Civil aircraft demand analysis method based on knowledge graph

Abstract

The civil aircraft demand analysis method based on the knowledge graph utilizes the characteristics of the knowledge graph and the graph database in explainability and high efficiency in multi-hop relationship search, finally improves the speed of demand link completeness calculation and demand link tracking; utilizes the characteristics that the knowledge graph can understand the demand sentence text features, finally can realize the recommendation of the demand link and the calculation of the demand design completeness.

Description

Technical Field

[0001] This invention relates to the field of civil aircraft demand analysis, specifically a method for itemized demand analysis using knowledge graphs as a technical means. Background Technology

[0002] Requirements management is a systematic approach to identifying, documenting, organizing, and tracking changes to system requirements. This approach aims to acquire, organize, and document system requirements, ensuring consistency between clients and project teams regarding changes in system requirements. Effective requirements management should include clear and explicit requirement statements, applicable attributes for each requirement type, and traceability between requirements and other requirements and project work, thereby achieving requirements integrity, traceability, and consistency. In the civil aircraft industry, requirements management is a crucial component of the development of airborne systems. It primarily involves engineering management of requirements generated during requirement capture, analysis, and validation, as well as those generated during functional analysis, design synthesis, integration, verification, and validation, ensuring that products strictly meet required specifications.

[0003] Traditional requirements analysis for civil aircraft systems is document-based, with requirements recorded in natural language. Designers manually link design results to the requirement text. If a requirement is not met, rework is necessary, leading to lengthy and costly iterative development. Currently, specialized requirements management tools are used in the aviation industry, enabling the migration and management of requirements from documents to databases. However, these tools lack semantic understanding and analysis of the requirement statements themselves. This prevents text-based requirements management from being tightly coupled with model-based design and development (MBD), significantly reducing the rate of requirement transmission in the "customer requirements - product design - product verification" chain. Consequently, it becomes difficult to identify influencing factors in a timely and complete manner when requirements change, resulting in frequent instances of requirements not being correctly implemented or fully verified. This greatly increases the system development cycle and cost.

[0004] Furthermore, current requirements are mostly expressed in unstructured statements, leading to a heavy reliance on engineering reviews for the analysis of requirement completeness and correctness. The chain-like transmission and compliance management of requirements in the "customer requirements - product design - product verification" process can only rely on manual intervention. Due to the knowledge asymmetry and varying skill levels of requirements managers, this significantly increases the risks during requirements development and maintenance.

[0005] In recent years, knowledge graphs, as a structured form of human knowledge, have attracted widespread attention from academia and industry. They are currently the most expressive, scalable, and self-learning-capable knowledge base construction solution. In the manufacturing sector, technological research and development and product development are knowledge-intensive activities; therefore, manufacturing knowledge engineering has become a focus for many researchers, and its application to product design, production management, and other fields has become one of the important directions for research and application in manufacturing. Knowledge graphs are a continuation of traditional knowledge engineering in the era of artificial intelligence and big data. Utilizing information extraction techniques, primarily document rule parsing and large-scale pre-trained language models, to graph civil aircraft requirement documents into knowledge graphs, designing the architecture of the domain knowledge graph, defining entity types and attributes, representing aircraft design requirement knowledge using triples, and constructing a structured aircraft design requirement domain knowledge graph, storing and utilizing requirements through reasonable knowledge representation methods, is a promising research direction. Summary of the Invention

[0006] To overcome the shortcomings of the existing technologies, this invention proposes a knowledge graph-based method for civil aircraft demand management and analysis. This method refines the granularity of demand management items and utilizes the high efficiency of graph databases in multi-hop relationship lookup to invent a new demand analysis method.

[0007] The technical solution of the present invention is as follows:

[0008] A knowledge graph-based method for demand analysis of civil aircraft, characterized by the following steps:

[0009] S1. Requirements Definition: Based on the civil aircraft requirements knowledge graph ontology model and civil aircraft requirements scenarios, create a model of civil aircraft requirements entities, types of relationships between entities, and entity attributes;

[0010] S2. Requirement Extraction: Extract knowledge from unstructured requirement text and convert it into knowledge graph data that conforms to the knowledge graph structure, including requirement category, parameter type, and parameter value;

[0011] S3. Demand Management: The Neo4j graph database is used to store and visualize the knowledge graph data. Cypher statements are used to operate on the database, and the functions of adding, deleting, modifying, and searching for requirements are completed to generate a knowledge graph of civil aircraft requirements.

[0012] S4. The civil aircraft demand knowledge graph is analyzed through big data correlation, including demand chain completeness calculation, demand chain tracking, demand chain recommendation, and demand design completeness calculation.

[0013] Furthermore, in S4.

[0014] a. Calculation of demand chain completeness, using the following formula:

[0015]

[0016] In the formula, the number of derived demands, the number of demands with recorded higher-level demands, and the total number of demands are retrieved from the civil aircraft demand knowledge graph.

[0017] b. Demand Path Tracing: The entire path from the demand node in the civil aircraft demand knowledge graph to the system node to which it belongs;

[0018] c. Demand Link Recommendation: For non-derived demand X without a superior demand record, first locate the demand type to which it belongs and the subsystem to which its component belongs in the civil aviation operation knowledge graph, and then find the demand set contained in the link from the subsystem to the demand type;

[0019] Connect X and Y i ∈Y serves as the input sentence pair matching model, and the cosine similarity between the output sentence pairs is used as the confidence score to rank the output results, returning the k results with the highest confidence.

[0020] If the results contain component-level requirements with high-level requirements, then the requirement chain is tracked and used as the requirement chain recommendation result for the current requirement X, and the cosine similarity is the confidence level of the recommendation.

[0021] d. Calculation of the completeness of the requirement design: First, for each requirement in the technical specification of the new model, find the corresponding requirement entity in the established knowledge graph. For each requirement x to be aligned, the equipment node and requirement type node of the requirement source have been determined. Use the graph query language Cypher to quickly locate the set of candidate requirement entities Y in the established graph.

[0022] For a requirement x to be aligned and a candidate requirement y∈Y, the input requirement statement matching model outputs the cosine similarity of the two texts; the requirement entity with similarity cos(θ)>ξ is selected as the alignment result, where ξ∈[0,1] is the threshold, and if there are multiple alignment results, the one with the highest similarity is selected.

[0023] Compared with the prior art, the beneficial effects of the present invention are: by utilizing the high efficiency of knowledge graphs and graph databases in interpretability and multi-hop relationship lookup, the speed of demand link completeness calculation and demand link tracking is ultimately improved; by utilizing the ability of knowledge graphs to understand the textual features of demand statements, demand link recommendation and demand design completeness calculation can be realized. Attached Figure Description

[0024] Figure 1 It is a knowledge graph ontology model for civil aircraft demand management.

[0025] Figure 2 It is a civil aircraft air conditioning system demand management tool built using the knowledge graph-based civil aircraft demand management and analysis method of this invention.

[0026] Figure 3 It is a requirement statement matching model framework.

[0027] Figure 4 It is a visualized demand link tracing achieved by using the knowledge graph-based civil aircraft demand analysis method of this invention.

[0028] Figure 5 This is a visualized demand link recommendation achieved by using the knowledge graph-based civil aircraft demand analysis method of this invention.

[0029] Figure 6 This is a schematic diagram of the knowledge graph-based civil aircraft demand analysis method of the present invention. Detailed Implementation

[0030] The present invention will now be described in detail with reference to the accompanying drawings and embodiments, but this should not be construed as limiting the scope of protection of the present invention.

[0031] ① Requirements Definition: When constructing a knowledge graph for civil aircraft requirements management, the first step is to define the ontology model of the requirements management knowledge graph, such as... Figure 1 Based on the established ontology model, the types and attributes of requirement entities and relationships between entities are defined; eight entity types and their corresponding entity attributes are defined, and Table 1 shows the entity types, descriptions, and their contained attributes; three types of relationships between requirement entities of the civil aircraft air conditioning system are defined, and Table 2 shows the relationship types and entity relationship triples.

[0032] Table 1 Entity Types, Descriptions, and Attributes

[0033]

[0034] Table 2 Relation Types and Triples

[0035]

[0036]

[0037] ② Requirement Extraction: Knowledge extraction from unstructured requirement text. Unlike typical information extraction tasks such as named entity recognition and entity relation extraction, the extraction goal is to automatically identify the requirement categories described in the requirement statements and the key parameters recorded within them. Its structure is not the traditional (head entity, relation, tail entity), but rather (requirement category, parameter type, parameter value). A type-parameter extraction model is constructed for this task. For civil aircraft valve equipment, some requirement types and hierarchies are shown in Table 3.

[0038] Table 3 Demand Types and Levels

[0039]

[0040] ③ Requirement Management: Neo4j graph database is used for the storage and visualization of knowledge graph data. Cypher statements can be used to operate on the database and complete basic management functions such as adding, deleting, modifying and searching requirements.

[0041] ④ Requirements Analysis: Leveraging the established knowledge graph for civil aircraft requirements management, various requirements analyses can be conducted, including requirements chain completeness calculation, requirements chain tracking, requirements chain recommendation, and requirements design completeness calculation.

[0042] The following sections will elaborate on the calculation of requirement chain completeness, requirement chain tracking, requirement chain recommendation, and requirement design completeness calculation.

[0043] Calculation of the completeness of the requirement chain: Due to the non-standard recording of requirements in the supplier's technical documents, there are cases where requirement attribute records and requirement chain records are missing. It is necessary to calculate the completeness of the requirement chain in the model design. Based on the civil aircraft requirement knowledge graph, to obtain the completeness of the requirement chain, we only need to search for ControlRequirement, SystemRequirement, and EquipmentRequirement in the current knowledge graph, calculate the number of derived requirements and requirements with higher-level requirements and the total number of requirements, and output the results according to formula (1).

[0044] This task can be quickly accomplished using the graph query language Cypher. The corresponding Cypher statement is as follows:

[0045] 1. Locate derived requirements and record the number of requirements with higher-level requirements:

[0046] The expression `match(n)where any(label in labels(n)where label in ['EquipmentRequirement', 'System Requirement', 'Control Requirement'])and not n.`requirement content (English)`contains 'Deleted'and(exists(n.`higher-level requirement`)or n.`requirement attribute`="DERIVEDREQUIREMENT")return count(n)`

[0047] 2. Find all required quantities:

[0048] match(n)where any(label in labels(n)where label in['EquipmentRequirement','System Requirement','Control Requirement'])and not n.`Requirement content (English)`contains'Deleted'return count(n)

[0049] Requirement Path Tracing: To quickly locate the superior and subordinate requirements when a requirement is modified, the requirement management tool needs to be able to return the complete requirement path. Based on the civil aircraft requirement knowledge graph, to obtain the requirement path of a certain requirement, it is only necessary to find all paths from the requirement node in the current knowledge graph to the system node to which it belongs.

[0050] This task can be quickly accomplished using the graph query language Cypher. The corresponding Cypher statement is as follows:

[0051] match(n),(m:"XXXXXX")where(n.code="XXXX")match data=(m)-[*1..5]->(n)return data

[0052] Requirement chain recommendation: For a non-derived requirement X without a parent requirement record, first locate its requirement type and the subsystem to which its component belongs in the knowledge graph, then find the requirement set Y contained in the links from the subsystem to the requirement type. The corresponding Cypher statement is as follows:

[0053] match(x),(m),(s:Subsystem),(y:`Equipment Requirement`)where(x.code="XXXX")match(x)-[r:`belongs to`]->(m)match(s)-[*1..5]->(x)match data=(s)-[*1..3]->(y)-[]->(m)return y

[0054] Connect X and Y i ∈Y as a sentence pair input Figure 3 The required statement matching model shown uses the cosine similarity between sentence pairs as a confidence score to rank the output results, returning the k results with the highest confidence. If any result contains a component-level requirement with a higher-level requirement, its requirement chain is traced and used as a requirement chain recommendation for the current requirement X, with the cosine similarity serving as the confidence score for the recommendation. This requirement chain can be updated to the graph after confirmation by engineers.

[0055] Requirements Design Completeness Calculation: When designing requirements for a system or component of a new aircraft model, ensure that the written requirements technical specifications include all necessary requirements for that system or component, without any omissions. Complete requirements design helps avoid or eliminate early errors in new model projects, thereby improving production efficiency, reducing development costs, and improving system quality. Requirements knowledge graphs can guide requirements design. Aligning existing content in the graph with new requirements designs can determine whether there are any missing requirements in the new design. Requirements design completeness can be defined by the following formula:

[0056]

[0057] Taking the requirements of the domestically produced C919 civil aircraft as a reference, we assume that its requirements design is complete. Based on the civil aircraft requirements knowledge graph built with C919 as the data source, we can quickly calculate the completeness of the requirements of the new design and the types of missing requirements, and provide guidance and suggestions to engineers when designing requirements.

[0058] First, for each requirement in the new model's technical specifications, the corresponding requirement entity is located in the established knowledge graph. Since the device node from which the requirement originates and the requirement type node are already determined for each requirement x to be aligned, the Cypher graph query language can be used to quickly locate the set of candidate requirement entities Y in the established graph. The corresponding Cypher statement is as follows:

[0059] match(n),(m)where n.name="XXXXX"and m.name="XXXXX"match(n)-[]-(Y)-[]->(m)return Y

[0060] For the requirement to be aligned x and the candidate requirement y∈Y, input them as follows: Figure 3 In the demand statement matching model, the cosine similarity of the two texts is output. We select the demand entity with similarity cos(θ)>ξ as the alignment result, where ξ∈[0,1] is the threshold. If there are multiple alignment results, the one with the highest similarity is selected.

[0061] The effectiveness of this invention was verified by a knowledge graph-based civil aircraft requirements management and analysis tool built using the C919 aircraft air conditioning system requirements document as the source.

[0062] Based on the currently constructed knowledge graph of C919 air conditioning system requirements, the requirement link completeness calculation method reveals 3744 derived requirements and requirements with recorded higher-level requirements, totaling 4962 requirements, resulting in a requirement link completeness of 75.5%. Taking the flow control valve requirement (REQ_70305_202) of the civil aircraft air conditioning system as an example, the requirement link tracing method can quickly locate its tracing link, such as... Figure 4 As shown. Taking the temperature control valve requirement X (REQ-70282-47_01) as an example, the requirement chain recommendation method finds that the requirement set Y contains only one requirement, REQ_70305_42. The corresponding sentence pair is input into the fine-tuned model, yielding a confidence score of 0.9767. Therefore, the requirement chain recommendation result for requirement REQ-70282-47_01 is as follows. Figure 5 As shown, the recommended confidence level is 0.9767. Taking the ARJ21 aircraft air conditioning system temperature control valve technical specification as an example, simulating the scenario of new model requirement design, the requirement design completeness calculation method shows that the ARJ21 aircraft air conditioning system temperature control valve technical specification contains 22 requirements, while the knowledge graph defines 42 requirements for temperature control valves. After inputting into the model, 18 requirement entities were successfully aligned. Partial alignment results are shown in Table 4.

[0063] Table 4 Example of Requirement Entity Alignment Results

[0064]

[0065]

[0066] Equation (2) shows that the requirement design completeness is 42.9%. The missing requirement types include pin, electromagnetic protection, valve travel limit, load capacity, sealing, durability, fatigue, drive type, failure mode, temperature limit, requirement conflict management, and torque.

Claims

1. A knowledge graph-based method for demand analysis of civil aircraft, characterized in that, Includes the following steps: S1. Requirement Definition: Based on the civil aircraft requirement knowledge graph ontology model and civil aircraft requirement scenarios, create a model of civil aircraft requirement entities, types of relationships between entities, and entity attributes; S2. Requirement Extraction: Extract knowledge from unstructured requirement text and convert it into knowledge graph data that conforms to the knowledge graph structure, including requirement category, parameter type, and parameter value; S3. Demand Management: The Neo4j graph database is used to store and visualize the knowledge graph data. Cypher statements are used to operate on the database, and the functions of adding, deleting, modifying, and searching for requirements are completed to generate a knowledge graph of civil aircraft requirements. S 4. The civil aircraft demand knowledge graph is analyzed through big data correlation, including demand chain completeness calculation, demand chain tracking, demand chain recommendation, and demand design completeness calculation; S4. a. The formula for calculating the completeness of the demand chain is as follows: （1） In the formula, the number of derived demands, the number of demands with recorded higher-level demands, and the total number of demands are retrieved from the civil aircraft demand knowledge graph. b. Demand Path Tracing: The entire path from the demand node in the civil aircraft demand knowledge graph to the system node to which it belongs; c. Requirement Path Recommendation: For non-derived requirements without parent requirement records. First, locate the demand type to which it belongs and the subsystem to which its component belongs in the civil aviation operation knowledge graph, and then find the demand set contained in the link from the subsystem to the demand type; Will and As a sentence pair input requirement sentence matching model, the cosine similarity between the output sentence pairs is used as a confidence score to sort the output results, and the k results with the highest confidence are returned; If the results contain component-level requirements with higher-level requirements, then trace the requirement chain for them as the current requirement. The recommendation results for the demand chain are calculated, and the cosine similarity is used as the confidence level of the recommendation. d. Calculation of requirement design completeness, using the following formula: （2） First, for each requirement in the new model's technical specifications, the corresponding requirement entity is found in the established knowledge graph, and each requirement to be aligned is... With the device nodes and demand type nodes of the demand source identified, the Cypher graph query language can be used to quickly locate the set of candidate demand entities in the established graph. ; For alignment requirements and candidate requirements In the input requirement statement matching model, the output is the cosine similarity of two texts; select the similarity The required entities are used as the alignment result, where It is a threshold; if there are multiple alignment results, the one with the highest similarity is selected.

Images