A Novel Approach to Risk Assessment for Occupational Health and

Safety Using Pythagorean Fuzzy AHP & Fuzzy Inference System

Esra Ilbahara,c , Ali Karaşanb,c , Selcuk Cebia,∗, Cengiz Kahramanc
a
Department of Industrial Engineering, Yildiz Technical University, Besiktas 34349, Istanbul, Turkey
b
Graduate School of Natural and Applied Sciences, Yildiz Technical University, Besiktas 34349, Istanbul,
Turkey
c
Department of Industrial Engineering, Istanbul Technical University, Macka 34367, Istanbul, Turkey

Abstract
Occupational health and safety involves systematic studies aimed at protecting employees
from harmful conditions that might be caused by various reasons during the execution of
work in the workplace. Different from the literature, in this study, a novel integrated ap-
proach, Pythagorean Fuzzy Proportional Risk Assessment (PFPRA), including Fine Kinney,
Pythagorean fuzzy analytic hierarchy process, and a fuzzy inference system is used for risk
assessment in the field of occupational health and safety. The main difference of the pro-
posed approach is the integration of these methods in a way providing a more accurate risk
assessment. The risks of an excavation process in a construction yard are assessed by the
proposed method. The results are compared with Pythagorean Fuzzy Failure Modes and
Effects Analysis (PFFMEA) and it is revealed that the proposed method produces reliable
and informative outcomes better representing the vagueness of decision making process.
Keywords: Pythagorean Fuzzy Sets, AHP, Fuzzy Inference System, FMEA, Fine Kinney

1 1. Introduction
2 Risk analysis is the utilization of the available information systematically to determine
3 hazards whereas risk evaluation involves judgments on the tolerability of the risk with re-
4 spect to some criteria. The whole process of risk analysis and risk evaluation is called risk
5 assessment [1]. Risk assessment techniques are divided into two groups as quantitative and
6 qualitative with respect to the type of parameters used. Techniques aiming at determining
7 the degree of risk based on purely numerical or statistical methods are called quantitative
8 techniques. The methods using parameters that determine the degree of risk based on ob-
9 servations, categorical evaluations, or non-numerical measurements are called qualitative
10 techniques. For example, if the severity parameter is used to determine the risk degree, the
11 relevant method is a qualitative method. Because there is no severity meter to measure

∗
Corresponding author:
Email address: scebi@yildiz.edu.tr (Selcuk Cebi)
Preprint submitted to Safety Science September 21, 2017
12 the severity numerically. If a risk assessment technique calculates the risk only based on
13 the probability parameter and the probability value is statistically determined from its past
14 records, it is a quantitative technique. Quantitative risk analysis techniques can also be
15 utilized by using categorical data. In this case, quantitative risk analysis techniques might
16 be called mixed techniques. The list of quantitative and qualitative risk analysis techniques
17 is given in Table 1 [2].

Table 1: Risk analysis techniques with respect to type of parameters

Quantitative Risk Analysis Techniques Qualitative Risk Analysis Techniques

• Fault Tree Analysis • Check List
• Event Tree Analysis • What if? Analysis
• Cause-Consequence Analysis • Preliminary Risk Analysis
• Management Oversight and Risk Tree • Job Safety Analysis
• Dynamic Event Tree Analysis • L Matrix Method
• Bow-Tie Risk Analysis • X Matrix Method
• 3T Matrix Method
• Fine Kinney Method
• Hazard and Operability Studies
• Failure Modes and Effects Analysis

18 Occupational health and safety (OHS) is described as anticipation, recognition, evalu-

19 ation and control of hazards that could harm the health of workers [3]. In other words,
20 OHS involves systematic studies aimed at protecting employees from harmful conditions
21 that might be caused by various reasons during the execution of work in the workplace.
22 The scope of OHS has expanded depending on social, political, economic and technological
23 factors [3]. Risk assessment techniques used in the field of OHS generally calculate the risk
24 value depending on the probability and severity factors. Since particularly the severity pa-
25 rameter can not be measured objectively, it is integrated to the evaluation process with the
26 help of subjective evaluation and categorical data.
27 Likewise, probability values can be evaluated as categorical given that historical data
28 are not available. Categorical data can not produce sensitive results in terms of computa-
29 tional accuracy. In the literature, the Analytic Hierarchy Process (AHP) method has been
30 frequently used to evaluate subjective and categorical parameters. In addition, there are
31 several studies in the literature that assess risk for OHS based on the AHP method. How-
32 ever, in the literature, the fuzzy set theory is used to take vagueness and impreciseness of
33 subjective evaluations into account.
34 In AHP, factors related to a decision making problem are categorized and consequently
35 form a hierarchy. After the hierarchy is built, linguistic terms are employed by experts
36 to make pairwise comparisons. These linguistic terms are converted to numerical values
37 by using fuzzy sets which are able to handle impreciseness and vagueness of evaluation
38 processes. It is thus possible to say that AHP is quite useful for modelling problems in the
39 absence of certain measures.

2
40 In this study, Pythagorean fuzzy AHP method is employed to make risk assessment
41 techniques work more effectively as different from the literature. Pythagorean fuzzy sets,
42 an extension of intuitionistic fuzzy sets, is developed for the purpose of providing more
43 freedom to experts in expressing their opinions about the vagueness and impreciseness of
44 the considered problem. Pythagorean fuzzy sets achieve this purpose because experts do not
45 have to assign membership and non-membership degrees whose sum is at most 1. However,
46 the sum of squares of these degrees must be at most 1. The weights obtained through
47 Pythagorean fuzzy AHP will be used as inputs for severity and probability parameter in risk
48 assessment techniques.
49 In this study, risk value in terms of OHS will be calculated by using an integrated method.
50 The proposed approach consists of Pythagorean fuzzy AHP, fuzzy inference system, and Fine
51 Kinney method. In the traditional Fine Kinney method, the magnitude of the risk is equal
52 to the scalar multiplication of probability, severity, and frequency parameters which are
53 directly obtained from experts. In the proposed integrated method, probability and severity
54 parameters will be determined by the Pythagorean fuzzy AHP (PFAHP) method. Along
55 with frequency parameter directly obtained from experts, the obtained values for probability
56 and severity parameters will be used as inputs for Fuzzy Inference System (FIS), and it will
57 provide a risk value as an output. Reliability of our integrated method will be discussed by
58 comparing it with the Failure Modes and Effects Analysis (FMEA) method. The FMEA
59 method calculates the risk value based on probability, severity and detectability parameters.
60 Likewise, probability and severity parameters obtained through PFAHP will be used for
61 FMEA.
62 The rest of this paper is organized as follows: A literature review on the combination of
63 AHP-Fine Kinney, and AHP-FIS is given in Section 2. Fine Kinney method, preliminaries
64 on Pythagorean fuzzy sets, steps of PFAHP, and FIS are examined in Section 3. Application
65 of the proposed method and conclusions are presented in Sections 4 and 5, respectively.

66 2. Literature Review
67 Fine Kinney method is a useful quantitative technique to estimate risks. In the Fine
68 Kinney method, probability, severity, and frequency are obtained for each identified risk.
69 Probability is the possibility of damage occurring over time whereas frequency refers to the
70 frequency of exposure to hazard. Severity represents the magnitude of the harm or damage
71 to human, workplace and environment if the hazard occurs. Risk score is calculated by
72 multiplying probability, severity and frequency. Then, whether the situation is acceptable
73 or not is assessed [4]. There was no study in the literature using both AHP and Fine Kinney
74 methods until 2017. Gül et al. [5] utilized a combination of fuzzy AHP, fuzzy VIKOR,
75 and Fine Kinney methods for ballast tank maintenance process. Kokangül et al. [6] used
76 both Fine Kinney and AHP methods to assess the risks in a large manufacturing company.
77 Hazards were prioritized through AHP and also evaluated using the Fine Kinney method.
78 The link between the results of these two methods was analyzed to identify the risk class
79 intervals for AHP.

3
 80 Fang et al. (2003) proposed a framework to select the most appropriate scaffolding for
81 a construction project by using AHP. The factors considered in the study of Fang et al.
82 are initial cost, running cost, safety risks, cost variation, speed of installation, efficiency of
83 other trades, project quality, and corporate image. Aminbakhsh et al.(2013) introduced a
84 framework involving cost of safety model and AHP to prioritize safety risks in construction
85 projects. Risks affecting construction safety are divided into three main groups as acci-
86 dent hazard, physical hazard, and chemical hazard. These main groups are further divided
87 into sub-groups as trips falls, electricity lighting, fire explosions, machinery equipment,
88 vibration, temperature, ventilation, burns, and neurological. Chan et al. (2004) adopted
89 AHP to identify the priority of processes for the Hong Kong construction industry. In this
90 study, AHP structure involves cost implication, development time, expertise required, client
91 requirements, corporate image. Podgrski (2015) utilized AHP to select the main key per-
92 formance indicators for assessing operational performance of occupational safety and health
93 system. The criteria used in this study are specific, measurable, achievable, relevant and
94 time-bound. Zheng et al. (2012) employed AHP to assess the work safety in hot and humid
95 environments. Trapezoidal fuzzy numbers are utilized to deal with uncertainty and impre-
96 cision of the data. The main factors used in this study are work, environment, and workers
97 whereas sub-factors are work nature, work intensity, and work duration, temperature, hu-
98 midity, airflow velocity, and heat radiation intensity, seniority structure, safety training and
99 personal protection. Qiaoxiu et al. (2016) adopted nonlinear fuzzy analytic hierarchy process
100 for risk assessment of a coal mine. Risk factors considered in this study include manage-
101 rial, environmental, operational and individual criteria. The logarithmic fuzzy preference
102 programming method is used for the analysis of the data. Guneri et al. (2015) employed
103 fuzzy AHP to select the best risk assessment method in occupational safety operations for
104 small and medium sized enterprises by taking scope, practicality, cost, and sensitivity cri-
105 teria into consideration. Janackovic et al. (2013) utilized fuzzy AHP to prioritize the main
106 occupational safety indicators of road construction companies. Fera and Macchiaroli (2010)
107 utilized AHP and fire dynamics simulator to evaluate fire safety in tunnels by considering
108 carbon monoxide, oxygen, temperature, and visibility criteria.
109 Topuz and van Gestel [7] developed an environmental risk assessment approach for the
110 usage of engineered nanoparticles by using both AHP and fuzzy inference method, and the
111 proposed method provides the risk class and its membership degree. Nieto-Morote and
112 Ruz-Vila [8] used fuzzy AHP and fuzzy inference for construction project risk assessment.
113 In the study of Rodriguez et al. [9], a new risk assessment method involving fuzzy AHP
114 and FIS was developed. FIS is employed to integrate the groups of risk factors, and then
115 these risk factors are utilized as the evaluation criteria for fuzzy AHP. Yang et al. [10]
116 used a combination of fuzzy AHP and FIS to prioritize environmental issues in offshore oil
117 and gas operations. In [11], FIS, fuzzy AHP and FMEA were used for risk management
118 in the construction industry. In the study of Zeng et al. [12], a risk assessment approach
119 was proposed to handle risks in complicated construction processes. AHP was utilized for
120 the prioritization of risk factors whereas fuzzy based decision making method was adopted
121 for the risk assessment of construction projects. On the other hand, Cebi [13] applied
122 fuzzy multiplication operation on probability and severity parameters to assess the risks in
4
123 construction projects. In [14], railway risk assessment was conducted by using the fuzzy
124 reasoning and fuzzy AHP approaches. In the study of Sabokbar et al. [15], AHP, FIS, and
125 dominance-based rough set approach were utilized to determine risks, examine rule base
126 structure and the impact of risks in a tourism destination. Acuner and Cebi [16] utilized
127 probability and severity parameters together with FIS for the evaluation of occupational
128 accidents at shipyards. Furthermore, a web based decision support system using the same
129 methodology was developed by Cebi et al. [17].
130 Different from the literature, an integrated method including Fine Kinney , PFAHP,
131 and FIS has been proposed in this study. The main difference of the proposed approach is
132 the integration of these methods in a way that facilitating a more accurate risk assessment.
133 In the proposed approach, instead of assigning one distinct number for severity parameter,
134 experts employ linguistic terms and Pythagorean fuzzy sets which provide larger freedom to
135 experts for their choices. Moreover, probability and severity values obtained from PFAHP
136 are not subjected to direct scalar multiplication as in the traditional Fine Kinney method.
137 Instead, FIS is utilized in the process of obtaining risk values of the potential hazards to
138 eliminate the high sensitivity of scalar multiplication to its multipliers (i.e. parameters).

139 3. Methodology
140 In this section, the methods integrated in the proposed approach, Fine Kinney, PFAHP
141 and FIS, are initially examined. Then, the steps of the proposed approach, Pythagorean
142 Fuzzy Proportional Risk Assessment (PFPRA), are described in detail.

143 3.1. Fine Kinney Method

144 Fine Kinney method which is also known as the Proportional Risk Assessment Technique
145 (PRAT) [18] calculates risk magnitude by using a proportional formula [19]. The formula
146 is based on exposure (frequencies of the actions), probability of the accident, and potential
147 consequences of an accident (severity). The risk magnitude is obtained by product of these
148 factors. The formula is as follows [18]:

RM = P × S × F (1)
149 where RM, P, S, and F are risk magnitude, probability, severity, and frequency (or
150 the exposure) of the event, respectively. The validity of the obtained results is based on the
151 decision on P, S and F factors. These values are estimated by the collected data, observation
152 of the work process, and discussion with the employees on their working activities [19].
153 These values are categorical and their numerical magnitudes are selected by standard tables
154 as given in Tables 2-4.
155 In Table 2, the linguistic scale for frequency of occurrence is presented [18]. In the
156 linguistic scale, six frequency levels are utilized and the selection from these levels has been
157 made by using experience of expert or observation. Then, Table 3 presents the linguistic
158 terms for the probability or likelihood parameter [18].

5
 Table 2: Linguistic scale for frequency of occurrence

Linguistic Terms
Description Very rarely Rarely Unusually Occasionally Frequently Continuously
Rating 0.5 1 2 3 6 10

Table 3: Linguistic scale for probability

Description Rating
Practically impossible 0.1
Extremely remote 0.5
Remotely possible 1
Unusual 3
Quite possible 6
Most likely 10

159 Based on injuries and damages, the consequence of any hazard or accident (i.e. severity)
160 changes. The methodology considers the consequence of any hazard by using linguistic scale
161 given in Table 4 [18].

Table 4: Linguistic scale for consequence

Description Rating
Minor cuts 1
Disabling injuries 5
Extremely serious 15
Fatality 25
Multiple fatalities 50
Catastrophe 100

162 The results (risk magnitude-RM) obtained based on probability, frequency, and conse-
163 quence parameters are evaluated by using the linguistic scale given in Table 5 and necessary
164 actions with respect to risk magnitude are determined. In Table 5, risk magnitudes are
165 classified into five categories [20].

Table 5: Linguistic scale for risk magnitude

Risk Level Description Risk Value

Negligible Risk Acceptable risk RM < 20
Minor Risk Monitoring 20 ≤ RM < 70
Medium Risk Measures to be taken 70 ≤ RM < 200
Major Risk Immediate improvement 200 ≤ RM < 400
Critical Immediate action 400 ≤ RM

6
166 3.2. Pythagorean Fuzzy AHP
167 Since the development of intuitionistic fuzzy sets [21], there are many extensions have
168 been employed on it such as neutrosophic sets [22], Pythagorean fuzzy sets [23], and orthopair
169 fuzzy sets [24]. These extensions aim to handle uncertainty and clarify the indeterminacy
170 with a high level of reliability. In this study, we work on Pythagorean fuzzy sets to reduce
171 vagueness and handle the impreciseness of the system environment. In this section, we
172 will give the principles and basic operations of Pythagorean fuzzy sets and the steps of
173 Pythagorean fuzzy AHP method.

174 3.2.1. Preliminaries of Pythagorean Fuzzy Sets (PFSs)

175 Pythagorean fuzzy sets are an extension of intuitionistic fuzzy sets aiming that decision
176 makers are not enforced to assign membership and non-membership degrees whose sum
177 must be at most 1. In Pythagorean fuzzy sets, unlike the intuitionistic fuzzy sets, the sum
178 of membership and non-membership degrees can exceed 1 but the sum of squares them can
179 not. It is expressed in Definition 1 as follows:
180 Definition 1. Let X be a fixed set. A Pythagorean fuzzy set P̃ is an object having the
181 form [25]:
P̃ ∼
= {hx, µP̃ (x), νP̃ (x)i ; xX} (2)
182 where the function µP̃ (x) : X 7→ [0, 1] defines the degree of membership and νP̃ (x) : X 7→
183 [0, 1] defines the degree of non-membership of the element xX to P, respectively, and, for
184 every xX, it holds:

0 6 µÃ (x)2 + νà (x)2 6 1 (3)

185 Here, also the degree of hesitancy condition is as follows:
q
πP̃ (x) = 1 − µP̃ (x)2 − νP̃ (x)2 (4)

186 Definition 2. Let à = hµ1 , ν1 i , B̃ = hµ2 , ν2 i be two PFNs, and λ > 0, then the
187 operations on these two PFNs are defined as follows [26]:
√

à ⊕ B̃ = µ1 + µ2 − µ1 µ2 , ν1 ν2 (5)
188 √

à ⊗ B̃ = µ1 µ2 , ν1 + ν2 − ν1 ν2 (6)
189 q 
λà = 1 − (1 − µ2 )λ , ν λ (7)
190  q 
λ λ 2 λ
à = µ , 1 − (1 − ν ) (8)

191 Definition 3. Let Ãi = hµi , νi i, i = (1, 2...., n) be a collection Pn of PFNs and w =
T
192 (w1 , w2 , ..., wn ) be the weight vector of Ãi , i = (1, 2, ..., n) with i=1 wi = 1, then the
193 Pythagorean fuzzy weighted power geometric (PFWPG) operator is [27]:
n
Y n
Y
P F W P G(Ã1 , Ã2 ...., Ãn ) = ((1 − (1 − µ2i )wi )1/2 , (1 − (1 − νi2 )wi )1/2 ) (9)
i=1 i=1
7
194 3.2.2. Steps of Pythagorean Fuzzy AHP
195 In this section, the steps of interval-valued Pythagorean fuzzy (IVPF) AHP are given.
196 Before giving the steps, we represent the weighting scale of IVPF AHP in Table 6.
Table 6: Weighting scale for the interval-valued Pythagorean Fuzzy AHP method

PFN equivalents
IVPF Numbers
Linguistic Terms
µL µU νL νU
Certainly Low Importance -CLI 0 0 0.9 1
Very Low Importance -VLI 0.1 0.2 0.8 0.9
Low Importance -LI 0.2 0.35 0.65 0.8
Below Average Importance -BAI 0.35 0.45 0.55 0.65
Average Importance -AI 0.45 0.55 0.45 0.55
Above Average Importance -AAI 0.55 0.65 0.35 0.45
High Importance -HI 0.65 0.8 0.2 0.35
Very High Importance -VHI 0.8 0.9 0.1 0.2
Certainly High Importance -CHI 0.9 1 0 0
Exactly Equal -K 0.1965 0.1965 0.1965 0.1965

197 Step 1. Construct the compromised pairwise comparison matrix R = (rik )mxm with
198 respect to experts opinions based on Table 6.

Table 7: Evaluation in matrix form

C1 ... Cm
C1 h[0.1965, 0.1965] , [0.1965, 0.1965]i ... h[µL1m , µU1m ] , [νL1m , νU1m ]i
.. .. .. ..
. . . .
Cm h[µLm1 , µUm1 ] , [νLm1 , νUm1 ]i ... h[0.1965, 0.1965] , [0.1965, 0.1965]i

199 Step 2. Find the differences matrix D = (dik )mxm between lower and upper values of
200 the membership and non-membership functions using Eq. (10) and Eq. (11):

dikL = µ2ikL − νik

2
L
(10)
201
dikU = µ2ikU − νik
2
U
(11)
202 Step 3. Find the interval multiplicative matrix S = (sik )mxm using Eq. (12) and Eq.
203 (13): √
sikL = 1000dL (12)
204 √
sikU = 1000dU (13)
205 Step 4. Calculate the determinacy value τ = (τik )mxm of the rik using Eq. (14):

τik = 1 − µ2ikU − µ2ikL − νik

2 2



U
− νik L
(14)
8
206 Step 5. Multiply the determinacy degrees with S = (sik )mxm matrix for finding the
207 matrix of unnormalized weights T = (tik )mxm using Eq. (15).
 
sikL + sikU
tik = τik (15)
2

208 Step 6. Find the priority weights wi by using Eq. (16).

Pm
t
wi = Pm k=1 Pmik (16)
i=1 k=1 tik

209 3.3. Fuzzy Inference System

210 In rules-based inference systems, the knowledge within the system determines the next
211 rule to execute. Fuzzy inference is composed of a formulating process and a mapping from
212 a set of inputs to an output by using fuzzy logic. The basis provided by the mapping is
213 used for decision-making. Fuzzy inference process includes membership functions, logical
214 operations, and if-then rules. A fuzzy inference system (FIS) is a systematic reasoning
215 process of formulating input/output mappings using fuzzy logic [28].
216 The types of FIS are Mamdani type and Sugeno type. Mamdani type FIS is one of the
217 first fuzzy control systems and the most used inference system in the literature. Mamdani
218 type FIS requires the outputs to be fuzzy sets. After the aggregation operation, the obtained
219 fuzzy output is defuzzified [28].
220 In a Mamdani type FIS, input variables are combined as follows: IF x1 is A1 AND x2
221 is A2 AND ... xn is An THEN y is B, where xi (i = 1, 2, 3, ..., m) are input variables and
222 y is the output variable, A1 , A2 , ..., An and B are the linguistic terms. The inputs may be
223 fuzzy or crisp in Mamdani type FIS [28]. Figure 1 illustrates a Mamdani type FIS with crisp
224 inputs and fuzzy rules.

Figure 1: Mamdani FIS with crisp inputs and two fuzzy rules

9
225 The main difference between Sugeno type FIS and Mamdani type FIS comes from the
226 generation of outputs. Mamdani type FIS is based on the defuzzification of a fuzzy output
227 whereas Sugeno type FIS is based on the weighted average of the outputs to calculate
228 the crisp output. Therefore, the defuzzification process is not used in Sugeno FIS. The
229 coefficients of the inputs have to be defined by the user in Sugeno FIS [29].

230 3.4. Proposed Integrated Approach: PFPRA

231 The proposed approach, Pythagorean Fuzzy Proportional Risk Assessment (PFPRA),
232 consists of PFAHP, Mamdani’s FIS, and Fine Kinney methods. Theoretical bases of these
233 methods are given above in detail. The framework of the proposed integrated approach is
234 provided in Figure 4. The main steps of the proposed approach are as follows:
235 Step 1. Classify hazards in the considered work activities to obtain a hierarchy.
236 Step 2. Evaluate the elements of the hierarchy with respect to probability and severity
237 through pairwise comparisons by experts. For probability, the hazards are compared in
238 pairs, and it is asked which one is more likely to occur and how much more likely than the
239 other. For severity, which one is more severe if hazards are realized and how much more
240 severe than the other is asked in pairwise comparisons. Responses to these questions are
241 obtained in linguistic terms.
242 Step 3. Transform linguistic terms to IVPFNs by using the scale given in Table 6. In
243 other words, determine two parameters of Fine Kinney method, probability and severity,
244 through PFAHP whose steps are explained in detail in Section 3.2.2.
245 Step 4. Using the hierarchical structure, find the probability weight of each hazard by
246 multiplying the weight of the main hazard group by the weight of each hazard belongs to
247 the main hazard group. Then, apply the same procedures to find the severity weights.
248 Step 5. Obtain normalized probability weights by dividing them with their maximum.
249 Then, apply the same procedures to find the normalized severity weights. Use these nor-
250 malized weights to find membership degrees of probability and severity of a hazard to the
251 terms in the linguistic scale using the functions given in Figure 2.

Figure 2: Membership functions of probability and severity inputs [12]

10
252 Step 6. Obtain frequencies of hazards in linguistic terms and derive their corresponding
253 membership degrees by using Table 8.

Table 8: Membership degrees of a hazard to different frequency levels

Frequency µF
levels µV L µL µM µH µV H
VL 1 0.5 0 0 0
L 0.5 1 0.5 0 0
M 0 0.5 1 0.5 0
H 0 0 0.5 1 0.5
VH 0 0 0 0.5 1

254 Step 7. Use membership degree of probability and severity of a hazard found in step 5
255 and membership degrees of a hazard to different frequency levels as inputs for FIS instead
256 of scalar multiplication of traditional Fine Kinney method. In order to achieve this, take
257 the minimum of membership degree of probability, severity and frequency of a hazard to
258 find Xijk values where i, j, and k represent probability, severity, and frequency, respectively
259 as in Eq. (17).

Xijk = min(µP , µS , µF ) (17)

260 where µP , µS , and µF represent membership degree of probability, severity, and frequency
261 of a hazard, respectively.
262 Step 8. Determine the class of a hazard as one of negligible, minor, major, or critical by
263 using Table 9. The same rules given in Table 9 were used by Cebi et al. [30] for evaluation
264 of occupational risks at a construction site.

11
 Table 9: Mamdani’s fuzzy inference system

P
F S
VL L M H VH
VL N N N N Mi
L N N N Mi Mi
VL M N N Mi Mi Mi
H Mi Mi Mi Mi Mi
VH Mi Mi Mi Ma Ma
VL N N N Mi Mi
L N N N Mi Ma
L M N Mi Mi Ma Ma
H Mi Mi Ma Ma Ma
VH Mi Mi Ma C C
VL N N Mi Mi Ma
L N Mi Mi Ma Ma
M M Mi Mi Ma Ma Ma
H Mi Ma Ma Ma C
VH Mi Ma Ma C C
VL N N Mi Ma Ma
L N Mi Mi Ma Ma
H M Mi Mi Ma Ma C
H Mi Ma Ma C C
VH Ma Ma C C C
VL N N Mi Ma Ma
L N Mi Ma Ma C
VH M Mi Ma Ma C C
H Ma Ma C C C
VH Ma C C C C
N: Negligible
Mi: Minor
Ma: Major
C: Critical

265 Step 9. Take the maximum of Xijk values which belongs to the same class to determine
266 N, M i, M a, and C values by using Eqs. (18)-(21) for defuzzification procedure.

N = max(Xijk ) ∀Xijk N (18)

267
M i = max(Xijk ) ∀Xijk MI (19)
268
M a = max(Xijk ) ∀Xijk MA (20)
269
C = max(Xijk ) ∀Xijk C (21)

12
270 Step 10. Defuzzify N, M i, M a, and C values by using Eq. (22) to obtain risk magnitude
271 [12].
1 × N + 4 × M i + 7 × M a + 10 × C
RM = (22)
N + Mi + Ma + C
272 Step 11. Risk magnitude obtained through Eq. (22) is assessed with respect to Figure 3
273 to find the corresponding membership degrees of the hazard to N, M i, M a, and C. According
274 to the results, it is determined whether the operation can continue or it should be terminated
275 immediately.

Figure 3: Membership functions of risk magnitude [12]

13
 Figure 4: Framework of the proposed integrated approach

276 4. Application
277 In this section, the proposed method is applied to evaluate the risks of excavation process
278 in a construction yard. After the risk assessment is performed with the proposed method, a
14
279 comparative analysis with PFFMEA is carried out.

280 4.1. Problem Definition

281 Excavation is a process including soil excavation and transport operations to control the
282 appropriateness of the land to be constructed. The process consists of excavation, chan-
283 neling, demolition, and debris operations. The process is carried out to obtain appropriate
284 ground for structures such as buildings, roads, tunnels, and dams. Various heavy equipment
285 including excavators, bulldozers, loaders, and dump trucks is used for excavation operations
286 and technical knowledge is required. Therefore, the process requires high level of expertise.
287 Before excavation operations, investigation on the excavation area is required, and precau-
288 tions which are vital for OHS must be taken. In particular, it must be investigated whether
289 there are any electricity lines, water sources, natural gas pipes, and sewage structures in the
290 area. In this section, the risks related to OHS have been considered by using the proposed
291 approach. In this perspective, it is aimed at determining the most dangerous processes. The
292 common risks observed in excavation processes are provided in Table 10.
Table 10: Classifications of hazards in construction yard

EF1. Existence of electricity, natural gas and wastewater installations in the

parcel to be excavated
EF2. Landslide due to lack of proper bevel or shoring
Environmental EF3. Gaps in the construction yard, high elevation differences
factors (EF) EF4. Electrical panels and cables in the construction yard
EF5. Noisy environment
EF6. Bad weather conditions
EF7. Formation of puddles
EF8. Piercing, cutting materials in the construction yard
SM1. Failure to provide health surveillance
SM2. Overworking the staff
Staff
SM3. Absence of observers during maneuvers of excavation vehicles
management (SM)
SM4. Leaving excavation vehicles without a driver within the construction
yard
SM5. Having drivers of excavation vehicles lacking a driver’s license
NB1. Existence of people who do not have jobs in the excavation area
NB2. Workers entering the maneuvering area during excavation
Non-secure
NB3. Failure to comply with the speed limit of excavation vehicles within the
behaviors (NB)
construction yard
NB4. Operation of the construction equipment by non-operator employees
NB5. Being at the high parts of the construction equipment (excavators)
HE1. Lack of light and sound warning devices in excavation vehicles
HE2. The absence of periodic inspections of construction equipment
HE3. Not performing necessary controls before working with a heavy equip-
Heavy ment (loader, backhoe loader , excavator, dozer, etc.)
equipment (HE) HE4. Leaving material on construction equipment
HE5. Fire, flare and explosion resulting from ignition of oil and diesel tanks of
construction equipment
HE6. Transfer of fuel to heavy equipment
HE7. Maintenance of construction equipment by untrained workers without
operator information on sloping land
15
 HE8. Replacement of construction equipment attachments without operator
information
CYM1. Stacking of materials and equipment at the end points of places where
there is a difference in elevation within the excavation area
Construction yard CYM2. Not specifying routes of excavation vehicles within the work area
management (CYM) CYM3. Absence of warning and warning signs on the construction equipment
and in the construction yard
CYM4. Absence of lightening or inadequate lightening
CYM5. Inappropriate loading locations and ramps
CYM6. Not specifying entry-exit, maneuvering and parking places of excava-
tion vehicles within the construction yard

293 4.2. Application of the Method

294 Hazards in a construction yard are classified as given in Table 10. Firstly, main hazard
295 groups are subjected to pairwise comparison, then the hazards in the same group are com-
296 pared in pairs by experts with respect to probability and severity. Pairwise comparisons
297 and the obtained weights (w) for probability and severity are given in Tables 11-22 together
298 with their consistency ratios (CR).

Table 11: Pairwise comparison of main hazards Table 12: Pairwise comparison of main hazards
with respect to probability with respect to severity

Goal EF SM NB HE CYM w Goal EF SM NB HE CYM w

EF K HI VHI HI AAI 0.407 EF K HI HI AAI BAI 0.245
SM LI K AAI BAI LI 0.072 SM LI K AAI BAI LI 0.082
NB VLI BAI K LI VLI 0.037 NB LI BAI K LI VLI 0.044
HE LI AAI HI K BAI 0.143 HE BAI AAI HI K LI 0.164
CYM BAI HI VHI AAI K 0.341 CYM AAI HI VHI HI K 0.465
CR=0.075 CR=0.081

Table 13: Pairwise comparison of hazards caused by environmental factors with respect to probability

EF EF1 EF2 EF3 EF4 EF5 EF6 EF7 EF8 w

EF1 K BAI HI VHI CHI AAI CHI VHI 0.289
EF2 AAI K VHI VHI CHI HI CHI CHI 0.376
EF3 LI VLI K AAI VHI BAI VHI HI 0.114
EF4 VLI VLI BAI K HI LI HI AAI 0.046
EF5 CLI CLI VLI LI K VLI AI BAI 0.012
EF6 BAI LI AAI HI VHI K VHI HI 0.129
EF7 CLI CLI VLI LI AI VLI K BAI 0.012
EF8 VLI CLI LI BAI AAI LI AAI K 0.023
CR=0.085

16
 Table 14: Pairwise comparison of hazards caused by environmental factors with respect to severity

EF EF1 EF2 EF3 EF4 EF5 EF6 EF7 EF8 w

EF1 K AAI CHI HI CHI HI VHI VHI 0.352
EF2 BAI K CHI HI CHI AAI HI VHI 0.305
EF3 CLI CLI K LI AAI VLI BAI BAI 0.020
EF4 LI LI HI K VHI BAI AAI HI 0.102
EF5 CLI CLI BAI VLI K VLI LI LI 0.010
EF6 LI BAI VHI AAI VHI K HI HI 0.150
EF7 VLI LI AAI BAI AAI LI K AAI 0.035
EF8 VLI VLI AAI LI AAI LI BAI K 0.027
CR=0.091

Table 15: Pairwise comparison of hazards caused Table 16: Pairwise comparison of hazards caused
by staff management with respect to probability by staff management with respect to severity

SM SM1 SM2 SM3 SM4 SM5 w SM SM1 SM2 SM3 SM4 SM5 w
SM1 K AAI VLI BAI LI 0.052 SM1 K HI VLI AAI BAI 0.096
SM2 BAI K CLI LI VLI 0.027 SM2 LI K CLI BAI VLI 0.025
SM3 VHI CHI K HI AAI 0.562 SM3 VHI CHI K VHI AAI 0.600
SM4 AAI HI LI K BAI 0.106 SM4 BAI AAI VLI K LI 0.048
SM5 HI VHI BAI AAI K 0.253 SM5 AAI VHI BAI HI K 0.232
CR=0.053 CR=0.066

Table 17: Pairwise comparison of hazards caused Table 18: Pairwise comparison of hazards caused
by non-secure behavior with respect to probabil- by non-secure behavior with respect to severity
ity
NB NB1 NB2 NB3 NB4 NB5 w
NB NB1 NB2 NB3 NB4 NB5 w NB1 K LI VLI BAI AAI 0.052
NB1 K VLI LI BAI AAI 0.052 NB2 HI K BAI AAI VHI 0.253
NB2 VHI K AAI HI CHI 0.562 NB3 VHI AAI K HI CHI 0.562
NB3 HI BAI K AAI VHI 0.253 NB4 AAI BAI LI K HI 0.106
NB4 AAI LI BAI K HI 0.106 NB5 BAI VLI CLI LI K 0.027
NB5 BAI CLI VLI LI K 0.027 CR=0.065
CR=0.064

Table 19: Pairwise comparison of hazards caused by heavy equipment with respect to probability

HE HE1 HE2 HE3 HE4 HE5 HE6 HE7 HE8 w

HE1 K VLI LI BAI AAI AAI HI LI 0.049
HE2 VHI K HI HI VHI CHI CHI AAI 0.389
HE3 HI LI K AAI HI VHI VHI BAI 0.166
HE4 AAI LI BAI K HI HI VHI BAI 0.114
HE5 BAI VLI LI LI K AAI HI VLI 0.040
HE6 BAI CLI VLI LI BAI K AAI VLI 0.022
HE7 LI CLI VLI VLI LI BAI K VLI 0.011
HE8 HI BAI AAI AAI VHI VHI VHI K 0.208
CR=0.098

17
 Table 20: Pairwise comparison of hazards caused by heavy equipment with respect to severity

HE HE1 HE2 HE3 HE4 HE5 HE6 HE7 HE8 w

HE1 K VLI LI BAI VLI BAI AAI AAI 0.028
HE2 VHI K HI VHI AAI HI CHI CHI 0.351
HE3 HI LI K HI BAI AAI VHI VHI 0.149
HE4 AAI VLI LI K LI BAI AAI HI 0.044
HE5 VHI BAI AAI HI K HI CHI CHI 0.305
HE6 AAI LI BAI AAI LI K HI VHI 0.092
HE7 BAI CLI VLI BAI CLI LI K AAI 0.019
HE8 BAI CLI VLI LI CLI VLI BAI K 0.011
CR=0.088

Table 21: Pairwise comparison of hazards caused by construction yard management with respect to proba-
bility

CYM CYM1 CYM2 CYM3 CYM4 CYM5 CYM6 w

CYM1 K CLI VLI LI LI BAI 0.023
CYM2 CHI K AAI HI HI VHI 0.482
CYM3 VHI BAI K AAI HI HI 0.241
CYM4 HI LI BAI K AAI HI 0.126
CYM5 HI LI LI BAI K AAI 0.085
CYM6 AAI VLI LI LI BAI K 0.043
CR=0.084

Table 22: Pairwise comparison of hazards caused by construction yard management with respect to severity

CYM CYM1 CYM2 CYM3 CYM4 CYM5 CYM6 w

CYM1 K VLI LI LI BAI AAI 0.043
CYM2 VHI K AAI HI HI CHI 0.482
CYM3 HI BAI K AAI HI VHI 0.241
CYM4 HI LI BAI K AAI HI 0.126
CYM5 AAI LI LI BAI K HI 0.085
CYM6 BAI CLI VLI LI LI K 0.023
CR=0.084

299 Linguistic terms in these matrices are converted to interval-valued Pythagorean fuzzy
300 numbers by using the scale given in Table 6. Then, PFAHP is applied to obtain probability
301 and severity values of hazards. Probability and severity values obtained through PFAHP
302 are provided in Tables 23-24, respectively.

18
 Table 23: Probability values obtained through Pythagorean fuzzy AHP

EF 0.407 SM 0.072 NB 0.037 HE 0.143 CYM 0.341

EF1 0.289 SM1 0.052 NB1 0.052 HE1 0.049 CYM1 0.023
EF2 0.376 SM2 0.027 NB2 0.562 HE2 0.389 CYM2 0.482
EF3 0.114 SM3 0.562 NB3 0.253 HE3 0.166 CYM3 0.241
EF4 0.046 SM4 0.106 NB4 0.106 HE4 0.114 CYM4 0.126
EF5 0.012 SM5 0.253 NB5 0.027 HE5 0.040 CYM5 0.085
EF6 0.129 HE6 0.022 CYM6 0.043
EF7 0.012 HE7 0.011
EF8 0.023 HE8 0.208

Table 24: Severity values obtained through Pythagorean fuzzy AHP

EF 0.245 SM 0.082 NB 0.044 HE 0.164 CYM 0.465

EF1 0.352 SM1 0.096 NB1 0.052 HE1 0.028 CYM1 0.043
EF2 0.305 SM2 0.025 NB2 0.253 HE2 0.351 CYM2 0.482
EF3 0.020 SM3 0.600 NB3 0.562 HE3 0.149 CYM3 0.241
EF4 0.102 SM4 0.048 NB4 0.106 HE4 0.044 CYM4 0.126
EF5 0.010 SM5 0.232 NB5 0.027 HE5 0.305 CYM5 0.085
EF6 0.150 HE6 0.092 CYM6 0.023
EF7 0.035 HE7 0.019
EF8 0.027 HE8 0.011

303 The probability weight of each hazard is calculated by multiplying the weight of the
304 main hazard group by the weight of each hazard belongs to the main hazard group. The
305 same procedures are applied to find the severity weights. After finding probability weights,
306 normalized probability weights are derived by dividing the obtained probability weights by
307 their maximum. The normalized severity weights are calculated in the same way. Normalized
308 probability and severity weights are given in Tables 25-26, respectively.

Table 25: Normalized probability weights of hazards

EF NPW SM NPW NB NPW HE NPW CYM NPW

EF1 0.715 SM1 0.023 NB1 0.012 HE1 0.042 CYM1 0.048
EF2 0.929 SM2 0.012 NB2 0.127 HE2 0.338 CYM2 1.000
EF3 0.282 SM3 0.246 NB3 0.057 HE3 0.144 CYM3 0.500
EF4 0.113 SM4 0.047 NB4 0.024 HE4 0.099 CYM4 0.261
EF5 0.029 SM5 0.111 NB5 0.006 HE5 0.035 CYM5 0.176
EF6 0.319 HE6 0.019 CYM6 0.089
EF7 0.029 HE7 0.010
EF8 0.056 HE8 0.181

19
 Table 26: Normalized severity weights of hazards

EF NSW SM NSW NB NSW HE NSW CYM NSW

EF1 0.384 SM1 0.035 NB1 0.010 HE1 0.020 CYM1 0.089
EF2 0.333 SM2 0.009 NB2 0.050 HE2 0.256 CYM2 1.000
EF3 0.021 SM3 0.221 NB3 0.111 HE3 0.109 CYM3 0.500
EF4 0.111 SM4 0.018 NB4 0.021 HE4 0.032 CYM4 0.261
EF5 0.011 SM5 0.085 NB5 0.005 HE5 0.222 CYM5 0.176
EF6 0.163 HE6 0.067 CYM6 0.048
EF7 0.038 HE7 0.014
EF8 0.029 HE8 0.008

309 Frequencies of the hazards, obtained from the experts in linguistic form, are given in
310 Table 27, and membership degrees of the hazards to different frequency levels are determined
311 by using Table 8.

Table 27: Hazards frequencies in linguistic terms

EF Frequency SM Frequency NB Frequency HE Frequency CYM Frequency

EF1 M SM1 L NB1 L HE1 H CYM1 M
EF2 H SM2 H NB2 L HE2 M CYM2 H
EF3 VH SM3 M NB3 M HE3 M CYM3 M
EF4 H SM4 L NB4 L HE4 M CYM4 M
EF5 VH SM5 VL NB5 H HE5 L CYM5 L
EF6 L HE6 VH CYM6 M
EF7 L HE7 L
EF8 H HE8 M

312 The normalized probability and severity weights obtained through PFAHP are used to
313 determine membership degrees of probability and severity of a hazard. The membership
314 degrees for probability and severity are used as inputs for FIS together with membership
315 degrees of a hazard to different frequency levels. After N, M i, M a, and C values are cal-
316 culated by using Eqs. (18)-(21), defuzzification is performed by using Eq. (22) to find risk
317 magnitudes. Results of the proposed integrated approach are given in Table 28.

20
 Table 28: Results of the proposed method

Hazards RM N Mi Ma C
CYM2 10 0.0% 0.0% 0.0% 100.0%
EF2 8.285 0.0% 0.0% 35.8% 64.2%
CYM3 5.999 0.0% 0.0% 100.0% 0.0%
EF1 5.043 0.0% 47.8% 52.2% 0.0%
CYM4 3.091 0.0% 100.0% 0.0% 0.0%
SM3 2.915 4.2% 95.8% 0.0% 0.0%
HE2 2.779 11.1% 88.9% 0.0% 0.0%
EF4 2.346 32.7% 67.3% 0.0% 0.0%
HE3 2.308 34.6% 65.4% 0.0% 0.0%
EF6 2.300 35.0% 65.0% 0.0% 0.0%
CYM5 2.245 37.7% 62.3% 0.0% 0.0%
NB3 1.874 56.3% 43.7% 0.0% 0.0%
NB2 1.851 57.5% 42.5% 0.0% 0.0%
EF3 1.819 59.0% 41.0% 0.0% 0.0%
CYM1 1.685 65.8% 34.2% 0.0% 0.0%
CYM6 1.685 65.8% 34.2% 0.0% 0.0%
HE4 1.527 73.6% 26.4% 0.0% 0.0%
EF8 1.391 80.4% 19.6% 0.0% 0.0%
HE5 1.385 80.7% 19.3% 0.0% 0.0%
EF7 1.359 82.0% 18.0% 0.0% 0.0%
SM1 1.289 85.6% 14.4% 0.0% 0.0%
HE6 1.282 85.9% 14.1% 0.0% 0.0%
HE1 1.268 86.6% 13.4% 0.0% 0.0%
NB4 1.255 87.3% 12.7% 0.0% 0.0%
SM4 1.238 88.1% 11.9% 0.0% 0.0%
EF5 1.143 92.8% 7.2% 0.0% 0.0%
HE8 1.132 93.4% 6.6% 0.0% 0.0%
NB1 1.125 93.8% 6.2% 0.0% 0.0%
HE7 1.121 93.9% 6.1% 0.0% 0.0%
SM2 1.110 94.5% 5.5% 0.0% 0.0%
NB5 1.064 96.8% 3.2% 0.0% 0.0%
SM5 1 100.0% 0.0% 0.0% 0.0%

318 According to Table 28, CYM2 - Not specifying routes of excavation vehicles within the
319 work area is the most critical hazard in the construction yard and CYM2 must be immedi-
320 ately eradicated. The second most dangerous hazard is EF2 - Landslide due to lack of proper
321 bevel or shoring with 64.2% critical and 35.8% major membership degrees. The process must
322 be stopped until these hazards are eliminated. CYM3 - Absence of warning and warning
323 signs on the construction equipment and in the construction yard and EF1 - Existence of
324 electricity, natural gas and wastewater installations in the parcel to be excavated are found
21
325 to be major risks with 100% and 53.3% membership degrees. It is thus possible to say that
326 the process may continue with extreme caution while taking necessary actions to eliminate
327 these hazards. CYM4 - Absence of lightening or inadequate lightening, SM3 - Absence of
328 observers during maneuvers of excavation vehicles, HE2 - Absence of periodic inspections of
329 construction equipment, EF4 - Electrical panels and cables in the construction yard, HE3 -
330 Not performing necessary controls before working with a heavy equip-ment (loader, backhoe
331 loader , excavator, dozer, etc.), EF6 - Bad weather conditions, and CYM5 - Inappropriate
332 loading locations and ramps are qualified as minor risks with 100%, 95.8%, 88.9%, 67.3%,
333 65.4%, 65.0%, and 62.3% membership degrees, respectively. These risks should be moni-
334 tored. The rest of the hazards are found to be negligible, which means they are acceptable.
335 4.3. Comparative Analysis
336 In this sub-section, after the basics of failure modes and effects analysis (FMEA) and its
337 use in the literature are provided, results obtained through PFPRA are compared with the
338 results derived from PFFMEA.
339 4.3.1. Failure modes and effects analysis
340 Failure mode and effective analysis (FMEA) is composed of successive steps for collecting
341 knowledge about possible failure modes in a design, manufacturing process, product or
342 service [31]. There are three components of the failure mode and effects analysis (FMEA):
343 Severity (S): If a failure occurs, what effect will it have on a service or on a product
344 quality?
345 Probability of occurrence (PO): What is the occurrence probability of a particular failure?
346 Detectability (D): What mechanisms exist to detect failures when they occur?
347 Each of the above metrics requires clear definitions and a corresponding scale to rank
348 or score the projected impact [31]. In addition, a composite score, namely Risk Priority
349 Number (RPN), would then need to be calculated as in Eq. (23).

RP N = S × P O × D (23)
350 In classical FMEA, using the linguistic variables in Tables 29-31, the crisp numerical
351 values of severity, occurrence, and detectability can be obtained [31].
Table 29: Frequency of occurrence evaluation [32]

Quantitative
Rank Occurrence Meaning
Failure Probability
1 Remote Failure is unlikely <1 in 106
2 1 in 20,000
Low Relatively few failures
3 1 in 4,000
4 1 in 1,000
5 Moderate Occasional failures 1 in 400
6 1 in 60
7 1 in 40
High Repeated failures
8 1 in 20
9 1 in 8
Very High Failure is almost inevitable
10 1 in 2

22
 Table 30: Detectability evaluation criteria [32]

Rank Detectability Meaning

1,2 Very high A potential design failure will be almost certainly detected.
3,4 High A potential design failure will be detected by a good chance.
5,6 Moderate A potential design failure may be detected.
7,8 Low A potential design failure is not likely detected.
9 Very low A potential design failure will not be probably detected.
10 Non-detection A potential design failure can not be detected.

Table 31: Severity evaluation criteria [32]

Severity
Rank Meaning
effect
1 Minor The failure will not have any real effect on the system performance and it will not
even noticed by the customers.
2,3 Low The failure will have a slight effect on the system performance and it will probably
noticed by the customers.
4,5,6 Moderate The failure will cause some dissatisfaction and it will make the customers uncom-
fortable.
7,8 High The failure will cause a high degree of customer dissatisfaction and an inoperable
vehicle or system. It does not involve noncompliance with regulations.
10 Very high The failure will affect the safe vehicle operation and it involves noncompliance with
regulations.

352 FMEA is a powerful technique for estimating risks and preventing failures. The potential
353 effects of failures are about what the customer will be aware of when the failure occurs. In
354 other words, it describes the response of the customer who encountered the failure. It
355 is necessary to give severity points for failure effects by using a scale, generally from 1
356 to 10. After the effects are identified, the causes of the failures must be identified. The
357 occurence score, the likelihood of failures occurring, must be determined. The detection
358 score is identified by the ability of the control method to detect a failure with a scale from
359 1 to 10. Then, risk priority number is calculated by multiplying severity, detection and
360 occurence [33]. In the literature, FMEA has been commonly combined with AHP. After
361 identifying the relevant factors under FMEA, the fuzzy AHP was employed by Hu et al. [34]
362 to determine the relative weights of factors,the declaration statement, likelihood of detection
363 by customhouse, the testing report and the frequency of green component used by project.
364 The risk priority number for a green component then can be calculated. Maheswaran and
365 Loganathan [35] used AHP to identify the weights for risk factors and utilized Preference
366 Ranking Organization Method for Enrichment Evaluation (PROMETHEE) for prioritizing
367 the failure mode. Thus, this integration enables decision makers to improve precision of
368 prioritizing failure modes [35]. In the study of Liu et al. [36], a combination of VIKOR,
369 DEMATEL and AHP was employed to prioritize the risk of the failure modes diagnosed
370 in FMEA for a diesel engines turbocharger system. The impacts of failure modes were
371 identified via the modified VIKOR whereas the influential relation map among the failure
372 modes and their causes was composed through the DEMATEL method. Then, AHP is
373 used to determine the weights and prioritize the failure modes. Shahrabi and Shojaei [37]
23
374 used the combination of AHP and FMEA for one of the largest finery stations of the world.
375 Fuzzy logic and AHP were utilized to deal with the limitations of traditional FMEA. The
376 relative weights of risk impacts on time, cost, quality and safety were determined through
377 AHP in [38]. Suebsomran and Talabgeaw [39] integrated AHP with FMEA to investigate the
378 maintenance priority for a thermal power plant. Priority of failure modes related to man per
379 hour working, maintenance cost and line priority were analyzed [39]. In [40], fuzzy sets were
380 integrated to FMEA by using the combination of fuzzy technique for order preference by
381 similarity to ideal solution (TOPSIS) and fuzzy AHP. Aslani et al. [41] used AHP and FMEA
382 for alternator failure causes. Liu et al. [42] used the combination of FMEA, AHP, GTM and
383 DEMATEL to identify the priorities of failure modes. Fuzzy approach was also incorporated
384 to this combination to manage immeasurable information in the evaluation process [42]. In
385 [43], FMEA and AHP were used to effectively investigate the risks in occupational health
386 and safety management system of the mining industry. In [44], AHP, fuzzy logic and FMEA
387 were used to analyze the main risks of highway construction projects. The criteria taken
388 into account to prioritize risks are cost, time and quality in the evaluation process [44].

389 4.3.2. Comparison of PFPRA and PFFMEA

390 In this sub-section, the results of PFPRA are compared with the result obtained from
391 PFFMEA. Due to space constraints, the steps of PFFMEA are not provided in this study.
392 Comparison of the results obtained from these two methods is given in Table 32.

Table 32: Results and comparison of both methods

FINE KINNEY FMEA FINE KINNEY FMEA

Hazard P F S RPN Rank D RM Rank Hazard P F S RM Rank D RPN Rank
CYM2 1.000 1.000 4 10 1 5 8 1 HE4 0.099 0.032 3 1.527 17 5 1 23
EF2 0.929 0.333 4 8.285 2 4 5.715 2 EF8 0.056 0.029 4 1.391 18 4 1.391 13
CYM3 0.500 0.500 3 5.999 3 5 4 4 HE5 0.035 0.222 2 1.385 19 3 1.417 12
EF1 0.715 0.384 3 5.043 4 3 5.043 3 EF7 0.029 0.038 2 1.359 20 4 1.359 15
CYM4 0.261 0.261 3 3.091 5 4 2.183 9 SM1 0.023 0.035 2 1.289 21 3 1.289 16
SM3 0.246 0.221 3 2.915 6 5 1 23 HE6 0.019 0.067 5 1.282 22 4 1.282 17
HE2 0.338 0.256 3 2.779 7 5 1.107 22 HE1 0.042 0.020 4 1.268 23 5 1 23
EF4 0.113 0.111 4 2.346 8 4 2.346 5 NB4 0.024 0.021 2 1.255 24 3 1.255 18
HE3 0.144 0.109 3 2.308 9 4 2.308 6 SM4 0.047 0.018 2 1.238 25 3 1.238 19
EF6 0.319 0.163 2 2.300 10 4 2.300 7 EF5 0.029 0.011 5 1.143 26 5 1 23
CYM5 0.176 0.176 2 2.245 11 4 2.245 8 HE8 0.181 0.008 3 1.132 27 2 1.132 20
NB3 0.057 0.111 3 1.874 12 4 1.874 10 NB1 0.012 0.010 2 1.125 28 3 1.125 21
NB2 0.127 0.050 2 1.851 13 2 1.851 11 HE7 0.010 0.014 2 1.121 29 5 1 23
EF3 0.282 0.021 5 1.819 14 4 1.378 14 SM2 0.012 0.009 4 1.110 30 5 1 23
CYM1 0.048 0.089 3 1.685 15 5 1 23 NB5 0.006 0.005 4 1.064 31 5 1 23
CYM6 0.089 0.048 3 1.685 15 5 1 23 SM5 0.111 0.085 1 1 32 5 1 23

393 Results of both methods indicate that CYM2 - Not specifying routes of excavation vehi-
394 cles within the work area is the most dangerous hazard in a construction yard, followed by
395 EF2 - Landslide due to lack of proper bevel or shoring.

24
 Table 33: Hazard assessment with respect to both methods

PFPRA (FINE KINNEY) PFFMEA

Hazards
RM N Mi Ma C RM N Mi Ma C
CYM2 10.000 0.0% 0.0% 0.0% 100.0% 8.000 0.0% 0.0% 50.0% 50.0%
EF2 8.285 0.0% 0.0% 35.8% 64.2% 5.715 0.0% 14.2% 85.8% 0.0%
CYM3 5.999 0.0% 0.0% 100.0% 0.0% 4.000 0.0% 100.0% 0.0% 0.0%
EF1 5.043 0.0% 47.8% 52.2% 0.0% 5.043 0.0% 47.8% 52.2% 0.0%
CYM4 3.091 0.0% 100.0% 0.0% 0.0% 2.183 40.9% 59.1% 0.0% 0.0%
SM3 2.915 4.2% 95.8% 0.0% 0.0% 1.000 100.0% 0.0% 0.0% 0.0%
HE2 2.779 11.1% 88.9% 0.0% 0.0% 1.107 94.7% 5.3% 0.0% 0.0%
EF4 2.346 32.7% 67.3% 0.0% 0.0% 2.346 32.7% 67.3% 0.0% 0.0%
HE3 2.308 34.6% 65.4% 0.0% 0.0% 2.308 34.6% 65.4% 0.0% 0.0%
EF6 2.300 35.0% 65.0% 0.0% 0.0% 2.300 35.0% 65.0% 0.0% 0.0%
CYM5 2.245 37.7% 62.3% 0.0% 0.0% 2.245 37.7% 62.3% 0.0% 0.0%
NB3 1.874 56.3% 43.7% 0.0% 0.0% 1.874 56.3% 43.7% 0.0% 0.0%
NB2 1.851 57.5% 42.5% 0.0% 0.0% 1.851 57.5% 42.5% 0.0% 0.0%
EF3 1.819 59.0% 41.0% 0.0% 0.0% 1.378 81.1% 18.9% 0.0% 0.0%
CYM1 1.685 65.8% 34.2% 0.0% 0.0% 1.000 100.0% 0.0% 0.0% 0.0%
CYM6 1.685 65.8% 34.2% 0.0% 0.0% 1.000 100.0% 0.0% 0.0% 0.0%
HE4 1.527 73.6% 26.4% 0.0% 0.0% 1.000 100.0% 0.0% 0.0% 0.0%
EF8 1.391 80.4% 19.6% 0.0% 0.0% 1.391 80.4% 19.6% 0.0% 0.0%
HE5 1.385 80.7% 19.3% 0.0% 0.0% 1.417 79.2% 20.8% 0.0% 0.0%
EF7 1.359 82.0% 18.0% 0.0% 0.0% 1.359 82.0% 18.0% 0.0% 0.0%
SM1 1.289 85.6% 14.4% 0.0% 0.0% 1.289 85.6% 14.4% 0.0% 0.0%
HE6 1.282 85.9% 14.1% 0.0% 0.0% 1.282 85.9% 14.1% 0.0% 0.0%
HE1 1.268 86.6% 13.4% 0.0% 0.0% 1.000 100.0% 0.0% 0.0% 0.0%
NB4 1.255 87.3% 12.7% 0.0% 0.0% 1.255 87.3% 12.7% 0.0% 0.0%
SM4 1.238 88.1% 11.9% 0.0% 0.0% 1.238 88.1% 11.9% 0.0% 0.0%
EF5 1.143 92.8% 7.2% 0.0% 0.0% 1.000 100.0% 0.0% 0.0% 0.0%
HE8 1.132 93.4% 6.6% 0.0% 0.0% 1.132 93.4% 6.6% 0.0% 0.0%
NB1 1.125 93.8% 6.2% 0.0% 0.0% 1.125 93.8% 6.2% 0.0% 0.0%
HE7 1.121 93.9% 6.1% 0.0% 0.0% 1.000 100.0% 0.0% 0.0% 0.0%
SM2 1.110 94.5% 5.5% 0.0% 0.0% 1.000 100.0% 0.0% 0.0% 0.0%
NB5 1.064 96.8% 3.2% 0.0% 0.0% 1.000 100.0% 0.0% 0.0% 0.0%
SM5 1.000 100.0% 0.0% 0.0% 0.0% 1.000 100.0% 0.0% 0.0% 0.0%

396 Risk assessment by both methods is given in Table 33. Pink, orange, yellow and green
397 cells represent critical, major, minor and negligible risks, respectively. According to Table
398 33, CYM2 - Not specifying routes of excavation vehicles within the work area is qualified
399 as critical by both PFPRA and PFFMEA methods. However, EF2 - Landslide due to lack
400 of proper bevel or shoring is found to be a major risk by PFFMEA. SM5 - Having drivers
401 of excavation vehicles lacking a driver’s license is found to be the least dangerous risk in
402 both methods. Moreover, major, minor and negligible risks are also mostly the same for
403 both methods. In other words, most of the risks were found in the same category by both
404 methods. In only a few cases, the results of the two methods are different. The reason
25
405 behind these differences might be the existence of non-common parameters (i.e. frequency
406 and detectability) in the two methods. Therefore, it is possible to say that the proposed
407 integrated method is reliable and informative.

408 5. Conclusions
409 Occupational health and safety requires systematic analysis to protect employees from
410 dangers that might be caused by various activities in the workplace. In this study, Pythagorean
411 fuzzy proportional risk assessment (PFPRA), a combination of Fine Kinney, Pythagorean
412 fuzzy analytic hierarchy process and a fuzzy inference system, is proposed and applied for
413 the hazards of excavation process in a construction yard. The proposed integrated method
414 has successfully evaluated the hazards in a construction yard based on experts’ opinions.
415 As a result of the analysis by the proposed method, risks that require observation, imme-
416 diate action or stopping the process are determined. The main difference of the proposed
417 novel approach is the integration of these methods in a way facilitating a more accurate risk
418 assessment.
419 The comparison with Pythagorean fuzzy failure modes and effects analysis is conducted
420 and it is revealed that the proposed method provides reliable and consistent outcomes con-
421 taining more information on the uncertainty of decision makers.
422 PRA (Fine Kinney) method considers probability, severity and frequency parameters
423 whereas FMEA takes probability, severity and detectability into account in the risk assess-
424 ment process. However, a method which includes all parameters related to risk assessment
425 does not exist, so development of a method involving probability, severity, frequency and
426 detectability parameters might be investigated for further research. Another suggestion for
427 further research is the extension of the proposed method through the usage of other fuzzy
428 set types such as hesitant fuzzy sets, neutrosophic sets, and type-2 fuzzy sets. Besides,
429 triangular fuzzy or trapezoidal fuzzy Pythagorean fuzzy sets can be alternatively used for
430 the future extensions of this study.

431 References
432 [1] M. Rausand, Risk assessment: theory, methods, and applications, Vol. 115, John Wiley & Sons, 2013.
433 [2] S. Cebi, Risk Management in Disasters, in: S. Cebi (Ed.), Risk and Crisis Management in Disasters,
434 Anadolu University Open Education Publications, Anadolu University, 2017, Ch. 6, pp. 1–20.
435 [3] Fundamental principles of occupational health and safety, accessed: 2017-06-16.
436 URL http://www.ilo.org/wcmsp5/groups/public/@dgreports/@dcomm/@publ/documents/publi
437 cation/wcms 093550.pdf
438 [4] G. F. Kinney, Practical risk analysis for safety management, Tech. rep., China Lake, CA: Naval Weapons
439 Center (1976).
440 [5] M. Gul, E. Celik, E. Akyuz, A hybrid risk-based approach for maritime applications: The case of ballast
441 tank maintenance., Human and Ecological Risk Assessment: An International Journal (accepted).
442 [6] A. Kokangül, U. Polat, C. Dağsuyu, A new approximation for risk assessment using the ahp and Fine
443 Kinney methodologies, Safety science 91 (2017) 24–32.
444 [7] E. Topuz, C. A. van Gestel, An approach for environmental risk assessment of engineered nanomaterials
445 using Analytical Hierarchy Process (AHP) and fuzzy inference rules, Environment international 92
446 (2016) 334–347.
26
447 [8] A. Nieto-Morote, F. Ruz-Vila, A fuzzy approach to construction project risk assessment, International
448 Journal of Project Management 29 (2) (2011) 220–231.
449 [9] A. Rodrı́guez, F. Ortega, R. Concepción, A method for the evaluation of risk in IT projects, Expert
450 Systems with Applications 45 (2016) 273–285.
451 [10] M. Yang, F. I. Khan, R. Sadiq, Prioritization of environmental issues in offshore oil and gas operations:
452 A hybrid approach using fuzzy inference system and fuzzy analytic hierarchy process, Process Safety
453 and Environmental Protection 89 (1) (2011) 22–34.
454 [11] M. Abdelgawad, A. R. Fayek, Risk management in the construction industry using combined fuzzy
455 FMEA and fuzzy AHP, Journal of Construction Engineering and Management 136 (9) (2010) 1028–
456 1036.
457 [12] J. Zeng, M. An, N. J. Smith, Application of a fuzzy based decision making methodology to construction
458 project risk assessment, International journal of project management 25 (6) (2007) 589–600.
459 [13] S. Cebi, Developing a fuzzy based decision making model for risk analysis in construction project,
460 Journal of Multiple-Valued Logic & Soft Computing 17 (4).
461 [14] M. An, S. Huang, C. Baker, Railway risk assessment-the fuzzy reasoning approach and fuzzy analytic
462 hierarchy process approaches: a case study of shunting at waterloo depot, Proceedings of the Institution
463 of Mechanical Engineers, Part F: Journal of rail and rapid transit 221 (3) (2007) 365–383.
464 [15] H. F. Sabokbar, A. Ayashi, A. Hosseini, A. Banaitis, N. Banaitienė, R. Ayashi, Risk assessment in
465 tourism system using a fuzzy set and dominance-based rough set, Technological and Economic Devel-
466 opment of Economy 22 (4) (2016) 554–573.
467 [16] O. Acuner, S. Cebi, An effective risk-preventive model proposal for occupational accidents at shipyards,
468 Brodogradnja 67 (1) (2016) 67–84.
469 [17] S. Cebi, E. Akyuz, Y. Sahin, Developing web based decision support system for evaluation occupational
470 risks at shipyards, Brodogradnja 68 (1) (2017) 17–30.
471 [18] W. T. Fine, Mathematical evaluations for controlling hazards, Tech. rep., Naval Ordnance Lab White
472 Oak MD (1971).
473 [19] P.-K. Marhavilas, D. Koulouriotis, V. Gemeni, Risk analysis and assessment methodologies in the
474 work sites: On a review, classification and comparative study of the scientific literature of the period
475 2000–2009, Journal of Loss Prevention in the Process Industries 24 (5) (2011) 477–523.
476 [20] M. Stankovic, V. Stankovic, Comparative analysis of methods for risk assessment-Kinney and Auva,
477 Safety Engineering (2013) 129–135.
478 [21] K. T. Atanassov, Intuitionistic fuzzy sets, Fuzzy sets and Systems 20 (1) (1986) 87–96.
479 [22] F. Smarandache, Neutrosophic logic and set, mss (1995).
480 [23] R. R. Yager, Pythagorean fuzzy subsets, in: IFSA World Congress and NAFIPS Annual Meeting
481 (IFSA/NAFIPS), 2013 Joint, IEEE, 2013, pp. 57–61.
482 [24] R. R. Yager, N. Alajlan, Approximate reasoning with generalized orthopair fuzzy sets, Information
483 Fusion 38 (2017) 65–73.
484 [25] R. R. Yager, Properties and applications of Pythagorean fuzzy sets, in: Imprecision and Uncertainty
485 in Information Representation and Processing, Springer, 2016, pp. 119–136.
486 [26] X. Zhang, Z. Xu, Extension of TOPSIS to multiple criteria decision making with pythagorean fuzzy
487 sets, International Journal of Intelligent Systems 29 (12) (2014) 1061–1078.
488 [27] R. R. Yager, A. M. Abbasov, Pythagorean membership grades, complex numbers, and decision making,
489 International Journal of Intelligent Systems 28 (5) (2013) 436–452.
490 [28] E. H. Mamdani, S. Assilian, An experiment in linguistic synthesis with a fuzzy logic controller, Inter-
491 national journal of man-machine studies 7 (1) (1975) 1–13.
492 [29] M. Sugeno, Industrial applications of fuzzy control, Elsevier Science Inc., 1985.
493 [30] A. Cebi, V. Togan, S. Cebi, A fuzzy inference based model for evaluation of occupational risks at
494 construction site, International Journal of Civil Engineering and Applications 6 (81) (2016) 61–77.
495 [31] C. Kahraman, İ. Kaya, Ö. Şenvar, Healthcare failure mode and effects analysis under fuzziness, Human
496 and Ecological Risk Assessment: An International Journal 19 (2) (2013) 538–552.
497 [32] C. E. Pelaez, J. B. Bowles, Using fuzzy logic for system criticality analysis, in: Reliability and Main-

27
498 tainability Symposium, 1994. Proceedings., Annual, IEEE, 1994, pp. 449–455.
499 [33] D. H. Stamatis, Failure mode and effect analysis: FMEA from theory to execution, ASQ Quality Press,
500 2003.
501 [34] A. H. Hu, C.-W. Hsu, T.-C. Kuo, W.-C. Wu, Risk evaluation of green components to hazardous
502 substance using FMEA and FAHP, Expert Systems with Applications 36 (3) (2009) 7142–7147.
503 [35] K. Maheswaran, T. Loganathan, A novel approach for prioritization of failure modes in FMEA using
504 MCDM, International Journal of Engineering Research and Applications 3 (4) (2013) 733–739.
505 [36] H.-C. Liu, J.-X. You, X.-F. Ding, Q. Su, Improving risk evaluation in FMEA with a hybrid multiple
506 criteria decision making method, International Journal of Quality & Reliability Management 32 (7)
507 (2015) 763–782.
508 [37] M. Shahrabi, A. A. Shojaei, Application of FMEA and AHP in lean maintenance, Int J Mod Eng Sci
509 3 (2014) 61–73.
510 [38] A. Mohammadi, M. Tavakolan, Construction project risk assessment using combined fuzzy and FMEA,
511 in: IFSA World Congress and NAFIPS Annual Meeting (IFSA/NAFIPS), 2013 Joint, IEEE, 2013, pp.
512 232–237.
513 [39] A. Suebsomran, S. Talabgeaw, Critical maintenance of thermal power plant using the combination
514 of failure mode effect analysis and AHP approches, King Mongkuts University of Technology North
515 Bangkok International Journal of Applied Science and Technology 3 (3) (2013) 1–6.
516 [40] A. C. Kutlu, M. Ekmekçioğlu, Fuzzy failure modes and effects analysis by using fuzzy TOPSIS-based
517 fuzzy AHP, Expert Systems with Applications 39 (1) (2012) 61–67.
518 [41] R. Aslani, H. Feili, H. Javanshir, A hybrid of fuzzy FMEA-AHP to determine factors affecting alternator
519 failure causes, Management Science Letters 4 (9) (2014) 1981–1984.
520 [42] H.-C. Liu, J.-X. You, M.-M. Shan, Q. Su, Systematic failure mode and effect analysis using a hybrid
521 multiple criteria decision-making approach, Total Quality Management & Business Excellence (2017)
522 1–28.
523 [43] J. Bao, J. Johansson, J. Zhang, An occupational disease assessment of the mining industrys occupational
524 health and safety management system based on FMEA and an improved AHP model, Sustainability
525 9 (1) (2017) 94.
526 [44] M. Ahmadi, K. Behzadian, A. Ardeshir, Z. Kapelan, Comprehensive risk management using fuzzy
527 FMEA and MCDA techniques in highway construction projects, Journal of Civil Engineering and
528 Management (2016) 1–11.

28