See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/225243838

Qualitative Model for Risk Assessment in Construction Industry: A Fuzzy

Logic Approach

Conference Paper  in  IFIP Advances in Information and Communication Technology · February 2010

DOI: 10.1007/978-3-642-11628-5_11 · Source: DBLP

CITATIONS READS

21 1,322

3 authors:

Abel Pinto Isabel L. Nunes

Safe@Work, Independent Safety Research Group NOVA School of Science and Technology, NOVA University of Lisbon (FCT-UNL)
31 PUBLICATIONS   604 CITATIONS    137 PUBLICATIONS   1,240 CITATIONS   

SEE PROFILE SEE PROFILE

Rita Almeida Ribeiro

Institute for the Development of New Technologies
246 PUBLICATIONS   10,524 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

eHealth Solutions View project

Multispectral image classification View project

All content following this page was uploaded by Abel Pinto on 26 October 2014.

The user has requested enhancement of the downloaded file.

 Qualitative Model for Risk Assessment in Construction
Industry: A Fuzzy Logic Approach

Abel Pinto1, Isabel L. Nunes1, and Rita A. Ribeiro2

1
Universidade Nova Lisboa / FCT, Caparica, 2829-516 Portugal
[email protected], [email protected]
2
Uninova, Campus UNL/FCT, Caparica, 2829-516 Portugal
[email protected]

Abstract. Risk Assessment for Health and Safety (RAH&S) of workers is a

complex task that entails the consideration of many parameters which are, more
often than not, difficult to quantify. RAH&S in the construction industry is
rampant with inadequate data and/or imprecise and incomplete information,
particularly in the design stage, for which traditional quantitative approaches do
not give adequate answers. In this work we outline the basic aspects for a Quali-
tative Risk Assessment Model (QRAM) based on elicited data and using a
fuzzy logic approach.

Keywords: Risk assessment, safety, fuzzy sets, construction industry.

1 Introduction

Safety is a prominent feature to be taken into account in the construction industry, and
Risk Assessment for Health and Safety (RAH&S) is a key step to achieve it, particu-
larly to support decision-making in safety programs (Ringdahl, 2001).
In general, risk assessment is a complex process that entails the consideration of
many qualitative parameters which are difficult to quantify. Risk assessment in the
construction industry, due the specific nature of the sector (Dedobbeleer e Beland,
1991; Ringen et al., 1995; Gillen et al, 1997; Laitinen, 1999; Loosemore, 2001; Tam
et al., 2004), must deal with ill-defined and imprecise data and information. So far
traditional approaches do not seem to provide adequate answers (Ringen et al., 1995)
to deal with this issue. Moreover, using probability (classical or Bayesian) or statistics
theories may mask other aspects of incomplete and imprecise knowledge and can lead
to a false sense of accuracy and precision, thus leading to inadequate and/or ineffi-
cient decisions.
To overcome some of the mentioned problems in assessing safety risks, we pro-
pose a qualitative model for risk assessment, hereinafter denoted QRAM (Qualitative
Risk Assessment Model), which based on elicited data and using a fuzzy logic ap-
proach. The goal is to contribute to work safety by improving risk assessment in con-
struction sites.

L.M. Camarinha-Matos, P. Pereira, and L. Ribeiro (Eds.): DoCEIS 2010, IFIP AICT 314, pp. 105–111, 2010.
© IFIP International Federation for Information Processing 2010
106 A. Pinto1, I.L. Nunes, and R.A. Ribeiro

2 Contribution to Technological Innovation

The majority of quantitative traditional methods use probabilistic techniques and / or
statistics to deal deal with the intrinsic uncertainty and imprecision in data and infor-
mation. This implies some limitations, such as requiring analysts to estimate parame-
ters, to ensure a sufficiently representative domain or make comparisons with other
sites (which departs from the real system under study).
Recently, there are also qualitative approaches using fuzzy logic, proposed in the
literature (see for example: Azadeh et al., 2008, Gurcanli, Mungen, 2009), but they
lack having a systematic survey of all parameters that should be taken into considera-
tion, together with a general formulation to obtain a complete risk assessment
measure.
Hence, the main contributions of the proposed QRAM are: to provide a framework
for factors that need to be evaluated in construction sites; to propose a formula ex-
pressing the relationships between the parameters (risk factors), thus obtaining an
overall risk measure; to provide a simple semantic scale to facilitate the elicitation of
information about the parameters. The final aim of QRAM is to obtain an overall
measure of risk for the construction site as well as to obtain partial risk calssifications
for important parameters included in the overall measure.
QRAM will focus on elicited data, thus avoiding estimates to allow assessment of
actual risks of the site. The data and information will be obtained by direct observa-
tion, interviews with workers and foremen and also from consultation and review of
documentation of the site (health and safety plan, reporting accidents and incidents,
records of meetings, work procedures...), among others. After we transform the col-
lected data, using fuzzy logic techniques, and then aggregate it to obtain the risks
level classification for the site.

3 Background on Risk Assessment for Health and Safety

(RAH&S)
Traditional RAH&S methods are typically based on information, which is subject to
uncertainty (usually handled by probability and/or statistics theories) or is just ill-
defined. Due to its intrinsic uncertainty and incompleteness traditional RAH&S meth-
ods present some limitations, as pointed by some authors [Karwowski & Mital,
(1986); Cornell (1996); Wang et Ruxton. (1997); Pender (2001); Sii et al. (2001);
Tixier (2002); Faber (2003); Nilsen (2003); Kentel et Aral. (2004)], such as:
o Inherent imprecision in human-centered systems;
o Difficult to generate a mathematical model due to intrinsic uncertainty in this
type of problems. For instance, it is difficult to represent and describe the
safety behavior, which causes work accidents involving operational proce-
dures, human error and decisions taken by designer and management;
o Difficult to quantify the effects and consequences of hazards because they
involve many factors with a high level of uncertainty, even when the physi-
cal processes are clearly understood;
 Qualitative Model for Risk Assessment in Construction Industry 107

o Large number of assumptions, judgments and opinions are involved in a risk

quantification process, hence, it requires considerable skills from a safety
analyst to interpret results;
o Construction projects are unique by definition. This reduces the relevance
and reliability of statistical aggregates derived from probability-based
analysis;
o Humans are limited in their ability to encompass and process the full range
of information required for holistic decisions;
o Uncertainty and ignorance may be found in temporal aspects of the flow of
knowledge, which are important in project planning;
o Project parameters and outcomes must be communicated to others and the
imprecision of our language is not good to express these.
Considering all above statements, it is apparent that RAH&S in the construction in-
dustry is rampant with inadequate data and/or imprecise, ill-defined, and incomplete
information, particularly in the design stage, for which traditional approaches do not
give adequate answers. Fuzzy logic (Zadeh, 1965, Zimmerman, 1991) provides a
natural way of modelling the intrinsic vagueness and imprecision of subjective as-
sessments, while also allows the inclusion of human creativity and intuition, which is
an essential ingredient for successful risk analysis (Ru, 1996). However, most qualita-
tive approaches neither consider a systematic framework for all parameters that can
influence the risk nor formalize the relationships between parameters (Liu et al., 2001,
Tixier et al., 2002, Mure et al., 2006).

4 Proposed Qualitative Model: QRAM

In this section we outline the basic aspects that a model for RAH&S in the construc-
tion industry should account for. The knowledge acquisition (i.e. data elicitation)
should be obtained by observation of reality, interviews with workers, foreman and
engineers, responsible and consultation of site documents (working procedures, re-
ports of work accident investigation ....). The collected data shall be transformed us-
ing fuzzy sets (Azadeh et al., 2008) and will be aggregated with specialized fuzzy
operators, to obtain a ranking of risks in the construction site.
The starting point is the following formulation for risk assessment in construction
sites, RAC , as shown in eq. 1.The risk is directly proportional to organizational inade-
quacies ( OI ), which are not related with a specific hazard, and other direct factors
( FD ), and additional factors ( FA ). The direct and additional factors are inversely pro-
portional to safety barriers factors ( S B ), which are also dependent on some hazard.

FD ⊕ FA
RAC = OI ⊗ (1)
SB

FD are the main direct factors related with the hazard under analysis (for example, in
falls from height, the factor is the height of the fall).
108 A. Pinto1, I.L. Nunes, and R.A. Ribeiro

FA are additional factors; they also depend on hazards under analysis and include
a variable number of parameters. Examples of aditional factors (for the same hazard)
are: tidiness of workplace, inclination of work floor, friction level between shoe´s
soles and soil.
Both direct factors and additional factors include a variable number of parameters,
which will depend on the site being evaluated. The risk specialist will choose which
factors will be evaluated for the specific site. In general the formulation will be:
F( D ∪ A) = ⊕( f1 , f 2 ,..., f n ) / N (2)

Where N is the number of parameters that will be considered for any of the two
factors.
OI express organizational inadequacies and are a function of the lack (or poor):
safety culture ( Sc ), safety organisation ( So ), work organisation ( Wo ), supervision
( S ), leadership ( L ), personal factors and communications ( C ). Formally,
OI = ⊕( S c , S o ,Wo , S , L, P, C ) / N (3)
with N being the number of parameters.
S B represent the Safety Barriers implemented in the site for controling the risk.
Their effectiveness should be measured by:
S B = ⊕( M , A, S , I ) / N (4)
where ( Hollnagel 1999):
o M -Material barriers physically prevent an action from being carried out or
the consequences from spreading;
o A -Functional (active or dynamic) barriers work by impeding the action to
be carried out, for instance by establishing a logical or temporal interlock.
o S -Symbolic barriers require an act of interpretation in order to achieve their
purpose, hence an “intelligent” agent that can react or respond to the barrier.
o I -Immaterial barriers are not physically present or represented in the
situation, but depend on the knowledge of the user to achieve their purpose.
The proposed process to obtain the final RAH&S measure, RAC (eq 1), will use
concepts and techniques from fuzzy set theory (Zadeh, 1965, 1975, 1983).
In this first work we use a single discrete fuzzy set for classifying the factors and
their respective parameters. The proposed fuzzy classification can be used to elicit
classifications from different sources (e.g. workers, engineers, safety experts), for all
parameters in a user-friendly semantic form. Formally, the general fuzzy parameter
classification membership function, to be used in this preliminary model is:
⎧ Excelent / 1, VeryGood / 0.8, Good / 0.6, ⎫
ParameterClassificat ion = ⎨ ⎬ (5)
⎩ Satisfacto ry / 0. 4, Mediocre / 0 . 2, Bad / 0. 001⎭
At this stage, the terms and respective values considered are just indicative and in
future work they will be validated and tuned. All parameters in eq. 2, eq. 3 and eq. 4
 Qualitative Model for Risk Assessment in Construction Industry 109

(e.g. material barriers M ) will be classified using the above discrete fuzzy set. To
obtain the values for the respective factors (e.g. S B ) we plan to use an aggregation
operator, ⊕ , such as arithmetic sum, average or any other compensatory operator
(Zimmerman, 1991). Finally to obtain the overall risk measure for the construction
site, we plan to use conventional arithmetic operators.
At this stage of the research work we are not yet sure about which operators are
more suitable for obtaining the risk assessment in the construction industry, but a
study about them is foreseen in the near future. Here, as mentioned, we just focused
on the conceptual model for determining the risk assessment.

5 Illustrative Example
In order to illustrate the applicability of the method, we will use a simple illustrative
case for the risk that causes more deaths in Portugal, in the sector: falls from height.
F
In this example we only describe the calculations for additional factors, A , because
the process is similar for the other factors in eq. 1.
Consider a construction work on a scaffold with three meters high, thus the single
direct factor, is FD = 0.001 , because we assumed the intervals: [0 to 0,5m] FD = 1 ;
FD = 0.8 ; ]1 to 1,4m], FD = 0.6 ; ]1,4 and 1,6 m] FD = 0.4 ; ]1,6 to 1,
]0,5 to 1m]
8m] FD = 0.3 ; ]1,8 to 2m] FD = 0.2 , and > 2m FD = 0.001 .
The additional factors considered and the respective semantic classification, using
the fuzzy membership semantic variables (eq. 5) are: (a little clean and tidy) = medio-
cre; (horizontal surface) = good; (foot boards, wood, presented in good repair) = satis-
factory; (when dry, the material floor provides good traction on the sole of the shoe) =
satisfactory; (lighting level well suited to work) = good, (irons waiting on the surface
of the collision) = bad; (dry weather and no wind) = good ( scaffolding meet all legal
and regulatory requirements, was well fitted and properly anchored) = good.
Now, aggregating the classifications for the additional factors (eq. 2), using the
membership values of eq. 5, we have:
F A = (0.2 + 0.6 + 0.4 + 0.4 + 0.6 + 0.001 + 0.6 + 0.6) / 8 = 0.4375

Notice that in this example we used arithmetic operators to aggregate the values, but
in the future other operators will be tested to improve the method.
Although the six factors are appropriate, in various degrees: good (4) and satisfac-
tory (2), there are two factors with inappropriate levels, one mediocre and another
bad. Hence, the combination is only satisfactory (closest semantic level in the classi-
fication fuzzy set for the result 0.4375). This result means the site safety has to be
improved, particularly to protect the irons on hold (bad) and clean and organize the
workplace (mediocre).
Regarding the safety barriers S B , the scaffold has railings around the outside pe-
rimeter of the platforms, properly fitted, good and robust (very good), the inner side
is 20 cm from the construction (very good), the access to work platforms is appro-
priate and also well protected (very good). The company implemented a procedure
110 A. Pinto1, I.L. Nunes, and R.A. Ribeiro

for working at heights which includes work on scaffolding, the procedure is well
prepared and written in language appropriate to their users (very good). The com-
pany provides training to all workers on scaffolding, however, the training records
do not show the knowledge acquired by trainees is properly understood, hence the
company did not prove the effectiveness of training (mediocre). Following the
above calculation for additional factors and using eq. 4,
S B = (0.8 + 0.8 + 0.8 + 0.8 + 0.2) / 5 = 0.68

This result shows that the efficiency in the safety barriers is good (closest semantic
value for the result 0.68) .
Regarding the organizational factors OI (eq. 2) and for simplicity we consider an
overall classification of excelent (membership value = 1).
Finally we can determine the level of risk assessment for the example, following
eq. 1,
0.1 + 0.4375
RAC = 1 × = 0.79
0.68
Notice we used again arithmetic operators to improve clarity of method´ explanation.
The final risk level in the construction site for the falling height, yields a value of very
good (corresponding to the closest semantic variable).
Although the criteria for acceptability of risk is appropriate and very good there is
room for improvements, namely implement new safety barriers and also improve
some additional factors (e.g. improve some aspects related to training).

6 Conclusion and Further Work

Fuzzy approaches for human-centred problems seem to be quite flexible, hence in this
work we introduce a preliminary version of a qualitative method for RAH&S, in the
construction industry, using fuzzy logic concepts and techniques. From all pointed
limitations and inadequacies of traditional RAH&S methods it seems that using this
type of fuzzy approach, to evaluate work safety factors, yields a more realistic repre-
sentation and solution for evaluation of risks in the construction industry.
The outlined qualitative model (QRAM) is still under development. It needs to be
further improved and refined, such as identifying other types of hazards and further
characterize all risks involved, as well as to set criteria for the tolerability of risk
analysis in construction sites. After these improvements it needs to be tested and vali-
dated with real case studies to assess its suitability and generality for the construction
sector.

References
1. Azadeh, A., Fam, I.M., Khoshnoud, M., Khoshnoud, M.: Design and implementation of a
fuzzy expert system for performance assessment of an integrated health, safety, environ-
ment (HSE) and ergonomics system: The case of a gas refinery. Information Sciences 178,
4280–4300 (2008)
 Qualitative Model for Risk Assessment in Construction Industry 111

2. Cornell, M.E.P.: Uncertainties in risk analysis: Six levels of treatment. Reliability Engi-
neering and System Safety 54, 95–111 (1996)
3. Dedobbeleer, N., e Beland, F.: A Safety Climate Measure for Construction Sites. Journal
of Safety Research 22, 97–103 (1991)
4. Faber, M.H., Stewart, M.G.: Risk assessment for civil engineering facilities: critical over-
view and discussion. Reliability Engineering and System Safety 80, 173–184 (2003)
5. Gurcanli, G.E., Mungen, U.: An occupational safety risk analysis method at construction
sites using fuzzy sets. International Journal of Industrial Ergonomics 39, 371–387 (2009)
6. Hollnagel, E.: Accidents and barriers. In: Proceedings of the European Conference on
Cognitive Science Approaches to Process Control, CSAPC (1999)
7. Karwowski, W., Mital, A.: Potential Applications of Fuzzy Sets in Industrial Safety Engi-
neering. Fuzzy Sets and Systems 19, 105–120 (1986)
8. Kentel, E., Aral, M.M.: Probabilistic-fuzzy health risk modeling. Stoch Envir. Res. and
Risk Ass. 18, 324–338 (2004)
9. Laitinen, H., Marjamaki, M., e Paivarinta, K.: The validity of the TR safety observation
method on building construction. Accident Analysis and Prevention 31, 463–472 (1999)
10. Liu, J., et al.: Fuzzy Rule-Based Evidential Reasoning Approach for Safety Analysis. In-
ternational Journal of General Systems 33(2-3), 183–204 (2004)
11. Loosemore, M., e Lee, P.: Communication problems with ethnic minorities in construction
industry. International Journal of Project Management 20, 517–524 (2001)
12. Mure, S., Demichela, M., Piccinini, N.: Assessment of the risk of occupational accidents
using a fuzzy approach. Cogn. Tech. Work 8, 103–112 (2006)
13. Nilsen, T., Aven, T.: Models and model uncertainty in the context of risk analysis. Reli-
ability Engineering and System Safety 79, 309–317 (2003)
14. Pender, S.: Managing incomplete knowledge: Why risk management is not sufficient. In-
ternational Journal of Project Management 19, 79–87 (2001)
15. Ringdahl, L.H.: Safety Analysis principles and pratice in occupational safety, 2nd edn.
Taylor & Francis, London (2001)
16. Ringen, K., Englund, A., Welch, L., Weeks, J.L., Seegal, J.L.: Why construction is differ-
ent. Occupational Medicine 10, 255–259 (1995)
17. Ru, W.G., Eloff, J.H.P.: Risk analysis modelling with the use of fuzzy logic. Computers &
Security 15(3), 239–248 (1996)
18. Sii, H.S., Wang, J., Ruxton, T.: Novel risk assessment techniques for maritime safety man-
agement system. International Journal of Quality & Reliability Management 18(8/9), 982–
999 (2001)
19. Tam, C.M., Zeng, S.X.E., Deng, Z.M.: Identifying elements of poor construction safety
management in China. Safety Science 42, 569–586 (2004)
20. Tixier, J., et al.: Review of 62 risk analysis methodologies of industrial plants. Journal of
Loss Prevention in the Process Industries 15, 291–303 (2002)
21. Wang, J., Ruxton, T.: A review of safety analysis methods applied to the design process of
large engineering products. J. Eng. Design 8(2), 131–152 (1997)
22. Zadeh, L.A.: Linguistic variables, approximate reasoning and dispositions. Med. In-
form. 8(3), 173–186 (1983)
23. Zadeh, L.A.: The concept of a linguistic variable and its application to approximate reason-
ing-part I. Information Sciences 8, 199–249 (1975)
24. Zadeh, L.A.: Fuzzy sets. Information and Control 8, 338–353 (1965)
25. Zimmerman, H.J.: Fuzzy Set Theory and its Applications, 2nd edn., pp. 36–43. Kluwer
Academic Publishers, London (1991)

View publication stats