Physical Fusion Bayesian Pipeline Leakage Evaluation Method, Device, Equipment and Medium

Abstract

The present application discloses a pipeline leakage evaluation method, device, equipment and medium based on physical fusion Bayesian, which relates to the fields of pipeline transportation and risk assessment. It includes constructing a leakage fault tree model and a leakage consequence event tree model of a transportation pipeline, and generating a bow-tie model by using the leakage fault tree model and the leakage consequence event tree model; combining the bow-tie model with a Bayesian network model and performing synchronous mapping to obtain a mapped Bayesian network model; training the mapped Bayesian network model by using a neural network based on physical information to obtain a dynamic Bayesian network model; using the dynamic Bayesian network model to perform probability calculation on the transportation pipeline to obtain the pipeline risk leakage probability, and performing pipeline leakage evaluation on the transportation pipeline based on the pipeline risk leakage probability. The present application can reduce the complexity of pipeline leakage evaluation, improve the accuracy and efficiency of pipeline leakage evaluation, so as to formulate more effective safety measures and risk management strategies subsequently.

Description

Technical Field

[0001] The present invention relates to the fields of pipeline transportation and risk assessment, and particularly to a pipeline leakage assessment method, device, equipment and medium that physically integrates Bayesian methods. Background Art

[0002] In daily life and industrial production, pipeline transportation plays a crucial role. The advantages of pipeline transportation are its large transportation volume, easy control during transportation, low cost, simple construction, and being unaffected by factors such as the ground and climate, with strong environmental adaptability. However, pipelines often have a large number of pipeline leakage problems due to factors such as aging, corrosion, weld defects, and third-party damage. Currently, domestic and foreign scholars have conducted extensive research on pipeline risk assessment. To measure pipeline leakage risk, the fault tree or analytic hierarchy process is often combined with the fuzzy method to obtain a quantitative risk result through causal analysis, achieving the quantification of pipeline risk. However, for very complex systems, the compiled fault tree will be very large, which brings certain difficulties to qualitative and quantitative analysis and requires very rich experience, thus reducing the accuracy of pipeline leakage assessment.

[0003] As can be seen from the above, how to reduce the complexity of pipeline leakage assessment, improve the accuracy and efficiency of pipeline leakage assessment, so as to formulate more effective safety measures and risk management strategies in the future is an issue to be solved in this field. Summary of the Invention

[0004] In view of this, the purpose of the present invention is to provide a pipeline leakage assessment method, device, equipment and medium that physically integrates Bayesian methods, which can reduce the complexity of pipeline leakage assessment, improve the accuracy and efficiency of pipeline leakage assessment, so as to formulate more effective safety measures and risk management strategies in the future. The specific solutions are as follows:

[0005] In a first aspect, the present application discloses a pipeline leakage assessment method that physically integrates Bayesian methods, including:

[0006] Construct a leakage fault tree model and a leakage consequence event tree model for the transportation pipeline, and generate a bow-tie model using the leakage fault tree model and the leakage consequence event tree model;

[0007] Combine the bow-tie model with a Bayesian network model and perform synchronous mapping to obtain the mapped Bayesian network model;

[0008] Train the mapped Bayesian network model using a neural network based on physical information to obtain a dynamic Bayesian network model;

[0009] Use the dynamic Bayesian network model to perform probability calculations on the pipeline to obtain the pipeline risk leakage probability, and evaluate the pipeline leakage based on the pipeline risk leakage probability.

[0010] Optionally, the construction of the leakage fault tree model and leakage consequence event tree model of the pipeline includes:

[0011] Taking pipeline leakage as the top event, identify the causes of pipeline risk leakage, and construct the leakage fault tree model of the pipeline;

[0012] Based on the top event, combine the preset accident consequence types, and use the triggering conditions of different accidents to determine the safety barriers, so as to construct the leakage consequence event tree model of the pipeline.

[0013] Optionally, the combination of the bow-tie model and the Bayesian network model and synchronous mapping to obtain the mapped Bayesian network model includes:

[0014] Convert each event, accident consequence, and safety barrier in the bow-tie model into corresponding nodes in the Bayesian network model to obtain each Bayesian node;

[0015] Convert the logical relationship in the bow-tie model into the conditional probability table of the Bayesian nodes, assign values, and connect each Bayesian node with a directed edge to obtain the mapped Bayesian network model.

[0016] Optionally, before training the mapped Bayesian network model using the neural network based on physical information, it further includes:

[0017] Calculate the prior probability of the root node of the mapped Bayesian network model;

[0018] Calculate the prior probability according to Bayes' theorem and using the occurrence probability calculation formula to obtain the occurrence probability of the leaf node;

[0019] Calculate the prior probability and the occurrence probability using the posterior probability calculation formula to obtain the posterior probability.

[0020] Optionally, the occurrence probability calculation formula is:

[0021]

[0022] Where P(Y|X n ) is the occurrence probability of any leaf node, Y is the leaf node, is the sum of the occurrence probabilities of the leaf nodes corresponding to all the root nodes under the occurrence condition, P(X i) is the prior probability of the root node;

[0023] The formula for calculating the posterior probability is:

[0024]

[0025] where P(Z) is the pipeline leakage probability, P(Z|X i ) is the pipeline leakage probability under the condition that the event leading to the leakage risk occurs, and P(X i |Z) is the posterior probability that may cause the leakage under pipeline leakage.

[0026] Optionally, training the mapped Bayesian network model using the physics-informed neural network includes:

[0027] Calculating the root mean square error loss function and the physical part loss function, and calculating the total loss function based on the root mean square error loss function and the physical part loss function;

[0028] Training the mapped Bayesian network model using the physics-informed neural network based on the total loss function, the prior probability, the occurrence probability, and the posterior probability.

[0029] Optionally, the pipeline leakage evaluation of the transportation pipeline based on the pipeline risk leakage probability includes:

[0030] Judging whether there is a leakage situation in the transportation pipeline based on the pipeline risk leakage probability;

[0031] If there is a leakage situation in the transportation pipeline, determining the pipeline leakage risk factor, and performing pipeline leakage evaluation on the transportation pipeline according to the pipeline leakage risk factor.

[0032] In a second aspect, the present application discloses a pipeline leakage evaluation device based on physical fusion Bayesian, including:

[0033] A model construction module, configured to construct a leakage fault tree model and a leakage consequence event tree model of a transportation pipeline, and generate a bow-tie model by using the leakage fault tree model and the leakage consequence event tree model;

[0034] A model combination and mapping module, configured to combine the bow-tie model with a Bayesian network model, and perform synchronous mapping to obtain the mapped Bayesian network model;

[0035] A model training module, configured to train the mapped Bayesian network model by using a physics-informed neural network to obtain a dynamic Bayesian network model;

[0036] A pipeline leakage evaluation module is used to perform probability calculations on the transportation pipeline using the dynamic Bayesian network model to obtain the pipeline risk leakage probability, and to evaluate the pipeline leakage of the transportation pipeline based on the pipeline risk leakage probability.

[0037] In a third aspect, the present application discloses an electronic device, including:

[0038] A memory for storing a computer program;

[0039] A processor for executing the computer program to implement the aforementioned pipeline leakage evaluation method of physical fusion Bayesian.

[0040] In a fourth aspect, the present application discloses a computer storage medium for storing a computer program; wherein, when the computer program is executed by a processor, the steps of the aforementioned disclosed pipeline leakage evaluation method of physical fusion Bayesian are implemented.

[0041] It can be seen that the present application provides a pipeline leakage evaluation method of physical fusion Bayesian, including constructing a leakage fault tree model and a leakage consequence event tree model of a transportation pipeline, and generating a bow-tie model using the leakage fault tree model and the leakage consequence event tree model; combining the bow-tie model with a Bayesian network model and performing synchronous mapping to obtain the mapped Bayesian network model; training the mapped Bayesian network model using a neural network based on physical information to obtain a dynamic Bayesian network model; performing probability calculations on the transportation pipeline using the dynamic Bayesian network model to obtain the pipeline risk leakage probability, and evaluating the pipeline leakage of the transportation pipeline based on the pipeline risk leakage probability. The present application generates a bow-tie model using a leakage fault tree model and a leakage consequence event tree model, combines the bow-tie model with a Bayesian network model, and performs synchronous mapping to obtain the mapped Bayesian network model, which can support probability calculations of the Bayesian network at different time nodes, trains the mapped Bayesian network model using a neural network based on physical information, thereby realizing the prediction of the leakage of a long-distance transportation pipeline for volatile and flammable liquids and the accurate identification of leakage risk factors. By using machine learning algorithms, a large amount of data can be analyzed and processed, thereby constructing a more accurate and reliable dynamic Bayesian network model, performing probability calculations using the dynamic Bayesian network model, and thus evaluating the pipeline leakage of the transportation pipeline, being able to identify the key factors leading to leakage, reducing the complexity of pipeline leakage evaluation, and helping to formulate more effective safety measures and risk management strategies in the future. Description of the Drawings

[0042] To more clearly illustrate the technical solutions in the embodiments of the present invention or in the prior art, the following will briefly introduce the accompanying drawings required for the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are only the embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can also be obtained based on the provided drawings.

[0043] Figure 1 Flowchart of a pipeline leakage evaluation method based on physical fusion Bayesian disclosed in this application;

[0044] Figure 2 Bayesian network diagram of a long-distance pipeline leakage of flammable and volatile liquids disclosed in this application;

[0045] Figure 3 Training process and PINN model structure diagram disclosed in this application;

[0046] Figure 4 Graph of the dynamic change of leakage probability disclosed in this application;

[0047] Figure 5 Posterior probability graph of basic events disclosed in this application;

[0048] Figure 6 Another flowchart of a pipeline leakage evaluation method based on physical fusion Bayesian disclosed in this application;

[0049] Figure 7 Example diagram of a leakage fault tree model for a long-distance pipeline of flammable and volatile liquids disclosed in this application;

[0050] Figure 8 Example diagram of a leakage event tree model for a long-distance pipeline of flammable and volatile liquids disclosed in this application;

[0051] Figure 9 Structure diagram of a pipeline leakage evaluation device based on physical fusion Bayesian disclosed in this application;

[0052] Figure 10 Structure diagram of an electronic device provided in this application. Detailed implementation manners

[0053] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0054] In daily life and industrial production, pipeline transportation plays a crucial role. The advantages of pipeline transportation are its large transportation volume, easy control during the transportation process, low cost, simple construction, and being unaffected by factors such as the ground and climate, with strong environmental adaptability. However, pipelines often have a large number of pipeline leakage problems due to factors such as aging, corrosion, weld defects, and third-party damage. At present, scholars at home and abroad have conducted extensive research on pipeline risk assessment. To measure the pipeline leakage risk, the fault tree or the analytic hierarchy process is often combined with the fuzzy method, and quantitative risk results are obtained through causal analysis to achieve the quantification of pipeline risk. However, for very complex systems, the compiled fault tree will be very large, which brings certain difficulties to qualitative and quantitative analysis and requires very rich experience, thus reducing the accuracy of pipeline leakage assessment. As can be seen from the above, how to reduce the complexity of pipeline leakage assessment, improve the accuracy and efficiency of pipeline leakage assessment, so as to formulate more effective safety measures and risk management strategies subsequently is an issue to be solved in this field.

[0055] See Figure 1 As shown, the embodiment of the present invention discloses a pipeline leakage assessment method integrating physics and Bayesian, which specifically may include:

[0056] Step S11: Construct a leakage fault tree model and a leakage consequence event tree model of the transportation pipeline, and generate a bow-tie model by using the leakage fault tree model and the leakage consequence event tree model.

[0057] Step S12: Combine the bow-tie model with the Bayesian network model and perform synchronous mapping to obtain the mapped Bayesian network model.

[0058] Specifically, each event, accident consequence, and safety barrier in the BT (Bow Tie) are transformed into corresponding nodes in the BN (Bayesian network). When there are multiple identical events, only one Bayesian node needs to be established; the logical relationship in the BT is transformed into a conditional probability table of the nodes for assignment, and directed edges are used to connect the nodes to represent the mutual relationship between the nodes.

[0059] The Bow Tie (Bow Tie analysis method) used in the present invention is initially used for risk analysis in the form of a bow tie based on the "Tripod Beta Models". The Bow Tie method is mainly used for risk assessment, risk management, accident investigation and analysis, risk auditing, etc. It can better illustrate the situation of specific risks to help people understand the risk system and the prevention and control measure system. In the Bow Tie analysis model, the analysis of causes (the left side of the bow tie) and consequences (the right side of the bow tie) is combined to conduct a detailed analysis of an event with safety risks (referred to as the top event, the center of the bow tie). The relationship between an accident (top event), the causes of the accident, the ways leading to the accident, the consequences of the accident, and the measures to prevent the accident is represented by drawing a bow tie diagram. A Bayesian network, also known as a belief network or a directed acyclic graphical model, is a probabilistic graphical model that connects variables or propositions with causal relationships (or non-conditionally independent) using arrows (i.e., the arrow connecting two nodes represents that these two random variables have a causal relationship or are non-conditionally independent). A Bayesian network consists of two parts: one is an undirected, acyclic, and connected graph, where each node represents one or more random variables; the other part is the probability information associated with each variable. The arrows in the graph represent the dependence relationships between these variables. Each node in the network has a CPT (conditional probability table) used to represent the probability distribution of a random variable given the states of its parent nodes. When the parent nodes of a node are observed and set, the probability distribution of this node will change. Such a network can be used in many fields, including fault diagnosis, image recognition, information retrieval, natural language processing, medical diagnosis, etc. A Bayesian network can also obtain the unknown conditional probability distribution through learning and be used for prediction and inference. Therefore, combining the BT model and the BN model and performing synchronous mapping is an effective means for pipeline leakage risk assessment, and a mapped Bayesian network model is obtained.

[0060] Step S13: Use a neural network based on physical information to train the mapped Bayesian network model to obtain a dynamic Bayesian network model.

[0061] In this embodiment, based on the above-mentioned BT model for long-distance transportation pipelines of flammable and volatile liquids, the risk assessment indicators and safety barriers for long-distance transportation pipelines of flammable and volatile liquids are used as nodes, and a Bayesian network diagram for long-distance transportation pipelines of flammable and volatile liquids is constructed in combination with the logical relationships between the nodes, as Figure 2 shown. After constructing the Bayesian network model, the fuzzy scientific scores of each expert evaluation are determined according to the expert scoring. Further, the results of the expert scoring are used as the prior probabilities of the root nodes ( Figure 2Substitute all the basic event data into the Bayesian network to calculate the leakage failure probability and risk of the long-distance pipeline for volatile and flammable liquids. The Bayesian network is the application and extension of Bayes' theorem. It is a directed acyclic graph that reflects the correlation between complex variables in a specific problem and is a network model based on probability theory. In analyzing the potential associations of events, the Bayesian network has its unique advantages and application scenarios.

[0062] The specific training process is as follows: Calculate the prior probability of the root node of the mapped Bayesian network model; Calculate the prior probability according to Bayes' theorem and using the occurrence probability calculation formula to obtain the occurrence probability of the leaf node; Calculate the prior probability and the occurrence probability using the posterior probability calculation formula to obtain the posterior probability.

[0063] After obtaining the posterior probability, calculate the root mean square error loss function and the physical part loss function, and calculate the total loss function based on the root mean square error loss function and the physical part loss function; Train the mapped Bayesian network model using a neural network based on physical information based on the total loss function, the prior probability, the occurrence probability, and the posterior probability.

[0064] Specifically, the occurrence probability calculation formula is:

[0065]

[0066] where, P(Y|X n ) is the occurrence probability of any leaf node, Y is the leaf node, is the sum of the occurrence probabilities of the leaf nodes corresponding to all the root nodes under the occurrence condition, P(X i ) is the prior probability of the root node;

[0067] The posterior probability calculation formula is:

[0068]

[0069] where, P(Z) is the pipeline leakage probability, P(Z|X i ) is the pipeline leakage probability under the condition that the event leading to the leakage risk occurs, P(X i |Z) is the posterior probability that may lead to the leakage under the pipeline leakage.

[0070] According to the occurrence probability calculation formula and the prior probability of the root node, the probabilities of all leaf nodes in the Bayesian network can be obtained. The conditional probability table is constructed using the prior probability of the root node and the probabilities of all leaf nodes. By using the reverse inference function of the BN model, the posterior probability is obtained from the prior probability and the conditional probability table, and the change in the leakage probability of the pipeline in the high-consequence area caused by a small change in the risk factor is calculated. Furthermore, it can be determined which types of risk factors are the main controlling factors leading to pipeline leakage. The occurrence probability of an event varies at different times, which will affect the development trend of the accident and its consequences. Therefore, the static Bayesian network is extended by adding a time dimension and incorporated with PINN (Physics-informed Neural Network) for training to form a DBN (Dynamic Bayesian Network). Different times are used as the input layer, and the training process and the PINN model structure are as Figure 3 shown.

[0071] If only weights and biases are used to process the input data, the output value is a linear combination of the input values. No matter how complex the neural network is, it belongs to a linear model and has a general approximation ability. Therefore, the sigmoid activation function is introduced into the fully connected layer to enable more complex non-linear relationships between neurons. The expression of the sigmoid activation function is as follows:

[0072]

[0073] where w is the weight and b is the bias parameter.

[0074] The prior probability prediction value is obtained by making assumptions about the parameters in the activation function, and combined with the true value p of the prior data obtained from expert scoring i The loss function of the data-driven part is calculated. The expression of the root mean square error loss function is as follows:

[0075]

[0076] Traditional neural networks based purely on data only consider the data level when calculating the loss function, that is, the mean square error between the predicted value and the true value. However, the PINN adopted in the present invention adds physical equations as constraints to the neural network to make the training results satisfy physical laws, and adds the difference before and after the iteration of the physical equations to the loss function of the neural network, allowing the physical equations to also participate in the training process. The physical equations followed are as follows:

[0077] f(P 1 (t), P 2 (t),..., P i (t)) = 0;

[0078] where Pi (t) is the prior probability, and f(·) is the function of the physical equation followed by the training.

[0079] The physical part of the loss function is obtained using the physical equation as follows:

[0080] L(P i (t)) = f(P 1 (t), P 2 (t),..., P i (t));

[0081] In summary, the total loss function is the sum of the two parts of the loss function:

[0082]

[0083] After calculating the loss function, the gradient descent method is used for iterative solution step by step to obtain the minimized loss function and the model parameter values. Subsequently, different times are input into the trained model to output the prior probabilities at different times, and then the posterior probabilities are calculated using the occurrence probability calculation formula and the posterior probability calculation formula, thereby forming a dynamic Bayesian network integrating PINN.

[0084] Taking the prior probabilities of the root nodes at different times obtained by a specific expert diagnosis as an example, the prior probability at the first time point is shown in Table 1:

[0085] Table 1

[0086]

[0087] The prior probability at the second time point is shown in Table 2:

[0088] Table 2

[0089]

[0090]

[0091] The prior probability at the third time point is shown in Table 3:

[0092] Table 3

[0093]

[0094] After PINN training, the parameters closest to the true prior probability are obtained, that is, w takes 3.56, b takes 20, and c takes 6. Substituting them into the fourth time node, the prior probability at time node 4 is obtained using the trained PINN model, as shown in Table 4:

[0095] Table 4

[0096]

[0097] Taking the root node Y1 and its child node Y as an example, the probabilities of the child nodes are obtained using the occurrence probability calculation formula as follows:

[0098] P(Y = 1|Y 1 = 1) = 0.89;

[0099] P(Y = 1|Y 1 = 0) = 0.19;

[0100] 1 indicates that the event occurs, and 0 indicates that the event does not occur. Similarly, the probabilities of other leaf nodes can be calculated. Table 5 shows the leaf node probabilities under the condition that all root node events occur, and Table 6 shows the leaf node probabilities under the condition that root node events do not occur:

[0101] Table 5

[0102]

[0103] Table 6

[0104]

[0105] Step S14: Use the dynamic Bayesian network model to perform probability calculation on the transportation pipeline to obtain the pipeline risk leakage probability, and perform pipeline leakage evaluation on the transportation pipeline based on the pipeline risk leakage probability.

[0106] In this embodiment, the dynamic Bayesian network model is used to perform probability calculation on the transportation pipeline to obtain the pipeline risk leakage probability, and then it is judged whether there is a leakage situation in the transportation pipeline based on the pipeline risk leakage probability. If there is a leakage situation in the transportation pipeline, the pipeline leakage risk factor is determined, and the pipeline leakage evaluation is performed on the transportation pipeline according to the pipeline leakage risk factor.

[0107] When a long-distance transportation pipeline for volatile flammable liquids leaks and the safety barriers fail layer by layer, accident consequences such as pool fire, jet fire, and vapor cloud explosion will be triggered. The occurrence probabilities of different accidents can be further calculated based on the leakage probability. Finally, the calculated probability data is input into the machine learning model for training to further improve the accuracy of the model. After training, the probability data of the subsequent operating pipeline can be input into the trained model to quickly and accurately judge whether there is a leakage situation, and further perform pipeline leakage evaluation on the transportation pipeline. Therefore, it can greatly improve the ability to identify pipeline leakage and risk assessment, provide strong support for safety management and monitoring work in related fields, and by continuously using this model, appropriate measures can be taken in time to avoid and respond to potential leakage risks.

[0108] In this embodiment, by using the forward reasoning function of the DBN, the leakage probabilities of the pipeline at different times are obtained as Figure 4 shown. By using the backward reasoning function of the DBN, the posterior probabilities of each basic event are obtained as Figure 5 shown.

[0109] In this embodiment, a leakage fault tree model and a leakage consequence event tree model of the transportation pipeline are constructed, and a bow-tie model is generated by using the leakage fault tree model and the leakage consequence event tree model; the bow-tie model is combined with a Bayesian network model and synchronously mapped to obtain the mapped Bayesian network model; a neural network based on physical information is used to train the mapped Bayesian network model to obtain a dynamic Bayesian network model; the dynamic Bayesian network model is used to perform probability calculation on the transportation pipeline to obtain the pipeline risk leakage probability, and the pipeline leakage evaluation is performed on the transportation pipeline based on the pipeline risk leakage probability. In this application, a bow-tie model is generated by using a leakage fault tree model and a leakage consequence event tree model, and the bow-tie model is combined with a Bayesian network model and synchronously mapped to obtain the mapped Bayesian network model, which can support the probability calculation of the Bayesian network at different time nodes. A neural network based on physical information is used to train the mapped Bayesian network model, so as to realize the prediction of the leakage of the long-distance transportation pipeline of volatile and flammable liquids and the accurate identification of leakage risk factors. By using machine learning algorithms, a large amount of data can be analyzed and processed, so as to construct a more accurate and reliable dynamic Bayesian network model. The dynamic Bayesian network model is used for probability calculation, so as to perform pipeline leakage evaluation on the transportation pipeline, identify the key factors leading to leakage, reduce the complexity of pipeline leakage evaluation, and help to formulate more effective safety measures and risk management strategies in the future.

[0110] See Figure 6 shown. An embodiment of the present invention discloses a pipeline leakage evaluation method based on physical fusion Bayesian, which may specifically include:

[0111] Step S21: Taking pipeline leakage as the top event, identifying the causes of pipeline risk leakage, constructing the leakage fault tree model of the transportation pipeline, and based on the top event, combining the preset accident consequence types and using the triggering conditions of different accidents to determine safety barriers, so as to construct the leakage consequence event tree model of the transportation pipeline.

[0112] In this embodiment, the process of constructing the BT model is as follows: In the BT model for the leakage of long-distance pipelines transporting volatile and flammable liquids, the pipeline leakage is taken as the top event, and the causes that may lead to the occurrence of risks are identified to form a leakage fault tree model; then, based on the top event, combined with the common accident consequence types of pipeline leakage, the safety barriers are determined using the triggering conditions of different accidents to form an event tree model, and finally, the BT model applied to the long-distance pipelines transporting flammable and volatile liquids in the present invention is constructed. As Figure 7 shown, the risk factors for pipeline operation leakage are as follows:

[0113] (1) Corrosion failure. Pipeline corrosion includes internal corrosion and external corrosion. The pipeline is in different surrounding environments, which are complex and changeable. Different media have different degrees of corrosion on the pipeline. The non-compliance of the external anti-corrosion layer of the pipeline and the peeling off of the internal coating of the pipeline may cause pipeline leakage due to the corrosion of the methanol medium transported by the pipeline.

[0114] (2) Natural disasters. Meteorological disasters such as floods, typhoons, blizzards, and sandstorms, volcanic and earthquake disasters, and geological disasters such as mountain collapses, landslides, and mudslides may all cause pipeline fractures and leaks.

[0115] (3) Fatigue fracture. Metal materials will cause local structural changes and the continuous development of internal defects under the repeated action of stress or strain, resulting in a decline in the mechanical properties of the materials and ultimately leading to the complete fracture of the product or material. Most pipelines are made of metal materials and may undergo fatigue fracture under the action of stress, thereby causing leakage.

[0116] (4) Misoperation. Pipeline leakage may also be caused by misoperation, including design misoperation, construction misoperation, operation misoperation, and maintenance misoperation. These misoperations may lead to damage to the pipeline structure, fatigue of materials, or instability of the pipeline system, thus triggering leakage incidents.

[0117] (5) Sabotage. Due to insufficient pipeline safety education, poor legal awareness among residents along the pipeline, and low frequency of pipeline patrol personnel, there is a high probability of oil theft by drilling holes; construction near the pipeline will also damage the pipeline.

[0118] (6) Design. According to statistics, the failure events are related to design, construction, and installation. When designing the pipeline, it is necessary to strictly consider pipeline stability, route design, safety factor, selection of anti-corrosion materials, etc.

[0119] The event tree model for the leakage of long-distance pipelines transporting volatile and flammable liquids is as Figure 8As shown in the figure, the common accident consequences of leakage in long-distance pipelines for volatile and flammable liquids are as follows: Once leakage occurs, if the pressure sensor-alarm system responds first and the operator receives the signal and performs an emergency shutdown, large-area leakage will not occur and it is relatively safe; if there is no emergency shutdown and volatile and flammable liquids leak, it may cause poisoning of personnel under the condition of no ignition; if ignition occurs immediately, it may cause pool fire or jet fire; delayed ignition will cause vapor cloud explosion or flash fire.

[0120] Step S22: Generate a bow-tie model by using the leakage fault tree model and the leakage consequence event tree model.

[0121] Step S23: Convert each event, accident consequence, and safety barrier in the bow-tie model into corresponding nodes in the Bayesian network model to obtain each Bayesian node, convert the logical relationship in the bow-tie model into the conditional probability table of the Bayesian node, and assign values. Connect each Bayesian node with a directed edge to obtain the mapped Bayesian network model.

[0122] Step S24: Train the mapped Bayesian network model by using a physics-informed neural network to obtain a dynamic Bayesian network model.

[0123] Step S25: Calculate the probability of pipeline risk leakage for the pipeline by using the dynamic Bayesian network model, and evaluate the pipeline leakage based on the pipeline risk leakage probability.

[0124] In actual situations, the occurrence probability of events is different in different time slices, which will affect the development trend of accidents and accident consequences. Therefore, the static Bayesian network is extended by adding a time dimension, and PINN is incorporated for training to form a dynamic Bayesian network, realizing risk assessment and analysis based on probability and uncertain knowledge. By using the reverse inference function of DBN, the posterior probability of each basic event is obtained.

[0125] The invention content of this application is as follows: First, construct a BT model and a BN model applicable to the leakage of long-distance pipelines for volatile and flammable liquids, and determine the topological relationships of various inducements and consequences during pipeline leakage. Among them, it includes the determination of risk factors for each node and leakage consequences; Second, determine the occurrence probabilities of the root node events of the Bayesian network in two steps. In the first step, determine the fuzzy scientific scores of each expert's evaluation according to expert scoring, and fuse the scores of each expert to obtain the prior probability of the root node. The prior probability generally changes over time. Therefore, in the second step, use PINN to train and update the prior probability, input the time data into PINN for training, and output the variation relationship of the prior probability with time. Couple the output results of the above two to form a dynamic Bayesian network, and then obtain the probabilities of each risk level of the pipeline at different times and the probabilities of several types of consequences caused by leakage. Finally, use the reverse inference function of DBN to obtain the posterior probability of each basic event, that is, the contribution of risk factors to pipeline leakage and the severity of consequences.

[0126] The main advantages are: (1) organically integrate the BT model and the BN model and apply them to the leakage risk assessment of long-distance pipelines for volatile and flammable liquids, connect the mapping relationship between the two with time, use the expert scoring method to evaluate the prior probability of the Bayesian network, and obtain the conditional probability and posterior probability according to the Bayesian calculation formula, ultimately making the leakage risk assessment of long-distance pipelines for volatile and flammable liquids more efficient and accurate; (2) combine the physics-informed neural network (PINN) with the Bayesian network (BN) to form a dynamic Bayesian network (DBN), which can support the probability calculation of the Bayesian network at different time nodes; (3) integrate machine learning technology into the model to achieve the prediction of the leakage of long-distance pipelines for volatile and flammable liquids and the accurate identification of leakage risk factors.

[0127] In this embodiment, a leakage fault tree model and a leakage consequence event tree model of a conveying pipeline are constructed, and a bow-tie model is generated by using the leakage fault tree model and the leakage consequence event tree model; the bow-tie model is combined with a Bayesian network model and synchronously mapped to obtain the mapped Bayesian network model; a neural network based on physical information is used to train the mapped Bayesian network model to obtain a dynamic Bayesian network model; the dynamic Bayesian network model is used to perform probability calculation on the conveying pipeline to obtain a pipeline risk leakage probability, and a pipeline leakage evaluation is performed on the conveying pipeline based on the pipeline risk leakage probability. This application generates a bow-tie model by using a leakage fault tree model and a leakage consequence event tree model, combines the bow-tie model with a Bayesian network model, and performs synchronous mapping to obtain the mapped Bayesian network model, which can support probability calculation of the Bayesian network at different time nodes. A neural network based on physical information is used to train the mapped Bayesian network model, so as to realize the prediction of leakage of a long-distance conveying pipeline for volatile and flammable liquids and the accurate identification of leakage risk factors. By using machine learning algorithms, a large amount of data can be analyzed and processed, so as to construct a more accurate and reliable dynamic Bayesian network model. The dynamic Bayesian network model is used for probability calculation, so as to perform a pipeline leakage evaluation on the conveying pipeline, identify the key factors leading to leakage, reduce the complexity of the pipeline leakage evaluation, and help formulate more effective safety measures and risk management strategies in the future.

[0128] See Figure 9 As shown, an embodiment of the present invention discloses a pipeline leakage evaluation device based on physical fusion Bayesian, which may specifically include:

[0129] A model construction module 11, configured to construct a leakage fault tree model and a leakage consequence event tree model of a conveying pipeline, and generate a bow-tie model by using the leakage fault tree model and the leakage consequence event tree model;

[0130] A model combination and mapping module 12, configured to combine the bow-tie model with a Bayesian network model and perform synchronous mapping to obtain the mapped Bayesian network model;

[0131] A model training module 13, configured to use a neural network based on physical information to train the mapped Bayesian network model to obtain a dynamic Bayesian network model;

[0132] A pipeline leakage evaluation module 14, configured to use the dynamic Bayesian network model to perform probability calculation on the conveying pipeline to obtain a pipeline risk leakage probability, and perform a pipeline leakage evaluation on the conveying pipeline based on the pipeline risk leakage probability.

[0133] In this embodiment, a leakage fault tree model and a leakage consequence event tree model of the conveying pipeline are constructed, and a bow-tie model is generated by using the leakage fault tree model and the leakage consequence event tree model; the bow-tie model is combined with a Bayesian network model and synchronously mapped to obtain the mapped Bayesian network model; a neural network based on physical information is used to train the mapped Bayesian network model to obtain a dynamic Bayesian network model; the dynamic Bayesian network model is used to perform probability calculation on the conveying pipeline to obtain the pipeline risk leakage probability, and the pipeline leakage evaluation is performed on the conveying pipeline based on the pipeline risk leakage probability. In this application, a bow-tie model is generated by using a leakage fault tree model and a leakage consequence event tree model, and the bow-tie model is combined with a Bayesian network model and synchronously mapped to obtain the mapped Bayesian network model, which can support the probability calculation of the Bayesian network at different time nodes. A neural network based on physical information is used to train the mapped Bayesian network model, so as to realize the prediction of the leakage of the long-distance conveying pipeline of volatile and flammable liquids and the accurate identification of leakage risk factors. By using machine learning algorithms, a large amount of data can be analyzed and processed, so as to construct a more accurate and reliable dynamic Bayesian network model. The dynamic Bayesian network model is used for probability calculation, so as to perform pipeline leakage evaluation on the conveying pipeline, identify the key factors leading to leakage, reduce the complexity of pipeline leakage evaluation, and help to formulate more effective safety measures and risk management strategies in the future.

[0134] In some specific embodiments, the model construction module 11 may specifically include:

[0135] A leakage fault tree model construction module, configured to identify the causes of pipeline risk leakage with pipeline leakage as the top event, and construct the leakage fault tree model of the conveying pipeline;

[0136] A leakage consequence event tree model construction module, configured to construct the leakage consequence event tree model of the conveying pipeline based on the top event, combine preset accident consequence types, and determine safety barriers by using the triggering conditions of different accidents.

[0137] In some specific embodiments, the model combination and mapping module 12 may specifically include:

[0138] A Bayesian node determination module, configured to convert each event, accident consequence, and safety barrier in the bow-tie model into corresponding nodes in the Bayesian network model to obtain each Bayesian node;

[0139] A model mapping module, configured to convert the logical relationships in the bow - tie model into the conditional probability tables of the Bayesian nodes, assign values, and connect the Bayesian nodes with directed edges to obtain the mapped Bayesian network model.

[0140] In some specific embodiments, the model training module 13 may specifically include:

[0141] A prior probability calculation module, configured to calculate the prior probability of the root nodes of the mapped Bayesian network model;

[0142] An occurrence probability calculation module, configured to calculate the occurrence probability of the leaf nodes by using the Bayesian theorem and the occurrence probability calculation formula to calculate the prior probability;

[0143] A posterior probability calculation module, configured to calculate the prior probability and the occurrence probability by using the posterior probability calculation formula to obtain the posterior probability.

[0144] In some specific embodiments, the occurrence probability calculation formula is:

[0145]

[0146] where P(Y|X n ) is the occurrence probability of any leaf node, Y is the leaf node, is the sum of the occurrence probabilities of the leaf nodes corresponding to all the root nodes under the occurrence condition, and P(X i ) is the prior probability of the root node;

[0147] The posterior probability calculation formula is:

[0148]

[0149] where P(Z) is the pipeline leakage probability, P(Z|X i ) is the pipeline leakage probability under the condition that the event leading to the leakage risk occurs, and P(X i |Z) is the posterior probability that may cause the leakage to occur under the condition of pipeline leakage.

[0150] In some specific embodiments, the model training module 13 may specifically include:

[0151] A loss function calculation module, configured to calculate the root - mean - square error loss function and the physical part loss function, and calculate the total loss function based on the root - mean - square error loss function and the physical part loss function;

[0152] A model training module, configured to train the mapped Bayesian network model by using a physics-informed neural network based on the total loss function, the prior probability, the occurrence probability, and the posterior probability.

[0153] In some specific embodiments, the pipeline leakage evaluation module 14 may specifically include:

[0154] A judgment module, configured to judge whether there is a leakage situation in the conveying pipeline based on the pipeline risk leakage probability;

[0155] A pipeline leakage evaluation module, configured to determine a pipeline leakage risk factor if there is a leakage situation in the conveying pipeline, and perform a pipeline leakage evaluation on the conveying pipeline according to the pipeline leakage risk factor.

[0156] Figure 10 The figure is a schematic structural diagram of an electronic device provided by an embodiment of the present application. The electronic device 20 may specifically include: at least one processor 21, at least one memory 22, a power supply 23, a communication interface 24, an input / output interface 25, and a communication bus 26. Among them, the memory 22 is used to store a computer program, and the computer program is loaded and executed by the processor 21 to implement the relevant steps in the pipeline leakage evaluation method of physical fusion Bayesian executed by the electronic device disclosed in any of the foregoing embodiments.

[0157] In this embodiment, the power supply 23 is used to provide a working voltage for each hardware device on the electronic device 20; the communication interface 24 can create a data transmission channel between the electronic device 20 and an external device, and the communication protocol it follows is any communication protocol applicable to the technical solution of the present application, and no specific limitation is imposed thereon here; the input / output interface 25 is used to obtain external input data or output data to the outside, and its specific interface type can be selected according to specific application needs, and no specific limitation is made here.

[0158] In addition, the memory 22, as a carrier for resource storage, may be a read-only memory, a random access memory, a disk, or an optical disc, etc. The resources stored thereon include an operating system 221, a computer program 222, and data 223, etc., and the storage method may be temporary storage or permanent storage.

[0159] Among them, the operating system 221 is used to manage and control each hardware device on the electronic device 20 and the computer program 222, so as to implement the operation and processing of the data 223 in the memory 22 by the processor 21. It can be Windows, Unix, Linux, etc. In addition to the computer program that can be used to complete the physical fusion Bayesian pipeline leakage evaluation method executed by the electronic device 20 disclosed in any of the foregoing embodiments, the computer program 222 may further include computer programs that can be used to complete other specific tasks. In addition to the data that can include the data transmitted by external devices received by the physical fusion Bayesian pipeline leakage evaluation device, the data 223 may also include the data collected by its own input / output interface 25, etc.

[0160] The steps of the methods or algorithms described in connection with the embodiments disclosed herein can be implemented directly in hardware, software modules executed by a processor, or a combination of both. The software module can be placed in a random access memory (RAM), memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, register, hard disk, removable disk, CD-ROM, or any other form of storage medium well known in the art.

[0161] Furthermore, the embodiments of the present application also disclose a computer-readable storage medium, in which a computer program is stored. When the computer program is loaded and executed by a processor, the steps of the physical fusion Bayesian pipeline leakage evaluation method disclosed in any of the foregoing embodiments are implemented.

[0162] Finally, it should also be noted that in this document, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not expressly listed, or further includes elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "including a..." does not exclude the existence of additional identical elements in the process, method, article or device including the element.

[0163] The above has introduced in detail a pipeline leakage evaluation method, device, equipment and storage medium of physical fusion Bayesian provided by the present invention. Specific examples are used in this article to elaborate on the principle and implementation manner of the present invention. The description of the above embodiments is only used to help understand the method and its core idea of the present invention; at the same time, for those of ordinary skill in the art, according to the idea of the present invention, there will be changes in the specific implementation manner and application scope. In summary, the content of this specification should not be construed as a limitation to the present invention.

Claims

1. A pipeline leakage assessment method integrating physical and Bayesian methods, characterized in that: include: Constructing a leakage accident tree model and a leakage consequence event tree model of the transmission pipeline, and generating a bow-tie model using the leakage accident tree model and the leakage consequence event tree model; Combining the bow-tie model with the Bayesian network model and performing synchronous mapping to obtain the mapped Bayesian network model; Using a neural network based on physical information to train the mapped Bayesian network model to obtain a dynamic Bayesian network model; Using the dynamic Bayesian network model to perform probability calculation on the transmission pipeline to obtain a pipeline risk leakage probability, and performing a pipeline leakage evaluation on the transmission pipeline based on the pipeline risk leakage probability; Before the neural network based on physical information is used to train the mapped Bayesian network model, the method further includes: calculating the prior probability of the root node of the mapped Bayesian network model; calculating the prior probability according to the Bayesian theorem and using the occurrence probability calculation formula to obtain the occurrence probability of the leaf node; calculating the prior probability and the occurrence probability using the posterior probability calculation formula to obtain the posterior probability; The method of using a neural network based on physical information to train the mapped Bayesian network model includes: calculating a root mean square error loss function and a physical part loss function, and calculating a total loss function based on the root mean square error loss function and the physical part loss function; and training the mapped Bayesian network model based on the total loss function, the prior probability, the occurrence probability, and the posterior probability and using a neural network based on physical information.

2. The pipeline leakage assessment method of physical fusion Bayesian according to claim 1 is characterized in that: The construction of the leakage accident tree model and leakage consequence event tree model of the transmission pipeline includes: Taking pipeline leakage as the top event, identifying the cause of pipeline risk leakage, and constructing the leakage accident tree model of the transmission pipeline; Based on the top event, combined with the preset accident consequence type, and using the triggering conditions of different accidents to determine the safety barrier, the leakage consequence event tree model of the transmission pipeline is constructed.

3. The pipeline leakage assessment method of physical fusion Bayesian according to claim 1 is characterized in that: The bow-tie model is combined with the Bayesian network model and synchronously mapped to obtain the mapped Bayesian network model, including: Convert each event, accident consequence and safety barrier in the bow-tie model into a corresponding node in the Bayesian network model to obtain each Bayesian node; The logical relationship in the bow-tie model is converted into a conditional probability table of the Bayesian nodes, and values ​​are assigned, and directed edges are used to connect the Bayesian nodes to obtain the mapped Bayesian network model.

4. The pipeline leakage assessment method of physical fusion Bayesian according to claim 1 is characterized in that: The occurrence probability calculation formula is: Among them, P(Y|X n ) is the occurrence probability of any leaf node, Y is the leaf node, is the sum of the occurrence probabilities of the leaf nodes corresponding to all the root nodes under the occurrence condition, P(X i ) is the prior probability of the root node; The posterior probability calculation formula is: Among them, P(Z) is the probability of pipeline leakage, P(Z|X i ) is the probability of pipeline leakage under the condition of the event leading to leakage risk, P(X i |Z) is the posterior probability that a leak may occur under pipeline leakage.

5. The pipeline leakage assessment method of physical fusion Bayesian according to any one of claims 1 to 4, characterized in that: The pipeline leakage evaluation of the transmission pipeline based on the pipeline risk leakage probability includes: Determine whether there is leakage in the transmission pipeline based on the pipeline risk leakage probability; If there is leakage in the transmission pipeline, a pipeline leakage risk factor is determined, and a pipeline leakage evaluation is performed on the transmission pipeline based on the pipeline leakage risk factor.

6. A pipeline leakage evaluation device integrating physical and Bayesian methods, characterized in that: include: A model building module, used to build a leakage accident tree model and a leakage consequence event tree model of the transmission pipeline, and generate a bow-tie model using the leakage accident tree model and the leakage consequence event tree model; A model combination and mapping module, used to combine the bow-tie model with the Bayesian network model and perform synchronous mapping to obtain the mapped Bayesian network model; A model training module, used to train the mapped Bayesian network model using a neural network based on physical information to obtain a dynamic Bayesian network model; A pipeline leakage evaluation module is used to perform probability calculation on the transmission pipeline using the dynamic Bayesian network model to obtain a pipeline risk leakage probability, and perform pipeline leakage evaluation on the transmission pipeline based on the pipeline risk leakage probability; Before the neural network based on physical information is used to train the mapped Bayesian network model, the method further includes: calculating the prior probability of the root node of the mapped Bayesian network model; calculating the prior probability according to the Bayesian theorem and using the occurrence probability calculation formula to obtain the occurrence probability of the leaf node; calculating the prior probability and the occurrence probability using the posterior probability calculation formula to obtain the posterior probability; The method of using a neural network based on physical information to train the mapped Bayesian network model includes: calculating a root mean square error loss function and a physical part loss function, and calculating a total loss function based on the root mean square error loss function and the physical part loss function; and training the mapped Bayesian network model based on the total loss function, the prior probability, the occurrence probability, and the posterior probability and using a neural network based on physical information.

7. An electronic device, characterized in that: include: Memory, used to store computer programs; A processor is used to execute the computer program to implement the pipeline leakage assessment method of physical fusion Bayesian as described in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that: Used to store a computer program; wherein, when the computer program is executed by a processor, the pipeline leakage assessment method of physical fusion Bayesian as described in any one of claims 1 to 5 is implemented.

Images