Method and system for establishing pipeline risk analysis model based on fuzzy bayesian network, and device

Abstract

This invention relates to a method, system, and equipment for establishing a pipeline risk analysis model based on fuzzy Bayesian networks, belonging to the field of pipeline risk analysis technology. The method for establishing the pipeline risk analysis model includes: constructing a pipeline failure analysis Bayesian network based on historical pipeline accident data and the Apriori algorithm; calculating the prior probabilities of basic risk factors based on expert opinions on basic pipeline risk factors and fuzzy comprehensive evaluation; obtaining the conditional probabilities of sub-nodes in the pipeline failure analysis Bayesian model based on expert pairwise evaluation of the relative importance of basic pipeline risk factors, as well as the analytic hierarchy process (AHP) and the ranking node method; and establishing the pipeline risk analysis model. This invention can calculate the failure probability of pipelines with complex multi-risk factors and identify key pipeline failure events. Based on the probabilities, it can assess the safety and reliability of pipeline systems, prevent accidents, reduce costs, and optimize resource allocation.

Description

Technical Field

[0001] This invention belongs to the field of pipeline risk analysis technology, and specifically relates to the method, system, and equipment for establishing a pipeline risk analysis model based on fuzzy Bayesian networks. Background Technology

[0002] Pipeline risk analysis is fundamental to improving pipeline transportation safety. Effective and scientific pipeline risk analysis helps identify pipeline risk indicators and is a guarantee of long-term pipeline safety. Existing pipeline risk analysis methods include the following:

[0003] First, real-time risk assessment factor data is acquired to obtain a real-time risk assessment factor set, which is then segmented. Based on the index data values ​​of each risk failure probability factor and the risk index scoring method, the index score of each risk failure probability factor is determined. Specifically, subjective weighting, objective weighting, and dynamic integrated weighting methods are used to calculate the dynamic integrated weight of the risk failure probability. Based on the dynamic integrated weight of the risk failure probability and the index scores of each risk failure probability factor, the probability of long-distance pipeline risk failure is calculated, and the consequences of risk failure are calculated to complete the pipeline risk assessment. This method combines expert scoring and grey relational analysis to calculate the subjective weights of the first and second level risk failure probability factors. However, the expert scoring method lacks a unified language, making it difficult to guarantee the accuracy of experts' consistent opinions. While the entropy weighting method is used to calculate objective weights, some first- and second-level indicators have non-linear relationships, and the entropy weighting method may not accurately reflect the weight of the indicators. Furthermore, this method requires a large amount of data, making accurate risk analysis difficult when historical data is insufficient.

[0004] Secondly, this method determines pipeline risk assessment indicators and calculates the weights of each factor within these indicators. Based on the indicators and their corresponding weights, it obtains the evaluation level corresponding to each pipeline risk element. This method considers the mutual influence between risk elements and between different levels, using triangular fuzzy numbers instead of specific numerical scales to represent the relative importance of two elements. Combined with fuzzy comprehensive evaluation technology, it constructs a complete subsea pipeline risk assessment system. However, this method relies on expert opinions and experience to determine weights through pairwise comparisons, resulting in high subjectivity. Pipeline risk analysis involves multiple indicators and levels, requiring numerous pairwise comparisons and a heavy workload for experts. Furthermore, while this method obtains the relative importance of pipeline risks through the analytic hierarchy process, it does not calculate the specific pipeline failure probability, leading to insufficient specificity and incomplete evaluation in the risk analysis.

[0005] Third, the method identifies risk sources for subsea pipelines; constructs a Bayesian network model of subsea pipeline risk accidents, and fuzzifies expert opinions; calculates the subjective and objective weights of each expert using the analytic hierarchy process (AHP) and the Pythagorean fuzzy weighting method, aggregates the weighted expert opinions based on the PTFEHG operator, calculates expert opinions under the influence of expert weights and location weights, and obtains the aggregated results of expert opinions; and uses a Bayesian network to calculate the failure probability, thus completing the pipeline risk analysis. This method mainly uses the Pythagorean fuzzy weighting method for risk weight calculation, but this method has high computational complexity and long computation time. Furthermore, the Pythagorean fuzzy weighting method assumes that the indicators are independent of each other and does not consider the correlation between indicators. In reality, there is a certain correlation between pipeline risk indicators, leading to deviations in weight allocation.

[0006] In view of the limitations of existing pipeline risk analysis methods, such as strong subjectivity, large workload, incomplete evaluation and lack of scientific rigor, there is an urgent need to propose an objective and scientifically complete pipeline risk analysis method. Summary of the Invention

[0007] To address the above problems, this invention provides a method, system, and equipment for establishing a pipeline risk analysis model based on fuzzy Bayesian networks.

[0008] The first objective of this invention is to provide a method for establishing a pipeline risk analysis model based on fuzzy Bayesian networks, comprising:

[0009] Based on historical pipeline accident data and the Apriori algorithm, a Bayesian network for pipeline failure analysis was built.

[0010] Based on expert opinions on the basic risk factors of pipelines and the fuzzy comprehensive evaluation method, the prior probabilities of the basic risk factors are obtained.

[0011] Based on the relative importance evaluation of basic risk factors for each pipeline by experts, as well as the analytic hierarchy process and the ranking node method, the conditional probabilities of sub-nodes in the Bayesian model for pipeline failure analysis are obtained.

[0012] A pipeline risk analysis model is established based on the prior probabilities of basic risk factors and the conditional probabilities of sub-nodes, as well as the Bayesian network for pipeline failure analysis.

[0013] In a specific embodiment of the present invention, the step of establishing a Bayesian model for pipeline failure analysis based on historical pipeline accident data and the Apriori algorithm includes:

[0014] Preprocessing, statistical analysis, classification analysis, and key field correlation analysis of historical pipeline accident data;

[0015] The Apriori algorithm was used to mine the potential patterns and hidden relationships of the causes of pipeline accidents from preprocessed historical pipeline accident data.

[0016] Based on the potential patterns and implicit relationships of pipeline accident causes, the interrelationships of pipeline accident causes and the basic risk factors of pipeline accidents are identified.

[0017] Classify failure types based on the interrelationships of causes of pipeline accidents;

[0018] A Bayesian model for pipeline failure analysis is constructed using basic risk factors as the root node, failure types as intermediate nodes, and pipeline failures as leaf nodes.

[0019] In a specific embodiment of the present invention, obtaining expert opinions on the basic risk factors of the pipeline includes:

[0020] Select experts from different business areas of the pipeline;

[0021] Based on the assessment of the probability of occurrence of the basic risk factors of the pipeline by the selected experts, expert opinions on the basic risk factors of the pipeline are obtained.

[0022] In a specific embodiment of the present invention, the step of obtaining expert opinions on the basic risk factors of the pipeline based on the assessment of the probability of occurrence of the basic risk factors by selected experts includes:

[0023] Based on the assessment of the probability of occurrence of the basic risk factors of the pipeline by the selected experts, the expert assessment opinions on the basic risk factors of the pipeline are obtained.

[0024] Using linguistic values ​​represented by triangular fuzzy functions and trapezoidal fuzzy number functions, a unified expression is provided for the expert assessment opinions on the basic risk factors of pipelines;

[0025] Based on the results of the unified presentation, the experts' second opinion on the basic risk factors of the pipeline was obtained.

[0026] By combining the preference ratio method and the consensus subjective weighting method, the expert opinions of the secondary opinions on the basic risk factors of the pipeline are weighted.

[0027] Based on the weighting of expert opinions, the summarized expert opinions are obtained, namely the expert opinions on the basic risk factors of the pipeline.

[0028] In a specific embodiment of the present invention, based on expert opinions on the basic risk factors of the pipeline and the fuzzy comprehensive evaluation method, the prior probabilities of the basic risk factors are calculated, including:

[0029] The fuzzy numbers corresponding to the expert opinions on the basic risk factors of the pipeline are converted into fuzzy probability scores representing the likelihood of events occurring.

[0030] Calculate the prior probability of basic risk factors based on the fuzzy probability score of the likelihood of an event occurring.

[0031] In a specific embodiment of the present invention, the formula for calculating the prior probability of the basic risk factor is as follows:

[0032]

[0033] in, FFP stands for fuzzy probability, i.e., the prior probability of the basic risk factors; FPS stands for fuzzy probability score, μ. A (u) is the membership function corresponding to the expert opinion, and c is a constant.

[0034] In a specific embodiment of the present invention, the process of obtaining the conditional probabilities of sub-nodes in the Bayesian model for pipeline failure analysis based on the relative importance evaluation of basic risk factors for each pair of pipelines by experts, and the analytic hierarchy process and the ranking node method, includes:

[0035] Based on the experts' evaluation of the relative importance of the basic risk factors of each pair of pipelines, a node importance assessment matrix is ​​formed;

[0036] Based on the node importance assessment matrix, the weights of each basic risk factor are calculated using the maximum eigenvalue method, which is the weight of each node in the Bayesian model for pipeline failure analysis.

[0037] Based on the obtained weights of each node, a corresponding mathematical expression is generated by combining the variance. The cumulative probability of each interval is calculated using the sorted node method to obtain the conditional probability of the child node.

[0038] In a specific embodiment of the present invention, before "calculating the weights of each basic risk factor using the maximum eigenvalue method based on the node importance evaluation matrix," the following step is also included:

[0039] The consistency of the node importance assessment matrix is ​​measured using the random consistency ratio.

[0040] Based on the results of the consistency assessment, the node importance evaluation matrix is ​​revised.

[0041] The second objective of this invention is to provide a pipeline risk analysis method based on fuzzy Bayesian networks, comprising:

[0042] Based on the event information of the identified basic risk factors and the pipeline risk analysis model described above, possible combinations of events leading to pipeline accidents are identified.

[0043] In a specific embodiment of the present invention, identifying possible combinations of events leading to pipeline accidents includes:

[0044] Using pipeline failure as an evidentiary event, and based on the event information of the basic risk factors, the pipeline risk analysis model described above updates the posterior probabilities of the basic risk factors.

[0045] Calculate the difference between the posterior probability and the prior probability of the basic risk factors;

[0046] Based on the magnitude of the difference, identify possible combinations of events that could lead to a pipeline accident.

[0047] The third objective of this invention is to provide a system for establishing a pipeline risk analysis model based on fuzzy Bayesian networks, comprising:

[0048] Building module: Used to establish a Bayesian model for pipeline failure analysis based on historical pipeline accident data and the Apriori algorithm;

[0049] Parent node module: used to calculate the prior probabilities of basic risk factors based on expert opinions on basic risk factors of pipelines and the fuzzy comprehensive evaluation method;

[0050] Sub-node module: Used for evaluating the relative importance of basic risk factors of pipelines in pairs based on experts, and for obtaining the conditional probability of sub-nodes in the Bayesian model of pipeline failure analysis using the analytic hierarchy process and the ranking node method.

[0051] The module is used to build a pipeline risk analysis model based on the prior probabilities of basic risk factors and the conditional probabilities of child nodes, as well as a Bayesian network for pipeline failure analysis.

[0052] The fourth objective of this invention is to provide a pipeline risk analysis system based on fuzzy Bayesian networks, comprising:

[0053] Calculation module: Used to update the posterior probability of the basic risk factors in the pipeline risk analysis model based on the event information of the basic risk factors in the above pipeline risk analysis model, taking pipeline failure as an evidentiary event; and calculate the difference between the posterior probability and the prior probability of the basic risk factors.

[0054] Analysis module: Used to identify possible combinations of events that could lead to a pipeline accident based on the magnitude of the difference.

[0055] A fifth object of the present invention is to provide an electronic device comprising: a processor coupled to a memory;

[0056] The memory is used to store computer programs;

[0057] The processor is configured to execute the computer program stored in the memory, so that the electronic device performs the method described above.

[0058] A computer-readable storage medium storing a program or instructions that, when executed on a computer, cause the computer to perform the methods described above.

[0059] The beneficial effects of this invention are:

[0060] The present invention relates to a pipeline risk analysis method, system, and equipment based on fuzzy Bayesian networks, which establishes a pipeline failure Bayesian network using the Apriori algorithm.

[0061] Based on fuzzy theory, the probability of basic events leading to pipeline failure is evaluated using fuzzy language. Fuzzy comprehensive evaluation is used to quantify and systematically analyze expert opinions, and the prior probability of basic risk factors (i.e., the parent node in the pipeline failure Bayesian network) is obtained.

[0062] Based on the key factors and logical relationships of pipeline failure, a hierarchical analysis structure for pipeline failure is established. Using the hierarchical analysis method, the conditional probabilities of sub-nodes in the Bayesian model for pipeline failure analysis are calculated.

[0063] Finally, based on the required prior probabilities, inter-node conditional probabilities, and the pipeline failure Bayesian network, a pipeline risk analysis model was established. This model diagnoses and analyzes possible events leading to pipeline accidents by continuously updating the node information (i.e., basic event information of basic risk factors) in the pipeline failure Bayesian model.

[0064] The method of this invention addresses the difficulties in risk analysis caused by insufficient accumulation of historical accident data for natural gas pipelines. It can calculate the failure probability of pipelines with complex and multi-risk factors and identify key failure events. Based on the probability, it assesses the safety and reliability of the pipeline system, preventing accidents, reducing costs, and optimizing resource allocation. Furthermore, this invention reduces the reliance on expert opinions and the subjectivity of conditional probability reasoning, effectively improving the accuracy of risk analysis results.

[0065] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures pointed out in the description, claims and drawings. Attached Figure Description

[0066] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0067] Figure 1 A flowchart illustrating a method for establishing a pipeline risk analysis model based on a fuzzy Bayesian network according to an embodiment of the present invention is shown.

[0068] Figure 2A diagram illustrating the pipeline failure hierarchy model according to an embodiment of the present invention is shown.

[0069] Figure 3 A Bayesian network diagram of pipeline failure according to an embodiment of the present invention is shown;

[0070] Figure 4 The diagram shows the function graphs of triangular fuzzy number and trapezoidal fuzzy number according to an embodiment of the present invention;

[0071] Figure 5 A diagram illustrating the main factors contributing to pipeline failure according to an embodiment of the present invention is shown.

[0072] Figure 6 A graph showing the difference between the prior and posterior probabilities of a basic event according to an embodiment of the present invention is shown.

[0073] Figure 7 A framework diagram of a pipeline risk analysis model based on a fuzzy Bayesian network according to an embodiment of the present invention is shown.

[0074] Figure 8 A framework diagram of a pipeline risk analysis system based on a fuzzy Bayesian network according to an embodiment of the present invention is shown.

[0075] Figure 9 A frame diagram of an electronic device according to an embodiment of the present invention is shown;

[0076] In the diagram: Module 10; Parent Node Module 20; Byte Node Module 30; Module 40; Calculation Module 50; Analysis Module 60; Electronic Device 300; Processor 301; Memory 302. Detailed Implementation

[0077] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0078] like Figure 1 As shown, a method for establishing a pipeline risk analysis model based on a fuzzy Bayesian network according to an embodiment of the present invention includes:

[0079] S1. Based on historical pipeline accident data and the Apriori algorithm, a Bayesian network for pipeline failure analysis is constructed.

[0080] S2. Based on expert opinions on the basic risk factors of the pipeline and the fuzzy comprehensive evaluation method, calculate the prior probabilities of the basic risk factors.

[0081] S3. Based on the relative importance evaluation of the basic risk factors of each pipeline in pairs by experts, and the analytic hierarchy process and the ranking node method, the conditional probabilities of the sub-nodes in the Bayesian model for pipeline failure analysis are obtained.

[0082] S4. Based on the prior probabilities of basic risk factors and the conditional probabilities of sub-nodes, as well as the Bayesian network for pipeline failure analysis, a pipeline risk analysis model is established.

[0083] In some embodiments of the present invention, step S1 includes:

[0084] Step S1-1: Preprocess, statistically analyze, classify, and perform key field correlation analysis on historical pipeline accident data, including:

[0085] Remove abnormal and useless information from historical pipeline accident data, preliminarily identify accident information, statistically analyze the frequency of accident causes, and summarize the main accident cause events.

[0086] Preliminary statistical analysis (using a section of the China-Myanmar gas pipeline in service as an example of historical pipeline accident data) categorizes the causes of pipeline failure into eight major factors: corrosion failure, equipment failure, excavation damage, misoperation, failure of pipeline or weld materials, damage by natural forces, other external forces, and other causes. The specific contents of each factor are shown in Table 1.

[0087] Table 1

[0088]

[0089] For example, in this embodiment of the invention, the PHMSA database is used to mine the accident causes of the corresponding pipe section field data, specifically including data preprocessing, statistical and classification analysis, and key field correlation analysis.

[0090] Step S1-2: Use the Apriori algorithm to mine the potential patterns and hidden relationships of the causes of pipeline accidents from the preprocessed historical pipeline accident data.

[0091] Steps S1-3: Based on the potential patterns and implicit relationships of pipeline accident causes, confirm the interrelationships of pipeline accident causes and the basic risk factors of pipeline accidents.

[0092] Step S1-4: Classify the failure types based on the interrelationships of the causes of pipeline accidents;

[0093] The specific process from steps S1-2 to S1-4 is exemplified as follows:

[0094] The Apriori algorithm is used to mine frequent itemsets in a hierarchical search order, generating (n+1) itemsets from n itemsets. The minimum support is set to 0.005, the minimum confidence to 0.2, and the boost to 1. All frequent itemsets that meet the requirements are simply counted iteratively. Then, based on the property that "all non-empty subsets of a frequent itemet must also be frequent itemsets," all subsets are tested. Pruning is performed by removing non-frequent itemsets to improve the efficiency of generating frequent itemsets.

[0095] By pruning, the causes of pipeline failure are simplified to corrosion, third-party damage, pipeline defects, misoperation, and natural disasters. The basic risk factors leading to pipeline failure are finally identified in Table 2 (meaning of pipeline failure event symbols).

[0096] Table 2

[0097]

[0098]

[0099] Based on the hierarchical relationship among pipeline failure factors, a pipeline failure hierarchy model was established, with pipeline failure (E) as the target layer, corrosion (M1), third-party damage (M2), pipeline defects (M3), misoperation (M4), and natural disasters (M5) as the criterion layers, and basic events X1 to X30 as the scheme layers. See details... Figure 2 .

[0100] Steps S1-5: Construct a Bayesian network for pipeline failure analysis, using basic risk factors as root nodes, failure types as intermediate nodes, and pipeline failures as leaf nodes. For example:

[0101] According to the pipeline failure hierarchy model (i.e.) Figure 2 Using basic risk factors as root nodes, failure types as intermediate nodes, and pipeline failures as leaf nodes, a Bayesian network for pipeline failure is constructed. See details below. Figure 3 .

[0102] In some embodiments of the present invention, obtaining expert opinions on the basic risk factors of the pipeline includes:

[0103] A1. Select experts from different business areas of the pipeline;

[0104] A2. Based on the selected experts' assessment of the probability of occurrence of the basic risk factors of the pipeline, expert opinions on the basic risk factors of the pipeline are obtained, including:

[0105] A2-1. Based on the selected experts' assessment of the probability of occurrence of basic pipeline risk factors, the expert assessment opinions on the basic pipeline risk factors are obtained, for example:

[0106] Experts from different business areas such as pipeline design, installation, operation and maintenance and management are selected to assess the likelihood of the occurrence of basic pipeline risk factors based on their experience;

[0107] More specifically: Eight experts from enterprises were invited to assess the probability of the basic event occurring, taking into account their position, working hours, education level, and age. The level of expert judgment was evaluated, and experts with higher scores were given higher weights. The expert scoring criteria are shown in Table 3.

[0108] Table 3

[0109]

[0110]

[0111] The weight of each expert is calculated using the ratio allocation method and formula (1).

[0112]

[0113] In equation (1), WS j W(Eu) and W(Eu) are the weight value and weight factor of expert j, respectively. The collected expert-related information and the calculated expert weights are shown in Table 4.

[0114] Table 4

[0115]

[0116]

[0117] A2-2. Using triangular fuzzy functions and trapezoidal fuzzy number functions to represent linguistic values, a unified expression of expert assessment opinions on basic pipeline risk factors is provided to prevent experts from using vague language such as "very low," "low," "high," and "quite high" to describe the probability of events. For details on triangular fuzzy number functions, please refer to [link to documentation]. Figure 4 For details on trapezoidal fuzzy number functions, see [reference 1]. Figure 4 b;

[0118] For example: The correspondence between fuzzy linguistic values ​​and fuzzy sets is shown in Table 5;

[0119] Table 5

[0120] Language value meaning Corresponding fuzzy set AL Almost impossible (0,0,0.1,0.2) QL It is unlikely to happen (0.1,0.2,0.2,0.3) L The event is unlikely to occur, but it is possible in very rare circumstances. (0.2,0.3,0.3,0.4) ML The probability of the possible event occurring is between low and medium. (0.3,0.4,0.4,0.5) M The probability of it happening is equal to the probability of it not happening. (0.4,0.5,0.5,0.6) MH The probability of the event occurring is between high and medium. (0.5,0.6,0.6,0.7) H The event is very likely to happen. (0.6,0.7,0.7,0.8) QH The probability of the event occurring is very high. (0.7,0.8,0.8,0.9) AH Events always happen (0.8,0.9,1,1)

[0121] A2-3. Based on the results of the unified expression, the expert secondary opinions on the basic risk factors of the pipeline are obtained. For example, the fuzzy language of the expert secondary opinions on basic events collected based on the results of the unified expression is shown in Table 6.

[0122] Table 6

[0123]

[0124]

[0125] A2-4. Combining the preference ratio method and the consensus subjective weight method, the expert opinions of the secondary opinions on the basic risk factors of the pipeline are weighted.

[0126] The reasons for performing steps A2-4 are: first, there are differences in the abilities of experts, so it is necessary to assign weights to expert opinions to further reduce the subjective factors in the analysis process; second, to eliminate discrete opinions caused by expert negligence or other reasons, further reducing the subjective factors in the analysis process.

[0127] Among them, the preference ratio method is used to assign weights to expert opinions. The preference ratio method is as follows: the level of experts is evaluated from the aspects of position, working hours, education level and age. Experts are scored according to their personal information based on the expert scoring criteria, and experts with higher scores are given higher weights.

[0128] The consensus degree subjective weighting method is applied to eliminate the discrete opinions of experts due to negligence or other reasons. The consensus degree subjective weighting method is as follows: to ensure the consistency of opinions, give higher weights to the opinions of experts with higher consistency, and perform consistency processing on the expert weights.

[0129] The combined application of the preference ratio method and the consensus subjective weight method, that is, to obtain the comprehensive weight based on the preference ratio method and the consensus subjective weight method, to assign weights to different expert opinions, and to summarize and process all expert opinions;

[0130] Example:

[0131] The consensus-based subjective weighting method is adopted, and weights are assigned according to the consistency of opinions. It is assumed that the fuzzy language values ​​given by any pair of experts u and v are... and Convert it into a trapezoidal fuzzy number. and The similarity of opinions between expert u and expert v is calculated using equation (2). The value ranges from 0 to 1, and the closer it is to 1, the higher the similarity between the opinions of the two experts.

[0132]

[0133] The average consistency AA of the opinion given by expert u compared with the opinions given by other experts is calculated according to formula (3), and the relative consistency RA of expert u with other experts is further calculated using formula (4).

[0134]

[0135] Taking basic event X1 as an example, the average consistency AA and relative consistency RA are calculated as shown in Table 7.

[0136] Table 7

[0137]

[0138]

[0139] The weights are obtained by combining the ratio allocation method and the consensus degree subjective weight method according to formula (5) to finally determine the weights of each expert's opinion.

[0140] CC(Eu)=β×W(Eu)+(1-β)×RA(Eu) (5)

[0141] In equation (5), β is a relaxation factor representing the importance of expert opinions, which is set to 0.5.

[0142] A2-5. Based on the weighting results of the expert opinions, the summarized expert opinions, namely the expert opinions on the basic risk factors of the pipeline, are obtained as follows:

[0143] Based on the weighting of expert opinions obtained by formula (5), the opinions of all experts are summarized and processed according to formula (6).

[0144]

[0145] In equation (6), This is the corresponding fuzzy value converted from the given fuzzy language value.

[0146] In some embodiments of the present invention, step S2 includes:

[0147] S2-1. The fuzzy numbers corresponding to the expert opinions of the basic risk factors of the pipeline are converted into fuzzy probability scores (FPS) representing the probability of the event. Commonly used methods in the conversion process include the maximum-minimum Delphi, data statistical algorithm, linear method and centroid method (COA). The maximum-minimum Delphi, data statistical algorithm, linear method and centroid method are commonly used methods in this technical field, and will not be described in detail in this example of the present invention.

[0148] For example, the centroid method is used in the transformation process to transform the fuzzy numbers corresponding to the expert opinions of the basic risk factors of the pipeline into fuzzy probability scores representing the possibility of the event. Since the centroid method can reduce the degree of information loss and make the analysis results more reasonable and reliable, the centroid method is preferred for defuzzification. The calculation formula is shown in Equation (8).

[0149]

[0150] In equation (8): FPS is the fuzzy probability fraction, μA (u) is the triangular fuzzy membership function and trapezoidal fuzzy number membership function corresponding to the expert opinion, calculated according to equations (9) and (10).

[0151]

[0152] S2-2. Calculate the prior probability of the basic risk factor based on the fuzzy probability score of the event's occurrence. The formula for calculating the prior probability of the basic risk factor is shown in equation (11):

[0153]

[0154] In equation (11), FFP stands for fuzzy probability, i.e., the prior probability of the basic risk factors; FPS stands for fuzzy probability score, μ. A (u) is the membership function corresponding to the expert opinion, and c is a constant. For example, c takes the value 2.301.

[0155] The prior probabilities of the basic risk factors are detailed in Table 8.

[0156] Table 8

[0157]

[0158]

[0159] In some embodiments of the present invention, step S3 includes:

[0160] S3-1. Based on the experts' evaluation of the relative importance of basic risk factors in each pair of pipelines, a node importance assessment matrix is ​​formed, specifically as follows:

[0161] When inviting experts to evaluate the relative importance of basic risk factors for each pipeline, to reduce the difference in standards used by different experts, a unified relative importance evaluation standard was developed for experts (see Table 9), using a scale of 1-9 to represent relative importance.

[0162] Table 9

[0163] <![CDATA[Scale a i,j > meaning 1 This indicates that the two factors are of equal importance (equal in importance). 3 This indicates that, compared to the latter, the former is slightly more important (stronger). 5 This indicates that, compared to the latter, the former is significantly more important (stronger). 7 This indicates that, compared to the latter, the former is significantly more important (stronger) than the latter. 9 This indicates that, compared to the latter, the former is extremely more important (absolutely stronger). 2,4,6,8 This represents the intermediate value of the above adjacent judgments. 1,1 / 2,1 / 3,…,1 / 9 <![CDATA[If factor x i and x j have an importance ratio of a i,j , then x j and x i have an importance ratio of 1 / a i,j >

[0164] Based on the relative importance of the basic risk factors of each pipeline evaluated by experts, an evaluation matrix was obtained. The general formula of the evaluation matrix is ​​detailed in Equation (12).

[0165]

[0166] Before proceeding to step S3-3, the following operations may also be performed, by way of example:

[0167] B1. Use the random consistency (i.e. consistency test coefficient CR) ratio to measure the consistency of the node importance judgment matrix, so as to overcome the subjective and one-sided influence that experts cannot avoid when making comparisons. The formula for calculating the consistency test coefficient CR is detailed in Equation (13).

[0168]

[0169] In equation (13), RI and n are the average consistency index value and the order of the matrix, respectively, and RI is calculated according to equation (14).

[0170]

[0171] In equation (14), λ max It is the largest eigenvalue.

[0172] The specific value of the average consistency index RI is selected based on the order of the evaluation matrix, as shown in Table 10.

[0173] Table 10

[0174]

[0175] B2. Based on the results of the consistency assessment, revise the node importance judgment matrix. That is, based on the relationship between the consistency index value and the preset value, if the consistency index value is less than the preset value, the judgment matrix is ​​considered to satisfy the consistency test; if the consistency index value is greater than or equal to the preset value, the judgment matrix is ​​considered to need to be revised.

[0176] For example, if the preset value is set to 0.1, then:

[0177] When CR < 0.1, the judgment matrix is ​​considered to satisfy the consistency test.

[0178] If CR≧0.1, then the judgment matrix needs to be corrected. Specifically, the judgment matrix is ​​modified based on the logical errors in it.

[0179] Steps S3-1 and B1-B2, exemplarily:

[0180] Taking the basic events X1 to X7 that cause corrosion as examples, the evaluation matrix A given by expert 1 is as follows: k =[a ij ] 7×7 ;

[0181] According to the formula in Table 10, the average consistency index RI of the evaluation matrix given by expert 1 is obtained. The RI is then substituted into formulas (13)-(14) to calculate the consistency test coefficient of the evaluation matrix given by expert 1, as shown in Table 11.

[0182] Table 11

[0183]

[0184]

[0185] The data in Table 11 shows that the CR value satisfies the condition that CR is less than 0.1, indicating that the evaluation matrix meets the consistency test and no correction is required.

[0186] S3-2. Based on the node importance assessment matrix, the weights of each basic risk factor are calculated using the maximum eigenvalue method, i.e., the weights of each node in the Bayesian model for pipeline failure analysis, including:

[0187] Based on the node importance evaluation matrix A k Using the largest eigenvalue (λ) max The method calculates the weight of each factor, λ. max The corresponding eigenvector P is calculated as shown in equation (15).

[0188] A k P = λ max P (15)

[0189] The largest eigenvalue λ max The corresponding eigenvector P is normalized to obtain the importance weights W of each factor in expert opinion k. k .

[0190] If there are multiple experts, the weight of the event is calculated according to equation (16).

[0191]

[0192] In equation (16): λ max It is the evaluation matrix A k Maximum eigenvalue, W k It is the judgment matrix A k The eigenvector corresponding to the largest eigenvalue.

[0193] Step S3-2, for example, the weights of each factor determined based on expert opinions are detailed in Table 12.

[0194] Table 12

[0195]

[0196]

[0197] S3-3. Based on the obtained node weights, generate corresponding mathematical expressions using the variance, and calculate the cumulative probability of each interval using the sorted node method to obtain the conditional probability of the child nodes, including:

[0198] The pipeline node status is associated with ordered scale intervals: the interval corresponding to pipeline failure is [0.5, 1], and the interval corresponding to non-failure is [0, 0.5].

[0199] Parent node X i Node state corresponding interval x i It cannot be directly substituted into the weight expression (i.e., equation (11)) for calculation in the state interval x i Take s equally spaced sampling points, and use these sampling points as node state values. The same state can be divided into s categories due to different sampling point selections. n Different function values.

[0200] The probability density distribution of an event is represented by a double-truncated normal distribution. The double-truncated normal distribution on [0,1] is represented by TNormal(μ,σ). 2 ,0,1) represents, where μ is the mean of the function, σ 2 It's the variance. Adjust σ 2 The value of σ is given by the double-truncated normal distribution, which can represent various probability distributions. Considering the pipeline failure probability distribution, σ... 2 Take 0.01.

[0201] To obtain a probability density distribution that conforms to the actual situation, the mean value μ of the double-truncated normal distribution is calculated based on the node weights, and the weighted average value (WMEAN) is obtained using equation (17).

[0202]

[0203] In equation (17): w i For parent node X i The weights, i.e., the weights of each node obtained in step S3-2 (in Table 12); z i,k For parent node X i Node state corresponding interval x i The value of a certain sampling point.

[0204] When the probability distribution is represented by a double-truncated normal distribution, the child node state z is represented by the cumulative probability of the corresponding interval. c The probability of occurrence, also known as the conditional probability, accumulates to s depending on the selection of sampling points. n The expression value of a certain node state is shown in equation (18).

[0205]

[0206] For s n kind Take the average value to obtain the state of the same parent node (x1…x). n The child node state z under the given conditions cThe probability of an event, i.e. the conditional probability of the child node state, is expressed in equation (19).

[0207]

[0208] In some embodiments of the present invention, step S4 is to obtain a pipeline risk analysis model based on the prior probability of the parent node (i.e., the prior probability of the basic risk factors), the conditional probability of the child node, and the pipeline failure analysis Bayesian network established in step S1.

[0209] In some embodiments of the present invention, the prior probabilities of the basic risk factors obtained in step S2 (see Table 8) and the conditional probabilities of the child nodes obtained in step S3 are imported into the Bayesian network for pipeline failure analysis established in step S1, and the pipeline failure probability is calculated to be 2.487 × 10⁻⁶. -4 times / (km·a);

[0210] This value matches the statistically obtained pipeline failure probability. Subsequently, based on the prior probabilities of the basic risk factors, the main factors leading to the failure of this pipeline section were analyzed, and the proportions of each factor contributing to pipeline failure were obtained (see [reference needed]). Figure 5 ;

[0211] This demonstrates that the pipeline risk analysis model established in this embodiment of the invention can accurately calculate the pipeline failure probability and analyze the main factors contributing to pipeline failure.

[0212] In some embodiments of the present invention, the pipeline risk analysis method includes:

[0213] Step C: Based on the event information of the identified basic risk factors and the pipeline risk analysis model described above, identify possible combinations of events that could lead to a pipeline accident.

[0214] In some embodiments of the present invention, step C includes:

[0215] C-1. Using pipeline failure as the evidentiary event, based on the event information of the identified basic risk factors (such as the known certainty of heavy rain or a future earthquake), and the pipeline risk analysis model in the above embodiments, update the posterior probability of the basic risk factors, i.e.:

[0216] Using pipeline failure as an evidentiary event, the probability of pipeline failure is set to 1. Based on the hierarchical relationship between failure and risk factors in the Bayesian network (and equations (11)-(19)), and the event information of the determined basic risk factors, the update is calculated in reverse to obtain the posterior probability of the basic risk factors.

[0217] C-2. Calculate the difference between the posterior probability and the prior probability of the basic risk factors. For example, see... Figure 6 ;

[0218] C-3. Based on the magnitude of the difference, identify possible combinations of events leading to pipeline accidents, specifically:

[0219] Events with a difference greater than a threshold are identified as potential events leading to pipeline accidents. The threshold can be set flexibly, and this invention does not impose specific limitations on it.

[0220] Alternatively, all differences can be compared together, and the basic events with larger differences can be identified as possible events leading to pipeline accidents. The present invention does not specifically limit how many basic events with larger differences are selected as possible events leading to pipeline accidents, that is, the specific number of basic events in the combination of possible events leading to pipeline accidents is not specifically limited.

[0221] In actual operation, the threshold value is set according to the safety level of the actual pipeline section, or the number of specific events that may lead to pipeline accidents is set according to the combination of possible events.

[0222] For example, according to Figure 6 The possible events that could lead to the pipeline accident were identified as X1 (water flow erosion corrosion), X15 (material defects), X16 (structural defects), X18 (laying and installation defects), and X19 (welding defects).

[0223] like Figure 7 As shown, a system for establishing a pipeline risk analysis model based on a fuzzy Bayesian network according to an embodiment of the present invention includes:

[0224] Module 10: Used to build a Bayesian network for pipeline failure analysis based on historical pipeline accident data and the Apriori algorithm;

[0225] Parent node module 20: Used to calculate the prior probability of basic risk factors based on expert opinions on basic risk factors of pipelines and fuzzy comprehensive evaluation method;

[0226] Sub-node module 30: Used for evaluating the relative importance of basic risk factors of pipelines in pairs based on experts, and for obtaining the conditional probability of sub-nodes in the Bayesian model of pipeline failure analysis using the analytic hierarchy process and the ranking node method.

[0227] Module 40 is used to establish a pipeline risk analysis model based on the prior probabilities of basic risk factors and the conditional probabilities of child nodes, as well as a Bayesian network for pipeline failure analysis.

[0228] like Figure 8 As shown, a pipeline risk analysis system based on fuzzy Bayesian networks according to an embodiment of the present invention includes:

[0229] Calculation module 50: Used to update the posterior probability of the basic risk factors in the pipeline risk analysis model based on the event information of the basic risk factors in the pipeline risk analysis model, taking pipeline failure as an evidentiary event; also used to calculate the difference between the posterior probability and the prior probability of the basic risk factors.

[0230] Analysis module 60: Used to identify possible combinations of events that could lead to a pipeline accident based on the magnitude of the difference.

[0231] like Figure 9 As shown, in some embodiments of the present invention, an electronic device is provided, the electronic device 300 including: a processor 301 coupled to a memory 302;

[0232] The memory 302 is used to store computer programs;

[0233] The processor 301 is configured to execute the computer program stored in the memory 302, so that the electronic device performs the method described in the above embodiments.

[0234] In some embodiments of the present invention, a computer-readable storage medium is provided that stores a program or instructions that, when executed on a computer, cause the computer to perform the methods described in the above embodiments.

[0235] According to embodiments of the present invention, the computer-readable storage medium may be a non-volatile computer-readable storage medium, such as including, but not limited to: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In the present invention, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, electronic device, or apparatus.

[0236] Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for establishing a pipeline risk analysis model based on fuzzy Bayesian networks, characterized in that, include: Based on historical pipeline accident data and the Apriori algorithm, a Bayesian network for pipeline failure analysis was built. Based on expert opinions on the basic risk factors of pipelines and the fuzzy comprehensive evaluation method, the prior probabilities of the basic risk factors are obtained. Based on the relative importance evaluation of basic risk factors for each pipeline by experts, as well as the analytic hierarchy process and the ranking node method, the conditional probabilities of sub-nodes in the Bayesian model for pipeline failure analysis are obtained. A pipeline risk analysis model is established based on the prior probabilities of basic risk factors and the conditional probabilities of sub-nodes, as well as the Bayesian network for pipeline failure analysis.

2. The method for establishing a pipeline risk analysis model based on fuzzy Bayesian networks according to claim 1, characterized in that, The Bayesian model for pipeline failure analysis, based on historical pipeline accident data and the Apriori algorithm, includes: Preprocessing, statistical analysis, classification analysis, and key field correlation analysis of historical pipeline accident data; The Apriori algorithm was used to mine the potential patterns and hidden relationships of the causes of pipeline accidents from preprocessed historical pipeline accident data. Based on the potential patterns and implicit relationships of pipeline accident causes, the interrelationships of pipeline accident causes and the basic risk factors of pipeline accidents are identified. Classify failure types based on the interrelationships of causes of pipeline accidents; A Bayesian model for pipeline failure analysis is constructed using basic risk factors as the root node, failure types as intermediate nodes, and pipeline failures as leaf nodes.

3. The method for establishing a pipeline risk analysis model based on fuzzy Bayesian networks according to claim 1, characterized in that, Obtaining expert opinions on the basic risk factors of the pipeline includes: Select experts from different business areas of the pipeline; Based on the assessment of the probability of occurrence of the basic risk factors of the pipeline by the selected experts, expert opinions on the basic risk factors of the pipeline are obtained.

4. The method for establishing a pipeline risk analysis model based on fuzzy Bayesian networks according to claim 3, characterized in that, The expert opinions on the basic risk factors of the pipeline, based on the assessment of the probability of occurrence by selected experts, include: Based on the assessment of the probability of occurrence of the basic risk factors of the pipeline by the selected experts, the expert assessment opinions on the basic risk factors of the pipeline are obtained. Using linguistic values ​​represented by triangular fuzzy functions and trapezoidal fuzzy number functions, a unified expression is provided for the expert assessment opinions on the basic risk factors of pipelines; Based on the results of the unified presentation, the experts' second opinion on the basic risk factors of the pipeline was obtained. By combining the preference ratio method and the consensus subjective weighting method, the expert opinions of the secondary opinions on the basic risk factors of the pipeline are weighted. Based on the weighting of expert opinions, the summarized expert opinions are obtained, namely the expert opinions on the basic risk factors of the pipeline.

5. The method for establishing a pipeline risk analysis model based on fuzzy Bayesian networks according to claim 1, characterized in that, Based on expert opinions on the basic risk factors of pipelines and using the fuzzy comprehensive evaluation method, the prior probabilities of the basic risk factors are calculated, including: The fuzzy numbers corresponding to the expert opinions on the basic risk factors of the pipeline are converted into fuzzy probability scores representing the likelihood of events occurring. Calculate the prior probability of basic risk factors based on the fuzzy probability score of the likelihood of an event occurring.

6. The method for establishing a pipeline risk analysis model based on fuzzy Bayesian networks according to claim 5, characterized in that, The formula for calculating the prior probability of the basic risk factors is as follows: in, FFP stands for fuzzy probability, i.e., the prior probability of the basic risk factors; FPS stands for fuzzy probability score, μ. A (u) is the membership function corresponding to the expert opinion, and c is a constant.

7. The method for establishing a pipeline risk analysis model based on fuzzy Bayesian networks according to any one of claims 1-6, characterized in that, The expert-based pairwise evaluation of the relative importance of basic pipeline risk factors, along with the analytic hierarchy process (AHP) and the ranking node method, yields the conditional probabilities of sub-nodes in the Bayesian model for pipeline failure analysis, including: Based on the experts' evaluation of the relative importance of the basic risk factors of each pair of pipelines, a node importance assessment matrix is ​​formed; Based on the node importance assessment matrix, the weights of each basic risk factor are calculated using the maximum eigenvalue method, which is the weight of each node in the Bayesian model for pipeline failure analysis. Based on the obtained weights of each node, a corresponding mathematical expression is generated by combining the variance. The cumulative probability of each interval is calculated using the sorted node method to obtain the conditional probability of the child node.

8. The method for establishing a pipeline risk analysis model based on fuzzy Bayesian networks according to claim 7, characterized in that, Before "calculating the weights of each basic risk factor using the maximum eigenvalue method based on the node importance assessment matrix," the following is also included: The consistency of the node importance assessment matrix is ​​measured using the random consistency ratio. Based on the results of the consistency assessment, the node importance evaluation matrix is ​​revised.

9. A pipeline risk analysis method based on fuzzy Bayesian networks, characterized in that, include: Based on event information of identified basic risk factors and the pipeline risk analysis model as described in any one of claims 1-8, possible combinations of events leading to pipeline accidents are identified.

10. The pipeline risk analysis method based on fuzzy Bayesian networks according to claim 9, characterized in that, The identification of possible combinations of events leading to pipeline accidents includes: Using pipeline failure as an evidentiary event, the posterior probability of the basic risk factors is updated based on the event information of the identified basic risk factors and the pipeline risk analysis model described in any one of claims 1-8. Calculate the difference between the posterior probability and the prior probability of the basic risk factors; Based on the magnitude of the difference, identify possible combinations of events that could lead to a pipeline accident.

11. A system for establishing a pipeline risk analysis model based on fuzzy Bayesian networks, characterized in that, include: Building module: Used to establish a Bayesian model for pipeline failure analysis based on historical pipeline accident data and the Apriori algorithm; Parent node module: used to calculate the prior probabilities of basic risk factors based on expert opinions on basic risk factors of pipelines and the fuzzy comprehensive evaluation method; Sub-node module: Used for evaluating the relative importance of basic risk factors of pipelines in pairs based on experts, and for obtaining the conditional probability of sub-nodes in the Bayesian model of pipeline failure analysis using the analytic hierarchy process and the ranking node method. The module is used to build a pipeline risk analysis model based on the prior probabilities of basic risk factors and the conditional probabilities of child nodes, as well as a Bayesian network for pipeline failure analysis.

12. A pipeline risk analysis system based on fuzzy Bayesian networks, characterized in that, include: Calculation module: used to update the posterior probability of the basic risk factors in the pipeline risk analysis model according to any one of claims 1-8, taking pipeline failure as an evidentiary event, and the event information of the basic risk factors in the pipeline risk analysis model according to any one of claims 1-8; and to calculate the difference between the posterior probability and the prior probability of the basic risk factors. Analysis module: Used to identify possible combinations of events that could lead to a pipeline accident based on the magnitude of the difference.

13. An electronic device, characterized in that, include: Processor, the processor being coupled to memory; The memory is used to store computer programs; The processor is configured to execute the computer program stored in the memory to cause the electronic device to perform the method as described in any one of claims 1 to 10.

14. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a program or instructions that, when executed on a computer, cause the computer to perform the method as described in any one of claims 1 to 10.

Images