Professional Documents
Culture Documents
Fuzzy SA For Project Time Cost Trade Off - Milenkovic
Fuzzy SA For Project Time Cost Trade Off - Milenkovic
DOI:10.3233/IFS-2012-0510
IOS Press
Abstract. Managers often face difficulties when making project decisions that involve large number of interrelated activities –
the planning and scheduling of which is project management. Most problems often arise in areas such as product development,
production planning and control and setting up the production facilities. One important aspect of project management is activity
crashing, i.e. reducing the activity time by adding more resources such as workers, overtime and so on. It is important to decide
the optimal crash plan to complete the project within the desired time period. A fuzzy simulated annealing crashing method
is introduced in this paper to evaluate project networks and determine the optimum crashing configuration that minimizes the
average project cost, due to lateness penalties and crashing costs in the presence of vagueness and uncertainty. Through a real
project example the proposed approach is shown effective and efficient in conducting time-cost tradeoff analysis.
Keywords: Project management, project crashing, fuzzy linear programming, simulated annealing
1064-1246/12/$27.50 © 2012 – IOS Press and the authors. All rights reserved
204 M. Milenkovic et al. / A Fuzzy Simulated Annealing approach for project time-cost tradeoff
project managers might be able to bring the project The paper is organized in six sections including the
back on track by incorporating additional resources. In first section for the introduction of the problem. Sec-
project management, this method of mitigating risk is tion 2 gives a brief summary of past work aimed at
known as crashing. project activity crashing methods. In Section 3 we made
Rosenau and Githens [24] state crashing is “spending an introduction to the fuzzy optimization, and present
more money on the project in order to speed up accom- a flexible fuzzification model for fuzzy optimization
plishment of scheduled activities”. Since crashing a problems. Then, in Section 4 we introduced the concept
project represents additional costs, crashing decisions of a fuzzy simulated annealing algorithm as a searching
need to be made in a cost-effective way. When crash- tool for obtaining the solution for fuzzy optimization
ing a project the tradeoff between the crashing cost and problems. In Section 5, we present the project time-
the penalty cost needs to be evaluated. The crashing cost tradeoff problem in the uncertainty environment
concept is focused on reducing the time of the activ- and we apply the proposed methodology on a real case
ities on the critical path. The critical path is the one study. Finally, in Section 6, we conclude this paper with
that can cause a delay of the project because there is no some perspectives on further developments.
slack on that path. The traditional method of crashing
CPM/PERT networks ([25, 27]) only considers aver-
age activity times for the calculation of the critical path,
ignoring the uncertainty related with the duration of the 2. Related work
activities.
Time-cost tradeoff problems can be modeled by Time/cost trade-offs in project networks have been
deterministic or uncertain approaches. Deterministic the subject of intensive research since the develop-
approaches assume that project parameters are pre- ment of the critical path method (CPM) in the late
sented as exact values. However, there are many 50 s. Time/cost behavior in a project activity basically
cases where a great uncertainty in project parameters describes the trade-off between the duration of the activ-
exists. One way to handle uncertainty is via stochas- ity and its amount of non-renewable resources (e.g.
tic time-cost tradeoff approach based on probability money) committed to it. It is generally accepted that the
theory. It requires a prior predictable regularity or a trade-off follows a non-increasing pattern, i.e. expedit-
posterior frequency distribution to construct the prob- ing an activity is possible by allocating more resources
ability distribution of activity times. However, in real (i.e. at a larger cost) to it. There are several contribu-
world applications, sometimes we have to use human tions in the literature, related with our work, and next
judgment instead of stochastic assumptions to deter- we summarize the most relevant ones.
mine activity times. An alternative way to deal with Demeulemeester et al. [8] described two new algo-
imprecise data is to employ the concept of fuzziness, rithms for optimally solving the discrete time-cost
where the vague activity times can be represented by trade-off problem in deterministic activity-on-arc net-
fuzzy sets [32]. The main advantages of methodolo- works under limited resource availability. Authors
gies based on fuzzy theory are that they do not require proposed two procedures, both based on a special proce-
prior predictable regularities or posterior frequency dure for finding the minimal number of node reductions
distributions, and they can deal with imprecise input needed to transform a general network to a series-
information containing feelings and emotions quanti- parallel network. The second algorithm minimizes the
fied based on the decision maker’s subjective judgment. computational effort in enumerating alternative modes
This paper adopts an approach combining the Sim- through a Branch and Bound search tree.
ulated Annealing algorithm and fuzzy set theory to Gutjahr et al. [12] introduce a stochastic optimiza-
develop a time-cost tradeoff model under uncertainty. tion approach to determine a set of efficient measures
A fuzzy linear programming (FLP) time-cost tradeoff that crash selected activities in a stochastic activity
model is presented and solved by Simulated Annealing network. The proposed approach modifies the conven-
(SA). The results obtained by SA are compared with tional stochastic Branch-and-Bound, using a heuristic
those of traditional approach for solving FLP problems. method instead of an exact one to solve the deterministic
The proposed method solves the time-cost tradeoff sub problem. This modification reduces computational
problems under an uncertain environment and proves time and by doing so provides an appropriate method for
suitable through a real example of franchising imple- solving various related applications of combinatorial
mentation project for the Serbian Postal Company. stochastic optimization.
M. Milenkovic et al. / A Fuzzy Simulated Annealing approach for project time-cost tradeoff 205
Kuhl and Tolentino-Pena [15] introduced a dynamic to imitate the probabilistic nature of project networks
simulation-based crashing method to evaluate project throughout the search of optimal solutions. The results
networks and determine the optimum crashing config- showed that GAs can be integrated with simulation
uration that minimizes the average project cost due to techniques to provide an efficient and practical means
lateness penalties and crashing costs. By reevaluating of assessing project time and cost risks.
the project network possible adjustments to the crashing Tareghian and Taheri [26] proposed a meta-heuristic
strategy may be identified and implemented. The out- solution procedure for the discrete time, cost and quality
put of the method includes a distribution of the project tradeoff problem. This problem involves the schedul-
completion time, a distribution of the project total cost, ing of project activities in order to minimize the total
and the project cost savings. cost of the project while maximizing the quality of the
Vanhoucke et al. [28] described a solution procedure project and also meeting a given deadline. They applied
for the deadline problem in which three special cases of electromagnetic scatter search to solve this problem.
time switch constraints are involved. These constraints Yang [29] developed and tested a particle swarm opti-
impose a specified starting time on the project activities mization (PSO) algorithm for project crashing analysis.
and force them to be inactive during specified time peri- The optimization results include direct and project cost
ods. They propose a branch-and-bound algorithm and curves, both of which can assist project managers in per-
a heuristic procedure which both make use of a lower forming what-if analyses on the project deadline and
bound calculation for the discrete time/cost trade-off budget. The application of the proposed algorithm is
problem. demonstrated through an example project and a real-life
Deckro et al. [7] developed a series of nonlinear case.
time/cost tradeoff models for project management. A Yang [30] proposed elitist PSO algorithm to facili-
base model with quadratic cost relations is developed tate bicriterion time-cost trade-off analysis. Algorithm
and then extended to include other operational consid- modifies a population based search procedure, parti-
erations. The use of nonlinear functions to describe cle swarm optimization, by adopting an elite archiving
time/cost relationships in project scheduling environ- scheme to store nondominated solutions and by aptly
ments permits a more accurate representation of the using members of the archive to direct further search.
operational setting than is possible with linear or piece- Ng and Zhang [19] presented a multiobjective
wise linear approximations. approach known as the ant colony system to model time
Moussourakis and Haksever [18] presented a flexible cost optimization problems so as to optimize the total
mixed integer-programming model for the solution of project duration and total cost simultaneously.
the time/cost tradeoff problem. The model presented in Zheng et al. [31] introduced a multiobjective
this paper makes minimal assumptions and accommo- approach that aims to optimize total time and total cost
dates any type of cost function that is linear, piecewise simultaneously by utilizing appropriate GAs concepts
linear or discrete. The model can be used for answering and tools. The model introduces a MAWA (modified
various “what if” questions that may be very helpful adaptive weight approach) to replace traditional fixed
to a project manager in making rational decisions. The or random weights and integrates time and total cost
basic model minimizes the total cost which is the sum into a single objective for simulation. The concept is
of direct and indirect costs, subject to a project deadline illustrated through a simple manual simulation and the
constraint. A simple modification in the model changes results indicate that the model could assist decision
the focus from minimizing total cost to minimizing makers in concurrently arriving at an optimal project
project completion time subject to a resource constraint. duration and total cost.
Phillips and Dessouky [20] proposed a solution Leu et al. [17] proposed a flexible approach employ-
procedure for solving the project time/cost tradeoff ing genetic algorithms (GAs) and fuzzy set theory to
problem of reducing project duration at a minimum develop a time-cost trade-off model for construction
cost. They obtained a solution by locating a minimal cut projects under uncertainty. They treat the activity dura-
in flow network derived from the original project net- tion as a fuzzy number. The main focus of the paper was
work. Minimal cut is then utilized to identify the project to determine the impact of uncertain activity durations
activities which should experience a duration modifica- on project direct costs.
tion in order to achieve the total project reduction. Ke et al. [13] proposed three new fuzzy time-cost
Feng et al. [10] developed a hybrid approach that trade-off models, the α-cost minimization model, the
combines simulation techniques and genetic algorithms expected cost minimization model and the credibility
206 M. Milenkovic et al. / A Fuzzy Simulated Annealing approach for project time-cost tradeoff
is the corresponding vector of the fuzzy limits of the 3.1. Flexible fuzzification model
resources.
The first model to solve decision problems in a fuzzy Ribeiro and Moura-Pires [21–23] proposed a method
environment was proposed by Bellman and Zadeh [4]. for fuzzy optimization problems that include fuzzy
These authors consider that the decision on each alter- coefficients either in the objective function or in the
native results from the intersection of the objectives constraints. The whole process is designated as “flex-
and the constraints and that the union of the decisions ible approach” for fuzzy optimization problems in its
on each alternative gives the “best” decision. A fuzzy formalization is: assuming the formulation for fuzzy
decision can then be seen as an intersection of goals and optimization problems defined in (4–6), we made a
constraints, which mean that it is a symmetric model. A transformation into the following system of equations:
maximization fuzzy decision is defined as a point in the
space of alternatives, in which the membership function max Z = cT x (10)
x
of a fuzzy decision reaches its maximum value.
In general, the process of solving a fuzzy optimiza- max M = ∩(µ0 (cT x), µ(Ai x), µ(Ai x)) (11)
tion problem consists on the following steps: Subject to:
• Depending on the problem type, formalize it in cT x ≤ Z0 + (1 − µ0 (cT x))p0 (12)
the form of a linear programming problem or in
the form of a multiobjective form, or even as a non Ai x ≤ bi + (1 − µi (Ai x))pi (i = 1, 2, ..., m)(13)
linear programming problem.
Ak x ≥ bk − (1 − µk (Ak x))pk (k = 1, 2, ..., n) (14)
• If we intend to fuzzify the objectives, define the
goal for those objectives. µ0 (cT x), µi (Ai x), µk (Ak x) ∈ [0, 1] (15)
• Select the membership functions to represent the
fuzzification of any parameter and constraints, x≥0 (16)
such as triangular, sinusoidal, trapezoidal, or other.
It should be noted that all the fuzzy parameters are
• Define the membership functions with the nec-
presented with a tilde on the top. The ∩ symbol repre-
essary parameters for the preference value and
sents the intersection of all membership values. The
tolerances.
second objective represents the maximization of the
• Define thresholds for the allowable degree
intersection of all membership values, i.e. this goal
of the deviations/violations in the constraints
gives the best value of the minimum (intersection) of
satisfaction.
the violations of constraints.
• Define the aggregation operator to combine the
The most frequently used membership functions -
constraints (and goals), e.g. the t-norm operator
to represent the deviations accepted for the fuzzy
minimum.
objectives (goals), constraints and/or coefficients - are
• Solve the problem.
triangular functions. In particular, let µ0 denote the
The first proposal to formalize a fuzzy linear pro- membership function for the objective function and µi
gramming problem is due to Zimmermann [32]. This or µk denote the membership functions for the ith or kth
author considers that there exist flexibility in the viola- constraint. Let p0 and pi and pk be the permissible tol-
tion of both constraints and objective(s). This model is erances for the objective function and the constraints.
based on the symmetric model of Bellman and Zadeh Then we may decide µ0 and µi or µk to be either
[4]. non-increasing or non-decreasing as well as continuous
membership functions as follows:
⎧
⎪ 1 Ai x bi
⎨ Ai x − b i
µi (Ai x) = 1 − bi < Ai x bi + pi (17)
⎪
⎩0 pi
Ai x > bi − pi
⎧
⎪ 1 Ai x bi
⎨ Ai x − b i
µi (Ai x) = 1 + bi − pi Ai x < bi (18)
⎪
⎩0 pi
Ai x < bi − pi
208 M. Milenkovic et al. / A Fuzzy Simulated Annealing approach for project time-cost tradeoff
⎧
⎪ 1 cT x < Z0
⎨ Z0 − c T x
T
µ0 (c x) = 1 + Z0 ≤ cT x ≤ Z0 + p0 (19)
⎪
⎩0 p0
cT x > Z0 + p0
still critical. If the path is still critical then the crashed be completed. Using events specifies the relationship
activity is the one with second lowest cost-time slopes between the activities. An event represents a point in
and this process is continued until the goal has been time that implies the completion of some activities and
reached. If, a reduction in a critical activity time causes the beginning of new ones, the beginning and end point
a non-critical path or paths to become critical, then of an activity are thus expressed by two events.
the crash will be continued along with the new criti- Now let’s define the variables of the problem:
cal path based on the lowest cost time slopes of the new
path. • ti – the time when an event i will occur, measured
Now, let us introduce a linear programming formula- from the start of the project, i = 1, ..., n;
tion for the optimal project crashing problem. The basic • tj – the time when an event j will occur, measured
data requirements are: from the start of the project, j = 1, ..., n;
• tij – Amount of time (in terms of days, weeks, or
• We have to know the project network with activity months or some other units) for each activity i − j;
time; • tijn – Normal time (in terms of days, weeks, or
• To what extent an activity can be crashed; months or some other units) for each activity
• Crash cost associated with per unit of time for all i−j ;
activities. • tiju – Crash time (in terms of days, weeks, or months
or some other units) for each activity i − j
As we know, a project is the combination of some • tij – Amount of time (in terms of days, weeks,
activities, which are interrelated in a logical sequence in or months or some other units) that each activity
the sense that the starting of some activities is dependent i − j can be crashed, i = 1, . . . , n, j = 1, . . . , n;
upon the completion of some other activities. These • Cij – crash cost per unit of time for activity i − j;
activities are jobs which require time and resources to • T – Proposed project deadline date;
210 M. Milenkovic et al. / A Fuzzy Simulated Annealing approach for project time-cost tradeoff
Hence, the linear programming formulation of the terms of starting and ending points (Fig. 2). Moreover,
time-cost tradeoff problem is as follows: as we can see the total number of events is 19.
n
n Before modeling the problem, it is required to deter-
Minimize Z = Cij tij (20) mine the extent of the project crashing time. According
i=1 j=1
to the Critical Path Method (CPM), the critical path for
normal length of activities is path B-H-M-O-P-Q-S-W,
Subject to: and the time length of this path is 280 time units (in
this case days). The critical path based on crash time
tij ≤ tijn − tiju , i = 1, ..., n, j = 1, ..., n, i =
/ j (21) of activities remain the same as previous B-H-M-O-P-
Q-S-W, and the time length is 262 days. Suppose now
tj − tij − ti ≥ tij , i = 1, ..., n, that we want to adjust the deadline of the project on its
expected completion date based on crash time.
j = 2, ..., n, i =
/ j (22)
According to LP formulation (20–23) of crashing
tn ≤ T (23) project time, the problem for this particular case study
can be interpreted as:
The objective (20) is to minimize the cost of crashing
the total project. Constraints (21) represent crash time Z = 30t12 + 30t13 + 20t24 + 20t28
constraints. The time to complete an activity can be + 15t25 + 16t36 + 20t39 + 18t37
reduced by increasing the resources or by improving the
productivity which also require additional resources. + 25t411 + 20t810 + 25t811 + 25t911
But it is not possible to reduce the time required to + 24t712 + 30t1112 + 120t1213 (24)
complete an activity after a certain threshold limit. Con-
straints (22) describe the network. As we mentioned, + 15t1314 + 10t1415 + 26.67t1416
activities of a project are interrelated, the starting of + 20t1518 + 20t1517 + 20t1618
some activities is dependent upon the completion of
some other activities. Since event 1 will start at the + 15t1619 + 16.67t1819 → Minimize
beginning of the project, we begin by setting the occur- Subject to:
rence time for event 1 equals to zero. Constraints (23) is Crashing time constraints:
the project completion constraint. This constraint will
recognize that the last event must take place before the t12 ≤ 1 (25)
project deadline date. However, in real world for many
projects we have to use human judgment for estimating. t13 ≤ 1 (26)
A way to deal with this imprecise data is to employ the t1819 ≤ 1 (27)
concept of fuzziness.
In the next chapter, for a real project, we fuzzified the Network constraints:
LP formulation according to the approach discussed in
t1 = 0 (28)
Chapter 3. Then the Simulated annealing algorithm is
used and the obtained results are compared with those t2 − t1 + t12 ≥ 5 (29)
of crisp linear programming and Zimmermann’s fuzzy
linear programming approach. t3 − t1 + t13 ≥ 5 (30)
t19 − t18 + t1819 ≥ 62.5 (31)
5.1. Fuzzy Simulated Annealing crashing
methodology applied to a project of Project completion constraint:
franchising implementation t19 ≤ 262 (32)
For the sake of testing our approach we used the The optimal solution for this exact formulation of
data from the project of franchising implementation project crashing problem, obtained using the CPLEX
managed by the Serbian Postal Company. All activi- software package is given in Table 3.
ties included in the project as well as needed time and Now, we will consider that in reality, unpredictable
costs, are given in Table 1. For the purpose of modeling situations may affect on the project activity times i.e.
the problem, it is necessary to define the activities in after obtaining crisp results we should relax the crashing
M. Milenkovic et al. / A Fuzzy Simulated Annealing approach for project time-cost tradeoff 211
Table 1
Activities, time, costs and relationships for the project network used for the example
Activity Description Depends on Normal time Crash time Normal cost Crash cost
(days) (days) (monetary units) (monetary units)
A Resource identification – 5 4 120 150
B Defining components and deadlines – 5 4 120 150
C Market segmentation by priority A 10 9 170 190
D Profile preparing for each market A 85 82 1470 1530
E Market analysis and franchise A 70 68 1210 1240
location identifying
F Identification of common outputs for B 22.5 20 390 430
rationalization
G Selling information preparation B 21 20 360 380
H Establishing the project budget and B 62.5 57.5 1080 1170
estimated costs
I Developing the process for C 30 28 740 790
categorization of outputs based on
functions preferred by Post
Company
J Developing the criteria for franchise D, E 5 4.5 120 130
selection
K Developing the process for franchise D, E 10 8 250 300
selection
L Developing the process of contract F, G 20 18 490 540
management
M Establishing the project budget and H 62.5 60 1540 1600
estimated costs
N Preparing the draft of plan I, J, K, L 10 9 250 280
O Presenting to superior M, N 5 4 0 120
P Evaluation of existing contract model O 20 18 300 330
Q Translation of contract on Serbian P 10 9 120 130
R Legal checking P 10 8.5 100 140
S Making necessary changes to adapt Q 52.5 50 1000 1050
in business environment
T Development of basic contract Q 50 47 1000 1060
models for each output
U Creating and implementation of R 30 25 600 700
franchise structure of payment
V Provision of all necessary R 45 43 600 630
permissions from higher instances
W Establishing the project budget and S, Y, U 62.5 59.5 1080 1130
estimated costs
Table 2
Tolerance intervals for project time-cost tradeoff problem
Constraints of “less or equal” type Constraints of “greater or equal” type
Lower bound Deviation Upper bound Lower bound Deviation Upper bound
(ci ) (di ) (ci +di ) (ci -di ) (di ) (ci )
Z = c0 = 350 100 450 4 d25 = 1 c25 = 5
c1 = 1 d1 = 1 2 4 d26 = 1 c26 = 5
c2 = 1 d2 = 5 6 9 d27 = 1 c27 = 10
c3 = 1 d3 = 1 2 69 d28 = 1 c28 = 70
c4 = 3 d4 = 2 5 20.5 d29 = 2 c29 = 22.5
c5 = 2 d5 = 1 3 60.5 d30 = 2 c30 = 62.5
c6 = 2.5 d6 = 5 7.5 82 d31 = 3 c31 = 85
c7 = 1 d7 = 0.5 1.5 −1 d32 = 1 c32 = 0
c8 = 5 d8 = 0.5 5.5 20 d33 = 1 c33 = 21
c9 = 2 d9 = 1 3 −1 d34 = 1 c34 = 0
c10 = 0.5 d10 = 0.5 1 3 d35 = 2 c35 = 5
c11 = 2 d11 = 1 3 9 d36 = 1 c36 = 10
c12 = 2 d12 = 0.5 2.5 −1 d37 = 1 c37 = 0
c13 = 2.5 d13 = 2 4.5 29 d38 = 1 c38 = 30
c14 = 1 d14 = 2 3 8 d39 = 2 c39 = 10
c15 = 1 d15 = 0.5 1.5 57.5 d40 = 5 c40 = 62.5
c16 = 2 d16 = 0.5 2.5 4 d41 = 1 c41 = 5
c17 = 1 d17 = 0.5 1.5 17 d42 = 3 c42 = 20
c18 = 1.5 d18 = 1 2.5 9 d43 = 1 c43 = 10
c19 = 2.5 d19 = 1 3.5 8 d44 = 2 c44 = 10
c20 = 3 d20 = 1 4 47 d45 = 3 c45 = 50
c21 = 5 d21 = 1 6 –2 d46 = 2 c46 = 0
c22 = 2 d22 = 1 3 52 d47 = 0.5 c47 = 52.5
c23 = 3 d23 = 1 4 29 d48 = 1 c48 = 30
c24 = 262 d24 = 5 267 62.5 d49 = 0 c49 = 62.5
44 d50 = 1 c50 = 45
time constraints and the network constraints to reflect considered project crashing problem. In the implemen-
the real situation during the project execution. Hence, tation of the SA algorithm it is needed to choose the
we fuzzified the problem, considering the tolerance following parameters:
intervals for the objective function and all constraints.
• How to generate a state y neighbor of x;
The total number of constraints is 51, and for each par-
• Which aggregation function to use;
ticular constraint, different permissible tolerances are
• The number of neighbours to generate - N(t);
defined in consultation with experts involved in real-
• The decreasing temperature function - T (t);
ization of considered project (Table 2).
• The stopping criterion.
We first solved the problem by using the Zimmer-
mann’s approach, which considers that in a fuzzy The choice of how to generate a state y as a neighbour
environment there is no difference between goals and of x, is done by defining a new state which is a random
constraints. Within this approach the objective moves point y where the distance to the point x is random and
into a constraint with a defined stretched boundary less than t, defined by
as any other fuzzy constraint. The resolution process ⎧
⎪
⎪ 0.5 if t < 100
proceeds to transform the problem into a crisp one, ⎪
⎪
using the objective for the maximization of the min- ⎨ 1.0 if 100 ≤ t < 200
η(t) =
imization of the membership values of the constraint ⎪
⎪ 1.5 if 200 ≤ t < 300
⎪
⎪
deviations, determined in functions like those in expres- ⎩ 2.5 if t ≥ 300
sions (17–19). The results of the considered problem
by using fuzzy Zimmermann approach are given in For the aggregation function the intersection of all
Tables 3 and 4. membership values of the constraints is selected, and
Next, on the base of flexible fuzzification model the operator used is the t – norm min. The intersection
presented in Chapter 3, we developed a suitable Sim- represents the logical “and” to signify that all con-
ulated Annealing algorithm for solving the model for straints must be satisfied to a certain degree (Min). The
M. Milenkovic et al. / A Fuzzy Simulated Annealing approach for project time-cost tradeoff 213
Table 3
Solutions of the Zimmermann’s approach and flexible fuzzification model for crashing project problem (α = 0.98)
Variable A B C Variable A B C Variable A B C
Z= 440.01 352.7 351.47
M 0.97 0.982
t12 0.00 0.00 0.54 t1314 2.00 2.01 0.44 t8 90.00 101.16 90.97
t13 1.00 1.13 0.54 t1415 1.00 1.01 0.26 t9 26.50 26.25 28.50
t24 0.00 0.00 0.60 t1416 0.00 0.00 0.57 t10 111.50 111.16 95.90
t28 0.00 0.00 0.09 t1518 2.50 2.53 0.02 t11 111.50 111.13 101.21
t25 0.00 0.00 0.15 t1517 0.00 0.00 0.29 t12 121.50 121.08 130.52
t36 0.00 0.00 0.85 t1618 0.00 0.00 2.61 t13 125.50 125.84 135.84
t39 0.00 0.00 0.00 t1619 0.00 0.00 0.85 t14 143.50 143.74 155.46
t37 5.00 5.01 1.19 t1819 3.00 3.02 1.88 t15 152.50 152.70 166.00
t411 0.00 0.00 0.52 t1 0.00 0.00 0.77 t16 172.50 172.69 165.49
t810 0.00 0.00 0.47 t2 5.00 4.97 5.99 t17 202.50 202.72 216.37
t811 0.00 0.00 0.06 t3 4.00 3.83 5.71 t18 202.50 202.66 219.43
t911 0.00 0.00 0.59 t4 81.50 81.16 15.50 t19 262.00 262.14 263.19
t712 2.50 2.55 0.16 t5 90.00 101.19 76.68
t1112 0.00 0.00 0.87 t6 26.50 26.28 28.29
t1213 1.00 0.21 0.59 t7 61.50 61.27 67.27
struction Engineering and Management, ASCE 134(9) (2008), [27] B.W. Taylor, Introduction to Management Science, Virginia
721–728. Polytechnic Institute and State University, USA, 2009 (pp.
[20] S. Phillips and M. Deesouky, Solving the project time/cost 320–389).
tradeoff problem using the minimal cut concept, Management [28] M. Vanhoucke, F. Demeuleester and W. Herroelen, Discrete
Science 24 (1977), 393–400. time/cost trade-offs in project scheduling with time-switch
[21] F.M. Pires, J.M. Pires, R.A. Ribeiro, Solving Fuzzy Opti- constraints, The Journal of the Operational Research Society
mization Problems: Flexible Appraches using Simulated 53 (2002), 741–751.
Annealing, Proceedings of the World Automation Congress, [29] I.T. Yang, Performing complex project crashing analy-
Monpellier, France, 1996 (pp. 1–10). sis with aid of particle swarm optimization algorithm,
[22] R.A. Ribeiro and F.M. Pires, Fuzzy Linear Programming via International Journal of Project Management 25 (2007),
Simulated Annealing, Kybernetika 35 (1999), 57–67. 637–646.
[23] R.A. Ribeiro and L.R. Varela, Fuzzy optimization using Sim- [30] I.T. Yang, Using elitist particle swarm optimization to facilitate
ulated Annealing: An example Set, In: Fuzzy Sets Based bicriterion time-cost tradeoff analysis, Journal of Construc-
Heuristics for Optimization, Studies in Fuzziness and Soft tion Engineering and Management, ASCE 133(7) (2007),
Computing Series, Springer 126(2003), 159–180. 498–505.
[24] M.D. Rosenau and G.D. Githens, Successful Project Man- [31] D.M. Zheng, S.T. Ng and M.M. Kumaraswamy, Applying
agement: A Step-by-step Approach with Practical Examples, a genetic algorithm-based multiobjective approach for time-
Wiley, Hoboken, NJ, 2005 (pp. 1–18). cost optimization, Journal of Construction Engineering and
[25] A.H. Taha, Operation Research: An Introduction, Prentice- Management 130 (2004), 168–176.
Hall, USA, 2003 (pp. 266–288). [32] H.J. Zimmermann, Fuzzy Set Theory - and its Applications,
[26] H.R. Tareghian and S.H. Taheri, A solution procedure for second edition, Kluwer Academic Publishers, Netherlands,
the discrete time, cost and quality tradeoff problem using 1991 (pp. 241–272).
electromagnetic scatter search, Applied Mathematics and
Computation 190(2) (2007), 1136–1145.
Copyright of Journal of Intelligent & Fuzzy Systems is the property of IOS Press and its content may not be
copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written
permission. However, users may print, download, or email articles for individual use.