Professional Documents
Culture Documents
Critical Sequence Crashing Heuristic For Resourced Contrained Discrete Time-Cost Trade-Off Problem
Critical Sequence Crashing Heuristic For Resourced Contrained Discrete Time-Cost Trade-Off Problem
Critical Sequence Crashing Heuristic For Resourced Contrained Discrete Time-Cost Trade-Off Problem
Abstract: Despite the importance of project deadlines and resource constraints in construction scheduling, very little success has been
achieved in solving the resource-constrained discrete time–cost trade-off problem (RCDTCTP), especially for large-scale projects. In this
paper a new heuristic method is designed and developed to achieve fast and high-quality solutions for the large-scale RCDTCTP. The pro-
posed method is based on the novel principles to enable effective exploration of the search space through adequate selection of the activities to
be crashed for a resource constrained schedule, by only crashing the activities with zero float in a resource constrained-schedule, which form
the critical sequence. The computational experiment results reveal that the new critical sequence crashing heuristic outperforms the state-
of-the-art methods, both in terms of the solution quality concerning project cost and computation time. Solutions with a deviation of 0.25%
from the best known solutions are achieved within seconds for the first time, for a large-scale project including up to 2,000 activities. The
main contribution of the new heuristic to practitioners and researchers is that it provides a fast and effective method for optimal scheduling of
real-life-size construction projects with project deadlines and resource constraints. DOI: 10.1061/(ASCE)CO.1943-7862.0001077. © 2015
American Society of Civil Engineers.
Author keywords: Project management; Resource management; Construction planning; Scheduling; Costs; Optimization; Cost and
schedule.
Introduction and does not provide any options for the time-cost trade-off
problem (Menesi et al. 2013).
The critical path method (CPM) is commonly used for scheduling Few studies have focused on the resource-constrained time-cost
of construction projects. However, the critical path method is not trade-off problem, which combines the time-cost trade-off problem
capable of optimal scheduling of projects when there are resource with the RCPSP. The majority of the resource-constrained discrete
constraints or project deadlines. Hence, extensive research efforts time–cost trade-off problem (RCDTCTP) studies used problem
have focused on the resource-constrained project scheduling prob- instances, including up to 50 activities in computational experi-
lem (RCPSP) and the time-cost trade-off problem. The general ments. Hegazy and Menesi (2012) reported the performance of
RCPSP aims to achieve the minimum project duration that satisfies a heuristic method for 360 activities. Menesi et al. (2013) used
both the precedence and resource constraints, whereas the time-cost large-size instances, including up to 2,000 activities in computa-
trade-off problem involves minimizing the total direct and indirect tional experiments.
costs without exceeding the project deadline. Because in practice Despite a large amount of the research being concentrated on
many resources (e.g., crews, equipment) are available in discrete designing heuristics and metaheuristics for the RCPSP and DTCTP,
units, much research has focused on the discrete version of the very few of the proposed methods can be applied to real-life con-
time-cost trade-off problem called the discrete time–cost trade- struction projects, which typically include more than 300 activities
off problem (DTCTP). (Liberatore et al. 2001). Besides, a few methods that are capable of
RCPSP and DTCTP are both crucial for planning and manage- solving large-scale problems usually require a significant amount
ment of construction projects because there are resource constraints of computation time to achieve high-quality solutions. The parallel
and project deadlines in the majority of projects. However, popular genetic algorithm (GA) of Kandil and El-Rayes (2006) required
commercial project management software has very limited capabil- 136.5 h on a single processor, and 19.7 h over a cluster of 20 pro-
ities for solving the RCPSP (Mellentien and Trautmann 2001; cessors to obtain the Pareto front for a DTCTP including 720
Hekimoglu 2007; Lu et al. 2008; Bettemir and Sonmez 2014) activities. The metaheuristics of Bettemir (2009) were able to
achieve a 2% deviation from the optimal in 73 min for DTCTP
1
Associate Professor, Dept. of Civil Engineering, Middle East Technical instances including 630 activities. The heuristic of Hegazy and
Univ., Ankara 06531, Turkey (corresponding author). E-mail: Menesi (2012) required 32 min for a RCDTCTP including 360 ac-
rsonmez@metu.edu.tr tivities (Menesi et al. 2013). The constraint programming model
2
Ph.D. Candidate, Dept. of Civil Engineering, Middle East Technical of Menesi et al. (2013) achieved a solution with 6.39% deviation
Univ., Ankara 06531, Turkey. from the upper bound (best known solution) in 120 min. Hence,
3
Expert, Ministry of Development, Necatibey Caddesi No: 110-A,
for the time-cost trade-off problem there is a significant gap be-
Kızılay, Ankara 06570, Turkey.
Note. This manuscript was submitted on March 11, 2015; approved on tween the literature and the needs of real-life construction project
August 26, 2015; published online on October 23, 2015. Discussion period management.
open until March 23, 2016; separate discussions must be submitted for The main objective of this research is to design and develop a
individual papers. This paper is part of the Journal of Construction En- heuristic that can achieve high-quality solutions in a short amount
gineering and Management, © ASCE, ISSN 0733-9364. of computation time for the large-scale RCDTCTP. The research
Hence numerous heuristic and metaheuristic methods were pro- the resource constraints. Leu and Yang (1999) obtained the nondo-
posed for optimal scheduling of projects under resource constraints minated solutions that minimized the sum of direct and indirect
or project deadlines. Priority rule-based scheduling heuristics costs for the RCDTCTP. Chen and Weng (2009) also focused
(Özdamar and Ulusoy 1994; Hegazy et al. 2000; Tormos and Lova on Pareto front optimization for the RCDTCTP and considered ac-
2001) and metaheuristics, including genetic algorithms (Chan et al. tivity interruption. Wuliang and Chengen (2009) presented a multi-
1996; Hartmann 1998; Hegazy 1999; Kim and Ellis 2008; Chen mode resource-constrained discrete time–cost trade-off model to
and Shahandashti 2009; Kim and Ellis 2010; Sonmez and Uysal achieve the Pareto front for the RCDTCTP.
2014), simulated annealing (Lee and Kim 1996; Bouleimen and
Lecocq 2003; Valls et al. 2005), tabu search (Deblaere et al.
2011), and particle swarm optimization (Lu et al. 2008; Wang Critical Sequence
and Qi 2009; Chen 2011), are among the methods proposed for
the RCPSP. In the critical path method, the project duration is calculated by
The methods proposed for the DTCTP include Siemens adding the durations of the activities on the longest path in the
approximation method (Siemens 1971), genetic algorithms (Feng project network called the critical path, which is determined by
et al. 1997; Hegazy 1999; Zheng et al. 2005; Kandil and El-Rayes the precedence relations. When there are resource constraints, the
2006; Fallah-Mehdipour et al. 2012; Sonmez and Bettemir 2012), critical path method is not sufficient to identify the sequence of
ant colony optimization (Ng and Zhang 2008; Xiong and Kuang activities with zero total float so-called critical sequence (Wiest
2008; Afshar et al. 2009), particle swarm optimization (Yang 1964) or critical chain (Goldratt 1997). Wiest (1964) presented
2007; Bettemir 2009; Fallah-Mehdipour et al. 2012), shuffled frog a procedure for calculation of floats in an early study to define
leaping (Elbeltagi et al. 2007), and tabu search (Vanhoucke and and identify the critical sequence. Lu and Li (2003) proposed
Debels 2007). the resource-activity critical-path method to calculate the floats
In an early attempt to integrate resource-constrained scheduling and to determine the sequence of critical activities for resource-
with the time-cost trade-off problem, Chua et al. (1997) proposed a constrained scheduling. Lu et al. (2008) developed a simplified
GA-based model. Leu and Yang (1999) presented a multicriteria simulation-based scheduling system to provide valid floats and
genetic algorithm for the RCDTCTP. Ahn and Erenguc (1998) optimum schedules for the RCPSP.
developed a multipass heuristic procedure for the resource- The significance of critical sequence in resource-constrained
constrained time-cost trade-off problem. Chen and Weng (2009) scheduling (RCS) is similar to the importance of critical path in
adopted a GA-based time-cost trade-off analysis for considering the critical path method. The precedence and resource feasible
resource-constrained scheduling along with time-cost trade-off. project duration can be shortened by crashing the activities that
Wuliang and Chengen (2009) developed a GA for the RCDTCTP. are on the critical sequence(s). Unlike the critical path method,
Hegazy and Menesi (2012) presented a heuristic method that in resource-constrained project scheduling it is sometimes possible
crashes the lowest-cost critical activities that are determined by to shorten the project duration by crashing the activities that are not
the critical path method and resolves any resource overallocation on the critical sequence (Wiest 1964). However, an efficient heu-
by imposing start-delay values to the activities to meet both project ristic method can be designed for the RCDTCTP by only consid-
deadlines and resource limits. In a recent study, Menesi et al. (2013) ering crashing of the activities that are on the critical sequence(s),
proposed a constraint programming model for the RCDTCTP and which is the main focus of this research.
implemented the model for large-scale projects including up to
2,000 activities. The constraint programming approach provides
a flexible method for the RCDTCTP and allows various practical Critical Sequence Crashing Heuristic
constraints such as different types of precedence relations.
A novel heuristic method that is based on crashing of the critical
sequence is designed and developed especially for the large-scale
Resource-Constrained Discrete Time–Cost Trade-Off RCDTCTP. The heuristic method consists of two parts: backward-
Problem forward resource-constrained scheduling, and critical sequence
crashing.
The objective of the resource-constrained time-cost trade-off prob-
lem is to determine a time, cost, or resource mode (option) and a
Backward-Forward Resource-Constrained Scheduling
start date for each activity in such a way that the precedence and
resource constraints are satisfied, and the total direct costs, indirect The critical sequence crashing heuristic (CSCH) starts the search
costs, and delay penalties (liquidated damages) are minimized. by using the normal (uncrashed) modes for the activities. Once
In the discrete version of this problem the relation between the the modes are selected, the start dates of the activities and project
duration of activities and the resources committed is discrete. completion can be determined by using a resource-constrained
the multiproject duration. In a recent study, the backward-forward with the larger activity number is selected as the third criterion.
scheduling method integrated hybrid genetic algorithm has outper- Backward-forward resource-constrained scheduling is per-
formed the state-of-the-art methods for a resource-constrained formed to determine the project duration once the activity to be
multiproject scheduling problem (Sonmez and Uysal 2014). crashed is determined. The project duration obtained by the latest
The backward-forward scheduling method performs resource- mode selections is compared with the project duration obtained by
constrained scheduling twice by using the serial scheduling scheme the previous mode selections (in the first cycle previous mode se-
(Kelley 1963). The serial scheduling scheme sequentially schedules lections includes the normal modes). Crashing is not executed and
the activities (one by one) at their earliest precedence and resource the selected activity is not crashed further if the project duration of
feasible start time, according to a priority list. In backward-forward the latest mode selections is larger than the project duration of the
scheduling, first backward scheduling is executed in the reverse previous mode selections. Finally, the crashing and backward-
time direction, then forward scheduling is performed. An arbitrary forward scheduling stages are executed until all of the activities
project completion time is selected to start backward scheduling in the latest critical sequence(s) are considered for crashing, and
because the exact duration of the feasible schedule is not known the solution with the minimum cost is reported. The flow chart
at the beginning. The resulting backward schedule is adjusted such of the critical sequence crashing heuristic method is illustrated
that the project completion start is equal to time instant zero. in Fig. 1.
In the backward scheduling phase of the proposed CSCH, total
floats of activities that are calculated by the critical path method are
used to determine the priority list. The activity with the smallest Case Example
CPM float is backward scheduled first, and in case of a tie the ac- A case example is presented in Fig. 2 and Table 1 to illustrate the
tivity with the larger activity number is selected. The CPM total proposed CSCH. The modes in Table 1 are the time, cost, and re-
float priority enables the activities on the critical path to be re- source options of the activities for different execution alternatives.
source-constrained scheduled first, and usually works well when For example, for Activity 1 the first mode (M-1) has a duration
the resource constraints are not tight. The forward scheduling is of 6 days, a resource requirement of 12 men, and a direct cost
performed in the order of start times that are obtained in backward of $17,000. The deadline for the case example is 36 days, and
scheduling, and in case of a tie the activity with the larger activity the indirect costs and the liquidated damages are $2,500=day
number is selected. and $5,000=day, respectively. The backward-forward resource-
constrained scheduling is initiated by selecting the normal modes
for all of the activities. The critical path method is performed to
Critical Sequence Crashing determine the CPM floats of the activities as shown in Fig. 3.
The critical sequence(s) is identified for the schedule determined in The backward scheduling priority list is determined as <9, 8, 7,
backward-forward scheduling to start crashing. In the proposed 3, 5, 2, 4, 1, 6 > based on the CPM total floats that are calculated
heuristic method, the critical sequence is defined as the sequence according to the critical path method and by selecting the activity
of activities that has zero total float for a precedence and resource with the larger activity number in case of a tie. The backward
feasible schedule. The critical sequence(s) is identified through the scheduling is performed according to the priority list by scheduling
following steps (Wiest 1964): the activities in the reverse time direction (one by one) at their latest
1. The activities of the schedule obtained by forward improvement precedence and resource feasible finish time, using an arbitrary
(forward schedule) are started as late as possible (right-shifted) project completion time of 50 days. The resulting backward sched-
one by one, in descending order of their finish times in the for- ule is adjusted such that the project start time is equal to Day 0, as
ward schedule (early finish times), without changing the project shown in Fig. 4, and the project duration is obtained as 47 days.
completion time to calculate the late start and late finish times. A priority list of <2, 6, 1, 3, 4, 7, 5, 8, 9 > is obtained for the
2. If several activities have the same early finish time, the initial forward scheduling phase using the start times of the activities in
late start times for these set of activities are calculated by right- the backward schedule of Fig. 4. The forward scheduling is also
shifting the activities so that they start as late as possible performed by using the serial scheduling scheme, according to
(assuming no other activity in the set has been shifted); then the priority list obtained from the backward scheduling phase.
the activities are right shifted in descending order of their initial The objective of the forward scheduling phase is to improve the
late start times, and their late start and late finish times are schedule obtained in the backward scheduling phase. The project
calculated. duration of the resulting schedule (Schedule 1), has decreased to
3. In case of a tie in the initial late start times, the activities are 40 days, at the end of forward scheduling improvement as shown
right-shifted in ascending order of their resource requirements. in Fig. 5.
4. If the tie is not broken, the tied activities are right-shifted in as- In Schedule 1, none of the activities can be right-shifted (started
cending order of their activity numbers. at a later start time without changing the project completion time),
No
No
No
Yes
so all of the activities are identified to be on the critical sequence. scheduling improvement are performed to obtain the next schedule
The crashing options for the activities that are on the critical se- (Schedule 2) as shown in Fig. 6.
quence in Schedule 1 are summarized in Table 2. Activity 5 was In Schedule 2 (Fig. 6), the activities are right-shifted one by one,
crashed first by changing the mode of this activity to Mode 1 (M-1) in descending order of their early finish times to identify the critical
because this activity had the least daily crashing cost. The critical sequence. Activity 9 has an early finish time of 40 days and is right-
path method is performed to determine the CPM floats for the new shifted first, followed by Activity 8, which has an early finish time
activity durations in which the mode of Activity 5 is changed to of 36 days. Activity 9 cannot finish later than Day 40 and Activity 8
M-1. The next backward scheduling priority list is determined cannot finish later than Day 36 without changing the project com-
based on the revised CPM floats. Backward scheduling and forward pletion time. Hence, for Activity 9 and Activity 8 the late finish
6 9
1. (5,13,$19600) 1. (2,6,$6900)
2 2. (8,10,$14500) 2. (4,2,$5800)
1. (7,5,$16500)
7
8
Downloaded from ascelibrary.org by UNIVERSITY OF FLORIDA on 10/26/15. Copyright ASCE. For personal use only; all rights reserved.
3 1. (13,12,$22500)
1. (3,4,$11500)
2. (5,2,$9000)
1. (6,13, $25500)
2. (8,9,$22500) NO
3. (10,6,$21000)
5
1. (Dur., Res., Cost)
1. (9,7,$19600) n. (Dur., Res., Cost)
2. (12,4,$18400)
Indirect costs=$2,500/Day
Liquidated damages=$5,000/Day
Deadline = 36 Days
Resource limit=20/Day
times are equal to the early finish times. Activity 5 and Activity 7 RCS total floats of the activities are calculated by subtracting
have the same early finish time of 31 days. Activity 5 and Activity 7 their late finish times from their early finish times. In Schedule
are right-shifted to calculate the initial late start times for these ac- 2, Activity 2 and Activity 6 are not on the critical sequence.
tivities, assuming that no other activity in the set of activities with The crashing options for the activities that are on the critical
an early finish time of 31 days has been right-shifted. The initial sequence in Schedule 2 are given in Table 3. Hence, the activity
late start time of Activity 5 is calculated as 22, and the initial late selected for next crashing is Activity 9, and Activity 9 was crashed
start time of Activity 7 is calculated as 18. Hence, Activity 5 is by changing the mode of this activity to M-1. Backward scheduling
right-shifted first because it has a later initial late start time, and and forward scheduling improvement are performed to obtain the
the late finish time of Activity 5 is calculated as 31. Activity 7 duration of the next schedule (Schedule 3) as 38 days. In Schedule
is right-shifted next, and the late finish of Activity 7 is also calcu- 3, Activity 3 has the least daily crashing cost, and the mode of this
lated as 31. Activity 4, Activity 1, and Activity 6 are respectively activity is changed to M-2 to obtain Schedule 4, which has a du-
right-shifted next. The late finish times for Activity 4 and Activity 1 ration of 36 days. Similarly, in Schedule 4, the mode of Activity 8 is
are also equal to their early finish times. However, Activity 6 can be changed to M-1 to obtain Schedule 5 with a duration of 34 days.
right-shifted 3 days and can finish at Day 18 without changing the Next, Activity 3 is considered for crashing, however, crashing
project completion time. Finally Activity 3 and Activity 2 are right- Activity 3 to its first mode increased the duration to 39 days in
shifted 0 and 3 days, respectively. Schedule 6, hence this crashing is not executed and Activity 3
is not crashed further. Next, the mode of Activity 1 is changed to
M-1 to obtain Schedule 7 with a duration of 33 days. Finally, the
mode of Activity 6 is changed to M-1; however, crashing Activity 6
Table 1. Activity Data of the Case Example to its first mode in Schedule 8 increased the duration to 38 days,
Duration Resource Direct hence this crashing is not executed. Because at this point all of the
Activity Mode (days) requirement (people) cost ($) activities in the latest critical sequence are considered for crashing,
1 1 6 12 17,000 the heuristic is terminated and the minimum cost solution is re-
2 8 10 12,500 ported. The proposed critical sequence crashing heuristic was able
2 1 7 5 16,500 to achieve the minimum cost of $216,700 in Schedule 5 for the case
3 1 6 13 25,500 example. The time, cost, and resource modes and the start dates of
2 8 9 22,500 the minimum cost solution are shown in Fig. 7.
3 10 6 21,000
4 1 4 6 5,200
5 1 9 7 19,600 Input-Output Interface
2 12 4 18,400
6 1 5 13 19,600 An input-output interface was developed in Microsoft Excel 2013
2 8 10 14,500 to enable simplified data input and output and to facilitate data
7 1 13 12 22,500 exchange with the commercial project management software to en-
8 1 3 4 11,500 hance the use of the proposed critical sequence crashing heuristic in
2 5 2 9,000 practice. The input screen of the interface for the case example is
9 1 2 6 6,900 illustrated in Fig. 8. The heuristic requires a dummy start and a
2 4 2 5,800
dummy finish activity. The successor information and time, cost,
Critical Path
0 7 7 7 8 15
2 6
3 3 10 20 13 28
28 4 32
9
28 0 32
10 13 23
Downloaded from ascelibrary.org by UNIVERSITY OF FLORIDA on 10/26/15. Copyright ASCE. For personal use only; all rights reserved.
7
0 10 10 10 0 23 23 5 28
3 8
0 0 10 23 0 28
10 12 22
5
11 1 23
20
19
18
17
16
1-M2 4-M1
15
5-M2
14
Resource Usage
13
12
11
10
9
8
7
7-M1
6
6-M2
5
4
3-M3
3 2-M1
2
8-M2 9-M2
1
0
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47
Time (Days)
and resource modes, the project deadline, the daily indirect cost, the RCDTCTP using benchmark instances. The proposed algo-
incentives, and the liquidated damages are entered in the input sheet rithm is coded in C# and is built within Visual Studio 2013
of the interface. Once the heuristic is executed, the time, cost, and for both 32- and 64-bit platforms. All of the tests were carried
resource modes and the start dates of activities for the minimum out on two different configurations. The first configuration (C1)
cost solution that satisfies the resource constraints can be obtained consisted of a desktop computer with a P9X79 Chipset (ASUSTeK
in the output sheet as shown in Fig. 9. An executable version of Computer, Taiwan) motherboard, 16 GB 667 MHz DDR3 RAM,
the critical sequence crashing heuristic and the input/output inter- Intel Core i7-3.40 GHz CPU (Intel Corporation, Santa Clara,
face can be obtained from http://www2.ce.metu.edu.tr/∼sonmez/ California), and 64-bit Windows 8.1 operating system. The second
research/csch. configuration (C2) included a laptop computer with a GM45 Chip-
set (Intel Corporation, Santa Clara, California) motherboard, 4 GB
400 MHz DDR2 RAM, Intel Core 2 Duo P8600 2.40-GHz CPU
Computational Experiments (Intel Corporation, Santa Clara, California), and 32-bit Windows
8.1 operating system. In the tests CSCH is executed solely (no other
Computational experiments are conducted to evaluate the perfor- application is ran simultaneously), and overclocking was not
mance of the proposed critical sequence crashing heuristic for performed.
13
7-M1
12
11
10
9
8
7
Downloaded from ascelibrary.org by UNIVERSITY OF FLORIDA on 10/26/15. Copyright ASCE. For personal use only; all rights reserved.
6
6-M2
5
4
4-M1
3 2-M1
5-M2
2
8-M2 9-M2
1
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40
Time (Days)
20
19 Critical
18 sequence
17 activity
16 5-M1
6-M2 4-M1
15
14
Resource Usage
13
12
11
10
9
2-M1
8
7
7-M1
6
1-M2
5
4
3-M3
3
2
8-M2 9-M2
1
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40
Time (Days)
1 M-2 to M-1 2,250 2 while considering a daily resource constraint of 30. The proposed
3 M-3 to M-2 750 3
heuristic achieved the best known solution of $244,000 in 0 s for
8 M-2 to M-1 1,250 2
9 M-2 to M-1 550 4
both C1 and C2. The best solution was also obtained by the genetic
algorithm of Chen and Weng (2009), which identified a Pareto front
20
19 Critical
18 sequence
17 activity
16 5-M1
6-M2 4-M1
15
14
13
Resource Usage
2-M1
12
11
10
9
8
7
7-M1
6
1-M2
5 3-M2
4
9-M1
3
8-M1
2
1
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34
Time (Days)
1 4
1. (12,16,$192)
1. (5,15,$75) 2. (15,10,$150)
3. (18,8,$144)
2 5 8
10
1. (4,16,$64) 1. (22,18,$396)
1. (14,7,$98)
2. (6,10,$60) 2. (24,16,$384) 1. (3,4,$12)
2. (15,6,$90)
3. (8,7,$56) 3. (26,14, $364) 2. (5,2,$10)
3. (16,4, $64)
4. (9,6,$54) 4. (28,12, $336)
3 6
1. (6,13, $78) 1. (14,20,$280)
2. (8,9,$72) 2. (18,15,$270)
3. (10,7,$70) 3. (24,8,$192)
7 9 NO
1. (15,5,$75)
1. (9,17,$153)
2. (18,4,$72) 1. (Dur., Res., Cost)
2. (10,14,$140)
3. (20,3, $60) n. (Dur., Res., Cost)
15 s, and a solution with a DUP value of 4.18% in 120 min. The of large-scale construction projects with project deadlines and re-
performance of the model for the problem including 2,000 activities source constraints. In the proposed heuristic, the backward-forward
worsened and had the model obtained a solution with a DUP value scheduling technique is adopted and crashing of the activities on
of 6.67% in 40 s, and a solution with a DUP value of 6.39% in the critical sequence are considered to present an effective method
120 min. for the resource-constrained discrete time–cost trade-off problem.
The computational results indicate that CSCH can achieve The computational experiments confirmed the effectiveness of the
high-quality solutions for the large-scale RCDTCTP in a short proposed critical sequence crashing heuristic.
computation time by searching a very small portion of the solution The computational tests reveal that the new heuristic achieved
space. CSCH’s effective exploration of the search space can be successful results for small-, medium-, and large-scale projects
mainly attributed to its adequate selection of the activities to be with project deadlines and resource constraints, and outperformed
crashed for a resource-constrained schedule by only considering the state-of-the-art methods with respect to both solution quality
crashing of the activities that are on the critical sequence(s). The and computation time requirement. High-quality solutions with mi-
backward-forward resource-constrained scheduling method also nor deviations from the best known solutions are obtained within
enables achieving an adequate project duration after the mode se- seconds for the large-scale resource-constrained discrete time–cost
lection for the activities are made. However, CSCH is developed for trade-off problem for the first time. The main contribution of the
the general RCDTCTP, integrating additional assumptions, such as new heuristic is that it provides adequate solutions for the real-
allowing splitting of resources, which may increase the processing life-size projects within seconds, and enables significant savings
time of the proposed method. In order to decrease the computation during planning of construction projects with project deadlines
time, only finish to start precedence relations are allowed in CSCH. and resource constraints.
Hence, CSCH requires transformation of the networks, including The test instances that are used to evaluate the performance
start to start, start to finish, and finish to finish relations, to finish of the proposed critical sequence crashing heuristic included up
to start networks (Lu and Lam 2009). to 2,000 activities and reflect the size of the real-life construction
CSCH was able to determine a solution with a total cost of projects. However, the instances do not reflect the complexity of
$48,919,400 for the project with 2,000 activities. The state-of-the-art the real-life construction projects because they are produced by
Acknowledgments
for scheduling with multi-skilled constrained resources.” J. Constr. Eng.
The authors gratefully acknowledge the financial support provided Manage., 10.1061/(ASCE)0733-9364(2000)126:6(414), 414–421.
by the Scientific and Technological Research Council of Turkey Hekimoglu, O. (2007). “Comparison of the resource allocation capabilities
(TÜBİTAK) under Grant #213M253. of project management software packages in resource constrained
project scheduling problems.” M.S. thesis, Middle East Technical
Univ., Ankara, Turkey.
Herroelen, W., and Leus, R. (2005). “Identification and illumination of
References popular misconception about project scheduling and time buffering
in a resource-constrained environment.” J. Oper. Res. Soc., 56(1),
Afshar, A., Ziaraty, A., Kaveh, A., and Sharifi, F. (2009). “Nondominated
102–109.
archiving multicolony ant algorithm in time-cost trade-off optimiza-
Kandil, A., and El-Rayes, K. (2006). “Parallel genetic algorithms for opti-
tion.” J. Constr. Eng. Manage., 10.1061/(ASCE)0733-9364(2009)
mizing resource utilization in large-scale construction projects.” J.
135:7(668), 668–674.
Constr. Eng. Manage., 10.1061/(ASCE)0733-9364(2006)132:5(491),
Ahn, T., and Erenguc, S. S. (1998). “The resource constrained project
491–498.
scheduling problem with multiple crashable modes: A heuristic pro-
Kelley, J. E., Jr. (1963). “The critical-path method: Resources planning and
cedure.” Eur. J. Oper. Res., 107(2), 250–259.
scheduling.” Industrial scheduling, J. F. Muth and G. L. Thompson,
Bettemir, O., and Sonmez, R. (2014). “Hybrid genetic algorithm with
eds., Prentice-Hall, Englewood Cliffs, NJ, 347–365.
simulated annealing for resource-constrained project scheduling.” J.
Manage. Eng., 10.1061/(ASCE)ME.1943-5479.0000323, 04014082. Kim, J. L., and Ellis, R. D. (2008). “Permutation-based elitist genetic algo-
rithm for optimization of large-sized resource-constrained project
Bettemir, Ö. H. (2009). “Optimization of time–cost–resource trade-off
problems in project scheduling using meta-heuristic algorithms.” scheduling.” J. Constr. Eng. Manage., 10.1061/(ASCE)0733-9364
Ph.D. thesis, Middle East Technical Univ., Ankara, Turkey. (2008)134:11(904), 904–913.
Blazewicz, J., Lenstra, J., and Rinnooy Kan, A. H. G. (1983). “Scheduling Kim, J. L., and Ellis, R. D. (2010). “Comparing schedule generation
subject to resource constraints: classification and complexity.” Discrete schemes in resource-constrained project scheduling using elitist genetic
Appl. Math., 5(1), 11–24. algorithm.” J. Constr. Eng. Manage., 10.1061/(ASCE)0733-9364
Bouleimen, K., and Lecocq, H. (2003). “A new efficient simulated (2010)136:2(160), 160–169.
annealing algorithm for the resource-constrained project scheduling Lee, J. K., and Kim, Y. D. (1996). “Search heuristics for resource con-
problem and its multiple modes version.” Eur. J. Oper. Res., 149(2), strained project scheduling.” J. Oper. Res. Soc., 47(5), 678–689.
268–281. Leu, S. S., and Yang, C. H. (1999). “A genetic-algorithm-based resource-
Chan, W. T., Chua, D. K. H., and Kannan, G. (1996). “Construction re- constrained construction scheduling system.” Constr. Manage. Econ.,
source scheduling with genetic algorithms.” J. Constr. Eng. Manage., 17(6), 767–776.
10.1061/(ASCE)0733-9364(1996)122:2(125), 125–132. Li, K., and Willis, R. (1992). “An iterative scheduling technique for re-
Chen, P. H., and Shahandashti, S. M. (2009). “Hybrid of genetic algorithm source-constrained project Scheduling.” Eur. J. Oper. Res., 56(3),
and simulated annealing for multiple project scheduling with multiple 370–379.
resource constraints.” Autom. Constr., 18(4), 434–443. Liberatore, M., Pollack-Johnson, B., and Smith, C. (2001). “Project man-
Chen, P.-H., and Weng, H. (2009). “A two-phase GA model for resource- agement in construction: Software use and research directions.” J.
constrained project scheduling.” Autom. Constr., 18(4), 485–498. Constr. Eng. Manage., 10.1061/(ASCE)0733-9364(2001)127:2(101),
Chen, R. M. (2011). “Particle swarm optimization with justification and 101–107.
designed mechanisms for resource-constrained project scheduling Lova, A., and Tormos, P. (2002). “Combining random sampling and
problem.” Exp. Syst. Appl., 38(6), 7102–7111. backward–forward heuristics for resource-constrained multi-project
Chua, D. K. H., Chan, W. T., and Govtndan, K. (1997). “A time-cost trade- scheduling.” Proc., 8th Int. Workshop on Project Management and
off model with resource consideration using genetic algorithm.” Civ. Scheduling, EURO Working Group, 244–248.
Eng. Syst., 14(4), 291–311. Lu, M., and Lam, H-C. (2009). “Transform schemes applied on nonfinish-
De, P., Dunne, E. J., Ghosh, J. B., and Wells, C. E. (1997). “Complexity of to-start logical relationships in project network diagrams.” J. Constr.
the discrete time-cost tradeoff problem for project networks.” Oper. Eng. Manage., 10.1061/(ASCE)CO.1943-7862.0000062, 863–873.
Res., 45(2), 302–306. Lu, M., Lam, H-C., and Dai, F. (2008). “Resource-constrained critical path
Deblaere, F., Demeulemeester, E., and Herroelen, W. (2011). “RESCON: analysis based on discrete event simulation and particle swarm optimi-
Educational project scheduling software.” Comput. Appl. Eng. Educ., zation.” Autom. Constr., 17(6), 670–681.
19(2), 327–336. Lu, M., and Li, H. (2003). “Resource-activity critical-path method for con-
Elbeltagi, E., Hegazy, T., and Grierson, D. (2007). “A modified shuffled struction planning.” J. Constr. Eng. Manage., 10.1061/(ASCE)0733-
frog-leaping optimization algorithm: applications to project manage- 9364(2003)129:4(412), 412–420.
ment.” Struct. Infrastruct. Eng., 3(1), 53–60. Mellentien, C., and Trautmann, N. (2001). “Resource allocation with
Fallah-Mehdipour, E., Bozorg Haddad, O., Rezapour Tabari, M. M., and project management software.” OR Spektrum, 23(3), 383–394.
Mariño, M. A. (2012). “Extraction of decision alternatives in construc- Menesi, W., Golzarpoor, B., and Hegazy, T. (2013). “Fast and near-
tion management projects: Application and adaptation of NSGA-II and optimum schedule optimization for large-scale projects.” J. Constr.
MOPSO.” Exp. Syst. Appl., 39(3), 2794–2803. Eng. Manage., 10.1061/(ASCE)CO.1943-7862.0000722, 1117–1124.