A machine learning-based design method for aircraft structures

Abstract

The invention discloses a method for designing aircraft structural parts based on machine learning, comprising the following steps: 1) inputting the design requirements and related parameters of the aircraft structural parts to be designed through a human-computer interaction interface; 2) OpenCv assisted feature recognition performs feature analysis on the aircraft wing structural parts; 3) retrieving similar instances based on rule selection and case reasoning and selecting instances with high similarity, on the basis of which the designer modifies the instances according to the requirements of the required design; 4) classifying the extracted features according to the k-nearest neighbor algorithm, and adding them to the training library of machine learning as the result of training and learning; 5) saving the modified instances and design solutions as new instances in the instance library as a new instance. The method expands the knowledge-model-instance library data, increases the utilization rate of the design parameters of aircraft structural parts, improves the design quality and efficiency of aircraft structural parts, and enhances the research and development capabilities.

Description

Technical Field

[0001] The present invention relates to the field of aircraft design, and in particular to an aircraft structural component design method based on machine learning. Background Art

[0002] With the progress of the times and the continuous development of digital technology, the performance of modern aircraft is getting better and better, and the technological content is getting higher and higher, which makes the role of aircraft in both civil and war very important. Aircraft structural parts are gradually developing in the direction of large-scale, integrated and complex. The efficient and high-quality design of aircraft structural parts is an important condition to ensure the performance and development progress of aircraft. Aircraft structural parts have typical characteristics of multiple varieties and small batches, which is not conducive to the accumulation and reuse of knowledge. The aircraft structure is mainly composed of fuselage, landing gear, wings, tail, flight operating system, power unit, etc. Each part of the structure is composed of machined parts, sheet metal parts, profiles and other structural parts.

[0003] At present, aircraft structural parts have problems such as complicated design, low efficiency, insufficient design capabilities and knowledge accumulation. At the same time, many problems have arisen in the modeling process of aircraft structural parts, such as a lot of repetitive work, poor preservation of design knowledge, and low design efficiency.

[0004] The current patent disclosures and literature data show: 1) Patent (CN201911112287.6) A collaborative design implementation method for complex aircraft structural parts based on MBD. This method uses multi-agent technology to design and implement a collaborative workflow to complete the collaborative design of aircraft structural parts to meet the needs of fast and accurate collaborative design of complex aircraft structural parts. However, because all dimensions are not parameterized, each dimension cannot be modified, the modeling design speed is relatively slow, and the modified instances cannot be stored in the instance library. 2) Patent (CN202011262345.6) A human-machine hybrid enhanced intelligent design method based on aircraft structural parts. This method improves the design efficiency and quality of aircraft structural parts through a human-machine hybrid enhanced intelligent design method, and organically combines designers with intelligent systems to collaboratively design products. However, this method has a narrow scope of application and cannot meet the design requirements of various aircraft structural parts, and has a low utilization rate.

[0005] In summary, although the existing research results and methods can meet the intelligent design of complex aircraft structural parts to a certain extent, they have relatively high fault tolerance, limited design methods, and cannot explain new examples and retain corresponding design parameters. In the field of aerospace, there are many types of aircraft structural parts with complex structures. The utilization rate of aircraft structural parts design knowledge is low, and the design efficiency is low, which cannot meet the current high efficiency and high design quality problems in my country's aviation field. Summary of the invention

[0006] The technical problem to be solved by the present invention is: based on the problems of low utilization rate of aircraft structural parts design knowledge, low design efficiency and low design quality of aircraft structural parts by designers or systems, the present invention proposes an aircraft structural parts design method based on machine learning, which not only expands the knowledge-model-case library data, but also increases the utilization rate of aircraft structural parts design parameters, improves the design quality and design efficiency of aircraft structural parts, and enhances R&D capabilities.

[0007] The present invention achieves the above-mentioned purpose through the following technical solutions: a method for designing aircraft structural parts based on machine learning, whose technical architecture is divided into five parts, namely, a knowledge acquisition module, an OpenCv auxiliary feature recognition module, an RBR and CBR reasoning module, a machine intelligent learning system module and an example automatic transfer system, specifically:

[0008] (1) Knowledge acquisition module

[0009] The human-computer interaction interface in the knowledge acquisition module is the main window for the dialogue between the system and designers. It allows designers to input the design requirements and related parameters of the aircraft structural parts to be designed, so that the computer can obtain the required knowledge.

[0010] (2) OpenCv auxiliary feature recognition module

[0011] The OpenCv auxiliary feature recognition module acts on aircraft structural parts and performs feature analysis on structural parts composed of different aircraft wings, so that it can classify the features of typical structural parts such as plate ribs, auxiliary panels, and skin aircraft, that is, it recognizes the features of aircraft structural parts in the form of matrix and point data of the function library CAA, and uses the convolutional neural network method to perform feature analysis and processing on the generated abnormal data, so as to identify the characteristics of each aircraft structural part.

[0012] (3) RBR and CBR reasoning modules

[0013] Retrieval is performed based on the template-model-instance library. The model instance is searched and found in the model library through the rule inference engine. Similar instances are retrieved and sorted by similarity. Instances with higher similarity are selected. On the basis of this instance, the designer modifies the instance according to the requirements of the required design, that is, the corresponding parameters are determined through the parameter calculation method provided by the design template, and then the deformation design is performed on it to achieve rapid model establishment and generate solutions.

[0014] (4) Machine intelligence learning system module

[0015] The machine intelligence learning system module has a self-learning function, and will perform feature information recognition and analysis based on the solution of the RBR and CBR reasoning modules. If the revised instance and solution meet the expected design requirements, the extracted features will be classified according to the k-nearest neighbor algorithm, and added to the machine learning training library as a result of training and learning as a new instance; if the solution does not meet the expected design requirements, it will interact between designers and computers through a human-computer collaborative working mode to achieve the design requirements that need to be met and further achieve the expected results. At the same time, the instance is also added to the machine learning training library as a type of training and learning.

[0016] (5) Instance automatic transfer system

[0017] Through machine learning, new instances and corresponding design parameters are retained, and the instances in the training library are stored in the knowledge-model-instance library. As an identification method of the knowledge-model-instance library, they can be directly extracted and used in similar designs in the future. This will increase the number of instances in the knowledge-model-instance library and provide new design solutions.

[0018] Furthermore, the knowledge in the part (1) refers to the description of the problem to be designed by the designer through the human-computer interaction interface.

[0019] Furthermore, the OpenCV auxiliary feature recognition module in the part (2) performs aircraft structural component feature recognition in the form of matrix and point data of the function library CAA, and uses a convolutional neural network method to perform feature analysis and processing on the generated abnormal data, thereby identifying the characteristics of each aircraft structural component.

[0020] Furthermore, the RBR and CBR reasoning modules in the part (3) are based on the designer's description of the design problem. If there is a similar instance that meets the problem description, the instance is reused, otherwise the case is revised and the corresponding parameters are re-determined.

[0021] Furthermore, in the variant design based on the instance, the rule-based reasoning engine retrieves some similar instances, and the design parameters in the instances may not fully meet the design requirements of the designer. It is necessary to modify some design parameters based on the retrieved instances. At this time, the rule-based reasoning method is used to perform rule-based reasoning on the parameters to be modified, so that the newly modified design parameters meet the design requirements of the designer.

[0022] Furthermore, the similarity calculation in the part (3) refers to the calculation of object attribute similarity, the calculation of numerical attribute similarity, and the calculation of global similarity.

[0023] Furthermore, the machine intelligence learning system in part (4) includes an aircraft structural component feature recognition module, an algorithm analysis module and a storage module.

[0024] Furthermore, the part (4) k-nearest neighbor algorithm specifies a weight assignment for each attribute of each instance in the instance library, and then calculates the maximum weight coefficient value to further find the model instance closest to the target instance.

[0025] Furthermore, the example automatic transfer system in part (5) stores the revised example into the example library, and at the same time uses the revised example as a new design solution, and saves the revised example and the reasonable design solution as new examples into the example library, thereby updating and enriching the example library.

[0026] Compared with the prior art, the present invention has the following beneficial effects and advantages:

[0027] 1. The aircraft structural parts design method based on machine learning of the present invention increases the utilization rate of aircraft structural parts design parameters by expanding the knowledge-model-case library data, improves the design quality and design efficiency of aircraft structural parts, and enhances research and development capabilities.

[0028] 2. The present invention combines designers and knowledge-model-case libraries, giving full play to their strengths and avoiding their weaknesses to carry out product design, which greatly improves design efficiency, and the utilization rate of design parameters is significantly improved, which can greatly shorten the development cycle. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 is a flow chart of a method for designing aircraft structural parts based on machine learning according to the present invention;

[0030] Figure 2 It is a schematic diagram of the human-computer interaction interface;

[0031] Figure 3 It is a schematic diagram of the rule base for rule reasoning in an instance;

[0032] Figure 4 It is a template-model-instance library framework;

[0033] Figure 5 It is the instance retrieval process;

[0034] Figure 6 It is a flowchart of machine learning. DETAILED DESCRIPTION

[0035] The present invention is further described in detail below in conjunction with the accompanying drawings. The accompanying drawings described herein are part of this application and are used to further explain the present invention, but do not constitute a limitation of the present invention.

[0036] The present invention provides an aircraft structural component design method based on machine learning, comprising the following steps:

[0037] (1) Knowledge acquisition is to input the design requirements and related parameters of the aircraft structure to be designed through the human-computer interaction interface, some of which are as follows Figure 2 shown.

[0038] (2) The OpenCv auxiliary feature recognition part acts on the aircraft structure, performs feature analysis on the structural parts of different aircraft wings, and classifies the features of typical aircraft structural parts. That is, the aircraft structure feature recognition is performed in the form of matrix and point data of the function library CAA, and the convolutional neural network method is used to perform feature analysis and processing on the generated abnormal data, so as to identify the features of each aircraft structure.

[0039] (3) RBR and CBR reasoning is based on the retrieval of the template-model-instance library. The rule-based reasoning engine searches for model instances in the model library, retrieves similar instances and sorts the retrieved instances by similarity, selects instances with high similarity, and reuses instances if there are similar instances that meet the problem description. That is, the corresponding parameters are determined by the parameter calculation method provided by the design template, and then the deformation design is performed to achieve rapid model establishment and generate solutions. Otherwise, the case is revised and the corresponding parameters are re-determined. Figure 3 , Figure 4 shown.

[0040] (4) The machine intelligence learning system includes an aircraft structural component feature recognition module, an algorithm analysis module, and a storage module. That is, feature information recognition and analysis is performed based on the solution of the RBR and CBR reasoning modules. If the corrected instance and solution meet the expected design requirements, the extracted features will be classified according to the k-nearest neighbor algorithm and added to the machine learning training library as a new instance, such as Figure 6 shown.

[0041] (5) The automatic instance transfer system stores the revised instance into the instance library and takes the revised instance as a new design solution. The revised instance and the reasonable design solution are saved as new instances in the instance library, thereby updating and enriching the instance library.

[0042] In summary, the present invention combines machine learning and aircraft structural parts design for product design, which greatly improves the design efficiency, and the utilization rate of design parameters is significantly improved, which can greatly shorten the development cycle.

[0043] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit the same. Although the present invention has been described in detail with reference to the above embodiments, a person skilled in the art can still modify or make equivalent substitutions to the specific implementation schemes of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention are within the scope of protection of the claims of the present invention to be approved.

Claims

1. A method for designing aircraft structural parts based on machine learning, characterized in that: It combines rule selection and case reasoning, including the following parts: (1) Knowledge acquisition module: The knowledge acquisition module inputs the design requirements and related parameters of the aircraft structure to be designed through the human-computer interaction interface; (2) OpenCv auxiliary feature recognition module: The OpenCv auxiliary feature recognition module acts on aircraft structural parts and performs feature analysis on different aircraft wing structural parts, so that it can classify the features of typical aircraft structural parts such as ribs, auxiliary panels, and skins; (3) RBR and CBR reasoning modules: Based on the template-model-instance library, the model instance is searched in the model library through the rule reasoning engine, similar instances are retrieved and sorted by similarity, and the instance with the highest similarity is selected. Based on this instance, the designer modifies the instance according to the design requirements, that is, the corresponding parameters are determined through the parameter calculation method provided by the design template, and then the deformation design is performed to achieve rapid model establishment and generate solutions; (4) Machine intelligence learning system module: The machine intelligence learning system module has a self-learning function and performs feature information recognition and analysis on the solution based on the RBR and CBR reasoning modules. If the modified instance and solution meet the expected design requirements, the extracted features will be classified according to the k-nearest neighbor algorithm and added to the machine learning training library as a new instance as a result of training and learning. If the solution does not meet the expected design requirements, the human-computer collaborative working mode will be used to interact between the designer and the computer to meet the design requirements and further achieve the expected results. At the same time, the instance will also be added to the machine learning training library as a kind of training and learning. (5) Automatic instance transfer system: Through machine learning, new instances and corresponding design parameters are retained, and instances in the training library are stored in the knowledge-model-instance library as a new instance of the knowledge-model-instance library. When similar designs are performed in the future, they can be directly extracted and used. This will increase the number of instances in the knowledge-model-instance library and provide new design solutions.

2. The method for designing aircraft structural parts based on machine learning according to claim 1, characterized in that: The knowledge in part (1) refers to the description of the problem to be designed by the designer through the human-computer interaction interface.

3. The method for designing aircraft structural parts based on machine learning according to claim 1, characterized in that: The OpenCV auxiliary feature recognition module in the part (2) performs aircraft structural component feature recognition in the form of matrix and point data of the function library CAA, and uses a convolutional neural network method to perform feature analysis and processing on the generated abnormal data, thereby identifying the characteristics of each aircraft structural component.

4. The aircraft structural component design method based on machine learning according to claim 1, characterized in that: The RBR and CBR reasoning modules in part (3) are based on the designer's description of the design problem. If there is a similar instance that meets the problem description, the instance is reused, otherwise the case is revised and the corresponding parameters are re-determined.

5. The aircraft structural component design method based on machine learning according to claim 1, characterized in that: In the example-based variant design of the rule-based reasoning machine, some similar examples are retrieved, and the design parameters in the examples may not fully meet the design requirements of the designer. It is necessary to modify some design parameters based on the retrieved examples. At this time, the rule-based reasoning method is used to perform rule-based reasoning on the modified parameters, so that the newly modified design parameters meet the design requirements of the designer.

6. The method for designing aircraft structural parts based on machine learning according to claim 1, characterized in that: The similarity calculation in the above-mentioned part (3) refers to the calculation of object attribute similarity, the calculation of numerical attribute similarity and the calculation of global similarity.

7. The method for designing aircraft structural parts based on machine learning according to claim 1, characterized in that: The machine intelligence learning system in part (4) includes an aircraft structural component feature recognition module, an algorithm analysis module and a storage module.

8. The method for designing aircraft structural parts based on machine learning according to claim 1, characterized in that: The k-nearest neighbor algorithm in part (4) is to assign a weight to each attribute of each instance in the instance library, and then calculate the maximum weight coefficient value to further find the model instance closest to the target instance.

9. The aircraft structural component design method based on machine learning according to claim 1, characterized in that: The automatic instance transfer system in part (5) stores the revised instance into the instance library, and at the same time uses the revised instance as a new design solution, and saves the revised instance and the reasonable design solution as new instances into the instance library, thereby updating and enriching the instance library.

Images