Intelligent aviation test platform

Abstract

The invention discloses an aviation intelligent test platform, and relates to the technical field of aviation test. According to the method, a locally deployed domestic large model and a knowledge base are introduced, a large model fine tuning technology is integrated to realize automatic generation from software requirements to test cases, then on the basis of graph or video identification of a multi-modal large model, characteristic graphs of avionics display control are identified through pictures, a test result is obtained, and the test efficiency is improved. Therefore, automatic execution of the whole testing process is achieved, many problems caused by manual testing are avoided, high efficiency is achieved by reducing labor input, the accuracy of a testing result is improved through an image characteristic recognition technology, and testing requirements of different industry fields can be met by building an exclusive knowledge base and a directional fine tuning model.

Description

Technical Field

[0001] The present invention relates to the field of aviation testing technology, and in particular to an aviation intelligent testing platform. Background Art

[0002] With the continuous advancement of aviation technology, the pace of version iteration is accelerating, and testing requirements are increasing. For example, during the development of a certain simulator model in 2024, the system version went through approximately 60 iterations. Each iteration required testing of basic functions and modified components based on new requirements. Under the existing testing process, a huge amount of manpower was invested to meet these testing requirements.

[0003] Although the domestic aviation software testing industry started late, it has developed rapidly, especially in the field of automated testing. However, it still faces the following problems: The generation of test cases still relies on manual decomposition of test requirements, and the quality of test cases is linked to the professional capabilities of testers.

[0004] Test script generation still relies on manual writing, requiring testers to have coding skills and requiring significant manpower. During project evaluation, automated testing is typically considered only when a module requires five or more iterations of testing.

[0005] Existing automated testing is unable to judge complex test results, which increases the complexity of testing and is prone to errors. For example, avionics testing still relies on manual image recognition to determine the correctness of the POP screen, and testers are required to observe the avionics display in the cockpit to confirm the test results.

[0006] Therefore, it is necessary to provide an aviation intelligent testing platform to solve the technical problems of existing technologies that rely on manual writing, the quality of test cases is linked to the professional ability of testers, and complex test results cannot be automatically judged. Summary of the Invention

[0007] The purpose of the present invention is to overcome the shortcomings of the background technology and provide an aviation intelligent testing platform, which aims to realize the automation of the entire process of intelligent test case generation - automated test execution (customized according to specific needs) - intelligent test result analysis - continuous integration and continuous delivery.

[0008] The purpose of the present invention is achieved through the following technical solutions: The present invention provides an aviation intelligent test platform for performing automated testing on aircraft, comprising: The presentation layer is used to interact with users and obtain user demand data; The business logic layer is used to execute corresponding business logic based on user demand data; it includes: Test case generation component, used to automatically generate test cases for aviation aircraft; A test result judgment component is used to automatically judge the test results of an aircraft; The AI ​​algorithm support layer is used to configure and run AI algorithm components and is called by the business logic layer; The data storage layer is used to provide data storage space and data call support for the presentation layer, business logic layer, and AI algorithm support layer; The system support layer is used to provide hardware equipment deployment, network environment configuration and security protection management for the entire aviation intelligent test platform.

[0009] As a further solution, the AI ​​algorithm support layer is provided with a domestic general large model, a multimodal large model, a document OCR module, a document parsing module and a knowledge vectorization module; wherein, the domestic general large model is used to provide natural language processing services; the multimodal large model is used to provide multimodal data processing services; the document OCR module is used to provide optical character recognition services and extract text information from images; the document parsing module is used to parse according to text information, extract key information and structured data, and the knowledge vectorization module is used to vectorize key information.

[0010] As a further solution, the data storage layer organizes the user requirement data into user requirement descriptions, obtains and stores corresponding requirement documents, stores expert knowledge through the expert knowledge base, provides knowledge support for the test case generation component, and stores the generated test cases.

[0011] As a further solution, the test case generation component pre-processes the requirement document through the document parsing module, including requirement document input processing and document intelligent parsing processing; wherein; Requirement document input processing: The presentation layer obtains the text information and / or image information provided by the user, and uses the document OCR module to convert the image information into text to obtain the corresponding user requirement data; the project knowledge used in the user requirement data is extracted and stored in the expert knowledge base, and the user requirement data is structured in table format and text format to obtain a structured requirement document and store it in the expert knowledge base; Intelligent document parsing and processing: The document parsing module is used to parse the table format and document format of the requirement document to obtain the corresponding element parsing data; the domestic general large model is used to extract the key elements of the element parsing data to obtain the corresponding key elements; the knowledge vectorization module is used to vectorize the key elements to obtain the corresponding key information.

[0012] As a further solution, the test case generation component automatically generates test cases through the following steps: Use predefined test step templates in the expert knowledge base to match key information in the requirements document with project knowledge and generate corresponding test steps; Generate detailed test step descriptions for each test step through the natural language generation capabilities of the domestic general large model; Use data generation tools and boundary value analysis tools to generate corresponding test input data, expected output data and boundary conditions for test steps; Generate the corresponding executable code file according to the test step description and boundary conditions through the code programming auxiliary component; Input the test input data into the executable code file and execute the code function to obtain the actual output data; Determine whether the actual output data meets the boundary conditions and whether the actual output data matches the expected output data; The executable code file that satisfies both the boundary conditions and matches the expected output data is output as an automatically generated test case.

[0013] As a further solution, the code programming assistance component is set in the business logic layer, which is obtained by conducting domain-specific programming training on domestic general-purpose large models; the business logic layer is also provided with an intelligent office assistance component, which is obtained by conducting domain-specific office training on domestic general-purpose large models, and is used to assist users in generating office operations for test cases, including information filling assistance, information checking assistance, operation guidance assistance and problem-solving assistance.

[0014] As a further solution, the presentation layer includes a user interface component and an interactive logic component, and is embedded with an intelligent office assistance component; wherein, the user interface component provides a visual interactive interface and obtains user operation information, and the interactive logic component performs interactive logic deconstruction on the user operation information to obtain corresponding user demand data.

[0015] As a further solution, the test result judgment component performs test result judgment through the following steps: Inject test cases into the target aircraft and perform corresponding operations according to the test cases; Collect image or video data from the target aircraft cockpit screen and perform preprocessing operations to obtain multimodal test data; Multimodal feature recognition and extraction are performed on multimodal test data through a multimodal large model to obtain corresponding feature data; The characteristic data is judged by the preset test result judgment logic to obtain the corresponding test judgment result; The multimodal test data, feature data and test judgment results are packaged and stored, and the corresponding data interface and result display are provided.

[0016] The aviation intelligent test platform of the present invention has at least the following beneficial effects: The present invention introduces a locally deployed domestic large model and knowledge base, and integrates large model fine-tuning technology to realize the automatic generation from software requirements to test cases. Then, based on the graphic or video recognition of the multimodal large model, the characteristic graphics of the avionics display and control are identified through pictures to obtain the test results, thereby realizing the automated execution of the entire test process, avoiding many problems caused by manual testing, achieving high efficiency by reducing labor input, improving the accuracy of test results through image feature recognition technology, and adapting to the testing needs of different industry fields by building a dedicated knowledge base and targeted fine-tuning model. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0018] Figure 1 A schematic diagram of the structure of an aviation intelligent test platform provided by the present invention; Figure 2 A schematic diagram of the automated test case generation process provided by the present invention; Figure 3 A schematic diagram of the test result automated decision process provided by the present invention; Figure 4 This is a schematic diagram of the existing general test process; The purpose, features and advantages of this application will be further explained with reference to the accompanying drawings in conjunction with the embodiments. DETAILED DESCRIPTION

[0019] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings herein can be arranged and designed in various different configurations.

[0020] Refer to the attached Figure 1 This embodiment provides an aviation intelligent testing platform for performing automated testing on aircraft, including: The presentation layer is used to interact with users and obtain user demand data; The business logic layer is used to execute corresponding business logic based on user demand data; it includes: Test case generation component, used to automatically generate test cases for aviation aircraft; A test result judgment component is used to automatically judge the test results of an aircraft; The AI ​​algorithm support layer is used to configure and run AI algorithm components and is called by the business logic layer; The data storage layer is used to provide data storage space and data call support for the presentation layer, business logic layer, and AI algorithm support layer; The system support layer is used to provide hardware equipment deployment, network environment configuration and security protection management for the entire aviation intelligent test platform.

[0021] It should be noted that the existing general testing process is as follows: Figure 4 As shown, this testing process involves a lot of manual operations, and its pain points are: 1. Test case design and execution require a large workload and manpower investment, and the quality of the test cases produced depends on the relevant skills of the testers.

[0022] 2. The test execution process requires testers to observe the avionics display in the cockpit to confirm the test results, which increases the complexity of the test and is prone to errors.

[0023] 3. The aviation industry has strong concerns about data security and privacy, and involving too many people will increase the potential risk of leaks.

[0024] To this end, this embodiment introduces a locally deployed domestic large model and knowledge base, and integrates large model fine-tuning technology to realize the automatic generation from software requirements to test cases. Then, based on the graphic or video recognition of the multimodal large model, the characteristic graphics of the avionics display and control are recognized through pictures to obtain test results, thereby realizing the automated execution of the entire test process, avoiding many problems caused by manual testing, achieving high efficiency by reducing labor input, improving the accuracy of test results through image feature recognition technology, and adapting to the testing needs of different industry fields by building a dedicated knowledge base and targeted fine-tuning model.

[0025] This embodiment can automatically generate available test cases based on the input project background and test requirements, comprehensively covering the functional logic of the aircraft simulator, including instructor control system testing, flight simulation system testing, avionics and mechanical weapon subsystem testing, battlefield simulation subsystem testing, visual simulation subsystem testing, etc.

[0026] This example leverages the natural language processing capabilities of a large AI model, combined with the specific needs of the target industry, to generate test cases through intelligent document parsing and structured data. The open source nature and deployment flexibility of the DeepSeek model make it more adaptable to local deployment needs.

[0027] The test result judgment component utilizes the image and video processing capabilities of a large multimodal model, combined with the specific needs of the target industry, to implement feature image recognition functions, extract key information such as latitude and longitude, altitude, direction and other graphic information, provide support for aircraft operation and monitoring in the target area, and can perform automated test result judgments well.

[0028] like Figure 1 As shown in the figure, the system architecture of the aviation intelligent test platform mainly consists of five layers: presentation layer, business logic layer, AI algorithm support layer, data storage layer, and system support layer. Each layer has its own specific internal components, which are described in detail below: Presentation Layer User interface: This is the interface through which users interact with the system, providing an intuitive operating experience and enabling users to easily access system functions.

[0029] Interaction logic: handles the logical relationship between user input and system output, ensuring the coherence and consistency between user operations and system responses.

[0030] Business Logic Layer Test case generation: Automatically generate test cases based on requirements to ensure the correctness and completeness of system functions.

[0031] Graphic or video recognition based on a multimodal large model: Leverage the image and video processing capabilities of the Qwen2.5-VL multimodal large model to achieve feature image recognition.

[0032] AI Algorithm Support Layer DeepSeek / Qwen: Large models deployed locally, providing powerful natural language processing and multimodal data processing capabilities.

[0033] Document OCR: Optical character recognition technology is used to extract text information from images.

[0034] Document parsing: Parse document content to extract key information and structured data.

[0035] Knowledge vectorization: Converting knowledge into vector representation to facilitate knowledge storage, retrieval, and application.

[0036] Data Storage Layer User requirement description: Stores user requirement documents to provide a basis for the implementation of system functions.

[0037] Test cases: Store generated test cases to support automation and standardization of the testing process.

[0038] Expert knowledge base: stores expert knowledge and provides knowledge support for test case generation.

[0039] System Support Layer Hardware deployment: Responsible for the hardware configuration and deployment of the system to ensure the stable operation of the system.

[0040] Network environment: Provides network support to ensure communication and data transmission between system components.

[0041] Security protection: Provide security measures to protect the system from external attacks and data leakage.

[0042] This architecture adopts a layered design with clear responsibilities between layers, which reduces system complexity and improves system maintainability and scalability.

[0043] Modular design: The internal components of each layer adopt modular design to facilitate independent development, testing and deployment.

[0044] AI-driven: Through the AI ​​algorithm support layer, the system has powerful intelligent capabilities and can realize functions such as automated test case generation and image recognition.

[0045] Data-driven: The data storage layer provides the system with rich data support, ensuring that the system's intelligent functions can operate effectively based on high-quality data.

[0046] Security: The system support layer provides support for hardware deployment, network environment, and security protection, ensuring system stability and data security.

[0047] like Figure 1 As shown, the AI ​​algorithm support layer is provided with a domestic general large model, a multimodal large model, a document OCR module, a document parsing module and a knowledge vectorization module; wherein, the domestic general large model is used to provide natural language processing services; the multimodal large model is used to provide multimodal data processing services; the document OCR module is used to provide optical character recognition services and extract text information from images; the document parsing module is used to parse according to text information, extract key information and structured data, and the knowledge vectorization module is used to vectorize key information.

[0048] It should be noted that when deploying large models locally, taking into account data security and privacy requirements, high-performance domestic open source large models such as DeepSeek or Qwen are given priority for local deployment to ensure data security and privacy protection. In addition, the local large model fine-tuning technology is combined to fine-tune the locally deployed large model according to the specific needs of the target industry to improve the adaptability and accuracy of the model in the target field.

[0049] The data storage layer uses retrieval-augmented generation (RAG) technology to build an expert knowledge base in the target field, solving the pain points of hallucination problems of large language models, lagging knowledge updates, and insufficient field specialization. It organizes user demand data into user demand descriptions, obtains corresponding demand documents and stores them. It stores expert knowledge in the expert knowledge base, provides knowledge support for the test case generation component, and stores the generated test cases.

[0050] like Figure 2 As shown, the test case generation module proceeds through three main steps: requirements document input, intelligent document parsing, and test case generation. This enables automated generation of test cases from requirements documents. Advanced technologies and tools are utilized in each step to ensure the system's efficiency and accuracy. Leveraging a locally deployed large model and RAG knowledge base, the system is able to process complex natural language text, extract key information, and generate high-quality test cases.

[0051] Specifically, the demand document input processing: obtain the text information and / or image information provided by the user through the presentation layer, and convert the image information into text through the document OCR module to obtain the corresponding user demand data; extract the project knowledge used in the user demand data and store it in the expert knowledge base, and structure the user demand data in table format and text format to obtain a structured demand document and store it in the expert knowledge base; Specifically, intelligent document parsing and processing: the document parsing module is used to parse the requirement document in table format (such as OpenPyXL) and document format to obtain the corresponding element parsing data; the domestic general large model is used to extract key elements from the element parsing data to obtain the corresponding key elements; the knowledge vectorization module is used to vectorize the key elements to obtain the corresponding key information.

[0052] Specifically, the test case generation component automatically generates test cases through the following steps: Use predefined test step templates in the expert knowledge base to match key information in the requirements document with project knowledge and generate corresponding test steps; Generate detailed test step descriptions for each test step through the natural language generation capabilities of the domestic general large model; Use data generation tools and boundary value analysis tools to generate corresponding test input data, expected output data and boundary conditions for test steps; Generate the corresponding executable code file according to the test step description and boundary conditions through the code programming auxiliary component; Input the test input data into the executable code file and execute the code function to obtain the actual output data; Determine whether the actual output data meets the boundary conditions and whether the actual output data matches the expected output data; The executable code file that satisfies both the boundary conditions and matches the expected output data is output as an automatically generated test case.

[0053] The business logic layer is equipped with a code programming auxiliary component and a smart office auxiliary component. The code programming auxiliary component can achieve: Code generation: Based on software requirements, users input functional requirements, programming languages, dependent libraries and other information, and the system automatically generates code to support specific programming requirements in target software development.

[0054] Code review: The user enters the code to be reviewed, and the system checks the errors and non-compliance with the code based on coding standards and grammatical knowledge, and provides modification suggestions to ensure that the code quality meets the target standards.

[0055] Code optimization: Optimize the code according to the optimization direction suggested by the user to improve code performance and maintainability.

[0056] Technical route: Utilize the code generation and review capabilities of DeepSeek or Qwen large models, combine with the characteristics of target software development, build an expert knowledge base, and implement code generation, review, and optimization functions.

[0057] Smart office auxiliary components can achieve: Knowledge Retrieval: Users upload internal company documents, which the system automatically categorizes, intelligently analyzes, and vectorizes to create an expert knowledge base. When users ask questions, the system extracts relevant data from the knowledge base, combines it with a large language model, and provides answers to users.

[0058] Solution generation: Based on the requirements input by the user, relevant data is retrieved from the knowledge base, and the data and user requirements are submitted to the large language model to generate solutions that meet the target industry standards, such as project bids, etc. The generated documents can also be checked for duplicates to ensure that the generated documents will not fail the duplicate check.

[0059] Technical implementation: Based on the knowledge retrieval and generation capabilities of DeepSeek or Qwen large models, an expert knowledge base in the target field is constructed to realize knowledge retrieval and solution generation functions.

[0060] Furthermore, the presentation layer includes a user interface component and an interactive logic component, and is embedded with an intelligent office assistance component; wherein, the user interface component provides a visual interactive interface and obtains user operation information, and the interactive logic component performs interactive logic deconstruction on the user operation information to obtain corresponding user demand data.

[0061] The test result judgment component implements feature image recognition based on the pictures or videos displayed on the cockpit screen, extracting key information such as latitude and longitude, altitude, direction, and other graphical information to support aircraft operation and monitoring in the target area. These components and the technologies used are described in detail below.

[0062] like Figure 3 As shown, the test result judgment component performs test result judgment through the following steps: Inject test cases into the target aircraft and perform corresponding operations according to the test cases; Collect image or video data from the target aircraft cockpit screen and perform preprocessing operations to obtain multimodal test data; Perform multimodal feature recognition and extraction on multimodal test data using a large multimodal model (such as the Qwen2.5-VL multimodal model) to obtain corresponding feature data; The characteristic data is judged by the preset test result judgment logic to obtain the corresponding test judgment result; The multimodal test data, feature data and test judgment results are packaged and stored, and the corresponding data interface and result display are provided.

[0063] The above are only some embodiments of the present application and are not intended to limit the patent scope of the present application. All equivalent structural transformations made using the contents of the present application specification and drawings under the technical concept of the present application, or direct / indirect application in other related technical fields are included in the patent protection scope of the present application.

Claims

1. An aviation intelligent test platform for performing automated testing on aircraft, characterized in that: include: The presentation layer is used to interact with users and obtain user demand data; The business logic layer is used to execute corresponding business logic based on user demand data; it includes: Test case generation component, used to automatically generate test cases for aviation aircraft; A test result judgment component is used to automatically judge the test results of an aircraft; The AI ​​algorithm support layer is used to configure and run AI algorithm components and is called by the business logic layer; The data storage layer is used to provide data storage space and data call support for the presentation layer, business logic layer, and AI algorithm support layer; The system support layer is used to provide hardware equipment deployment, network environment configuration and security protection management for the entire aviation intelligent test platform.

2. The aviation intelligent test platform according to claim 1, characterized in that: The AI ​​algorithm support layer is provided with a domestic general large model, a multimodal large model, a document OCR module, a document parsing module and a knowledge vectorization module; wherein, the domestic general large model is used to provide natural language processing services; the multimodal large model is used to provide multimodal data processing services; the document OCR module is used to provide optical character recognition services and extract text information from images; the document parsing module is used to parse according to text information, extract key information and structured data, and the knowledge vectorization module is used to vectorize key information.

3. The aviation intelligent test platform according to claim 2, characterized in that: The data storage layer organizes the user demand data into user demand descriptions, obtains corresponding demand documents and stores them, stores expert knowledge through the expert knowledge base, provides knowledge support for the test case generation component, and stores the generated test cases.

4. The aviation intelligent test platform according to claim 3, characterized in that: The test case generation component pre-processes the requirement document through the document parsing module, including requirement document input processing and document intelligent parsing processing; in; Requirement document input processing: The presentation layer obtains the text information and / or image information provided by the user, and uses the document OCR module to convert the image information into text to obtain the corresponding user requirement data; the project knowledge used in the user requirement data is extracted and stored in the expert knowledge base, and the user requirement data is structured in table format and text format to obtain a structured requirement document and store it in the expert knowledge base; Intelligent document parsing and processing: The document parsing module performs table format parsing and document format parsing on the requirement document to obtain the corresponding element parsing data; the domestic general large model is used to extract key elements from the element parsing data to obtain the corresponding key elements; The key elements are vectorized through the knowledge vectorization module to obtain the corresponding key information.

5. The aviation intelligent test platform according to claim 4, characterized in that: The test case generation component automatically generates test cases through the following steps: Use predefined test step templates in the expert knowledge base to match key information in the requirements document with project knowledge and generate corresponding test steps; Generate detailed test step descriptions for each test step through the natural language generation capabilities of the domestic general large model; Use data generation tools and boundary value analysis tools to generate corresponding test input data, expected output data and boundary conditions for test steps; Generate the corresponding executable code file according to the test step description and boundary conditions through the code programming auxiliary component; Input the test input data into the executable code file and execute the code function to obtain the actual output data; Determine whether the actual output data meets the boundary conditions and whether the actual output data matches the expected output data; The executable code file that satisfies both the boundary conditions and matches the expected output data is output as an automatically generated test case.

6. The aviation intelligent test platform according to claim 5, characterized in that: The code programming assistance component is set in the business logic layer and is obtained by performing domain-specific programming training on a domestic general-purpose large model; the business logic layer is also provided with an intelligent office assistance component, which is obtained by performing domain-specific office training on a domestic general-purpose large model, and is used to assist users in performing test case generation office operations, including information filling assistance, information checking assistance, operation guidance assistance and problem answering assistance.

7. The aviation intelligent test platform according to claim 6, characterized in that: The presentation layer includes a user interface component and an interactive logic component, and is embedded with an intelligent office assistance component; wherein, the user interface component provides a visual interactive interface and obtains user operation information, and the interactive logic component performs interactive logical deconstruction on the user operation information to obtain corresponding user demand data.

8. The aviation intelligent test platform according to claim 2, characterized in that: The test result judgment component performs test result judgment through the following steps: Inject test cases into the target aircraft and perform corresponding operations according to the test cases; Collect image or video data from the target aircraft cockpit screen and perform preprocessing operations to obtain multimodal test data; Multimodal feature recognition and extraction are performed on multimodal test data through a multimodal large model to obtain corresponding feature data; The characteristic data is judged by the preset test result judgment logic to obtain the corresponding test judgment result; The multimodal test data, feature data and test judgment results are packaged and stored, and the corresponding data interface and result display are provided.

Images