Aircraft multidisciplinary optimization design integrated system based on OpenMDAO

Abstract

The invention discloses an aircraft multidisciplinary optimization design integration system based on OpenMDAO, and belongs to the field of multidisciplinary optimization design. The system comprises the following modules: a subject establishment module, a coupling analysis module, an experiment design module, an optimization design module and a data modeling module. The subject establishment module automatically generates an OpenMDAO component through input, output and computational logic defined by a user. And the coupling analysis module automatically generates an expansion structure matrix and an N2 graph according to the subject selected by the user, so that the user can carry out multidisciplinary coupling analysis. The experiment design module carries out experiment design and sensitivity analysis according to the experiment design method of the selected subject and system integration. And the optimization design module allows a user to select a subject, set a design variable, an optimization target and a constraint condition, and perform multidisciplinary optimization. The data modeling module uses historical data generated by experimental design to carry out data modeling through an integrated modeling method, and the model is used for experimental design and optimization to balance the contradiction between calculation precision and quick response. According to the method, automatic establishment and calling of the OpenMDAO assembly in the overall design process of the aircraft are achieved, a user does not need to deeply know an OpenMDAO frame and an algorithm thereof, analysis and optimization of multidisciplinary problems can be rapidly completed only by defining input and output information of the disciplinary, and an efficient and convenient tool is provided for multidisciplinary optimization design.

Description

Technical Field

[0001] The present invention belongs to the field of multidisciplinary optimization design, and in particular relates to an aircraft multidisciplinary optimization design integrated system based on OpenMDAO. Background Art

[0002] The overall design of an aircraft is a high-dimensional and complex design process involving multiple systems, multiple disciplines, multiple variables, and multiple constraints. It directly affects the performance, development cycle, and funding of the aircraft, and therefore plays an important role in the development process of the aircraft. Multidisciplinary Design Optimization (MDO) is a method for designing complex systems and subsystems by fully exploring and utilizing the interactive synergy mechanisms in the system. The multidisciplinary design optimization in the overall design of an aircraft aims to organically integrate the credible analysis models and optimization techniques of various disciplines based on the MDO concept and the application of advanced computer technology to find the best design solution.

[0003] In order to improve the efficiency of the overall design of aerospace vehicles, accurately manage the multidisciplinary design status, and efficiently process multidisciplinary models and design data, it is particularly important to develop a multidisciplinary interactive integration software for aerospace vehicles. Based on the design concept of deep integration and comprehensive coverage of all stages of vehicle design, the software aims to promote close collaboration between disciplines, thereby meeting the complex needs of the new generation of aircraft design, achieving integrated optimization of multidisciplinary design, and significantly improving design capabilities and efficiency.

[0004] Aircraft MDO involves the combined use of multiple analysis programs and complex data interaction. Therefore, it is necessary to establish an MDO framework and develop an efficient and general software platform through a suitable integration method. As early as the 1990s, NASA proposed the importance of the MDO framework for MDO research. In the early days of the development of aircraft MDO, the MDO framework was mainly developed by research institutions, universities and industries according to their own needs or specific design objects. Among them, the OpenMDAO tool developed by NASA provides efficient and convenient multidisciplinary integration, data management and optimization analysis capabilities. However, OpenMDAO lacks a graphical user interface (GUI), making it relatively difficult to understand and use. Despite this, the flexibility and freedom of open source software are particularly important in the field of overall aircraft design. Compared with commercial software, it allows users to freely experiment and customize solutions to meet their unique design needs. Therefore, the development of a multidisciplinary design optimization system based on open source software is of great significance to promoting innovation and efficiency in aircraft design. It not only promotes technological progress, but also provides an open and collaborative environment for researchers and engineers to cope with various challenges in design and promote the development of the aerospace industry. Summary of the invention

[0005] In view of this, the purpose of the present invention is to provide an aircraft multidisciplinary optimization design integration system based on OpenMDAO. By modularizing the methods in the overall design of the aircraft and combining visual interface configuration and front-end and back-end interaction technology, users can quickly establish OpenMDAO components without understanding the underlying code and complex algorithm principles of OpenMDAO, complete multidisciplinary analysis, optimization and experimental design, and use integrated modeling methods for data modeling, providing strong support for the overall design of the aircraft.

[0006] The system achieves efficient integration of multidisciplinary design optimization through layered structural design. Each layer independently undertakes specific functions and communicates through a unified interface to ensure the efficiency and flexibility of the overall operation of the system. The entire framework includes four main levels and five core functional modules:

[0007] 1. Data management layer: Located in the basic layer of the system, it is responsible for unified management of various types of data in multidisciplinary design optimization and experimental design, ensuring accurate and timely data transmission between different disciplinary models. This layer provides efficient data transmission and storage mechanisms, while managing historical data to provide support for sensitivity analysis and agent model training. In addition, the data management layer also needs to persist the XDSM graph and N2 graph generated by the coupling analysis, and manage the data generated by the discipline establishment module and the data modeling module.

[0008] 2. Discipline coupling layer: The core of the system, responsible for the integration and coupling of various discipline models. Through the OpenMDAO component, the various discipline models are integrated into a unified solution process to ensure synchronous and consistent data transmission and realize iterative solutions among multiple disciplines until the optimization problem converges.

[0009] 3. Optimization decision layer: responsible for executing optimization algorithms and user decisions to ensure the realization of global optimization. It integrates multiple optimization algorithms, experimental design algorithms, and data modeling algorithms, allowing users to flexibly choose the most appropriate path and tools according to their needs. The optimization decision layer also supports the rapid construction of proxy models, replacing some high-computation simulation models with proxy models, reducing computing costs and accelerating the optimization convergence process.

[0010] 4. User interaction layer: Based on PyQt interface design technology, it provides an intuitive and easy-to-use operation interface, making multidisciplinary design optimization more convenient and efficient. Through QThread and QtCore multi-threading technology, it realizes the parallel processing of user operations and system background calculations, supports the visual display of modeling and optimization iteration process, and provides users with a highly customized interactive experience.

[0011] The five functional modules include:

[0012] 1. Discipline establishment module: Encapsulate each discipline model into an OpenMDAO component, implement data transmission and calling through standardized interfaces, and design a flexible pluggable model integration and update mechanism to meet the needs of model updating or replacement.

[0013] 2. Coupling analysis module: It provides a variety of coupling solvers and configures hyperparameters through a graphical interface to ensure that the coupling solution of multidisciplinary models is simple and efficient, and provides visualization tools to help users analyze the coupling relationship between disciplines.

[0014] 3. Experimental design module: Integrates multiple sampling methods, supports configuring hyperparameters through a graphical interface to complete experimental design, and provides an open interface for users to integrate new sampling algorithms.

[0015] 4. Design Optimization Module: It performs multidisciplinary design optimization based on subject coupling, integrates multiple optimization algorithms, automatically identifies and configures optimization variables through a graphical interface, simplifies the multidisciplinary optimization process, and enables non-professional users to use advanced optimization tools.

[0016] 5. Data modeling module: Integrates BPNN, DDNet and Kriging models, supports data modeling and real-time viewing of modeling results through a graphical interface, and persistently stores the model through pickle and PyTorch serialization tools to ensure the reusability and reliability of the model. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is the overall framework diagram of the present invention.

[0018] Figure 2 A module diagram is created for the subject of the present invention.

[0019] Figure 3 The subject matrix of the present invention is dynamically more detailed and maintains the diagram.

[0020] Figure 4 This is a module diagram of the experimental design of the present invention.

[0021] Figure 5 This is a library diagram of the design optimization algorithm of the present invention.

[0022] Figure 6 This is a data modeling module diagram of the present invention. DETAILED DESCRIPTION

[0023] The following will be combined with the embodiments of the present invention Figure 1 , clearly and completely describe the technical solutions in the embodiments of the present invention.

[0024] The present invention provides an aircraft multidisciplinary optimization design integration system based on OpenMDAO. Each functional module in the system is written in Python. OpenMDAO is used as a core coupling analysis tool, and an interactive page is developed in combination with PyQt5. The interactive page is finally run on a Windows operating system in the form of a client.

[0025] The user first defines the input, output and operation logic of the subject in the interface through the subject establishment module. If the operation logic is a formula calculation, the user needs to enter the formula in Python format in the text box of the interface, and the system will automatically convert it to the form required by OpenMDAO; if the operation logic is an executable program (EXE), the user needs to upload the program and provide the necessary configuration, the system will automatically parse and call the program through the subprocess module for calculation; if the operation logic is a proxy model, the user can upload a pre-trained model or select a model in the local model library. After the configuration is completed, the system will generate the corresponding OpenMDAO component and save it using pickle.

[0026] After the components are established, users can select the required components through the interface, select the coupling analysis method provided by the coupling analysis module through the drop-down menu, and configure the relevant hyperparameters for decoupling. The system automatically generates the corresponding XDSM diagram and N2 diagram to help users understand the calculation order and data dependencies between disciplines. These diagrams are saved in PDF and HTML formats and visualized through QWebEngineView.

[0027] After the coupling analysis is completed, the system enters the optimization decision layer. During the optimization process, users can first perform experimental design and data modeling to cope with the long-term computing requirements that may be brought about by complex problems. Users can select the components that need to be experimentally designed, and select the appropriate sampling algorithm through the interface, configure the hyperparameters to complete the experimental design, and the experimental design results will be automatically saved as a CSV file for subsequent analysis and use.

[0028] After the experimental design is completed, the user can open the data modeling window, select the appropriate modeling algorithm and configure the relevant parameters for modeling training, and view the convergence curve of the loss function in real time to determine the training effect of the model. The established model will be stored locally in pth and pkl formats for subsequent use.

[0029] After the model is established, users can use these models as subject components in subsequent optimization modules. In the optimization module, users select the components to be optimized, and the system will automatically identify all variables in the components, including design variables, constraint variables, and response variables, and configure relevant parameters through the interface, select the optimization algorithm, and the system will automatically open a thread for the optimization task. During the optimization process, users can view the changing trends of each variable in real time. After the optimization is completed, the system will save the results as a CSV file.

[0030] The above is a detailed introduction to the aircraft multidisciplinary optimization design integration system based on OpenMDAO provided by the present invention. The description of the embodiments in this article is only used to help understand the method of the present invention and its core idea, but not to limit it. Therefore, any other similar implementation methods derived by technical personnel in this field based on the technical solution of the present invention also belong to the protection scope of the present invention.

Claims

1. An aircraft multidisciplinary optimization design integration system based on OpenMDAO, written in Python, characterized by: include: Data management layer: responsible for unified management of various data in multidisciplinary design optimization and experimental design, ensuring accurate and timely data transmission between different disciplinary models; Discipline coupling layer: includes discipline establishment module and coupling analysis module, responsible for the integration and coupling of various discipline models; Optimization decision layer: includes design optimization module, experimental design module and data modeling module, responsible for executing experimental design and optimization algorithms and supporting user modeling decisions; User interaction layer: Based on PyQt interface design technology, it provides an intuitive and easy-to-use operation interface, making multidisciplinary design optimization more convenient and efficient, and realizes parallel processing of user operations and system background calculations through QThread and QtCore multi-threading technology.

2. The aircraft multidisciplinary optimization design integrated system based on OpenMDAO according to claim 1, characterized in that: The subject establishment module uses the factory mode and template method design mode to integrate the call of OpenMDAO's ExplicitComponent. It can automatically generate OpenMDAO's subject components according to the input information, output information, derivative calculation method and operation logic defined by the user in the system interface, and serialize and save the component as a pkl file to achieve persistent storage. At the same time, it supports the "plug and unplug" call of subject components for subsequent multidisciplinary analysis and optimization. The input information and output information are defined by the user by adding input variables and output variables. Each variable contains a variable name, variable value, variable type, unit and description. This information is called into the setup() function of the OpenMDAO component; the derivative calculation method supports finite differences, complex step lengths, precise derivatives and no derivatives, and all relevant information is called into the setup_partials() function of the OpenMDAO component; the operation logic includes formula calculation, model calculation and EXE program calculation, which are called into the compute() function of the OpenMDAO component, which are responsible for different types of subject calculation tasks respectively, ensuring that various types of calculation tasks are executed independently, in line with the single responsibility principle of software development.

3. The aircraft multidisciplinary optimization design integrated system based on OpenMDAO according to claim 1, characterized in that: The adaptive coupling analysis module can automatically generate and dynamically update the visualization matrix through the proxy design mode, including the design structure matrix (DSM), the extended design structure matrix (XDSM) and the N2 diagram, which are stored locally in PNG, PDF and HTML formats respectively, and these visualization results are displayed through QwebEngineView to ensure the efficiency and adaptability of the multidisciplinary coupling analysis process.

4. The aircraft multidisciplinary optimization design integrated system based on OpenMDAO according to claim 1, characterized in that: The experimental design module integrates a variety of experimental design methods, including full factorial design, Box-Behnken design, uniform design and Latin hypercube design. It supports users to easily configure and call experimental design schemes according to problem requirements. The system automatically identifies the variables of the selected subject and the hyperparameters of the experimental design, uses multi-threading technology to improve calculation efficiency, and uses matplotlib to visualize the experimental design results. The result data is saved locally in CSV format to facilitate users to perform parameter sensitivity analysis and optimization design.

5. The aircraft multidisciplinary optimization design integration system based on OpenMDAO according to claim 1 is characterized in that: The optimization design module specifically includes an integrated optimization algorithm, a linear multidisciplinary coupling solver, and a nonlinear multidisciplinary coupling solver. After the user configures the optimization algorithm and coupling solver parameters, sets the design variable value range, constraints, and optimization objectives, multidisciplinary optimization can be quickly performed; wherein, The optimization algorithms include SLSQP (sequential quadratic programming), CONMIN (constrained minimization) and IPOPT (interior point optimizer) in ScipyOptimize, as well as genetic algorithm (GA) and particle swarm optimization (PSO) integrated by inheriting openmdao.api.Driver. Users can choose the appropriate problem optimization algorithm according to their needs, and the optimization results are saved locally in CSV format; The linear multidisciplinary coupling solver has six options for users to choose from: None, linear block Gauss-Seidel (linearblockGS), linear block Jacobi (linearblockJac), direct solver (directsolver), SciPy Krylov solver (scipykrylov) and PETSc Krylov solver (PETScKrylov), and the user configures the hyperparameters of the linear solver through the system interface; The nonlinear multidisciplinary coupling solver has five options for users to choose from: None, nonlinear block Gauss-Seidel (NonlinearblockGS), nonlinear block Jacobi (NonlinearblockJac), Newton solver (NewtonSolver) and BroydenSolver (BroydenSolver). Users configure the hyperparameters of the nonlinear solver through the system interface.

6. The aircraft multidisciplinary optimization design integrated system based on OpenMDAO according to claim 1, characterized in that: The data modeling module specifically includes three modeling methods: dendritic neural network (DDNet), back propagation neural network and Kriging algorithm. It supports users to dynamically configure independent variables and dependent variables, as well as hyperparameters required by the model algorithm in the interface, adopts multi-threading technology to improve modeling efficiency, and provides real-time loss function monitoring and feedback mechanism to help users dynamically adjust model parameters, improve modeling accuracy and the ability to capture nonlinear features. The model is stored locally in pth or pkl format.

7. The aircraft multidisciplinary optimization design integration system based on OpenMDAO according to claim 1 is characterized in that: The system has an open interface, which allows users to quickly integrate new optimization algorithms, experimental design methods and data modeling methods through the interface, and uses object-oriented design patterns to achieve high scalability and modularity. Users can expand system functions according to their needs.

Images