A fast simulation method for dent resistance analysis of vehicle body

Abstract

The application relates to a rapid simulation method for analyzing the dent resistance of a vehicle body, and belongs to the technical field of automobile structure resistance analysis. The method comprises the following steps: obtaining a CAE simulation model according to a CAD geometric model of a vehicle roof; obtaining a high-fidelity data set after adjusting the parameter values of key geometric features and performing simulation analysis through the CAE simulation model; constructing a prediction model and training the prediction model according to the high-fidelity data set to obtain a rapid simulation model; and obtaining a target dynamic displacement field through the rapid simulation model according to the parameter values of a set of key geometric features of the latest design. The method trains the prediction model by using the high-fidelity data set and obtains the rapid simulation model. When the parameter values of the key geometric features of the latest design are input into the rapid simulation model, the dynamic displacement field can be obtained, the calculation time is effectively reduced, the analysis efficiency is improved, and the product development efficiency is improved.

Description

Technical Field

[0001] This invention belongs to the field of automotive structural resistance analysis technology, specifically relating to a rapid simulation method for vehicle body anti-dent analysis. Background Technology

[0002] Analysis of the dent resistance of the engine cover is a key part of the design process. Among the many factors that enhance automotive R&D capabilities, Computer-Aided Engineering (CAE) is an indispensable numerical calculation tool for product structural performance analysis and optimization design. It mainly uses the finite element method for simulation analysis, and can analyze the structural mechanical properties of complex products through computer-aided solutions. By using CAE simulation to analyze the mechanical properties of the engine cover, and evaluating its mechanical characteristics based on its displacement under dynamic and static loads, it plays a vital role in the evaluation of the engine cover's dent resistance and its structural design.

[0003] The basic process of CAE simulation analysis is as follows: the CAD geometric model of the hood is proportionally generated into a CAE finite element model. The finite element model discretizes the continuous structure into a grid distribution. The grid is a combination of quadrilaterals and triangles. Theoretically, the smaller the size of the grid, the more numerous the elements, the denser the distribution, and the higher the accuracy, but the higher the computational resources required. After dividing the finite element mesh model, load conditions are applied to the model according to the actual stress conditions. All degrees of freedom of the hood and the vehicle body mounting points are constrained, and dynamic loads perpendicular to the surface are applied to the outer surface of the hood. Through CAE analysis, the displacement values ​​at different positions of the hood, i.e., the deformation, can be obtained. If the maximum deformation is less than the required fixed value, the hood's anti-dent performance is considered to meet the analysis requirements.

[0004] Based on the differences in their specific structural shapes, hairnets can be broken down into several key geometric features. These key geometric features directly affect their mechanical properties. Therefore, to obtain the optimal geometric structure, it is necessary to use rapid calculation tools to analyze the mechanical properties corresponding to different geometric feature parameters, and then optimize the dimensional parameters based on these mechanical properties to obtain the optimal geometric parameters. However, to ensure analysis accuracy, the mesh size is usually very small, especially when dealing with multi-parameter value groups. If existing commercial finite element software is used, each parameter change requires re-meshing, and CAE analysis needs to be performed again for each parameter change. From re-meshing to performing CAE analysis to finally obtaining the mechanical properties corresponding to the geometric feature parameters, this process requires a lot of computational resources and time for CAE analysis. To obtain the optimal geometric parameters, multiple adjustments are needed, which is extremely time-consuming and leads to low product development efficiency. Summary of the Invention

[0005] To address this issue, the present invention provides a rapid simulation method for vehicle body dent resistance analysis, thereby solving the problem that obtaining optimal geometric parameters using existing commercial finite element software requires a huge amount of time, resulting in low product development efficiency.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: In a first aspect, the present invention provides a rapid simulation method for vehicle body dent resistance analysis, comprising: A CAE simulation model is obtained based on the CAD geometric model of the vehicle's hood. After adjusting the parameter values ​​of key geometric features and performing simulation analysis through the CAE simulation model, a high-fidelity dataset is obtained. The key geometric features include the thickness of the plate material in four regions near the edge of the hair cover, the thickness of the reinforcing rib in the middle region of the hair cover, the width of the reinforcing rib in the middle region of the hair cover, and the width of the reinforcing plate inside the hair cover. The high-fidelity dataset includes multiple high-fidelity data points, each of which includes the parameter values ​​of the key geometric features, the coordinate values ​​of all nodes in the CAE simulation model, and the dynamic displacement field. The dynamic displacement field includes the displacement values ​​of each node at a preset number of time points. A prediction model is constructed and trained based on the high-fidelity dataset to obtain a fast simulation model; The target dynamic displacement field is obtained by using the rapid simulation model based on the parameter values ​​of a set of key geometric features designed in the latest version.

[0007] Furthermore, the high-fidelity dataset obtained after adjusting the parameter values ​​of the key geometric features and performing simulation analysis through the CAE simulation model includes: Multiple sets of parameter values ​​are obtained by acquiring the parameter values ​​of key geometric features through various methods; each set of parameter values ​​includes the thickness value of the plate in four areas around the hair cover near the edge, the thickness value of the reinforcing rib in the middle area of ​​the hair cover, the width value of the reinforcing rib in the middle area of ​​the hair cover, and the width value of the reinforcing plate inside the hair cover. After performing simulation analysis on each of the multiple sets of parameter values, multiple high-fidelity data points are obtained as a high-fidelity dataset, with each set of parameter values ​​corresponding to one high-fidelity data point.

[0008] Furthermore, the simulation analysis operation includes: The first set of parameter values ​​and the preset time load data are input into the CAE simulation model for simulation analysis to obtain the corresponding dynamic displacement field; the CAE simulation model includes the coordinate values ​​of all nodes; The high-fidelity data corresponding to the first set of parameter values ​​is obtained by associating and storing the coordinate values ​​of all nodes in the CAE simulation model with the dynamic displacement field corresponding to the first set of parameter values.

[0009] Furthermore, the construction of the prediction model includes: A prediction model is constructed based on a deep operator network. The prediction model includes two feedforward neural networks, namely a branch network and a backbone network. The branch network is used to encode the discrete input function space, and the backbone network is used to encode the output function domain.

[0010] Furthermore, the step of training the prediction model based on the high-fidelity dataset to obtain a fast simulation model includes: The high-fidelity dataset is divided into a training dataset and a test dataset according to a preset ratio; The prediction model is trained and iterated based on the training dataset; A fast simulation model is obtained by testing the prediction model on the test dataset.

[0011] Furthermore, the model training and model testing include: The model training process is as follows: the parameter values ​​of the corresponding key geometric features in the high-fidelity data and the preset time load data are used as the input of the branch network of the prediction model, and the coordinate values ​​of all nodes in the CAE simulation model are used as the input of the backbone network of the prediction model. The prediction model will be trained and iterated based on the corresponding dynamic displacement field output and the dynamic displacement field in the high-fidelity data. The model testing process is as follows: the parameter values ​​of the corresponding key geometric features in the high-fidelity data and the preset time load data are used as inputs to the prediction model. The prediction model will output the corresponding dynamic displacement field. The prediction accuracy is obtained based on the dynamic displacement field in the high-fidelity data and the dynamic displacement field output by the prediction model. If the prediction accuracy is greater than or equal to the preset prediction accuracy, the model test is completed and the tested prediction model is used as the rapid simulation model.

[0012] Furthermore, the target dynamic displacement field is obtained by passing the parameter values ​​of a set of key geometric features based on the latest design through the rapid simulation model, including: The branch network of the rapid simulation model is input with the parameter values ​​of a set of key geometric features of the latest design and the preset time load data. The coordinate values ​​of all nodes in the CAE simulation model are used as the backbone network input of the rapid simulation model. The rapid simulation model will output the target dynamic displacement field, which includes the displacement value of each node in the CAE simulation model at a preset number of time points.

[0013] Furthermore, the method of obtaining parameter values ​​for key geometric features through multiple means yields multiple sets of parameter values, including: The specific steps for obtaining the parameter values ​​of key geometric features using the single-feature independent change method are as follows: within the design range of each key geometric feature, select the corresponding preset typical values ​​according to engineering experience, and set other key geometric features as benchmark values. The benchmark values ​​are reliable values ​​of the corresponding geometric features obtained through multiple experiments. The specific steps for obtaining the parameter values ​​of key geometric features using the multi-feature collaborative transformation method are as follows: identify key geometric features with physical coupling relationships, generate the parameter values ​​of key geometric features with physical coupling relationships in the form of orthogonal combination or uniform combination, and set other key geometric features as reference values. The specific steps for obtaining the parameter values ​​of key geometric features using a random generation method are as follows: take the parameter values ​​of key geometric features obtained by using a single feature independent change method or a multi-feature collaborative change method as known parameter values, and randomly generate the parameter values ​​of key geometric features among the known parameter values ​​using a uniform random method.

[0014] Furthermore, the CAE simulation model obtained based on the CAD geometric model of the vehicle's hood includes: The CAE finite element model generated based on the CAD geometric model of the vehicle's hood is used as the CAE simulation model; The CAE simulation model is subjected to initial simulation analysis based on the preset time load data to obtain the evaluation results; Based on the evaluation results, determine whether the CAE simulation model meets the standards.

[0015] Furthermore, the method also includes: The maximum displacement value within the preset circular region at the center of the hood is obtained based on the target dynamic displacement field; If the maximum displacement value is less than the preset displacement value, then the parameter values ​​of the latest set of key geometric features meet the standard.

[0016] The present invention, by adopting the above technical solution, has at least the following beneficial effects: This paper presents a rapid simulation method for vehicle body anti-dentition analysis. Based on the CAD geometric model of the vehicle's hood, a CAE simulation model is obtained. After adjusting the parameter values ​​of key geometric features and performing simulation analysis through the CAE model, a high-fidelity dataset is obtained. A prediction model is constructed and trained using the high-fidelity dataset to obtain a rapid simulation model. Finally, based on the parameter values ​​of a newly designed set of key geometric features, the target dynamic displacement field is obtained through the rapid simulation model. This method initially adjusts the parameter values ​​of key geometric features and uses CAE to analyze these values ​​to obtain a high-fidelity dataset. The prediction model is then trained using this high-fidelity dataset to obtain the rapid simulation model. Subsequently, with newly designed key geometric feature parameter values, CAE analysis is no longer required; the dynamic displacement field can be directly input into the rapid simulation model, effectively reducing computation time, improving analysis efficiency, and consequently, enhancing product development efficiency.

[0017] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit the invention. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a flowchart illustrating a rapid simulation method for vehicle body anti-dent analysis, as shown in an exemplary embodiment of the present invention. Figure 2 This is a schematic diagram illustrating a hair cover structure according to an exemplary embodiment of the present invention; Figure 3 This is a schematic diagram illustrating the time-load-geometric feature matrix according to an exemplary embodiment of the present invention; Figure 4 This is a schematic diagram of the architecture of a prediction model shown in an exemplary embodiment of the present invention; Figure 5 This is an exemplary embodiment of the present invention illustrating the prediction results of a set of key geometric feature parameter values ​​using a fast simulation model.

[0020] The present invention will be further described below with reference to the accompanying drawings and specific embodiments. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be described in detail below. Obviously, the described embodiments are merely some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other implementation methods obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0022] Existing prediction models are inadequate: their predictive capabilities are limited, confined to "point prediction." Existing prediction models based on shallow neural networks (such as Response Surface Modeling (RSM), Kriging, etc.) can only predict single-point responses or a small number of features, such as predicting scalars like the maximum indentation depth, and cannot provide a continuous and complete dynamic displacement field.

[0023] This invention provides a rapid simulation method for vehicle body anti-dent analysis. By adjusting the parameter values ​​of key geometric features and analyzing the parameter values ​​using CAE, a high-fidelity dataset is obtained. The prediction model is then trained using the high-fidelity dataset to obtain a rapid simulation model. In the future, if there are parameter values ​​of newly designed key geometric features, there is no need to use CAE analysis; the dynamic displacement field can be directly input into the rapid simulation model, effectively reducing computation time and improving analysis efficiency.

[0024] The method of the present invention will be described below through specific embodiments.

[0025] Please see Figure 1 , Figure 1 This is a flowchart illustrating a rapid simulation method for vehicle body dent resistance analysis, as shown in an exemplary embodiment of the present invention. See also... Figure 1 The method includes: Step S11: Obtain the CAE simulation model based on the CAD geometric model of the vehicle's hood; Step S12: Adjust the parameter values ​​of key geometric features and perform simulation analysis through a CAE simulation model to obtain a high-fidelity dataset; Step S13: Construct a prediction model and train the prediction model based on a high-fidelity dataset to obtain a fast simulation model; Step S14: Based on the parameter values ​​of a set of key geometric features in the latest design, the target dynamic displacement field is obtained through a rapid simulation model.

[0026] It should be noted that the technical solution provided in this embodiment is applicable to scenarios including but not limited to: dent resistance analysis of automotive hoods.

[0027] It should be noted that the high-fidelity dataset includes multiple high-fidelity data points. Each high-fidelity data point includes parameter values ​​of key geometric features, coordinate values ​​of all nodes in the CAE simulation model, and a dynamic displacement field. The dynamic displacement field includes the displacement values ​​of each node at a preset number of time points. The preset number can be set according to specific business requirements. The default preset number is 80. When displayed dynamically, it represents the displacement values ​​of each node at 80 time points. After being displayed in chronological order, an 80-frame dynamic displacement map is obtained.

[0028] It is understood that the method provided in this embodiment obtains a high-fidelity dataset by adjusting the parameter values ​​of key geometric features and analyzing the parameter values ​​using CAE. The high-fidelity dataset is then used to train the prediction model and obtain a fast simulation model. In the future, if there are parameter values ​​of newly designed key geometric features, there is no need to use CAE analysis. The dynamic displacement field can be obtained by directly inputting them into the fast simulation model, which effectively reduces the calculation time, improves the analysis efficiency, and thus improves the product development efficiency.

[0029] In practical application, please refer to Figure 2 , Figure 2 This is a schematic diagram of a hairnet structure illustrating an exemplary embodiment of the present invention. See also: Figure 2 The key geometric features include areas 1, 2, 3, 4, 5, 6, and 7 near the edges of the hair cover. The thickness of the plate in areas 1, 2, 3, and 4 are four key geometric features, corresponding to different areas in front, back, left, and right of the hair cover, respectively. They are considered independently because each area has different stress and support conditions. Areas 5 and 6 are the reinforcing ribs in the middle area of ​​the hair cover. The thickness of the reinforcing ribs in the middle area of ​​the hair cover is a key geometric feature, as is the width of the reinforcing ribs in the middle area of ​​the hair cover. Area 7 is the internal reinforcing plate of the hair cover. The width of the internal reinforcing plate of the hair cover is a key geometric feature.

[0030] It should be noted that the seven key geometric features cover core design variables such as the thickness distribution of the hood's skin and the dimensions of the internal reinforcing structure. The value of each key geometric feature is set as an independent design parameter. Each parameter has its corresponding design range due to space layout and process limitations on the vehicle. By combining the parameter values ​​of different key geometric features, any kind of hood geometry can be fully characterized.

[0031] Specifically, the parameter value design ranges for the seven key geometric features are shown in Table 1 below. Table 1 is an explanation of the variables for the seven key geometric features of the hood.

[0032] Table 1 In practice, step S11, "obtaining a CAE simulation model based on the CAD geometric model of the vehicle's hood," includes: using the CAE finite element model generated based on the CAD geometric model of the vehicle's hood as the CAE simulation model; performing an initial simulation analysis on the CAE simulation model based on preset time load data to obtain evaluation results; and determining whether the CAE simulation model meets the standards based on the evaluation results.

[0033] It should be noted that the preset time load data can specifically be a time-load diagram, in which dynamic impact loads are applied at different positions on the outer surface of the hood and the time is recorded. The preset time load data can be set according to specific business needs. Select the CAD model of the hood of a certain vehicle under development, use existing CAE technology to generate the corresponding CAE finite element model, set the preset time load data, that is, apply dynamic impact loads at different positions on the outer surface of the hood to simulate the situation where the hood is impacted by small stones on the road surface during vehicle driving, and perform initial simulation analysis to obtain the evaluation result. The evaluation result is the maximum displacement value of the 100 mm circular area at the center of the hood. If the maximum displacement of the 100 mm circular area at the center of the hood is less than or equal to 6 mm, it is considered that the dent resistance is qualified, that is, the CAE simulation model meets the standard and can be used.

[0034] In practice, step S12, "adjusting the parameter values ​​of key geometric features and obtaining a high-fidelity dataset through simulation analysis using a CAE simulation model," includes: obtaining multiple sets of parameter values ​​by acquiring the parameter values ​​of key geometric features through various methods; each set of parameter values ​​includes the thickness values ​​of the plates in the four regions around the hair cover near the edge, the thickness values ​​of the reinforcing ribs in the middle region of the hair cover, the width values ​​of the reinforcing ribs in the middle region of the hair cover, and the width values ​​of the reinforcing plates inside the hair cover; and performing simulation analysis on each set of parameter values ​​to obtain multiple high-fidelity data as a high-fidelity dataset, with each set of parameter values ​​corresponding to one high-fidelity data.

[0035] Specifically, multiple sets of parameter values ​​are obtained by acquiring the parameter values ​​of key geometric features through various methods, including: The specific steps for obtaining the parameter values ​​of key geometric features using a single-feature independent variation method are as follows: within the design range of each key geometric feature, corresponding preset typical values ​​are selected according to engineering experience, and other key geometric features are set as benchmark values. The benchmark values ​​are reliable values ​​obtained from multiple experiments for the corresponding geometric features. The specific steps for obtaining the parameter values ​​of key geometric features using a multi-feature collaborative variation method are as follows: key geometric features with physical coupling relationships are identified, and parameter values ​​of key geometric features with physical coupling relationships are generated according to orthogonal combination or uniform combination, with other key geometric features set as benchmark values. The specific steps for obtaining the parameter values ​​of key geometric features using a random generation method are as follows: the parameter values ​​of key geometric features obtained using the single-feature independent variation method or the multi-feature collaborative variation method are used as known parameter values, and parameter values ​​of key geometric features are randomly generated among the known parameter values ​​using a uniform random method.

[0036] It should be noted that the single-feature independent variation method for obtaining parameter values ​​specifically involves: within the parameter design range of each key geometric feature, selecting 5 to 10 typical values ​​based on engineering experience, such as minimum, maximum, median, and several key intermediate values, while keeping the parameter values ​​of other key geometric features as baseline values, and generating parameter value sets for analyzing the sensitivity of a single feature; approximately 50 sets can be set. The multi-feature collaborative variation method for obtaining parameter values ​​specifically involves: identifying feature pairs with physical coupling relationships, for example, rib thickness and rib width as feature pairs, and plate thickness in adjacent areas as feature pairs, and generating parameter values ​​for the collaborative variation of feature pairs using orthogonal or uniform combinations; for example, rib thickness and width are often adjusted simultaneously to maintain the section modulus; approximately 150 such parameter value sets can be set. The random generation method for parameter values ​​specifically involves: based on the aforementioned structured parameter value sets, to fully cover areas within the design space not covered by system sampling, supplementing with a certain number (approximately 240) of random parameter value sets. The parameter values ​​of these parameter value sets are uniformly and randomly distributed within the design range, but do not rely on specific sampling algorithms such as Latin hypercube; instead, they are generated using simple uniform random generation.

[0037] It is understood that the method provided in this embodiment differs from traditional methods that blindly generate parameter value groups using statistical sampling (such as Latin hypercube, full factorial design, etc.). This invention proposes a parameter value group space design method based on the engineering significance of geometric features. This method makes full use of the understanding of the anti-concavity mechanism of the hair mask to perform a structured and systematic parameter value group combination design for 7 geometric features, ensuring that the parameter value group space can not only cover the design range, but also reflect the physical synergistic relationship between features.

[0038] Specifically, the simulation analysis operation includes: inputting the first set of parameter values ​​and preset time load data into the CAE simulation model for simulation analysis to obtain the corresponding dynamic displacement field; the CAE simulation model includes the coordinate values ​​of all nodes; and associating and storing the first set of parameter values, the coordinate values ​​of all nodes in the CAE simulation model, and the dynamic displacement field corresponding to the first set of parameter values ​​to obtain high-fidelity data corresponding to the first set of parameter values.

[0039] It should be noted that CAE simulation analysis was performed on each set of parameter values ​​one by one. Each set of parameter values ​​was submitted to commercial CAE software (such as Abaqus / LS-DYNA) for calculation. The same mesh generation standard (no less than 139,372 nodes), material parameters, load conditions (dynamic impact load, output 80 time steps) and boundary conditions were used as the CAE simulation model used for initial verification.

[0040] It should be noted that after the CAE simulation is completed, the following data is extracted from each set of parameter values. The input part includes: the specific values ​​of the 7 geometric feature parameters and the preset time load data, i.e., the time load curve; the spatial coordinate part includes: the X / Y / Z coordinate values ​​of all nodes in the finite element model; the output part includes: the X / Y / Z displacement values ​​of each node in the 80 time frames; the parameter values, the coordinate values ​​of all nodes in the corresponding CAE simulation model, and the dynamic displacement field corresponding to each set of parameter values ​​are associated and stored, and finally multiple sets of high-fidelity data are obtained and stored as a dataset.

[0041] In practice, step S13, “constructing a prediction model”, includes: constructing a prediction model based on a deep operator network. The prediction model includes two feedforward neural networks, namely a branch network and a backbone network. The branch network is used to encode the discrete input function space, and the backbone network is used to encode the output function domain.

[0042] It should be noted that in constructing the improved DeepONet proxy model, the inputs to the branch networks are the parameter values ​​of 7-dimensional key geometric features and preset time load data, which are generally time-load curves, while the inputs to the backbone network are node coordinates.

[0043] In practice, step S13, "training the prediction model based on the high-fidelity dataset to obtain a fast simulation model", includes: dividing the high-fidelity dataset into a training dataset and a test dataset according to a preset ratio; training and iterating the prediction model based on the training dataset; and testing the prediction model based on the test dataset to obtain a fast simulation model.

[0044] It should be noted that the preset ratio of the training dataset to the test dataset is 4:1, and they are randomly drawn from the high-fidelity dataset. The training dataset is used for training and iterating the prediction model parameters, and the test dataset is used to test the prediction accuracy of the prediction model.

[0045] It should be noted that, in order to make the model training more stable and improve the convergence speed and model accuracy during training, the high-fidelity data in each dataset is first normalized.

[0046] It should be noted that the prediction model includes two feedforward neural networks: one that encodes the discrete input function space (i.e., the branch net), and the other that encodes the output function domain (i.e., the trunk net). The branch net represents the number of discrete output functions required, almost to an infinite power, to generalize to any solution. Its function can be expressed as: Trunk net: Represents the conditions that need to be met for each function, such as initial conditions and boundary conditions. Its function can be represented as: The nonlinear operator form can be approximated by the dot product of the two networks mentioned above, and can be expressed as: , The loss function can be expressed as: , in the above formula, This represents the input function, used to characterize structural geometric parameters, dynamic load information, or operating condition input information; This represents the set of discrete sampling points of the input function; This indicates the number of sampling points for the input function; Indicates the input function In the set of sampling points Discrete values ​​on. Represents a branch network. This represents the feature vector output by the branch network; and These represent the weight matrix and bias vector of the branch network, respectively. Indicates the backbone network. This represents the evaluation point of the output function, i.e., the spatial or spatiotemporal location corresponding to the displacement value to be predicted; This represents the feature vector output by the backbone network; and These represent the weight matrix and bias vector of the backbone network, respectively. This represents a non-linear activation function. Represents a real nonlinear operator. This represents the approximate operator learned by DeepONet; The input function is: At the evaluation point The actual displacement response at that location, This represents the displacement response predicted by the model. The dimension representing the output features of the branch network and the backbone network. and These represent the first and second eigenvectors of the corresponding feature vectors. Each component. This represents the operator learning loss function. This indicates the number of training samples or the number of input functions. This indicates the number of output evaluation points in each sample. Indicates the first One input sample, Indicates the first The first sample Each output evaluation point. Minimize the predicted displacement value. Compared with the actual displacement value The mean square error between them enables rapid prediction of the displacement field of the cover under dynamic load.

[0047] In practical application, please refer to Figure 3 , Figure 3 This is a schematic diagram of the time-load-geometric feature matrix shown in an exemplary embodiment of the present invention. See also: Figure 3 To predict the displacement field of the hood under continuous dynamic load, this project extracts the time and load information at all time steps and matches it with the parameter value groups of the corresponding key geometric features to obtain the time-load-geometric feature matrix. The geometric feature matrix needs to be expanded from (n,7) to (n×j,7) according to the number of calculation time steps j, and then combined with the time and load matrix (n×j,2) to form the time-load-geometric feature matrix (n×j,9). The time-load-geometric feature matrix is ​​used as the input of the prediction model and the fast simulation model.

[0048] In practical application, please refer to Figure 4 , Figure 4 This is a schematic diagram of the architecture of a prediction model illustrated in an exemplary embodiment of the present invention. See also: Figure 4 In the prediction model architecture, parameter value groups are extracted from the total sample in batches for network training based on the batch size (batch size or batch size, which in deep learning refers to the number of parameter value groups input during a single iteration of training and is a key hyperparameter affecting model training).

[0049] Specifically, model training and model testing include: The model training process involves using the parameter values ​​of the corresponding key geometric features in the high-fidelity data and the preset time load data as inputs to the branch network of the prediction model, and using the coordinate values ​​of all nodes in the CAE simulation model as inputs to the backbone network of the prediction model. The prediction model will be trained and iterated based on the output dynamic displacement field and the dynamic displacement field in the high-fidelity data. The model testing process involves using the parameter values ​​of the corresponding key geometric features in the high-fidelity data and the preset time load data as inputs to the prediction model. The prediction model will output the corresponding dynamic displacement field. The prediction accuracy is obtained based on the dynamic displacement field in the high-fidelity data and the dynamic displacement field output by the prediction model. If the prediction accuracy is greater than or equal to the preset prediction accuracy, the model testing is complete, and the tested prediction model is used as the rapid simulation model.

[0050] It should be noted that the parameter values ​​of key geometric features and the preset time load number are converted into a time-load-geometric feature matrix, which is used as the input of the branch network of the prediction model. The prediction model outputs the prediction accuracy at a maximum displacement moment based on the dynamic displacement field in the high-fidelity data and the dynamic displacement field output by the prediction model. The preset prediction accuracy can be set according to specific business needs, with a default value of 0.9946. When the prediction accuracy is greater than 0.9946, the height model prediction result is close to the CAE analysis result.

[0051] In practice, step S14, "obtaining the target dynamic displacement field based on the parameter values ​​of a set of key geometric features in the latest design through a rapid simulation model," includes: inputting the parameter values ​​of a set of key geometric features in the latest design and the preset time load data into the branch network of the rapid simulation model; using the coordinate values ​​of all nodes in the CAE simulation model as the backbone network input of the rapid simulation model; and outputting the target dynamic displacement field, which includes the displacement value of each node in the CAE simulation model at a preset number of time points.

[0052] It should be noted that the parameter values ​​of the latest set of key geometric features and the preset time load data are converted into a time-load-geometric feature matrix, which is used as the input of the branch network of the rapid simulation model. The coordinate values ​​of all nodes in the CAE simulation model are used as the input of the backbone network of the rapid simulation model. The rapid simulation model will then output the target dynamic displacement field. After multiple tests, this process is basically completed within 6 seconds. Compared with the traditional CAE calculation process: CAD parameter change → CAE mesh re-division (10–30 min) → nonlinear explicit solution (6–8 h) → post-processing to view deformation results and evaluate anti-dip performance (30 min), it is much faster. Compared with the traditional anti-dip performance analysis that relies on commercial CAE software such as Abaqus and LS-DYNA, the calculation time of a single CAE simulation analysis is 390 s. A single 390 s is only the time of a single-core CPU, which also improves the analysis speed.

[0053] It is understood that the method provided in this embodiment has parameter values ​​of key geometric features of the new design. Without the need for CAE analysis, the dynamic displacement field can be obtained by directly inputting them into the fast simulation model, which effectively reduces the calculation time and improves the analysis efficiency.

[0054] In practice, this also includes: obtaining the maximum displacement value within the preset circular region of the center of the hood based on the target dynamic displacement field; if the maximum displacement value is less than the preset displacement value, then the parameter values ​​of the latest set of key geometric features meet the standards.

[0055] It should be noted that the preset circular area and preset displacement value can be set according to specific business requirements. The preset circular area is generally 100mm by default, and the preset displacement value is generally 6mm by default. When the maximum displacement value is less than 6mm, the concavity resistance is considered to be qualified, that is, the parameter values ​​of the latest set of key geometric features meet the standards.

[0056] In one specific embodiment, a set of designed key geometric feature parameter values ​​are used as input to a rapid simulation model. The rapid simulation model outputs a target dynamic displacement field. Simultaneously, the dynamic displacement field is obtained by analyzing the set of key geometric feature parameter values ​​using existing commercial CAE software. It can be understood that the dynamic displacement field obtained by CAE software analysis is close to the real dynamic displacement field, which is used to verify the prediction effect of the rapid simulation model.

[0057] Specifically, please refer to Figure 5 , Figure 5 This is an exemplary embodiment of the present invention illustrating the prediction results of a set of key geometric feature parameter values ​​using a fast simulation model. See [link to example]. Figure 5The first column shows the frame images of the dynamic displacement field from time t0 to time t17 in the X direction obtained after analysis by commercial CAE software. It represents the displacement value of each node in the X direction from time t0 to time t17 in the dynamic displacement field. At time t17, there is a node with the largest dynamic displacement value in the X direction. Blue indicates that the displacement value is relatively small, green indicates that the displacement value is average, and red indicates that the displacement value is relatively large. True_Dis_x_case1_T0…True_Dis_x_case1_T17 are the displacement field frame images from time t0 to time t17, respectively.

[0058] See Figure 5 The second column shows the frame images of the target dynamic displacement field from time t0 to time t17 in the X direction output by the fast simulation model. It represents the displacement value of each node in the target dynamic displacement field from time t0 to time t17 in the X direction. At time t17, there is a node with the largest dynamic displacement value in the X direction. Blue indicates that the displacement value is relatively small, green indicates that the displacement value is average, and red indicates that the displacement value is relatively large. Pred_Dis_x_case1_T0…Pred_Dis_x_case1_T17 are the displacement field frame images from time t0 to time t17, respectively.

[0059] See Figure 5 The third column represents the absolute error of the displacement in the X direction from time t0 to time t17 in the first two columns. Ae_Dis_x_case1_T0…Ae_Dis_x_case1_T17 are the absolute error values ​​of the displacement in the X direction from time t0 to time t17 in the first two columns, respectively. Blue indicates the same absolute error value, green indicates a very small absolute error value, and red indicates a relatively large absolute error value. R² is the prediction accuracy value, which is obtained based on the absolute error value using existing algorithms, specifically using the formula… The prediction accuracy value is calculated, where R is in the formula. 2 This represents the prediction accuracy value. These are the actual values; in this method, the values ​​are obtained from analysis using commercial CAE software. These are the model prediction values; in this method, they are the values ​​obtained using a fast simulation model. This is the average of the true values. In this method, it refers to the average of the values ​​obtained from analysis using commercial CAE software, specifically the average of the displacement values ​​from t0 to t17. The sum of squared prediction errors, also known as SSE; The total fluctuation of the true value relative to the mean is also called SST; when R 2 When R = 1, the prediction is completely accurate; 2 When R = 0, the model performance is similar to "predicting all values ​​as the average", R 2When R < 0, the model is worse than predicting using the average value directly, indicating a very poor prediction; 2 The closer the result is to 1, the better the model's prediction.

[0060] It should be noted that the prediction accuracy R2 of the displacement field along the X direction gradually increases with the number of time frames, and the prediction accuracy at the maximum displacement time T17 reaches 0.9946. This indicates that the dynamic displacement field output by the fast simulation model is very close to the dynamic displacement field given by commercial CAE software analysis. The prediction time using the fast simulation model is 4.12s, which is 93 times more efficient than the 390s of commercial CAE software analysis.

[0061] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0062] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0063] It should also be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for display, data used for analysis, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0064] The various embodiments in this specification are described in a related manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0065] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0066] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A rapid simulation method for analyzing the dent resistance of car bodies, characterized in that, The method includes: A CAE simulation model is obtained based on the CAD geometric model of the vehicle's hood. After adjusting the parameter values ​​of key geometric features and performing simulation analysis through the CAE simulation model, a high-fidelity dataset is obtained. The key geometric features include the thickness of the plate material in four regions near the edge of the hair cover, the thickness of the reinforcing rib in the middle region of the hair cover, the width of the reinforcing rib in the middle region of the hair cover, and the width of the reinforcing plate inside the hair cover. The high-fidelity dataset includes multiple high-fidelity data points, each of which includes the parameter values ​​of the key geometric features, the coordinate values ​​of all nodes in the CAE simulation model, and the dynamic displacement field. The dynamic displacement field includes the displacement values ​​of each node at a preset number of time points. A prediction model is constructed and trained based on the high-fidelity dataset to obtain a fast simulation model; The target dynamic displacement field is obtained by using the rapid simulation model based on the parameter values ​​of a set of key geometric features designed in the latest version.

2. The rapid simulation method according to claim 1, characterized in that, The high-fidelity dataset obtained by adjusting the parameter values ​​of key geometric features and performing simulation analysis through the CAE simulation model includes: Multiple sets of parameter values ​​are obtained by acquiring the parameter values ​​of key geometric features through various methods; each set of parameter values ​​includes the thickness value of the plate in four areas around the hair cover near the edge, the thickness value of the reinforcing rib in the middle area of ​​the hair cover, the width value of the reinforcing rib in the middle area of ​​the hair cover, and the width value of the reinforcing plate inside the hair cover. After performing simulation analysis on each of the multiple sets of parameter values, multiple high-fidelity data points are obtained as a high-fidelity dataset, with each set of parameter values ​​corresponding to one high-fidelity data point.

3. The rapid simulation method according to claim 2, characterized in that, The simulation analysis operation includes: The first set of parameter values ​​and the preset time load data are input into the CAE simulation model for simulation analysis to obtain the corresponding dynamic displacement field; the CAE simulation model includes the coordinate values ​​of all nodes; The high-fidelity data corresponding to the first set of parameter values ​​is obtained by associating and storing the coordinate values ​​of all nodes in the CAE simulation model with the dynamic displacement field corresponding to the first set of parameter values.

4. The rapid simulation method according to claim 3, characterized in that, The construction of the prediction model includes: A prediction model is constructed based on a deep operator network. The prediction model includes two feedforward neural networks, namely a branch network and a backbone network. The branch network is used to encode the discrete input function space, and the backbone network is used to encode the output function domain.

5. The rapid simulation method according to claim 4, characterized in that, The process of training the prediction model based on the high-fidelity dataset to obtain a fast simulation model includes: The high-fidelity dataset is divided into a training dataset and a test dataset according to a preset ratio; The prediction model is trained and iterated based on the training dataset; A fast simulation model is obtained by testing the prediction model on the test dataset.

6. The rapid simulation method according to claim 5, characterized in that, The model training and model testing include: The model training process is as follows: the parameter values ​​of the corresponding key geometric features in the high-fidelity data and the preset time load data are used as the input of the branch network of the prediction model, and the coordinate values ​​of all nodes in the CAE simulation model are used as the input of the backbone network of the prediction model. The prediction model will be trained and iterated based on the corresponding dynamic displacement field output and the dynamic displacement field in the high-fidelity data. The model testing process is as follows: the parameter values ​​of the corresponding key geometric features in the high-fidelity data and the preset time load data are used as inputs to the prediction model. The prediction model will output the corresponding dynamic displacement field. The prediction accuracy is obtained based on the dynamic displacement field in the high-fidelity data and the dynamic displacement field output by the prediction model. If the prediction accuracy is greater than or equal to the preset prediction accuracy, the model test is completed and the tested prediction model is used as the rapid simulation model.

7. The rapid simulation method according to claim 4, characterized in that, The target dynamic displacement field is obtained by passing the parameter values ​​of a set of key geometric features based on the latest design through the rapid simulation model, including: The branch network of the rapid simulation model is input with the parameter values ​​of a set of key geometric features of the latest design and the preset time load data. The coordinate values ​​of all nodes in the CAE simulation model are used as the backbone network input of the rapid simulation model. The rapid simulation model will output the target dynamic displacement field, which includes the displacement value of each node in the CAE simulation model at a preset number of time points.

8. The rapid simulation method according to claim 2, characterized in that, The method of obtaining parameter values ​​for key geometric features through multiple means yields multiple sets of parameter values, including: The specific steps for obtaining the parameter values ​​of key geometric features using the single-feature independent change method are as follows: within the design range of each key geometric feature, select the corresponding preset typical values ​​according to engineering experience, and set other key geometric features as benchmark values. The benchmark values ​​are reliable values ​​of the corresponding geometric features obtained through multiple experiments. The specific steps for obtaining the parameter values ​​of key geometric features using the multi-feature collaborative transformation method are as follows: identify key geometric features with physical coupling relationships, generate the parameter values ​​of key geometric features with physical coupling relationships in the form of orthogonal combination or uniform combination, and set other key geometric features as reference values. The specific steps for obtaining the parameter values ​​of key geometric features using a random generation method are as follows: take the parameter values ​​of key geometric features obtained by using a single feature independent change method or a multi-feature collaborative change method as known parameter values, and randomly generate the parameter values ​​of key geometric features among the known parameter values ​​using a uniform random method.

9. The rapid simulation method according to claim 3, characterized in that, The CAE simulation model obtained based on the CAD geometric model of the vehicle's hood includes: The CAE finite element model generated based on the CAD geometric model of the vehicle's hood is used as the CAE simulation model; The CAE simulation model is subjected to initial simulation analysis based on the preset time load data to obtain the evaluation results; Based on the evaluation results, determine whether the CAE simulation model meets the standards.

10. The rapid simulation method according to claim 1, characterized in that, The method further includes: The maximum displacement value within the preset circular region at the center of the hood is obtained based on the target dynamic displacement field; If the maximum displacement value is less than the preset displacement value, then the parameter values ​​of the latest set of key geometric features meet the standard.

Images