A method for quickly solving structural assembly stress

Abstract

The application belongs to the field of aircraft strength design, and particularly relates to a structure assembly stress rapid solving method, which selects a specific surface as a reference surface according to structure surface point cloud data, performs point cloud matching, and obtains node coordinates in a specified coordinate system; performs distance calculation on the point cloud at a specific position before assembly and after assembly according to the node coordinates, obtains displacement of the specific position and a global position field; performs finite element simulation analysis on various assembly states of an assembly part, stores analysis result data in a result database, extracts displacement and stress data in the result database, performs neural network training, and obtains a radial basis function neural network model; inputs the displacement of the specific position into the radial basis function neural network model, performs structure stress calculation, and obtains structure assembly stress distribution prediction results. Interference caused by different 3D scanning data can be effectively coped with, point displacement can be reasonably identified, and assembly stress field calculation of the entire structure surface can be completed.

Description

Technical Field

[0001] This application belongs to the field of aircraft strength design, and specifically relates to a method for rapid calculation of structural assembly stress. Background Technology

[0002] Aircraft assembly is the final stage of aircraft manufacturing, accounting for more than half of the total workload. It involves numerous steps and demands high precision, making assembly quality control crucial for ensuring aircraft performance and structural integrity. However, due to the large number of components, their large size, high requirements, and complex forming processes, assembly gaps or forced assembly problems are unavoidable in aircraft assembly. This can lead to structural assembly stress, reducing the aircraft's fatigue life and causing stress corrosion cracking or vibration cracking, thus affecting the aircraft's strength and quality. Assembly stress control involves design, materials, processes, and management, with assembly stress detection during the assembly process being the most direct and effective measure for controlling assembly stress.

[0003] Currently, the methods for controlling assembly stress during aircraft structural assembly are relatively traditional, mainly involving on-site stress measurement combined with finite element analysis: assembly workers use techniques such as strain gauges and speckle to measure assembly stress or assembly gaps during assembly, and then feed the measurement results back to strength designers; designers establish simulation models to conduct numerical simulation analysis to analyze the impact of assembly stress on strength; and finally, the handling methods are fed back to the assembly workers for processing.

[0004] The entire assembly stress decision-making process is lengthy, time-sensitive, and relatively cumbersome, making it difficult to ensure that aircraft structural assembly stress control is implemented throughout the aircraft structural assembly process. Therefore, how to reduce the risks of using aircraft in the field and reduce hidden dangers and maintenance probability is a problem that needs to be solved. Summary of the Invention

[0005] To address the aforementioned issues, this application provides a rapid method for calculating structural assembly stress, thereby resolving the problem in the prior art that it is difficult to ensure that stress control in aircraft structural assembly is consistently implemented throughout the aircraft structural assembly process.

[0006] The technical solution of this application is: a method for rapid calculation of structural assembly stress, including:

[0007] 3D scanners are used to scan the structural geometry of aircraft parts, components and assemblies before and after assembly to obtain point cloud data of the structural surface.

[0008] Based on the point cloud data of the structural surface, a specific curved surface is selected as the reference surface, and point cloud matching is performed to obtain the node coordinates in the specified coordinate system.

[0009] Distance calculations are performed on the point cloud before and after assembly based on the node coordinates to obtain the displacement at the specific location and the global position field.

[0010] Finite element simulation analysis was performed on various assembly states of the assembly components. The analysis results were stored in a result database. Then, displacement and stress data were extracted from the result database and used to train a neural network to obtain a radial basis function neural network model.

[0011] A certain number of displacements at specific locations are input into a radial basis function neural network model to calculate structural stress and obtain the predicted results of structural assembly stress distribution.

[0012] Preferably, the specific method for point cloud matching is as follows:

[0013] The structural feature surface is selected as the reference surface. The reference surface and the point cloud data before and after assembly are read. The data is downsampled by 30% to reduce the amount of data. Then, the spatial position of the reference surface and the point cloud data is matched by the NICP algorithm to obtain the node coordinates in the specified coordinate system.

[0014] Preferably, the specific method for calculating the distance between a specific location before assembly and the point cloud after assembly is as follows:

[0015] Based on the matched node coordinates, a plane is fitted with three neighboring points of the measured point to calculate the normal vector and normal line of the measured point before assembly. In the point cloud after assembly, the point closest to the normal line is selected as the corresponding position, and the displacement at the specific position is obtained by subtracting the coordinates. After traversing all measured points, the global displacement field before and after the structure is assembled is constructed.

[0016] Preferably, the specific method for performing finite element simulation is as follows:

[0017] First, a parametric finite element model of the target assembly is established, and material properties, constraints and contact relationships are set. Finite element simulation calculations are performed to simulate typical working conditions with different bolt preload and assembly clearance. Displacement and stress field data of structural nodes under each working condition are extracted, and a results database is constructed.

[0018] Preferably, the specific method for training the neural network is as follows:

[0019] The displacement field features of the result database are used as the input vector and the stress field features are used as the output vector. The training set and the test set are divided. The Gaussian function is used as the activation function. The network weights are iteratively optimized by the least squares method to continuously correct the prediction error. After convergence, the radial basis function neural network model is obtained.

[0020] Preferably, the specific method for calculating structural stress is as follows:

[0021] The global displacement field data obtained from actual assembly is preprocessed according to the model requirements and then input into the trained radial basis function neural network model for inference calculation. The output is the global assembly stress distribution data of the structural surface, which serves as the prediction result of the structural assembly stress distribution.

[0022] The rapid stress calculation method for structural assembly in this application has the following advantages:

[0023] It can effectively cope with interference from different 3D scanning data, reasonably identify the displacement of measuring points, and complete the assembly stress field calculation of the entire structural surface. For scenarios with different bolt preloads, the calculation results can effectively identify changes in assembly stress, accurately reflect the loading of bolt preloads, and promptly detect unreasonable positions during assembly, thereby improving the quality of structural assembly and demonstrating practicality. Attached Figure Description

[0024] Figure 1 This is a schematic diagram of the overall process of this application;

[0025] Figure 2 This is a schematic diagram of the NICP algorithm flow in this application;

[0026] Figure 3 This is a schematic diagram of point cloud matching in this application;

[0027] Figure 4 This is a schematic diagram of the structural displacement field calculation method of this application;

[0028] Figure 5 This is a schematic diagram of the radial basis function neural network model of this application;

[0029] Figure 6 This is a schematic diagram of the predicted assembly stress distribution results for this application. Detailed Implementation

[0030] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions in the embodiments of this application will be described in more detail below with reference to the accompanying drawings. In the drawings, the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The described embodiments are only some, not all, of the embodiments of this application. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application. The embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0031] The first aspect of this application provides a method for rapid calculation of structural assembly stress, including the following steps:

[0032] Step S100: Use a 3D scanner to scan the structural geometry of aircraft parts, components and assemblies before and after assembly to obtain point cloud data of the structural surface.

[0033] Step S200: Based on the point cloud data of the structural surface, select a specific curved surface as the reference surface, perform point cloud matching, and obtain the node coordinates in the specified coordinate system.

[0034] Preferably, the specific method for point cloud matching is as follows:

[0035] The structural feature surface was selected as the reference surface. Point cloud data of the reference surface and before and after assembly were read. The data was downsampled by 30% to reduce the data volume. Then, the NICP (Normal Iterative Closest Point) algorithm was used. (See...) Figure 2 The reference plane and point cloud data are matched to obtain the node coordinates in the specified coordinate system. See [link to documentation]. Figure 3 .

[0036] By selecting the structural feature surface as the reference surface, the spatial reference consistency of point cloud matching is ensured, avoiding subsequent calculation errors caused by reference deviation; 30% downsampling simplifies the data volume while retaining the core geometric features of the point cloud, greatly reducing the computational load of point cloud matching and improving matching efficiency.

[0037] Step S300: Based on the node coordinates, calculate the distance between the specific position before assembly and the point cloud after assembly to obtain the displacement at the specific position and the global position field, such as... Figure 4 .

[0038] Preferably, the specific method for calculating the distance between a specific location before assembly and the point cloud after assembly is as follows:

[0039] Based on the matched node coordinates, a plane is fitted with three neighboring points of the measured point to calculate the normal vector and normal line of the measured point before assembly. In the point cloud after assembly, the point closest to the normal line is selected as the corresponding position, and the displacement at the specific position is obtained by subtracting the coordinates. After traversing all measured points, the global displacement field before and after the structure is assembled is constructed.

[0040] By fitting the plane with three neighboring points, the normal vector and normal line of the measuring point are calculated, ensuring the accuracy of the normal direction calculation and conforming to the curved surface characteristics of the aircraft structure. The point closest to the normal line in the point cloud after assembly is selected as the corresponding position, realizing accurate matching of the measuring points before and after assembly and avoiding the deviation of the measuring points caused by the deformation of the structural surface.

[0041] Step S400: Perform finite element simulation analysis on various assembly states of the assembly components, store the analysis results in a result database, then extract the displacement and stress data from the result database, train the neural network, and obtain the radial basis function neural network model, such as... Figure 5 .

[0042] Preferably, the specific method for performing finite element simulation is as follows:

[0043] First, a parametric finite element model of the target assembly is established, and material properties, constraints and contact relationships are set. Finite element simulation calculations are performed to simulate typical working conditions with different bolt preload and assembly clearance. Displacement and stress field data of structural nodes under each working condition are extracted, and a results database is constructed.

[0044] The specific methods for training neural networks are as follows:

[0045] The displacement field features of the result database are used as the input vector and the stress field features are used as the output vector. The training set and the test set are divided. The Gaussian function is used as the activation function. The network weights are iteratively optimized by the least squares method to continuously correct the prediction error. After convergence, the radial basis function neural network model is obtained.

[0046] By establishing a parametric finite element model, model parameters can be flexibly adjusted to adapt to different assembly conditions, improving the efficiency and versatility of simulation modeling; by setting a vector with displacement field as input and stress field as output, accurate mapping from displacement field to stress field can be achieved, meeting the actual needs of assembly stress calculation; dividing the training set and test set ensures the effectiveness and generalization ability of model training, avoiding overfitting.

[0047] Step S500: Input a certain number of displacements at specific locations into the radial basis function neural network model to calculate the structural stress and obtain the predicted results of the structural assembly stress distribution, such as... Figure 6 .

[0048] Preferably, the specific method for calculating structural stress is as follows:

[0049] The global displacement field data obtained from actual assembly is preprocessed according to the model requirements and then input into the trained radial basis function neural network model for inference calculation. The output is the global assembly stress distribution data of the structural surface, which serves as the prediction result of the structural assembly stress distribution.

[0050] By preprocessing the actual assembly's global displacement field data according to the model requirements before inputting it into the model, the consistency between the input data and the model training samples is ensured, thus avoiding prediction errors caused by data format deviations.

[0051] In summary, this application has the following advantages:

[0052] This integrated process—acquiring point cloud data through 3D scanning, obtaining node coordinates through point cloud matching, calculating displacement and global displacement field, combining finite element simulation with neural network training and modeling, and solving for stress distribution by inputting displacement data—breaks through the cumbersome traditional model of combining on-site measurement with finite element analysis. It achieves rapid, end-to-end calculation of aircraft structural assembly stress, from deformation data acquisition to stress distribution prediction. It eliminates the need for repeated manual data transfer, significantly shortening the overall cycle of assembly stress analysis, improving calculation efficiency, and increasing the number of aircraft structures whose assembly stress can be monitored within the same timeframe. Simultaneously, it enables direct and accurate capture of assembly stress, providing complete data support for stress control during aircraft structural assembly, ensuring high-quality assembly of aircraft structures overall, and reducing risks associated with field use.

[0053] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for rapid calculation of structural assembly stress, characterized in that, include: 3D scanners are used to scan the structural geometry of aircraft parts, components and assemblies before and after assembly to obtain point cloud data of the structural surface. Based on the point cloud data of the structural surface, a specific curved surface is selected as the reference surface, and point cloud matching is performed to obtain the node coordinates in the specified coordinate system. Distance calculations are performed on the point cloud before and after assembly based on the node coordinates to obtain the displacement at the specific location and the global position field. Finite element simulation analysis was performed on various assembly states of the assembly components. The analysis results were stored in a result database. Then, displacement and stress data were extracted from the result database and used to train a neural network to obtain a radial basis function neural network model. A certain number of displacements at specific locations are input into a radial basis function neural network model to calculate structural stress and obtain the predicted results of structural assembly stress distribution.

2. The method for rapid calculation of structural assembly stress as described in claim 1, characterized in that, The specific method for point cloud matching is as follows: The structural feature surface is selected as the reference surface. The reference surface and the point cloud data before and after assembly are read. The data is downsampled by 30% to reduce the amount of data. Then, the spatial position of the reference surface and the point cloud data is matched by the NICP algorithm to obtain the node coordinates in the specified coordinate system.

3. The method for rapid calculation of structural assembly stress as described in claim 2, characterized in that, The specific method for calculating the distance between a specific location before assembly and the point cloud after assembly is as follows: Based on the matched node coordinates, a plane is fitted with three neighboring points of the measured point to calculate the normal vector and normal line of the measured point before assembly. In the point cloud after assembly, the point closest to the normal line is selected as the corresponding position, and the displacement at the specific position is obtained by subtracting the coordinates. After traversing all measured points, the global displacement field before and after the structure is assembled is constructed.

4. The method for rapid calculation of structural assembly stress as described in claim 3, characterized in that, The specific method for performing finite element simulation is as follows: First, a parametric finite element model of the target assembly is established, and material properties, constraints and contact relationships are set. Finite element simulation calculations are performed to simulate typical working conditions with different bolt preload and assembly clearance. Displacement and stress field data of structural nodes under each working condition are extracted, and a results database is constructed.

5. The method for rapid calculation of structural assembly stress as described in claim 4, characterized in that, The specific methods for training neural networks are as follows: The displacement field features of the result database are used as the input vector and the stress field features are used as the output vector. The training set and the test set are divided. The Gaussian function is used as the activation function. The network weights are iteratively optimized by the least squares method to continuously correct the prediction error. After convergence, the radial basis function neural network model is obtained.

6. The method for rapid calculation of structural assembly stress as described in claim 3, characterized in that, The specific method for calculating structural stress is as follows: The global displacement field data obtained from actual assembly is preprocessed according to the model requirements and then input into the trained radial basis function neural network model for inference calculation. The output is the global assembly stress distribution data of the structural surface, which serves as the prediction result of the structural assembly stress distribution.

Images