Airborne system digital forward design method, device, equipment and medium

Abstract

The invention relates to the technical field of airborne system design, and discloses an airborne system digital forward design method, device, equipment and medium, a design process from demand analysis, architecture design, architecture tradeoff to simulation verification is established, integration and optimization of the airborne system design process are effectively realized, machine learning and multidisciplinary design optimization are combined, and the design efficiency is improved. According to the method, knowledge in multiple fields such as machinery, electronics and software is fused, design rules are mined from historical design cases through deep learning, a multi-target tradeoff analysis model is established, automatic generation and optimization of an architecture scheme are achieved, dependence on artificial experience is remarkably reduced, airborne system design efficiency is improved, design change cost is reduced, and the method is suitable for large-scale popularization and application. And the reliability of the airborne system is improved.

Description

Technical Field

[0001] This invention relates to the field of airborne system design technology, and discloses a digital forward design method, apparatus, equipment and medium for airborne systems. Background Technology

[0002] Currently, with the continuous increase in the complexity of airborne systems, the number of functional modules, data processing requirements, and inter-system collaboration requirements are all growing exponentially. Traditional airborne system development models primarily rely on experience accumulation and manual design. While these can meet basic needs at certain stages, they have gradually revealed numerous bottlenecks when facing mission requirements of high reliability, high security, high real-time performance, and rapid iteration. The main shortcomings of existing technologies include: (1) Dispersed design process: Each design stage (such as requirements analysis, architecture design, simulation verification, etc.) uses independent tools, resulting in data silos and making it difficult to achieve full-process collaboration; (2) Lack of intelligent support: The design process relies on human experience and lacks automatic optimization and decision support based on big data and artificial intelligence, resulting in low design efficiency and high error rate; (3) Insufficient multidisciplinary coupling: Airborne systems involve multiple disciplines such as mechanics, electronics, and software. Existing methods are difficult to effectively integrate multidisciplinary data and perform collaborative optimization. These shortcomings result in long development cycles and high costs for airborne systems, and make it difficult to achieve optimal system performance. Summary of the Invention

[0003] The purpose of this invention is to provide a digital forward design method, apparatus, equipment, and medium for airborne systems, which integrates and optimizes the airborne system design process, achieves multidisciplinary coupling, improves the design efficiency of airborne systems, reduces R&D costs, and enhances the reliability of airborne systems.

[0004] To achieve the above-mentioned technical effects, the technical solution adopted by the present invention is as follows: A digital forward design method for airborne systems includes: Based on a multi-source database of target design objects in the airborne system, a structured requirement model of the target design objects is established using knowledge graphs and natural language processing models; the multi-source database includes design requirements, design specifications and standards, airworthiness provisions, and historical design cases; The structured requirement model is transformed into a SysMLv2 architecture model using a generative AI model; the SysMLv2 architecture model includes a behavioral model, a structural model, and a parametric model. Using the structured requirements model, the SysMLv2 architecture model, and historical design cases as input, a set of system-level architecture solutions is generated using a multi-objective optimization algorithm. Based on a given quality quantification model, the TOPSIS method is used to score the quality of each system-level architecture solution, and the system-level architecture solution with the highest quality score is selected as the recommended architecture solution. The recommended architecture scheme is integrated with a multi-domain simulation model through the functional prototype interface standard to establish a system-level simulation model of the target design object, and joint simulation is performed based on a preset verification test case library. Then, it is determined whether the simulation results meet the preset verification rules. If they do, the system-level simulation model is deemed to have passed the verification and is output as the design result. Otherwise, the system-level simulation model is deemed to have failed the verification, and the parameters of the system-level simulation model are adaptively adjusted using a Bayesian optimization algorithm. Then, the joint simulation is performed again until the design result is output.

[0005] Furthermore, based on a multi-source database of target design objects in the airborne system, the method for establishing a structured requirement model of the target design objects using knowledge graphs and natural language processing models includes: Based on the multi-source database, an ontology modeling tool is used to construct an ontology model of the target design object, and the core classes of the target design object ontology model and the relationships between the core classes are defined; the core classes include requirement classes, function classes, performance classes, component classes, constraint classes, interface classes, and security classes; The text data in the multi-source database is semantically parsed using a natural language processing model to identify requirements that comply with design specifications and airworthiness clauses from the design requirements document of the target design object, and entities that comply with design specifications and airworthiness clauses are extracted from the design requirements of the target design object using a BiLSTM-CRF model. Based on the identified needs and extracted entities, a structured needs knowledge graph is built using a graph database to establish traceability relationships between needs; The structured requirement knowledge graph is converted into a structured requirement model and output.

[0006] Furthermore, before outputting the structured requirement model, the structured requirement knowledge graph is subjected to requirement conflict detection using the SPARQL query language, and the integrity constraint verification of the structured requirement knowledge graph is performed using a rule engine. If the structured requirement knowledge graph has requirement conflicts and / or incomplete constraints, the structured requirement knowledge graph is manually adjusted using the graph database, and requirement conflict detection and integrity constraint verification are performed again until there are no requirement conflicts and the constraints are complete.

[0007] Furthermore, the method for converting the structured requirement model into a SysMLv2 architecture model using a generative AI model is as follows: based on historical design cases in the multi-source knowledge base, the generative AI model learns the architecture design paradigm of the historical design cases, and then the generative AI model is used to convert the structured requirement model into a SysMLv2 architecture model.

[0008] Furthermore, after converting the structured requirement model into a SysMLv2 architecture model using a generative AI model, a formal verification tool is used to verify the architecture logic of the SysMLv2 architecture model. At the same time, the SysMLv2 architecture model is checked according to the given architecture design rules. If the architecture logic verification passes and the SysMLv2 architecture model conforms to the architecture design rules, the SysMLv2 architecture model is deemed qualified. Otherwise, the structural model, behavior, or parameter model of the SysMLv2 architecture model is manually adjusted until the SysMLv2 architecture model is qualified.

[0009] Furthermore, the quality quantification model includes preset design parameters for the target design object, such as weight design parameters, power consumption design parameters, cost design parameters, and reliability design parameters.

[0010] Furthermore, the multi-domain simulation model includes mechanical domain simulation models, electronic domain simulation models, and software domain simulation models corresponding to the target design object.

[0011] An airborne system digital intelligent forward design device, based on the aforementioned airborne system digital forward design method, includes: The requirements analysis module is used to establish a structured requirements model of the target design objects based on a multi-source database of the target design objects in the airborne system, using knowledge graphs and natural language processing models; the multi-source database includes design requirements, design specifications and standards, airworthiness clauses, and historical design cases; An architecture design module is used to convert the structured requirement model into a SysMLv2 architecture model using a generative AI model; the SysMLv2 architecture model includes a behavioral model, a structural model, and a parametric model. The architecture trade-off module is used to take the SysMLv2 architecture model and the given quality quantification model as input, according to the pattern definition of the airborne system architecture composition, driven by the component knowledge base corresponding to the target design object, guided by the design rules of the target design object, and generate a set of system-level architecture schemes using a multi-objective optimization algorithm. The TOPSIS method is used to score the quality of each system-level architecture scheme, and the system-level architecture scheme with the highest quality score is selected as the recommended architecture scheme. The simulation verification module is used to integrate the recommended architecture scheme with multi-domain simulation models through the functional prototype interface standard, establish a system-level simulation model of the target design object, and perform joint simulation based on a preset verification test case library. It determines whether the simulation results meet the preset verification rules. If the preset verification rules are met, the system-level simulation model is deemed to have passed the verification and is output as the design result. Otherwise, the system-level simulation model is deemed to have failed the verification. The parameters of the system-level simulation model are adaptively adjusted using a Bayesian optimization algorithm, and then the joint simulation is performed again until the design result is output.

[0012] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor implements the aforementioned airborne system digital forward design method when executing the computer program.

[0013] A computer-readable storage medium storing a computer program that performs the aforementioned digital forward design method for airborne systems.

[0014] Compared with the prior art, the beneficial effects of this invention are: This invention establishes a design process from requirements analysis, architecture design, architecture trade-offs to simulation verification, effectively realizing the integration and optimization of the airborne system design process. It adopts a combination of machine learning and multidisciplinary design optimization, integrating knowledge from multiple fields such as mechanics, electronics, and software. Through deep learning, it mines design patterns from historical design cases, establishes a multi-objective trade-off analysis model, and realizes the automatic generation and optimization of architecture solutions. This significantly reduces the dependence on human experience, improves the efficiency of airborne system design, reduces design change costs, and enhances the reliability of airborne systems. Attached Figure Description

[0015] Figure 1 This is a flowchart of the digital forward design method for airborne systems in Example 1; Figure 2 This is a structural block diagram of the airborne system digital forward design method device in Example 1; Figure 3 This is a structural block diagram of the computer device in Example 1; Wherein, 1-memory, 2-processor. Detailed Implementation

[0016] The present invention will now be described in further detail with reference to the embodiments and accompanying drawings. However, this should not be construed as limiting the scope of the above-described subject matter of the present invention to the following embodiments; all technologies implemented based on the content of the present invention fall within the scope of the present invention.

[0017] Example 1 See Figure 1This embodiment provides a digital forward design method for airborne systems, including: Step 1: Based on a multi-source database of target design objects in the airborne system, establish a structured requirement model for the target design objects using knowledge graphs and natural language processing models. The multi-source database includes design requirements, design specifications and standards, airworthiness clauses, and historical design cases. The target design object can be one or more of the avionics system, electromechanical system, and flight control system within the airborne system; it can also be a subsystem within the avionics system, electromechanical system, or flight control system.

[0018] Specifically, natural language processing technology is used to intelligently parse the textual requirements of the target design object, automatically identify key elements such as functionality, performance, and security, and establish a structured requirement model for the target design object. Semantic association analysis ensures the consistency and completeness of the requirements, providing accurate input for subsequent design activities. The detailed steps are as follows: Step 1.1: Based on the multi-source database, construct the target design object ontology model using the Protégé ontology modeling tool, and define the core classes of the target design object ontology model and the relationships between them. The core classes include requirement classes, functional classes, performance classes, component classes, constraint classes, interface classes, and security classes. Requirement classes include functional and non-functional requirements; functional classes include navigation and display functions; performance classes include positioning accuracy; component classes include Global Positioning System (GPS) receivers and Flight Management Computers (FMC); and constraint classes include weight constraints and power consumption constraints. Relationships between core classes include inheritance, dependency, and constraint. Inheritance indicates a "one-to-one" relationship between core classes; for example, functional and non-functional requirements both inherit from the requirement class. Similarly, functional classes can have sub-functions, and component classes can have sub-components. Dependency indicates that one core class uses another core class; for example, a functional class depends on a requirement class, a performance class depends on both functional and component classes, and a component class depends on both functional and performance classes. Constraint relationships represent the constraints that one core class imposes on another core class. For example, a constraint class may impose constraints on a component class, such as a weight constraint on a GPS receiver; a constraint class may impose constraints on a function class, such as a computation delay constraint on a navigation function.

[0019] Step 1.2: Use the BERT-Transformer architecture-based Natural Language Processing (NLP) model to perform semantic parsing on the text data in the multi-source database, identify the requirements that conform to the design specifications and airworthiness clauses from the design requirement document of the target design object, and use the BiLSTM-CRF model to extract entities belonging to each core class and conforming to the relevant design specifications and airworthiness clauses from the design requirements of the target design object; it should be noted that the entity refers to the key information unit with corresponding semantics and type extracted from the design requirements and specifications.

[0020] Step 1.3: Based on the identified requirements and extracted entities, a structured requirement knowledge graph is built using the Neo4j graph database to establish traceability relationships between requirements.

[0021] Step 1.4: Perform requirement conflict detection on the structured requirement knowledge graph using the SPARQL query language, and verify the integrity constraints of the structured requirement knowledge graph using a rule engine. If the structured requirement knowledge graph has requirement conflicts and / or incomplete constraints, return to Step 1.3, manually adjust the structured requirement knowledge graph, and then re-perform requirement conflict detection and integrity constraint verification until there are no requirement conflicts and the constraints are complete.

[0022] Step 1.5: Convert the structured requirement knowledge graph into XML format, outputting it as a structured requirement model. Also outputting a requirement traceability matrix and a requirement consistency analysis report. It should be noted that the requirement traceability matrix originates from the requirement traceability relationships in Step 1.3. All information in the requirement traceability matrix comes from the nodes (entities) and edges (relationships) in the knowledge graph.

[0023] Step 2: Use a generative AI model to convert the structured requirement model into a SysMLv2 architecture model; the SysMLv2 architecture model includes a behavioral model, a structural model, and a parametric model.

[0024] Specifically, based on multi-objective optimization algorithms and generative AI technologies, along with model formal verification and cross-tool integration development techniques, and under the SysMLv2 text language and ontology model library mapping mechanism, the generative AI model is used to transform the structured requirement model into a behavioral model, a structural model, and a parametric model. The detailed steps are as follows: Step 2.1: Based on historical design cases in the multi-source knowledge base of the target design object, a generative AI model is used to learn the architectural design paradigm of the historical design cases. Then, the generative AI model takes the structured requirement model and the SysMLv2 meta-model as input to convert the structured requirement model into a SysMLv2 architecture model. The SysMLv2 architecture model includes a behavioral model, a structural model, and a parametric model. It should be noted that the structured requirement model includes the functional requirements, performance requirements, and interface requirements of the target design object. The generative AI model is used to parse the structured requirement model to identify the components, functions, performance indicators, and constraints in the design object. Based on the components and their relationships mentioned in the requirements, a SysMLv2 structural model is generated; based on the functional and behavioral descriptions in the requirements, a SysMLv2 behavioral model is generated; and based on the performance indicators and constraints in the requirements, a SysMLv2 parametric model is generated.

[0025] Step 2.2: Verify the architectural logic of the SysMLv2 architecture model using formal verification tools: (1) Perform structural consistency verification: check whether the components, interfaces, ports and other elements in the SysMLv2 architecture model are consistent. For example, whether the connection between components follows the interface definition and whether the port types match.

[0026] (2) Perform behavioral consistency verification: Check whether the SysMLv2 behavioral model (such as state machine, activity diagram) is consistent with the SysMLv2 structural model. For example, whether the state transitions defined in the state machine are correctly implemented by the components in the structural model.

[0027] (3) Perform parameter consistency verification: check whether the equations and constraints in the parameter model are consistent with the structural model and the behavioral model. For example, whether the variables in the parameter model have corresponding attributes in the structural model, and whether the actions in the behavioral model affect the variables in the parameter model. When the SysMLv2 architecture model meets the predefined verification rules above, it indicates that the SysMLv2 architecture model is consistent in terms of structure, behavior, and parameters, and the SysMLv2 architecture model is deemed to have passed verification. If the SysMLv2 architecture model violates any of the above three verification rules, the SysMLv2 architecture model logic verification is deemed to have failed. Simultaneously, the formal verification tool generates a verification report detailing the specific reasons for and location of the failure, allowing designers to modify the SysMLv2 architecture model. Then, returning to step 2.1, based on the errors indicated by the formal verification tool, manual input is given to the generative AI model to generate an adjusted SysMLv2 architecture model. The architecture logic verification is then performed again until it passes. The formal verification tool can be a SysML-based architecture modeling tool, such as M-Design.

[0028] Step 2.3: Based on the given architecture design rules, check the SysMLv2 architecture model. If the SysMLv2 architecture model conforms to the architecture design rules, output the SysMLv2 architecture model, architecture verification report, and cross-domain interface specification. Otherwise, return to Step 2.1, manually input instructions into the generative AI model to generate an adjusted SysMLv2 architecture model, and then perform architecture logic verification and architecture design rule checks until the SysMLv2 architecture model conforms to the architecture design rules.

[0029] It should be noted that the architecture design rules include interface consistency rules, security rules, and reliability rules, among which: (1) The interface consistency rule aims to ensure that the components within the target design object in the airborne system and the target design object and the external environment can interact correctly and unambiguously. For example, the OWL inference engine of the network ontology model language is used to check whether the component interfaces in the SysMLv2 behavioral model match. (2) Safety rules are a set of mandatory design constraints aimed at ensuring that the target design object will not lead to catastrophic consequences in the event of a failure, and can continue to operate safely or enter a predetermined safe state. Safety rules originate from Functional Hazard Analysis (FHA) and system safety assessment processes. For example, functional isolation is verified based on the ARP4754 standard, that is, functions and dependencies are extracted from the SysMLv2 architecture model, the Development Assurance Level (DAL) of each function is determined based on the ARP4754 standard, then graph algorithms are used to detect paths, model checkers are used to verify resource conflicts, and formal tools are used to prove isolation, ensuring that high-critical functions are not affected by low-critical functions.

[0030] (3) Reliability rules focus on the ability of a target design object to continuously perform its intended functions without failure under specified conditions and within a specified time. Reliability rules quantify the durability and availability of a system through probabilistic and statistical methods. For example, FMEA analysis can be used to analyze whether a single point of failure exists in the SysMLv2 architecture model. A "single point of failure" refers to a single component, part, channel, signal, logic, or functional unit in the SysMLv2 architecture model whose independent failure would directly lead to the loss of the top-level function of the SysMLv2 architecture model and cause catastrophic or dangerous consequences. When applying FMEA analysis, firstly, the SysMLv2 architecture model is decomposed into the required levels to identify all components that need to be analyzed, i.e., potential "single points," such as sensors, controllers, valves, pumps, software modules, etc. Secondly, for each "single point" component, all foreseeable failure modes are systematically listed. For example: controller (hardware): power failure, processor failure, memory error. Finally, for each failure mode, the consequences are deduced from the bottom up along the functional flow and data flow in the system architecture model, proactively identifying and fixing system design flaws, thereby fundamentally improving the system's security and reliability.

[0031] Step 3: Using the structured requirements model, the SysMLv2 architecture model, and historical design cases as input, based on the known system-level functional architecture, according to the pattern definition of airborne system architecture composition, driven by the component knowledge base corresponding to the target design object, and guided by the design rules of the target design object, a set of system-level architecture schemes is generated using a multi-objective optimization algorithm. Based on the given quality quantification model, the TOPSIS method is used to score the quality of each system-level architecture scheme, and the system-level architecture scheme with the highest quality score is selected as the recommended architecture scheme.

[0032] Specifically, based on the known system-level functional architecture, and according to the pattern definition of the system architecture composition of the target design object, driven by the component knowledge base of the target design object, and guided by the relevant design rules of the target design object in the airborne system architecture, a complete set of system-level architecture solutions for the target design object is constructed. Then, based on the quality quantification model and the design parameters of the system-level architecture solutions, a quality score is calculated for each system-level architecture solution, and the optimal system-level architecture solution is selected as the recommended architecture solution according to the quality score. Detailed steps are as follows: Step 3.1: Using the structured requirements model of the target design object, the SysMLv2 architecture model, and historical design cases of the target design object as input, the NSGA-II multi-objective optimization algorithm is used to generate a set of system-level architecture solutions. When generating the set of system-level architecture solutions, the NSGA-II multi-objective optimization algorithm seeks the Pareto front across three optimization objectives: weight, power consumption, and cost.

[0033] Step 3.2: Based on the architecture trade-off rules, use modeling and simulation tools to verify each system-level architecture scheme. If a system-level architecture scheme does not meet the requirements of the architecture trade-off rules, adjust the corresponding system-level architecture scheme until all system-level architecture schemes meet the requirements of the architecture trade-off rules. The modeling and simulation tools can be MATLAB / Simulink, ANSYS, ModelCenter, etc. The architecture trade-off rules include performance weighting rules, cost constraint rules, and reliability allocation rules. Performance weighting rules refer to allocating performance index weights based on task criticality. Cost constraint rules refer to applying linear programming to solve for the cost-optimal solution. Reliability allocation rules refer to allocating reliability indexes based on the AGREE method. If the system-level architecture scheme does not meet the architecture trade-off rules, return to step 3.1 and regenerate the system-level architecture scheme set using NSGA-II multi-objective optimization algorithm.

[0034] It should be noted that the "task" in the performance index weighting based on mission criticality refers to both airborne system-level tasks and component-level functional tasks. System-level tasks include, for example, the flight control system's task is to maintain attitude stability and control; the fuel system's task is to ensure engine fuel supply and center of gravity balance; and the hydraulic system's task is to provide power for lowering the landing gear and speed brakes. Component-level functional tasks include, for example, the fuel pump's task is to provide fuel at a specified flow rate and pressure; and the fuel control valve's task is to respond to a shut-off command within 2 seconds to cut off fuel supply. The task weighting process is as follows: First, a list of system- and component-level critical tasks is decomposed based on the task profile; then, domain experts, based on criteria such as safety and mission success criticality, use the analytic hierarchy process (AHP) to compare tasks pairwise and construct a judgment matrix; by calculating the matrix's eigenvectors, the initial weights of each task are obtained; finally, normalization is performed to form a priority weight set for weighted evaluation of the architecture scheme.

[0035] Step 3.3: Based on the given quality quantification model, calculate the quality score of each system-level architecture scheme using the TOPSIS method according to the system-level architecture scheme design parameters; the quality quantification model includes preset weight design indicators, power consumption design indicators, cost design indicators, reliability design indicators, etc. of the target design object.

[0036] Step 3.4: Output the architecture scheme quality score matrix, and select the system-level architecture scheme with the highest quality score as the recommended architecture scheme. It should be noted that the architecture scheme quality score matrix includes the quality score of each architecture scheme on various preset indicators.

[0037] Step 4: Integrate the recommended architecture scheme with the multi-domain simulation model through the Functional Prototype Interface (FMI) standard to establish a system-level simulation model of the target design object. Perform co-simulation based on a preset verification test case library and determine whether the simulation results meet the preset verification rules. If the preset verification rules are met, the system-level simulation model is deemed to have passed verification and is output as the design result. Otherwise, the system-level simulation model is deemed to have failed verification. Adaptive parameter adjustment of the system-level simulation model is performed using a Bayesian optimization algorithm, and then co-simulation is performed again until the design result is output.

[0038] Specifically, step 4.1: Recommend an architecture scheme, multiphysics model, and verification based on the target design object. The verification test case library serves as input. Through the Functional Prototype Interface (FMI) standard, the recommended architecture scheme is integrated with multi-domain simulation models to establish a system-level simulation model of the target design object. Co-simulation is then performed based on the pre-set verification test case library. The multiphysics model includes fluid models, thermal models, and control models, each integrated with the recommended architecture scheme through the FMI standard. The verification test case library includes pre-set verification test cases for subsequent co-simulation; examples include extreme temperature conditions and fault injection test cases. The multi-domain simulation models include mechanical domain simulation models (e.g., structural simulation models, strength simulation models), electronic domain simulation models (e.g., electromagnetic simulation models), and software domain simulation models (e.g., control simulation models) corresponding to the target design object.

[0039] Step 4.2: During the co-simulation process, an LSTM-based anomaly detection algorithm is applied to identify whether there is abnormal data in the simulation process. Abnormal data includes time series abnormal data and multivariate relationship abnormal data. The time series abnormal data is data that deviates from the normal pattern. For example, during the operation of the system-level simulation model, data such as fuel system temperature, hydraulic system pipeline pressure, and electrical system current suddenly show unreasonable data jumps. The multivariate relationship abnormal data is data where there is a dynamic coupling relationship between multiple sensor variables, but the data does not change according to the original coupling relationship. For example, the hydraulic flow rate of the hydraulic system increases, but the output pressure value does not increase accordingly.

[0040] After the co-simulation is completed, the system-level simulation model is assessed based on the results to determine whether it meets the performance compliance rules, safety boundary rules, and real-time rules. The performance compliance rules verify whether the system-level simulation model's response time, accuracy, and other performance indicators meet the design requirements. The safety boundary rules check whether the parameters of the system-level simulation model exceed safety thresholds. The real-time rules verify the time limit requirements through time logic, i.e., whether the actual computation time of the system-level simulation model is less than the simulation step size.

[0041] Step 4.3: If there is no abnormal data in the co-simulation process, and the co-simulation results meet the performance compliance rules, safety boundary rules, and real-time rules, then the design of the target design object is completed, and the system-level simulation model is output as the design result; otherwise, the parameters of the system-level simulation model are adaptively adjusted using the Bayesian optimization algorithm, and then the system-level simulation model is re-performed in co-simulation until the design is completed and the design result is output.

[0042] This invention combines machine learning with multidisciplinary design optimization, integrating knowledge from multiple fields such as mechanics, electronics, and software. Through deep learning, it extracts design patterns from historical design cases, establishes a multi-objective trade-off analysis model, and realizes the automatic generation and optimization of architectural solutions. This significantly reduces the reliance on human experience and improves the scientific nature and efficiency of complex system design.

[0043] This invention leverages knowledge graphs and natural language processing models to deeply mine and analyze multi-source databases, enabling the accurate and comprehensive construction of a structured requirement model for the target design object, providing a solid foundation for subsequent design. Secondly, a generative AI model is used to transform the structured requirement model into a SysMLv2 architecture model containing behavioral, structural, and parametric models, making the design architecture clearer, more standardized, and scalable. Furthermore, using multiple models and historical cases as input, a multi-objective optimization algorithm generates a set of system-level architecture solutions, and the TOPSIS method is used to select the optimal recommended architecture solution, effectively improving the quality and efficiency of the design. Finally, the recommended architecture solution is integrated with multi-domain simulation models through a functional prototype interface standard, performing joint simulation and adaptively adjusting parameters based on the verification results until a design result that meets the requirements is output, ensuring the reliability and accuracy of the design. This invention establishes a design process from requirements analysis, architecture design, architecture trade-offs to simulation verification, effectively realizing the integration and optimization of the airborne system design process, improving airborne system design efficiency, reducing design change costs, and enhancing airborne system reliability.

[0044] Based on the same inventive concept, see Figure 2 This embodiment also provides an airborne system digital intelligent forward design device, implemented based on the aforementioned airborne system digital forward design method, including: The requirements analysis module is used to establish a structured requirements model of the target design objects based on a multi-source database of the target design objects in the airborne system, using knowledge graphs and natural language processing models; the multi-source database includes design requirements, design specifications and standards, airworthiness clauses, and historical design cases; An architecture design module is used to convert the structured requirement model into a SysMLv2 architecture model using a generative AI model; the SysMLv2 architecture model includes a behavioral model, a structural model, and a parametric model. The architecture trade-off module is used to take the SysMLv2 architecture model and the given quality quantification model as input, according to the pattern definition of the airborne system architecture composition, driven by the component knowledge base corresponding to the target design object, guided by the design rules of the target design object, and generate a set of system-level architecture schemes using a multi-objective optimization algorithm. The TOPSIS method is used to score the quality of each system-level architecture scheme, and the system-level architecture scheme with the highest quality score is selected as the recommended architecture scheme. The simulation verification module is used to integrate the recommended architecture scheme with multi-domain simulation models through the functional prototype interface standard, establish a system-level simulation model of the target design object, and perform joint simulation based on a preset verification test case library. It determines whether the simulation results meet the preset verification rules. If the preset verification rules are met, the system-level simulation model is deemed to have passed the verification and is output as the design result. Otherwise, the system-level simulation model is deemed to have failed the verification. The parameters of the system-level simulation model are adaptively adjusted using a Bayesian optimization algorithm, and then the joint simulation is performed again until the design result is output.

[0045] This embodiment also provides a computer device, such as... Figure 3 As shown, the system includes a memory 1, a processor 2, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the aforementioned digital intelligent forward design of the airborne system. Specifically, the computer device can be a computer terminal, a server, or a similar computing device.

[0046] This embodiment also provides a computer-readable storage medium storing a computer program that performs the above-described digital intelligent forward design of the airborne system.

[0047] Specifically, computer-readable storage media, including both permanent and non-permanent, removable and non-removable media, can store information using any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer-readable storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable storage media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0048] Example 2 This embodiment takes an aircraft environmental control system as the target design object and performs a forward design to further illustrate the invention. An aircraft environmental control system needs to meet various requirements, including temperature control range, airflow, pressure regulation, and emergency oxygen supply, while also considering constraints such as weight, power consumption, and reliability. In traditional design methods, requirements analysis, architecture design, and simulation verification are disconnected, leading to numerous design iterations and long cycles. Therefore, the forward design method of this invention includes: Step 1.1: Construct a multi-source database for the aircraft environmental control system using natural language design requirements documents (such as "cabin temperature maintained at 20-25℃", "maximum air supply flow rate of 0.5kg / s", etc.), design specifications (such as DO-160 environmental conditions), historical environmental control system design case data, and airworthiness clauses (such as CS-25.831 ventilation system). Use the Protégé ontology modeling tool to construct a domain ontology model of the aircraft environmental control system, defining core classes such as "temperature control", "air circulation", and "pressure regulation" and the relationships between these core classes.

[0049] Step 1.2: The BERT-Transformer model is used to perform semantic parsing on the text data in the multi-source database to identify requirements that comply with design specifications and airworthiness clauses from the design requirements document of the aircraft environmental control system. These requirements include functional requirements (such as "cooling" and "heating"), performance requirements (such as temperature range and flow rate), and safety requirements (such as "emergency oxygen supply"). Simultaneously, the BiLSTM-CRF model is used to extract entities that comply with design specifications and airworthiness clauses from the design requirements document, such as "temperature sensor" and "compressor".

[0050] Step 1.3: Use the Neo4j graph database to build a structured requirement knowledge graph to establish traceability relationships between the requirements, such as associating the requirement of "cabin temperature" with "temperature sensor" and "controller".

[0051] Step 1.4: Perform a requirement consistency check: Use SPARQL queries to check for requirement conflicts in the structured requirement knowledge graph. For example, simultaneously requiring "maximum gas supply flow rate 0.5 kg / s" and "minimum gas supply flow rate 0.6 kg / s" will trigger a conflict. Perform an integrity check: All interactions between components must have clearly defined interface protocols and data types. For example, if the requirement description is "temperature sensor" connecting to "environmental control computer," and the communication protocol (e.g., ARINC 429), data update frequency, and signal type are not defined, the constraint is incomplete. If the structured requirement knowledge graph has requirement conflicts and / or incomplete constraints, return to Step 1.3, manually adjust the structured requirement knowledge graph, and then re-perform requirement conflict detection and integrity constraint verification until there are no requirement conflicts and the constraints are complete.

[0052] Step 1.5: Convert the structured requirement knowledge graph into XML format and output it as a structured requirement model. At the same time, output the requirement traceability matrix and the requirement consistency analysis report. The requirement traceability matrix shows the relationship between requirements and design elements, and the requirement consistency analysis report is used to list all conflicts and inconsistencies found in the requirement consistency check.

[0053] Step 2.1: Based on historical design cases of environmental control systems (including common component models such as air circulators and heat exchangers), a generative AI model is used to learn the architectural design paradigm of these historical cases. Then, the generative AI model takes the structured requirement model output in Step 1.5 and the SysMLv2 meta-model as input to convert the structured requirement model into a SysMLv2 architecture model. For example, temperature control requirements are converted into closed-loop control behavior. The SysMLv2 architecture model includes a structural model, a behavioral model, and a parameter model. The structural model includes the components of the environmental control system, the behavioral model includes the control logic of the environmental control system, and the parameter model includes the performance parameters of the environmental control system.

[0054] Step 2.2: Use formal verification tools to verify the architectural logic of the SysMLv2 architecture model, such as verifying whether the control logic will cause temperature oscillations. If the architectural logic verification fails, return to step 2.1, input instructions into the generative AI model to generate an adjusted SysMLv2 architecture model, and then perform architectural logic verification again until the architectural logic verification passes.

[0055] Step 2.3: Execute the architecture design rules: Interface consistency rule: Check the component interfaces for matching using the OWL inference engine, such as checking the diameter and pressure rating of the air circulator outlet and pipe interface. Safety rule: Verify the architectural isolation between the refrigeration system and the emergency oxygen supply system based on the ARP4754 standard. Reliability rule: Apply FMEA automated analysis to identify single-point failures, such as the impact of a single temperature sensor failure. If the SysMLv2 architecture model conforms to the architecture design rules, output the SysMLv2 architecture model, architecture verification report, and cross-domain interface specifications. Otherwise, return to step 2.1, input instructions to the generative AI model to generate an adjusted SysMLv2 architecture model, and then perform architecture logic verification and architecture design rule checks until the SysMLv2 architecture model conforms to the architecture design rules. The cross-domain interface specifications include mechanical interface specifications, electrical interface specifications, data interface specifications, interface lists, interface control documents, etc. The architecture verification report includes a verification execution summary, detailed verification results (interface consistency verification results, security rule verification results, and reliability rule verification results), and model quality metrics (structural complexity, behavioral compliance, and parameter rationality), etc.

[0056] Step 3.1: Using the structured requirements model of the environmental control system, the SysMLv2 architecture model, the environmental control system component knowledge base (containing performance parameters and cost data of various compressors, turbines, and heat exchangers), and historical design cases as input, the NSGA-II multi-objective optimization algorithm is used to generate a set of system-level architecture schemes for the environmental control system. When generating the set of system-level architecture schemes, the NSGA-II multi-objective optimization algorithm seeks the Pareto front on the three optimization objectives of weight, power consumption, and cost.

[0057] Step 3.2: Implement architectural trade-off rules: Performance weighting rule: Assign performance metrics weights based on task criticality. For example, comfort (temperature control accuracy) has a high weight in passenger aircraft, while weight and power consumption have high weights in cargo aircraft. Cost constraint rule: Apply linear programming to solve for the optimal solution under a given cost budget. Reliability allocation rule: Assign system-level reliability metrics to each component based on the AGREE method.

[0058] Step 3.3: Based on the given quality quantification model, calculate the quality score of each system-level architecture scheme using the TOPSIS method, taking into account the achievement of multiple objectives. The quality quantification model includes preset design indicators for the environmental control system, such as weight design indicators (e.g., total system weight), power consumption design indicators (e.g., maximum steady-state power consumption), cost design indicators (e.g., single-unit production cost target), and reliability design indicators (e.g., mean time between failures).

[0059] Step 3.4: Use the random forest algorithm to predict the feasibility of the schemes, and predict the probability of success of each scheme in the detailed design and experimental stages based on historical data; Step 3.5: Output the quality evaluation matrix for each architecture scheme. The quality evaluation matrix lists the score of each scheme on each indicator. At the same time, select the system-level architecture scheme with the highest quality score as the recommended architecture scheme.

[0060] Step 4.1: Using the recommended architecture scheme and multiphysics model (including fluid model, thermal model, and control model) output in Step 3.5 as input, integrate the recommended architecture scheme with the multi-domain simulation model through the Functional Prototype Interface (FMI) standard to establish a system-level simulation model of the environmental control system; and perform joint simulation based on a preset verification test case library. The verification test case library includes extreme temperature condition verification test cases, fault injection test cases, etc.

[0061] Step 4.2: During the co-simulation process, an LSTM-based anomaly detection algorithm is applied to monitor the simulation data in real time and identify abnormal behaviors, such as temperature runaway. After the co-simulation, based on the results, it is determined whether the system-level simulation model meets the performance compliance rules, safety boundary rules, and real-time rules. Specifically, when determining whether the system-level simulation model of the environmental control system meets the performance compliance rules, the temperature control response time of the system-level simulation model is verified to be within the required range (e.g., the time to drop from 25℃ to 20℃ does not exceed 5 minutes). When determining whether the system-level simulation model of the environmental control system meets the safety boundary rules, the pressure of the system-level simulation model is checked to ensure that it remains within the safety threshold, for example, whether the cabin pressure does not exceed 1.1 atmospheres. When determining whether the system-level simulation model of the environmental control system meets the real-time rules, the control cycle of the system-level simulation model is verified through time logic to ensure that it meets the real-time requirements, for example, whether the control algorithm executes once every 100ms.

[0062] Step 4.3: If there is no abnormal data in the co-simulation process, and the co-simulation results meet the performance compliance rules, safety boundary rules, and real-time rules, then the environmental control system design is completed, and the system-level simulation model is output as the design result; otherwise, the parameters of the system-level simulation model are adaptively adjusted using a Bayesian optimization algorithm, for example, adjusting the PID controller parameters to optimize performance, and then the system-level simulation model is re-performed in co-simulation until the design is completed and the design result is output.

[0063] It should be noted that step 4.3 also outputs a simulation verification report of the system-level simulation model, which includes test cases that passed / failed.

[0064] Through the implementation of this embodiment, the design cycle of the environmental control system is shortened by reducing manual iteration and repetitive work, the design change cost is reduced because the change of requirements can be automatically transmitted to subsequent stages, and the reliability of the environmental control system is improved by discovering design defects in advance through intelligent trade-offs and simulation verification.

[0065] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A digital forward design method for airborne systems, characterized in that, include: Based on a multi-source database of target design objects in the airborne system, a structured requirement model of the target design objects is established using knowledge graphs and natural language processing models; the multi-source database includes design requirements, design specifications, airworthiness clauses, and historical design cases; The structured requirement model is converted into a SysMLv2 architecture model using a generative AI model; Using the structured requirements model, the SysMLv2 architecture model, and historical design cases as input, a set of system-level architecture solutions is generated using a multi-objective optimization algorithm. Based on a given quality quantification model, the TOPSIS method is used to score the quality of each system-level architecture solution, and the system-level architecture solution with the highest quality score is selected as the recommended architecture solution. The recommended architecture scheme is integrated with a multi-domain simulation model through the functional prototype interface standard to establish a system-level simulation model of the target design object, and joint simulation is performed based on a preset verification test case library; Then, it is determined whether the simulation results meet the preset verification rules. If they do, the system-level simulation model is deemed to have passed the verification and is output as the design result. Otherwise, the system-level simulation model is deemed to have failed the verification, and the parameters of the system-level simulation model are adaptively adjusted using a Bayesian optimization algorithm until the design result is output.

2. The airborne system digital forward design method according to claim 1, characterized in that, The methods for establishing structured requirement models of target design objects based on multi-source databases of target design objects in airborne systems, using knowledge graphs and natural language processing models, include: Based on the multi-source database, an ontology modeling tool is used to construct an ontology model of the target design object, and the core classes of the target design object ontology model and the relationships between the core classes are defined; the core classes include requirement classes, function classes, performance classes, component classes, constraint classes, interface classes, and security classes; The text data in the multi-source database is semantically parsed using a natural language processing model to identify requirements that comply with design specifications and airworthiness clauses from the design requirements of the target design object. The BiLSTM-CRF model is then used to extract entities that comply with design specifications and airworthiness clauses from the design requirements of the target design object. Based on the identified needs and extracted entities, a structured needs knowledge graph is built using a graph database to establish traceability relationships between needs; The structured requirement knowledge graph is converted into a structured requirement model and output.

3. The airborne system digital forward design method according to claim 2, characterized in that, Before outputting the structured requirement model, the structured requirement knowledge graph is subjected to requirement conflict detection using the SPARQL query language, and the integrity constraint verification of the structured requirement knowledge graph is performed using a rule engine. If the structured requirement knowledge graph has requirement conflicts and / or incomplete constraints, the structured requirement knowledge graph is manually adjusted using the graph database, and requirement conflict detection and integrity constraint verification are performed again until there are no requirement conflicts and the constraints are complete.

4. The airborne system digital forward design method according to claim 3, characterized in that, The method for converting the structured requirement model into a SysMLv2 architecture model using a generative AI model is as follows: based on historical design cases in the multi-source knowledge base, the generative AI model learns the architecture design paradigm of the historical design cases, and then the generative AI model is used to convert the structured requirement model into a SysMLv2 architecture model.

5. The airborne system digital forward design method according to claim 4, characterized in that, After converting the structured requirements model into a SysMLv2 architecture model using a generative AI model, a formal verification tool is used to verify the architecture logic of the SysMLv2 architecture model. At the same time, the SysMLv2 architecture model is checked according to the given architecture design rules. If the architecture logic verification passes and the SysMLv2 architecture model conforms to the architecture design rules, the SysMLv2 architecture model is considered qualified. Otherwise, the structural model, behavioral model, or parameter model of the SysMLv2 architecture model is manually adjusted until the SysMLv2 architecture model is qualified.

6. The airborne system digital forward design method according to claim 5, characterized in that, The quality quantification model includes preset design parameters for the weight, power consumption, and cost of the target design object.

7. The airborne system digital forward design method according to claim 6, characterized in that, Multi-domain simulation models include mechanical domain simulation models, electronic domain simulation models, and software domain simulation models corresponding to the target design object.

8. A digital intelligent forward design device for airborne systems, implemented based on the digital forward design method for airborne systems according to any one of claims 1-7, characterized in that, include: The requirements analysis module is used to establish a structured requirements model for the target design objects based on a multi-source database of target design objects in the airborne system, using knowledge graphs and natural language processing models. The multi-source database includes design requirements, design specifications, airworthiness clauses, and historical design cases. The target design objects include one or more of the environmental control system and avionics system in the airborne system. An architecture design module is used to convert the structured requirement model into a SysMLv2 architecture model using a generative AI model; the SysMLv2 architecture model includes a behavioral model, a structural model, and a parametric model. The architecture trade-off module is used to take the SysMLv2 architecture model and the given quality quantification model as input, according to the pattern definition of the airborne system architecture composition, driven by the component knowledge base corresponding to the target design object, guided by the design rules of the target design object, and generate a set of system-level architecture schemes using a multi-objective optimization algorithm. The TOPSIS method is used to score the quality of each system-level architecture scheme, and the system-level architecture scheme with the highest quality score is selected as the recommended architecture scheme. The simulation verification module is used to integrate the recommended architecture scheme with multi-domain simulation models through the functional prototype interface standard, establish a system-level simulation model of the target design object, and perform joint simulation based on a preset verification test case library. It determines whether the simulation results meet the preset verification rules. If the preset verification rules are met, the system-level simulation model is deemed to have passed the verification and is output as the design result. Otherwise, the system-level simulation model is deemed to have failed the verification. The parameters of the system-level simulation model are adaptively adjusted using a Bayesian optimization algorithm, and then the joint simulation is performed again until the design result is output.

9. A computer device comprising a memory (1), a processor (2), and a computer program stored in the memory (1) and executable on the processor (2), characterized in that, When the processor (2) executes the computer program, it implements the airborne system digital forward design method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that performs the digital forward design method for airborne systems according to any one of claims 1 to 7.

Images