Aerospace equipment number-real fusion agile test method based on multi-modal learning and recommendation algorithm

Abstract

The invention discloses an aerospace equipment number-real fusion agile test method based on a multi-modal learning and recommendation algorithm, and belongs to the field of spacecraft digitization and computer science, and the method comprises the steps: carrying out the test demand analysis of an aerospace equipment test task according to the aerospace equipment design provided by a design department, and generating a structured test demand; according to structured test requirements, test schemes are generated in batches, and the test requirements and the test schemes are verified, iterated and screened on a digital simulation test platform; establishing a data-real fusion test scene: according to a structured test requirement, matching a proper aerospace equipment virtual model and a virtual environment model from an existing multi-granularity aerospace equipment test scene model library, configuring a physical test bed and setting connection; and performing a test task in the established data-real fusion test scene and analyzing an evaluation result. According to the invention, the agile test is greatly supported, and the comprehensiveness of the aerospace equipment test is improved.

Description

Technical Field

[0001] This invention belongs to the fields of spacecraft digitization and computer science, specifically relating to an agile testing method for data-real fusion of aerospace equipment based on multimodal learning and recommendation algorithms. Background Technology

[0002] Aerospace equipment refers to the collective term for technical equipment and related support systems capable of performing missions such as flight, exploration, transportation, combat, and support within the Earth's atmosphere, near space, or outside the atmosphere, including aircraft, satellites, and rockets. With the deep integration of the aerospace field, accelerated technological innovation, upgraded safety requirements, and the rapid development of emerging industries such as the low-altitude economy and commercial aerospace, the importance of aerospace equipment is becoming increasingly prominent. Testing and verification are the core means to ensure its performance. However, in traditional waterfall-style development, testing is often a separate stage, occurring after design completion. Detailed test plans are designed according to mission requirements, requiring a considerable amount of time to construct test scenarios. If test requirements change or problems are discovered during testing, each change necessitates rebuilding the test scenario and re-running the test process. This approach leads to late problem discovery, high repair costs, and long iteration times, failing to meet the demands of rapid aerospace equipment development. Furthermore, traditional physical testing, due to factors such as cost, risk, and environment, cannot complete tests on certain aspects of aerospace equipment, resulting in incomplete testing.

[0003] Current aerospace equipment testing suffers from the following problems: ① Existing aerospace equipment testing methods encounter numerous problems during testing due to changes in requirements and defects in testing schemes. This necessitates re-executing the testing process from the initial requirements analysis, significantly increasing time costs and hindering agile testing. ② Existing testing methods require manual design of test cases and cannot quickly respond to changes in testing requirements, resulting in low testing efficiency. ③ When matching aerospace equipment virtual models with virtual environment models, existing aerospace equipment testing relies on manual operation and suffers from insufficient richness in the aerospace equipment testing model library, making it difficult to guarantee model accuracy and reliability. ④ Existing physical testing methods for aerospace equipment are unsuitable for conducting tests on high-cost, high-risk subjects, failing to meet the requirements for comprehensive testing. Summary of the Invention

[0004] To address the aforementioned problems and meet the demands for rapid equipment development and urgent mission testing, this invention proposes an agile testing method for aerospace equipment based on multimodal learning and recommendation algorithms, which improves the timeliness, economy, and comprehensiveness of aerospace equipment testing to a certain extent. The technical problem solved by this invention is achieved through the following technical solution:

[0005] A method for agile testing of aerospace equipment based on multimodal learning and recommendation algorithms, comprising:

[0006] Step S101: Based on the aerospace equipment design provided by the design department, conduct test requirement analysis on the aerospace equipment test mission and generate structured test requirements.

[0007] Step S102: Based on the structured test requirements, generate test plans in batches using multimodal learning, and verify, iterate, and screen the test requirements and test plans on the digital simulation test platform;

[0008] Step S103: Build a data-real fusion test scenario: Based on the content-based recommendation algorithm, match suitable aerospace equipment virtual models and virtual environment models from the existing multi-granularity aerospace equipment test scenario model library according to the structured test requirements, configure the physical test bench and set up the connection.

[0009] Step S104: Conduct experimental tasks in the established data-real fusion experimental scenario and analyze and evaluate the results. Feed back the experimental results to the design department to drive design iteration.

[0010] A computing device includes: at least one processor and a memory storing program instructions; when the program instructions are read and executed by the processor, the computing device performs the method.

[0011] Beneficial effects:

[0012] (1) The experimental requirements and experimental plans are tested and verified in the digital simulation platform, which drives the modification of the experimental requirements and experimental plans. This exposes the problems of the experimental requirements and experimental plans in the early stage, and the experimental requirements and experimental plans can be modified and verified as early as possible. In this process, after multiple rounds of iterative closed loop, the user's experimental requirements are basically determined, which reduces the problems of the experimental plan and avoids the exposure of a large number of problems in the later stage of the experiment, which leads to rework of R&D and wastes a lot of time and cost of rebuilding the experimental scenario. This greatly supports agile testing.

[0013] (2) After the test requirements and test plan are iterated, check the integrity of the test cases and generate test cases based on the multimodal large model. This can enable the test cases to respond quickly to changes in the design requirements of aerospace equipment, reduce problems caused by test cases in the later test process, and become a basic element of agile testing.

[0014] (3) Based on recommendation algorithms and multimodal large models, it can quickly match and generate virtual models and virtual environment models of aerospace equipment, significantly reducing manual operation and judgment, building test environments quickly, responding to needs quickly, and suitable for emergency aerospace equipment test missions.

[0015] (4) Building a digital-real integrated test scenario does not require a lot of manpower, material resources and financial resources, which greatly saves costs. Furthermore, some high-risk test subjects can be carried out in a virtual environment, which reduces test risks and improves the comprehensiveness of aerospace equipment testing. Attached Figure Description

[0016] Figure 1 This is a flowchart of the agile test method for data-real fusion of aerospace equipment based on multimodal learning and recommendation algorithms according to the present invention.

[0017] Figure 2 This is a flowchart of the recommendation algorithm for the virtual model and virtual environment model of aerospace equipment in this invention. Detailed Implementation

[0018] 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. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the protection scope of the present invention.

[0019] This invention provides an agile testing method for data-real fusion of aerospace equipment based on multimodal learning and recommendation algorithms. Figure 1 This is a flowchart of the agile testing method for aerospace equipment based on multimodal learning, according to the present invention. Figure 2 This is a flowchart of the algorithm for recommending virtual models and virtual environment models for aerospace equipment according to the present invention. See below for reference. Figure 1 and Figure 2 The method is described. This description uses the ultimate load test of a civil aircraft wing as an example.

[0020] like Figure 1 As shown, the method specifically includes:

[0021] Step S101: Based on the aerospace equipment design provided by the design department, conduct a test requirement analysis for the aerospace equipment test mission and generate structured test requirements. In one embodiment, the aerospace equipment can be a civil aircraft, and the test mission can be a civil aircraft wing ultimate load test mission. Step S101 specifically includes:

[0022] S101-1, Input the design of each part of aerospace equipment, including relevant documents, models and data generated from conceptual design, overall design, subsystem design and integration design;

[0023] S101-2, based on natural language processing, the test requirement document in the document is parsed, semantic analysis is performed, and a pre-trained aerospace equipment named entity recognition model is used to extract key named entities and classify them into predefined categories, including equipment type, actions, components, environmental conditions, and performance indicators, to generate structured test requirements. In one embodiment, the equipment type is a civil airliner wing and a static test bench; the actions are progressive load increase and maintenance, data acquisition and monitoring; the main component is the wing's main wing box structure; the environmental conditions are simulated maneuvering flight loads; and the performance indicators are the structured test requirements for ultimate load, holding time, and whether there is permanent deformation.

[0024] S101-3, combining the knowledge graph of the aerospace equipment testing field, links the extracted named entities to corresponding nodes in the knowledge graph. Based on the relationships between entities in the knowledge graph, it infers the associated knowledge in the aerospace equipment testing field, generating knowledge-enhanced structured test requirements, including test type, test object, test scenario, test objective, performance indicators, and resource requirements. In one embodiment, the test type is structural static test and ultimate load verification; the test object is the main wing box structure; the test scenario is a windless, interference-free environment simulating maneuvering flight loads; the test objective is to verify the structural integrity of the wing structure under ultimate load and obtain strain distribution data of key parts; the performance indicators are ultimate load-bearing capacity, holding time, and presence of permanent deformation; and the resource requirements are a static test bench, a loading system, and a data acquisition system.

[0025] Step S102: Based on the structured test requirements, generate test plans in batches using multimodal learning, and verify, iterate, and screen the test requirements and test plans on a digital simulation test platform. Specifically, this includes:

[0026] S102-1, Based on the test requirements analysis results, generate test plans in batches based on multimodal large models, and determine the test type, test level, test method, test technology, test sequence, and test cases;

[0027] S102-2, From the multi-granularity aerospace equipment test scenario model library, call coarse-grained, low-fidelity aerospace equipment virtual models and virtual environment models to build a digital simulation platform for pre-testing the test plan and perform dynamic verification. In one embodiment, the aerospace equipment test scenario is a civil aircraft wing ultimate load test scenario. The aerospace equipment virtual model and virtual environment model can be a low-fidelity civil aircraft wing finite element model and a complex working condition environment virtual model. After calling the model, load the system model, set boundary conditions to simulate test bench constraints, build a digital simulation platform for pre-testing the test plan, and perform dynamic verification.

[0028] S102-3, Based on the pre-simulation results of the test plan on the digital simulation platform, analyze the structured test requirements and the problems existing in the test plan;

[0029] S102-4 If the structured test requirements analysis is complete and the requirements coverage meets the requirements, proceed to step S102-5; otherwise, report the problem to step S101-2 to supplement and modify the test requirements.

[0030] S102-5, Check and screen the batch-generated test plans. If there are feasible test plans, select the feasible test plan and proceed to step S102-6. Otherwise, proceed to step S102-1 to adjust the test plan or find an alternative test plan.

[0031] S102-6, systematically check the test case library for aerospace equipment, evaluate the mapping relationship between test requirements and test cases, check whether the coverage and redundancy ratio of test cases meet the requirements, and ensure that key test requirements are covered by high-intensity tests;

[0032] S102-7 If the test case coverage and redundancy meet the requirements, proceed to step S102-9 and input the test cases generated by the multimodal large model into the test case library in a certain format to supplement and optimize the test case library; otherwise, proceed to step S102-8.

[0033] S102-8, based on the understanding of test requirements input, equipment specifications, system design, failure mode library and historical test data by multimodal large model, generates missing test cases, including test objectives, test scope, preconditions, input conditions and parameters, test steps and expected results;

[0034] S102-9 generates an executable test plan based on the test requirements and test plan after iterative closure.

[0035] Step S103: Construct a data-real fusion test scenario. Based on the recommendation algorithm and according to the structured test requirements, match suitable virtual models and virtual environment models of aerospace equipment from the existing multi-granularity aerospace equipment test scenario model library, configure the physical test bench and set up connections. As mentioned above, the multi-granularity aerospace equipment test scenario model library can be a multi-granularity civil aircraft wing ultimate load test scenario model library. Step S103 is detailed below:

[0036] S103-1, a content-based recommendation algorithm, takes structured test requirements as input and matches fine-grained, high-fidelity aerospace equipment virtual models and virtual environment models from a multi-granularity aerospace equipment test scenario model library based on dimensions such as functional matching degree, model accuracy, computational efficiency, and interface compatibility. In one embodiment, the aerospace equipment virtual model and virtual environment model are a high-fidelity civil aircraft wing finite element model and a complex working condition environment virtual model;

[0037] S103-2, If the matching degree of the recommended model meets the requirements, proceed to step S103-4; otherwise, proceed to step S103-3.

[0038] S103-3 generates suitable virtual models of aerospace equipment and virtual environment models based on multimodal large models according to the needs of structured experiments, including model code, parametric models and geometric models, and integrates a physics engine to simulate physical environmental factors.

[0039] S103-4, Verify the virtual model of aerospace equipment and the virtual environment model, and identify whether there are any parts of the model that need to be corrected;

[0040] S103-5 If the model does not need to be modified, update the modified new model, new configuration, and new results to the model library to enrich the system intelligence, and proceed to step S103-7; otherwise, proceed to step S103-6.

[0041] S103-6, Perform model correction operation on the model to be corrected according to the model correction strategy. After the correction is completed, proceed to step S103-4.

[0042] S103-7, a data-physical fusion test bench is built according to the testing requirements of aerospace equipment. The physical test bench is configured according to the subsystem hierarchy, and data-physical interfaces are set up to connect the corresponding aerospace equipment virtual models and virtual environment models. The interface architecture is designed, and the connection parameters between all virtual components and the physical test bench are configured to ensure real-time data interaction. In one embodiment, the physical test bench includes a real wing specimen and a static test bench. Data-physical interfaces are set up to connect the corresponding high-fidelity civil aircraft wing finite element model and complex working condition virtual model. The interface architecture is designed to include command channels and feedback channels.

[0043] Step S104: Conduct experimental tasks in the established data-real fusion test scenario and analyze and evaluate the results. Feedback the experimental results to the design department to drive design iteration. This specifically includes:

[0044] S104-1. According to the generated test plan, test cases are obtained from the test case library, and the test cases are input into the aerospace equipment virtual model and virtual environment model. Control commands (which can be load grading loading commands) are generated and transmitted to the physical test bench to carry out the corresponding physical test (civil aircraft wing ultimate load test). Real physical test data is collected and fed back to the aerospace equipment virtual model and virtual environment model to generate new control commands for physical test.

[0045] S104-2 collects and analyzes physical and digital test data collected during the test, calculates key performance indicators, evaluates the performance of aerospace equipment (limit load performance of civil aircraft wings), and proposes optimization suggestions for the prototype design of aerospace equipment (design of civil aircraft wings).

[0046] S104-3, the analysis and optimization suggestions of the test results are fed back to the design department to drive the design department to modify the design.

[0047] like Figure 2 As shown, step S103-1, based on a content-based recommendation algorithm, takes structured test requirements as input and matches fine-grained, high-fidelity aerospace equipment virtual models and virtual environment models in a multi-granularity aerospace equipment test scenario model library from the dimensions of functional matching degree, model accuracy, computational efficiency, and interface compatibility. This includes: calculating the model matching degree using a content-based recommendation algorithm. This step specifically includes:

[0048] S103-1-1. Using historical aerospace equipment test requirement documents and the current test requirement document, calculate the current test requirement vector:

[0049] ① Review historical aerospace equipment test (civil aircraft wing ultimate load test) requirement documents and current test requirement documents. This constitutes a set of test requirements documents;

[0050] ② Use natural language processing to analyze all test requirement documents in the test requirement document set and extract the named entity keywords for aerospace equipment tests;

[0051] ③ Calculate the keywords in the test requirements document Weighted scores in: , , ,in This indicates the current test requirements document. Named entity keywords extracted through natural language processing. Keywords In the test requirements document Frequency of occurrence in Keywords In the test requirements document The number of times it appears in Documentation indicating test requirements Total word count Inverse document frequency, used to measure keywords The rarity of the test requirements document set, This indicates the total number of test requirement documents. Indicates the presence of keywords Number of test requirement documents Keywords In the test requirements document The weighted score in the middle;

[0052] ④ Construct a keyword library containing four dimensions: ,in This represents a keyword library related to functional matching. This represents a keyword library related to model accuracy. This represents a keyword library related to computational efficiency. This refers to a keyword library related to interface compatibility.

[0053] ⑤ Calculate the weights of each dimension of demand: ,in For dimension Demand weight, A database of relevant keywords for a specific dimension. for The sum of the weight scores of all keywords in the text. For the entire keyword database, for The sum of the weight scores of all keywords in the text;

[0054] ⑥ The demand weights calculated above constitute the current experimental demand feature vector: Demand feature vector These are the expected weights of the model obtained from the requirements analysis, where... Weighting based on functional matching requirements. Weights are required for model accuracy. To calculate the efficiency requirement weight, The weight for interface compatibility requirements ranges from 0 to 1.

[0055] S103-1-2. Calculate the model feature vectors of each model in the multi-granularity aerospace equipment test scenario model library based on historical information, and use this to determine the final recommended model:

[0056] ① Obtain the scores of each model in each dimension of the aerospace equipment test from the historical information in the model library;

[0057] ②Constructing the characteristic matrix: ,in The matrix formed by the scores of each dimension of the model. The elements in the vector represent the model feature fit scores. The elements in the vector represent the model accuracy score. The elements in the vector represent computational efficiency scores. The elements in the vector represent the interface compatibility score;

[0058] ③ Perform min-max normalization on each dimension to obtain the feature matrix: ,in The characteristic matrix, The elements in the vector represent the model function matching feature values. The elements in the vector represent the model accuracy feature values. The elements in the vector represent computational efficiency eigenvalues. The elements in the vector represent interface compatibility feature values; the normalization formula is: ,in These are the normalized eigenvalues. For the feature values ​​of this dimension, The minimum value of the feature values ​​in this dimension. The maximum value of the feature value in this dimension;

[0059] ④ Each row of the feature matrix represents a feature vector of a model: , Let be the feature vector of the model to be matched, where This represents the feature value of functional matching. Represents the model accuracy feature value. This represents the computational efficiency eigenvalue. Indicates interface compatibility characteristic value;

[0060] S103-1-3, Based on the experimental requirement vector and the model feature vector, calculate the model matching degree to determine the final recommendation model:

[0061] ① Calculate the model matching degree: , , , ,in , Representing vectors respectively , 2-norm, Indicates the number of elements in the vector. , For elements in the vector, The cosine similarity between the demand feature vector and the model feature vector is given. The degree of match between the final recommended model and the model required for the experiment;

[0062] ② The model with the highest matching degree is selected as the recommended virtual model for aerospace equipment and virtual environment. In one embodiment, the virtual model for aerospace equipment and virtual environment is a high-fidelity finite element model of a civil aircraft wing and a virtual model of a complex working environment.

[0063] In summary, this invention discloses an agile testing method for aerospace equipment based on multimodal learning and recommendation algorithms, including aerospace equipment testing requirements analysis, generation of aerospace equipment testing schemes based on multimodal large models, construction of data-real fusion testing scenarios, and testing evaluation process. It can solve the current problems in aerospace equipment testing to a certain extent and improve the timeliness, economy, and comprehensiveness of aerospace equipment testing.

[0064] The contents not described in detail in this specification are existing technologies known to those skilled in the art.

[0065] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for agile testing of aerospace equipment based on multimodal learning and recommendation algorithms, characterized in that, include: Step S101: Based on the aerospace equipment design provided by the design department, conduct test requirement analysis on the aerospace equipment test mission and generate structured test requirements. Step S102: Based on the structured test requirements, generate test plans in batches using multimodal learning, and verify, iterate, and screen the test requirements and test plans on the digital simulation test platform; Step S103: Build a data-real fusion test scenario. Based on the content recommendation algorithm, match the appropriate aerospace equipment virtual model and virtual environment model from the existing multi-granularity aerospace equipment test scenario model library according to the structured test requirements, configure the physical test bench and set up the connection. Step S104: Conduct experimental tasks in the established data-real fusion experimental scenario and analyze and evaluate the results. Feed back the experimental results to the design department to drive design iteration.

2. The agile experimental method for data-real fusion of aerospace equipment based on multimodal learning and recommendation algorithms according to claim 1, characterized in that, Step S101 includes: S101-1, Input the design of each part of aerospace equipment, including relevant documents, models and data generated from conceptual design, overall design, subsystem design and integration design; S101-2, based on natural language processing, parses the test requirement document in the document, performs semantic analysis, uses a pre-trained aerospace equipment named entity recognition model to extract key named entities and classify them into predefined categories, including equipment type, action, component, environmental conditions, performance indicators, and generates structured test requirements; S101-3 combines the knowledge graph of the aerospace equipment testing field, links the extracted named entities to the corresponding nodes in the knowledge graph, and infers the associated knowledge in the aerospace equipment testing field based on the relationships between entities in the knowledge graph, generating knowledge-enhanced structured test requirements, including test type, test object, test scenario, test objective, performance index, and resource requirements.

3. The agile experimental method for data-real fusion of aerospace equipment based on multimodal learning and recommendation algorithms according to claim 1, characterized in that, Step S102 includes: S102-1, Based on the test requirements analysis results, generate test plans in batches based on multimodal large models, and determine the test type, test level, test method, test technology, test sequence, and test cases; S102-2: Call coarse-grained, low-fidelity virtual models of aerospace equipment and virtual environment models from the multi-granularity aerospace equipment test scenario model library, build a digital simulation platform to pre-test the test plan, and perform dynamic verification. S102-3, Based on the pre-simulation results of the test plan on the digital simulation platform, analyze the structured test requirements and the problems existing in the test plan; S102-4 If the structured test requirements analysis is complete and the requirements coverage meets the requirements, proceed to step S102-5; otherwise, report the problem to step S101-2 to supplement and modify the test requirements. S102-5, Check and screen the batch-generated test plans. If there are feasible test plans, select the feasible test plan and proceed to step S102-6. Otherwise, proceed to step S102-1 to adjust the test plan or find an alternative test plan. S102-6, systematically check the test case library for aerospace equipment, evaluate the mapping relationship between test requirements and test cases, check whether the coverage and redundancy ratio of test cases meet the requirements, and ensure that key test requirements are covered by high-intensity tests; S102-7 If the test case coverage and redundancy meet the requirements, proceed to step S102-9 and input the test cases generated by the multimodal large model into the test case library in a certain format to supplement and optimize the test case library; otherwise, proceed to step S102-8. S102-8, based on the understanding of test requirements input, equipment specifications, system design, failure mode library and historical test data by multimodal large model, generates missing test cases, including test objectives, test scope, preconditions, input conditions and parameters, test steps and expected results; S102-9 generates an executable test plan based on the test requirements and test plan after iterative closure.

4. The agile experimental method for data-real fusion of aerospace equipment based on multimodal learning and recommendation algorithms according to claim 1, characterized in that, Step S103 includes: S103-1, a content-based recommendation algorithm, takes structured test requirements as input and matches fine-grained, high-fidelity aerospace equipment virtual models and virtual environment models in a multi-granularity aerospace equipment test scenario model library from the dimensions of functional matching degree, model accuracy, computational efficiency, and interface compatibility. S103-2, If the matching degree of the recommended model meets the requirements, proceed to step S103-4; otherwise, proceed to step S103-3. S103-3, based on a multimodal large model, generates suitable aerospace equipment virtual models and virtual environment models according to structured test requirements, including model code, parametric models and geometric models, and integrates a physics engine to simulate physical environmental factors; S103-4, Verify the virtual model of aerospace equipment and the virtual environment model, and identify whether there are any parts of the model that need to be corrected; S103-5 If the model does not need to be modified, update the modified new model, new configuration, and new results to the model library to enrich the system intelligence, and proceed to step S103-7; otherwise, proceed to step S103-6. S103-6, Perform model correction operation on the model to be corrected according to the model correction strategy. After the correction is completed, proceed to step S103-4. S103-7: Based on the testing requirements of aerospace equipment, a data-physical fusion test bench is built. The physical test bench is configured according to the subsystem level, and the data-physical interface is set to connect the corresponding aerospace equipment virtual model and virtual environment model. The interface architecture is designed and the connection parameters between all virtual components and the physical test bench are configured to ensure real-time data interaction.

5. The agile experimental method for data-real fusion of aerospace equipment based on multimodal learning and recommendation algorithms according to claim 1, characterized in that, Step S104 includes: S104-1. According to the generated test plan, test cases are obtained from the test case library, the test cases are input into the aerospace equipment virtual model and virtual environment model, control commands are generated and transmitted to the physical test bench to carry out the corresponding physical test, real physical test data is collected, the physical test data is fed back to the aerospace equipment virtual model and virtual environment model, and new control commands are generated to carry out the physical test. S104-2: Collect and analyze physical and digital test data collected during the test, calculate key performance indicators, evaluate the performance of aerospace equipment, and propose optimization suggestions for the prototype design of aerospace equipment. S104-3, the analysis and optimization suggestions of the test results are fed back to the design department to drive the design department to modify the design.

6. The agile experimental method for data-real fusion of aerospace equipment based on multimodal learning and recommendation algorithms according to claim 4, characterized in that, S103-1 includes: calculating the model matching degree through a content-based recommendation algorithm, specifically including: S103-1-1. Call up the historical aerospace equipment test requirement documents and the current test requirement documents, and calculate the current test requirement vector; S103-1-2. Calculate the model feature vector of each model in the multi-granularity aerospace equipment test scenario model library based on historical information, and determine the final recommended model accordingly. S103-1-3 Calculate the model matching degree based on the experimental requirement vector and the model feature vector, and then determine the final recommended model.

7. The agile experimental method for data-real fusion of aerospace equipment based on multimodal learning and recommendation algorithms according to claim 6, characterized in that, S103-1-1 includes: ① Review historical aerospace equipment test requirement documents and current test requirement documents This constitutes a set of test requirements documents; ② Use natural language processing to analyze all test requirement documents in the test requirement document set and extract the named entity keywords for aerospace equipment tests; ③ Calculate the keywords in the test requirements document Weighted scores in: , , ,in This indicates the current test requirements document. Named entity keywords extracted through natural language processing. Keywords In the test requirements document Frequency of occurrence in Keywords In the test requirements document The number of times it appears in Documentation indicating test requirements Total word count Inverse document frequency, used to measure keywords The rarity of the test requirements document set, This indicates the total number of test requirement documents. Indicates the presence of keywords Number of test requirement documents Keywords In the test requirements document The weighted score in the middle; ④ Construct a keyword library containing four dimensions: ,in This represents a keyword library related to functional matching. This represents a keyword library related to model accuracy. This represents a keyword library related to computational efficiency. This refers to a keyword library related to interface compatibility. ⑤ Calculate the weights of each dimension of demand: ,in For dimension Demand weight, A database of relevant keywords for a specific dimension. for The sum of the weight scores of all keywords in the text. For the entire keyword database, for The sum of the weight scores of all keywords in the text; ⑥ The demand weights calculated above constitute the current experimental demand feature vector: Demand feature vector These are the expected weights of the model obtained from the requirements analysis, where... Weighting based on functional matching requirements. Weights are required for model accuracy. To calculate the efficiency requirement weight, The weight for interface compatibility requirements ranges from 0 to 1.

8. The agile experimental method for data-real fusion of aerospace equipment based on multimodal learning and recommendation algorithms according to claim 7, characterized in that, S103-1-2 includes: ① Obtain the scores of each model in each dimension of the aerospace equipment test from the historical information in the model library; ②Constructing the characteristic matrix: ,in The matrix formed by the scores of each dimension of the model. The elements in the vector represent the model feature fit scores. The elements in the vector represent the model accuracy score. The elements in the vector represent computational efficiency scores. The elements in the vector represent the interface compatibility score; ③ Perform min-max normalization on each dimension to obtain the feature matrix: ,in The characteristic matrix, The elements in the vector represent the model function matching feature values. The elements in the vector represent the model accuracy feature values. The elements in the vector represent computational efficiency eigenvalues. The elements in the vector represent interface compatibility feature values; the normalization formula is: ,in These are the normalized eigenvalues. For the feature values ​​of this dimension, The minimum value of the feature values ​​in this dimension. The maximum value of the feature value in this dimension; ④ Each row of the feature matrix represents a feature vector of a model: , Let be the feature vector of the model to be matched, where This represents the feature value of functional matching. Represents the model accuracy feature value. This represents the computational efficiency eigenvalue. This indicates the interface compatibility characteristic value.

9. The agile experimental method for data-real fusion of aerospace equipment based on multimodal learning and recommendation algorithms according to claim 8, characterized in that, S103-1-3 includes: ① Calculate the model matching degree: , , , ,in , Representing vectors respectively , 2-norm, Indicates the number of elements in the vector. , For elements in the vector, The cosine similarity between the demand feature vector and the model feature vector is given. The degree of match between the final recommended model and the model required for the experiment; ② The model with the highest matching degree is selected as the recommended virtual model for aerospace equipment and virtual environment.

10. A computing device, characterized in that, include: At least one processor and a memory storing program instructions; When the program instructions are read and executed by the processor, the computing device performs the method as described in any one of claims 1-9.

Images