PEO ETI PP
Problem Cmax NPV Cmax NPV Cmax NPV Cmax NPV
J=10 7.1 0.15 7.1 2.2 0 0 0 1.28
J=14 2.4 0.25 2.4 0.006 2.4 0.18 4.8 0.27
J=18 1.6 0.44 0.53 0.097 4.8 0.74 0.53 0.21
J=20 1.5 0.36 1.5 0.07 3.1 0.09 4.7 8.8
J=25 3.7 0.26 3.1 0.06 4.6 0.11 1.5 3.3
The average
3.7 0.29 2.95 0.5 3 0.22 2.33 2.27
percentage difference

Table 12. implementation time from the GAMS software and NSGA-II
Problem J=10 J=14 J=18 J=20 J=25 J=35
NSGA-II 228.76 252.48 282.55 265.6 327.09 707.15
GAMS 414.31 923.63 2475.43 3627.94 5162.57 -

Completion time of the project and reaching to a solution time for

different problems with different activities with different payments
methods are presented in Table 10. This problem is solved by Meta
heuristic simulation method, too. Result of differences for proposed
method and precise method are presented in Table 11 in which their
differences are very small and less than 3%. Also based on Table 12
and Figure 1, time to reach a solution is constant in the proposed
method, but it increases as a quadratic function whereas above results
shows the proposed method is a convergence to an optimum and
solution algorithm.

Fig. 1. Time of sample problems solving in the proposed NSGA-II algorithm and GAMS software
680 (IJMS) Vol. 8, No. 4, October 2015

Conclusion
In this research, the scheduling section limited portion of the
construction of a refinery by using a meta-heuristic approach
investigated. The objectives of this model have been considered,
minimizing project completion time and maximizing the net present
value of the project. Also, the major Constraints and multi-objective
model and Computational time complexity are classified as in the
group NP-Hard problems. Therefore, in this paper, NSGA-II, MOPSO
and MOSA algorithms are used in order to achieve optimal
scheduling. Since every algorithm must be validated before use, the
current study is applied for a real project which is progressive; we
cannot also compare algorithm results with project results; therefore,
to prove efficiency of the algorithm, the algorithm results are
compared with results of solving the problem which is solved by
GAMS software for some problems in small scales. Results represent
that they are almost similar and smaller than 3%. Also, time to reach a
solution in proposed method is constant, but in GAMS software
increases as quadratic function. These results show that the proposed
method is a convergence algorithm to optimal and efficient solution.
The results of the NSGA-II, MOPSO and MOSA algorithm were
investigated with comparative indices. The results of the NSGA-II,
MOPSO and MOSA algorithm for the main problem and sample
problems Indicates NSGA-II algorithm in the different criteria, have
performed better than the other algorithms. For example, the NSGA-II
algorithm in the number of Pareto solutions in problems all have been
more of MOPSO, and MOSA algorithm provides more options for
the decision makers. In the diversity, Maximum Spread (MS) and
Spacing (S) index in the overwhelming of cases performed better than
the other algorithms that Indicates are considered the extent and
greater distribution of response space and uniformity between the
solutions.
Suggestion that can be implemented in process of the project:
 Considering other objectives, such as robust, resources leveling,
and project quality in the objective function
A multi-objective resource-constrained optimization of time-cost trade-off … 681

 Considering the time and activities Implementation cost and the

amount of resources consumed as a fuzzy matter
 Considering non-renewable resource constraints
 Using the other algorithms for improvement and production
solutions
682 (IJMS) Vol. 8, No. 4, October 2015

References
Aboutalebi, R.S.; Najafi, A.A. & Ghorashi, B. (2012). “Solving multi-mode
resource-constrained project scheduling problem using two multi
objective evolutionary algorithms”. African Journal of Business
Management, 6, 4057-4065.
Afshar, A.; Ziaraty, A.K.; Kaveh, A. & Sharifi, F. (2009). “Non-dominated
Archiving Multi colony Ant Algorithm in Time-Cost Trade-Off
Optimization”. Journal of Construction Engineering and
Management, 135, 668–674.
Azimi, F.; Aboutalebi, R. S. & Najafi, A. A. (2011). “Using Multi-Objective
Particle Swarm Optimization for Bi-objective Multi-Mode Resource
Constrained Project Scheduling Problem”. World Academy of Science,
Engineering and Technology, 54, 285-289.
Aladini, K.; Afshar, A. & Kalhor, E. (2011). “Discount Cash Flow Time-
Cost Trade off Problem Optimization; ACO Approach”. Asian
Journal of Civil Engineering (Bulding and Housing), 12, 511-522.
Bashiri, M.; Kazemzadeh, R.; Atkinson, A.C. & Karimi, H. (2011). “Met-
heuristic Based Multiple Response Process Optimization”. Journal of
Industrial Engineering. University of Tehran, Special Issue, 13-23.
Chen, W.N.; Zhang, J. & Liu, H. (2010). A Monte-Carlo Ant Colony System
for Scheduling Multi-mode Project with Uncertainties to Optimize
Cash flows. proceeding of IEEE Congress on Evolutionary
Computation (CEC), 1-8.
Dayanand, N. & Padman, R. (2001). “Project contracts and payment
schedules: the client’s problem”. Management Science, 47, 1654-67.
D¨orner, K.F.; Gutjahr, W.J.; Hartl, R.F.; Strauss, C. & Stummer, C. (2008).
“Nature-inspired meta-heuristics for multi objective activity
crashing”. Omega, 36, 1019–1037.
Dayanand, N. & Padman, R. (1998). Project contracts and payment
schedules: The client's problem. Working Paper, The Heinz School,
Carnegie Mellon University, Pittsburgh, Pennsylvania.
El-Rayes, K. & Kandil, A. (2005). “Time-Cost-Quality Trade-Off Analysis
for Highway Construction”. Journal of Construction Engineering and
Management, 131, 477-486.
Elloumi, S. & Fortemps, P. (2010). “A hybrid rank-based evolutionary
algorithm applied to multi-mode resource-constrained project
scheduling problem”. European Journal of Operational Research,
205, 31–41.
Feng, C.; Liu, L. & Burns, S.A. (1997). “Using Genetic Algorithms to Solve
A multi-objective resource-constrained optimization of time-cost trade-off … 683

Construction Time-Cost Trade-Off Problems”. Journal of Computing

in Civil Engineering, 11, 184-189.
Herroelen, W.S.; De Reyck, B. & Demeulemeester, E.L. (1997). “Project
network models with discounted cash flows: A guided tour through
recent developments”. European Journal of Operational Research,
100, 97-121.
Jongyul, K.; Changwook, K. & Inkeuk, H. (2012). “A practical approach to
project scheduling: considering the potential quality loss cost in the
time–cost tradeoff problem”. International Journal of Project
Management, 30, 264–272.
Kwan, W.K.; Mitsuo, G. & Genji,Y. (2003). “Hybrid genetic algorithm with
fuzzy logic for resource-constrained project scheduling”. Applied Soft
Computing, 2, 174–188.
Kashif Gill, M.; Kaheil, H.Y.; Khalil, A.; McKee, M. & Bastidas, L. (2006).
“Multi objective particle swarm optimization for parameter estimation
in hydrology”. Water Resources Research, 42(Iss.7).
Kim, J.Y.; Kang, C.W. & InKeuk, H. (2012). “A practical approach to
project scheduling: considering the potential quality loss cost in the
time–cost tradeoff problem”. International Journal of Project
Management, 30, 264–272.
Khalilzadeh, M.; Kianfar, F. & Ranjbar, M. (2011). “A Scatter Search
Algorithm for the RCPSP with Discounted Weighted Earliness-
Tardiness Costs”. Life Science Journal, 8, 634-641.
Liu, L.; Burns, S.A. & Feng, C. (1995). “Construction Time-Cost Trade-Off
Analysis Using LP/IP Hybrid Method”. Journal of Construction
Engineering and Management, 121, 446-454.
Luong, D.L. & Ario, O. (2008). “Fuzzy critical chain method for project
scheduling under resource constraints and uncertainty”. International
Journal of Project Managemen, 26, 688–698.
Moselhi, O. (1993). “Schedule compression using the direct stiffness
method”. Canadian Journal of Civil. Engineering, 20(1), 65-72.
Siemens, N. (1971). “A Simple CPM Time-Cost Tradeoff Algorithm”.
Management Science, 17, 354-363.
Moussourakis, J. & Haksever, C. (2004). “Flexible Model for Time/Cost
Tradeoff Problem”. Journal of Construction Engineering and
Management, 130, 307-314.
Marek, M.; Grzegorz, W. O. & Wezglarz, J. (2005). “Simulated annealing
and tabu search for multi-mode resource-constrained project
scheduling with positive discounted cash flows and different payment
models”. European Journal of Operational Research, 164, 639–668.
684 (IJMS) Vol. 8, No. 4, October 2015

Möhring, R.H. & Stork, F. (2000). “Linear pre selective policies for
stochastic project scheduling”. Math Methods Operation Research,
52, 501–15.
Najafi, A.A. & Niaki, S.T.A. (2006). “A genetic algorithm for resource
investment problem with discounted cash flows”. Applied
Mathematics and Computation, 183, 1057–1070.
Pan, H.; Robert, J. & Wilish, C.H. (2008). “Resource Constrained Project
Scheduling with Fuzziness”. Industrial engineering and management
systems conference.
Ritwik, A. & Paul, G. (2013). “A Heuristic Algorithm for Resource
Constrained Project Scheduling Problem with Discounted Cash
Flows”. International Journal of Innovative Technology and
Exploring Engineering (IJITEE), 3, 99-102.
Rifat, S. & Önder Halis, B. (2012). “A hybrid genetic algorithm for the
discrete time–cost trade-off problem”. Expert Systems with
Applications, 39, 11428–11434.
Shu-Shun, L. & Chang-Jung, W. (2008). “Resource-constrained construction
project scheduling model for profit maximization considering cash
flow”. Automation in Construction, 17, 966–974.
Smith-Daniels, D.E.; Padman, R. & Smith- Daniels, V.L. (1996). “Heuristic
scheduling of capital constrained projects”. Journal of Operations
Management, 14, 241–254.
Seifi, M. & Tavakkoli-Moghaddam, R. (2008). “A new bi-objective model
for a multi-mode resource-constrained project scheduling problem
with discounted cash flows and four payment models”. IJE
Transactions A: Basics, 21, 347-360.
Salimi, R.; Bazrkar, N. & Nemati, M. (2013). “Task Scheduling for
Computational Grids Using NSGA- II with Fuzzy Variance Based
Crossover”. Advances in Computing, 3, 22-29.
Ulusoy, G. & Cebelli, S. (2000). “An equitable approach to the payment
scheduling problem in project management”. European Journal of
Operational Research, 127, 262–278.
Varadharajan, T.K. & Rajendran, C. (2005). “A multi-objective simulated
annealing algorithm for scheduling in flowshops to minimize the
makespan and total flowtime of jobs”. European Journal of
Operational Research, 167, 772–795.
Xu, S. (2011). “Applying Ant Colony System to Solve Construction Time-
Cost Trade off Problem”. Advances Materials Research, 179, 1390-
1395.
Xiong, Y. & Kuang, Y. (2008). “Applying an Ant Colony Optimization
A multi-objective resource-constrained optimization of time-cost trade-off … 685

Algorithm-Based Multi-objective Approach for Time-Cost Trade-

Off”. Journal of Construction Engineering and Management, 134,
153-156.
Zhengwen, H. & Yu, X. (2008). “Multi-mode project payment scheduling
problems with bonus–penalty structure”. European Journal of
Operational Research, 189, 1191–1207.
Zheng, D. X. M.; Ng, S.T. & Kumaraswamy, M. M. (2005). “Applying
Pareto Ranking and Niche Formation to Genetic Algorithm-Based
Multi objective Time--Cost Optimization”. Journal of Construction
Engineering and Management, 131, 81-91.
Zitzler, E.; Deb, K. & Thiele, L. (2000). “Comparison of multi objective
evolutionary algorithms: Empirical results”. Evolutionary
Computation Journal, 8, 125–148.