A Bayesian analysis method and apparatus based on sysML and knowledge graph

Abstract

This invention discloses an automatic Bayesian network construction method and apparatus based on SysML models and knowledge graphs. The method includes: parsing the SysML system model, automatically extracting the system structure, interfaces, and redundancy relationships, and generating Bayesian network node connections; performing logical optimization on the Bayesian network node connections, merging redundant nodes, identifying voting / reserve gates, and obtaining a minimized optimized graph structure; introducing a knowledge graph to supplement fault causal relationships and probability parameters, and constructing a Bayesian network model; and using Bayesian back-inference to quantify the contribution of each node to the top-level fault, outputting the fault propagation path, weak points, and redundant design schemes. This invention replaces the traditional method of manually building fault models, achieving full automation of the quantitative analysis of reliability in complex systems, significantly improving the efficiency of fault prediction, location, and design optimization, and can be widely applied in high-safety fields such as aerospace, rail transportation, and nuclear power.

Description

Technical Field

[0001] This invention relates to the field of systems engineering data processing technology, and in particular to a Bayesian analysis method and apparatus based on sysML and knowledge graphs. Background Technology

[0002] With the widespread research and application of MBSE-related digital development methods, technologies, and software tools both domestically and internationally, its application has gradually expanded from functional performance requirement definition and design analysis to professional fields such as project management, cost management, and general quality characteristics. Currently, MBSE research and application in various industries in my country mainly focus on system / product design and analysis, and is gradually expanding to areas such as general quality characteristics. However, the overall general quality characteristic assurance based on MBSE is still in its initial, exploratory, and preliminary application stages.

[0003] The System Modeling Language (SysML) provides a powerful tool for MBSE (Balanced Entity Sequence) to accurately represent complex systems and address related technical issues such as security and reliability. However, because SysML itself does not support quantitative analysis methods such as Bayesian network analysis, this significantly limits the direct construction of Bayesian networks using SysML. Nevertheless, thanks to SysML's extension mechanisms, some research has extended the SysML model to extract the logical information needed to generate Bayesian networks from internal module diagrams and state machine diagrams. However, this approach has a low level of automation and cannot directly generate Bayesian networks.

[0004] Meanwhile, in modern engineering and reliability analysis of complex systems, the knowledge graph core undertakes the functions of knowledge recording and relation storage. It can generate Bayesian networks (BN) on its own or be integrated with SysML to construct BN. The combination of the two can significantly improve the overall capabilities of system fault analysis, risk prediction and decision support. Summary of the Invention

[0005] The technical problem this invention aims to solve is to provide a Bayesian analysis method and apparatus based on SysML and knowledge graphs. This includes parsing a SysML system model, automatically extracting system structure, interfaces, and redundancy relationships, and generating Bayesian network node connections; logically optimizing the Bayesian network node connections by merging redundant nodes and identifying voting / reserve gates to obtain a minimized optimized graph structure; introducing knowledge graphs to supplement fault causal relationships and probability parameters, and constructing a Bayesian network model; and using Bayesian back-inference to quantify the contribution of each node to the top-level fault, outputting the fault propagation path, weak points, and redundant design schemes. This invention replaces the traditional method of manually building fault models, achieving full automation of the quantitative analysis of the reliability of complex systems, significantly improving the efficiency of fault prediction, location, and design optimization, and can be widely applied in high-safety fields such as aerospace, rail transportation, and nuclear power.

[0006] To address the aforementioned technical problems, a first aspect of the present invention discloses a Bayesian analysis method, the method comprising: Obtain the data to be processed; The data to be processed is subjected to model building processing to obtain model building information; The model building information is subjected to structural recognition processing to obtain model recognition result information; The model recognition results are post-processed to obtain Bayesian results.

[0007] As an optional implementation, in the first aspect of the present invention, the step of performing structural recognition processing on the model building information to obtain model recognition result information includes: The model building information is processed to extract model node information; the model node information includes several node information. The model node information is processed to obtain model recognition result information.

[0008] As an optional implementation, in the first aspect of the present invention, the post-processing of the model recognition result information to obtain Bayesian result information includes: The model recognition result information is optimized to obtain optimized model data information; The optimized model data is processed to obtain Bayesian results.

[0009] A second aspect of this invention discloses a Bayesian analysis method based on sysML and knowledge graphs, the method comprising: S1, Obtain the modeling language model information of the system to be processed; S2, process the modeling language model information of the system to be processed to obtain Bayesian network node connection relationship information; S3, process the connection relationship information of the Bayesian network nodes to obtain the target Bayesian network information; S4, process the target Bayesian network to obtain the target Bayesian network information model result information.

[0010] As an optional implementation, in a second aspect of the present invention, processing the modeling language model information of the system to be processed to obtain Bayesian network node connection relationship information includes: S21, perform information extraction processing on the modeling language model information of the system to be processed to obtain model node information; the model node information includes several node information. S22, The model node information is processed to obtain Bayesian network node connection relationship information.

[0011] As an optional implementation, in a second aspect of the present invention, processing the model node information to obtain Bayesian network node connection relationship information includes: S221, The model node information is traversed to obtain logic gate node information; the logic gate node information is the node information in the model node information that does not have a predecessor node information. S222, add the logic gate node information to the logic gate queue information; S223, determine whether the number of successor node information in the logic gate node information is greater than 1, and obtain the first judgment result; When the first judgment result is negative, determine whether the number of successor node information in the logic gate node information is less than 1, and obtain the second judgment result; When the second judgment result is negative, the logic gate node information and the successor node information are first marked and merged to obtain OR gate information; the OR gate information is added to the logic gate list information, and S226 is executed. When the second judgment result is yes, the logic gate node information is determined to be basic event information; the basic event information is added to the logic gate list information, and S227 is executed; When the first judgment result is yes, execute S224; S224, determine whether there is redundant information in any of the subsequent node information, and obtain a third judgment result; When the third judgment result is yes, the logic gate node information and all the successor node information are subjected to a second marking and merging process to obtain the reserve gate information; the reserve gate information is added to the logic gate list information, and S226 is executed; If the third judgment result is negative, execute S225; S225, determine whether voting information exists in any of the subsequent node information, and obtain the fourth determination result; When the fourth judgment result is yes, the logic gate node information and all the successor node information are subjected to a third marking and merging process to obtain the voting gate information; the voting gate information is added to the logic gate list information; When the fourth judgment result is negative, the logic gate node information and all the successor node information are subjected to a fourth marking and merging process to obtain parallel information; the parallel information is added to the logic gate list information. S226, add all the successor node information to the logic gate queue information; S227, Remove the logic gate node information from the logic gate queue information; S228, determine whether the logic gate queue information contains the successor node information, and obtain the fifth determination result; When the fifth judgment result is yes, determine any of the successor node information in the logic gate queue information as the logic gate node information, and execute S223; If the result of the fifth judgment is negative, execute S229; S229, Process the logic gate list information to obtain Bayesian network node connection relationship information.

[0012] As an optional implementation, in a second aspect of the present invention, processing the Bayesian network node connection relationship information to obtain target Bayesian network information includes: S31, The Bayesian network node connection relationship information is processed to obtain optimized Bayesian network node connection relationship information; S32, process the optimized Bayesian network node connection relationship information to obtain the target Bayesian network information.

[0013] A third aspect of this invention discloses a Bayesian analysis method based on knowledge graphs, the method comprising: Obtain the knowledge graph information to be processed; The knowledge graph information to be processed is processed to obtain knowledge graph information; The knowledge graph information is optimized to obtain the target knowledge graph information; The target knowledge graph information is processed to obtain the target knowledge graph result information.

[0014] As an optional implementation, in a third aspect of the present invention, processing the knowledge graph information to be processed to obtain knowledge graph information includes: The knowledge graph information to be processed is subjected to information extraction processing to obtain model node information; the model node information includes several node information. The model node information is processed to obtain knowledge graph information.

[0015] The fourth aspect of the present invention discloses a computer-readable storage medium storing computer instructions, which, when invoked, are used to perform some or all of the steps of the Bayesian analysis method based on sysML and knowledge graphs as described in any one of the second aspects of the present invention.

[0016] Compared with the prior art, the embodiments of the present invention have the following beneficial effects: In fields such as reliability design of complex systems, engineering project design and manufacturing, and reliability expertise, SysML can automatically construct efficient and accurate Bayesian networks, providing strong support for related work. At the same time, by constructing Bayesian network information through knowledge graphs, the system's fault analysis, risk prediction, and decision support capabilities can be further improved.

[0017] By using Bayesian analysis to perform reverse reasoning on the knowledge graph and the target Bayesian network, it is possible to quantify the contribution of each potential fault node to the target fault, thus providing scientific decision support. Furthermore, by automatically performing reverse reasoning on the Bayesian network model based on known fault information and logical relationships, Bayesian analysis helps to quantify the contribution of each potential fault node to the target fault, thereby providing scientific decision support. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a flowchart of a Bayesian analysis method disclosed in an embodiment of the present invention. Figure 2 This is a satellite component structure module definition diagram (BDD) for a Bayesian analysis method disclosed in an embodiment of the present invention. Figure 3This is an internal module diagram (IBD) illustrating the interconnectivity of individual satellite components within a Bayesian analysis method disclosed in an embodiment of the present invention. Figure 4 This diagram illustrates the .bayes (left) and .json (right) format files automatically generated by a Bayesian analysis method disclosed in an embodiment of the present invention. Figure 5 This is a diagram illustrating the Bayesian results of a Bayesian analysis method disclosed in an embodiment of the present invention. Figure 6 This is a flowchart illustrating a Bayesian analysis method based on sysML and knowledge graphs disclosed in an embodiment of the present invention. Figure 7 This is a schematic diagram of the structure of a Bayesian analysis device based on sysML and knowledge graph disclosed in an embodiment of the present invention; Figure 8 This is a schematic diagram of another Bayesian analysis device based on sysML and knowledge graph disclosed in an embodiment of the present invention; Figure 9 This is a schematic diagram of the three data domains of a Bayesian analysis method disclosed in an embodiment of the present invention. Detailed Implementation

[0020] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] The terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this invention are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, apparatus, product, or device that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or devices.

[0022] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0023] The knowledge graph in the embodiments and claims of this invention refers to knowledge graph information, which is a type of data information that stores and displays entities, relationships, and their attributes in a graph structure. In the context of sysML analysis and Bayesian networks, knowledge graph information provides a visual representation of system structure, fault propagation paths, node characteristics, and their interrelationships, supporting fault diagnosis, reasoning, prediction, and optimization.

[0024] This invention discloses a Bayesian analysis method and apparatus based on sysML and knowledge graphs. Based on sysML and knowledge graphs, it can automatically construct efficient and accurate Bayesian networks, which is beneficial for supporting the reliability design of complex systems. This provides strong support for engineering project design, manufacturing, and reliability-related work. By constructing Bayesian network information through knowledge graphs, it can improve the system's fault analysis, risk prediction, and decision support capabilities. Detailed descriptions follow.

[0025] Example 1 See Figures 1-5 This invention discloses a Bayesian analysis method. This Bayesian analysis method is applied in a Bayesian analysis device, such as a local server or cloud server used for Bayesian analysis optimization management, etc., and this invention is not limited thereto. The Bayesian analysis method may include the following operations: Data to be processed: The data to be processed includes three aspects: database / data lake, knowledge base, and model library. The database / data lake primarily stores, processes, and applies multi-source, multi-dimensional data throughout the entire lifecycle of the application system. The knowledge base mainly extracts knowledge to form a system that balances expert experience with objective product laws. The model library mainly constructs and stores various task assurance models that serve business applications. Based on the database / data lake, knowledge base, and model library, an innovative "three-domain, four-fusion" data technology solution is proposed: the integration of big data across the product domain, usage domain, and ecosystem domain (top-down, front-end, internal / external, and data-real-world), empowering task assurance technology to achieve enhanced capabilities in both high quality and high efficiency.

[0026] (1) Database / Data Lake This primarily includes various types of data collected extensively throughout the entire process of aerospace systems engineering mission assurance. Considering the wide range of sources, diverse data types, and large volume of aerospace engineering data, and to facilitate data management and analysis, the relevant data information for aerospace systems engineering mission assurance is categorized from the perspectives of data source and application benefits, forming "three-domain" data: product domain big data, usage domain big data, and ecosystem domain big data. Specifically, as follows... Figure 9 As shown.

[0027] Product domain big data refers to data generated during the development process of system design, manufacturing, and testing. It mainly includes user needs, qualitative and quantitative indicator requirements, attributes such as functions, performance, and general quality characteristics, architectural composition, development processes such as design, manufacturing, and testing, requirement compliance, quality issues, quality management system, and development costs.

[0028] Domain big data refers to the data generated during the use of the system. It mainly includes the system's usage tasks, usage environment such as meteorology, geography, and human factors, usage process, failures, damages, and accidents, usage support, maintenance support, and supply support processes, support resources such as personnel, spare parts, and support equipment and tools, support organization, support model, support effectiveness, and usage and support costs.

[0029] Ecosystem domain big data encompasses data from all stakeholders and business processes across the entire system's value chain. It primarily involves stakeholders such as system users, developers, and managers, as well as different value chain players including product manufacturers, suppliers, partners, and competitors. Ecosystem domain data includes the needs, capabilities, and relationships of each party, along with business processes and data dimensions related to materials, personnel, funding, and information exchange.

[0030] Furthermore, based on the data structure type, multi-source heterogeneous data can be divided into a limited number of categories to facilitate data processing. These categories can generally be divided into structured rule knowledge, tabular semi-structured data, text unstructured data, and spacecraft test and telemetry data, etc.

[0031] 1) Structured rule knowledge Generally, this refers to data that can be represented and stored using relational databases and logically expressed using two-dimensional tables. This type of data storage and arrangement conforms to certain rules. For example, in spacecraft fault knowledge, expert rules are typically stored in the form of structured data. Expert rules consist of relevant rules set by domain experts based on design specifications or experience. The spacecraft expert rules discussed in this book are stored in a relational database with column names such as rule code, discriminant expression, fault phenomenon, and fault level. Because each row of structured rule data represents one entity, and each column of data has the same attributes, there is no need for extensive preprocessing of the structured rule knowledge; a knowledge graph can be directly generated based on the constructed ontology model.

[0032] 2) Semi-structured tabular data This refers to data that does not conform to the relational database model structure. Its characteristics include a mixture of data structure and content, with no strict distinction between the two, but containing relevant tags that can be used to separate semantic elements and hierarchically categorize records and fields. For example, spacecraft fault knowledge and risk management knowledge often contain semi-structured data in tabular form, such as FMEA analysis tables and risk analysis tables. This type of data cannot be directly mapped to a knowledge graph, but it is easier to extract structured knowledge than unstructured data. For this type of data, manual analysis is used to construct wrappers to extract rules, thereby transforming semi-structured data into structured data.

[0033] 3) Unstructured text data Unstructured data refers to data without a fixed structure. The large amount of unstructured data generated on spacecraft, primarily in text format, originates from design documents, spacecraft logs, fault records, and anomaly messages. This unstructured data from spacecraft is text-intensive, with fault knowledge implicitly embedded within it. Documents contain potential spacecraft fault phenomena, causes, and maintenance measures. Fault text is recorded according to certain standards, facilitating text preprocessing and knowledge acquisition. Text processing methods will be used to mine entities, relationships, and attributes from this unstructured data. Furthermore, unstructured data also includes images and video data generated from various inspections. This type of data requires tools such as graph neural networks for information extraction.

[0034] 4) Spacecraft test and telemetry data Besides the expert experience provided by design and reliability specialists, the vast amounts of data collected during ground testing and experimentation, as well as telemetry data generated during operation, are crucial supplements to mission assurance. Ground testing and experimentation data includes data from functional performance tests, reliability tests, and environmental adaptability tests. Spacecraft operation and maintenance data, in addition to written materials such as spacecraft logs and anomaly messages, also includes a large amount of operational telemetry data. Anomalies in telemetry data can reveal spacecraft fault information. As data changes over time, its anomalies are directly correlated with spacecraft fault modes. However, telemetry data suffers from limitations such as large volume, wide scope, unclear entities, ambiguous relationships, and incomplete attributes. Therefore, data mining techniques are needed to uncover its implicit anomaly rules to expand fault knowledge sources and enrich the fault knowledge graph.

[0035] (2) Knowledge base Based on the data resource database, the interrelationships and structured information of the data are extracted to form knowledge graphs or large language models that can be autonomously identified and efficiently retrieved by computers, such as system quality risk analysis graphs, system cost analysis graphs, product assurance large language models, and relevant defense patents and standards.

[0036] (3) Model library Digital models, including demand models, design models, manufacturing models, operation and maintenance models, management evaluation models, and data fusion models, are hierarchically categorized and stored for management, providing digital application services on demand. The demand and design models include not only functional and performance requirements and design models, but also reliability, maintainability, security, testability, supportability, and environmental adaptability (collectively referred to as the "six properties") requirements and design models. The operation and maintenance models include not only conventional operation and maintenance optimization models, but also fault prediction and health management (PHM) models. The management evaluation models include risk assessment models, cost analysis models, and management models.

[0037] Obtain the data to be processed; It should be noted that the data to be processed is either the system modeling language (SysML) model information or the knowledge graph information. By fusing SysML with the knowledge graph, a highly efficient and accurate Bayesian Network (BN) model is generated. This model not only strongly supports the reliability design of complex systems, providing a solid guarantee for engineering project design and manufacturing, reliability expertise, and other related fields, but also allows for reverse reasoning of the knowledge graph and the target Bayesian network through Bayesian analysis, quantifying the contribution of each potential fault node to the target fault, further strengthening scientific decision support. Examples 2 and 3 fully demonstrate the effectiveness of this fusion path: both use the system modeling language (SysML) model information and the knowledge graph information as inputs, respectively, following a basically consistent data processing flow, and both successfully generate Bayesian networks. Ultimately, both achieve support for the reliability design of complex systems, effectively improving the system's fault analysis, risk prediction, and decision support capabilities, thereby providing strong technical support for the aforementioned related fields.

[0038] It should be noted that the data to be processed consists of various types of engineering project design information from the early stages.

[0039] The data to be processed is modeled to obtain model building information. For example, the above process is as follows: 1. Engineering project model building Module Definition Graphs (BDDs) can express information and relationships in different types of system structures. They can also display model elements from other SysML graphs, including modules, actors, value types, constraint modules, interfaces, etc. Internal Module Graphs (IBDs) can represent information that cannot be expressed in BDDs: first, the connections between component attributes and referencing attributes; second, the types of events, energy, and data flowing between connections; and third, the services provided and requested through connections. BDDs and IBDs can complement each other. BDDs define modules and their attributes, while IBDs represent the proper configuration of modules (a specific series of connections between module attributes).

[0040] See Figure 2 Taking a satellite as an example, firstly, a SysML module definition diagram (BDD) is created based on the satellite's composition structure, mainly including the satellite's individual components and redundancy backup. Then, a "Proxy Port" is created for the individual components that exchange information with other individual components. Interface blocks are created according to the "Proxy Port" of each individual component to input the proxy port, and "Flow Property" elements are created for the interface blocks. The flow attributes can be modified according to the actual signal flow direction, providing a basis for representing signal transmission between individual components in the IBD.

[0041] See Figure 3 Then, using the modules and proxy ports created in BDD, combined with the signal transmission relationships required when the satellite provides services, an internal module diagram (IBD) of the satellite is created, which is the model building information.

[0042] The model building information is processed for structural recognition to obtain the model recognition result information; The model recognition results are post-processed to obtain Bayesian results.

[0043] In an optional embodiment, the model building information is subjected to structural recognition processing to obtain model recognition result information, including: The model building information is processed to extract model node information; the model node information includes several node information. For example, after obtaining the model file, the code is run to analyze the model, which consists of three main steps: model and data structure extraction, logic gate identification and storage, and sequential generation of logic gate relationships. Based on the input model file, the system automatically decomposes and rearranges the initial model, enabling it to be reorganized into a logical and data structure that conforms to the actual structure of a Bayesian network. The specific steps are explained below: Model and data structure extraction 1. Preparatory work First, define the data structure, which includes nodes, logical gates, a data dictionary, logical gate type enumerations, and basic methods for retrieving fixed fields from XML data.

[0044] 2. Obtain the basic structure 2.1 Basic Information Extraction The system reads a specified XML file, retrieves the model text with the 'Model' tag, and then obtains the specified graph. Based on the model file, a series of data extractions are performed, resulting in `rongyu_mom_son_dict` to record the parent-child relationships of redundant single machines (e.g., {B:[B1,B2]}) and `property_xmiId_name_dict` to record the correspondence between node IDs and Chinese names in the graph, serving as a dictionary for easy lookup. Simultaneously, all underlying node data (such as reliability data) is stored in the data dictionary.

[0045] 2.2 Constructing the list of direct successor nodes Based on the IBD image, the direction of the connection and associated nodes are identified by the start-end order of the connection, and recorded in `node_dict`. Additionally, if a node is followed by multiple nodes, the number and names of these multiple nodes are recorded in `son_num_dict`. These two methods are used respectively for finding the direct successor node, finding the first common node of branches, and establishing gates.

[0046] The model node information is processed to obtain the model recognition result information.

[0047] For example, the above process is as follows: Logical recognition and storage 1. Identify parallel relationships (AND gates) Traverse `son_num_dict`. If a node has more than one successor node, it is considered that the node will be followed by an AND gate. Create a `LogicGate` instance for this node, and the gate's children are all the successor nodes in `son_num_dict`. Simultaneously, use `get_first_together_node` to calculate the first common node. For all input nodes to be calculated, use the `get_chuan_between_node` method to calculate the node sequence strings corresponding to the specified start and end nodes, and then use the `make_calculate_first_node` method to calculate the first common node (end_node) of all node sequence strings, storing this node in the `son_num_dict`. Finally, add the AND gates to `and_gate_list`, which stores all AND gates.

[0048] 2. Identify serial relationships It is assumed that each child node of an AND gate will have subsequent concatenations, so concatenation (OR gate) is identified based on the AND gates. Traverse the and_gate_list, store the nodes from the AND gate's child node to the AND gate's end_node into the OR gate's child node, generate a new OR gate, and add it to the or_gate_list that stores all OR gates.

[0049] 3. Identify voting relationships Voting is equivalent to "selecting k from n", meaning that if k out of n nodes are normal, the system is considered normal. By identifying structural relationships and included nodes, we can obtain all nodes in the subsystem and the required number of nodes k. Combining these two pieces of information yields the basic voting structure, which is then stored in vote_gate_list.

[0050] 4. Identify reserve relationships Reserve relationships can be analogized to parallel relationships, the difference being that the type of reserve may lead to different final calculation methods. Following the parallel identification method described above, all reserve relationships are stored in `reserve_gate_list`.

[0051] 5. Identify the top-level logic gate Based on methods 2.1 and 2.2 above, the top-level logic gate is constructed using `and_gate_list` (all AND gates) and `node_dict` (direct successor node relationships). First, determine the logic gate type: If the input port 'start' is the starting point of the AND gate and the end node (end_node) of the AND gate is the output port 'end', then the IBD is considered to contain only one large parallel relationship, the top-level logic gate is identified as an AND gate, the corresponding relationship is recorded and returned; Otherwise, it is considered a non-simple parallel system. Then, starting from 'start', the successor node is matched one by one from the basic logical relationship set and node_dict in the loop, and recorded as the child node of the top-level logic gate.

[0052] Logical gate relations are generated in sequence Generate logic gate set structure in sequence First, define a stack as a container, initially placing the top-level logic gate inside it. Then, perform the following operations within the loop: First, record the number of nodes (gates) contained in the container as tmp_num. Then, perform tmp_num operations. In each operation, match each child node of the gate with parallel gates, serial gates, voting gates, reserve gates, and simple redundancies. If a match is found, the node is a gate; if no match is found, it is a basic event (single-machine failure). If it is a gate, put it into the container. After tmp_num operations, pop the container nodes (the identified gates) tmp_num times and add them to result_gate_list as the final gate record. Then, start matching newly added gates until the container contains no more gates.

[0053] In an optional embodiment, the model recognition result information is post-processed to obtain Bayesian result information, including: The model recognition results are optimized to obtain optimized model data. It should be noted that the above processing can be optimized using tools such as GPT and DeepSeek to obtain optimized model data. This mainly involves optimizing the model's recognition results by removing default values ​​and outliers.

[0054] The data information of the optimization model is processed to obtain Bayesian results.

[0055] For example, the above processing is post-processing such as visualization and quantitative analysis.

[0056] Table 1. Some supported visualization and storage formats The final visualization or display format generated by the method is not limited to a specific format or software; here, a Bayesian network is used as an example. Specifically, after generating the logic and data organization conforming to the Bayesian network structure in the previous step, the `result_gate_list` is traversed. If each gate belongs to a corresponding basic logical relationship, it can be found in its respective set, indicating that this gate belongs to a basic logical type. The basic logical gates are then stored in the gate order list in the traversal order. This ultimately forms a sequential structure of top-level logical gate - level 1 logical gate - level 2 logical gate... If a node in a gate is the source of other gates, it can be found below; otherwise, it represents a basic event. Writing the gate order list and data dictionary into JSON format facilitates interaction with other formats and supports API calls, making it naturally suitable for network transmission, local storage, and human collaboration.

[0057] After obtaining the JSON file, the file is converted into a .bayes file. The content of the .bayes file is the Bayes result information, which is generated through code programming in a form that can be clearly understood by software and personnel and meets the actual requirements of engineering, so as to be displayed and subsequently quantitatively calculated and analyzed, thereby meeting the reliability professional needs of related or other fields.

[0058] As can be seen, implementing the Bayesian analysis method described in the embodiments of the present invention can automatically construct an efficient and accurate Bayesian network model, which is beneficial for supporting the reliability design of complex systems. This provides strong support for engineering project design, manufacturing, reliability, and other related fields. By constructing Bayesian network information through knowledge graphs, the system's fault analysis, risk prediction, and decision support capabilities can be improved.

[0059] Example 2 Please see Figure 6 , Figure 6 This is a flowchart illustrating a Bayesian analysis method based on sysML and knowledge graphs disclosed in an embodiment of the present invention. Figure 6 The described Bayesian analysis method based on sysML and knowledge graphs is applied to a Bayesian analysis device based on sysML and knowledge graphs, such as a local server or cloud server for optimization and management of Bayesian analysis based on sysML and knowledge graphs. This invention is not limited to specific applications. Figure 6 As shown, this Bayesian analysis method based on sysML and knowledge graphs can include the following operations: S1, Obtain the modeling language model information of the system to be processed; It's important to note that SysML stands for System Modeling Language, a general-purpose, systems engineering-oriented modeling language designed to support the design, analysis, and verification of complex systems. It's an extension and modification of UML (Unified Modeling Language), focusing on the needs of systems engineering, especially complex multidisciplinary systems. A SysML model is a complete system model represented using SysML. It includes all aspects of the system: structure, behavior, constraints, and requirements. SysML models describe a system's design, behavior, interaction with the external environment, and the relationships between its components. SysML models facilitate collaboration, communication, and understanding among team members, ensuring effective coordination of system components during the design and implementation phases. The information in the SysML model to be processed contains detailed information about the system's structure, behavior, and requirements, comprehensively describing the system's functions, performance, interfaces, components, and their interrelationships. It is commonly used in the design, analysis, verification, and optimization phases of complex systems. Bayesian networks are used to analyze and describe system failure modes, causes, and their interrelationships. It is typically represented in the form of a tree structure, where the root node represents the final failure event of the system, and the leaf nodes represent the basic events that lead to these failures. Bayesian networks are widely used in system reliability analysis, risk assessment, and troubleshooting to identify potential hazards or failure modes and their likelihood of occurrence.

[0060] S2, process the modeling language model information of the system to be processed to obtain Bayesian network node connection relationship information; S3, process the connection relationship information of the Bayesian network nodes to obtain the target Bayesian network information; S4, process the target Bayesian network to obtain the target Bayesian network information model result information.

[0061] As can be seen, implementing the Bayesian analysis method based on sysML and knowledge graphs described in the embodiments of this invention, by automatically constructing efficient and accurate Bayesian networks, is beneficial for supporting the reliability design of complex systems, thereby providing strong support for engineering project design, manufacturing, and reliability-related work. Furthermore, through Bayesian analysis and reverse reasoning on the Bayesian network, it helps quantify the contribution of each potential fault node to the target fault, providing scientific decision support.

[0062] In an optional embodiment, the language model information of the system to be processed is processed to obtain Bayesian network node connection relationship information, including: S21, perform information extraction processing on the modeling language model information of the system to be processed to obtain model node information; the model node information includes several node information; the node information includes node ID information, predecessor node information, successor node information, redundancy information and voting information; It should be noted that the above extraction process can be performed using tools such as Capella and Rhapsody; however, the specific implementation of this invention is not limited thereto.

[0063] It should be noted that the predecessor node information in the node information refers to the node information directly connected to the current node information and located "upstream" of it, that is, the output of which node information becomes the input of the current node information. The successor node information refers to the node information directly connected to the current node information and located "downstream" of it, that is, the node information that takes the output of the current node information as its input. Both the predecessor node information and the successor node information are node information in the model node information.

[0064] For example, the above extraction process reads the modeling language model information of the system to be processed, obtains the model text information of the 'Model' tag, and then obtains the specified graph information. Based on the model file information, a series of data extractions are performed, resulting in `rongyu_mom_son_dict` to record the parent-child relationship of redundant single machines (e.g., {B:[B1, B2]}) and `property_xmiId_name_dict` to record the correspondence between node IDs and Chinese names in the image, serving as a dictionary for easy lookup. Simultaneously, all underlying node data (such as reliability) is stored in the data dictionary. Based on the IBD image, identification is performed, and the direction of the connection and associated nodes are identified by the start-end order of the connection, recorded in `node_dict`. If a node subsequently connects to multiple nodes, `son_num_dict` records the number and names of these multiple nodes. These two are respectively used for finding subsequent direct successor nodes, finding the first common node of branches, and establishing gates.

[0065] It's important to note that node information, in a system modeling language model, is a collection of detailed descriptions of the attributes and states of each node. This information identifies the node's role, its relationship with other nodes, and design features such as redundancy and fault tolerance. Node ID information is a unique identifier for each node, used to distinguish each node in the system. It can be a simple number, name, or other unique marker. Predecessor nodes refer to other node information that must occur before the current node's occurrence. They typically represent input conditions or prerequisites for a node's failure. Successor node information refers to other node information or events that will result from the occurrence of the current node's information. Successor nodes typically represent the outcome of an event or the next step in fault propagation. Voting information is often used to represent the "k-in-n" rule of a redundant system, meaning that the system will only fail if at least k nodes out of n nodes fail. Voting information describes the specific rules for system redundancy and fault tolerance design. Redundancy information describes whether nodes employ redundant design to ensure the system can continue to operate normally in the event of a node failure. Redundancy information typically indicates whether the system improves reliability through backup components or backup paths.

[0066] It should be noted that not every node's information includes predecessor node information, successor node information, redundancy information, and voting information. This depends on the node's type and function. The specific information of each node will vary according to its role and function in the system, specifically determined by the system modeling language model information; this embodiment of the invention does not impose limitations on this.

[0067] S22, process the model node information to obtain the Bayesian network node connection relationship information.

[0068] As can be seen, implementing the Bayesian analysis method based on sysML and knowledge graphs described in the embodiments of this invention, by automatically constructing efficient and accurate Bayesian networks, is beneficial for supporting the reliability design of complex systems, thereby providing strong support for engineering project design, manufacturing, and reliability-related work. Furthermore, through Bayesian analysis and reverse reasoning on the Bayesian network, it helps quantify the contribution of each potential fault node to the target fault, providing scientific decision support.

[0069] In another optional embodiment, the model node information is processed to obtain Bayesian network node connection relationship information, including: S221, Traverse the model node information to obtain logic gate node information; the logic gate node information is the node information that does not have a predecessor node information among all the node information in the model node information. It should be noted that when there is no predecessor node information in the node information, it means that the node information is the first node information in the system modeling language model information. For Bayesian networks, this node information corresponds to the top-level logic gate.

[0070] S222, add the logic gate node information to the logic gate queue information; S223, determine whether the number of successor node information in the logic gate node information is greater than 1, and obtain the first judgment result; If the first judgment result is negative, determine whether the number of successor node information in the logic gate node information is less than 1, and obtain the second judgment result; When the second judgment result is negative, the logic gate node information and the successor node information are marked and merged to obtain the OR gate information; the OR gate information is added to the logic gate list information, and S226 is executed. It should be noted that the first marking and merging mentioned above marks the logical relationship between the logical gate node information and the successor node information as an OR gate, that is, the logical relationship between the logical gate node information and the successor node information is an OR gate relationship, and merges the logical gate node information, the successor node information and the OR gate relationship information between them to obtain the OR gate information.

[0071] When the second judgment result is yes, the logic gate node information is determined to be basic event information; the basic event information is added to the logic gate list information, and S227 is executed; If the result of the first judgment is yes, execute S224; S224, determine whether there is redundant information in the information of any successor node, and obtain the third judgment result; When the third judgment result is yes, the logic gate node information and all successor node information are marked and merged in the second way to obtain the reserve gate information; the reserve gate information is added to the logic gate list information and S226 is executed. It should be noted that the above-mentioned second marking and merging involves marking the logical relationship between the logical gate node information and all subsequent node information as a reserve gate, that is, the logical relationship between the logical gate node information and all subsequent node information is a reserve gate relationship, and merging the logical gate node information, the subsequent node information and the reserve gate relationship information between them to obtain the reserve gate information.

[0072] If the result of the third judgment is negative, execute S225; S225, determine whether voting information exists in the information of any successor node, and obtain the fourth judgment result; When the fourth judgment result is yes, the logic gate node information and all successor node information are marked and merged in the third process to obtain the voting gate information; the voting gate information is added to the logic gate list information. It should be noted that the third marking and merging mentioned above marks the logical relationship between the logical gate node information and all subsequent node information as a voting gate, that is, the logical relationship between the logical gate node information and all subsequent node information is a voting gate relationship, and merges the logical gate node information, the subsequent node information and the voting gate relationship information between them to obtain the voting gate information.

[0073] When the fourth judgment result is negative, the logic gate node information and all successor node information are marked and merged to obtain parallel information; the parallel information is added to the logic gate list information. It should be noted that the fourth marking and merging mentioned above marks the logical relationship between the logical gate node information and all subsequent node information as parallel, that is, the logical relationship between the logical gate node information and all subsequent node information is parallel, and merges the logical gate node information, subsequent node information and the parallel relationship information between them to obtain parallel information.

[0074] S226, add all successor node information to the logic gate queue information; S227, Remove the logic gate node information from the logic gate queue information; S228, determine whether the logic gate queue information has successor node information, and obtain the fifth judgment result; When the result of the fifth judgment is yes, determine any successor node information in the logic gate queue information as logic gate node information, and execute S223; If the result of the fifth judgment is negative, execute S229; S229, process the logic gate list information to obtain the Bayesian network node connection relationship information.

[0075] It should be noted that the above processing can be achieved through Reliability Workbench, OpenFTA, etc., and the specific implementation of this invention is not limited thereto.

[0076] For example, the steps of the above process are as follows: Step 1: Establish node relationships For each logic gate in the CPT table of the Bayesian network nodes, traverse the table to extract the input node (i.e., its predecessor node) and output node (i.e., its successor node), and construct the connections between nodes based on these relationships. The connections between each node can be represented by information from its predecessor and successor nodes.

[0077] Step 2: Determine the type of CPT table between nodes Determine the logical relationships between each node based on the relationships between them (such as AND gates, OR gates, voting gates, etc.). For example: AND gate: All input events must occur simultaneously for the parent node event to occur.

[0078] OR gate: Whenever an input event occurs, the parent node event will also occur.

[0079] Voting gate: Among multiple input events, a specific "k out of n" condition must be met for the parent node event to occur.

[0080] Step 3: Construct Bayesian network node connection information Based on the node connections and logical relationships obtained in steps 1 and 2, construct the Bayesian network node connection information. The top-level nodes of the Bayesian network typically represent the top events of the entire system, while the bottom-level (edge) nodes represent basic events (such as failures of system components or subsystems). The CPT table connects these events, forming different event propagation paths.

[0081] The method of this invention can automatically process the logical relationships in complex system modeling language model information without manual intervention. At the same time, it can decompose a large problem of constructing a Bayesian network into smaller subproblems for solving, which can effectively reduce the amount of computation, avoid redundant calculations, and improve computational efficiency.

[0082] As can be seen, implementing the Bayesian analysis method based on sysML and knowledge graphs described in the embodiments of this invention, by automatically constructing efficient and accurate Bayesian networks, is beneficial for supporting the reliability design of complex systems, thereby providing strong support for engineering project design, manufacturing, and reliability-related work. Furthermore, through Bayesian analysis and reverse reasoning on the Bayesian network, it helps quantify the contribution of each potential fault node to the target fault, providing scientific decision support.

[0083] In another optional embodiment, the Bayesian network node connection relationship information is processed to obtain the target Bayesian network information, including: S31, Process the Bayesian network node connection relationship information to obtain optimized Bayesian network node connection relationship information; S32, process the connection relationship information of the optimized Bayesian network nodes to obtain the target Bayesian network information.

[0084] As can be seen, implementing the Bayesian analysis method based on sysML and knowledge graphs described in the embodiments of this invention, by automatically constructing efficient and accurate Bayesian networks, is beneficial for supporting the reliability design of complex systems, thereby providing strong support for engineering project design, manufacturing, and reliability-related work. Furthermore, through Bayesian analysis and reverse reasoning on the Bayesian network, it helps quantify the contribution of each potential fault node to the target fault, providing scientific decision support.

[0085] In an optional embodiment, the Bayesian network node connection relationship information is processed to obtain optimized Bayesian network node connection relationship information, including: S311, integrate all logic gate information in the Bayesian network node connection relationship information to obtain the first Bayesian network information; It should be noted that the above integration process merges logically redundant gates. Specifically, tools such as FaultTree+ and OpenFTA can be used for this process. This embodiment of the invention does not limit the scope of the invention. For example, if two AND gates or two OR gates share the same set of child nodes, the two AND gates or two OR gates can be merged into one AND gate or one OR gate.

[0086] It's important to note that redundant logic gates typically increase the size and complexity of Bayesian networks. Merging these redundant gates can make Bayesian networks simpler, thus easier to understand and analyze. This allows for a clearer representation of the system's structure and potential failure paths, while also reducing the number of nodes and computational load, thereby accelerating Bayesian network analysis.

[0087] S312, perform redundant path elimination processing on the first Bayesian network information to obtain optimized Bayesian network node connection relationship information.

[0088] It should be noted that the above-mentioned redundant path elimination process is used if a node in the logic gate information is simultaneously used as the input of multiple gates. Unnecessary redundancy can be eliminated through Boolean algebra, thereby eliminating unnecessary redundant calculations. Specifically, the embodiments of the present invention do not limit this.

[0089] It should be noted that the above-mentioned redundant path elimination process can improve computational efficiency, reduce system complexity, optimize resource utilization, and enhance analysis accuracy.

[0090] An example of redundant path elimination processing steps is as follows: Shared input merging: If both AND gate 1 and AND gate 2 have N as input, and OR gate 1 also has N as input, the outputs of these gates can be combined to reduce unnecessary redundant calculations.

[0091] Boolean algebra simplification: Some commonly used theorems in Boolean algebra can simplify logical expressions. For example: If N is the input to AND gate 1 and AND gate 2, and these two AND gates share the same input, then these two AND gates can be combined into a simpler expression.

[0092] For example, suppose: O1 = N∧A1 and O3 = N∧C1; Therefore, they can be merged into: OAND=N∧(A1∧C1); In this way, the outputs of two AND gates can be combined into one.

[0093] Besides AND and OR, the AND gate is used in scenarios where multiple bottom events must occur simultaneously to cause a top event, while the OR gate is used in scenarios where any bottom event can independently cause a top event.

[0094] Also includes the following: The NOT gate is used in scenarios where a parent event occurs only if a specific event does not occur. The NAND gate is used in scenarios where multiple underlying events occurring simultaneously do not lead to a parent event, but at least one event must not occur for a parent event to occur. The NOR gate is used in scenarios where any underlying event occurring does not lead to a parent event, but all underlying events must not occur for a parent event to occur. The XOR gate is used in scenarios where exactly one underlying event occurs for a parent event to occur. The XNOR gate is used in scenarios where all underlying events occur simultaneously or simultaneously do not occur for a parent event to occur. The AND-OR-NOT gate is used in scenarios where, in at least one group of events, simultaneous occurrence of all events does not lead to a parent event, and none of the groups meet the condition that simultaneous occurrence of all events leads to a parent event. The Majority Gate is used in scenarios where more than half of the underlying events must occur for a parent event to occur. The Inhibit Gate is used in scenarios where, when a conditional event occurs, it prevents an underlying event from leading to a parent event. The Implication Gate... A gate is used in scenarios where a lower-level event will trigger a higher-level event only if a lower-level event occurs when a lower-level event does not occur.

[0095] As can be seen, implementing the Bayesian analysis method based on sysML and knowledge graphs described in the embodiments of this invention, by automatically constructing efficient and accurate Bayesian networks, is beneficial for supporting the reliability design of complex systems, thereby providing strong support for engineering project design, manufacturing, and reliability-related work. Furthermore, through Bayesian analysis and reverse reasoning on the Bayesian network, it helps quantify the contribution of each potential fault node to the target fault, providing scientific decision support.

[0096] In an optional embodiment, the optimized Bayesian network node connection relationship information is processed to obtain the target Bayesian network information, including: S321, Obtain the basic event reliability probability information of the optimized Bayesian network node connection relationship information; the basic event reliability probability information includes several basic event reliability probability values; the basic event reliability probability values ​​correspond one-to-one with the basic event information in the optimized Bayesian network node connection relationship information; It should be noted that the basic event reliability probability value, in Bayesian network analysis, is typically used to represent the probability level of reliability for each basic event. This probability value reflects the probability that the event will not fail within a given time period.

[0097] S322, Calculate and process the node connection relationship information and basic event reliability probability information of the optimized Bayesian network to obtain intermediate event reliability probability information; the intermediate event reliability probability information includes several intermediate event reliability probability values; the intermediate event reliability probability values ​​correspond one-to-one with the intermediate event information in the node connection relationship information of the optimized Bayesian network. It should be noted that the intermediate event reliability probability value, in Bayesian network analysis, is typically used to represent the probability level of reliability for each intermediate event. This probability value reflects the probability that the event will not fail within a given time period.

[0098] S323, Calculate and process the node connection relationship information and intermediate event reliability probability information of the optimized Bayesian network to obtain intermediate event importance information; the intermediate event importance information includes several intermediate event importance values; the intermediate event importance values ​​correspond one-to-one with the intermediate event information in the node connection relationship information of the optimized Bayesian network; It's important to note that intermediate event importance information in a Bayesian network refers to the importance measures of events that, while not directly causing system failure, still impact the overall behavior and reliability of the system. Intermediate event information is typically a combination of different basic event information (such as components or subsystems), connected to other events via logic gates. In a Bayesian network, intermediate event information serves as a transitional element between "basic event information" (the lowest level, indivisible fault / normal units) and "system target events" (the top level, such as system faults / normalities). Essentially, it consists of "event nodes formed by combining several basic events / sub-intermediate events through logical relationships (AND / OR, etc.)." It describes the "intermediate links" in the system's fault propagation process and acts as a crucial bridge connecting the state of lower-level components with the overall system state.

[0099] Intermediate event importance information reflects the degree of importance of intermediate event information to the overall system failure in Bayesian network analysis, and is an indicator for measuring the role of intermediate event information in system failure. It usually takes into account the probability of event occurrence, the propagation impact of failure paths, and the overall reliability of the system.

[0100] S324 calculates and processes the node connection relationship information, intermediate event reliability probability information, and intermediate event importance information of the optimized Bayesian network to obtain the target Bayesian network information.

[0101] As can be seen, implementing the Bayesian analysis method based on sysML and knowledge graphs described in the embodiments of this invention, by automatically constructing efficient and accurate Bayesian networks, is beneficial for supporting the reliability design of complex systems, thereby providing strong support for engineering project design, manufacturing, and reliability-related work. Furthermore, through Bayesian analysis and reverse reasoning on the Bayesian network, it helps quantify the contribution of each potential fault node to the target fault, providing scientific decision support.

[0102] In an optional embodiment, the node connection relationship information and intermediate event reliability probability information of the optimized Bayesian network are calculated to obtain intermediate event importance information, including: S3231, Obtain Bayesian network fault path information from the optimized Bayesian network node connection relationship information; Bayesian network fault path information includes several fault path information; It should be noted that fault path information is a logical chain of relationships from the top of the system (representing events that cause system failures or malfunctions) to the most basic events at the bottom level. It represents the possible paths in which a system failure can occur. Each fault path is composed of a series of event information and their logical relationships (such as AND, OR, and other logic gates) connected together.

[0103] S3232 utilizes a Bayesian network importance calculation model to calculate and process Bayesian network fault path information, optimized Bayesian network node connection relationship information, and intermediate event reliability probability information to obtain intermediate event importance information. The Bayesian network importance calculation model is as follows: ; ; ; In the formula, This provides information on the importance of intermediate events. This represents the importance value of the i-th intermediate event in the intermediate event importance information. Let be the reliability probability value of the i-th intermediate event in the intermediate event reliability probability information. This represents the number of fault path messages in the Bayesian network fault path information that include information about the i-th intermediate event. ZD represents the number of intermediate event information entries for the j-th fault path among all fault path information entries containing information about the i-th intermediate event, and ZD represents the number of intermediate event information entries in the node connection relationship information of the optimized Bayesian network. and These represent the third, fourth, and fifth weight parameters, respectively.

[0104] It should be noted that the third, fourth, and fifth weight parameters can be set by the user or obtained from historical data, and this embodiment of the invention does not impose any limitations.

[0105] It should be noted that the third, fourth, and fifth weighting parameters incorporate the influence of multiple factors, allowing them to flexibly adapt to changing circumstances. By adjusting these coefficients, the influence of different aspects can be emphasized or weakened, thus more accurately describing the importance information of intermediate events. The Bayesian network importance calculation model can more comprehensively evaluate the contribution of intermediate events to system reliability, thereby providing a basis for optimizing Bayesian networks; through the several terms following the fifth weighting parameter... The introduction of this concept enhances the nonlinear expression of the formula, making the importance of intermediate events more sensitive in certain situations. It is particularly suitable for handling the effects of probability and information uncertainty, thereby enabling more accurate acquisition of information on the importance of intermediate events.

[0106] As can be seen, implementing the Bayesian analysis method based on sysML and knowledge graphs described in the embodiments of this invention, by automatically constructing efficient and accurate Bayesian networks, is beneficial for supporting the reliability design of complex systems, thereby providing strong support for engineering project design, manufacturing, and reliability-related work. Furthermore, through Bayesian analysis and reverse reasoning on the Bayesian network, it helps quantify the contribution of each potential fault node to the target fault, providing scientific decision support.

[0107] In an optional embodiment, the node connection relationship information, intermediate event reliability probability information, and intermediate event importance information of the optimized Bayesian network are calculated and processed to obtain the target Bayesian network information, including: S3241 uses a Bayesian network redundancy importance calculation model to calculate and process the reliability probability information and importance information of intermediate events to obtain redundancy importance information; the redundancy importance information includes several redundancy importance values; the redundancy importance values ​​correspond one-to-one with the intermediate event information in the optimized Bayesian network node connection relationship information; The Bayesian network redundancy importance calculation model is as follows: ; ; In the formula, RY represents the redundancy importance information. For the i-th redundancy importance value in the redundancy importance information, This represents the importance value of the i-th intermediate event in the intermediate event importance information. Let be the reliability probability value of the i-th intermediate event in the intermediate event reliability probability information. and These are the first and second weighting parameters, respectively, and N is the number of intermediate event importance values ​​in the intermediate event importance information. It should be noted that the first weight parameter and the second weight parameter can be set by the user or obtained from historical data, and this embodiment of the invention does not limit them.

[0108] It should be noted that the redundancy importance value of each intermediate event information is calculated. This allows us to identify which intermediate event information requires redundant design to improve the overall reliability of the system. For example, intermediate events with high redundancy importance values ​​are usually those with high importance but low reliability. These intermediate events play a crucial role in system failures, thus requiring redundancy to reduce the probability of failure. Furthermore, by appropriately selecting weighting parameters, the redundancy design can be flexibly optimized according to system requirements, thereby improving system reliability.

[0109] It should be noted that the above calculation and processing can also be implemented using weighted adaptive algorithms, analytic hierarchy process, etc., and the specific implementation of this invention is not limited thereto.

[0110] S3242, Sort all redundancy importance values ​​in the redundancy importance information in descending order to obtain the redundancy importance information to be processed; S3243, perform an extraction operation on the redundancy importance information to be processed to obtain preprocessed redundancy importance information; the preprocessed redundancy importance information includes the first M redundancy importance values ​​in the redundancy importance information to be processed; M is a positive integer less than N; S3244 performs redundancy processing on the intermediate event information corresponding to all redundant importance values ​​in the preprocessed redundant importance information to obtain the target Bayesian network information.

[0111] It should be noted that the above-mentioned redundancy processing adds redundant paths to intermediate event information, so that when intermediate event information fails, its corresponding redundant path can take over the work and continue to work, thereby reducing the probability of system failure. Specifically, the embodiments of the present invention do not limit this.

[0112] For example, an intermediate event A is connected to other events through an OR gate, causing the system to fail. However, A has a high redundancy importance value, indicating that A is crucial to the reliability of the system. In order to enhance reliability, redundancy processing is performed: a redundant event A1 is added to A in the Bayesian network and connected to A and other events through an AND gate to form a redundant path. In this way, the Bayesian network after redundancy processing improves the overall reliability of the system by adding redundant paths.

[0113] Redundant paths refer to the addition of extra paths or backup components in system design to ensure that when one path fails, other paths can continue to function normally, thereby enhancing the reliability of the system.

[0114] As can be seen, implementing the Bayesian analysis method based on sysML and knowledge graphs described in the embodiments of this invention, by automatically constructing efficient and accurate Bayesian networks, is beneficial for supporting the reliability design of complex systems, thereby providing strong support for engineering project design, manufacturing, and reliability-related work. Furthermore, through Bayesian analysis and reverse reasoning on the Bayesian network, it helps quantify the contribution of each potential fault node to the target fault, providing scientific decision support.

[0115] In an optional embodiment, the target Bayesian network is processed to obtain the target Bayesian network information model result information, including: S41, Obtain knowledge graph information; S42, Transform the target Bayesian network to obtain Bayesian data information; It should be noted that the above conversion process can be performed using GenIe / SMILE, Hugin, OpenBUGS, PyMC3, or Netica. Specifically, the embodiments of the present invention do not limit the specific conversion process.

[0116] S43, analyze and process the Bayesian data and knowledge graph information to obtain the target Bayesian network information model result information.

[0117] It should be noted that the above analysis and processing of Bayesian data and knowledge graph information to obtain the target Bayesian network information model result information is obtained by using Bayesian software (such as GenIe Modeler, Hugin Expert, Netica, BayesiaLab). Specifically, the embodiments of the present invention are not limited.

[0118] For example, the analysis and processing of Bayesian data and knowledge graph information to obtain the target Bayesian network information model result information is achieved by performing the following steps using Bayesian software: Step 1: Convert Bayesian data into a Bayesian network; Node definition: Each node structure or common relationship (basic event information, intermediate event information, logic gates, etc.) in a Bayesian network is represented as a node.

[0119] Edge connection: Based on the logical relationships (such as AND, OR, etc.) in the Bayesian network, connect nodes in the Bayesian network to represent the causal relationship between events.

[0120] CPT settings: Set up a conditional probability table (CPT) for each node, which defines the probability relationship between parent and child nodes.

[0121] Step 2: Integrate the knowledge graph information into the Bayesian network.

[0122] Extract structural information, causal relationships, and influencing factors related to the system from knowledge graph information.

[0123] The entities and relationships extracted from the knowledge graph are mapped to the corresponding nodes and edges in the Bayesian network.

[0124] Step 3: Perform reasoning calculations to determine the probability of other nodes failing and the overall reliability of the system given certain known events (such as a node failing).

[0125] Sampling and computation: Bayesian inference methods (such as sampling method, maximum a posteriori method, etc.) are used to perform inference computation on the Bayesian network to obtain the overall failure probability of the system.

[0126] Update node probability: Given the failure of certain nodes (e.g., the failure of a component), the failure probability of other nodes and the failure propagation path are inferred backward through the reasoning algorithm.

[0127] Analyze edge node probabilities: By adjusting the probability values ​​of edge nodes, observe changes in system reliability to identify weak links and critical nodes in the system.

[0128] Step 4: Use backward reasoning to find the path of the failure. For example, given a known faulty node, backward reasoning can reveal the potential root cause of the failure and identify the possible factors that led to it.

[0129] Set known nodes: Set the failure probability of a certain node to 1 (representing the occurrence of failure), and calculate the failure probability of other nodes.

[0130] Backward reasoning: Using the backward reasoning function in Bayesian networks, identify other nodes that caused the failure and calculate their failure probabilities.

[0131] Step 5: Based on the results of Bayesian inference, generate updated Bayesian network information, namely, the target Bayesian network model result information. The target Bayesian network model result information includes the latest failure probability of each node in the system, as well as the failure propagation path between nodes.

[0132] Update Bayesian Network: Based on the Bayesian analysis results, update the failure probability of each node in the Bayesian network, and transform the updated Bayesian network into the target Bayesian network model result information.

[0133] It should be noted that the results of the target Bayesian network model, through the analysis and optimization of the Bayesian network, can identify key nodes and potential risks in the system, and guide decision-makers to improve the system, optimize the design, and prevent failures. For example, it can simulate the reverse reasoning of the cause after a failure occurs in the project, find possible failure propagation links, and check weak links and weak processes in the engineering design.

[0134] It should be noted that using Bayesian methods for analysis can significantly improve a system's fault diagnosis, reliability assessment, decision support, and fault prevention capabilities. The reasoning power, dynamic updating capabilities, and ability to handle complex relationships within a system make Bayesian methods a powerful tool for analyzing faults in complex systems. Combining Bayesian network construction and analysis can effectively identify weaknesses in a system, improve system reliability, and provide support for optimization design, risk assessment, and predictive maintenance.

[0135] It should be noted that by advancing through three layers—Bayesian network construction, Bayesian network optimization, and Bayesian network analysis—the reliability and stability of the system can be comprehensively improved. Bayesian network construction can accurately identify and assess failure risks; Bayesian network optimization can identify weak points and implement redundant design and system optimization; Bayesian analysis, through real-time analysis and backward reasoning, helps improve fault prevention and emergency response capabilities. Through this three-layered approach, system reliability can be significantly improved, failure rate reduced, design optimized, and operational efficiency increased.

[0136] As can be seen, implementing the Bayesian analysis method based on sysML and knowledge graphs described in the embodiments of this invention, by automatically constructing efficient and accurate Bayesian networks, is beneficial for supporting the reliability design of complex systems, thereby providing strong support for engineering project design, manufacturing, and reliability-related work. Furthermore, through Bayesian analysis and reverse reasoning on the Bayesian network, it helps quantify the contribution of each potential fault node to the target fault, providing scientific decision support.

[0137] As demonstrated by the embodiments of this invention, the Bayesian analysis method based on SysML and knowledge graphs does not rely on large amounts of cumbersome manual information filtering and extraction. It requires only a small number of non-professionals to efficiently, accurately, systematically, and clearly build a system model by following clearly defined steps. While supporting the digital work of designers, it also automates the conversion and generation of different forms of Bayesian networks using coding tools. This method significantly reduces manpower and time investment, facilitates batch processing, and the built logical model can serve as an intermediary interface for internal and external interactions, enabling collaboration between staff and other professionals.

[0138] This method effectively saves time for professionals, comprehensively improves the reliability and efficiency of professional work, reduces human error in analysis, and saves costs in many aspects.

[0139] It should be noted that after obtaining the target Bayesian network information model result, this method can also perform visualization post-processing on the target Bayesian network, as follows: The specific method involves generating a target Bayesian network conforming to the Bayesian network structure in the previous step. Then, the target Bayesian network is traversed. If each gate belongs to a corresponding basic logical relationship, it can be found in its respective set, indicating that the gate belongs to a basic logical type. The basic logical gates are then stored in the gate order list in the traversal order. This ultimately forms a sequential structure of top-level logical gates, level 1 logical gates, level 2 logical gates, and so on. If a node in a gate is the source of other gates, it can be found in the target Bayesian network; otherwise, it is a basic event. The gate order list and data dictionary are written in JSON format, facilitating interaction with other formats and supporting API calls, making it naturally suitable for network transmission, local storage, and personnel collaboration.

[0140] After obtaining the JSON file, the file is converted into a .bayes file. This involves using Python code to generate a Bayesian network that can be clearly understood by software and personnel and meets the actual engineering requirements. This allows for visualization and subsequent quantitative calculations and analysis, thereby satisfying the reliability requirements of related or other fields.

[0141] This application enables efficient and accurate professional operation of converting to Bayesian networks while supporting relevant digital work in the field of support systems engineering.

[0142] Compared to the drawbacks of traditional fully manual analysis, such as being time-consuming and labor-intensive, having potentially inaccurate results, and being unable to reuse work content, the method in this application is efficient, rigorous, and has a clear sequence, making it easy for staff to perform batch workflows. Furthermore, the constructed target Bayesian network can serve as an intermediate interface for input and output interaction, facilitating collaboration between staff and other professionals.

[0143] Especially in the field of complex engineering design, this method can be applied to obtain different forms of target Bayesian networks through post-processing operations such as model building, software analysis, visualization and quantitative analysis, which can strongly support the work in engineering project design, manufacturing and reliability.

[0144] Example 3 This invention discloses a knowledge graph-based Bayesian analysis method. This method is applied in a knowledge graph-based Bayesian analysis device, such as a local server or cloud server for optimizing and managing knowledge graph-based Bayesian analysis; however, this invention is not limited to this specific device. The knowledge graph-based Bayesian analysis method may include the following operations: Obtain the knowledge graph information to be processed; The knowledge graph information to be processed is processed to obtain the knowledge graph information; The knowledge graph information is optimized to obtain the target knowledge graph information; It should be noted that the above optimization process involves opening the generated Bayesian file (knowledge graph information) using Bayesian software. The model can be fine-tuned by setting the number of samples, adjusting the probability values ​​of edge nodes, and modifying the node CPT table. The calculated data is the target knowledge graph information and is automatically updated to the MySQL database.

[0145] The target knowledge graph information is processed to obtain the target knowledge graph result information.

[0146] It should be noted that the above process uses libraries such as pymysql to connect to MySQL, filters the nodes of a specified model and state in a specified table and calculates their probability values ​​using SQL statements, updates the knowledge graph using Cypher statements with information from libraries such as py2neo, and then uses Python to convert the updated data into a Bayesian network, thus obtaining the target knowledge graph result information.

[0147] As an optional implementation method, the knowledge graph information to be processed is processed to obtain knowledge graph information, including: Information extraction processing is performed on the knowledge graph information to be processed to obtain model node information; the model node information includes several node information. It should be noted that the above extraction process uses libraries such as py2neo to link the code with the knowledge graph (such as neo4j), querying all relevant relationships and nodes based on a specified node. If the "relationship required for analysis" (such as "cause_AND", "cause_OR", etc.) does not exist, the system is considered empty and the process exits; if it exists, the node is pushed into a queue, and iteration begins in the queue. In each iteration, the first element in the queue is retrieved, and the existence of the "relationship required for analysis" is checked. If it does not exist, the node is considered to have no lower-level structure, and iteration continues; if it exists, a branching approach is used to determine which type of "relationship required for analysis" it is, the corresponding gate structure is stored in a JSON list, the direct nodes of the relationship are pushed into the queue sequentially, and the probability information of all edge nodes (the lowest-level nodes, with occurrence probability values) in the direct nodes is added to the probability dictionary. Finally, the next iteration begins.

[0148] The model node information is processed to obtain knowledge graph information.

[0149] It should be noted that the above processing involves extracting all gate structures and probability dictionary data from the JSON list using a Bayes file generation algorithm, automatically organizing the Bayesian node hierarchy structure according to the order of the JSON list and calculating the corresponding display position information, and automatically generating the corresponding Bayesian node CPT table according to the gate type and filling in the specific probability value information to complete the generation of the Bayesian file.

[0150] It is evident that implementing the knowledge graph-based Bayesian analysis method described in the embodiments of the present invention, and constructing Bayesian network information through knowledge graphs, can improve the system's fault analysis, risk prediction, and decision support capabilities.

[0151] Example 4 Please see Figure 7 , Figure 7 This is a schematic diagram of a Bayesian analysis device based on sysML and knowledge graphs disclosed in an embodiment of the present invention. Figure 7 The described sysML and knowledge graph-based Bayesian analysis device is applied to sysML and knowledge graph-based Bayesian analysis optimization systems, such as local servers or cloud servers used for sysML and knowledge graph-based Bayesian analysis. This invention is not limited to these applications. Figure 7 As shown, this Bayesian analysis apparatus based on sysML and knowledge graphs includes: Module 201 is used to acquire the modeling language model information of the system to be processed; The first calculation module 202 is used to process the information of the modeling language model of the system to be processed to obtain the connection relationship information of the Bayesian network nodes; The second calculation module 203 is used to process the connection relationship information of the Bayesian network nodes to obtain the target Bayesian network information. The third calculation module 204 is used to process the target Bayesian network to obtain the target Bayesian network information model result information.

[0152] As can be seen, implementing the Bayesian analysis device based on sysML and knowledge graphs described in the embodiments of the present invention can help support the reliability design of complex systems by automatically constructing efficient and accurate Bayesian networks, thereby providing strong support for engineering project design, manufacturing, reliability and other related fields.

[0153] Example 5 Please see Figure 8 , Figure 8 This is a schematic diagram of another Bayesian analysis device based on sysML and knowledge graphs disclosed in an embodiment of the present invention. Figure 8 The described sysML and knowledge graph-based Bayesian analysis device is applied to sysML and knowledge graph-based Bayesian analysis optimization systems, such as local servers or cloud servers used for sysML and knowledge graph-based Bayesian analysis. This invention is not limited to these applications. Figure 8 As shown, this Bayesian analysis apparatus based on sysML and knowledge graphs includes: Processor 301; A memory 302 containing executable program code is coupled to the processor 301; The processor 301 calls the executable program code stored in the memory 302 to execute some or all of the steps of the Bayesian analysis method based on sysML and knowledge graph in Embodiment 2.

[0154] As can be seen, implementing the Bayesian analysis device based on sysML and knowledge graphs described in the embodiments of the present invention can help support the reliability design of complex systems by automatically constructing efficient and accurate Bayesian networks, thereby providing strong support for engineering project design, manufacturing, reliability and other related fields.

[0155] Example 6 This invention discloses a computer-readable storage medium storing computer instructions. When the computer instructions are invoked, they are used to execute some or all of the steps of the Bayesian analysis method based on sysML and knowledge graphs in Embodiment 2.

[0156] Example 7 This invention discloses a computer program product, which includes a non-transitory computer-readable storage medium storing a computer program, and the computer program is operable to cause a computer to perform some or all of the steps in the Bayesian analysis method based on sysML and knowledge graphs described in Embodiment 2.

[0157] The system embodiments described above are merely illustrative. The modules described as separate components may or may not be physically separate, and the components shown as modules may or may not be physical modules; that is, they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0158] Through the detailed description of the above embodiments, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, including read-only memory (ROM), random access memory (RAM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), one-time programmable read-only memory (OTPROM), electrically-Erasable Programmable Read-Only Memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, disk storage, magnetic tape storage, or any other computer-readable medium that can be used to carry or store data.

[0159] Finally, it should be noted that the Bayesian analysis method and apparatus based on sysML and knowledge graph disclosed in the embodiments of the present invention are merely preferred embodiments of the present invention and are only used to illustrate the technical solutions of the present invention, not to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A Bayesian analysis method characterized in that, The method includes: Obtain the data to be processed; The data to be processed is subjected to model building processing to obtain model building information; The model building information is subjected to structural recognition processing to obtain model recognition result information; The model recognition results are post-processed to obtain Bayesian results.

2. The Bayesian analysis method of claim 1, wherein, The structural recognition processing of the model building information to obtain model recognition result information includes: The model building information is processed to extract model node information; the model node information includes several node information. The model node information is processed to obtain model recognition result information.

3. The Bayesian analysis method of claim 1, wherein, The post-processing of the model recognition result information to obtain Bayesian result information includes: The model recognition result information is optimized to obtain optimized model data information; The optimized model data is processed to obtain Bayesian results.

4. A Bayesian analysis method based on sysML and knowledge graph, characterized in that, The method includes: S1, Obtain the modeling language model information of the system to be processed; S2, process the modeling language model information of the system to be processed to obtain Bayesian network node connection relationship information; S3, process the connection relationship information of the Bayesian network nodes to obtain the target Bayesian network information; S4, process the target Bayesian network to obtain the target Bayesian network information model result information.

5. The sysML and knowledge graph based Bayesian analysis method of claim 4, wherein, The process of processing the modeling language model information of the system to be processed to obtain Bayesian network node connection relationship information includes: S21, perform information extraction processing on the modeling language model information of the system to be processed to obtain model node information; the model node information includes several node information. S22, The model node information is processed to obtain Bayesian network node connection relationship information.

6. The sysML and knowledge graph based Bayesian analysis method of claim 5, wherein, The process of processing the model node information to obtain Bayesian network node connection relationship information includes: S221, The model node information is traversed to obtain logic gate node information; the logic gate node information is the node information in the model node information that does not have a predecessor node information. S222, add the logic gate node information to the logic gate queue information; S223, determine whether the number of successor node information in the logic gate node information is greater than 1, and obtain the first judgment result; When the first judgment result is negative, determine whether the number of successor node information in the logic gate node information is less than 1, and obtain the second judgment result; When the second judgment result is negative, the logic gate node information and the successor node information are first marked and merged to obtain OR gate information; the OR gate information is added to the logic gate list information, and S226 is executed. When the second judgment result is yes, the logic gate node information is determined to be basic event information; the basic event information is added to the logic gate list information, and S227 is executed; If the first judgment result is yes, execute S224; S224, determine whether there is redundant information in any of the subsequent node information, and obtain a third judgment result; When the third judgment result is yes, the logic gate node information and all the successor node information are subjected to a second marking and merging process to obtain the reserve gate information; the reserve gate information is added to the logic gate list information, and S226 is executed; If the third judgment result is negative, execute S225; S225, determine whether voting information exists in any of the subsequent node information, and obtain the fourth determination result; When the fourth judgment result is yes, the logic gate node information and all the successor node information are subjected to a third marking and merging process to obtain the voting gate information; the voting gate information is added to the logic gate list information; When the fourth judgment result is negative, the logic gate node information and all the successor node information are subjected to a fourth marking and merging process to obtain parallel information; the parallel information is added to the logic gate list information. S226, add all the successor node information to the logic gate queue information; S227, Remove the logic gate node information from the logic gate queue information; S228, determine whether the logic gate queue information contains the successor node information, and obtain the fifth determination result; When the fifth judgment result is yes, determine any of the successor node information in the logic gate queue information as the logic gate node information, and execute S223; If the result of the fifth judgment is negative, execute S229; S229, Process the logic gate list information to obtain Bayesian network node connection relationship information.

7. The Bayesian analysis method based on sysML and knowledge graphs according to claim 4, characterized in that, The process of processing the Bayesian network node connection relationship information to obtain the target Bayesian network information includes: S31, The Bayesian network node connection relationship information is processed to obtain optimized Bayesian network node connection relationship information; S32, process the optimized Bayesian network node connection relationship information to obtain the target Bayesian network information.

8. A Bayesian analysis method based on knowledge graphs, characterized in that, The method includes: Obtain the knowledge graph information to be processed; The knowledge graph information to be processed is processed to obtain knowledge graph information; The knowledge graph information is optimized to obtain the target knowledge graph information; The target knowledge graph information is processed to obtain the target knowledge graph result information.

9. The Bayesian analysis method based on knowledge graphs according to claim 8, characterized in that, The process of processing the knowledge graph information to obtain the knowledge graph information includes: The knowledge graph information to be processed is subjected to information extraction processing to obtain model node information; the model node information includes several node information. The model node information is processed to obtain knowledge graph information.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that, when invoked, are used to execute the Bayesian analysis method based on sysML and knowledge graphs as described in any one of claims 4-7.

Images