Sei sulla pagina 1di 12

Improved Critical Chain Project Management Framework

for Scheduling Construction Projects


Guofeng Ma 1; Aimin Wang 2; Nan Li, M.ASCE 3; Lingyun Gu 4; and Qi Ai 5

Abstract: Construction projects are subject to a wide range of constraints, such as project complexity, resource scarcity, and duration
uncertainty. The critical chain project management (CCPM) has emerged as a method for construction scheduling. This paper proposes
Downloaded from ascelibrary.org by University of Melbourne on 10/13/14. Copyright ASCE. For personal use only; all rights reserved.

an improved CCPM framework to enhance the implementation of CCPM in construction project management practices. The framework
addresses two major challenges in CCPM-based construction scheduling, including buffer sizing and multiple resources leveling. Buffers
play a key role in ensuring successful schedule management. However, buffers generated by the existing sizing methods are either unnec-
essarily large, which wastes resources, or insufficiently robust against various uncertainties. Resource leveling is another critical challenge in
CCPM-based construction scheduling because it requires a fundamentally different approach from the resource leveling used in traditional
scheduling methods. The proposed framework improves buffer sizing by integrating into the buffer sizing process various uncertainties that
affect construction scheduling but are not factored in by current practice. These uncertainties are assessed in five dimensions with their
respective metrics developed in the framework. Furthermore, the framework explores the feasibility of multiple resources leveling in
CCPM-based construction scheduling, with a novel method that manages the trade-offs between activity duration and resource usages based
on a multimodal activity execution structure. Three case studies were undertaken in this paper. The results showed that the proposed frame-
work outperformed existing buffer sizing methods by generating buffers with reasonable sizes and sufficient robustness against uncertainties.
The results also proved the feasibility and effectiveness of performing multiple resources leveling in CCPM-based construction scheduling.
DOI: 10.1061/(ASCE)CO.1943-7862.0000908. © 2014 American Society of Civil Engineers.
Author keywords: Critical chain project management; Construction management; Buffer sizing; Scheduling; Uncertainty; Multiple
resources leveling; Cost and schedule.

Introduction The CCPM develops a sound schedule using buffer management


to avoid project overruns (Cohen et al. 2004). To manage the
Construction projects are subject to a wide range of constraints, project, the CCPM uses time buffers, a basic TOC tool, and
such as project complexity, resource scarcity, and duration uncer- provides a simple tool for monitoring projects and setting realistic
tainty. Failing to manage any of these constraints can create sched- deadlines (Rozenes et al. 2006). The main advantages of the CCPM
uling distortions. The critical path method (CPM) and the program include its simplicity, its explicit protection against stochastic varia-
evaluation and review technique (PERT) are scheduling methods tion, and a set of reasonable control guidelines it provides (Trietsch
that have been extensively researched and widely used to develop and Baker 2012). The CCPM has increasingly been used in con-
and manage project schedules. These methods, however, are not struction scheduling and the method has been subject to a number
always justifiable and effective due to their reliance on certain of refinements and extensions (Ahlemann et al. 2013).
assumptions that become more questionable as projects become There are two major challenges with the implementation of
increasingly complex and demands for resources grow (Kastor and the CCPM, namely, buffer sizing and multiple resources leveling.
Sirakoulis 2009; Hall 2012). To address such challenges, the theory The CCPM does not assign safety time to each individual activity.
of constraints (TOC), introduced by Glodratt (1997), was applied to Instead, it generates an aggregated safety time and places it at
project management, resulting in the development of a new sched- certain positions on the critical chain to act as a buffer. The use
uling method called critical chain project management (CCPM). of buffers, which takes into account both precedence constraints
and resource constraints, serves to shorten the project duration and
1
Associate Professor, Dept. of Construction Management and Real
enhance the robustness of project scheduling. Buffers play a key
Estate, Tongji Univ., Shanghai 200092, China. role in the CCPM (Raz et al. 2003), and sizing the buffers is crucial
2
Assistant Professor, Dept. of Innovation and Strategy, Shanghai Jiao to enhancing their effectiveness and, ultimately, to ensuring suc-
Tong Univ., Shanghai 200240, China. cessful project schedule management. Two methods are currently
3
Assistant Professor, Dept. of Construction Management, Tsinghua used for buffer sizing: the cut and paste method (C&PM) and the
Univ., Beijing 100084, China (corresponding author). E-mail: nanl@ root square error method (RSEM). However, the C&PM adopts a
usc.edu linear procedure and the size of the calculated buffer increases
4
Graduate Student, Dept. of Construction Management and Real Estate, linearly with the length of the critical chain, which often results
Tongji Univ., Shanghai 200092, China. in unnecessarily long schedules and leads to waste of resources
5
Graduate Student, Dept. of Construction Management and Real Estate,
(Herroelen and Leus 2001). The RSEM is less influenced than the
Tongji Univ., Shanghai 200092, China.
Note. This manuscript was submitted on March 4, 2014; approved on C&PM by the length of the critical chain (Tukel et al. 2006), but
June 6, 2014; published online on July 3, 2014. Discussion period open may lead to schedules that are not sufficiently robust against
until December 3, 2014; separate discussions must be submitted for indi- various uncertainties.
vidual papers. This paper is part of the Journal of Construction Engineer- Moreover, resource leveling, which is important to successfully
ing and Management, © ASCE, ISSN 0733-9364/04014055(12)/$25.00. implement construction schedules, has remained a significant

© ASCE 04014055-1 J. Constr. Eng. Manage.

J. Constr. Eng. Manage.


challenge for the CCPM. Resource leveling problems arise when- Some researchers have argued that buffer sizing is a nondeter-
ever it is expedient to reduce the fluctuations in patterns of resource ministic polynomial (NP) problem (Kolisch and Drexl 1997), and a
utilization over time while maintaining compliance with the pre- number of algorithms have been proposed to address this issue.
scribed project completion date. These problems are further com- Long and Ohsato (2008) developed a fuzzy CCPM for project
plicated when more than one type of resource is involved. CPM and scheduling under resource constraints and uncertainty. The size
PERT base multiple resources leveling on adjusting the free floats of the project buffer is determined by computations with fuzzy
of individual activities. However, CCPM does not involve free numbers. Zhao et al. (2010) developed a genetic algorithm to iden-
floats and its multiple resources leveling requires a fundamentally tify the critical chain and obtain the optimal start time of each
different approach that has been barely examined in the literature. activity under the most optimistic duration of each activity and re-
In order to address the previous two challenges and enhance the source constraint. Hazir et al. (2010) proposed a two-stage robust
implementation of CCPM in construction project management scheduling algorithm that concentrates on slack-based measures to
practices, this paper proposes an improved CCPM framework. assess schedule robustness. The slack of an operation, which is de-
The framework introduces an uncertainty-aware method (UAM) to fined as the time by which an operation can be delayed without
Downloaded from ascelibrary.org by University of Melbourne on 10/13/14. Copyright ASCE. For personal use only; all rights reserved.

determine reasonable buffer sizes by taking into account various slowing down performance of the entire schedule, is used as a
factors that affect construction scheduling but are not factored in buffering time to absorb the effects of disruptions (Al-Hinai and
by existing buffer sizing methods. The UAM is expected to outper- EIMekkawy 2011). Peng and Xu (2012) proposed an active critical
form existing buffer sizing methods by generating buffers with chain method. Unlike traditional methods, their method was de-
reasonable sizes and sufficient robustness. The framework also in- signed such that all activities, including those in noncritical chains,
tegrates a multimode-based method (MBM) that explores the fea- are executed in the earliest possible manner.
sibility of multiple resources leveling in CCPM-based construction The algorithms described offer feasible methods of buffer siz-
scheduling, which has thus far remained a significant challenge ing. However, there are limitations that have yet to be addressed.
with a lack of effective methods. The MBM performs multiple re- The buffer is relevant both to the duration of the activity, which is
sources leveling by managing the trade-offs between maximizing the main focus of these algorithms, and to managerial experience
resource utilization and minimizing project duration that is made and preference, project circumstances, past execution experience,
possible with the introduction of a multimodal activity execution and the capabilities of personnel and equipment, which have yet
structure. to be taken into consideration. To address this limitation, Shou and
Yao (2000) analyzed the different levels of uncertainty of activities
for different types of projects and management attitudes toward
Literature Review risk, and proposed an enhanced method for estimating the size
of project buffers. Tukel et al. (2006) introduced two feeding buffer
sizing methods that integrate project characteristics into the formu-
Buffer Sizing for Critical Chain Scheduling lation. One of these approaches incorporates resource tightness,
There are three types of buffers distinguished by their positions and whereas the other uses network complexity. Chu (2008) raised the
functions: the resource buffer (RB), the feeding buffer (FB), and the indicators of managers’ risk preferences. Bie et al. (2012) intro-
project buffer (PB). RBs are set to protect the critical chain from the duced a method for determining buffer sizes with dependence
scarcity of critical resources (Rand 2000). RBs are used as warn- assumption between activities assuming that only one risk factor
ings and do not consume any time. FBs are set to protect the critical is present in a given project.
chain from the variation of tasks not on the critical chain and
are scheduled where noncritical and critical activities converge
(Bevilacqua et al. 2009). PBs are placed at the end of the critical Multiple Resources Leveling
chain to protect against exceeding the project delivery date. The The majority of the prior research on resource leveling is based on
key to the successful implementation of critical chain scheduling is the CPM and PERT methods. Early research assumed a single re-
the proper determination of the size of the buffers. Two approaches source with an unlimited quantity, whereas later research began to
have been proposed in the previous literature: the C&PM and the hone its focus on multiple resources with limited quantities shared
RSEM. The C&PM estimates the safety time of each activity based by multiple projects. Crandall (1985) made one of the earliest ef-
on conventional methods, reduces the safety time of each activity forts to address the resource leveling problem, defining the basic
by 50%, and uses the sum of the safety times of all activities as a concept of resource leveling as a procedure that allocates available
buffer. The major advantage of this method is its simplicity. resources to activities that meet a set of project- and schedule-
However, the C&PM adopts a linear procedure, and the size of related criteria. When multiple activities meet the applicable criteria
the calculated buffer increases linearly with the length of the critical at a given time and compete for resources, Chang et al. (1990) pro-
chain. As a result, a schedule developed by the C&PM may contain posed weighting the criteria and prioritizing the activities using a
an unnecessarily large amount of protection, which may lead to fuzzy system and an expert system. Their approaches also allowed
wasted resources, uncompetitive proposals, and in turn lost project teams to manage weights and priorities and to develop a
business opportunities (Herroelen and Leus 2001). The RSEM variety of schedules based on changing project environments. Easa
produces two estimates for each activity in the critical chain, a safe (1989) proposed an integer-linear optimization model for resource
estimate and an average estimate, and calculates the difference leveling in construction projects that minimizes the absolute devia-
between the two estimates. The square root of the sum of squares tions between resource requirements and a uniform resource level
of this difference for each activity is then used as the buffer between consecutive resource requirements and/or between
(Newbold 1998). The RSEM is less influenced than the C&PM resource requirements and desirable nonuniform resource levels.
by the length of the critical chain. Tukel et al. (2006) reported that Leu et al. (1999) proposed another multiple resources leveling
compared with the C&PM, the RSEM has the distinct advantage of method for projects that involve activities with uncertain durations.
not generating very large or very small buffer sizes; this advantage Fuzzy theory is used to model the uncertainties of activity duration,
becomes clearer as the problem size increases (Herroelen and and a genetic-algorithms-based technique is used to search for the
Leus 2001). optimal resource leveling indexes under decision makers’ different

© ASCE 04014055-2 J. Constr. Eng. Manage.

J. Constr. Eng. Manage.


risk levels. Jun and El-Rayes (2011) proposed a multiobjective op- conducted to extend multiple resources leveling to CCPM. Because
timization model for measuring and minimizing undesired resource CCPM does not calculate free floats, which are the basis on which
fluctuations that consists of three modules: a startup module, a resource leveling is performed in CPM and PERT, resource leveling
scheduling module, and a multiobjective genetic algorithm module. with CCPM requires a fundamentally different approach that has
Case studies have demonstrated the model’s ability to manage the yet to be explored. Thus, this paper’s second contribution to the
trade-offs between maximizing resource utilization and minimizing literature is the MBM for performing multiple resources leveling
project duration. Genetic-algorithm-based resource leveling models in CCPM-based construction scheduling. The MBM is based on
were also proposed by Hegazy (1999) and Senouci and Eldin a multimodal structure for activity execution that allows manage-
(2004). Chen and Shahandashti (2009) argued that a genetic algo- ment of the trade-off between the type and amount of resources an
rithm could be integrated with a simulated annealing algorithm to activity requires and the time the activity takes to complete. The
achieve improved applicability. Damci et al. (2013) proposed that method first reduces the multiple resources leveling problem to
multiple resources leveling should be performed based on princi- a single resource leveling problem; then, it introduces a multimode
ples of optimum crew size that utilize a utility data curve and the execution structure to the project and identifies the optimal sched-
Downloaded from ascelibrary.org by University of Melbourne on 10/13/14. Copyright ASCE. For personal use only; all rights reserved.

principle of natural rhythm that allows the start times of an activity ule by iteratively updating the execution modes of individual
to be shifted forward or backward at different units of production activities and comparing the quality of the resulting schedules.
by changing the number of crews employed. The model yielded a The proposed CCPM framework is validated with three case
smoother resource utilization histogram while maintaining opti- studies from real-world settings. The results demonstrate the advan-
mum productivity. Unlike most models that do not permit activities tages of the proposed framework over traditional buffer sizing meth-
to be split, Son and Mattila (2004) proposed a linear program ods because it ensures on-time project completion while avoiding
binary variable model for resource leveling that adds to project unnecessarily large buffers. The results also demonstrate the fea-
teams’ flexibility in managing the resource balance by permitting sibility and effectiveness of balancing multiple resources in CCPM-
selected activities to stop and restart. based construction scheduling with the proposed framework.

Research Objectives and Contributions Uncertainty-Aware Buffer Sizing

By proposing the improved CCPM framework, this paper aims to This section introduces the uncertainties that influence schedule
achieve two specific objectives. First, the paper improves on current management, examined from five dimensions. These dimensions
buffer sizing practices in CCPM-based construction scheduling by include environment, network, activity, resources, and manage-
developing a sizing method that can generate buffers with reason- ment. The definition and metrics of these uncertainties are ex-
able sizes and sufficient robustness against various uncertainties. plained in detail subsequently. The effects of these uncertainties
Buffers generated by existing sizing methods are either unneces- on buffer sizes are then discussed, which leads to the introduction
sarily large, which results from overestimated uncertainties and of the UAM.
leads to wasted resources, or not large enough and insufficiently
robust against uncertainties. To generate higher quality buffers,
Environmental Uncertainty
it is crucial to factor in various distinctive project attributes, such
as managerial experience and preference, project circumstances, Environmental uncertainty (EU) refers to the uncertainty of exter-
experience with past execution, and personnel and equipment capa- nal project environments that can result from a wide variety of
bilities. Some of these attributes have been discussed in prior factors ranging from project-specific factors, such as contractual
research (Tukel et al. 2006); however, these discussions have in- risks and supply chain stability, to non-project-specific factors,
volved only a preliminary assessment of the impact of a limited such as natural disasters and macroeconomic conditions. To depict
number of attributes. There has been no comprehensive assessment the external environment under which a project is executed, the
of these attributes, and no method has been developed to reason- political, economic, social, and technological (PEST) framework
ably reflect their impact on the buffer sizing process. Thus, the first can be applied to help analyze the impact of external environmental
contribution this paper makes to the literature is a comprehensive factors. The PEST framework is a widely used tool for analyzing
assessment of the various factors that can introduce uncertainties the macroenvironment of organizations and evaluating the environ-
into construction scheduling, and the development of the UAM mental impacts (Yingfa and Hong 2010). Based on the PEST rat-
within the proposed CCPM framework to integrate impacts of such ings, the overall external environment can be categorized as one of
factors into the buffer sizing process. The factors are categorized three types—favorable, normal, and unfavorable—each of which is
and assessed in the following five dimensions: environment, net- assigned an EU value that illustrates the level of uncertainty to
work, activity, resources, and management. The applicable metrics which the project is exposed. As Eq. (1) shows, lower PEST ratings
for these factors are developed in this paper, and the magnitudes of indicate higher uncertainties and are, therefore, associated with
the impacts of these factors are reasonably measured and integrated larger EU values
into the buffer sizing process. 8 P
>
> 0.9 μi ≥ 3
Second, this paper explores the feasibility of multiple resources >
< i¼1 P
leveling in CCPM-based construction scheduling, with the MBM EU ¼ 1 1.5 ≤ μi ≤ 3 ð1Þ
developed within the proposed CCPM framework. Constraints >
> P i¼1
>
: 1.1 μi ≤ 1.5
imposed by the availability of various resources, including labor i¼1
and materials, constitute significant challenges to schedule man-
agement in construction projects, and it is critically important to where μi ði ¼ 1; 2; 3; 4Þ respectively represent the experts’ evalua-
develop and maintain resource-balanced schedules to ensure the tions of the political, economic, social, and technological environ-
on-time completion of such projects. However, despite the fact ments. Eq. (1) is the authors’ suggestion. Other means of measuring
that the resource leveling problem has been discussed extensively the EU can be discussed in future research. The values of EU
in conjunction with CPM and PERT, limited research has been (0.9, 1, 1.1) are arbitrary. These values can be modified based

© ASCE 04014055-3 J. Constr. Eng. Manage.

J. Constr. Eng. Manage.


X
m 
on managerial experience of project teams to manage the magni-
tude of the impact of EU variations on resulting buffer sizes. RTi ¼ max rkt =Rt ; t ∈ ½STi ; SFi  ð4Þ
k¼1

Activity Complexity where Rt = availability of a resource at time t; rkt = possible re-


source demand for resource type k at time t; m = number of activ-
Activity complexity (AC) denotes the complexity of a particular
ities executed at time t; and STi and SFi = start and finish times of
activity, which is determined by the complexity of the chain that
activity i, respectively. When the RT values are high, the resource
contains that activity (Tukel et al. 2006). When an activity has more
constraint becomes more obvious. Accordingly, larger buffers are
precedent activities in the chain, it is more likely to be influenced
required to prevent the occurrence of resource-related delays.
by delays and distortions that affect its precedent activities. Thus, a
larger buffer is desirable for this activity. The AC can be measured
as follows: Risk Preference

ACi ¼ N pre;i =N total;i ð2Þ In practice, given identical situations different project teams may
Downloaded from ascelibrary.org by University of Melbourne on 10/13/14. Copyright ASCE. For personal use only; all rights reserved.

yield different risk assessments due to their varied judgments of


where N pre;i = number of the predecessors of activity i; and N total;i = situations, managerial skills, and risk-bearing capacities. Thus,
number of activities in the chain that contains activity i. Let {CC} the project team’s impact on project implementation adds to the
be the critical chain activities set, let {NC} be the noncritical chain uncertainty of the project and should arguably be internalized in
activities set, and let fNCi g be the noncritical chain set that con- schedule management. This type of uncertainty is represented by
tains activity i. ACCC is the activity complexity of the critical chain risk preference (RP) in this paper. To measure RP, assume the
and is equal to the sum of the AC of all the activities in the critical project duration follows a normal distribution T ∼ Nðμ; σÞ based
chain. Similarly, ACNCi is the activity complexity of the noncritical on the central limit theorem (Herroelen and Leus 2001), and set
chain that contains activity i and is equal to the sum of the AC of all the standard risk level ε at 5%, which guarantees a project comple-
the activities in the chain. tion rate (1 − ε) of 95%. RP can then be calculated as follows:
RP ¼ ðzε − μÞ=2σ ð5Þ
Activity Flexibility
where 2σ = project buffer size under the RSEM; and zε = z-score in
Activity flexibility (AF) indicates the flexibility of a particular ac-
normal distribution.
tivity. Prior research has shown that activities with less AF are less
likely to be completed on time (Yang et al. 2009) and therefore
require larger buffers. Define a flexibility variable fi as f i ¼ Buffer Sizing Considering the Uncertainties
TFi ln TFi using entropy (Kumar 1987), where TFi refers to the In the remainder of this paper, the following project is used to
total float of activity i. The value of the AF of activity i can then discuss the UAM without losing generality. The project is depicted
be determined as follows: by an activity-on-node network N ¼ ðV; AÞ, where V denotes
8 the set of vertices and A denotes the set of arcs. Vertex set V ¼
< 1.1 0 ≤ f i ≤ f 0
f0; 1; : : : ; n; n þ 1g consists of n activities 1; 2; : : : ; n that must
AFi ¼ 1 f 0 ≤ fi ≤ f 0 0 ð3Þ
: be executed without interruption and two dummy start and finish
0.9 f 0 0 ≤ fi
activities, i ¼ 0 and i ¼ n þ 1. Each activity must not be initiated
where f 0 and f 0 0 = thresholds that are determined based on the prior to the scheduled start time and must be completed by project
project teams’ analyses and judgments of the project environment. termination. Let {CC} be the critical chain activities set, let {NC}
When f i is below threshold f 0 , activity i has a small total float, and be the noncritical chain activities set, and let fNCi g be the non-
thus low flexibility; therefore, it imposes higher risk on the on-time critical chain set that contains activity i. Assume that the duration
completion of successive activities. In contrast, when f i is above of activity i has a normal distribution, which is denoted by
threshold f 0 0 , activity i has a large total float, and thus high flex- di ∼ Nðμi ; σi Þ, and that σi stands for the safety time of activity
ibility, which is less likely to delay successive activities. An activity i. The RSEM calculates the PB and the FB as follows (Herroelen
with f i between the two thresholds has a reasonable flexibility that and Leus 2001):
does not require particular managerial attention. Eq. (3) is the !1=2
Xn
2
authors’ suggestion. Other means of measuring the AF can be PB ¼ σi ð6Þ
discussed in future research. The values of AF (1.1, 1, 0.9) in i∈fCCg
Eq. (3) are arbitrary. These values can be modified based on mana- !1=2
gerial experience of project teams to manage the magnitude of the X
n
impact of AF variations on resulting buffer sizes. FBi ¼ σ2j ð7Þ
j∈fNCi g

Resource Tightness The UAM is built on the RSEM but considers the influences of
Implementing a project may require various resources that are typ- various uncertainties that are not addressed by the RSEM. The
ically shared by multiple activities. A particular activity must com- UAM integrates chain-specific uncertainties that have global effects
pete with other activities for required resources, and is therefore on the project span, including EU, RP, and AC, and activity-
subject to resource scarcity and delays. Resource scarcity can result specific uncertainties that impact individual activities alone, includ-
from a range of factors, such as insufficient supplies, poor inventory ing AF and RT. The buffer sizes are calculated as follows:
management, and damage. Resource tightness (RT) is introduced to ( )1=2
represent uncertainties associated with resource availability and X n
PB ¼ EU × RP × ACCC × ½ð1 þ RTi Þ × AFi × σi  2
measures how accessible a resource is to a particular activity, given
i∈fCCg
the concurrent demands for this resource by other activities. The RT
value of activity i is measured as follows: ð8Þ

© ASCE 04014055-4 J. Constr. Eng. Manage.

J. Constr. Eng. Manage.


( )1=2
X
n requirement of activity i, which is denoted as ri , can be calcu-
FBi ¼ EU × RP × ACNCi × ½ð1 þ RTi Þ × AFi × σi 2 lated as
i∈fNCi g

ð9Þ X
K
ri ¼ ðrik × ωk Þ; i∈V ð11Þ
k¼1
The PB that is calculated based on Eq. (8) will be shifted to the
end of the critical chain, whereas the FB that is calculated based on where the weight, ωk , P is determined based on the importance of the
Eq. (9) will be placed at the intersections at which a noncritical resource, 0 ≤ ωk ≤ 1, Kk¼1 ωk ¼ 1. The introduction of ri converts
chain joins the critical chain. Both the PB and the FB are used the multiple resources leveling problem into a single resource
as safety times to mitigate the adverse effects of uncertainties on leveling problem.
the project. Eqs. (8) and (9) internalize the effects of the uncertain- Resource leveling is frequently performed to reduce the fluc-
ties on project schedules, and the resulting buffers and associated tuation of resource usage over the project life span. However, this
project schedules are expected to yield better performance against objective cannot always be achieved concurrently with another
Downloaded from ascelibrary.org by University of Melbourne on 10/13/14. Copyright ASCE. For personal use only; all rights reserved.

various uncertainties than schedules developed using traditional critical objective, i.e., reducing the deviation between planned daily
buffer sizing methods. It must be emphasized that buffer sizes de- usage of resources and available resources. When a project team
veloped using the proposed method are not necessarily longer than can predict the availability of resources over the life span of a
those developed with traditional methods. When a project is per- project, it is critical for the project team to establish the resource
formed under little uncertainty, where the relevant coefficients, profile of the project schedule in such a way that the planned
such as the EU and the RP, may be lower than 1, the proposed resources best match the resource availability. Therefore, multiple
method reports shorter buffers, which is valuable for developing resources leveling is performed in this paper by minimizing the
competitive proposals and securing business opportunities. When deviations between planned usage of resources and available re-
the uncertainties are significant, the reported buffers render the pro- sources. Accordingly, an indicator M is introduced as follows to
posals less appealing. However, the schedules have sufficient built- evaluate the quality of resource profiles. The M indicator is the
in robustness against various foreseen uncertainties, which reduces sum of the absolute values of deviations between the actual usage
the risk of project overruns and helps avoid associated losses. of resources and the available amount of resources over the life
span of a project

Multiple Resources Leveling in Critical Chain X X


D
Scheduling M¼ Ck jZkt − Rk j ð12Þ
k∈R t¼0
In CPM- and PERT-based schedule management, resource leveling
is substantially based on the amount of available free floats.
Resource leveling is undertaken by adjusting the start time of non- Multimodal Structure for Activity Execution
critical activities to an extent allowed by the amount of their free This paper introduces a multimodal structure for activity execution
floats for purposes of balancing resource usage. However, CCPM is to perform resource leveling with schedules developed under
fundamentally different from CPM and PERT methods and does CCPM. Under this structure, activities in the activity-on-node net-
not involve calculating free floats for each activity. Therefore, work N ¼ ðV; AÞ can be performed in one of the following three
resource leveling with schedules developed by CCPM requires a modes: urgent mode, normal mode, and deferred mode. For each
different approach that has not yet been fully studied. With the in- activity, its three modes are associated with respective combina-
creased popularity of CCPM resulting from its proven effectiveness tions of multiple resources that are allocated to support the activity,
in schedule management, this paper examines the method neces- and the activity is initially assigned the normal mode. The possibil-
sary for performing multiple resources leveling with schedules ity of switching among different modes is a better reflection of
developed using the critical chain method and the UAM. real-world scenarios, in which certain activities are assigned higher
priorities, and these assignments are subject to constant change.
Conversion from Multiple Resources Leveling to Single More importantly, the multimodal structure allows for resource
Resource Leveling usage rates and the duration of a particular activity to be adjusted,
which provides project management teams with a promising tool
The multiple resources leveling problem can be converted into a for schedule management and resource balancing.
single resource leveling problem with a normalization method Define the following variable
(Rothe et al. 1996). In the activity-on-node network N ¼ ðV; AÞ,

assume that resources 1; 2; : : : ; K are required for executing the 1; if activity i is performed in mode m
project. Let Ck ≥ 0 be the cost incurred per unit of resource k xim ¼ ð13Þ
0; otherwise
per time unit, let Rk > 0 be the capacity of resource k available,
and let rik ≥ 0 be the amount of resource k used by activity i where the urgent, normal, and deferred modes are denoted by
ðk ¼ 1; 2; : : : ; K; i ∈ VÞ, where 0 ≤ rik ≤ Rk and r0k ¼ rnþ1;k ¼ m1 ; m2 ; m3 , respectively.
0. For the sake of simplicity, Rk and rik are assumed to be constant.
Let Zkt be the usage of resources k at time t within a certain period.
It is assumed that a variety of resources can be used for each activity. Multimode-Based Multiple Resources Leveling
Define CCPM relies on managing the trade-off between the resources an
r activity requires and the time it takes to complete the activity. This
rik ¼ ik ; i∈V ð10Þ trade-off is undertaken by switching the execution mode of an ac-
Rk
tivity from normal to either urgent or deferred, which increases the
Thus, the range of the need for each resource is between 0 and 1. amount of required resources and reduces the activity duration or
With the normalization method, the weighted multiresource reduces the amount of required resources and increases the activity

© ASCE 04014055-5 J. Constr. Eng. Manage.

J. Constr. Eng. Manage.


duration, respectively. The proposed multimode-based resource project done in the same city. Case 3 was developed by including a
level framework can be undertaken in the following steps: subset of activities with reduced complexity of precedence relation-
1. Select the normal execution mode for each activity, generate ships and the same types of resources from Case 2. The general
the baseline schedule S, and output the resource profile. information of these cases is summarized in Table 1. The rest of
2. Identify the critical chain and noncritical chains, and identify this section provides further details of project settings and CCPM
positions for inserting the PB and the FB. Assess the uncer- framework implementation for Case 1 only; however, results from
tainties and calculate the size of the PB and the FB based on implementing the framework in all of the three cases are reported
Eqs. (8) and (9). and compared in the next section.
3. Reduce the multiple resources leveling problem to a single The precedence relationships between activities in Case 1 are
resource leveling problem using the normalization method. shown in Fig. 1. Case 1 involves three types of resources, i.e., per-
4. Select the time section with the largest resource usage devia- sonnel, material, and machinery, with limited daily availability of
tion and adjust the mode of activities within the time section. R ¼ ð48; 53; 42Þ. The parameters of each activity, including its
Resource usage deviation refers to the deviation between durations, the standard deviation of its duration, and its require-
Downloaded from ascelibrary.org by University of Melbourne on 10/13/14. Copyright ASCE. For personal use only; all rights reserved.

planned weighted resource usages of the period and the aver- ments for each type of resources, are summarized in Table 2.
age weighted resource usages over the entire project life span.
The adjustment of execution mode follows the following rule:
if the resources are underused (planned usages below the aver- Generate Base Schedules
age usages over the entire project life span), change the mode Based on the durations of the activities and their precedence rela-
from deferred to normal or urgent, or from normal to urgent; if tionships, a baseline schedule was developed for Case 1. The base
the resources are overused (planned usages over the average schedule included a critical path ABCDEHIJLNTU with a duration
usages over the entire project lifespan), change the mode from of 350 days, and four noncritical paths, F, KP, RS (QM), and GO.
urgent to normal or deferred, or from normal to deferred. Then, based on the resource requirements of the activities and
The resource usages cannot exceed the maximum amount of the availability of all resources, a resource profile of the baseline
available resources after the adjustments. schedule for Case 1 was developed, as shown in Fig. 2.
5. Repeat Step 4 until the M indicator is minimized. Then insert
the buffers and output the resource profile.
Determine Buffer Sizes

Case Studies The next step was to determine the buffer size using the UAM.
Based on the identified critical chain and noncritical chains, it
Three case studies from real-world settings are used in this paper to was determined that FB should be inserted after activities F, P,
validate the proposed CCPM framework. This section provides S, and O, and the PB should be inserted after activity U. The fol-
details of the cases and presents steps taken to implement the lowing arbitrary parameter values were used when sizing the FB
framework in these cases. and the PB: EU ¼ 1, risk preference ε ¼ 5%, RP ¼ 0.8225, and
the AC of the critical chain ABCDEHIJLNTU and noncritical
chains F, KP, RS (QM), and GO were 0.92, 0, 0.5, 1 and 0.5, re-
Project Description
spectively. With parameter values f 0 ¼ 5 and f 0 0 ¼ 10, the AF of
Case 1 was developed based on an infrastructure project done in a activities A, B, C, D, E, H, I, J, L, N, T, U is 1.1, and the AF of
midsized city in China. Case 2 was developed based on an airport activities F, K, P, Q, R, S, M, G, O is 0.9. The RT values could be

Table 1. General Information of Three Case Studies


Case study Baseline project Number of Types of Daily availability of
number Type of project duration activities resources resources
1 Tramway and associated 350 days 21 Personnel, material, 48, 53, 42
engineering systems and machinery
2 Airport and associated infrastructure 64 months 24 Personnel, material, 100, 30, 15
and machinery
3 Airport and associated infrastructure 50.5 months 17 Personnel, material, 100, 26, 12
and machinery

Fig. 1. Precedence relationships in Case 1

© ASCE 04014055-6 J. Constr. Eng. Manage.

J. Constr. Eng. Manage.


Table 2. Activity Parameters in Case 1 calculated based on Eq. (4) and the activity resource requirements
Resource Standard in Table 2.
Act Duration (day) requirements deviation Based on the previous parameter values, the PB and the FB were
determined as follows:
A 5 38, 18, 5 0.70
7 35, 16, 3 PB ¼ 19; FBF ¼ 0; FBP ¼ 2; FBS ¼ 9; FBO ¼ 10 ð14Þ
8 32, 14, 2
B 50 37, 15, 6 3.36
56 35, 12, 4
65 32, 8, 3 Balance Multiple Resources
C 35 28, 15, 4 4.32 The multiple resources leveling was then carried out with the
40 24, 13, 2 MBM. After the problem was reduce to a single resource leveling
45 22, 10, 1 problem using the normalization method, a resource balanced
D 8 25, 18, 8 0.92 schedule was developed by minimizing the value of the M indica-
Downloaded from ascelibrary.org by University of Melbourne on 10/13/14. Copyright ASCE. For personal use only; all rights reserved.

12 20, 16, 5
tor. Starting with an M indicator of 0.240, the procedure is
16 18, 14, 3
explained in detail in the following steps:
E 30 25, 12, 23 5.58
1. Search for a period with the largest resource usage deviation.
45 20, 8, 20
50 15, 5, 18
This period was identified to be days 339–350, and the re-
F 170 13, 13, 12 15.3
sources were underused during this period. The execution
180 9, 10, 9 mode of activities T and U were therefore changed from nor-
190 7, 8, 7 mal to urgent, which increased the resource usages of these
G 145 14, 15, 10 9.40 activities, shortened their durations, and reduced the M indi-
150 10, 12, 8 cator to 0.235.
156 8, 10, 8 2. After the adjustment in the last step, days 328–338 (when ac-
H 38 30, 11, 21 5.13 tivity M was carried out) became the period with the largest
45 27, 11, 21 resource usage deviation. The resources were underused dur-
50 25, 10, 20 ing this period. However, changing activity M from normal to
I 35 13, 16, 9 6.17 urgent would result in a resource demand that exceeded the
45 10, 16, 8 maximum amount of available resources. Therefore, no adjust-
55 9, 16, 8 ment was made in this step.
J 18 12, 17, 9 1.52 3. The period of days 291–295 had the second largest resource
20 10, 16, 8 usage deviation, and the resources were overused. The execu-
23 10, 16, 8 tion mode of activities M, O, P, and Q were therefore changed
K 10 3, 6, 3 1.82 from normal to deferred, which reduced the resource usages
15 1, 4, 3 of these activities, increased their durations, and reduced
20 1, 3, 3 the M indicator to 0.202.
L 15 9, 10, 3 0.84 4. After the adjustment in the last step, days 161–205 became
20 7, 8, 3
the period with the largest resource usage deviation. The re-
25 7, 6, 3
sources were overused during this period. The execution
M 8 11, 4, 3 1.17
10 2, 2, 3
mode of activities F and G were therefore changed from
13 5, 2, 3
normal to deferred, which reduced the resource usages of these
N 45 10, 6, 3 3.70
activities, increased their durations, and reduced the M indi-
48 7, 6, 3 cator to 0.177. The mode of activity H was not changed to
50 5, 5, 3 urgent because it would otherwise increase the M indicator
O 80 18, 18, 17 8.19 to 0.179.
90 16, 17, 17 5. Similar to Step 4, activities A and B were changed from nor-
95 14, 14, 17 mal to urgent, reducing the M indicator to 0.157; activities C
P 35 8, 8, 3 4.16 and D were then adjusted, reducing the M indicator to 0.139.
38 5, 6, 3 6. After the previous steps, no further adjustment could be made
42 3, 5, 3 to reduce the M indicator. The PBs and FBs were inserted to
Q 20 5, 6, 3 2.65 the adjusted schedule and a final schedule was developed. The
25 3, 5, 3 M indicator slightly increased to 0.159 after the buffers were
30 1, 4, 3 inserted.
R 10 14, 5, 0 1.30 Resource profiles after adjustments in Steps 1 and 3 are shown
12 12, 2, 0 in Figs. 3 and 4, respectively, to illustrate the resource leveling pro-
15 9, 2, 0 cess. The resource profile of the final schedule is shown in Fig. 5.
S 20 6, 4, 0 5.45 The previous multiple resources leveling procedure was also per-
25 4, 2, 0 formed using MBM in Cases 2 and 3 to develop resource balanced
30 2, 2, 0 schedules in both cases.
T 5 16, 4, 0 1.28
7 12, 3, 0
10 10, 3, 0
Findings
U 3 15, 0, 0 1.02
5 12, 0, 0
The previous implementation of the CCPM framework is evaluated
8 9, 0, 0
in this section. First, the effectiveness of the UAM in terms of the

© ASCE 04014055-7 J. Constr. Eng. Manage.

J. Constr. Eng. Manage.


Fig. 2. Resource profile of the baseline schedule in Case 1
Downloaded from ascelibrary.org by University of Melbourne on 10/13/14. Copyright ASCE. For personal use only; all rights reserved.

Fig. 3. Resource profile after adjustment in Step 1

Fig. 4. Resource profile after adjustment in Step 3

Fig. 5. Resource profile of the final schedule

length and robustness of the buffers is examined based on actual Monte Carlo Simulation of Actual Project Durations
project durations obtained from Monte Carlo simulations. More-
The Monte Carlo method, which is commonly used in the risk
over, the feasibility of performing multiple resources leveling using
analysis of scheduling problems (Usábel 1998; Schuhmacher et al.
the MBM has been proven in the last section. This section provides
further evaluation of the MBM in terms of its effectiveness in 2001; Pozzi 2003), was used to simulate the actual durations of the
reducing resources usage deviation and project duration, and usage project and to analyze the risks of schedule overruns to evaluate the
fluctuation of individual resources. The results are reported and dis- performance of the UAM. The Monte Carlo simulation in this paper
cussed subsequently. was undertaken using the following steps:

© ASCE 04014055-8 J. Constr. Eng. Manage.

J. Constr. Eng. Manage.


Table 4. Changes in the Schedules after Multiple Resources Leveling
Resource Resource usage fluctuation (%)
Case usage deviation Project
study (M indicator) duration
number (%) (%) Resource 1 Resource 2 Resource 3
1 −23.52 −5.98 −27.03 −17.39 −21.00
2 −1.66 −6.25 −36.68 −16.56 −26.94
3 −10.83 −1.99 −13.46 0.33 −24.97

differences indicated a substantial waste of buffers by the


Fig. 6. Distribution graph of the project duration C&PM, which may further increase with project size, considering
the linear procedure of the C&PM that simply calculates the buffer
Downloaded from ascelibrary.org by University of Melbourne on 10/13/14. Copyright ASCE. For personal use only; all rights reserved.

size by adding together the buffers of all activities on the critical


1. Determine the distribution function for each activity of the chain. The buffers developed using the RSEM had the smallest
construction project. The distribution of the activity can be buffer size in all cases. However, the results showed that the proj-
a normal distribution, a triangular distribution, or a beta ects with schedules containing these buffers were the least likely to
distribution. be completed on time. The average on-time completion rate re-
2. Determine the number of repetitions of the simulation. ported by the RSEM was 95.9%, which was 4.0 and 0.4% lower
3. Calculate the three time estimations of the duration of each than the on-time completion rates reported by the C&PM and the
activity—including the optimistic duration, pessimistic dura- UAM, respectively. In Case 3 in particular, the on-time completion
tion, and most likely duration—and determine the distribution rate reported by the RSEM was noticeably lower than those re-
of the duration. ported by the other two methods and close to failing the guaranteed
4. For each activity, generate a random sampling of its actual 95% threshold. In general, it could be concluded that the UAM
duration based on the distribution of its duration. outperformed the C&PM and the RSEM by generating buffers
5. Calculate the actual project duration based on the sampling that not only had reasonable sizes to avoid waste but also provided
durations of the activities and their logical relationships. sufficient protection and robustness to the schedules to ensure a
6. Repeat Steps 3 and 4 until all of the repetitions are completed. guaranteed on-time project completion rate.
In this paper, the durations of all activities were assumed to fol-
low a normal distribution and the simulation was repeated 1,000
times. The results indicated that the distribution of the actual Evaluation of the MBM in Case Studies
project durations in the simulation generally followed a normal dis- Multiple resources leveling in CCPM-based construction schedul-
tribution. Take Case 1 as an example. As illustrated in Fig. 6, the ing has thus far remained a significant challenge with a lack of
95th percentile line in the figure has a duration of 347.8 days. It effective methods. The MBM is developed within the proposed
suggests that the schedule in Case 1 developed using the UAM, CCPM framework to address this challenge, and its feasibility
with a duration of 348 days, ensured a 95.6% probability of on-time has been proven by the implementation of the framework demon-
project completion, which exceeded the guaranteed 95% threshold strated in the last section. This section provides further evaluation
and therefore could satisfy managerial requirements. Similar results of the MBM by measuring its impact on the overall resource usage
were observed from Cases 2 and 3. Their schedules developed deviation, the project duration, and the usage fluctuation of indi-
using the UAM ensured 97.1 and 95.1% probabilities of on-time vidual resources. The results are summarized in Table 4.
project completion, respectively. As can be seen in Table 4, after multiple resources leveling
using the MBM, the M indicator was reduced by 12.00% on aver-
age. The magnitude of the reduction varied between the three cases,
Evaluation of the UAM in Case Studies
suggesting that the extent to which a baseline schedule can be opti-
Table 3 shows a performance comparison of the UAM with the mized relies on the stringency of resource constraints of a particular
C&PM and RSEM. The projects with critical chain-based sched- project. More stringent resource constraints, like those seen in
ules developed using C&PM were the most likely to be completed Case 2, are likely to leave less room for schedule optimization.
on time in all three cases, with an average on-time completion rate Moreover, the multiple resources leveling has an impact on the
of 99.8% in the simulation. Buffers generated using the C&PM had project duration because the duration of an individual activity
respective total sizes of 175 days, 28.5 months, and 21.5 months in changes with its execution mode. As can be seen in Table 4, the
the three cases. These sizes were approximately five times the durations of all projects were reduced, with an average reduction
buffer sizes reported by the other two methods in Case 1, three of 4.74%. The results suggested that resources were more effi-
times in Case 2, and four times in Case 3. Such significant ciently used in the balanced schedules so that some activities could

Table 3. Comparison of the UAM with the C&PM and the RSEM
Case 1 Case 2 Case 3
Buffer sizing Total buffer Probability of on-time Total buffer Probability of on-time Total buffer Probability of on-time
method size (day) completion (%) size (30 days) completion (%) size (30 days) completion (%)
C&PM 175 99.5 28.5 100 21.5 100
RSEM 36 96.1 8 96.3 5 95.2
UAM 37 96.2 10 96.6 6 96.1

© ASCE 04014055-9 J. Constr. Eng. Manage.

J. Constr. Eng. Manage.


Fig. 7. Individual resource histograms before resource leveling in Case 1: (a) Resource 1; (b) Resource 2; (c) Resource 3
Downloaded from ascelibrary.org by University of Melbourne on 10/13/14. Copyright ASCE. For personal use only; all rights reserved.

Fig. 8. Individual resource histograms after resource leveling in Case 1: (a) Resource 1; (b) Resource 2; (c) Resource 3

be executed in an urgent mode with more intensive resource usages, completion. The case study results also demonstrated the feasibility
leading to reduced project durations. of performing multiple resources leveling within the proposed
The impact of multiple resources leveling on the usage of indi- framework. The MBM was proven to be an effective multiple re-
vidual resources was also examined. The histograms of daily usage sources leveling method, which was capable of concurrently reduc-
of each individual resource before and after the resource leveling in ing the resource usage deviation, the project duration, and the usage
Case 1 are shown in Figs. 7 and 8, respectively. The results showed fluctuation of individual resources in most cases.
that the sum of the absolute deviations between daily resource The chances for the proposed CCPM framework to be appli-
usage and average resource usages, which is a widely used metric cable to and accepted by the construction industry are promising
for resource usage fluctuation, was reduced by 27.03, 17.39, and because of several reasons. First, the framework is based on more
21.00% after the resource leveling for Resources 1, 2, and 3, re- realistic assumptions than existing methods. For example, the
spectively. Similar results were observed in Cases 2 and 3, except framework recognizes that an activity may be executed in different
for Resource 2 in Case 3, whose fluctuation slightly increased by modes with different resource demands, which is a better descrip-
0.33% after the resource leveling. The results provide evidence for tion of jobsite and managerial practices and allows for more flex-
the assumption that the two objectives, namely, reducing the devi- ible adjustments of schedules when project environments change.
ations between planned daily resource usages and available resour- Second, the framework integrates the impact of network complex-
ces on the one hand, and reducing the fluctuation of resource ity and resource tightness into the buffer sizing process; therefore,
usages over the project lifespan on the other, are not always con- the framework can be applied to construction projects with varying
sistent with one another. In general, the results showed that MBM is complexities and resource demands. Third, the framework takes
an effective method for performing multiple resources leveling into account not only objective constraints in project environments
within the proposed CCPM framework. but also subjective judgments by project teams, and hence provides
a comprehensive assessment of various factors that are impactful on
project schedules in practice.
Discussions and Conclusions The proposed CCPM framework has substantial potential to en-
hance the implementation of the CCPM in construction project
CCPM is an emerging scheduling method that relies on buffers for management practices. In the past decade, CCPM has obtained
protecting schedules from overruns and requires resource leveling noticeable significance as an emerging construction scheduling
to develop feasible and competitive schedules. This paper proposes method. It is considered advantageous over traditional scheduling
an improved framework for CCPM-based construction scheduling methods mainly because of its capability of mitigating impacts of
to address two challenges, including buffer sizing and multiple re- uncertainties by protecting deterministic schedules with buffers and
sources leveling. The results from three case studies showed that its emphasis on the integration of activity durations, precedence
the UAM outperformed existing buffer sizing methods because relations, and resource requirements and availabilities in the sched-
buffers generated by the UAM had not only smaller sizes than uling process. The proposed CCPM framework is expected to fur-
buffers generated by the C&PM, leading to less waste of resources ther increase the adoption of the CCPM in construction scheduling
and improved competitiveness of schedules, but also more robust- by providing justifiable and operable instructions on buffer sizing
ness against uncertainties compared with buffers generated by and multiple resources leveling so that schedules are given a more
the RSEM, leading to higher possibilities of on-time project reasonable amount of protection and resources are planned and

© ASCE 04014055-10 J. Constr. Eng. Manage.

J. Constr. Eng. Manage.


utilized more efficiently. The CCPM has not been as widely Damci, A., Arditi, D., and Polat, G. (2013). “Multiresource leveling in line-
adopted as traditional construction scheduling methods and the of-balance scheduling.” J. Constr. Eng. Manage., 10.1061/(ASCE)CO
commercialization of the CCPM is still in its infancy. Some .1943-7862.0000716, 1108–1116.
may even argue that its establishment poses an unnecessary choice Easa, S. (1989). “Resource leveling in construction by optimization.”
between CCPM and mainstream practices (Raz et al. 2003). How- J. Constr. Eng. Manage., 10.1061/(ASCE)0733-9364(1989)115:2(302),
302–316.
ever, as its methodological development advances, the CCPM will
Goldratt, E. M. (1997). Critical chain, North River Press, Great Barrington,
eventually evolve into a mainstream method that complements
MA.
existing methods for the schedule management of construction Hall, N. G. (2012). “Project management: Recent developments and
projects challenged by ever increasing complexities. research opportunities.” J. Syst. Sci. Syst. Eng., 21(2), 129–143.
To further advance research in the area of schedule management Hazir, O., Haouari, M., and Erel, E. (2010). “Robust scheduling and robust-
and to improve the effectiveness of the proposed framework, future ness measuresfor the discrete time/cost trade-off problem.” Eur. J.
research should be carried out focusing on increasing the resolu- Oper. Res., 207(2), 633–643.
tions of the uncertainty metrics to better assess the magnitude of Hegazy, T. (1999). “Optimization of resource allocation and leveling using
Downloaded from ascelibrary.org by University of Melbourne on 10/13/14. Copyright ASCE. For personal use only; all rights reserved.

the effects of various uncertainties and increasing the number of genetic algorithms.” J. Constr. Eng. Manage., 10.1061/(ASCE)0733-
activity execution modes to better reflect how activities are per- 9364(1999)125:3(167), 167–175.
formed in practice and to enrich the options that project teams have Jun, D. H., and El-Rayes, K. (2011). “Multiobjective optimization
in adjusting resource profiles. Moreover, further research should of resource leveling and allocation during construction scheduling.”
examine the dynamic updating of buffer sizes during project im- J. Constr. Eng. Manage., 10.1061/(ASCE)CO.1943-7862.0000368,
plementation such that project teams can be prepared when risks 1080–1088.
Herroelen, W., and Leus, R. (2001). “On the merits and pitfalls of critical
change due to changing project environments and ongoing work
chain scheduling.” J. Oper. Manage., 19(5), 559–577.
progress. Moreover, future research can be performed to examine
Kastor, A., and Sirakoulis, K. (2009). “The effectiveness of resource
the potential of multiobjective optimization that considers not only levelling tools for resource constraint project scheduling problem.”
project duration but also other objectives, such as cost control and Int. J. Proj. Manage., 27(5), 493–500.
logistics management, to achieve informed and comprehensive Kolisch, R., and Drexl, A. (1997). “Local search for nonpreemptive multi-
project management. mode resource-constrained project scheduling.” IIE Trans., 29(11),
987–999.
Kumar, V. (1987). “Entropic measures of manufacturing flexibility.” Int. J.
Acknowledgments Prod. Res., 25(7), 957–966.
Leu, S. S., Chen, A. T., and Yang, C. H. (1999). “A fuzzy optimal model for
This work is supported by the National Natural Science Foundation construction resource leveling scheduling.” Can. J. Civ. Eng., 26(6),
of China (NNSFC) under Grants #70802045 and #71102142. The 673–684.
authors thank the NNSFC for its support. Any opinions, findings, Long, L. D., and Ohsato, A. (2008). “Fuzzy critical chain method for
conclusions, or recommendations expressed in this material are project scheduling under resource constraints and uncertainty.” Int. J.
Proj. Manage., 26(6), 688–698.
those of the authors and do not necessarily reflect the views of
Newbold, R. C. (1998). Project management in the fast lane: Applying the
the NNSFC. theory of constraints, Saint Lucie Press, Boca Raton, FL.
Peng, W. L., and Xu, H. (2012). “The scheduling problem of active critical
chain method.” Inf. Technol. J., 11(7), 829–839.
References Pozzi, L. (2003). “The coefficient of relative risk aversion: a Monte Carlo
study investigating small sample estimator problems.” Econ. Modell.,
Ahlemann, F., ElArbi, F., Kaiser, M. G., and Heck, A. (2013). “A process 20(5), 923–940.
framework for theoretically grounded prescriptive research in the Rand, G. K. (2000). “Critical chain: The theory of constraints applied to
project management field.” Int. J. Proj. Manage., 31(1), 43–56. project management.” Int. J. Proj. Manage., 18(3), 173–177.
Al-Hinai, N., and ElMekkawy, T. Y. (2011). “Robust and stable flexible job Raz, T., Barnes, R., and Dvir, D. (2003). “A critical look at critical chain
shop schedulingwith random machine breakdowns using hybrid genetic project management.” Proj. Manage. J., 34(4), 24–32.
algorithm.” Int. J. Prod. Econ., 132(2), 279–291. Rothe, I., Susse, H., and Voss, K. (1996). “The method of normalization to
Bevilacqua, M., Ciarapica, F. E., and Giacchetta, G. (2009). “Critical chain determine invariants.” IEEE Trans. Pattern Anal. Mach. Intell., 18(4),
and risk analysis applied to high-risk industry maintenance: A case 366–376.
study.” Int. J. Proj. Manage., 27(4), 419–432. Rozenes, S., Vitner, G., and Spraggett, S. (2006). “Project control:
Bie, L., Cui, N., and Zhang, X. (2012). “Buffer sizing approach with Literature review.” Proj. Manage. J., 37(4), 5–14.
dependence assumption between activities in critical chain scheduling.” Schuhmacher, M., Meneses, M., Xifró, A., and Domingo, J. L. (2001).
Int. J. Prod. Res., 50(24), 7343–7356. “The use of Monte-Carlo simulation techniques for risk assessment:
Chang, T., Ibbs, C., and Crandall, K. (1990). “Network resource allocation Study of a municipal waste incinerator.” Chemosphere, 43(4–7),
with support of a fuzzy expert system.” J. Constr. Eng. Manage., 787–799.
10.1061/(ASCE)0733-9364(1990)116:2(239), 239–260. Senouci, A., and Eldin, N. (2004). “Use of genetic algorithms in resource
Chen, P. H., and Shahandashti, S. M. (2009). “Hybrid of genetic algorithm scheduling of construction projects.” J. Constr. Eng. Manage., 10.1061/
and simulated annealing for multiple project scheduling with multiple (ASCE)0733-9364(2004)130:6(869), 869–877.
resource constraints.” Autom. Constr., 18(4), 434–443. Shou, Y. Y., and Yao, K. T. (2000). “Estimation of project buffers in critical
Chu, C. C. (2008). “Buffer sizing and critical chain project management.” chain project management.” Proc., 2000 IEEE Int. Conf. on Manage-
Comput. Integr. Manuf. Syst., 18(5), 1029–1035 (in Chinese). ment of Innovation and Technology, Vol. 1, Institution of Engineers,
Cohen, I., Mandelbaum, A., and Shtub, A. (2004). “Multi-project Singapore, 162–167.
scheduling and control: A process-based comparative study of the criti- Son, J., and Mattila, K. (2004). “Binary resource leveling model: Activity
cal chain methodology and some alternatives.” Proj. Manage. J., 35(2), splitting allowed.” J. Constr. Eng. Manage., 10.1061/(ASCE)0733-
39–50. 9364(2004)130:6(887), 887–894.
Crandall, K. C. (1985). Resource allocation with project manager control, Trietsch, D., and Baker, K. R. (2012). “PERT 21: Fitting PERT/CPM for
ASCE, Reston, VA, 1–17. use in the 21st century.” Int. J. Proj. Manage., 30(4), 490–502.

© ASCE 04014055-11 J. Constr. Eng. Manage.

J. Constr. Eng. Manage.


Tukel, O. I., Rom, W. O., and Eksioglu, S. D. (2006). “An investigation of Yingfa, S., and Hong, Y. (2010). “The risk study of e-governance based
buffer sizing techniques in critical chain scheduling.” Eur. J. Oper. Res., on PEST analysis model.” Proc., Int. Conf. on E-Business and
172(2), 401–416. E-Government (ICEE), The Korean Institute of Electrical Engineers
Usábel, M. A. (1998). “Applications to risk theory of a Monte Carlo (KIEE), Seoul, Korea, 563–566.
multiple integration method.” Insur. Math. Econ., 23(1), 71–83. Zhao, Z. Y., You, W. Y., and Zuo, J. (2010). “Application of innovative
Yang, L., Li, S., and Huang, X. (2009). “A buffer sizing approach in critical critical chain method for project planning and control under resource
chain scheduling with attributes dependent.” Ind. Eng. Manage., 1, constraints and uncertainty.” J. Constr. Eng. Manage., 10.1061/
11–14 (in Chinese). (ASCE)CO.1943-7862.0000209, 1056–1060.
Downloaded from ascelibrary.org by University of Melbourne on 10/13/14. Copyright ASCE. For personal use only; all rights reserved.

© ASCE 04014055-12 J. Constr. Eng. Manage.

J. Constr. Eng. Manage.

Potrebbero piacerti anche