A complex system risk assessment method based on complex network

Abstract

The application relates to a complex system risk assessment method based on a complex network, and belongs to the field of complex system risk assessment. For all risk nodes, a risk propagation probability matrix between the risk nodes is generated through a P-I probability matrix. Meanwhile, according to the characteristics of the risk nodes, a suitable complex network model is selected as the basis of a risk propagation network graph of the risk nodes. The risk propagation probability matrix between the nodes is taken as the output to output the risk propagation network graph of the system. Whether the actual situation is met is judged. If not, the P-I probability matrix is reconstructed, the complex network model is reselected, and the operation is performed again until a risk propagation network graph more in line with the actual situation is output. Meanwhile, the Monte Carlo thought is used to analyze the degree distribution of each node, and the degree distribution of the node is taken as an evaluation index for evaluating the influence size of the risk node in the system. The application realizes clearer understanding of the risk propagation in the complex system, and simplifies the risk assessment of the complex system.

Description

Technical Field

[0001] This invention belongs to the field of risk assessment for complex systems, and relates to a method for risk assessment of complex systems based on complex networks. Background Technology

[0002] Networks possessing some or all of the properties of self-organization, self-similarity, attractors, small-world characteristics, and scale-free operation are called complex networks. Complex networks offer a perspective and methodology for studying complex systems, primarily focusing on the interactions between individuals within the system (a type of topology). With current development, various systems are becoming increasingly complex, requiring consideration of more and more factors for risk assessment, making risk assessment increasingly difficult. Traditional risk assessment methods based on the PRA framework, often employing methods like FTA and ET, primarily focus on risk assessment at a single point, which becomes extremely cumbersome when assessing complex systems. Summary of the Invention

[0003] The technical problem solved by this invention is to overcome the shortcomings of the prior art and propose a risk assessment method for complex systems based on complex networks, which enables a clearer understanding of risk propagation within complex systems and simplifies risk assessment of complex systems.

[0004] The solution of the present invention is:

[0005] A risk assessment method for complex systems based on complex networks, comprising:

[0006] Step 1: Divide the complex system into multiple interconnected subsystems, with each subsystem serving as a risk node;

[0007] Step 2: Perform FMECA analysis on each risk node to obtain the impact and probability of occurrence of each risk node; construct a PI probability matrix based on the probability of occurrence and the degree of impact of each risk node.

[0008] Step 3: Obtain the risk propagation probability matrix between each risk node using the PI probability matrix;

[0009] Step 4: Based on the propagation characteristics between risk nodes, select a suitable complex network model;

[0010] Step 5: Using the risk propagation probability matrix between risk nodes as input to the complex network model, improve the selected complex network model and output the risk propagation network graph.

[0011] Step Six: Evaluate the obtained risk propagation network diagram; if it does not match the actual situation, return to Step Three and re-calculate the risk propagation probability matrix or reselect a complex network model; continue until the obtained risk propagation network diagram matches the actual situation, then proceed to Step Seven;

[0012] Step 7: Use the Monte Carlo algorithm to calculate the risk propagation network diagram that conforms to the actual situation multiple times, and use the average of the multiple calculation results as the degree distribution of each risk node in the risk propagation network diagram; use the degree distribution of each risk node as the evaluation index of each risk node; and use the evaluation index of each risk node to evaluate the risk impact of each subsystem.

[0013] In the aforementioned risk assessment method for complex systems based on complex networks, in step one, when the complex system is divided into subsystems, the subsystem structure is divided according to function or composition.

[0014] In the aforementioned risk assessment method for complex systems based on complex networks, step two, the method for constructing the PI probability matrix, is as follows:

[0015] The probability and impact of each risk node are classified into different levels to obtain the rows and columns of the PI probability matrix, thus realizing the construction of the PI probability matrix.

[0016] In the aforementioned risk assessment method for complex systems based on complex networks, the rules for constructing the PI probability matrix are as follows:

[0017] Its risk probability range and impact level include the probability range and impact scope of all identified risk factors.

[0018] In the aforementioned risk assessment method for complex systems based on complex networks, the method for constructing the risk propagation probability matrix between risk nodes in step three is as follows:

[0019] Five levels are defined: very high, high, medium, low, and very low. Different levels correspond to different P values. The degree of impact is comprehensively evaluated by considering multiple factors, including the impact on quality, technical indicators, schedule, and cost. The degree of impact is divided into levels based on comprehensive consideration, and different levels correspond to different I values. Finally, the risk score is the value P*I of the risk propagation probability matrix. Obtaining all values ​​P*I of the risk propagation probability matrix completes the construction of the risk propagation probability matrix.

[0020] In the above-mentioned risk assessment method for complex systems based on complex networks, step four involves four types of complex network models: regular graph, ER random network model, BA scale-free network model, and small-world network model.

[0021] Rule graph: Specifies the number of nodes and the number of edges connected to each node;

[0022] ER stochastic network model: The connections between network nodes are randomly arranged, resulting in strong randomness;

[0023] BA scale-free network model: characterized by the degree distribution of nodes following a power law, and its node size is scalable compared to ER random network;

[0024] Small-world network models: Networks exhibiting small-world characteristics include the WS small-world model and the NS small-world model. The WS small-world model is a single-parameter small-world network model that falls between regular and completely random networks. This model reflects the two phenomena of small average path length and large clustering coefficient in social networks. The NS model replaces the randomized reconnection in the WS small-world model with randomized edge addition. The advantage of the NS model is that it simplifies theoretical analysis because the WS small-world model has isolated nodes, but the NS model does not.

[0025] In the aforementioned risk assessment method for complex systems based on complex networks, when the user is sufficiently small and N is sufficiently large, the NS small-world model is equivalent to the WS small-world model.

[0026] In the aforementioned risk assessment method for complex systems based on complex networks, step five involves improving the selected complex network model as follows:

[0027] When it is a rule graph, the corresponding rule is matched by inputting the range of values ​​in the risk propagation probability matrix;

[0028] When it is an ER random network model, modify its fixed edge probabilities to the values ​​of the risk propagation probability matrix, and output the risk propagation network graph of the system;

[0029] When using a scale-free BA network model, a system risk network diagram is generated by starting from a certain node and using the risk propagation probability matrix as the risk propagation path.

[0030] When using a small-world network model, the randomized edge addition strategy is changed to adding edges using a risk probability propagation matrix to generate a system risk network graph.

[0031] In the aforementioned risk assessment method for complex systems based on complex networks, step six involves evaluating the risk propagation network graph as follows:

[0032] We use a method of comparing the risk propagation algorithm diagram with the system fault tree diagram and combining it with historical fault data as a reference to evaluate whether the risk propagation algorithm diagram matches the actual situation.

[0033] In the aforementioned risk assessment method for complex systems based on complex networks, step seven involves performing multiple calculations using the Monte Carlo algorithm to reduce the impact of randomness on the result error, with the number of calculations not less than 100.

[0034] The advantages of this invention compared to the prior art are:

[0035] (1) The present invention is a risk assessment method for complex systems based on complex networks. It is applicable to complex systems such as traditional complex equipment and aerospace. When assessing the risk of a complex system, it can clearly understand the propagation process of internal risks and identify subsystems with high risk factors within the complex system. This provides a certain reference for subsequent risk analysis and can also reduce the difficulty of risk assessment for complex systems.

[0036] (2) Before using this method, the complex system needs to be decomposed into multiple interrelated subsystems. Each subsystem serves as a risk node of the system, thereby realizing the point-dispersed processing of the complex system and improving the adaptability of different complex systems.

[0037] (3) For the output network graph, the present invention uses the Monte Carlo method to analyze the degree distribution of each node and uses the degree distribution of the node as an evaluation index to assess the impact of risk nodes on the system, thus enabling the analysis of complex systems that could not be evaluated in the past; at the same time, it simplifies the risk assessment of complex systems.

[0038] (4) The evaluation of the results of this invention adopts the Monte Carlo method to perform multiple calculations, which can reduce the influence of randomness on the result error. Therefore, the number of experiments should not be less than 100, so as to minimize the influence of randomness on the result error. Attached Figure Description

[0039] Figure 1 This is a flowchart of the risk assessment process for complex systems according to the present invention;

[0040] Figure 2 This is a schematic diagram showing the system breakdown according to an embodiment of the present invention;

[0041] Figure 3 This is a risk propagation network diagram according to an embodiment of the present invention. Detailed Implementation

[0042] The present invention will be further described below with reference to the embodiments.

[0043] This invention provides a risk assessment method for complex systems based on complex networks, which enables a clearer understanding of risk propagation within complex systems and simplifies risk assessment of complex systems.

[0044] Risk assessment methods for complex systems based on complex networks, such as Figure 1 As shown, the specific steps include the following:

[0045] Step 1: Divide the complex system into multiple interconnected subsystems, with each subsystem serving as a risk node. When dividing the complex system into subsystems, the subsystem structure should be divided according to function or composition.

[0046] Step 2: Perform FMECA analysis on each risk node to obtain the impact and probability of occurrence of each risk node; construct the PI probability matrix based on the probability of occurrence and the degree of impact of each risk node.

[0047] The method for constructing the PI probability matrix is ​​as follows:

[0048] The probability and impact of each risk node are classified into different levels to obtain the rows and columns of the PI probability matrix, thus realizing the construction of the PI probability matrix.

[0049] The rules for constructing the PI probability matrix are as follows:

[0050] Its risk probability range and impact level include the probability range and impact scope of all identified risk factors.

[0051] Step 3: Obtain the risk propagation probability matrix between each risk node using the PI probability matrix; the method for constructing the risk propagation probability matrix between each risk node is as follows:

[0052] Five levels are defined: very high, high, medium, low, and very low. Different levels correspond to different P values. The degree of impact is comprehensively evaluated by considering multiple factors, including the impact on quality, technical indicators, schedule, and cost. The degree of impact is divided into levels based on comprehensive consideration, and different levels correspond to different I values. Finally, the risk score is the value P*I of the risk propagation probability matrix. Obtaining all values ​​P*I of the risk propagation probability matrix completes the construction of the risk propagation probability matrix.

[0053] Step 4: Select a suitable complex network model based on the propagation characteristics between risk nodes.

[0054] Complex network models are classified into four types: regular graph, ER random network model, BA scale-free network model, and small-world network model.

[0055] Rule graph: Specifies the number of nodes and the number of edges connected to each node.

[0056] ER stochastic network model: The connections between network nodes are randomly arranged, resulting in strong randomness.

[0057] BA scale-free network model: characterized by the degree distribution of nodes following a power law, and its node size is scalable compared to ER random network.

[0058] Small-world network models: Networks exhibiting small-world characteristics include the WS small-world model and the NS small-world model. The WS small-world model is a single-parameter small-world network model that falls between regular and completely random networks. This model reflects the two phenomena of small average path length and large clustering coefficient in social networks. The NS model replaces the randomized reconnection in the WS small-world model with randomized edge addition. The advantage of the NS model is that it simplifies theoretical analysis because the WS small-world model has isolated nodes, but the NS model does not.

[0059] When the user is small enough and N is large enough, the NS small-world model is equivalent to the WS small-world model.

[0060] Step 5: Using the risk propagation probability matrix between risk nodes as input to the complex network model, improve the selected complex network model and output the risk propagation network graph.

[0061] The method for improving the selected complex network model is as follows:

[0062] When it is a rule graph, the range of values ​​in the risk propagation probability matrix is ​​used to match the corresponding rule.

[0063] When using an ER stochastic network model, modify its fixed edge probabilities to the values ​​of the risk propagation probability matrix, and output the risk propagation network graph of the system.

[0064] When using a scale-free network model (BA), a system risk network diagram is generated by starting from a certain node and using the risk propagation probability matrix as the risk propagation path.

[0065] When using a small-world network model, the randomized edge addition strategy is changed to adding edges using a risk probability propagation matrix to generate a system risk network graph.

[0066] Step Six: Evaluate the obtained risk propagation network diagram; if it does not match the actual situation, return to Step Three and re-calculate the risk propagation probability matrix or reselect a complex network model; continue until the obtained risk propagation network diagram matches the actual situation, then proceed to Step Seven.

[0067] The method for evaluating risk transmission network diagrams is as follows:

[0068] We use a method of comparing the risk propagation algorithm diagram with the system fault tree diagram and combining it with historical fault data as a reference to evaluate whether the risk propagation algorithm diagram matches the actual situation.

[0069] Step 7: Use the Monte Carlo algorithm to calculate the risk propagation network diagram that conforms to the actual situation multiple times, and use the average of the multiple calculation results as the degree distribution of each risk node in the risk propagation network diagram; use the degree distribution of each risk node as the evaluation index of each risk node; and use the evaluation index of each risk node to evaluate the risk impact of each subsystem.

[0070] This invention reduces the impact of randomness on the result error by performing multiple calculations using the Monte Carlo algorithm, with the number of calculations not less than 100.

[0071] Example

[0072] Taking the space station construction mission system as the research object, the specific steps include:

[0073] S1. Based on function, the system is divided into 13 subsystems: space station system, CZ-5B launch vehicle system, astronaut system, space application system, CZ-7 launch vehicle system, cargo spacecraft system, manned spacecraft system, CZ-2F launch vehicle system, Jiuquan launch site system, Hainan launch site system, telemetry and communication system, landing site system, and optical module system. (For example...) Figure 2 As shown.

[0074] S2. Based on the risk factor identification method, the PI probability index tables of the PI probability matrix diagrams formulated after risk identification of different subsystems are shown in Table 1 and Table 2, and the obtained PI matrix is ​​shown in Table 3.

[0075] Table 1

[0076] probability of occurrence Possible incidence Sorting and retrieving values Very high It will definitely happen; a project may happen more than once. 0.9 high It often happens once in 10 projects. 0.7 medium Sometimes it happens once in 100 projects. 0.5 Low It rarely happens once in 1000 projects. 0.3 Extremely low It is highly unlikely that this will happen once in more than 10,000 projects. 0.1

[0077] Table 2

[0078]

[0079]

[0080] Table 3

[0081]

[0082] S3. For the risk items of the subsystem, combine historical failure data and FMECA analysis method to analyze the probability of risk occurrence of each subsystem and its impact on different subsystems, and then combine the PI probability matrix to obtain the risk propagation probability matrix of the system.

[0083] S4. Based on the characteristics of system risk propagation, the ER random network model is selected as the risk propagation network model for this system.

[0084] S5. The fixed edge probabilities in the ER random network are improved to the probability values ​​of the risk propagation probability matrix between subsystems. Using the subsystems as risk nodes and the subsystems as unit inputs, the risk propagation network graph of the system is obtained, as shown below. Figure 3 As shown.

[0085] S6. The obtained risk propagation network diagram is evaluated, mainly by comparing the propagation paths of subsystem risks in the risk propagation network diagram with the results of historical failure data and FMECA analysis. The results show that the risk propagation network diagram generated by the improved ER random network can represent the risk propagation between subsystems to a certain extent and meets the requirements.

[0086] S7. For this network diagram, Monte Carlo simulation was used to conduct multiple experiments and take the average value. The average degree distribution of the subsystems is shown in Table 4, where the results are rounded to the nearest whole number.

[0087] Table 4

[0088] Subsystem Average In-degree Average out-degree Degree distribution Landing site system 0 2 2 CZ-2 launch vehicle system 2 1 3 Manned spacecraft system 6 2 8 Jiuquan Launch Site System 0 1 1 Astronaut System 5 3 8 Measurement and control communication system 0 8 8 Space station system 6 4 10 Space application system 5 5 10 CZ-7 launch vehicle system 2 1 3 Cargo spacecraft system 5 3 8 Optical cabin system 4 2 6 Hainan Launch Site System 0 2 2 CZ-5B launch vehicle system 2 2 4

[0089] S8. Based on the degree distribution of subsystems, the importance ranking of the subsystems in a complex system to the overall risk of the system can be obtained, which can clearly identify the subsystems in the system that have a greater impact on the overall risk.

[0090] Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make possible changes and modifications to the technical solutions of the present invention by utilizing the methods and techniques disclosed above without departing from the spirit and scope of the present invention. Therefore, any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solutions of the present invention shall fall within the protection scope of the technical solutions of the present invention.

Claims

1. A risk assessment method for complex systems based on complex networks, characterized in that: include: Step 1: Divide the complex system into multiple interconnected subsystems, with each subsystem serving as a risk node; Step 2: Perform FMECA analysis on each risk node to obtain the impact and probability of occurrence of each risk node; construct a PI probability matrix based on the probability of occurrence and the degree of impact of each risk node. Step 3: Obtain the risk propagation probability matrix between each risk node using the PI probability matrix; Step 4: Based on the propagation characteristics between risk nodes, select a suitable complex network model; Complex network models are classified into four types: regular graph, ER random network model, BA scale-free network model, and small-world network model. Rule graph: Specifies the number of nodes and the number of edges connected to each node; ER stochastic network model: The connections between network nodes are randomly arranged, resulting in strong randomness; BA scale-free network model: characterized by the degree distribution of nodes following a power law, and its node size is scalable compared to ER random network; Small-world network models: Networks exhibiting small-world characteristics include the WS small-world model and the NS small-world model. The WS small-world model is a single-parameter small-world network model that falls between regular and completely random networks. This model reflects the small average path length and large clustering coefficient of social networks. The NS model replaces the randomized reconnection in the WS small-world model with randomized edge addition. The advantage of the NS model is that it simplifies theoretical analysis because the WS small-world model has isolated nodes, while the NS model does not. Step 5: Using the risk propagation probability matrix between risk nodes as input to the complex network model, improve the selected complex network model and output the risk propagation network graph. The method for improving the selected complex network model is as follows: When it is a rule graph, the corresponding rule is matched by inputting the range of values ​​in the risk propagation probability matrix; When it is an ER random network model, modify its fixed edge probabilities to the values ​​of the risk propagation probability matrix, and output the risk propagation network graph of the system; When using a scale-free network model (BA), a system risk network diagram is generated by starting from a certain node and using the risk propagation probability matrix as the risk propagation path. When it is a small-world network model, its randomized edge addition strategy is changed to adding edges by using a risk probability propagation matrix to generate a system risk network graph. Step Six: Evaluate the obtained risk propagation network diagram; if it does not match the actual situation, return to Step Three and re-calculate the risk propagation probability matrix or reselect a complex network model; continue until the obtained risk propagation network diagram matches the actual situation, then proceed to Step Seven; Step 7: Use the Monte Carlo algorithm to calculate the risk propagation network diagram that conforms to the actual situation multiple times, and use the average of the multiple calculation results as the degree distribution of each risk node in the risk propagation network diagram; use the degree distribution of each risk node as the evaluation index of each risk node; and use the evaluation index of each risk node to evaluate the risk impact of each subsystem.

2. The method for risk assessment of complex systems based on complex networks according to claim 1, characterized in that: In step one, when breaking down a complex system into subsystems, the subsystem structure is broken down according to function or composition.

3. The method for risk assessment of complex systems based on complex networks according to claim 1, characterized in that: In step two, the method for constructing the PI probability matrix is ​​as follows: The probability and impact of each risk node are classified into different levels to obtain the rows and columns of the PI probability matrix, thus realizing the construction of the PI probability matrix.

4. The method for risk assessment of complex systems based on complex networks according to claim 3, characterized in that: The rules for constructing the PI probability matrix are as follows: Its risk probability range and impact level include the probability range and impact scope of all identified risk factors.

5. The method for risk assessment of complex systems based on complex networks according to claim 4, characterized in that: In step three, the method for constructing the risk propagation probability matrix between risk nodes is as follows: Five risk levels are defined: very high, high, medium, low, and very low. Different levels correspond to different P-values. The degree of impact is comprehensively assessed based on multiple factors, including quality, technical specifications, schedule, and cost. The impact level is then categorized, with different levels corresponding to different I-values. The final risk score is the value P of the risk propagation probability matrix. I; Obtain all values ​​P of the risk propagation probability matrix. I, that is, the completed risk propagation probability matrix.

6. The method for risk assessment of complex systems based on complex networks according to claim 1, characterized in that: In step six, the method for evaluating the risk transmission network diagram is as follows: We use a method of comparing the risk propagation algorithm diagram with the system fault tree diagram and combining it with historical fault data as a reference to evaluate whether the risk propagation algorithm diagram matches the actual situation.

7. The method for risk assessment of complex systems based on complex networks according to claim 1, characterized in that: In step seven, the Monte Carlo algorithm is used to perform multiple calculations to reduce the impact of randomness on the result error, with the number of calculations not less than 100.

Images