Formation of Dataset for Fuzzy Quantitative Risk Assessment of LNG Bunkering SIMOPs

: New international regulations aimed at decarbonizing maritime transportation are positively contributing to attention being paid to the use of liqueﬁed natural gas (LNG) as a ship fuel. Scaling up LNG-fueled ships is highly dependent on safe bunkering operations, particularly during simultaneous operations (SIMOPs); therefore, performing a quantitative risk assessment (QRA) is either mandated or highly recommended, and a dynamic quantitative risk assessment (DQRA) has been developed to make up for the deﬁciencies of the traditional QRA. The QRA and DQRA are both data-driven processes, and so far, the data of occurrence rates (ORs) of basic events (BEs) in LNG bunkering SIMOPs are unavailable. To ﬁll this gap, this study identiﬁed a total of 41 BEs and employed the online questionnaire method, the fuzzy set theory, and the Onisawa function to the investigation of the fuzzy ORs for the identiﬁed BEs. Purposive sampling was applied when selecting experts in the process of online data collection. The closed-ended structured questionnaire garnered responses from 137 experts from the industry and academia. The questionnaire, the raw data and obtained ORs, and the process of data analysis are presented in this data descriptor. The obtained data can be used directly in QRAs and DQRAs. This dataset is ﬁrst of its kind and could be expanded further for research in the ﬁeld of risk assessment of LNG bunkering.


Introduction
Maritime shipping, which represents 80-90% of international trade, is less carbon intensive than other forms of transport [1]; however, due to the large volumes of freight and long distances travelled, maritime shipping is responsible for about 3% of total global anthropogenic greenhouse gas (GHG) emissions on a carbon dioxide equivalent basis [2]. In recent times, the maritime industry has increased its efforts against global GHG emissions. The International Maritime Organization (IMO) has set the target to cut the carbon intensity of all ships by at least 40% by 2030, and to reduce total GHG emissions from global shipping by 50% (compared to 2008 levels) by 2050 [3]. Further stringent requirements are expected from the climate change agenda of the IMO's Marine Environment Protection Committee (MEPC), upon release of the Intergovernmental Panel on Climate Change (IPCC) report after the 26th UN Climate Change Conference (COP 26) in Glasgow [4]; therefore, the need to switch to low-carbon or zero-carbon alternative fuels seems urgent for the maritime industry [5,6]. Given that zero-emission fuels, with the relevant technologies, such as green hydrogen, green ammonia, and green methanol, are premature and might be introduced after 2028 [7,8], liquified natural gas (LNG) is considered to be a suitable practicable Data 2022, 7, 60 3 of 13 commonly used fuzzy numbers are triangular and trapezoidal fuzzy numbers [38]. In the existing literature, studies have demonstrated that the applications of the fuzzy set theory contribute to risk assessment methods and safety-related decision-makings in the real-world [39][40][41][42][43][44].
The dataset provided in this data descriptor is aimed at filling the gap for the lack of the ORs required for the BEs in the field of the risk assessment of LNG bunkering SIMOPs.

Methods
This section presents the design strategy and methods applied to data collection and data analysis.

Ethics Approval
The online questionnaire method used in this work involves human information, and therefore, an ethics application was submitted to the University of Tasmania's Social Sciences Human Research Ethics Committee on 13 January 2021, and it was approved on 9 February 2021 (Project ID:23903).
The survey invitations were distributed in China, where the authors had an extended network of maritime and LNG-related expertise. The participants' professional fields, professional positions, service years, and education levels were sought in the process of data collection. The participants are referred to by numeric pseudonyms for confidentiality and anonymity purposes. In other words, personal information is de-identified before analysis and replaced by a code. The participants' information is kept in password-protected computers, which are separated from other data. All participants are assured that their private information is kept strictly confidential.

Online Questionnaire
A closed-ended structured questionnaire was developed and made available to potential participants through online access with the Microsoft Forms tool. The questionnaire includes the following sections: • Section A: demographics including four variables (professional field, professional role, service years, education level). • Section B: Likert scale single-choice questions about possibilities of BEs that might occur during LNG bunkering SIMOPs. The questions were classified into seven subsections, in accordance with Table 1.
For the closed-ended questions, a seven-point Likert type scale was employed and anchored with a range from "very low" to" very high". The main reason for using seven descriptors is that humans' unidimensional judgment span is usually seven plus or minus two (i.e., five to nine) [46]. The questionnaire is designed to have mandatory and optional questions. The questions in G1 were set to be mandatory; the questions in G2-G7 were set to be optional, considering that different experts have different levels of familiarity with different ship types. See Document S1 available online.

Participants
Purposive sampling (also referred to as a judgmental or expert sample) was applied in this study to gain representative samples [47]. This sampling strategy enables the authors to utilize the participant's expertise and familiarity in this research field. The criterion for the selection of experts was that LNG bunkering-related knowledge or experience was necessary. All participants had expertise in LNG bunkering-related fields from shipping Data 2022, 7, 60 5 of 13 companies, energy companies, maritime safety administrations, port authorities, classification societies, ship design companies, shipyards, equipment manufacturers, and academia.

Sample Size
LNG bunkering is an emerging industry, and at the time of data collection, a conservative estimate stipulates that there are about 500 qualified experts in China. A sample size of 82 is recommended using the formula for the sample size in the literature [48]. In this calculation, the population size is taken as 500, the margin of error is taken as 5%, the confidence level is taken as 90% as a result of using the purposive sampling, and the sample proportion is taken as 90%.

Data Collection
A total of 152 invitations were distributed via emails and social media messaging apps including WeChat and WhatsApp. A total of 137 responses were received, as shown in Figure 1, where there is a reasonable distribution of experts from different sectors. The responses appeared to have higher percentage from energy companies, and a lower percentage from maritime safety administrations and port authorities. This is because experts from energy companies had accumulated more experience in LNG bunkering practices, whereas maritime safety administrations and port authorities focused more on safety management and had limited experience in the details of risk assessment. The respondents were 78% Chinese companies or institutes and 22% international companies or institutes based in China. It is believed that acquired data from the diversity of these affiliations ensured a better reflection of the actual situation of the LNG bunkering industry. The minimum number of responses to the individual questions for the optional part in the questionnaire was 97, which meets the sample size criterion described in Section 2.5.
ping companies, energy companies, maritime safety administrations, port auth classification societies, ship design companies, shipyards, equipment manufacture academia.

Sample Size
LNG bunkering is an emerging industry, and at the time of data collection, servative estimate stipulates that there are about 500 qualified experts in China. A s size of 82 is recommended using the formula for the sample size in the literature this calculation, the population size is taken as 500, the margin of error is taken as 5 confidence level is taken as 90% as a result of using the purposive sampling, and th ple proportion is taken as 90%.

Data Collection
A total of 152 invitations were distributed via emails and social media mes apps including WeChat and WhatsApp. A total of 137 responses were received, as in Figure 1, where there is a reasonable distribution of experts from different secto responses appeared to have higher percentage from energy companies, and a low centage from maritime safety administrations and port authorities. This is because e from energy companies had accumulated more experience in LNG bunkering pra whereas maritime safety administrations and port authorities focused more on management and had limited experience in the details of risk assessment. The re ents were 78% Chinese companies or institutes and 22% international companies o tutes based in China. It is believed that acquired data from the diversity of these tions ensured a better reflection of the actual situation of the LNG bunkering ind The minimum number of responses to the individual questions for the optional the questionnaire was 97, which meets the sample size criterion described in sub 2.5.

Data Analysis
This subsection presents the process of data analysis.

. Experts' Qualitative Expressions
A Likert scale is a psychometric scale that is used to represent people's opinions and attitudes on a topic or subject matter. When designing a Likert scale, the number of points on the scale must be specified. In this study, a seven-point Likert scale is used to represent the likelihood of a certain BE with 1 = "very low", 2 = "low", 3 = "fairly low", 4 = "medium", 5 = "fairly high", 6 = "high", 7 = "very high". An expert's expression of a certain question represents the occurrence possibility of the BE per bunkering operation (a whole bunkering process from start to finish).
There are two main reasons for using seven-point Likert scale in this study: (1) Miller concluded that humans' unidimensional judgment span is usually seven plusminus two, which means the suitable number of comparisons for a human to judge at a time is between five and nine [46], and seven is the median. (2) The IMO has introduced a seven by four risk matrix for formal safety assessment (FSA) for use in the maritime industry, it reflects seven potential variations for frequencies and four potential variations for consequences; therefore, using seven linguistic terms is in line with the habit of the experts in the maritime industry [49].

Converting the Experts' Qualitative Expressions into Quantitative Fuzzy Corresponding Numbers
The trapezoidal fuzzy number is used in this study whose membership functions are defined as Equation (1) [38], and is plotted in Figure 2.
where µ A is the membership function of the fuzzy set A.

Data Analysis
This subsection presents the process of data analysis.

Experts' Qualitative Expressions
A Likert scale is a psychometric scale that is used to represent people's opinions and attitudes on a topic or subject matter. When designing a Likert scale, the number of points on the scale must be specified. In this study, a seven-point Likert scale is used to represent the likelihood of a certain BE with 1 = "very low", 2 = "low", 3 = "fairly low", 4 = "medium", 5 = "fairly high", 6 = "high", 7 = "very high". An expert's expression of a certain question represents the occurrence possibility of the BE per bunkering operation (a whole bunkering process from start to finish).
There are two main reasons for using seven-point Likert scale in this study: (1) Miller concluded that humans' unidimensional judgment span is usually seven plusminus two, which means the suitable number of comparisons for a human to judge at a time is between five and nine [46], and seven is the median. (2) The IMO has introduced a seven by four risk matrix for formal safety assessment (FSA) for use in the maritime industry, it reflects seven potential variations for frequencies and four potential variations for consequences; therefore, using seven linguistic terms is in line with the habit of the experts in the maritime industry [49].

Converting the Experts' Qualitative Expressions into Quantitative Fuzzy Corresponding Numbers
The trapezoidal fuzzy number is used in this study whose membership functions are defined as Equation (1) [38], and is plotted in Figure 2.
where μ is the membership function of the fuzzy set .  [38].
A numerical approximation system is proposed to systematically convert linguistic terms into corresponding fuzzy numbers. Chen and Hwang proposed eight conversion scales (Scale one to Scale eight) which were based on previous extensive empirical studies on the use of linguistic terms [50]. In the present study, the Scale six scale, which includes seven linguistic terms, are adopted for estimating the occurrence possibilities of BEs as A numerical approximation system is proposed to systematically convert linguistic terms into corresponding fuzzy numbers. Chen and Hwang proposed eight conversion scales (Scale one to Scale eight) which were based on previous extensive empirical studies on the use of linguistic terms [50]. In the present study, the Scale six scale, which includes seven linguistic terms, are adopted for estimating the occurrence possibilities of BEs as shown in Figure 3 [50]. The effectiveness of the Scale six scale has been proven in the literature [51][52][53]. shown in Figure 3 [50]. The effectiveness of the Scale six scale has been proven in the literature [51][52][53].

Converting the Fuzzy Numbers about a Certain Question into an Aggregated Fuzzy Number
The geometric mean and the linear opinion pool method could be employed to aggregate experts' opinions. The linear opinion pool is adopted in this study due to its advantage of taking experts' weights into account [54]. The occurrence possibility of a certain BE is affected by various factors, for example, the reliability of equipment, the environmental conditions, safety culture of the industry, psychological factors of the operators, etc. In this paper, experts' engineering judgments are used to assess the possibility of occurrence by considering these factors; however, an expert's background and engineering experience determine his/her judgment, therefore, the weight of each expert must be considered.
Based on the linear opinion pool method [55], the aggregated fuzzy number can be expressed by Equation (2).
where is the weight given to the expert , and ∑ = 1 ; is a fuzzy number obtained from the expression of the expert about BE , is the total number of experts, whereas is the total number of BEs. For example, the 1st expert's linguistic expression on the 1st BE (BE1) is "very low" (see Figure 3), then is (0.0, 0.0, 0.1, 0.2), and the weight of the 1st expert to BE1 is 0.00657462, therefore, is (0.0, 0.0, 0.007, 0.0013). See Table S2 for the detailed calculation in the Supplemental Material available online.

Converting the Fuzzy Numbers about a Certain Question into an Aggregated Fuzzy Number
The geometric mean and the linear opinion pool method could be employed to aggregate experts' opinions. The linear opinion pool is adopted in this study due to its advantage of taking experts' weights into account [54]. The occurrence possibility of a certain BE is affected by various factors, for example, the reliability of equipment, the environmental conditions, safety culture of the industry, psychological factors of the operators, etc. In this paper, experts' engineering judgments are used to assess the possibility of occurrence by considering these factors; however, an expert's background and engineering experience determine his/her judgment, therefore, the weight of each expert must be considered.
Based on the linear opinion pool method [55], the aggregated fuzzy number P j can be expressed by Equation (2).
where W i is the weight given to the expert i, and ∑ n i=1 w i = 1; P ij is a fuzzy number obtained from the expression of the expert i about BE j, n is the total number of experts, whereas m is the total number of BEs. For example, the 1st expert's linguistic expression on the 1st BE (BE1) is "very low" (see Figure 3), then P 11 is (0.0, 0.0, 0.1, 0.2), and the weight of the 1st expert to BE1 is 0.00657462, therefore, P 1 is (0.0, 0.0, 0.007, 0.0013). See Table S2 for the detailed calculation in the Supplemental Material available online.
where WS i is the weight score of the expert i, WS i = PPS i + STS i + ELS i , PPS i , STS i , and ELS i represent the professional position score (PPS), the service time score (STS), and the education level score (ELS) of the expert i, respectively.  By defuzzification, the fuzzy occurrence possibilities of the BEs per operation can be obtained so that it is easy to handle the fuzzy ORs. The defuzzification of a trapezoidal fuzzy number (a, b, c, d) is given by Equation (4) based on the center of area method [63].
where FOP is the fuzzy occurrence possibility of the Bes per operation, and x COA is the abscissa of the center of area of the trapezoid as shown in Figure 2. Thereafter, the function developed by Onisawa is used for converting FOPs into fuzzy ORs per operation [36], in which the fuzzy OR can be expressed by Equations (5)- (7).
where K is a constant value, P SC is taken as 5 × 10 −3 which implies a safety criterion [35].

Data Description
The data were collected over three months, from 21 February 2021 to 21 May 2021. The raw data comprises four main parts, where details of the data are demonstrated in a sampled table for further explanation in the following sub-sections.
The questionnaire form, the raw data and obtained fuzzy ORs, and the calculation sheet for data analysis are published online at https://doi.org/10.5281/zenodo, accessed on 4 April 2022.

Demographic Information of the Experts
In the first part of the raw data, experts' demographic information is presented. Table 3 shows the demographic information and their data types. The first row shows the attributes' names, and the second row is the descriptions of the corresponding data, where expert E1 is taken as an example.

Expressions of the Experts to the Occurrence Possibilities of the BEs
In the second part of the raw data, the experts' expressions are presented. Table 4 shows the data types, where expert E1 is taken as an example.

Weight of Each Expert to Each BE
In the third part of the raw data, the weight of each expert to each BE (WBE) is presented in Table 5, where expert E1 is taken as an example.

Fuzzy Occurrence Rates of BEs
In the fourth part of the dataset, the fuzzy ORs per operation of the BEs are presented in Table 6.  Figure 4 presents the fuzzy ORs of the BEs per operation. These values could be directly used as input data when performing quantitative risk assessments of LNG bunkering SIMOPs.

User Notes
The existing numerical risk criteria in performing quantitative risk assessment uses "per year" as a unit. This is given in Table 7, which presents the IMO's individual risk

User Notes
The existing numerical risk criteria in performing quantitative risk assessment uses "per year" as a unit. This is given in Table 7, which presents the IMO's individual risk criteria [49,64]; therefore, the duration of each LNG bunkering operation in hours, and the number of LNG bunkering operations in a year for a specific LNG bunkering project, should be estimated prior to converting ORs per operation into occurrence probabilities per year. Thereafter, with the exponential distributions expressed by Equation (8), the annual occurrence probability for a BE can be calculated.
where OP is the annual occurrence probability of a BE, λ is the occurrence rate per hour of a BE, t is the operational hours in a year. For example, for BE 3 (Failure of LNG bunkering breakaway couplings), the OR per operation is 0.00123, if the duration of each LNG bunkering operation is 5 h, then λ = 0.00123 5 = 0.000246, if the operational hours in a year t = 124, then OP = 0.03, which means the annual failure probability of LNG bunkering breakaway couplings is 0.03. Table 7. The IMO's individual risk criteria [49,64].

Limitation
Firstly, in this investigation there is a limited source for data collection, and the reader should bear in mind that the study is based on experts from companies established in China. In spite of its limitations, the study adds to our understanding in terms of the quality required for generating a dataset of a dynamic risk assessment of an LNG SIMPOS operation. A natural progression of this work would be to add more data to this dataset obtained from experts in different region of the world.
Secondly, there are limitations to the purposive sampling method in this study. On one hand, to some extent, the investigators are making subjective judgements when choosing participants. On the other hand, it is hard to evaluate the reliability of the experts; this means it is difficult to determine if there is a sampling error in the information that investigators present. To minimize these limitations, the inclusion criteria for experts are defined for screening the eligibilities of experts.
Thirdly, the weighting criteria of experts adopted in this study have limitations. There are inhomogeneities of gaps between metrics, for example, the difference between a PhD and Masters might be larger than that between vocational education and high school. Thus, it deserves further study on quantifying the difference between metrics.
Supplementary Materials: The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/data7050060/s1, Document S1: The questionnaire form; Table S1: The raw data and obtained ORs; Table S2: The calculation sheet for data analysis. Informed Consent Statement: Informed consent was obtained from all subjects involved in the study.

Data Availability Statement:
The data used for this paper can be found at https://doi.org/10.5281/ zenodo.6527869, accessed on 4 April 2022.
Acknowledgments: The authors would like to express their gratitude to the experts who participated in the online questionnaire survey.

Conflicts of Interest:
The authors declare no conflict of interest.