An n-valued neutrosophic set method for the assessment of an offshore oil spill risk

With the rapid expansion of the coastal oil reserve base, the wharf's oil throughput and wharf's ship numbers increase sharply, then the risk of oil spills rose and oil spill accidents occurred occasionally, which bring enormous pressure to large influence on human health, safety, and economy, tourism and trade of coastline et al. The hazards of oil spills are as follows: (1) when crude oil or oil products are leaked at sea, due to the different oil composition, an oil film of varying thickness will be formed on the sea surface, which will reduce the photosynthesis of phytoplankton and affect the oxygen content of seawater. (2) The oil spill will stick to the feathers of seabirds and make them lose their ability to keep warm and fly. Seabirds that cannot fly will eventually die due to drowning, hunger, or loss of body temperature. (3) Harmful substances contained in the oil spill will accumulate in fish, shrimp, shells, and other organisms, and continue to pass through the food chain. If people eat contaminated seafood, these harmful substances will harm human health. In addition, the oil spill will also cause different degrees of impact on fishing, tourism, and so on.

The consequences of oil spill accidents are very serious, and oil spill risk assessment has become a hot research in the field of marine environmental protection. Domestic and foreign scholars have adopted different methods to analyze their respective research subjects. Common risk analysis methods include the neural network method, grey system method, Bayesian network method, accident tree method, analytic hierarchy process, risk matrix method, multilevel fuzzy comprehensive evaluation method, dynamic Bayesian model and machine learning, and so on. Xiao et al. proposed a method for the identification of oil spill risk degree in ship areas (Xiao et al. 2002). Yan & Ma proposed a grey fuzzy comprehensive evaluation (Yan & Ma 2011). Al-Amin et al. developed a new risk model to evaluate the possibility of ship collision, sinking, grounding, and other accidents (Baksh et al. 2018). Ozkan did some analysis by using fault tree analysis (Özkan 2016). Wu improved the traditional artificial neural network and used the double-layer particle swarm algorithm to evaluate the ship risk and reduce the error (Wu 2016). Maisa et al. analyzed the ecological impact of oil spills by a probabilistic framework (Nevalainen et al. 2016). Moonjin et al. built a risk matrix to assess the worst-case scenario for oil and toxic and hazardous substances in Korea's future (Lee & Jung 2013). Singh et al. developed an oil potential risk model and used a space model to identify the key areas of potential oil spill risk (Singh et al. 2015). Sun and Han established the evaluation module about the marine single-ship oil spill at Yangshan port according to the risk value and divided the risk level into four grades, and then used Visual Basic computer programming language to develop corresponding software tools for quantifying the monitoring and evaluation of ships in Shanghai Yangshan Port (Sun & Han 2010). Jin et al. analyzed the tanker accident probability by using some machine learning methods (Jin et al. 2019). Liu et al. used the analytical hierarchy process to make a method for ship oil spill risk (Liu et al. 2017). Zhang et al. used the fuzzy evaluation in the assessment model about marine oil spill pollution (Zhang et al. 2016). Pan used the fuzzy evaluation in the risk assessment of the oil tanker spill (Pan 2020).

We have too much dependence on the objective phenomenon, objective data, and objective indicators in oil spill risk assessment methods. However, in the actual oil spill risk assessment, we need to consider various factors, especially some uncertain factors, such as crew responsibilities, crew business skills and experience, crew health, and maintenance status. Therefore, this paper introduced neutrosophic sets (NSs) (Zadeh 1965; Smarandache 1998), because NS not only has true and false components but also has indeterminable components. To introduce NS theory into engineering applications, a single-valued neutrosophic set (SVNS) was proposed (Wang et al. 2010). When the index has multiple levels, we should need NVNS (Smarandache 2013) to settle these problems. NVNS consists of three independent subjection functions and can quantify the qualitative evaluation, solve the problems of ambiguity, difficulty in quantification and uncertainty, and can improve the prediction accuracy. Now NVNS plays an important tool in many domains (Ye & Ye 2014; Said et al. 2015; Fan & Ye 2017; Smarandache 2017; Fan et al. 2018, 2019; Fan 2019). The evaluation indexes of offshore oil spill risk sources generally have multiple levels, so we can use NVNS to establish the risk assessment model and use the NVNWAA operator to objectively evaluate and analyze to provide some advice for offshore oil spill risk evaluation. The new method proposed in this paper can represent uncertainty evaluation, and consider the subjective expression of some important people and key experts in shipping, to achieve better control of risk assessment.

Other parts are as follows: Section 2 constructs an index system of risk assessment about the three risk sources. Section 3 describes some concepts. Section 4 defines the NVNWAA operator. Section 5 establishes offshore oil spill risk assessment methods. Section 6 presents an example of this method and gives some suggestions. Section 7 concludes this paper.

Risk identification in oil spill is to comprehensively consider the possible factors, and then select valuable influencing factors to establish an index system of risk assessment. The main risk sources of offshore oil spill include ship oil spill, offshore oil platforms, submarine pipelines, oil processing enterprises, and offshore oil transportation. We choose the first three risk sources for oil spill risk assessment.

Combining the port channel conditions, and the natural environment, the risk factors of ship oil spills are analyzed to obtain the main factors that may affect ship oil spill risk. Then, through consulting experts, captains, and other important ship personnel to adjust and modify the evaluation indicators, the factors that ultimately affect the risk of oil spill are human, ship, ship management, and environment. Table 1 lists some indexes.

The evaluation index system of oil spills on an offshore oil platform is shown in Table 2.

The evaluation index system of the submarine pipeline oil spill is shown in Table 3.

According to the common classification principle of oil spill risk assessment, five risk grades can be classified: minimal, minor, general, great, and extremely great, which are, respectively, represented by ranges between 0 and 1, and Table 4 lists the corresponding relationship.

The relations between Z and F are listed:

• Ⅰ ⁠, just ⁠, for i= 1, 2,…, m;

• Ⅱ

• Ⅲ

For convenience, NVNN is denoted by ⁠.

Proof:

• (a)

  

• (b)

  

• (c)

  

• (d)

  

Equation (9) is true based on (a)−(d).

Using the entropy weight method, each index's weight can be got by the following four steps.

Step 1: For m samples and n indexes, then is the value of the index b of sample a, if each index's normalized value is ⁠, then for the positive index, the normalization formula is ⁠.

While the indicators are negative, the normalization formula is ⁠.

Step 2: The information entropy ⁠, in which ⁠.

Step 3: According to the calculation formula of information entropy, the information entropy of each index is calculated as ⁠. Each index's weight can be calculated by ⁠.

Step 4: Calculate the composite index ⁠, in which express the previous level number, express the number of indexes of the previous level. By repeating steps 1 ∼ 4, we can calculate the weight of each index.

Some indexes are multilevel in risk assessment of offshore oil spills, thus, we can use NVNS to represent them. Assuming that an index set is ⁠, contains sub-indexes, denoted by a sub-index set ⁠. Then, each index can be expressed by the NVNN ⁠. Table 5 shows the NVNN assessment matrix.

Table 5

Assessment matrix

z1
z2
…	…
zm

z1
z2
…	…
zm

View Large

According to the assessment matrix of Table 3, we establish a risk assessment method, it has four steps:

Step 1: According to the information entropy, the sub-indexes’ weight vector and the indexes’ weight vector can be calculated.

Step 3: The score values can be calculated by Equation (6).

Step 4: According to Table 4, risk levels can be determined.

In this part, we simulated the basic data of three ships according to the risk assessment indexes to verify the feasibility of new methods. In the future, these methods can be tested in a more realistic environment. Here ship 1 to ship 3 are used to represent three different ships. Risk assessment index data of the three ship areas are shown in Table 6 under n-value neutrosophic information.

Table 7 lists the decision matrix.

Table 7

NVNN matrix of ship

.		.	.	.

.		.	.	.

View Large

In Table 7, we can see that every number is smaller than 1, so when we calculate the weight, we can skip Step 1 in Section 5.1, with Steps 2–4 in Section 5.1 we calculate the information entropy E, the results are shown in Table 8 (we only put T value to calculate).

expresses z₃ of S₁, expresses z₃ of S₂ and expresses z₃ of S₃. The weight vector of the sub-indexes of z₃ is 1.

By comparing with Table 4, we can conclude that ships 1 and 3 are at minor risk, and ship 3 is in the best condition. According to the historical data, the model results are the same as the real results. It is effective to verify the risk assessment with n-value neutrosophic comprehensive evaluation method, which can get an obvious effect and high credibility. The evaluation results can be applied to ship oil spill-related aspects.

Here P1 to P3 are used to represent three different oil platforms. Risk assessment index data of the three platforms are shown in Table 9 under n-value neutrosophic information.

Table 10 lists the decision matrix.

Table 10

NVNN matrix of offshore oil platform

.	.	.	.

.	.	.	.

View Large

In Table 10, we can see that every number is smaller than 1, so when we calculate the weight, we can skip Step 1 in Section 5.1, with Steps 2–4 in Section 5.1 we calculate the information entropy E, the results are shown in Table 11 (we only put T value to calculate).

By comparing with Table 4, we can conclude that platforms 2 and 3 are at minor risk, and platform 3 is in the best condition.

Here L1 to L3 are used to represent three different submarine pipelines. Risk assessment index data of the three submarine pipelines are shown in Table 12 under n-value neutrosophic information.

Table 13 lists the decision matrix.

Table 13

NVNN matrix of submarine pipeline

.		.	.	.

.		.	.	.

View Large

In Table 13, we can see that every number is smaller than 1, so when we calculate the weight, we can skip Step 1 in Section 5.1, with Steps 2–4 in Section 5.1 we calculate the information entropy E, the results are shown in Table 14 (we only put T value to calculate).

By comparing with Table 4, we can conclude that L1 and L3 are at minor risk, and L3 is in the best condition.

The leading cause of ship oil spill are as follows: In ship 1, the crews’ responsibility factor and ship factor. In ship 2, crews’ business skills and experience and natural environmental factors. In ship 3, crews’ physical condition, staffing and training, and management supervision and inspection.

Except for ship factors, it is found that in the correlation analysis results of the three ships, the first and second factors are always the responsibility of the crew and the business skills and experience of the crew are among human factors. The responsibility index of the crew is always the biggest factor affecting the oil spill risk value. Therefore, strengthening the control of human factors has a significant effect on reducing the risk of oil spill. Shipping companies should strengthen the quality education of seafarers and require them to fulfill relevant obligations while on duty. Technical training and psychological quality training of the crew should be strengthened to increase their responsibility. The health status of the crew should be concerned to ensure the health of the crew from the aspects of mental health and physical health, to avoid the oil spill risk caused by the health of crew. In the ship factor, strengthen the maintenance of the ship to improve the technical state of the ship. Among the environmental factors in the port, the navigation command of ships in the port should be strengthened to clear the port in time. Meanwhile, navigable conditions in the port should be designated to strengthen real-time observation of changes in the marine environment. Among the management factors, we should strengthen the training and management of the crew, formulate a perfect safety management system for the ship company, and at the same time, the management should always perform the duties of supervision and inspection.

The leading cause of offshore oil platform oil spill are as follows: In P1, viscosity factor in oil characteristic element and coastal area category factor in oil spill location element. In P2, toxicity, persistence and flammability factors in oil characteristic elements. In P3, surface flow direction and wave height factors in hydrology element.

It is found that in the correlation analysis results of the three offshore oil platforms, oil spill factors are closely related to oil characteristics. In order to effectively reduce the risk of oil spill, operators must strictly implement operating procedures and make thorough drilling plans before drilling. The platform should be equipped with safe and effective blowout prevention equipment and good kill materials and well-control equipment. At the same time, it is necessary to provide professional and technical training to operators in key positions, adhere to the certification of the job, and establish a sound well-control management system. It should also strengthen drilling observation and detect problems on time, implement effective control according to correct shut-in procedures, and organize well killing servicing timely. Setting up a fire extinguishing system and providing portable fire extinguishers in key places, developing a rigorous oil spill contingency plan, and responding in the event of a blowout.

Third-party damage, erosion and suspension, natural disasters and corrosion are the main causes of submarine pipeline leakage accidents. In order to reduce the risk of oil spill in the submarine pipelines, first improve the design and installation level of the submarine pipelines, select high-quality materials, optimize pipeline design, and formulate the correct construction schemes, strict operation management and other comprehensive measures. Second, improve the ability of emergency response to oil spill events: improve the ability to predict oil spill point, oil spill volume, and oil spill drift path. and strengthen the allocation capacity of petroleum emergency equipment. Again, use online detection technology to prevent the oil spill of submarine pipelines. Through regular detection of the possible leakage potential points of submarine pipelines, and timely maintenance measures are taken to reduce the risk of oil spill. Finally, define the use function of the sea area to ensure that anchoring and fishing operations are prohibited in the submarine oil pipeline area, so as to reduce the risk of submarine oil pipeline leakage caused by external forces.

How to determine the influencing factors and their weights is the key and difficult part when dealing with the problem of the offshore oil spill. In this paper, a set of evaluation index system for ship oil spill, offshore oil platform oil spill and submarine pipeline oil spill is constructed, and an n-valued neutrosophic comprehensive evaluation model is built. In addition, entropy is used for calculating the weight, and the oil spill risk values of each sample can be calculated by the NVNWAA operator. Finally, the risk grades of each sample are analyzed and some suggestions are given. Compared with the literature (Zhang et al. 2016; Liu et al. 2017; Pan 2020), we use a neutrosophic number to represent the evaluation data, which can better reflect the integrity of the evaluation data, because it can represent not only the membership degree value and non-membership degree value, but also the uncertainty value, so that our evaluation results are more accurate. However, in the literature (Zhang et al. 2016; Liu et al. 2017; Pan 2020), only membership degree value is used to represent the evaluation data, which will cause a little error in the experimental results. In addition, we use the entropy weight method to calculate the weight value, which is more accurate and objective, and can better explain the obtained results. The model provides the basis and support for the safety management of offshore oil spills, and can be used for risk assessment in the real environment while only adjusting the data according to the actual situation of the assessment object.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.