Risk assessment method and system for airborne hydrogen storage system of airplane

Abstract

The invention discloses a risk assessment method and system for an airborne hydrogen storage system of an airplane, and belongs to the technical field of civil aviation safety analysis, and the method comprises the steps: determining the node information and state information of each node according to the composition architecture and component working principle of the airborne hydrogen storage system of the airplane, and constructing a fuzzy dynamic Bayesian network graph structure; determining a risk state evaluation result of each node based on a fuzzy dynamic Bayesian network graph structure, and converting the risk state evaluation result of each node into a fuzzy probability as probability information of a fuzzy dynamic node; constructing a fuzzy dynamic Bayesian network model according to the fuzzy dynamic Bayesian network graph structure and the probability information of the fuzzy dynamic nodes; and solving the fuzzy dynamic Bayesian network model, and performing risk assessment on the airborne hydrogen storage system. According to the method, quantitative risk assessment of the risk diffusion and evolution characteristic process in a complex and dynamic airborne hydrogen storage system can be realized, so that the risk assessment result is more practical.

Description

Technical Field

[0001] The present invention relates to the technical field of civil aviation safety analysis, and more specifically to a risk assessment method and system for an aircraft onboard hydrogen storage system. Background Art

[0002] In the field of civil aviation safety analysis, safety analysis of aircraft onboard hydrogen storage systems is essential. Fuzzy Dynamic Bayesian Networks (FDBN) is a method based on the combination of Dynamic Bayesian Networks (DBN) and fuzzy set theory. DBN is an extension of Bayesian networks in time and can display the evolutionary state of node state changes. DBN uses the Bayesian principle as its mathematical basis and can dynamically evaluate the evolution of system risks. Fuzzy set theory is used to process expert evaluations into fuzzy probabilities as the input of the DBN model, which is used to solve the problem of inaccurate and insufficient empirical data in most systems.

[0003] At present, the research on airborne hydrogen storage systems in the existing technology has only stopped at the system architecture design, and no detailed research on system safety has been carried out, which has caused the existing research on hydrogen-powered aircraft to always remain at the theoretical innovation stage and have a low degree of recognition. However, due to the complex design architecture of aircraft airborne hydrogen storage systems, the traditional research on airborne hydrogen storage systems lacks empirical data for dynamic safety performance analysis, that is, the risk assessment of existing airborne hydrogen storage systems fails to achieve dynamic risk evolution assessment of complex, dynamic and data-lacking airborne hydrogen storage systems.

[0004] In summary, the existing aircraft onboard hydrogen storage systems are complex, dynamic and lack of empirical data, making it impossible to conduct accurate and efficient risk assessments, resulting in stagnation in research on airborne hydrogen storage systems and affecting the recognition of airborne hydrogen storage systems. Summary of the invention

[0005] In response to the problems existing in the above-mentioned fields, the present invention proposes a risk assessment method and system for aircraft onboard hydrogen storage systems, which can solve the technical problems that the existing aircraft onboard hydrogen storage systems are complex, dynamic and lack of empirical data, making it impossible to conduct accurate and efficient risk assessments, resulting in stagnation in research on onboard hydrogen storage systems and affecting the recognition of onboard hydrogen storage systems.

[0006] In order to solve the above technical problems, the present invention discloses a risk assessment method for an aircraft onboard hydrogen storage system, comprising the following steps:

[0007] According to the composition architecture and component working principle of the aircraft onboard hydrogen storage system, the component failure event is taken as the root dynamic node, the abnormal event of the component set that completes a specific function is taken as the intermediate dynamic node, the hydrogen state parameter of the physical process of the onboard hydrogen storage system is taken as the static node, and the functional failure event of the onboard hydrogen storage system is taken as the result node, and the risk state and normal state of each node are determined; according to the node information and state information of each node, the fuzzy dynamic Bayesian network graph structure is constructed;

[0008] Based on the fuzzy dynamic Bayesian network graph structure, the risk status evaluation results of each node are determined, and the risk status evaluation results of each node are converted into fuzzy probabilities as the probability information of the fuzzy dynamic nodes;

[0009] According to the fuzzy dynamic Bayesian network graph structure and the probability information of fuzzy dynamic nodes, a fuzzy dynamic Bayesian network model is constructed;

[0010] The fuzzy dynamic Bayesian network model is solved, and the risk status of the risk node is determined by solving the probability information of each node, and the risk assessment of the onboard hydrogen storage system is carried out.

[0011] Preferably, the risk assessment of the airborne hydrogen storage system specifically includes reasoning diagnosis, sensitivity analysis and risk trend analysis, wherein:

[0012] Reasoning diagnosis is to set the probability of the risk state of the risk node to 1, and use the reverse reasoning ability of the dynamic Bayesian network to update the fuzzy dynamic Bayesian network model to obtain the posterior probability of each dynamic node; according to the posterior probability of each dynamic node, the possibility of each dynamic node causing the risk to occur is judged;

[0013] Sensitivity analysis is to calculate the failure probability change rate based on the prior probability and posterior probability of each dynamic node, and determine the degree of dependence of the risk node on each dynamic node;

[0014] Risk trend analysis is based on the results of reasoning calculation of the fuzzy dynamic Bayesian network model. It combines the results of reasoning diagnosis and sensitivity analysis to identify the risk nodes that cause failure events, analyze the risk probability of the risk nodes, and complete the risk trend analysis.

[0015] Preferably, the probability information as the fuzzy dynamic node comprises the following steps:

[0016] Collect information of each expert and calculate expert weight;

[0017] According to the expert evaluation results of each node of the dynamic Bayesian network, the consistency of expert opinions between every two experts is calculated;

[0018] Combining the expert opinion consistency and expert weight, calculate the weighted consistency of every two experts' opinions, and so on, calculate the weighted consistency of all experts;

[0019] The ratio of the weighted agreement between every two experts to all weighted agreements is taken as the relative agreement of the experts’ opinions. The consistency coefficient of each expert’s opinion is calculated by combining the expert weights with the relative agreement of the experts’ opinions.

[0020] Through the consistency coefficient of each expert's opinion, the integrated result corresponding to the expert evaluation result of each node risk status is obtained, that is, the overall fuzzy number;

[0021] The overall fuzzy number is defuzzified, and the fuzzy probability is obtained by calculating the fuzzy possible value, and the fuzzy probability is used as the probability information of the fuzzy dynamic node.

[0022] Preferably, the step of calculating the weighted consistency of all experts specifically includes the following steps:

[0023] Experts evaluate the possibility of the risk status of each node based on the fuzzy dynamic Bayesian network graph structure. The contribution of each expert's evaluation result to the integrated evaluation result depends on the expert weight and the consistency of expert opinions.

[0024] Compare the opinions of each two experts on each node to determine the consistency of their evaluations of the node. The higher the consistency, the closer the evaluation results of the two experts are.

[0025] Combining the expert opinion consistency and expert weights, calculate the weighted consistency of every two expert opinions until the calculation of the weighted consistency of all expert opinions is completed.

[0026] Preferably, the calculation formula for the expert opinion consistency is:

[0027]

[0028] In the formula, and They represent the evaluation fuzzy intervals of the x-th expert and the y-th expert, respectively.

[0029] Preferably, the calculation formula for the weighted consistency of the expert opinions is:

[0030]

[0031] In the formula, n represents the number of experts, A W (E x ) indicates expert E x The weighted agreement of expert opinions; E x and E y represent the xth expert and the yth expert respectively, W(Ex ) and W(E y ) are experts E x and E y The weight of .

[0032] Preferably, the calculation formula of the overall fuzzy number is:

[0033]

[0034] In the formula, C c (E 1 ) indicates the first expert E 1 The expert opinion consistency coefficient is Indicates the first expert E 1 Evaluation results of C c (E n ) represents the nth expert E n The expert opinion consistency coefficient is Indicates the nth expert E n The evaluation results;

[0035] The xth expert E x The expert opinion consistency coefficient C c (E x )for:

[0036] C c (E x )=βW(E x )+(1-β)A R (E x )

[0037] The xth expert E x The relative consistency of expert opinions A R (E x )for:

[0038]

[0039] In the formula, C c (E x ) represents the xth expert E x is the expert opinion consistency coefficient, and β is the relaxation factor.

[0040] Preferably, a risk assessment system for an aircraft onboard hydrogen storage system is also included, comprising:

[0041] A fuzzy dynamic Bayesian network graph structure construction module is used to take component failure events as root dynamic nodes, component set abnormal events that complete specific functions as intermediate dynamic nodes, hydrogen state parameters of the physical process of the onboard hydrogen storage system as static nodes, and functional failure events of the onboard hydrogen storage system as result nodes according to the composition architecture and component working principles of the aircraft onboard hydrogen storage system, and determine the risk state and normal state of each node; construct a fuzzy dynamic Bayesian network graph structure according to the node information and state information of each node;

[0042] The probability information acquisition module is used to determine the risk status evaluation result of each node based on the fuzzy dynamic Bayesian network graph structure, and convert the risk status evaluation result of each node into fuzzy probability as the probability information of the fuzzy dynamic node;

[0043] A fuzzy dynamic Bayesian network model building module is used to build a fuzzy dynamic Bayesian network model according to the fuzzy dynamic Bayesian network graph structure and the probability information of the fuzzy dynamic nodes;

[0044] The risk assessment module is used to solve the fuzzy dynamic Bayesian network model, determine the risk status of the risk node by solving the probability information of each node, and conduct risk assessment on the airborne hydrogen storage system.

[0045] Preferably, a computer device is also included, the computer device comprising a memory and a processor, the memory storing a computer program, and when the computer program is executed by the processor, the processor executes the following steps:

[0046] According to the composition architecture and component working principle of the aircraft onboard hydrogen storage system, the component failure event is taken as the root dynamic node, the abnormal event of the component set that completes a specific function is taken as the intermediate dynamic node, the hydrogen state parameter of the physical process of the onboard hydrogen storage system is taken as the static node, and the functional failure event of the onboard hydrogen storage system is taken as the result node, and the risk state and normal state of each node are determined; according to the node information and state information of each node, the fuzzy dynamic Bayesian network graph structure is constructed;

[0047] Based on the fuzzy dynamic Bayesian network graph structure, the risk status evaluation results of each node are determined, and the risk status evaluation results of each node are converted into fuzzy probabilities as the probability information of the fuzzy dynamic nodes;

[0048] According to the fuzzy dynamic Bayesian network graph structure and the probability information of fuzzy dynamic nodes, a fuzzy dynamic Bayesian network model is constructed;

[0049] The fuzzy dynamic Bayesian network model is solved, and the risk status of the risk node is determined by solving the probability information of each node, and the risk assessment of the onboard hydrogen storage system is carried out.

[0050] Preferably, it further comprises a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the processor executes the following steps:

[0051] According to the composition architecture and component working principle of the aircraft onboard hydrogen storage system, the component failure event is taken as the root dynamic node, the abnormal event of the component set that completes a specific function is taken as the intermediate dynamic node, the hydrogen state parameter of the physical process of the onboard hydrogen storage system is taken as the static node, and the functional failure event of the onboard hydrogen storage system is taken as the result node, and the risk state and normal state of each node are determined; according to the node information and state information of each node, the fuzzy dynamic Bayesian network graph structure is constructed;

[0052] Based on the fuzzy dynamic Bayesian network graph structure, the risk status evaluation results of each node are determined, and the risk status evaluation results of each node are converted into fuzzy probabilities as the probability information of the fuzzy dynamic nodes;

[0053] According to the fuzzy dynamic Bayesian network graph structure and the probability information of fuzzy dynamic nodes, a fuzzy dynamic Bayesian network model is constructed;

[0054] The fuzzy dynamic Bayesian network model is solved, and the risk status of the risk node is determined by solving the probability information of each node, and the risk assessment of the onboard hydrogen storage system is carried out.

[0055] Compared with the prior art, the present invention has the following beneficial effects:

[0056] The risk assessment method of the aircraft onboard hydrogen storage system proposed by the present invention constructs a fuzzy dynamic Bayesian network graphic structure according to the composition architecture and component working principle of the aircraft onboard hydrogen storage system, and can realize the quantitative risk assessment of the risk diffusion and evolution characteristics process in the complex and dynamic onboard hydrogen storage system through the risk transmission causal relationship network, so that the risk assessment result is more in line with reality. The risk transmission and diffusion path of the onboard hydrogen storage system is described by a complex network system, and the risk state evaluation result of each node is determined based on the fuzzy dynamic Bayesian network graphic structure, and the risk state evaluation result of each node is converted into fuzzy probability, which is used as the probability information of the fuzzy dynamic node. The risk evolution process can be directly quantified, and the probability change of the risk node can be intuitively expressed. It has a high degree of fit with the actual risk evolution law. By quantifying the expert evaluation results, a relatively objective data set can be obtained as the data basis for dynamic risk assessment without relying on the data source, and the system can also be accurately evaluated for risk evolution when the data is insufficient and uncertain. The constructed fuzzy dynamic Bayesian network model is inferred and calculated, and the risk state of the risk node is determined by solving the probability information of each node to complete the risk assessment. This method is based on the constructed fuzzy dynamic Bayesian network model and can provide a risk assessment template based on the fuzzy dynamic Bayesian network results. Compared with traditional risk assessment methods, this method avoids the technical defects of low risk assessment efficiency and poor accuracy caused by lack of empirical data and data analysis methods, and effectively improves the accuracy of risk assessment. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 A flow chart of the risk assessment method for an aircraft onboard hydrogen storage system proposed by the present invention;

[0058] Figure 2 A schematic diagram of a fuzzy dynamic Bayesian network graphical structure for the present invention;

[0059] Figure 3 The failure probability trend of each node obtained by completing the inference calculation of the fuzzy dynamic Bayesian network model constructed by the present invention;

[0060] Figure 4 This is the failure probability trend of each node obtained by taking the risk trend of D2-1 state "2" as an example in the embodiment of the present invention. DETAILED DESCRIPTION

[0061] The following will be combined with the attached embodiment of the present invention Figure 1-Figure 4 , the technical solutions in the embodiments of the present invention are clearly and completely described. It should be understood that the terms described in the present invention are only used to describe specific implementation methods and are not used to limit the present invention.

[0062] like Figure 1As shown, the present invention proposes a risk assessment method for an aircraft onboard hydrogen storage system, which can break through the current risk assessment of the onboard hydrogen storage system, which is not capable of conducting quantitative risk assessment taking into account the complexity and dynamic characteristics of the system and the lack of data sources, thereby causing the risk assessment to deviate from reality and the current situation where the reference value of the evaluation results is low.

[0063] This method establishes a dynamic Bayesian model for the airborne hydrogen storage system according to the system architecture, describes the risk transmission and diffusion path of the airborne hydrogen storage system with a complex network system, and converts the evaluation opinions of the dynamic Bayesian network model into fuzzy probabilities through expert evaluation to obtain the input probability of the fuzzy dynamic Bayes, and then constructs a fuzzy dynamic Bayesian network model. The fuzzy dynamic Bayesian probability information obtained through expert opinions can obtain a more objective data set as the data basis for dynamic risk assessment without relying on data sources. Using fuzzy dynamic Bayesian networks as the main tool for risk assessment can also achieve quantitative risk assessment of risk diffusion and evolution characteristics in complex and dynamic airborne hydrogen storage systems through risk transmission causal networks, making the risk assessment results more realistic.

[0064] The method specifically comprises the following steps:

[0065] S1: According to the composition architecture and component working principle of the aircraft onboard hydrogen storage system, the component failure event is taken as the root dynamic node, the abnormal event of the component set that completes a specific function is taken as the intermediate dynamic node, the hydrogen state parameter of the physical process of the onboard hydrogen storage system is taken as the static node, and the functional failure event of the onboard hydrogen storage system is taken as the result node, and the risk state and normal state of each node are determined; according to the node information and state information of each node, the fuzzy dynamic Bayesian network graph structure is constructed;

[0066] S2: Based on the fuzzy dynamic Bayesian network graph structure, determine the risk status evaluation results of each node, and convert the risk status evaluation results of each node into fuzzy probability as the probability information of the fuzzy dynamic node;

[0067] S3: Construct a fuzzy dynamic Bayesian network model based on the fuzzy dynamic Bayesian network graph structure and the probability information of the fuzzy dynamic nodes;

[0068] S4: Solve the fuzzy dynamic Bayesian network model, determine the risk status of the risk node by solving the probability information of each node, and conduct risk assessment on the onboard hydrogen storage system.

[0069] Specifically, in step S2, determining the risk status evaluation result of each node includes the following steps:

[0070] The expert ratings were divided into seven levels of language value, including very low (VL), low (L), low (ML), medium (M), high (MH), high (H) and very high (VH);

[0071] According to the level of speech quantity value, each node of the dynamic Bayesian network is evaluated, and according to the fuzzy interval corresponding to the speech quantity value shown in Table 1, the expert evaluation results are converted into corresponding fuzzy intervals.

[0072] Table 1 Corresponding fuzzy intervals of quantity values

[0073]

[0074] Expert information is collected and expert weights are calculated, indicating that since each expert has different levels of depth in their research fields, the reliability of the expert evaluation results needs to be expressed according to the corresponding weights.

[0075] The calculation of expert weights includes the following steps:

[0076] As shown in Table 2, this is the scoring standard adopted by the present invention. According to the scoring standard shown in Table 2, the total score of each expert (ie, the weight score) and the proportion of the weight score in all expert weight scores (ie, the expert weight) are calculated.

[0077] Table 2 Expert scoring criteria

[0078]

[0079] According to the evaluation results of each expert, the consistency of the expert opinions of every two experts is calculated. That is, the experts evaluate the possibility of the risk state of each node based on the fuzzy dynamic Bayesian network graph structure. The contribution of each expert's evaluation result to the integrated evaluation result depends on the expert weight and the consistency of the expert opinions.

[0080] The opinions of every two experts on each node are compared to determine the consistency of their evaluations of the node. The higher the consistency, the closer the evaluation results of the two experts are.

[0081] Combining the consistency of expert opinions with expert weights, calculate the weighted consistency of every two expert opinions. Considering the respective weights of the two experts, calculate the weighted consistency of the two expert opinions until the calculation of the weighted consistency of all experts is completed.

[0082] Expert opinion agreement for:

[0083]

[0084] In the formula, and They represent the evaluation fuzzy intervals of the x-th expert and the y-th expert, respectively.

[0085] Will is called the deviation value of the opinions of the two experts, and the fuzzy interval and Subtract the corresponding four values ​​from each other and find the average of the four differences.

[0086] E i (i=1,2,…,n) represents the i-th expert, n represents the number of experts, and expert E x The weighted agreement of expert opinions A W (E x )for:

[0087]

[0088] In the formula, E x and E y represent the xth expert and the yth expert respectively, W(E x ) and W(E y ) are experts E x and E y The weight of .

[0089] The ratio of the weighted agreement between every two experts to all weighted agreements was taken as the relative agreement of the experts’ opinions.

[0090] Relative consistency of expert opinions A R (E x )for:

[0091]

[0092] The expert opinion consistency coefficient is calculated by combining the expert weight and the relative consistency of expert opinions.

[0093] Expert opinion consistency coefficient C c (E x )for:

[0094] C c (E x )=βW(E x )+(1-β)A R (E x )

[0095] Wherein, β is a relaxation factor, and the value of the present invention is 0.5.

[0096] Through the expert opinion consistency coefficient, the integrated result corresponding to the expert evaluation results of the risk status of each node can be obtained, that is, the overall fuzzy number.

[0097] Total fuzzy number for:

[0098]

[0099] In the formula, C c (E 1 ) indicates the first expert E 1 The expert opinion consistency coefficient is Indicates the first expert E 1 Evaluation results of C c (E n ) represents the nth expert E n The expert opinion consistency coefficient is Indicates the nth expert E n evaluation results.

[0100] The steps of calculating "expert weight", "expert opinion consistency", "expert opinion weighted consistency", "expert opinion relative consistency", and "expert opinion consistency coefficient" are methods of integrating expert opinions in fuzzy theory, called "similarity aggregation method". The process of converting expert opinions into overall fuzzy sets through similarity aggregation method is called expert opinion fuzzification. The detailed process of processing expert evaluation results using similarity aggregation method is as follows:

[0101] Assume that a node is D, and there are 10 nodes in total. A total of 3 experts score the risk status of the nodes. Each node is required to be evaluated by an expert. Each node corresponds to three evaluation results. The expert scores are divided into seven language value levels, including very low (VL), low (L), low (ML), medium (M), high (MH), high (H) and very high (VH).

[0102] Based on these three evaluation results, according to Table 1, each corresponds to a fuzzy interval, and each fuzzy interval corresponds to four values ​​a, b, c and d.

[0103] Expert weight is a calculation of the confidence of each expert, reflecting the credibility of the expert in this field.

[0104] Then calculate the consistency of expert opinions. Compare the three expert opinions at node D in pairs. Subtract the fuzzy intervals corresponding to these three opinions and then find the average value to calculate the degree of deviation of the opinions of the two experts. Expert opinion consistency = 1-the degree of deviation of the opinions of the two experts. And so on. Each node needs to be solved, and finally we will get the evaluation results of 10 groups of experts comparing each other.

[0105] According to the consensus of 10 groups of experts' opinions, the weighted consensus of experts' opinions is calculated. For example, the weighted consensus of expert 1's opinion is calculated using node D as follows:

[0106] (Consistency between expert 1 and expert 2 × weight of expert 2 + consistency between expert 1 and expert 3 × weight of expert 3) / (weight of expert 2 + weight of expert 3) = weighted consistency of expert 1's opinion

[0107] Each node corresponds to three weighted consensus of expert opinions.

[0108] Calculate the relative consistency of expert opinions. For example, node D calculates the relative consistency of expert 1's opinion as:

[0109] Weighted consistency of expert 1's opinion / (weighted consistency of expert 1's opinion + weighted consistency of expert 2's opinion + weighted consistency of expert 3's opinion) = relative consistency of expert 1's opinion

[0110] According to the relative consistency of expert opinions and expert weights, the expert opinion consistency coefficient can be calculated to indicate the degree of use of each expert opinion. Taking node D as an example, the expert 1 opinion consistency coefficient is:

[0111] β×Expert 1 weight+(1-β)×Expert 1 relative consistency=Expert 1 opinion consistency coefficient

[0112] Among them, β is the relaxation factor, which is generally taken as 0.5.

[0113] The overall fuzzy number is calculated as:

[0114] The overall fuzzy number = expert 1's opinion consistency coefficient × expert 1's evaluation result (corresponding fuzzy interval) + expert 2's opinion consistency coefficient × expert 2's evaluation result + expert 3's opinion consistency coefficient × expert 3's evaluation result.

[0115] A set of four numbers is obtained, which is the total fuzzy number.

[0116] Defuzzification is the process of calculating fuzzy probability through overall fuzzy numbers. The present invention obtains fuzzy probability by calculating fuzzy possibility values; average values ​​are calculated through overall fuzzy numbers, and fuzzy probability is to convert each calculated average value into a failure probability value.

[0117] The overall fuzzy number is defuzzified, and the fuzzy probability is obtained by calculating the fuzzy possible value. The obtained fuzzy probability is used as the probability information of the fuzzy dynamic node.

[0118] Fuzzy probability value FP S for:

[0119]

[0120] Where a, b, c and d are all fuzzy intervals of the overall fuzzy number, and the fuzzy interval is the summary result of each expert's evaluation results on each node.

[0121] Using the fuzzy possibility value, we can find the fuzzy probability FP R for:

[0122]

[0123] The evaluation results of each node by experts based on the dynamic Bayesian network model are converted into fuzzy probabilities as the probability information of fuzzy dynamic nodes, including: root dynamic node prior probability, root dynamic node transfer probability, intermediate dynamic node conditional probability, static node conditional probability, and consequence node conditional probability.

[0124] In step S3, a fuzzy dynamic Bayesian network model is constructed

[0125] The fuzzy dynamic Bayesian network constructed by the present invention is completed based on software, and the graphical structure and probability information are input into the software. In the GeNle software, the node graphical position and node state are set, and each node is connected in the order of "root dynamic node-intermediate dynamic node-static node-consequence node". The probability value (i.e., probability information) of the corresponding state of each node is set in the software, and a time slice is set (the system research duration is divided, and a time slice is defined as a time slice at a certain time interval. The number of time slices used in the present invention is 500, indicating that the aircraft experiences 500 repeated flight missions). Automatic calculation by software completes the construction and calculation of the fuzzy dynamic Bayesian model.

[0126] In step S4, the risk status of the risk node is determined by solving the probability information of each node, and the risk assessment of the airborne hydrogen storage system is performed, which specifically includes reasoning diagnosis, sensitivity analysis and risk trend analysis, wherein:

[0127] Reasoning diagnosis is to set the probability of the risk state of the risk node to 1, and use the reverse reasoning ability of the dynamic Bayesian network to update the fuzzy dynamic Bayesian network model to obtain the posterior probability of each dynamic node; according to the posterior probability of each dynamic node, the possibility of each dynamic node causing the risk to occur is judged;

[0128] Sensitivity analysis is to calculate the failure probability change rate based on the prior probability and posterior probability of each dynamic node, and determine the degree of dependence of the risk node on each dynamic node;

[0129] Risk trend analysis is based on the calculation results of fuzzy dynamic Bayesian network model reasoning. It combines the results of reasoning diagnosis and sensitivity analysis to identify the risk nodes that cause failure events, analyze the risk probability of risk nodes, and complete risk trend analysis.

[0130] Specifically, it is necessary to determine the system operation time and set the time slice T; set the probability of the risk state of the risk node to 1, and calculate the posterior probability of each root dynamic node; according to the posterior probability of each dynamic node, complete the reasoning diagnosis;

[0131] The failure probability change rate RoV is:

[0132]

[0133] Where, X i Represents the root dynamic node event, RoV(X i ) is X i RoV value, θ(X i ) is X i The prior probability of π(X i ) is X i The posterior probability of

[0134] Complete sensitivity analysis based on the RoV value of each dynamic node.

[0135] Based on the calculation results of fuzzy dynamic Bayesian network reasoning, combined with the results of reasoning diagnosis and sensitivity analysis, the main risk events that lead to dangerous events are identified, the risk probability is further analyzed, and the risk trend analysis is completed.

[0136] The present invention also proposes a risk assessment system for an aircraft onboard hydrogen storage system, comprising:

[0137] A fuzzy dynamic Bayesian network graph structure construction module is used to take component failure events as root dynamic nodes, component set abnormal events that complete specific functions as intermediate dynamic nodes, hydrogen state parameters of the physical process of the onboard hydrogen storage system as static nodes, and functional failure events of the onboard hydrogen storage system as result nodes according to the composition architecture and component working principles of the aircraft onboard hydrogen storage system, and determine the risk state and normal state of each node; construct a fuzzy dynamic Bayesian network graph structure according to the node information and state information of each node;

[0138] The probability information acquisition module is used to determine the risk status evaluation result of each node based on the fuzzy dynamic Bayesian network graph structure, and convert the risk status evaluation result of each node into fuzzy probability as the probability information of the fuzzy dynamic node;

[0139] A fuzzy dynamic Bayesian network model building module is used to build a fuzzy dynamic Bayesian network model according to the fuzzy dynamic Bayesian network graph structure and the probability information of the fuzzy dynamic nodes;

[0140] The risk assessment module is used to solve the fuzzy dynamic Bayesian network model, determine the risk status of the risk node by solving the probability information of each node, and conduct risk assessment on the airborne hydrogen storage system.

[0141] Example

[0142] In order to verify the feasibility of the risk assessment method for the aircraft onboard hydrogen storage system proposed in the present invention, a risk assessment is carried out by taking the onboard hydrogen storage system of a hydrogen-powered aircraft as an example. The onboard hydrogen storage system of a hydrogen-powered aircraft comprises: a hydrogen refueling tank group (including a tank body, a pressure sensor, and a temperature sensor), a bottle mouth combination valve (including a filter, a flow stabilizing valve, a solenoid valve, and a manual shut-off valve), a bottle tail combination valve (pressure relief device, a flame arrester), connecting pipelines, etc.

[0143] The embodiment of the present invention provides a risk assessment method for an onboard hydrogen storage system of a hydrogen-powered aircraft based on a fuzzy dynamic Bayesian network. The onboard hydrogen storage system architecture is shown in Table 3.

[0144] Table 3 Airborne hydrogen storage system architecture

[0145]

[0146]

[0147] The above-mentioned risk assessment method for the hydrogen storage system onboard a hydrogen-powered aircraft based on the fuzzy dynamic Bayesian network comprises the following steps:

[0148] Step 1: Clarify the architecture of the onboard hydrogen storage system and construct the fuzzy dynamic Bayesian network graph structure.

[0149] Specifically, according to the system architecture of the airborne hydrogen storage system, the root dynamic node, intermediate dynamic node, static node and consequence node of the fuzzy dynamic Bayesian network are selected; and according to the possible failure modes of each node, its risk state and normal state are determined.

[0150] Among them, the determination results of the root dynamic node and the intermediate dynamic node are shown in Table 4, the determination results of the static node are shown in Table 5, and the determination results of the consequence node are shown in Table 6.

[0151] Table 4 Determination results of root dynamic nodes and intermediate dynamic nodes

[0152]

[0153]

[0154] Table 5 Determination results of static nodes

[0155] Node Symbol Static Node Name state S1 Hydrogen supply pressure 0: too low; 1: normal S2 Hydrogen supply flow 0: too large; 1: normal; 2: too small S3 Hydrogen temperature in tank 0: too high; 1: normal S4 Hydrogen pressure in tank 0: too low; 1: normal

[0156] Table 6 Determination results of consequence nodes

[0157]

[0158]

[0159] Select the consequence node as "Insufficient hydrogen supply" and set the number of time slices to 500, which means that the aircraft repeats the same flight mission 500 times.

[0160] Combining the system composition architecture and component working principles of the airborne hydrogen storage system, a fuzzy dynamic Bayesian network graph structure is constructed, such as Figure 2 shown.

[0161] Step 2: Based on the fuzzy dynamic Bayesian network graph structure constructed in step 1, obtain the evaluation results through expert evaluation of the possibility of risk events; quantify the evaluation results through expert evaluation, first convert them into fuzzy sets through similarity aggregation method, and then convert the fuzzy sets into fuzzy probabilities through defuzzification, and use the fuzzy probabilities as dynamic Bayesian network probability information, where the probability information that needs to be obtained includes: root dynamic node prior probability, root dynamic node transfer probability, intermediate dynamic node conditional probability, static node conditional probability, and consequence node conditional probability.

[0162] Collect expert information and combine Table 1 with Table 2 to calculate the expert weight of each expert as shown in Table 7.

[0163] Table 7 Expert information and weights

[0164] expert Professional positions Study duration age Weighted score Weight 1 Technical Consultant 15-19 ≥50 24 0.218 2 Technical Supervisor ≥20 40-49 26 0.236 3 Technical Consultant 10-14 30-39 18 0.164 4 engineer 10-14 30-39 16 0.145 5 Technical Consultant ≥20 ≥50 26 0.236 sum 110 1

[0165] Taking node D1-1 “pipeline hydrogen embrittlement cracking” as an example, the fuzzy probability obtained through expert evaluation is converted into the root dynamic node prior probability process, as shown in Table 8.

[0166] Table 8D1-1 Expert evaluation results and root dynamic node prior probability

[0167]

[0168] Taking node D1-1 “pipeline hydrogen embrittlement cracking” as an example, the fuzzy probability obtained through expert evaluation is converted into the root dynamic node transfer probability process, as shown in Table 9.

[0169] Table 9D1-1 Node expert evaluation results and transfer probability

[0170]

[0171] Taking node D1 “rupture at the connection between the pipeline and the valve” as an example, the fuzzy probability obtained through expert evaluation is converted into the conditional probability process of the intermediate dynamic node, static node or consequence node, as shown in Table 10.

[0172] Table 10D1 node expert evaluation results and conditional probability

[0173]

[0174] Step 3: Combine the fuzzy dynamic Bayesian network graph structure with probability information to construct a fuzzy dynamic Bayesian network model and complete the inference calculation of the model. Based on the calculation results of the fuzzy dynamic Bayesian network model, complete the risk assessment.

[0175] Specifically, refer to the examples of probability acquisition process in step 2, calculate all probability information (including root dynamic node prior probability, root dynamic node transfer probability, intermediate dynamic node conditional probability, static node conditional probability and consequence node conditional probability). Combine all the calculated probability information with the fuzzy dynamic Bayesian network graph structure, build a fuzzy dynamic Bayesian network model, and complete the inference calculation to obtain the failure probability trend of each node, such as Figure 3 shown.

[0176] Based on the above calculation results, the reasoning diagnosis, sensitivity analysis and risk trend analysis are completed in sequence.

[0177] (1) Inference diagnosis

[0178] The probability of node C state "0" is set to 1, indicating that "insufficient hydrogen supply" occurs, and the posterior probabilities of each dynamic node are shown in Table 11.

[0179] Table 11 Posterior probability of root dynamic node

[0180]

[0181]

[0182] By analyzing the data in Table 11, it can be seen that the posterior probabilities of node D2-2 state "2" "manual valve opening too small", D2-1 state "2" "solenoid valve opening too small", and D2-3 "filter blockage" are 0.5106, 0.3442, and 0.1306, respectively, which are much higher than the posterior probabilities of other nodes, indicating that these events are the main events leading to "insufficient hydrogen supply" and should be taken seriously in aircraft maintenance work to reduce their failure probability.

[0183] (2) Sensitivity analysis

[0184] The value of the failure probability change rate RoV can be obtained through the prior probability and the posterior probability, as shown in Table 12.

[0185] Table 12 Root dynamic node RoV value

[0186]

[0187]

[0188] The RoV values ​​of nodes D2-3, D2-2 state "2", and D2-1 state "2" are 934.38, 934.10, and 933.88, respectively, which are much higher than those of other nodes. Therefore, "filter blockage", "manual shut-off valve opening too small", and "solenoid valve opening too small" are the main contributing events to the consequence "insufficient hydrogen supply".

[0189] (3) Risk trend analysis

[0190] According to the results of reasoning diagnosis and sensitivity analysis, the nodes that cause the consequence of "insufficient hydrogen supply" are D2-1 state "2", D2-2 state "2", and D2-3.

[0191] Taking the node D2-1 state "2" "solenoid valve opening is too small" as an example, explain the reasons for the formation of risk trends.

[0192] like Figure 4 As shown, this is the risk trend of D2-1 status "2". Figure 4 The horizontal axis is the number of time slices, and the vertical axis is the probability. The risk trend is divided into three stages with the trend change point as the dividing point, among which:

[0193] In stage 1, the risk of valve opening being too small increases, indicating that the aircraft needs to perform flight missions requiring the solenoid valve to repeatedly open and close the valve, causing the solenoid valve to age. This is the initial stage of valve aging, and the electronic components in the solenoid valve have reduced ability to recognize the valve opening. The valve is slightly loose, causing the valve to open too large or too small, and this error gradually increases over time.

[0194] In stage 2, the probability trend of valve opening being too small begins to decrease, indicating that as the valve ages, the electronic components age seriously, the valve becomes seriously loose and cannot resist high-pressure hydrogen.

[0195] In stage 3, the probability trend of the valve opening being too small decreases and then converges to a lower probability value, indicating that although the valve is seriously aged, the valve opening resistance may increase or lose the ability to open due to corrosion, deformation and other reasons, and there is still a certain probability that the valve opening is too small.

[0196] The evaluation method for an aircraft onboard hydrogen storage system proposed in the present invention can realize quantitative risk evaluation of the risk diffusion and evolution characteristics process in a complex and dynamic onboard hydrogen storage system, making the risk evaluation result more in line with reality.

[0197] The risk transmission and diffusion path of the airborne hydrogen storage system is described by a complex network system. The risk status evaluation results of each node are determined based on the fuzzy dynamic Bayesian network graph structure, and the risk status evaluation results of each node are converted into fuzzy probability. As the probability information of the fuzzy dynamic node, the risk evolution process can be directly quantified. The obtained risk status evaluation results of each node are quantified through expert evaluation, which can intuitively show the change of risk probability and have a high degree of fit with the actual risk evolution law. By quantifying the expert evaluation results, a more objective data set can be obtained as the data basis for dynamic risk assessment without relying on the data source, which realizes accurate risk evolution dynamic assessment of the system when the data is insufficient and uncertain. The constructed fuzzy dynamic Bayesian network model is inferred and calculated, and the risk status of the risk node is determined by solving the probability information of each node to complete the risk assessment.

[0198] This method is based on the constructed fuzzy dynamic Bayesian network model and can provide a risk assessment template based on the fuzzy dynamic Bayesian network results. Compared with traditional risk assessment methods, this method avoids the technical defects of low risk assessment efficiency and poor accuracy caused by lack of empirical data and data analysis methods, and effectively improves the accuracy of risk assessment.

[0199] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

[0200] In addition, unless otherwise specified, all technical and scientific terms used in the present invention have the same meanings as commonly understood by those skilled in the art to which the present invention belongs. All documents mentioned in this specification are incorporated by reference to disclose and describe the methods related to the documents. In the event of any conflict with any incorporated document, the content of this specification shall prevail.

Claims

1. A risk assessment method for an aircraft onboard hydrogen storage system, characterized in that: The following steps are involved: According to the composition architecture and component working principle of the aircraft onboard hydrogen storage system, the component failure event is taken as the root dynamic node, the abnormal event of the component set that completes a specific function is taken as the intermediate dynamic node, the hydrogen state parameter of the physical process of the onboard hydrogen storage system is taken as the static node, and the functional failure event of the onboard hydrogen storage system is taken as the result node, and the risk state and normal state of each node are determined; According to the node information and state information of each node, a fuzzy dynamic Bayesian network graph structure is constructed; Based on the fuzzy dynamic Bayesian network graph structure, the risk status evaluation results of each node are determined, and the risk status evaluation results of each node are converted into fuzzy probabilities as the probability information of the fuzzy dynamic nodes; According to the fuzzy dynamic Bayesian network graph structure and the probability information of fuzzy dynamic nodes, a fuzzy dynamic Bayesian network model is constructed; The fuzzy dynamic Bayesian network model is solved, and the risk status of the risk node is determined by solving the probability information of each node, and the risk assessment of the onboard hydrogen storage system is carried out.

2. The risk assessment method for an aircraft onboard hydrogen storage system according to claim 1, characterized in that: The risk assessment of the airborne hydrogen storage system specifically includes reasoning diagnosis, sensitivity analysis and risk trend analysis, wherein: Reasoning diagnosis is to set the probability of the risk state of the risk node to 1, and use the reverse reasoning ability of the dynamic Bayesian network to update the fuzzy dynamic Bayesian network model to obtain the posterior probability of each dynamic node; according to the posterior probability of each dynamic node, the possibility of each dynamic node causing the risk to occur is judged; Sensitivity analysis is to calculate the failure probability change rate based on the prior probability and posterior probability of each dynamic node, and determine the degree of dependence of the risk node on each dynamic node; Risk trend analysis is based on the results of reasoning calculation of the fuzzy dynamic Bayesian network model. It combines the results of reasoning diagnosis and sensitivity analysis to identify the risk nodes that cause failure events, analyze the risk probability of the risk nodes, and complete the risk trend analysis.

3. The risk assessment method for an aircraft onboard hydrogen storage system according to claim 2, characterized in that: The probability information as the fuzzy dynamic node includes the following steps: Collect information of each expert and calculate expert weight; According to the expert evaluation results of each node of the dynamic Bayesian network, the consistency of expert opinions between every two experts is calculated; Combining the expert opinion consistency and expert weight, calculate the weighted consistency of every two experts' opinions, and so on, calculate the weighted consistency of all experts; The ratio of the weighted agreement between every two experts to all weighted agreements is taken as the relative agreement of the experts’ opinions. The consistency coefficient of each expert’s opinion is calculated by combining the expert weights with the relative agreement of the experts’ opinions. Through the consistency coefficient of each expert's opinion, the integrated result corresponding to the expert evaluation result of each node risk status is obtained, that is, the overall fuzzy number; The overall fuzzy number is defuzzified, and the fuzzy probability is obtained by calculating the fuzzy possible value, and the fuzzy probability is used as the probability information of the fuzzy dynamic node.

4. The risk assessment method for an aircraft onboard hydrogen storage system according to claim 3, characterized in that: The calculation of the weighted consistency of all experts specifically includes the following steps: Experts evaluate the possibility of the risk status of each node based on the fuzzy dynamic Bayesian network graph structure. The contribution of each expert's evaluation result to the integrated evaluation result depends on the expert weight and the consistency of expert opinions. Compare the opinions of each two experts on each node to determine the consistency of their evaluations of the node. The higher the consistency, the closer the evaluation results of the two experts are. Combining the expert opinion consistency and expert weights, calculate the weighted consistency of every two expert opinions until the calculation of the weighted consistency of all expert opinions is completed.

5. The risk assessment method for an aircraft onboard hydrogen storage system according to claim 4, characterized in that: The calculation formula for the expert opinion consistency is: In the formula, and They represent the evaluation fuzzy intervals of the x-th expert and the y-th expert, respectively.

6. The risk assessment method for an aircraft onboard hydrogen storage system according to claim 5, characterized in that: The calculation formula of the weighted consistency of the expert opinions is: In the formula, n represents the number of experts, A W (E x ) represents the weighted consistency of expert opinion of expert Ex; Ex and E y represent the xth expert and the yth expert respectively, W(E x ) and W(E y ) are experts E x and E y The weight of .

7. The risk assessment method for an aircraft onboard hydrogen storage system according to claim 6, characterized in that: The calculation formula of the overall fuzzy number is: In the formula, C c (E1) represents the expert opinion consistency coefficient of the first expert E1, represents the evaluation result of the first expert E1; C c (E n ) represents the nth expert E n The expert opinion consistency coefficient is Indicates the nth expert E n The evaluation results; The consistency coefficient C of the expert opinion of the xth expert Ex c (E x )for: C c (AND x )=βW(E x )+(1-β)A R (AND x ) The relative consistency of the expert opinion of the xth expert Ex is A R (E x )for: In the formula, C c (E x ) represents the expert opinion consistency coefficient of the xth expert Ex, and β is the relaxation factor.

8. A risk assessment system for an aircraft onboard hydrogen storage system, characterized in that: include: A fuzzy dynamic Bayesian network graph structure building module is used to take component failure events as root dynamic nodes, abnormal events of component sets that complete specific functions as intermediate dynamic nodes, hydrogen state parameters of the physical process of the onboard hydrogen storage system as static nodes, and functional failure events of the onboard hydrogen storage system as result nodes according to the composition architecture and component working principles of the aircraft onboard hydrogen storage system, and determine the risk state and normal state of each node; According to the node information and state information of each node, a fuzzy dynamic Bayesian network graph structure is constructed; The probability information acquisition module is used to determine the risk status evaluation result of each node based on the fuzzy dynamic Bayesian network graph structure, and convert the risk status evaluation result of each node into fuzzy probability as the probability information of the fuzzy dynamic node; A fuzzy dynamic Bayesian network model building module is used to build a fuzzy dynamic Bayesian network model according to the fuzzy dynamic Bayesian network graph structure and the probability information of the fuzzy dynamic nodes; The risk assessment module is used to solve the fuzzy dynamic Bayesian network model, determine the risk status of the risk node by solving the probability information of each node, and conduct risk assessment on the airborne hydrogen storage system.

9. A computer device, characterized in that: The computer device includes a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor performs the following steps: According to the composition architecture and component working principle of the aircraft onboard hydrogen storage system, the component failure event is taken as the root dynamic node, the abnormal event of the component set that completes a specific function is taken as the intermediate dynamic node, the hydrogen state parameter of the physical process of the onboard hydrogen storage system is taken as the static node, and the functional failure event of the onboard hydrogen storage system is taken as the result node, and the risk state and normal state of each node are determined; According to the node information and state information of each node, a fuzzy dynamic Bayesian network graph structure is constructed; Based on the fuzzy dynamic Bayesian network graph structure, the risk status evaluation results of each node are determined, and the risk status evaluation results of each node are converted into fuzzy probabilities as the probability information of the fuzzy dynamic nodes; According to the fuzzy dynamic Bayesian network graph structure and the probability information of fuzzy dynamic nodes, a fuzzy dynamic Bayesian network model is constructed; The fuzzy dynamic Bayesian network model is solved, and the risk status of the risk node is determined by solving the probability information of each node, and the risk assessment of the onboard hydrogen storage system is carried out.

10. A computer-readable storage medium, characterized in that: The computer readable storage medium stores a computer program, and when the computer program is executed by a processor, the processor performs the following steps: According to the composition architecture and component working principle of the aircraft onboard hydrogen storage system, the component failure event is taken as the root dynamic node, the abnormal event of the component set that completes a specific function is taken as the intermediate dynamic node, the hydrogen state parameter of the physical process of the onboard hydrogen storage system is taken as the static node, and the functional failure event of the onboard hydrogen storage system is taken as the result node, and the risk state and normal state of each node are determined; According to the node information and state information of each node, a fuzzy dynamic Bayesian network graph structure is constructed; Based on the fuzzy dynamic Bayesian network graph structure, the risk status evaluation results of each node are determined, and the risk status evaluation results of each node are converted into fuzzy probabilities as the probability information of the fuzzy dynamic nodes; According to the fuzzy dynamic Bayesian network graph structure and the probability information of fuzzy dynamic nodes, a fuzzy dynamic Bayesian network model is constructed; The fuzzy dynamic Bayesian network model is solved, and the risk status of the risk node is determined by solving the probability information of each node, and the risk assessment of the onboard hydrogen storage system is carried out.

Images