(Asce) Co 1943-7862 0000664 PDF

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

Risk-Based Protocol for Inspection of Transportation

Construction Projects Undertaken by State


Departments of Transportation
Downloaded from ascelibrary.org by CEPT - Centre for Environmental Planning & Tech University on 11/07/19. Copyright ASCE. For personal use only; all rights reserved.

Ali Mostafavi 1; Dulcy Abraham 2; Samy Noureldin 3; Greg Pankow 4; Joseph Novak 5;
Ronald Walker 6; Kevin Hall 7; and Bren George 8

Abstract: In the last decade, the state departments of transportation in the US have experienced an increase in their construction projects,
while the level of their in-house inspection staff and resources has either remained the same or declined. Previous studies have found that one
strategy that may reduce the inspection workload is prioritizing construction activities for inspection. However, reducing the number of
inspections also has its risks, such as functional failures and reduced design life. Thus, available inspection resources should be allocated
to the activities with significant risk consequences if inspection is reduced. The objective of this paper is to develop a risk-based inspection
protocol to facilitate efficient allocation of available inspection resources to minimize the risks associated with reduced inspection. First, the
risk consequences associated with reduced inspection are identified for various construction activities linked to transportation projects.
Based on data collected from 23 state departments of transportation, 58 engineers and inspectors from the Indiana Department of
Transportation and 20 inspection consultants in the Midwest, the subjective perceived probabilities associated with the occurrence of each
risk consequence are encoded by using fuzzy analysis, from which the risk impacts due to reduced inspection are obtained. The construction
activities are prioritized based on the risk impacts associated with reduced inspection into five priority levels. The greater the risk impacts are
as a result of reduced inspection, the higher the priority would be for inspection of that activity. Thus, the proposed protocol can be used for
resource allocation based on risk impacts. The proposed list of prioritized construction activities can assist project and program managers in
state departments of transportation in better allocating their limited inspection resources while reducing the risks due to reduced inspection.
DOI: 10.1061/(ASCE)CO.1943-7862.0000664. © 2013 American Society of Civil Engineers.
CE Database subject headings: Inspection; Risk management; Transportation management; Construction; Probability; State government.
Author keywords: Risk-based inspection; Risk analysis; Transportation construction probability encoding; Inspection protocol.

Introduction aging transportation infrastructure in the US. The emergence of


the Transportation Equity Act of 21st Century (TEA-21), State
State departments of transportation have experienced a growth in Infrastructure Banks, and the Transportation Infrastructure Finance
their funding for construction projects to restore and expand the and Innovation Act (TIFIA) allowed state departments of transpor-
tation to expand the number of construction projects. For instance,
1
Ph.D. Candidate and Graduate Research Assistant, School of Civil the Texas Department of Transportation (TXDOT) experienced an
Engineering, Purdue Univ., 550 Stadium Mall Dr., West Lafayette, increase in funding for construction projects of more than 44% dur-
IN 47909-2051 (corresponding author). E-mail: amostafa@purdue.edu ing the last decade (Jagars-Cohen et al. 2009). In the last decade,
2
Professor, School of Civil Engineering, Purdue Univ., 550 Stadium
the Indiana Department of Transportation (INDOT) increased the
Mall Dr., West Lafayette, IN 47907-2051. E-mail: dulcy@ecn.purdue.edu
3
Transportation Systems Section Manager, Division of Research and number of construction projects funded by capital made available
Development, Indiana Dept. of Transportation, West Lafayette, IN through the leasing of the Indiana Toll Road. However, during
47906. E-mail: snoureldin@indot.in.gov the same time period, the level of personnel available for construc-
4
State Construction Engineer, Indiana Dept. of Transportation, tion inspection has either remained the same or declined. The limi-
Indianapolis, IN 46204. E-mail: gpankow@indot.in.gov tation of available inspection resources is partly attributable to:
5
Construction Director, Crawfordsville District, Indiana Dept. of (1) retirement of experienced inspectors, (2) departure of experi-
Transportation, IN 47933. E-mail: jnovak@indot.in.gov
6 enced inspectors to private firms, and (3) insufficient training of
Manager, Office of Materials Management, Indiana Dept. of Transpor-
tation, Indianapolis, IN 46219. E-mail: rwalker@indot.in.gov new inspectors (Martin 2001). To address this problem, “the state
7
Construction Manager, Parsons Brinckerhoff, Indiana Area, IN 46204. Departments of Transportation are addressing their workforce
E-mail: hallk1@pbworld.com challenges by outsourcing key project responsibilities that were
8
Construction Program Oversight Manager, Federal Highway previously performed by in-house state DOT forces and adapting
Administration, Indiana Division, Indianapolis, IN 46204. E-mail: bren their practices to perform construction administration more effi-
.george@dot.gov ciently” (Jagars-Cohen et al. 2009).
Note. This manuscript was submitted on May 10, 2012; approved on
In a search for strategies for inspection workload reduction,
December 5, 2012; published online on December 7, 2012. Discussion
period open until January 1, 2014; separate discussions must be submitted Jagars-Cohen et al. (2009) identified the best strategy to be the
for individual papers. This paper is part of the Journal of Construction creation of checklists for the prioritization of different construc-
Engineering and Management, Vol. 139, No. 8, August 1, 2013. © ASCE, tion activities that help inspectors prioritize inspection elements.
ISSN 0733-9364/(10)/$25.00. This strategy seeks allocation of the available resources for the

JOURNAL OF CONSTRUCTION ENGINEERING AND MANAGEMENT © ASCE / AUGUST 2013 / 977

J. Constr. Eng. Manage., 2013, 139(8): 977-986


Micro-consequences Macro-consequences
inspection of the most critical construction activities. However,
there is no formal approach to determining whether the inspec-
tion efforts of the state DOTs are focused on the most critical Longitudinal Reduced
activities. To facilitate efficient allocation of inspection resources and transverse design life
to the most critical activities, this paper presents a risk-based cracks

protocol for prioritization of the inspection of construction


activities. First, an overview is presented of the current state of Settlement Functional
Limited Reduced
inspection practices in state DOTs. Next, the research methodol- failures
Downloaded from ascelibrary.org by CEPT - Centre for Environmental Planning & Tech University on 11/07/19. Copyright ASCE. For personal use only; all rights reserved.

Inspection inspection
ogy, the data collection and probability encoding processes, and Resources
the risk analysis methodology are described. A discussion of the
results of the analysis and the summary and conclusions close Rebar Reduced
corrosion safety
the paper.
Increased
Structural maintenance
State of Practice in State DOTs deformation costs

To evaluate the current state of practice regarding the inspection


of construction activities, a survey was deployed by the research Other micro- Other macro-
consequences consequences
team of this study to all state DOTs in the summer of 2010, of
which 23 responded. The results of the survey showed that that
74% of the state DOTs that responded had experienced changes in
their inspection staffing level over the last five years. On the other
Fig. 1. Micro and macro-risk consequences due to missed/reduced
hand, the Ohio, Illinois, and North Carolina DOTs reported in-
inspection
spection staff growth. However, this growth was not proportional
to the growth in the number of construction projects. For DOTs in
some states, such as Indiana and Texas, which experienced reduc-
tion of inspection workforce, the workforce typically assigned to longitudinal cracks in asphalt or soil settlement in an embank-
maintenance activities was now performing construction inspec- ment. One or more microconsequences will lead to macroconse-
tion activities. Respondents to the survey stated that the lack of quences such as short-term functional failures or reduced
experience and the differing expertise of the maintenance work- design life.
force reduced the efficiency of their construction inspections. There are two categories of risks: known risks and unknown
Eighty-three percent of the responding state DOTs stated that they risks (Pawson et al. 2011). Known risks are those whose out-
implement their construction inspection using both in-house and comes are known but their probability of occurrence is unknown.
outsourced inspectors. Forty-four percent of the DOT respondents In unknown risks, both the nature of the outcome and the prob-
do not consider their current inspection practices to be efficient, ability of occurrence are unknown. In this study and for the
implying that inspection resources are not necessarily allocated assessment of the risks due to missed/reduced inspection, only
efficiently to the most critical activities. Sixty-five percent of state known risks have been evaluated. Thus, it is assumed that the
DOTs seek full observation of certain construction activities and nature of the risk outcomes due to missed/reduced inspection
inspect other activities when resources are available; the remaining is detectable and known.
35% require contractor certification with a quality control (QC)
program and provide random inspection for quality assurance
(QA). Also, 74% of responding state DOTs indicated that they Risk-Based Inspection
do not have a protocol for prioritizing the inspection of construc- Risk-based inspection is a widely used concept for pipe systems
tion activities. Among the remaining 26%, Nevada DOT indicated in oil and gas infrastructure. Reynolds (1996), Dey et al. (2004),
having informal guidelines for prioritizing inspection of construc- Nalli (2007), and Tien et al. (2007) presented risk-based inspection
tion activities. The prioritization of construction activities for in- frameworks for oil and gas infrastructure, and Straub and Faber
spection is left to the experience and judgment of the inspectors (2006) discussed the computational aspects of risk-based inspec-
on the construction site. With a high rate of retirement and depar- tion planning. The two components of risk-based inspection assess-
ture of experienced inspectors from the state DOTs, the significant ment are (1) risk consequences and (2) probabilities of occurrences
challenge facing new inspectors is the prioritization of construc- of risk consequences. The risk impact is calculated by using Eq. (1).
tion activities for inspection. Thus, the greater the risk consequences and the probability of oc-
currence of risk consequences, the greater the risk impact, leading
to higher priority for inspection:
Research Methodology

To minimize the risks associated with reduced inspection due Risk impact ¼ Risk consequences × Probability of occurrence
to insufficient resources, inspection of construction activities ð1Þ
should be prioritized. Construction activities that present signifi-
cant risks as a result of reduced/missed inspection should be given The risk consequences and their likelihood can be recorded
a higher priority for inspection. The decision regarding whether to from historical data. Such data are not readily available in the trans-
inspect a construction activity now, later, or never, should be made portation infrastructure domain. Data related to defects (such as
based on the subsequent risks. If the inspection of the activity can- cracks) and the frequencies of defects in transportation infrastruc-
not be implemented at a later time, there will be micro and macro- ture facilities are recorded. However, it is not known to what extent
risk consequences associated with the missed/reduced inspection, the consequences can be attributed to missed inspection, which
as shown in Fig. 1. Microconsequences are consequences such as may be an impediment to employing risk-based inspection for

978 / JOURNAL OF CONSTRUCTION ENGINEERING AND MANAGEMENT © ASCE / AUGUST 2013

J. Constr. Eng. Manage., 2013, 139(8): 977-986


Identification of the macro consequences due to missed/reduced inspection Analysis of Transportation Construction Activities

The research methodology discussed in the previous section was


Encoding the probabilities of the macro consequences used to develop an inspection protocol for the prioritization of
transportation construction activities. An elaboration of the steps
through which the protocol is obtained is presented in the remain-
Evaluation of the risk impacts due to missed/reduced inspection
der of this section.
Downloaded from ascelibrary.org by CEPT - Centre for Environmental Planning & Tech University on 11/07/19. Copyright ASCE. For personal use only; all rights reserved.

Prioritization of construction activities based on risk impacts due to missed/reduced inspection Identification of the Macroconsequences of Reduced
Inspection
Fig. 2. Methodological framework of the research To identify the macroconsequences of reduced inspection, the
Delphi method was adopted. The Delphi method is a communica-
tion technique designed to obtain the insights of a panel of experts
through two rounds of interviews. The results of each round of in-
transportation infrastructure. An alternate approach to address terviews are summarized and given to the experts in the next round
the lack of appropriate data would be to obtain the information of interviews so the experts can modify their judgment. The process
from subject matter experts based on their experience. To elicit stops when the panel of experts reaches a consensus on the subject
the required data from subject matter experts, the level of detail under investigation.
should be limited to avoid the overestimation of risk consequences In this study, seventeen site visits were conducted at five high-
and their likelihoods (Anderson 1998). Thus, it would be better to way construction projects in the state of Indiana. The projects in-
assess the macroconsequences due to missed/reduced inspections cluded different construction activities such as asphalt paving,
to limit the level of details and the number of events for which concrete paving, embankment, and bridge construction (Fig. 3).
subject matter experts estimate probabilities. Fig. 2 shows the The information regarding the site visits is summarized in Table 1.
methodological framework of this paper. As shown in Fig. 2, the In addition to observing the construction processes on sites, inter-
risk-based assessment of the inspection of construction activities views were conducted, individually and in groups, with 12 project
can be implemented through the following steps: (1) identification engineers and inspectors of INDOT and consulting companies in-
of the macroconsequences instead of microconsequences; (2) prob- specting INDOT projects. The interviews included providing an
ability encoding to extract the “perceived (subjective) probabilities” overview of the objectives of the study and asking open-ended
of macroconsequences instead of actual probabilities; (3) evaluation questions to facilitate discussion of the emergent topics within
of risk impacts due to missed/reduced inspection; and (4) prioriti- and across the interviews. Examples of questions include: “What
zation of construction activities based on risk impacts due to would be the consequences of missing the inspection of asphalt
missed/reduced inspection. compaction? What would be its short- and long-term consequen-
ces?” Each interview took between 60 and 120 min, and often also
included discussions between/among the interviewees. The inter-
Probability Encoding views were recorded and analyzed through transcription and coding
The concept of perceived (subjective) probability was introduced for use in identifying the macroconsequences. After the first round
by De Finetti (1964). Perceived probability refers to the likelihood of interviews, a list of the macroconsequences of reduced/missed
that one assigns to a particular uncertain consequence base. One of inspection were identified. The input from all 12 interviewees was
the characteristics of human reasoning is to form judgments from combined and provided to the same 12 interviewees in the second
uncertain and incomplete evidence (Duda et al. 1976). The process round of site interviews. The interviewees reached a consensus that
of extracting and quantifying individual judgment about the like- there are six groups of macroconsequences attributable to missed
lihood of an uncertain consequence is called probability encoding inspection: short-term functional failures, long-term functional
(Spetzler and Von Holsten 1975). The encoded probability gets failures, increased user costs, decreased design life, increased main-
closer to the actual probability if: (1) the occurrence of the uncer- tenance costs, and decreased safety. Although, in reality, the iden-
tified risk consequences are not independent and mutually
tain consequence is frequent, (2) the uncertain consequence is a
exclusive, for simplifying the risk analysis and the probability en-
result of few causes, (3) the individual has sufficient knowledge
coding process in this study, they are assumed to be independent
regarding the consequence, and (4) the individual is neither risk-
(e.g., the occurrence of increased maintenance costs is independent
averse or risk-taking.
of the occurrence of short-term functional failures), but they are not
The assessment of subjective probability is based on certain
assumed to be mutually exclusive.
heuristics (i.e., experience-based mental models) such as availabil-
ity, representativeness, and anchoring (Tversky and Kahneman
1974, 1982; Von Winterfeldt and Edwards 1986). Availability re- Encoding the Perceived Probabilities of
fers to having memories of an event taking place, representative- Macroconsequences of Reduced Inspection
ness refers to making judgments based on the similarity of a sample In this study, the individuals from whom the perceived probabilities
of events to the population, and anchoring refers to having previous were derived are engineers and inspectors who have sufficient
knowledge regarding the occurrence of an event (Spetzler and Von knowledge regarding the consequences of reduced inspection.
Holsten 1975; Barnes 1984; Kahneman et al. 1974; Kahneman and However, the consequences of reduced inspection usually manifest
Tversky 1996). Due to these heuristics, the encoded probability after the project is completed and when the inspectors are no longer
would not be equal to the actual probability, and cognitive biases on the job. Thus, the frequency of observing the consequences of
exist. However, probability encoding is useful in understanding reduced inspection is low. In addition, the problems with function-
which consequence is more likely to occur, even when the order ality may be the result of other causes, such as problems due to poor
of magnitude of the likelihood is different from actual probabilities design or severe weather conditions. Furthermore, different inspec-
derived from historical data. tors have different risk attitudes (e.g., risk averse, risk neutral, and

JOURNAL OF CONSTRUCTION ENGINEERING AND MANAGEMENT © ASCE / AUGUST 2013 / 979

J. Constr. Eng. Manage., 2013, 139(8): 977-986


Downloaded from ascelibrary.org by CEPT - Centre for Environmental Planning & Tech University on 11/07/19. Copyright ASCE. For personal use only; all rights reserved.

Excavation - SIR -30843-A On SR 25 – visited on 7-1-2010 Culvert excavation - SRS-31918-A on I-65 – visited on 6-10-2010

Aggregate base course - SRS-31918-A on I-65 – visited on 8-4-2010 Bridge construction - SIR-30843-A On SR 25 – visited on 7-1-2010

MSE wall construction -SB-28901-A US 52 Norfolk – visited on Concrete paving - R-30576-A on SR 38 – visited on 8-2-2010
8-12-2010

Fig. 3. Site visits for data collection (images by the authors)

Table 1. Summary of Site Visits for Data Collection


Date of Organization conducting
Number Project site visit the inspection Type of interviews and activities observed
1 SIR-30843-A on SR 25 5-28-2010 HNTB corporation Interviewing project inspectors regarding
consequences of missed/reduced inspection
2 SIR-30843-A on SR 25 6-22-2010 HNTB corporation Interviewing project inspectors regarding
consequences of missed/reduced inspection
3 SIR-30843-A on SR 25 7-1-2010 HNTB corporation Bridge construction and base course earthwork
4 SRS-31918-A on I-65 6-3-2010 Indianapolis Testing Lab Interviewing project inspectors regarding
consequences of missed/reduced inspection
5 SRS-31918-A on I-65 6-10-2010 Indianapolis Testing Lab Base course earthwork, base stabilization, bridge
construction, and culvert construction
6 SRS-31918-A on I-65 6-16-2010 Indianapolis Testing Lab Base course earthwork, base stabilization, bridge
construction, and culvert construction
7 SRS-31918-A on I-65 7-08-2010 Indianapolis Testing Lab Base stabilization, asphalt paving, bridge
construction, and culvert construction
8 SRS-31918-A on I-65 8-04-2010 Indianapolis Testing Lab Base stabilization, asphalt paving, and bridge
construction
9 SB-28901-A on US 52 6-1-2010 PBWorld Interviewing project inspectors regarding
consequences of missed/reduced inspection
10 SB-28901-A on US 52 6-30-2010 PBWorld Base course earthwork and bridge construction
11 SB-28901-A on US 52 8-12-2010 PBWorld Base course earthwork and bridge construction
12 SB-28901-A on US 52 9-23-2010 PBWorld Base course earthwork and bridge construction
13 R-30576-A on SR 38 6-2-2010 INDOT Interviewing project inspectors regarding
consequences of missed/reduced inspection
14 R-30576-A on SR 38 6-18-2010 INDOT Base course earthwork and pipe installation
15 R-30576-A on SR 38 7-09-2010 INDOT Base course cement stabilization
16 R-30576-A on SR 38 8-02-2010 INDOT Concrete paving
17 R-31484-A on SR 38 6-24-2010 INDOT Interviewing project inspectors regarding
consequences of missed/reduced inspection

980 / JOURNAL OF CONSTRUCTION ENGINEERING AND MANAGEMENT © ASCE / AUGUST 2013

J. Constr. Eng. Manage., 2013, 139(8): 977-986


risk taking). Thus, the encoded perceived probabilities from dif- survey deployment, ACEC Indiana had 83 active members. The
ferent individuals are not the same and may not reflect the actual target respondents included construction managers in the compa-
probabilities. The objective of this paper is not to obtain an accurate nies. Twenty responses were collected from 20 companies (one
estimate of the probability distributions of the occurrence of risk response per company). The response rate was 24.1%. The low
consequences due to reduced inspection, but rather to use the en- response rate can be attributed to the roles of the target respondents.
coded probability estimates to identify the construction activities in Baruch and Holtom (2008) found that the average response rate
which observing a risk consequence due to reduced inspection is from individuals in management and executive positions is 35.7%,
more likely. Tversky and Koehler (1994) refer to this as “the assign- which is considerably lower than that from other respondents
Downloaded from ascelibrary.org by CEPT - Centre for Environmental Planning & Tech University on 11/07/19. Copyright ASCE. For personal use only; all rights reserved.

ment of probabilities by experts to the description of an event rather (such as employees in nonmanagerial positions). For the survey
than the event itself.” Thus, probability encoding may be useful in deployed to INDOT area engineers and inspectors, INDOT had
understanding which consequence is more likely to occur, although approximately 290 project engineers and supervisors, and 58 from
the order of magnitude of the likelihood is different from the actual this group responded to the survey. The response rate was 20%.
probabilities. Although the response rate was not high, there were at least five
There are many approaches for probability encoding, and select- responses for each of the six districts of INDOT.
ing an approach depends on the nature of the problem. For instance, In the surveys, the experts were asked to comment on the typical
if the nature of the problem requires the evaluation of an individ- inspection practices in their organizations and on different inspec-
ual’s risk attitudes and perception, direct interviews with the sub- tion workload reduction strategies used by their organizations.
jects are appropriate, as discussed by Spetzler and Von Holsten These questions were asked to implement the structuring and con-
(1975). However, if the problem requires the assessment of the like- ditioning stages of probability encoding, as introduced by Spetzler
lihood of the occurrence of a certain consequence based on the per- and Von Holsten (1975). Structuring refers to clearly defining the
ception of a large population of experts, the use of direct interviews uncertain variable for the experts, and conditioning refers to making
is not viable. The choice of experts is the most important step of the experts think about the uncertain variable. The respondents
group subjective probability encoding (Daneshkhah 2004). were asked to assign subjective probabilities to the likelihood of
Three groups of individuals are involved in planning and imple- risk consequences due to reduced inspection. The use of verbal
menting the inspection of transportation construction activities: expressions is an appropriate approach to elicit the perceptions
(1) state construction engineers, who plan strategies for inspection of uncertainty from experts (Clark 1990). When subjective proba-
of projects; (2) consulting companies that implement inspection for bilities are collected by using survey questionnaires from a group
state DOTs; and (3) area engineers, project engineers, and inspec- of experts in which experts communicate their perceptions regard-
tors, who implement inspection on construction sites. The collec- ing the likelihood of events using verbal expressions, probability
tion of data from the area engineers and inspectors of all state encoding using fuzzy logic is viable. Fuzzy set theories are power-
DOTs was not feasible in this study because of the large number ful mathematical tools for modeling uncertain systems (Mostafavi
of engineers and inspectors in each state. Thus, only state construc- and Karamouz 2010). These tools facilitate probability encoding
tion engineers in state DOTs were targeted. The collection of data in the absence of precise and complete information. Fig. 4 shows
from project engineers and inspectors at the state level was limited the steps of the probability encoding process and risk analysis. The
to INDOT engineers and inspectors. Also, the targeted consulting following sections present the fuzzy probability encoding and the
companies included those in the Midwest. steps through which the risk impacts are derived.
A web-based survey was used and deployed through an online
survey panel service between August of 2010 and January of 2011. Step 1: Fuzzification of Subjective Probabilities
The target respondents included state construction engineers in Linguistic terms such as “likely” or “probable” are acceptable ways
state DOTs, construction managers in contracting companies in the to express the notion of uncertainty (Clark 1990). These terms carry
Midwest, and area engineers and inspectors of INDOT. An opt-in meaning for communicating degrees of uncertainty, but they are
mechanism for participation was utilized in the surveys, giving less precise than numbers. These verbal expressions can be quan-
participants the choice to participate. The objective of the study was tified by using fuzzy numbers to assist in probability assessment
described in the introduction of the survey, and the survey was (Beyth-Marom 1982; Bonissone et al. 1987; Clark 1990; Van der
qualified for Institutional Review Board Research Exemption by Gaag et al. 2002). A fuzzy number does not refer to one single
Purdue University. The reliability of the survey was tested through value, but rather to a continuous set of possible values, where each
content validity, which consisted of a review of the contents of
the survey by five subject matter experts (SMEs) who were not
included in the sample of survey respondents. The SMEs who
Step 1: Fuzzification of the subjective probabilities
evaluated the reliability of the surveys included four INDOT project
engineers and one project engineer from a consulting company.
The SMEs had a minimum of 10 years of experience in construc- Step 2: Aggregation of the probabilities
tion inspection. These surveys were deployed in August 2010,
September 2010, and January 2011.
For the survey deployed to the state DOTs, the population in- Step 3: Defuzzification of the fuzzy probabilities
cluded 49 state DOTs (all US states minus Indiana). The target
respondents included the construction engineer from each state.
Step 4: Evaluation of the risk impacts based on
Twenty-three responses were received from 23 state construction individual risk consequences
engineers (one response per state). Thus, the response rate was
46.9%, which is very close to the average response rate of 48%
Step 5: Evaluation of overall risk impact for a
in organizational research studies (Baruch and Holtom 2008). construction activity, based on all risk consequences
For the survey deployed to inspection consultants in the Midwest,
members of the American Consulting Engineers Council (ACEC)
Fig. 4. Steps in probability encoding
in Indiana were chosen as the sample population. At the time of

JOURNAL OF CONSTRUCTION ENGINEERING AND MANAGEMENT © ASCE / AUGUST 2013 / 981

J. Constr. Eng. Manage., 2013, 139(8): 977-986


possible value has its own weight between 0 and 1. This weight is One of the most commonly used approaches is the linear opinion
called the membership function (Hanss 2005). A triangular fuzzy pool (Stone 1961). With the linear opinion pool, the aggregated
number is represented using three components, as shown in Eq. (2). probability was obtained by using Eq. (3):
Values less than the left-hand side component and greater than the
right-hand side component have a membership function of zero. pðmacro risk consequencejmissed inspectionÞ ¼ pðθÞ
The values between the left and right-hand values have membership X
5
functions between 0 and 1. The middle component signifies the ¼ wi pi ðθÞ ð3Þ
value with the membership function of 1: i¼1
Downloaded from ascelibrary.org by CEPT - Centre for Environmental Planning & Tech University on 11/07/19. Copyright ASCE. For personal use only; all rights reserved.

pðθÞ∶ðθ1 ; θ2 ; θ3 Þ ð2Þ where pi ðθÞ represents the probability fuzzy numbers (assigned
by individual experts) presented in Table 2; wi = percentage of
Van der Gaag et al. (2002) proposed a scale for transforming experts who assigned pi ðθÞ to the uncertain consequence θ;
probability linguistic terms to fuzzy numbers (Table 2). The trans- pi ðθÞ = aggregated fuzzy number corresponding to the probability
formation is called fuzzification. The scale has not been proven to of consequence θ occurs due to reduced inspection of a construc-
be context-specific (Clark 1990). In this study, the scale presented tion activity pðmacro risk consequencejmissed inspectionÞ.
in Table 2 was used for the fuzzification of the linguistic terms Step 3: Defuzzification of Fuzzy Probabilities
corresponding to the perceived probabilities of risk consequences The probability fuzzy numbers need to be defuzzified so they
due to reduced inspection. For instance, Table 2 shows that 50% is can be used as probability point estimates for risk analysis.
the most representative probability corresponding to the medium Defuzzification refers to transforming a fuzzy number into a regular
likelihood probability linguistic term, which has a membership crisp number. The method used for defuzzification in this study
value of 1 in the triangular fuzzy number. As probabilities move is the centroid method. The centroid of a triangular fuzzy number
farther from 50%, they become less representative of the medium is equal to the average of the three components of the fuzzy
likelihood probability linguistic term and their membership values number [Eq. (4)]:
decrease. As the probabilities become less than 40% or greater than
60%, they are no longer representative of the medium likelihood θ1 þ θ2 þ θ3
Centroid½PðθÞ∶ðθ1 ; θ2 ; θ3 Þ ¼ ð4Þ
probability linguistic term. Thus, they have a membership value 3
of 0. In the survey questionnaires, the experts were asked to assign
probabilities of occurrence of the risk consequences from reduced Step 4: Evaluation of Risk Impacts
inspection by using verbal expressions. The assigned probabilities
As shown in Eq. (1), the risk impact is the product of the risk
by each expert were fuzzified using the scale shown in Table 2
consequence multiplied by the probability of occurrence of the
and Fig. 5.
risk consequence. Risk consequences are usually evaluated by
using dollar values; however, in this assessment of the macrocon-
Step 2: Aggregation of Probabilities
sequences of reduced inspection, it was difficult to assign dollar
The assessments of several experts should be combined to capture
values to the risk consequences. Thus, it is assumed that all the
the wisdom of the crowd and to normalize the differences in the risk
macroconsequences are of equal significance (e.g., similar dollar
attitudes of the experts (Winkler et al. 1992). There are various
values) so that the risk impacts are derived solely from the prob-
methods for aggregating the perceived probabilities of several ex-
ability of occurrence of the risk consequences. Hence, the risk
perts (Lindley et al. 1979; West 1988; Clemen and Winkler 1999).
impacts for a given risk consequence in a construction activity
can be represented by the value of the probability of risk conse-
quences, pi ðθÞ.
Table 2. Fuzzy Numbers Corresponding to Probabilities
Linguistic term Fuzzy number ½pðθÞ∶ðθ1 ; θ2 ; θ3 Þ Step 5: Evaluation of Average Risk Impact for a
Construction Activity
Very unlikely (0.15, 0.17, 0.25) In the formulas, pi ðθÞ is the probability of occurrence of risk
Unlikely (0.25, 0.30, 0.40)
consequence j due to reduced inspection in a construction activity.
Medium likelihood (0.40, 0.50, 0.60)
Likely (0.60, 0.70, 0.75)
Because in the previous step it was assumed that all risk consequen-
Very likely (0.75, 0.83, 0.90) ces are of equal significance, the average risk impact (considering
all risk consequences) attributable to reduced inspection in a
construction activity is equal to the average of the probabilities of
occurrences of the six identified risk consequences (i.e., short-term
Very Unlikely Unlikely Medium Likelihood Likely Very Likely functional failures, long-term functional failures, increased user
1.0 costs, decreased design life, increased maintenance costs, and
decreased safety) and can be evaluated by using Eq. (5). In other
Degree of Membership

words, the average risk impact is equal to the average risk


probability:
P6
j¼1 pðθj Þ
Average probability of risk impacts ¼ ð5Þ
6

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Triangular Fuzzy Numbers (probability) Prioritization of Construction Activities for Inspection
The results of the risk analysis performed by using Steps 1–5 of the
Fig. 5. Fuzzy numbers corresponding to the probability linguistic
probability encoding process are summarized in Table 3. The re-
terms
sults presented in Table 3 are based on the responses of 101 experts

982 / JOURNAL OF CONSTRUCTION ENGINEERING AND MANAGEMENT © ASCE / AUGUST 2013

J. Constr. Eng. Manage., 2013, 139(8): 977-986


Table 3. Average Risk Impacts Based on Responses from Different Expert To assess whether the obtained results are sensitive to the re-
Groups sponses from different groups of experts, the analyses were per-
All formed separately for the responses of different groups of experts.
responses DOT Consultants INDOT Three sets of surveys were deployed to state DOTs, inspection
Construction activity (%) (%) (%) (%) consultants in the Midwest, and INDOT area engineers and inspec-
Traffic control setup 43 48 46 40 tors. These surveys contained different sections to collect data re-
Clearing site 32 34 30 32 garding the inspection practices of different organizations. Further,
Stripping 37 38 34 38 all three sets of surveys included a similar section related to assess-
Downloaded from ascelibrary.org by CEPT - Centre for Environmental Planning & Tech University on 11/07/19. Copyright ASCE. For personal use only; all rights reserved.

Clearing site: bridge 34 37 31 34 ment of the subjective probabilities of risk consequences of missed/
Installing soil erosion/sediment 46 46 47 45 reduced inspections for different construction activities. The data
control items related to the subjective probabilities of risk outcomes were evalu-
Excavation 46 46 46 46 ated both individually and in conjunction with other construction
Blasting 44 44 44 45
activities. As shown in Table 3, the results of the individual analy-
Handling/removal of regulated 45 42 46 47
waste
ses of different groups of experts were very similar (e.g., for em-
Aggregate base courses 58 57 57 59 bankment activity, the encoded probabilities obtained from the state
Embankment 57 56 55 58 DOTs, consultants, and INDOT surveys were equal to 56, 55, and
Milling 42 41 39 44 58%, respectively). This result implies that: (1) there is no signifi-
Asphalt paving 63 65 58 64 cant difference in the risk attitude of the group of experts from state
Concrete paving 64 64 59 65 DOTs, consultants, and INDOT; and (2) the methodology used
Concrete forms (structures) 49 45 48 52 in the study was successful in eliciting the beliefs of the experts.
Reinforcement steel in structures 57 55 56 58 Because the findings from the surveys deployed to three groups of
Placement of concrete in structures 61 61 59 62 experts involved in the inspection of construction activities are very
Structure rehabilitation (repairs to 60 61 57 60
similar, the findings could be considered to be generalizable.
concrete deck)
Drilled shafts 51 57 52 49 To prioritize inspection activities based on the level of risks
Driven piles 58 57 56 60 due to missed inspections, a risk-based inspection protocol can be
Sheet piles 49 50 45 49 created. The greater the average risk impact of a construction ac-
Cofferdams 45 44 37 48 tivity, the higher the priority of the activity for inspection. Initially,
Beam erection 54 52 51 56 three categories of priorities of construction activities for inspection
Bolting structural connections 58 59 58 58 were defined: high, medium, and low. The boundaries of the differ-
Posttensioning (prestressed 58 60 59 57 ent categories were set based on the initial analysis of the results,
structures) which indicated that 95% of the average values of the encoded per-
Painting steel 50 57 49 48
ceived probabilities of risk outcomes were greater than 30% and
Guardrail/cable rail 50 56 50 49
Barrier curb 48 49 44 48
less than 65%. Therefore, the range (i.e., 30–65%) was divided into
Sidewalk 46 44 44 47 three intervals (below 40%, between 40 and 55%, and above 55%).
Drainage 54 54 54 54 If the average probability of risk consequences was greater than
Traffic stripes/traffic markings 52 56 50 51 55% and less than 65%, the activity was considered to be high
Fence 37 41 35 37 priority; if the average probability of risk consequences was greater
Electrical conduit and wiring 46 51 40 46 than 40% and less than 55%, the activity was considered to be
ITs: fiber optic conduit and cable 46 54 39 45 medium priority; if the average probability of risk consequences
Highway lighting (foundations and 49 50 47 50 was less than 40% and greater than 30%, the activity was consid-
poles) ered to be low priority. Further analysis revealed that there are many
Traffic signals (foundations and 50 52 49 49
activities whose average perceived probability of risk consequences
poles)
Overhead sign structures 49 55 47 48 due to missed inspection was close to the boundary values, which
Landscape plantings 37 43 38 35 made it difficult to judge the priority category in which they would
Pipe placement 56 54 51 59 be appropriate. Thus, two additional intermediate priority catego-
Seal coating 47 50 44 46 ries (i.e., medium-low and medium-high) were defined to address
Sound wall post placement 46 47 42 47 this issue. Table 4 summarizes the list of prioritized construction
Sound wall panel placement 45 46 40 46 activities. The construction activities were prioritized into five cat-
Placement of lighting features 44 47 37 46 egories based on the risks associated with reduced inspection: high,
Subgrade treatment 59 54 53 63 medium-high, medium, medium-low, and low.
Retaining walls 57 56 56 58
The proposed protocol provides a tool for prioritizing inspection
of construction activities. The higher the priority of an activity for
inspection, the greater the risk impacts due to reduced inspection.
from the state DOTs, consultants, and INDOT. The values in the Examples of high priority activities include, but are not limited to,
table indicate the average perceived risk impacts due to reduced aggregate base course placement, asphalt paving, and concrete pav-
inspection for different construction activities. For instance, for ing. Inspection of these activities includes supervising many critical
concrete paving, the average perceived risk impact of reduced in- items. The added values of proper inspection for high priority
spection is 64% based on all responses. This result implies that if activities include, but are not limited to, executing tests, ensuring
the inspection of concrete paving is reduced or missed, it is per- correct implementation of processes, and evaluating completion of
ceived that the likelihood of occurrence of macroconsequences activities according to specifications. For instance, in the aggregate
would be 64%. Although these values do not reflect the actual base course activity, inspectors should check and ensure that the
risk impacts due to the existence of biases, they can be used to iden- required density and moisture content have been achieved accord-
tify the construction activities with greater risk impacts due to ing to the specifications. If these items are not inspected according
reduced inspection. to the specifications, the likelihood increases of not detecting

JOURNAL OF CONSTRUCTION ENGINEERING AND MANAGEMENT © ASCE / AUGUST 2013 / 983

J. Constr. Eng. Manage., 2013, 139(8): 977-986


Table 4. Prioritized Construction Activities for Inspection
Inspection priority level
Medium-low
High priority Medium-high priority Medium priority priority Low priority
Construction activities
Aggregate base courses Beam erection Barrier curb Cofferdam Clearing site
Electrical conduit
Downloaded from ascelibrary.org by CEPT - Centre for Environmental Planning & Tech University on 11/07/19. Copyright ASCE. For personal use only; all rights reserved.

Asphalt paving Pipe placement Blasting and wiring Clearing site: bridge
Bolting structural connections Subgrade treatment Concrete forms (structures) Fence Stripping
ITs: fiber optic
Concrete paving Drilled shafts Drainage conduit and cable
Driven piles Guardrail Excavation Landscape plantings
Handling/removal of
Embankment Overhead sign structure regulated waste Milling
Placement of concrete Highway lighting Placement of
in structures Painting steel (foundations and poles) lighting features
Posttensioning Installing soil erosion/sediment
(prestressed structures) Traffic marking control items Seal coating
Sound wall panel placement Sound
Reinforcement steel in structures wall post placement Sheet piles
Retaining walls Traffic control: setup Sidewalk
Structure rehabilitation Traffic signals (foundations
(repair concrete deck) and poles)

defects that could lead to settlement and longitudinal cracks. In the allocate their limited inspection resources when many activities
asphalt paving activity, inspection activities should be implemented (those whose inspection could not be performed at a later time,
such as sampling according to QC/QA, ensuring the correct asphalt regardless of the level of inspection required) are taking place con-
mix temperature, and asphalt density testing. Failure to check the currently on the job site.
measurements of temperature, density, and other properties of the The estimated risk impacts and the prioritized list of construc-
materials being placed or the process being performed may lead to tion activities are based on the typical conditions that exist in trans-
the lack of discovery of defects that may lead to functional failures portation construction projects. The existence of QC/QA programs,
such as rutting, potholes, and alligator, longitudinal, and transverse warranties, accuracy of design, and level of experience of the
cracks. In concrete paving, examples of critical items to be in- contractors are the most significant factors affecting the reduction
spected include, but are not limited to, concrete sampling and test- of inspection in a project. Another factor that affects the allocation
ing, concrete placement and compaction, and curing. Missed or of the inspection workforce is the schedule of the project. For
reduced inspection of these items may lead to the lack of discovery instance, an activity may be fully inspected if it is the only task
of defects that may lead to plastic shrinkage, drying shrinkage, underway in the project at the time; however, if it is implemented
surface cracks, pop-outs, and delamination. On the other hand, site concurrently with other construction activities, it may be inspected
clearing and stripping are categorized as low priority inspection randomly. The proposed risk-based inspection protocol may assist
activities based on the risk analysis. Site clearing only requires inspectors in deciding whether a certain activity is more critical
checking the clearing area and the depth of clearing. Failure to in- to be inspected when there are different construction activities
spect these items is not likely to lead to the lack of discovery of the occurring concurrently on the job site.
defects. Inspection of stripping includes checking the depth of re-
moval of the top soil. If these items are not inspected as specified or
inspected at a later time, the likelihood of a major risk outcome is Summary and Conclusions
low. Thus, the priority for the inspection of these activities is low.
Testing and safety requirements are typically specified when The retirement of experienced inspectors, the departure of experi-
there is an increased perceived probability of macro-risk conse- enced inspectors to private firms, and insufficient inspection train-
quences due to missed inspections. For instance, activities such ing have led to increased workloads as a result of insufficient
as asphalt paving, concrete paving, aggregate base course, and em- resources for the inspection of construction projects for state DOTs
bankment require testing. Activities such as structure rehabilitation in the US. The study presented in this paper proposes a risk-based
and bolting structural connections require testing to ensure reliabil- inspection protocol for the inspection of transportation construction
ity and safety (e.g., safety of workers and public during the con- activities as a strategy for the reduction of inspection workload. The
struction phase and safety of facility users after the construction theory behind the proposed protocol is that the activities that ex-
phase). Thus, these activities are perceived to experience greater perience greater risks from missed/reduced inspection should be
risk impacts due to reduced inspection. In addition, activities such given a higher priority for inspection.
as installing reinforcement steel in structures in which the work is Risk analysis was performed to identify the risk impacts of
covered upon completion of the activity (i.e., these activities cannot missed/reduced inspection. The risk consequences due to reduced
be inspected later unless the work is destroyed) are perceived to inspection were identified through site visits and interviews with
entail greater risk impacts due to missed inspections. The proposed inspectors. The subjective probabilities corresponding to the per-
risk-based inspection protocol can be used for resource allocation ceived probability of risk consequences due to reduced inspection
based on the risk impacts. The proposed list of prioritized construc- were encoded. The subjective probability encoding process in-
tion activities may assist project and program mangers to optimally cluded the deployment of three separate sets of surveys to the state

984 / JOURNAL OF CONSTRUCTION ENGINEERING AND MANAGEMENT © ASCE / AUGUST 2013

J. Constr. Eng. Manage., 2013, 139(8): 977-986


DOTs, consultants, and INDOT. A total of 101 expert responses References
were elicited through a fuzzy probability encoding approach and
the risk impacts for different construction activities were calculated. Anderson, J. L. (1998). “Embracing uncertainty: The interface of Bayesian
Based on the calculated risk impacts, transportation construction statistics and cognitive psychology.” Conservat. Ecol., 2(1), 2.
activities were prioritized. Barnes, J. H. (1984). “Cognitive biases and their impact on strategic
planning.” Strat. Manage. J., 5(2), 129–137.
In this study, the list of prioritized construction activities was
Baruch, Y., and Holtom, B. C. (2008). “Survey response rate levels and
validated through discussions with SMEs. The validation of models
trends in organizational research.” Human Relat., 61(8), 1139–1160.
by SMEs has been widely used in different research domains (Law
Beyth-Marom, R. (1982). “How probable is probable? A numerical
Downloaded from ascelibrary.org by CEPT - Centre for Environmental Planning & Tech University on 11/07/19. Copyright ASCE. For personal use only; all rights reserved.

2006). The validation of models by SMEs requires face to face translation of verbal probability expressions.” J. Forecast., 1(3),
meetings to describe the objectives of the study and the components 257–269.
of the model. The proposed risk-based inspection protocol was Bonissone, P. P., Gans, S. S., and Decker, K. S. (1987). “RUM: A layered
evaluated and verified by three experienced SMEs in separate face architecture for reasoning with uncertainty.” Proc., 10th Int. Joint Conf.
to face meetings that lasted more than an hour for each SME. Two of Artificial Intelligence, Morgan Kaufmann, Los Altos, CA, 891–896.
of the SMEs were INDOT project engineers and one was an engi- Clark, D. A. (1990). “Verbal uncertainty expressions: A critical re-
neer from a consulting company. Although there is no consensus in view of two decades of research.” Curr. Psychol. Res. Rev., 9(3),
the literature regarding the number of experts required for valida- 203–235.
tion, the level of experience of the experts involved in the validation Clemen, R. T., and Winkler, R. L. (1999). “Combining probability distri-
process is greatly emphasized (Sandelowski 1998; Sargent 2010). butions from experts in risk analysis.” Risk Anal., 19(2), 187–203.
In this study, the SMEs had a minimum of 10 years of experience in Daneshkhah,, A. R. (2004). “Uncertainty in probabilistic risk assessment:
transportation construction inspection. Nevertheless, the proposed A review.” Univ. of Sheffield, 〈http://www.shef.ac.uk/beep/publications
protocol should be further validated after its adoption and applica- .html〉 (Jun. 2011).
tion in construction projects over time. Currently, INDOT plans De Finetti, B. (1964). “Foresight: Its logical laws, its subjective sources.”
Studies in subjective probability, H. Kyburg and H. Smokler, eds.,
to use the proposed inspection protocol and to include it in its gen-
Wiley, New York.
eral instructions to field employees. After the field employees of
Dey, P. K., Ogunlana, S. O., and Naksuksakul, S. (2004). “Risk-based
INDOT and other state DOTs implement the proposed protocol, the maintenance model for offshore oil and gas pipelines: A case study.”
validity of the protocol may be further evaluated by analyzing the J. Qual. Maint. Eng., 10(3), 169–183.
outcomes related to: (1) efficiency of inspection resource alloca- Duda, R. O., Hart, P. E., and Nilsson, N. J. (1976). “Subjective Bayesian
tion, (2) reduction of the risks due to reduced/missed inspection, methods for rule-based inference systems.” Proc., AFIPS National
and (3) consistency of inspection practices across projects. Computer Conf., Vol. 45, New York, 1075–1082.
The list of prioritized construction activities will be helpful in Hanss, M. (2005). Applied fuzzy arithmetic: An introduction with engineer-
addressing the inspection challenges on job sites. The greater the ing applications, Springer-Verlag, Berlin.
risk impacts due to reduced inspection, the higher the priority for Jagars-Cohen, C. A., Menches, C. L., Jangid, Y. K., and Caldas, C. H.
inspection. This result implies that, while facing limited inspection (2009). “Priority-ranking workload reduction strategies to address
resources, state DOTs can allocate their available resources for the challenges of transportation construction inspection.” Transportation
inspection of their high priority activities. This will be helpful in Research Record 2098, Transportation Research Board, Washington,
reducing the inspection workload while reducing the risks associ- DC, 13–17.
ated with missed/reduced inspection. The proposed protocol is Kahneman, D., Slovic, P., and Tversky, A. (1974). “Judgment under
intended for use by state DOTs as a strategy to address their current uncertainty: Heuristics and biases.” Science, 185(4157), 1124–1131.
challenges of the reduction of inspection workforce and the in- Kahneman, D., and Tversky, A. (1996). “On the reality of cognitive
illusions.” Psychol. Rev., 103(3), 582–592.
crease in construction inspection workload. Using the proposed
Law, A. M. (2006). “How to build valid and credible simulation models.”
protocol, available inspection resources can be allocated to more
Proc., 2006 Winter Simulation Conf., S. J. Mason, R. Hill, L. Moench,
critical activities when two or more activities are underway on a and O. Rose, eds., IEEE, Piscataway, NJ, 39–47..
project site. Lindley, D. V., Tversky, A., and Brown, R. V. (1979). “On the reconcilia-
tion of probability assessments.” J. R. Stat. Soc., 142(2), 146–180.
Martin, C. (2001). “Help wanted: Meeting the needs for tomorrow’s trans-
Acknowledgments portation work force.” Public Roads, 65(1), 2–12.
Mostafavi, A., and Karamouz, M. (2010). “Selecting appropriate project
This research work was supported by the Joint Transportation delivery systems: A fuzzy approach with risk analysis.” J. Constr.
Research Program administered by INDOT and Purdue University Eng. Manage., 136(8), 923–930.
through Project No. SPR-3400. The authors would like to thank the Nalli, K. (2007). “Risk-based inspection—An efficient maintenance tool.”
engineers and inspectors of the state Departments of Transporta- J. Hydroc. Process., 86(12), 91–94.
tion, the engineers and inspectors of INDOT, and project personnel Pawson, R., Wong, G., and Owen, L. (2011). “Known knowns, known
from Parsons Brinckerhoff, HNTB, and Indianapolis Testing Lab unknowns, unknown unknowns: The predicament of evidence-based
for their support and insightful input for this study. The authors policy.” Am. J. Eval., 32(4), 518–546.
also acknowledge Professor Julio Martinez for his insightful feed- Reynolds, J. T. (1996). “Risk and material performance for petroleum, pro-
cess and power.” Proc., of the 1996 ASME Pressure Vessels and Piping
back on this study. The contents of this paper reflect the views
Conf., Vol. 336, Montreal, QB, Canada, 125–134.
of the authors, who are responsible for the facts and the accuracy
Sandelowski, M. (1998). “The call to experts in qualitative research.”
of the data presented herein. The contents do not necessarily reflect Res. Nursing Health, 21(5), 467–471.
the official views or policies of the Federal Highway Administra- Sargent, R. G. (2010). “Verification and validation of simulation
tion and INDOT, nor do the contents constitute a standard, speci- models.” Proc., 2010 Winter Simulation Conf., B. Johansson, S. Jain,
fication, or regulation. Any opinions, findings, and conclusions or J. Montoya-Torres, J. Hugan, and E. Yücesan, eds., 166–183.
recommendations expressed in this paper are those of the authors Spetzler, C. S., and Staël Von Holstein, C. S. (1975). “Probability encoding
and do not reflect necessarily the views of the organizations or the in decision analysis.” Manage. Sci., 22(3), 340–358.
individuals listed here. Stone, M. (1961). “The opinion pool.” Ann. Math. Stat., 32(4), 1339–1342.

JOURNAL OF CONSTRUCTION ENGINEERING AND MANAGEMENT © ASCE / AUGUST 2013 / 985

J. Constr. Eng. Manage., 2013, 139(8): 977-986


Straub, D., and Faber, M. H. (2006). “Computational aspects of risk-based Van der Gaag, L. C., Renooij, S., Witteman, C. L. M., Aleman, B. M. P.,
inspection planning.” Comput. Aided Civ. Infrastruct. Eng., 21(3), and Taal, B. G. (2002). “Probabilities for a probabilistic network:
179–192. A case-study in esophageal cancer.” Artif. Intell. Med., 25(2),
Tien, S.-W., Hwang, W.-T., and Tsai, C.-H. (2007). “Study of a risk-based 123–148.
piping inspection guideline system.” ISA Trans., 46(1), 119–126. Von Winterfeldt, D., and Edwards, W. (1986). Decision analysis and
Tversky, A., and Kahneman, D. (1974). “Judgment under uncertainty: behavioral research, Cambridge University Press, New York.
Heuristics and biases.” Science, 185(4157), 1124–1131. West, M. (1988). “Modelling expert opinion.” Bayesian statistics,
Tversky, A., and Kahneman, D. (1982). “Belief in the law of small J. M. Bernardo, J. O. Berger, A. P. Dawid, and A. F. M. Smith, eds.,
numbers.” Judgment under uncertainty: Heuristics and biases, Oxford University Press, New York.
Downloaded from ascelibrary.org by CEPT - Centre for Environmental Planning & Tech University on 11/07/19. Copyright ASCE. For personal use only; all rights reserved.

D. Kahneman, P. Slovic, and A. Tverskey, eds., Cambridge University Winkler, R. L., Hora, S. C., and Baca, R. G. (1992). “The quality of expert
Press, Cambridge, MA, 23–31. judgment elicitations.” Nuclear Regulatory Commission Contract
Tversky, A., and Koehler, D. J. (1994). “Support theory: A nonextensional NRC-02-88-005, Center for Nuclear Waste Regulatory Analyses,
representation of subjective probability.” Psychol. Rev., 101(4), 547–567. San Antonio, TX.

986 / JOURNAL OF CONSTRUCTION ENGINEERING AND MANAGEMENT © ASCE / AUGUST 2013

J. Constr. Eng. Manage., 2013, 139(8): 977-986

You might also like