A reliability evaluation method for submarine pipelines based on neutrosophic fuzzy Bayesian networks

Abstract

The present invention relates to a method for evaluating the reliability of a submarine pipeline based on a neutrosophic fuzzy Bayesian network, comprising the following steps: Step 1, construction of a Bayesian network: firstly analyzing a submarine pipeline failure accident, analyzing the root cause of system failure, identifying the fault event, and constructing a Bayesian network structure: converting expert comments into corresponding single-valued neutrosophic hesitant fuzzy elements; determining expert weights based on expert authority superiority and expert average opinion similarity, and aggregating opinions of different experts; converting the single-valued neutrosophic hesitant fuzzy elements obtained by aggregating expert opinions into a priori failure probabilities through defuzzification; Step 3, Bayesian network reasoning and sensitivity analysis; Step 4, system risk analysis and troubleshooting.

Description

Technical Field

[0001] The present invention relates to the technical field of submarine pipeline failure accident assessment, and in particular to a submarine pipeline failure accident risk assessment method based on a single-valued neutrosophic hesitant fuzzy Bayesian network. Background Art

[0002] With the increase in energy demand, the speed of marine oil and gas exploration is accelerating. Submarine pipelines are the main mode of transportation after submarine oil and gas exploration and development. However, due to water flow, wave scouring, third-party damage, earthquakes or design defects, the probability of leakage and failure of submarine pipelines is relatively high. Once a submarine pipeline leaks, it may cause serious fires and explosions due to the escalation of the accident, posing a threat to human safety and the environment. The risk of submarine pipeline leakage and failure cannot be eliminated, but it can be reduced through preventive measures. Therefore, the risk analysis of submarine pipeline failure accidents is very important for improving safety.

[0003] In the risk analysis of the system, it is necessary to use statistical methods and historical data to calculate the failure probability and reliability of the system. When historical data is insufficient or inaccurate, hesitant fuzzy theory provides an effective technology. The single-valued neutrosophic hesitant fuzzy linguistic element that takes into account uncertainty and hesitation can transform expert knowledge from qualitative language evaluation to relatively reliable failure probability and reliability. Bayesian networks have outstanding capabilities in dealing with complex causal relationships and uncertainty reasoning.

[0004] At present, since submarine pipeline leakage failure accidents with catastrophic consequences rarely occur, there is not enough data to estimate the probability of such events. In addition, for submarine pipeline leakage failure accidents, related research is mostly focused on the fault tree method, which still has limitations in representing the uncertainty of event logical relationships and probability updates. Summary of the invention

[0005] The purpose of the present invention is to provide a more practical and reasonable method for evaluating the reliability of submarine pipelines that can express uncertain information and perform probabilistic reasoning and updating. The technical solution is as follows:

[0006] A method for evaluating the reliability of a submarine pipeline based on a neutrosophic fuzzy Bayesian network comprises the following steps:

[0007] Step 1, construction of Bayesian network: first analyze the submarine pipeline failure accident, analyze the root cause of the system failure, identify the fault event, and complete the modeling of the submarine pipeline fault tree; convert the fault tree model into a Bayesian network structure, and construct the node logical relationship by converting it into a conditional probability table through the "AND / OR" logic gate, map the basic event to the root node, map the intermediate event to the intermediate node, and map the top event to the leaf node T;

[0008] Step 2, acquisition and conversion of evaluation data: convert expert comments into corresponding single-valued neutral-intelligent hesitant fuzzy elements; determine expert weights based on expert authority and similarity of average expert opinions, and aggregate the opinions of different experts; the single-valued neutral-intelligent hesitant fuzzy elements obtained by aggregating expert opinions are converted into prior failure probabilities through defuzzification;

[0009] Among them, the expert weight is determined based on the superiority of expert authority and the similarity of average expert opinions, and the method of aggregating the opinions of different experts includes:

[0010] (1) Calculate the expert's subjective weight vector;

[0011] (2) Based on the similarity of average opinions as the order-inducing variable, O'Hagan's entropy-orness optimization model is used to calculate the expert objective correlation weight vector;

[0012] (3) Using single-valued neutrosophic hesitant fuzzy mixed weighted geometric operators to aggregate expert opinions;

[0013] Step 3: Bayesian network reasoning and sensitivity analysis: Based on the constructed Bayesian network model, the prior probability and conditional probability of each root node are integrated, and the probability value of submarine pipeline failure accident is obtained through forward reasoning, and the reliability is calculated; the posterior probability of each basic event is inferred based on the reverse diagnosis technology of the Bayesian network; sensitivity analysis is carried out to clarify the impact of each fault event on the occurrence of the accident;

[0014] Step 4: System risk analysis and troubleshooting; predict the probability of risk events in advance through the forward reasoning results of the Bayesian network; use the posterior probability results obtained through the reverse diagnosis technology of the Bayesian network to determine the importance of risk factors in affecting the occurrence of accidents; use sensitivity analysis to determine key risk factors.

[0015] Furthermore, in step 1, the identified fault events include: setting the "submarine pipeline failure" event as the top event, and the submarine pipeline failure accident is mainly caused by "external factors" or "internal factors". The causes of "external factors" include "corrosion, external load, overhang phenomenon, natural disasters", and the causes of "internal factors" include "material defects, weld defects, auxiliary equipment failure". "Corrosion" is caused by "internal corrosion" or "external corrosion", and "overhang phenomenon" is caused by "insufficient burial depth" and "harsh environmental conditions", and the intermediate events are determined.

[0016] The technical effects of the present invention include: (1) Bayesian networks can more scientifically express the complex logical relationships between nodes in a complex system, and the constructed model is more in line with reality; (2) The introduction of single-valued neutral-intelligence hesitant fuzzy theory transforms expert knowledge from qualitative language evaluation into relatively reliable probabilistic data, and considers the uncertainty and hesitation of expert cognition to make the result more reasonable; (3) The combination of single-valued neutral-intelligence hesitant fuzzy language and Bayesian networks can more accurately express uncertain information and perform probabilistic reasoning and updating; (4) The bidirectional reasoning and sensitivity analysis technology of the Bayesian network is used to calculate the failure probability of the system, identify the key events of the system, and guide risk prevention and control to reduce the possibility of accidents. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 is a flow chart of the evaluation method of the present invention;

[0018] Figure 2 It is the Bayesian network model of submarine pipeline failure accident in the present invention;

[0019] Figure 3 It is the prior probability and posterior probability value of the basic event of submarine pipeline failure accident in the present invention. DETAILED DESCRIPTION

[0020] The specific structure and implementation process of this solution are described in detail below through specific embodiments and drawings.

[0021] like Figure 1 As shown, in one embodiment of the present invention, a method for evaluating the reliability of a submarine pipeline based on a single-valued neutrosophic hesitant fuzzy Bayesian network is disclosed, comprising the following steps:

[0022] Step 1: Construction of Bayesian network. First, analyze the submarine pipeline failure accident, analyze the root cause of the system failure, identify the fault event, and complete the modeling of the submarine pipeline fault tree; convert the fault tree model into a Bayesian network structure, and construct the node logical relationship by converting it into a conditional probability table through "AND / OR" logic gates.

[0023] Among them, through the analysis of submarine pipeline failure accidents, the "submarine pipeline failure" event is set as the top event, and the submarine pipeline failure accident is mainly caused by "external factors" or "internal factors". The causes of "external factors" include "corrosion, external load, overhang phenomenon, natural disasters", and the causes of "internal factors" include "material defects, weld defects, auxiliary equipment failures". "Corrosion" is caused by "internal corrosion" or "external corrosion", and "overhang phenomenon" is caused by "insufficient burial depth" and "harsh environmental conditions". Continue to analyze on the basis of the above intermediate events to identify all possible basic events (root nodes). The analysis results show that there are 24 basic events and 13 intermediate events in the submarine pipeline failure fault tree.

[0024] Table 1 Intermediate events of submarine pipeline failure accidents

[0025]

[0026] Table 2 Basic events of submarine pipeline failure accidents

[0027]

[0028]

[0029] Based on the established fault tree model, the submarine pipeline fault tree is mapped to a Bayesian network structure, and the logical relationship of the Bayesian network nodes is constructed by converting the "AND / OR" logic gate into a conditional probability table. A total of 24 basic events are mapped to 24 root nodes, 13 intermediate events are mapped to 13 intermediate nodes, and the top event is mapped to the leaf node T.

[0030] Step 2, acquisition and conversion of evaluation data. Domain experts evaluate the possibility of occurrence of basic events of submarine pipeline leakage failure through defined evaluation terms; convert expert comments into corresponding single-valued neutrosophic hesitant fuzzy elements; determine expert weights based on expert authority and similarity of expert average opinions, and aggregate the opinions of different experts; and convert the single-valued neutrosophic hesitant fuzzy elements obtained by aggregating expert opinions into prior probabilities through defuzzification.

[0031] The comment variable is divided into 7 levels, namely S = {s0 = very low, s1 = very low, s2 = low, s3 = medium, s4 = high, s5 = very high, s6 = very high}. According to the language term set S, the expert opinion in the form of single-valued neutrosophic hesitant fuzzy element can be obtained. are three sets of values ​​in [0, 1], representing the membership hesitation degree from x to n, the uncertain membership hesitation degree, and the non-membership hesitation degree respectively.

[0032] The expert weight is determined based on the superiority of expert authority and the similarity of expert average opinion, and the single-valued neutrosophic hesitant fuzzy mixed weighted geometric operator is used to aggregate the expert opinions. The main process is:

[0033] Step 2.1, calculate the expert's subjective weight vector. The subjective weight vector is obtained by measuring the relative authority of the expert through the fuzzy optimization theory. The expert's authority is measured from two aspects: one is hard indicators, including the expert's position, working time, educational background and age; the other is soft indicators, including practical experience and professional proficiency. The hard indicators are scored according to the scoring table, and the soft indicators are filled in by the expert himself in the form of a percentage system. Finally, the expert authority measurement matrix Y = [y ij ] m×n , where y ijis the score of the jth expert for the ith measurement indicator. After normalizing the matrix Y, the relative superiority matrix of the experts is obtained as R = [r ij ] m×n ,in sup(y ij ) and inf(y ij ) are the index values ​​y of different experts under the same measurement index. ij The maximum and minimum values ​​in r ij is the membership degree of expert j’s measurement index i to the best: 0≤r ii ≤1. According to the optimization principle of "the comprehensive minimum of the square of the weighted distance and the square of the weighted distance" in the comprehensive evaluation of system engineering, the objective function is established: Solution Get j The formula for calculating the optimal value of is:

[0034]

[0035] p is the distance parameter, p = 2 can be taken as the Euclidean distance; l i is the weight of index i. We can get the superiority vector v=(v1, v2, …, v n ), and normalize it to get the expert subjective weight vector λ = (λ1, λ2, ..., λ k ) T .

[0036] Step 2.2, calculate the expert objective correlation weight vector. The objective correlation weight vector is determined based on the similarity of the average opinion as a sequence induction variable, and a set of evaluation variables {a1, ..., a n}, and an associated weight vector in The order-induced variable is defined as:

[0037]

[0038] in, is the average value of each evaluation, and ε is a sufficiently small positive number to ensure that s i The denominator is not 0. s-index: {1, 2, ..., n} → {1, 2, ..., n} When the arrangement satisfies i = 1, 2, ..., n-1, s s-index(i) >s s-index(i+1) , that is, s s-index(i) is {s1,…,s n}, so that the new order of the average variables can be obtained. Then the orness value is selected, see Table 3, and the O'Hagan entropy-orness optimization model is used to determine the weight according to the position.

[0039] Table 3 Meaning of orness

[0040]

[0041] On this basis, set As an additional constraint of the entropy-orness optimization model, when α∈(0.5, 1), the closer the position is, the higher the weight is. The optimization model is established as:

[0042]

[0043]

[0044] Solve the entropy-orness optimization model, and the optimal solution is the objective association weight vector w = (w1, w2, ..., w k ) T .

[0045] Step 2.3, use the single-valued neutrosophic hesitant fuzzy mixed weighted geometric operator to aggregate the expert opinions. The adjusted single-valued neutrosophic hesitant fuzzy mixed weighted geometric operator is defined as:

[0046]

[0047] Among them, λ = (λ1, λ2,..., λ k ) T n j (j=1, 2, ..., k) weight vector, and w=(w1,w2,…,w k ) T is the association weight vector, ε: {1, 2, ..., k} → {1, 2, ..., k} is a permutation such that n j is the ε(j)th largest element in the set. And

[0048] The defuzzification process of the aggregated single-valued neutrosophic hesitant fuzzy element to failure probability is divided into two parts: the conversion of the single-valued neutrosophic hesitant fuzzy element to fuzzy probability and the conversion of fuzzy probability to failure probability. The single-valued neutrosophic hesitant fuzzy element is converted into a definite value through defuzzification. Fuzzy Probability Then convert FPS into failure probability FFP: FFP = 1 / 10 K , K = -(FPS×10-12)×0.5.

[0049] Step 3: Bayesian network reasoning and sensitivity analysis. Based on the constructed Bayesian network model, the prior probability and conditional probability of each root node are combined to obtain the probability value of submarine pipeline leakage failure accident through forward reasoning, and the reliability is calculated; the posterior probability of each basic event is inferred based on the reverse diagnosis technology of the Bayesian network; and sensitivity analysis is carried out to clarify the impact of each fault event on the occurrence of the accident.

[0050] Given the causes and evidence, Bayesian forward reasoning techniques can be used to calculate the probability of risk events or results and the reliability of the system. Given the prior probability distribution of the root node and the conditional probability distribution of other nodes, the joint probability distribution between nodes is given by the chain rule: X i The probability of The probability that the system works normally, that is, the reliability of the system, is: R=1-P(T=Occur).

[0051] When the highest accident risk occurs at a certain moment, the posterior probability of each risk factor can be inferred through the reverse diagnosis technology of the Bayesian network. The posterior probability of the event is estimated by Bayes' theorem. Given a new observation, that is, evidence E, X i The posterior probability under the condition of evidence E is:

[0052]

[0053] By using the prior probability φ(X i ) and the posterior probability Sensitivity analysis was performed, and the sensitivity value

[0054] Step 4: System risk analysis and troubleshooting. The probability of risk events can be predicted in advance through the forward reasoning results of the Bayesian network, and preventive measures can be taken as early as possible when the safety risk level is high. The posterior probability results obtained through the reverse diagnosis technology of the Bayesian network are used to determine the importance of risk factors in affecting the occurrence of accidents. The key risk factors are determined by sensitivity analysis, which is helpful for system risk troubleshooting.

[0055] This implementation method first analyzes the causes of submarine pipeline failure accidents, identifies fault events to construct a fault tree, and establishes a Bayesian network structure based on "AND / OR" logic gate transformation. Then, the language evaluation of the root node event by domain experts is converted into relatively reliable failure probability data through the single-valued neutrosophic hesitant fuzzy theory. Finally, the bidirectional reasoning and sensitivity analysis of the Bayesian network are used to obtain the failure probability of the system and the ranking of key events.

[0056] Compared with the prior art, this embodiment has the following beneficial effects:

[0057] (1) Bayesian networks can more scientifically express the complex logical relationships between nodes in complex systems, and the constructed models are more realistic;

[0058] (2) The single-valued neutrosophic hesitant fuzzy theory is introduced to transform expert knowledge from qualitative language evaluation into relatively reliable probabilistic data, and the uncertainty and hesitation of expert cognition are considered to make the results more reasonable;

[0059] (3) The combination of single-valued neutrosophic hesitant fuzzy language and Bayesian network can more accurately express uncertain information and perform probabilistic reasoning and updating;

[0060] (4) Use the bidirectional reasoning and sensitivity analysis techniques of Bayesian networks to calculate the failure probability of the system, identify the key events of the system, and guide risk prevention and control to reduce the possibility of accidents.

[0061] The following is a detailed description of the evaluation method of the present invention by taking a submarine pipeline failure accident as an example and combining with the accompanying drawings.

[0062] 1. Based on the identification and analysis of submarine pipeline leakage accidents, we build Figure 2 The Bayesian network model of the submarine pipeline leakage accident is shown.

[0063] 2. Invite three experts to judge the possibility of all root node events in the Bayesian network based on the defined language variables, and aggregate the opinions of different experts.

[0064] According to the fuzzy optimization theory, the expert subjective weight vector is obtained as λ = (0.54, 0.10, 0.36) T Based on the similarity of the average opinions among experts, the expert objective correlation weight vector can be determined. The orness value α = 0.7 is taken, and the expert opinions are aggregated using a single-valued neutral hesitant fuzzy mixed weighted geometric operator. And through the defuzzification technology, it is converted into a clear numerical fuzzy probability FPS, and further converted into the corresponding prior probability FFP. The aggregation results and probability conversion results are shown in Table 4.

[0065] Table 4 Expert opinion aggregation and transformation results

[0066]

[0067]

[0068] 3. Based on the constructed Bayesian network model, combined with the prior probability of each root node and the conditional probability table of other nodes, the probability value of submarine pipeline leakage failure accident can be obtained by forward reasoning P = 1.49 × 10 -2 , then the reliability of the submarine pipeline is R = 1-P (T = Occur) = 98.51 × 10-2 .

[0069] In the diagnostic analysis (reverse reasoning) of the Bayesian network, the posterior probability reflects the most likely cause of the system accident. Assuming that an accident occurs, the probability of occurrence of the leaf node T is instantiated as 1, that is, P(T=Occur)=1. Then, the posterior probability of the root node P(X i = Occur|T = Occur), the posterior probability value of the key event is as follows Figure 3 as shown in .

[0070] The sensitivity value Rov of each root node is calculated to determine the influence of the leaf node input on the root node output. The results of the key nodes that have a greater impact on the occurrence of submarine pipeline leakage failure accidents are shown in Table 5.

[0071] Table 5 Sensitivity values ​​of key nodes

[0072]

[0073] 4. The reverse reasoning results of the Bayesian network show that when a submarine pipeline leakage failure accident occurs, the root node event auxiliary equipment aging (X 23 ), artificial drilling and oil theft (X9) and seabed movement (X 17 ) has a larger posterior probability and is the most likely cause of the accident.

[0074] According to the sensitivity analysis results, typhoon (X 18 ), unscheduled cleaning (X3) and anchoring (X7) are the key factors that managers should focus on. In actual safety management, the purpose of managers' continuous risk control is to find out the key factors that have a greater impact on the occurrence of accidents, so that they can pay more attention to these key points in the management process, take more prevention and control measures, and reduce the probability of accidents as much as possible.

[0075] At this point, those skilled in the art should recognize that, although multiple exemplary embodiments of the present invention have been shown and described in detail herein, many other variations or modifications that conform to the principles of the present invention can still be directly determined or derived based on the content disclosed in the present invention without departing from the spirit and scope of the present invention. Therefore, the scope of the present invention should be understood and recognized as covering all these other variations or modifications.

Claims

1. A method for evaluating the reliability of a submarine pipeline based on a neutrosophic fuzzy Bayesian network, comprising the following steps: Step 1, construction of Bayesian network: first analyze the submarine pipeline failure accident, analyze the root cause of system failure, identify the fault event, and complete the modeling of the submarine pipeline fault tree; convert the fault tree model into a Bayesian network structure, and construct the node logical relationship by converting it into a conditional probability table through "AND / OR" logic gates, map the basic event to the root node, map the intermediate event to the intermediate node, and map the top event to the leaf node T; Step 2, acquisition and conversion of evaluation data: convert expert comments into corresponding single-valued neutral-intelligent hesitant fuzzy elements; determine expert weights based on expert authority and similarity of average expert opinions, and aggregate the opinions of different experts; the single-valued neutral-intelligent hesitant fuzzy elements obtained by aggregating expert opinions are converted into prior failure probabilities through defuzzification; in, The expert weight is determined based on the superiority of expert authority and the similarity of average expert opinions. The methods for aggregating the opinions of different experts include: (1) Calculate the expert's subjective weight vector; (2) Based on the similarity of average opinions as the order-inducing variable, O'Hagan's entropy-orness optimization model is used to calculate the expert objective correlation weight vector; (3) Using single-valued neutrosophic hesitant fuzzy mixed weighted geometric operators to aggregate expert opinions; Step 3: Bayesian network reasoning and sensitivity analysis: Based on the constructed Bayesian network model, the prior probability and conditional probability of each root node are integrated, and the probability value of submarine pipeline failure accident is obtained through forward reasoning, and the reliability is calculated; the posterior probability of each basic event is inferred based on the reverse diagnosis technology of the Bayesian network; sensitivity analysis is carried out to clarify the impact of each fault event on the occurrence of the accident; Step 4: System risk analysis and troubleshooting; use the forward reasoning results of the Bayesian network to predict the probability of occurrence of risk events in advance; use the posterior probability results obtained through the reverse diagnosis technology of the Bayesian network to determine the importance of risk factors in affecting the occurrence of accidents; use sensitivity analysis to determine key risk factors.

2. The method for evaluating the reliability of a submarine pipeline according to claim 1, characterized in that: In step 1, the identified fault events include: setting the "submarine pipeline failure" event as the top event, the submarine pipeline failure accident is mainly caused by "external factors" or "internal factors"; the causes of "external factors" include "corrosion, external load, overhang phenomenon, natural disasters", and the causes of "internal factors" include "material defects, weld defects, auxiliary equipment failure"; "corrosion" is caused by "internal corrosion" or "external corrosion", and "overhang phenomenon" is caused by "insufficient burial depth" and "harsh environmental conditions", and the intermediate events are determined.

Images