Safety Science 70 (2014) 387–396

Contents lists available at ScienceDirect

Safety Science

journal homepage: www.elsevier .com/locate /ssc i

Fuzzy logic-based user interface design for risk assessment consideringhuman factor: A case study for high-voltage cell

http://dx.doi.org/10.1016/j.ssci.2014.07.0130925-7535/� 2014 Elsevier Ltd. All rights reserved.

⇑ Corresponding author. Tel.: +90 5335262596; fax: +90 3805240099.E-mail addresses: faruk.aras@kocaeli.edu.tr (F. Aras), karakas@kocaeli.edu.tr

(E. Karakas�), yunusbicen@duzce.edu.tr, yunusbicen@gmail.com (Y. Biçen).

Faruk Aras a, Ercüment Karakas� b, Yunus Biçen c,⇑a Aircraft Electrical and Electronics Department, Kocaeli University, 41285 _Izmit, Kocaeli, Turkeyb Department of Electrical Education, Kocaeli University, 41380 Umuttepe, Kocaeli, Turkeyc Department of Industrial Electronics, Duzce University, 81010 Uzunmustafa, Duzce, Turkey

a r t i c l e i n f o

Article history:Received 2 October 2013Received in revised form 20 May 2014Accepted 13 July 2014

Keywords:Fuzzy logicHuman factorHigh voltageRisk assessmentInterface

a b s t r a c t

This paper presents a novel risk assessment model considering human factor based on the fuzzy logicapproach. For the contribution of the literature, not only the number of people is included in the processof risk assessment, but also with the human factor as a quantitative entry in this study. A flexible anduser-friendly risk assessment interface is developed using LabVIEW program, which puts at disposal dif-ferent applications for the course material. Designed interface gives an opportunity to users to assessrisks in a wide range of consequences containing many different combinations and options. The interfaceis tested for a 100-kV high-voltage cell as a case study. As a result, it is seen that the interface assessesplenty of input elements and possibilities in a short time. For this reason, the fuzzy logic approach is sug-gested as a suitable method for risk assessment.

� 2014 Elsevier Ltd. All rights reserved.

1. Introduction

In the last two decades, there has been a dramatic increasing ofhuman contribution to accident development, reaching levels ofpercentages of as high as 70–80%, independently of the technolog-ical domain of application. There are two main reasons for such rel-evant increasing, namely: (a) the very high reliability andrefinement of mechanical and electronic components; and (b) thecomplexity of the system and the role assigned to human operatorin the control loop (Cacciabue, 2000). It is now widely acceptedthat the majority of accidents in industry generally are in someway attributable to human as well as technical factors in the sensethat actions by people caused to accidents, or people could haveacted better to avert them (Jon Espen and Jan Erik, 2011).

In general, the term ‘‘human factor’’ is used to describe accidentcausality when cause is attributed to the characteristics or behav-ior of an individual or organization, rather than structural ormechanical failure or some environmental or other contextual fac-tors that are outside our control. ‘‘Human errors,’’ on the otherhand, are the mistakes people make often resulting from thesehuman factors (Elise and Sierra, 2006). Human factors may referto various traits or ‘‘elements of the human’’ as individuals, which

should be considered for safe and effective results from engineeredsystems. Or, the term may mean the applied science technologyrelating fundamental human sciences (like anatomy, physiology,neuro-psychology) to industrial systems (Cadick et al., 2006).

Despite the growing awareness of the significance of humanfactors in safety, particularly major accident safety, the focus ofmany sites is almost exclusively on engineering and hardwareaspects, at the expense of ‘people’ issues (Health and SafetyExecutive, 2005). Careful consideration of human factors at workcan reduce the number of accidents and cases of occupational ill-health. It can also pay dividends in terms of a more efficient andeffective workforce (Health and Safety Executive, 2009). It isimportant to decide if the risks vary due to human influences.For example, there is a higher likelihood of human error between2.00 and 5.00 am when physiology dictates that the human bodyshould be asleep. The risks will also be influenced by how well-trained people are, whether they have had sufficient rest beforestarting a shift, and whether they have taken alcohol or used drugs.You may find useful information in your company’s own accidentreports and analyses (Health and Safety Executive, 2009).

The process of risk analysis and assessment does not includepredefined definite steps. Risk assessment is an evaluation of thoselikelihoods and consequences. A risk assessment can either bequalitative or quantitative, although the emphasis in the systemsafety process typically is on qualitative risk assessment (Hardy,2010). Risk factors and assessments are difficult to describe

388 F. Aras et al. / Safety Science 70 (2014) 387–396

mathematically. However, if you can describe a system risk assess-ment qualitatively, you can use fuzzy logic. The fuzzy system canserve as a useful tool for risk analysis to consider organizationaland human factors so as to enhance their study and highlight theuncertainty related to human performance variability(Kirytopoulos et al., 2014). The advantages of fuzzy logic controlinclude the integration of human expertise, experience and knowl-edge into the rule base which has qualitative, descriptive and lin-guistic quantities (Wang and Liu, 2001). Although many studiesusing various methods for risk assessment are available in the lit-erature, it is too difficult to compare them in terms of concludedrisk assessments for various systems using different models. Forthis reason, some risk assessment standards have been developed,and are summarized in the study (Rouhiainen and Gunnerhed,2002; Hale et al., 1990).

Statistics and probabilistic approaches are based on two quali-ties, frequency and severities, which are mostly, applied in riskassessment studies (Rouhiainen and Gunnerhed, 2002; Haleet al., 1990; Cuny and Lejeune, 2003; John Garrick and Robert,2002). However, these models are subjective because availabilityof objective data is very rare and inadequate for risk assessment.Therefore, subjective judgment emerges as a consequence ofassessment. If there are no prior data about the system or the sys-tem has been installed recently, risk can only be assessed in light ofinformation given by the experts who are aware of the possiblehazards.

On the other hand, in the fuzzy logic method, qualitative andquantitative risk methodologies are combined and the structurebecomes more flexible. Thus, the risk rate can be stated by bothnumerical values as in the qualitative risk analysis and definitionsas in the quantitative risk analysis in the fuzzy logic approach. Bythis means, the risk rate can be determined using many inputs suchas possibility of the hazard, frequency of the exposure, and degreesof possible harm. In addition, it can easily be applied to any compli-cated system by means of changing the rule base. The fuzzy logicmethod can also incorporate expert human judgment to definethose variables and their relationships. Thus, it can be closer to real-ity and can be site specific as compared to some of the other meth-ods. For this reason, the fuzzy logic is getting increasingly popularfor risk assessment nowadays. Various applications have been car-ried out recently. Sii et al. (2001) have developed a security modelrelated to marine environment and marine security systems usingthe fuzzy logic approach. The developed model gives out moreeffective results compared to previous risk models. Nieto-Moroteand Ruz-Vila, 2011 have presented risk assessment based on thetheory of fuzzy set indicating that fuzzy logic is used as an effectiveanalyzing tool in the case of excessive amounts of risky situations inthe analytic hierarchy process (AHP). A fuzzy logic-based riskassessment tool has been developed to assess the risk of river-basedhydroelectric power plant projects by Kucukali (2011). Fuzzy logicmethodology enables multi-criterion decision analysis and pro-vides an easy and understandable way to analyze the possible risksthat emerge in the projects. Bajpai et al. (2010) developed a methodin which two linguistic fuzzy scales are used at the base of trapezoi-dal fuzzy numbers in the modification of the early developedsecurity risk factor table (SRFT) model by using the concept of fuzzylogic. This method was tested at a refinery and compared to for-merly used methods so that it could be explained. Cho et al.(2002) emphasize that conventional risk assessments involveambiguities. In order to get rid of these ambiguities, a new methodis suggested to assess risks more securely by using fuzzy concepts.Cho et al. present new forms of fuzzy membership curves as well.Markowski et al. (2011) have indicated that workers happen to bea potentially risk group in an explosive environment and theirsafety and health conditions are based on regulations publishedby ANSI/AIHA in the United States and ATEX in the European Union.

They emphasize that risk assessment is a must for ATEX but is notso for ANSI/AIHA, and state that in order to assess the risks, theassessment must come into existence in a semi-quantitative explo-sion layer of protection analysis (ExLOPA) that was implementedfor the purpose of developing a standard method that did not earlierexist. Hu et al. (2007) developed a methodology named formalsafety assessment (FSA) in order to increase marine security. Quan-titative risk assessment and a comprehensive modeling of possiblerisk, along with the extent of frequency and severity especially inthe navigation of seagoing vessel, were accomplished by analysesas a result of FSA approach as well. Hadjimichael (2009) not onlydeveloped a risk modeling methodology capable of stating the riskfactors by using fuzzy expert systems named, The Flight OperationsRisk Assessment System (FORAS), but also estimated cumulativeeffects of possible dangers in single-flight operations by using aquantitative relative risk index defined by the FORAS risk model.Li et al. (2008) developed a new risk assessment method that couldassess the possible risk at a power system by determining the out-age of power system components using a hybrid model consistingof fuzzy clumps and Monte Carlo simulation. Elsayed (2009) accom-plished a multiple-attribute risk assessment by using a fuzzy infer-ence system based on the usage of fuzzy clumps, rule base, andfuzzy inference engine. The suggested method was designed forseagoing vessel-operating modes as open sea and/or port input/output transit and was tested at a terminal during the loading–unloading process of a natural gas-loaded seagoing vessel.Lavasani et al. (2011) assessed the risk of hazards using a basic riskitem (BRI) composed of fuzzy numbers because of the emphasizedreason that obtained data cause uncertainty in risk assessmentowing to complex and incomprehensible hazard mechanisms. Inthis study, a flexible and applicable risk assessment interface isdeveloped using the fuzzy logic method involving the human fac-tor. While in the classic methods the human factor is generallyadded to risk assessment as a multiplier, in the suggested approach,behavior attributes, as well as the number of persons, are also con-sidered. Thus, elements originating from human behaviors that arelikely to affect the possibilities of occurrence of dangers to animportant degree are evaluated via the fuzzy logic approach.

2. Model description

Inelastic conventional methods are not suitable for dialectics dueto the fact that an object either belongs to a clump or not, whichmeans that the underlying logic is 1 or 0 and which can be statedby certain expressions as open-closed or hot–cold. Fuzzy logic is amathematical method of processing uncertain and vague data. Byusing the basic properties and operations defined for fuzzy sets,any compound rule structure may be decomposed and reduced toa number of simple canonical rules (Ross, 2004; Morari et al., 2010).

In this study, the rules of fuzzy logic risk assessment are desig-nated by availing the PILZ method; however, the boundary condi-tions are designated by taking advantage of the smooth passing offuzzy logic unlike PILZ method that reduces the number of rulebase.

Two main assessment units constituted as hazard and humanfactors include three inputs each that are likelihood (LO), fre-quency of exposure (FE), degree of possible harm (DPH), and fati-gue and attention deficit (F/AD), stress (S), technical competence(CW/II), respectively, as seen in Fig. 1. Interim values are carriedout with the fuzzy logic-based interface by reason of the fact thatthe number of linguistic labels belonging to inputs of the two mainassessment units was reduced compared to conventional methodsas variable gaps increased.

Membership functions of the inputs are seen in Fig. 2. Sharptransitions between the linguistic inputs labels are eliminated as

Fuzzy Block

LO

FE

DPHHRN

Fuzzy Block

F/AD

S

CW/IICHF

x

NP

x THRN

FUZZY RULE BASED HAZARD FACTOR

FUZZY RULE BASED HUMAN FACTOR

Fig. 1. Structure of fuzzy logic risk assessment.

Fig. 2. (a) Membership functions of hazard factor for likelihood of occurrence (LO). (b) Membership functions of hazard factor for frequency of exposure (FE). (c) Membershipfunctions of hazard factor for degree of possible harm (DPH).

F. Aras et al. / Safety Science 70 (2014) 387–396 389

can be seen from the figure. For example, in Fig. 2(a), linguisticlabels are highly unlikely, possible, probable, and certain. Althoughthe given value of 15 for a certain linguistic label is a full member-ship value, linguistic label members of both probable and certainlinguistic labels lie between 12 and 13. These variable ranges areattained by program.

In order to notify the situation the people are in, human factorlinguistic labels are set in three different agents and three differentranges as good, normal, and bad. Resolution of membership func-tions is increased as these ranges vary between 1 and 3. For thisreason, membership functions were constituted by the programfor stress in Fig. 3(a); for fatigue, attention, and deficit in

Fig. 3. (a) Membership functions of human factor for fatigue, attention deficit (F/AD). (b) Membership functions of human factor for Stress (S). (c) Membership functions ofhuman factor for Technical Competence (CW/II).

390 F. Aras et al. / Safety Science 70 (2014) 387–396

Fig. 3(b); and for technical competence as triangle in Fig. 3(c). Thisstructure in which human factor is involved is an innovation of themodel. Risk is calculated by assessing the human factor as just amultiplier together with the factors in machine or environmentin the literature generally. Agents changing the risk emergencepossibility as humane features such as stress condition, technicalcompliance, and physical fitness for work, sleeplessness, fatigue,and inattentiveness are evaluated in this model apart from thenumber of humans.

Fig. 4 shows rule-based windows scripted in fuzzy design forhazard and human factor used in the risk assessment phase. Rela-tions between input and output are enabled via the rules scriptedin this module. LO, FE, and DPH are the input variables and HRN isthe output variable for the hazard factor in the model shown as anexample below:

IF LO is (highly unlikely) and FE is (annually) and DPH is (lacera-tion) THEN HRN is (negligible).

Depending on the values taken by the LO, FE, and DPH inputs,the fuzzy value of the HRN output is designated in this rule.Thirty-six rules can be scripted in the model involving three inputsin total for the hazard factor. Rules are correlated with eachother as providing all probabilities between inputs and outputs.Similarly, F/AD, S, and CW/II are input variables and CHF is outputvariable for human factor in the model shown as an examplebelow:

IF F/AD is (good) and S is (good) and CW/II is (good) THEN CHF is(low).

Thus, the fuzzy value of the CHF output is designated dependingon the F/AD, S, and CW/II values. Twenty-seven rules can bescripted in the model involving three inputs in total for the humanfactor.

Fig. 5 presents outputs of the model given as the hazard factor(HRN) and the human factor (CHF). Fig. 5(a) presents a bell shapeused for the hazard factor membership function. Linguistic labelsare determined as negligible, low significant, high significant, andunacceptable, respectively. For example, the values determinedbetween 40 and 150 are full members to high significant linguisticlabel. The trapezoidal shape was chosen as the membership func-tion for the human factor as seen in Fig. 5(b). Linguistic labels aredetermined as low, medium, and high, respectively. Values deter-mined between 2.25 and 2.75 are full members of the medium lin-guistic label as an example for linguistic label scales. Others aredetermined by fuzzy logic as interim values.

Then, main model outputs, human factor (CHF) and hazard fac-tor (HRN) are defuzzificated. For this purpose, the fuzzy logicmethod uses the following equations to calculate a weighted aver-age of the geometric center of area for all membership functions.

On the brink of U = {u1, . . .,u‘};

HRN ¼P‘

i¼1ui � lHRNðuiÞP‘

i¼1lHRNðuiÞð1Þ

Fig. 4. Rule-based windows scripted in fuzzy design for hazard and human factor.

F. Aras et al. / Safety Science 70 (2014) 387–396 391

CHF ¼P‘

i¼1ui � lCHFðuiÞP‘

i¼1lCHFðuiÞð2Þ

On the other hand, total risk (THRN) is found by multiplyingCHF and HRN outputs obtained from the equalities above by thenumber of people (NP). As ranges of number of persons (NP) needto be integers, they were added to the assessment as a multiplier inthe program as given in Table 1.

The inference surfaces in 3D for the three fuzzy logic subsys-tems for outputs HRN and CHF are given in Figs. 6 and 7.

Fig. 5. (a) Output of hazard factor (HRN)

3. User interface and case study

The interface of the system to be executed scripted by LabVIEWPackage Software is shown in Fig. 8. The program offers a numberof different combinations and options to users for risk assessment.States of distress concerning the desired sectors can easily beadded to the program except the options determined as ‘‘default.’’

In every stage, users can benefit from the HELP menu or guid-ance text in windows. The program includes two main stages inwindows such as the selection module and the assessment module.

. (b) Output of human factor (CHF).

Table 1Ranges of number of persons (NP).

Factor NP (number of persons under risk)

1 1–2 persons2 3–7 persons4 8–15 persons8 16–50 persons

12 50+ persons

392 F. Aras et al. / Safety Science 70 (2014) 387–396

The first stage is the selection module which places several optionsabout the definition of hazards and human factors.

Boundary gaps for hazards and human factors can be deter-mined visually by users through the instrument of setting knobsstructured in their own fields in the interface. Ten risk groups arepossible to take in and out on demand, including a considerablenumber of hazard subgroups generated as main titles in that phase.This phase consists of prepared questions sampled as seen below,thus any risk can easily be identified by the user. It is possible toadd new windows and questions. Humane situations are possibleto be set from good to bad in the human factor section.

In this section, with a bar graph showing the relative size, haz-ard is involved with the number of the exposed people selected.Another feature of the interface is the section that assessmentresults related to the hazard question under each category scaledlogarithmically at the bottom of the window as a bar graph. Riskdegree is expressed by a chromatic spectrum in order to be under-stood visually better.

3.1. Examples of hazard definition questions:

3.1.1. Workplace environment

1. Is there neglected use of personal protective equipment orinsufficient safety signs?

2. Is it possible for hazards to be generated by neglecting ergo-nomic principles in machinery design?

3. Are there thermal hazards resulting in burns, scalds, and otherinjuries by a possible contact of persons with objects or materi-als with extremely high or low temperature, flame or explosion,or radiation from heat source?

4. Are there hazards generated by vibration or noise resultingfrom use of handheld machines, resulting from a variety of neu-rological and vascular diseases or hearing loss (deafness), orother physiological disorder (e.g. loss of balance, loss ofawareness)?

5. Is there inadequate local lighting?

3.1.2. Electrical

1. Is it possible to approach live parts under high voltage?2. Are you sure that safety devices and switches are in place and

that they work?3. Is it possible to work in dangerous proximity to electrical

systems?4. Are there any exposed conductive parts not connected to the

grounding system?5. Are there any electrostatic charges (such as when refueling)?

3.1.3. Mechanical

1. Are there machine parts or work pieces (mass and stability –potential energy or mass and velocity – kinetic energy ofelements)?

2. Are all machine guards secured firmly and not easilyremovable?

3. Are there any unguarded gears, sprockets, pulleys, or flywheels?4. Is it easy for an operator to reach ON/OFF controls?

3.1.4. Hazardous substances

1. Are there hazards from contact with or inhalation of harmfulfluids, gases, mists, fumes, or dust?

2. Is there any risk tending or serving to explode or characterizedby explosion or sudden outburst of an ‘‘an explosive device,’’‘‘explosive gas,’’ ‘‘explosive force,’’ ‘‘explosive violence,’’ or ‘‘anexplosive temper’’?

3. Is there any toxic substance containing poison or somethingharmful to the body that is characterized by a diffuse macularerythema, vomiting, diarrhea, severe myalgia, or shock?

4. Are there any machinery parts having high temperature or highpressure that can harm humans?

3.1.5. Fire

1. Are oxidizing or flammable substances, such as paint, finishes,adhesives, and solvents used?

2. Are there any sources of ignition (e. g., open fire, electricalequipment, electrostatic charges, or high temperature)?

3. Are fire-fighting equipments such as fire detector, gas sensor,fire extinguisher in place and are they suitable (serviceable, ser-viced and regularly easily accessible)?

4. Are there emergency alarms and escape plans, and is fire-fight-ing training provided?

The second phase is the window that obtained results presentednot only as a numerical but also as a chromatic spectrum, as seenin Fig. 9.

A high-voltage test cell is designed as a course material forstudents to recognize electrode systems and to test solid–liquidinsulators using the high-voltage technique in the laboratory(Aras and Keles, 2007). The system is also used for some academicstudies in the electrical department. The system is active twice aweek for two hours on course days for a number of attendingstudents ranging between 8 and 15. The tests are performed underhigh voltage and high temperatures in oil or air. The most signifi-cant hazard that might result in death or serious injury in the sys-tem is electric shock due to working under high voltage. Besidesthis, other serious hazards are fire due to electric arc and seriousburns owing to the flammable scorching oil if it comes into contactduring the course of experimentation.

A low probability of the occurrence of hazards, such as jam-ming, failure, impact due to system mechanics in placing or dis-placing the specimens into the test cell, might result inscorching; moderate damage can also result.

Environmental hazards, such as rising of ambient temperaturedue to high heat, gas and fumes during extraction of oil, slipperyground due to spillage of oil, are considered other hazardous situ-ations with dangerous results.

The most important hazards are the unauthorized interventionand inattentiveness as human factors in the system. In addition,uncontrolled behavior, stress, and technical incompetence areother considered human factors in the system. There is no occupa-tional accident background in the system. Improper practicesmight be performed by forgetting the provided information, whichincludes required occupational safety training and operatinginstructions. Uncontrolled behaviors are probable because of theexistence of students apart from expert crew. Unclamping the sys-tem gate in the course of experiment is a rare behavior, amongothers.

There is a strong probability of an accident in the system, espe-cially electric shock due to human error, which is considered very

Fig. 6. 3D inference surfaces for HRN.

Fig. 7. 3D inference surfaces for CHF.

Fig. 8. Risk assessment user interface.

F. Aras et al. / Safety Science 70 (2014) 387–396 393

Fig. 9. Result window of the risk assessment program.

Fig. 10. Risk results after application.

394 F. Aras et al. / Safety Science 70 (2014) 387–396

serious. For this reason, even the slightest possibility of hazard isassessed. Personal protective equipment is also an insecure behav-ior among human factors and is assessed as a condition that canresult in serious injury or burns. The existence of risk increases sig-nificantly when considering human factor as a result of assess-ment. For example, one of the electrical hazards is the possibilityof working in dangerous proximity to electrical systems.

The risk assessment of this hazard is determined as ‘‘high’’ bythe program without taking into consideration the human factor.If human factors such as restlessness, distractibility, stress, andtechnical incompetence are taken into consideration as a ‘‘bad’’mark, then the result is marked as ‘‘extreme.’’ A similar situationoccurred in all hazards regardless of the amount of people. Becausethis situation would cause more specified risks to rise, significant

Fig. A. (a) Control and preservation unit of electrode system, (b) realized electrode system.

F. Aras et al. / Safety Science 70 (2014) 387–396 395

measures need to be taken. For this purpose, the following precau-tions were taken for the sample application.

� Automatic overall energy cut off feature was provided by inte-grating a limit switch to the low-voltage side containing thevariable transformer.� During the test, against the probability of opening the high-

voltage cell cover because of absent-mindedness or careless-ness, the security switch was placed on its cover. Thus, whenopening the cover, power to the high-voltage transformer wascut off and on the side of low voltage; by rapidly dropping to‘‘zero’’ the voltage level of auto-transformer, circuit voltagewas interrupted.� A cut out switch was integrated to cut off the overall system

energy automatically at the end of experiment in case someoneforgot to shut it down.� An additional precaution is grounding. The grounding rod

must be connected to the system before the beginning ofthe experiment. Otherwise, the system will not work. Theoutput of the high-voltage transformer was grounded via acontact till to the system restart if the system voltage is zeroat the end of experiment. In order to control the systemgrounding continuously, an instruction manual was preparedfor the staff.� The heater turns off automatically at the end of the

experiment as a precaution for temperature hazards. WarningLEDs glow at high-temperature conditions and all actions arepassive in order to prevent intervention if the cooling require-ments are not satisfied even if any experiment was not inprogress.� As a measure to detect fire hazard in the system, smoke detec-

tors, fire alarms, and halocarbon portable fire extinguisherswere made available.� An instruction manual regarding the usage of personal protec-

tive equipment and fire hazard was prepared.� The cell is confined with a PVC-coated metal barrier to prevent

unauthorized access in addition to all precautions.

Risks are significantly reduced as can be seen in Fig. 10.Negative case of human factor comes up as a great factor as oneof the results for the risk increment when risk factors are low inthe system. Positive or negative case of human factor can varythe total of risk factor in a wider range in the model as well. Forexample, if the system risk factor is high and the human factor isnegative, the total risk (THRN) will increase even more.

4. Conclusion

In the developed program of this study, regardless of the num-ber of people, the human factor has been included in the process ofrisk assessment. The designed interface gives an opportunity tousers to assess risks in a wide range that contains many differentcombinations and options. The interface includes flexible, user-friendly, and useful modules for users. Future work has beenplanned to make the interface a suitable web media in order toprovide easier access.

Appendix A

The experiments on solid/liquid insulation materials have beenconducted by student groups under the control of a supervisor (seeFig. A).

References

Aras, F., Keles, K., 2007. Laboratory experiments for electrical insulation coursesusing a new high-voltage cell. Int. J. Electr. Eng. Ed. 44 (1), 84–94.

Bajpai, S., Sachdeva, A., Gupta, J.P., 2010. Security risk assessment: applying theconcepts of fuzzy logic. J. Hazard. Mater. 173, 258–264.

Cacciabue, P.C., 2000. Human factors impact on risk analysis of complex systems. J.Hazard. Mater. 71 (1), 101–116.

Cadick, J., Capelli-Schellpfeffer, M., Neitzel, D.K., 2006. Electrical Safety: Handbook.Cho, H.-N., Choi, H.-H., Kim, Y.-B., 2002. A risk assessment methodology for

incorporating uncertainties using fuzzy concepts. Reliability Eng. Syst. Saf. 78,173–183.

Cuny, X., Lejeune, M., 2003. Statistical modeling and risk assessment. Saf. Sci. 41 (1),29–51.

Elise, D.C., Sierra, F., 2006. An Assessment of the Role of Human factors in Oil Spillsfrom Vessels, Nuka Research & Planning Group, LLC, PWSRCAC-Report.

Elsayed, T., 2009. Fuzzy inference system for the risk assessment of liquefied naturalgas carriers during loading/offloading at terminals. Appl. Ocean Res. 31, 170–185.

Hadjimichael, M., 2009. A fuzzy expert system for aviation risk assessment. ExpertSyst. Appl. 36, 6512–6519.

Hale, A.R., de Loor, M., van Drimmelen, D., Huppes, G., 1990. Safety standards, riskanalysis and decision making on prevention measures: implications of somerecent European legislation and standards. J. Occupational Accidents 13 (3),213–231.

Hardy, T., 2010. The Role of Human Factors in Safety Risk Assessment, Great CircleAnalytics LLC, pp. 1–10.

Health and Safety Executive (HSE), 2005. Inspectors Toolkit, Human factors in themanagement of major accident hazards.

Health and Safety Executive, 2009. Reducing error and influencing behaviour.Hu, S., Fang, Q., Xia, H., Xi, Y., 2007. Formal safety assessment based on relative risks

model in ship navigation. Reliability Eng. Syst. Saf. 92, 369–377.John Garrick, B., Robert, F.C., 2002. Probabilistic risk assessment practices in the USA

for nuclear power plants, Safety Science, vol. 40, issues 1–4, February–June2002, pp. 177–201.

396 F. Aras et al. / Safety Science 70 (2014) 387–396

Jon Espen, S., Jan Erik, V., 2011. Quantitative risk analysis offshore—human andorganizational factors. Reliability Eng. Syst. Saf. 96 (4), 468–479.

Kirytopoulos, K., Konstandinidou, M., Nivolianitou, Z., Kazaras, K., Embedding thehuman factor in road tunnel risk analysis. Process Safety and EnvironmentalProtection. <http://dx.doi.org/10.1016/j.psep.2014.03.006> 27 March 2014.

Kucukali, S., 2011. Risk assessment of river-type hydropower plants using fuzzylogic approach. Energy Policy 39, 6683–6688.

Lavasani, S.M.M., Yang, Z., Finlay, J., Wang, J., 2011. Fuzzy risk assessment of oil andgas offshore wells. Process Saf. Environ. Prot. 89, 277–294.

Li, W., Zhou, J., Xie, K., Xiong, X., 2008. Power systems risk assessment using ahybrid method of fuzzy set and Monte Carlo simulation. IEEE Trans. Power Syst.23 (2), 336–343.

Markowski, A.S., Mannan, M.S., Kotynia, A., Pawlak, H., 2011. Application of fuzzylogic to explosion risk assessment. J. Loss Prev. Process Ind. 24 (6), 780–790.

Morari, M., et al., 2010. Fuzzy-control of a model helicopter, Swiss Federal Instituteof Technology (ETH) Zurich. Lab Manual, Page last modified on September 16.

Nieto-Morote, A., Ruz-Vila, F., 2011. A fuzzy approach to construction project riskassessment. Int. J. Project Manage. 29, 220–231.

Ross, T.J., 2004. Fuzzy Logic with Engineering Applications, second ed. Wiley, WestSussex.

Rouhiainen, V., Gunnerhed, M., 2002. Development of international riskanalysis standards, Safety Science, vol. 40, issues 1–4, February–June 2002,pp. 57–67.

Sii, H.S., Ruxton, T., Wang, J., 2001. A fuzzy-logic-based approach to qualitativesafety modeling for marine systems. Reliability Eng. Syst. Saf. 73, 19–34.

Wang, X.G., Liu, W., 2001. A Fuzzy Fault Diagnosis Scheme with Application, IFSAWorld Congress and 20th NAFIPS Int. Conference, vol. 3, pp. 1489–1493.