A fender structural strength evaluation method and system for new energy vehicles
Through the combination of finite element analysis and neural network model, the problem of inefficiency of traditional evaluation methods is solved, and the rapid and accurate evaluation of the structural strength of the fender panel of new energy vehicles is achieved, which meets the requirements of lightweight and safety, and improves the evaluation. Efficiency and accuracy.
Patent Information
- Application Number
- CN202311751954.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-12-19
- Publication Date
- 2024-11-08
- Estimated Expiration
- 2043-12-19
AI Technical Summary
Traditional structural strength evaluation methods are time-consuming and inefficient, making them difficult to meet the dual requirements of lightweight and safety of fenders in new energy vehicles, and it is difficult to quickly and accurately evaluate structural strength.
Using finite element analysis combined with neural network model, the structural strength of the fender, including stress and strain, is evaluated by establishing three-dimensional geometric model, mesh division, loading and boundary condition definition, finite element analysis, mesh convergence analysis and neural network training. , displacement and fatigue life assessment.
It has achieved flexible and comprehensive performance evaluation of the structural strength of the fender structure of new energy vehicles, reduced the number of experiments, improved the evaluation efficiency, reduced costs, and accelerated product development.
Smart Images

Figure CN118094977B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of structural strength evaluation, and in particular to a method and system for evaluating the structural strength of a fender of a new energy vehicle. Background Art
[0002] The fender of a new energy vehicle is a body component that plays a key role in improving vehicle performance and aerodynamic performance. Components with insufficient structural strength may not be able to withstand the load during normal use, resulting in sudden failure of mechanical components or the entire system, causing safety risks. Therefore, new energy vehicles must meet the requirements of structural strength while meeting the lightweight requirements to ensure the safety of the vehicle.
[0003] In traditional structural strength assessment, it usually takes a lot of time and high costs, and the efficiency is relatively low. Therefore, the engineering community is increasingly using computer-aided engineering and numerical simulation technology to replace or assist traditional physical test methods to predict and evaluate the performance of structures, in order to improve efficiency, reduce costs and provide more comprehensive performance evaluation. These methods are usually more flexible, can simulate a variety of conditions, reduce the number of experiments, and accelerate product development. Based on the above advantages, a relevant evaluation method for the structural strength of new energy vehicle fenders has become the content that needs to be studied at present. Summary of the invention
[0004] The present invention provides a method and system for evaluating the structural strength of a fender of a new energy vehicle, which can effectively solve the problems in the background technology.
[0005] In order to achieve the above object, the technical solution adopted by the present invention is:
[0006] A method for evaluating the structural strength of a fender of a new energy vehicle comprises:
[0007] Establish the 3D geometric model of the fender;
[0008] Performing mesh division on the three-dimensional geometric model to obtain a model mesh division result;
[0009] Define loading conditions and boundary conditions for finite element analysis;
[0010] Based on the model meshing result, performing finite element analysis on the three-dimensional geometric model according to the loading conditions and boundary conditions to obtain a finite element analysis result;
[0011] The structural strength of the fender is evaluated based on the finite element analysis result.
[0012] Furthermore, after meshing the three-dimensional geometric model and before performing finite element analysis, a mesh convergence analysis is used to determine a suitable mesh density, including:
[0013] Create several grids of different densities and obtain a grid sequence;
[0014] Run finite element analysis of the three-dimensional geometric model using meshes of different densities in the mesh sequence to obtain a number of mesh analysis results;
[0015] An analysis result threshold is set based on the grid analysis result. When the grid density is gradually refined, if the change between the analysis results of the three-dimensional geometric model for two adjacent grid densities is less than or equal to the analysis result threshold, the smaller of the two adjacent grid densities is determined to be the appropriate grid density.
[0016] Furthermore, define the loading conditions for finite element analysis, including:
[0017] Acquiring the working condition of the fender;
[0018] According to the working condition of the fender, a load combination is established, and the load on the fender is calculated, and the loading condition of the finite element analysis is defined according to the load obtained by calculation.
[0019] Furthermore, the boundary conditions for finite element analysis are defined, including:
[0020] The structural characteristics of the fender are collected, the support points and constraint conditions of the fender are determined, and the boundary conditions of the finite element analysis are defined according to the structural characteristics, the support points and the constraint conditions.
[0021] Further, evaluating the fender structural strength according to the finite element analysis result includes:
[0022] Analyzing stress and strain data in the finite element analysis results to obtain stress and strain evaluation results;
[0023] Analyzing the displacement data in the finite element analysis results to obtain a displacement evaluation result;
[0024] The safety factor of the fender is calculated. If the safety factor is less than or equal to 1, the structure is determined to be unqualified. If the safety factor is greater than 1, the structure is determined to be qualified. The formula of the safety factor is:
[0025] Safety factor = maximum allowable stress / maximum withstand stress;
[0026] The maximum allowable stress is the maximum allowable stress specified in the design or specification of the structure; the maximum bearing stress is the maximum stress value that the structure can withstand when subjected to external loads, obtained by calculation;
[0027] performing fatigue life estimation on the qualified fender to obtain a fatigue life assessment result;
[0028] Based on the stress-strain evaluation result, the displacement evaluation result and the fatigue life evaluation result, the structural strength of the fender is evaluated to obtain a structural strength evaluation result.
[0029] Further, evaluating the fender structural strength according to the finite element analysis result includes:
[0030] Acquire the finite element analysis results and the structural parameters of the fender to form a data set, wherein each sample in the data set may include a finite element analysis result sequence and structural parameters related thereto;
[0031] Design a neural network model, the architecture of the neural network model includes:
[0032] The input layer receives a comprehensive input sequence formed by the series connection of the finite element analysis result sequence and the structural parameters of the fender;
[0033] a gated recurrent unit layer that receives the integrated input sequence at a current time step and a hidden state at a previous time step and generates a hidden state at a current time step;
[0034] A hidden layer, comprising a plurality of neurons, wherein the neurons receive the hidden state generated by the gated recurrent unit layer as input;
[0035] The output layer generates structural strength assessment results through a single neuron;
[0036] A loss function, calculating the difference between the predicted value of the neural network model and the actual structural strength;
[0037] An activation function is applied to the neurons of the hidden layer and the output layer to introduce nonlinear properties;
[0038] It also includes transmitting the comprehensive input sequence to the neural network model to train the neural network model;
[0039] Validate and test the trained model.
[0040] Furthermore, the calculation formula for the neuron to receive the hidden state generated by the gated recurrent unit layer as input is as follows:
[0041]
[0042] Among them, Z j is the weighted input of neuron j in the hidden layer;
[0043] W ij Enter X for the connection i and the weight of neuron j;
[0044] X i is an element from the hidden state of the gated recurrent unit layer;
[0045] b j is the bias term of neuron j, used to adjust the output;
[0046] n is the number of elements in the hidden state.
[0047] Furthermore, the method further includes adjusting the weight and bias term according to the gradient information of the loss function by a back propagation algorithm until the condition for stopping training is met, including:
[0048] In the initial stage of the neural network model, weights and bias items are randomly initialized;
[0049] In each training iteration, the current weights and biases are first used for forward propagation to obtain the prediction results, and the gradients of the weights and biases are calculated respectively;
[0050] Updating weights and bias terms, updating weights using the gradient of the weights, appending a term related to the bias term when updating, updating bias terms using the gradient of the bias term, appending a term related to the weight when updating;
[0051] Repeat the calculation of the gradient of the weight and the gradient of the bias term, and the process of updating the weight and bias term until the condition for stopping training is met.
[0052] A fender structural strength evaluation system for a new energy vehicle, the system comprising:
[0053] Model building module, building a three-dimensional geometric model of the fender;
[0054] A meshing module is used to mesh the three-dimensional geometric model to obtain a model meshing result;
[0055] Condition definition module, which defines the loading conditions and boundary conditions for finite element analysis;
[0056] A finite element analysis module, based on the model meshing result, performs finite element analysis on the three-dimensional geometric model of the fender according to the loading condition and the boundary condition to obtain a finite element analysis result;
[0057] The structural strength evaluation module evaluates the structural strength of the fender according to the finite element analysis result.
[0058] Furthermore, the grid division module includes:
[0059] A grid sequence creation unit creates several grids of different densities to obtain a grid sequence;
[0060] A mesh analysis unit, which uses meshes of different densities in the mesh sequence to run a finite element analysis of a three-dimensional geometric model to obtain a number of mesh analysis results;
[0061] The grid selection unit sets an analysis result threshold based on the grid analysis result. When the grid density is gradually refined, if the change between the analysis results of the three-dimensional geometric model for two adjacent grid densities is less than or equal to the analysis result threshold, the smaller of the two adjacent grid densities is determined to be the appropriate grid density.
[0062] The technical solution of the present invention can achieve the following technical effects:
[0063] The present invention provides a more flexible and comprehensive performance evaluation of the structural strength of the fender of a new energy vehicle, can simulate a variety of conditions, reduce the number of experiments, and at the same time, can effectively improve efficiency, reduce costs, and accelerate product development.
[0064] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS
[0065] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0066] Figure 1 A schematic diagram of a fender structural strength evaluation method and system for new energy vehicles;
[0067] Figure 2 Schematic diagram of the process for determining the appropriate mesh density;
[0068] Figure 3 Schematic diagram of the process for defining loading conditions and boundary conditions for finite element analysis;
[0069] Figure 4 A schematic diagram of the process for evaluating the structural strength of a fender;
[0070] Figure 5 A schematic diagram of the process of adjusting weights and biases according to the gradient information of the loss function through the back-propagation algorithm until the conditions for stopping training are met;
[0071] Figure 6 The structure diagram of a fender structural strength evaluation system for a new energy vehicle is shown; DETAILED DESCRIPTION
[0072] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.
[0073] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art of the present invention. The terms used in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The term "and / or" used herein includes any and all combinations of one or more related listed items.
[0074] Embodiment 1
[0075] like Figure 1 As shown, a method for evaluating the structural strength of a fender of a new energy vehicle comprises:
[0076] S100: Establishing a three-dimensional geometric model of the fender;
[0077] S200: Meshing the three-dimensional geometric model to obtain a model meshing result;
[0078] S300: Define loading conditions and boundary conditions for finite element analysis;
[0079] S400: Based on the model meshing result, finite element analysis is performed on the three-dimensional geometric model according to the loading conditions and boundary conditions to obtain finite element analysis results;
[0080] S500: Evaluate the fender structural strength based on the finite element analysis results.
[0081] Specifically, choosing to perform finite element analysis on the fenders of new energy vehicles can flexibly and accurately evaluate the structural strength of the fenders of new energy vehicles. First, it is necessary to select appropriate engineering software to establish a three-dimensional geometric model of the fender. The model should include the geometric shape, material properties, etc. of the fender, and then divide the geometric shape of the fender into small units to construct a discrete grid. These small units constitute a model, which is used to approximate the complex shape of the fender structure. The fineness of the grid directly affects the accuracy of the analysis, but it will also increase the complexity and cost of the calculation. Larger grid elements may reduce the complexity of the calculation, but may sacrifice accuracy. Therefore, in order to ensure the accuracy of the analysis, it is necessary to select an appropriate grid density.
[0082] Before conducting finite element analysis, it is necessary to clearly define loading conditions and boundary conditions. Appropriate loading conditions should be defined, including aerodynamic loads, gravity, vibration, wind loads, etc. These conditions should simulate the working conditions of the fender in actual use. At the same time, boundary conditions and the way of applying boundary conditions should be defined to simulate the force conditions of the fender. Then, using the divided grid, considering the loading conditions and boundary conditions, finite element analysis is performed on the fender of new energy vehicles to simulate the behavior of the fender under various working conditions and obtain finite element analysis results. Then, relevant information is extracted from the obtained finite element analysis results to evaluate the structural strength of the fender of new energy vehicles, for example, the stress, strain, displacement, etc. of the fender under different working conditions.
[0083] The present invention provides a more flexible and comprehensive performance evaluation of the structural strength of the fender of a new energy vehicle, can simulate a variety of conditions, reduce the number of experiments, and at the same time, can effectively improve efficiency, reduce costs, and accelerate product development.
[0084] Furthermore, after meshing the three-dimensional geometric model of the fender, and before conducting finite element analysis, a mesh convergence analysis is used to determine the appropriate mesh density, such as Figure 2 As shown, including:
[0085] S210: creating a number of grids with different densities to obtain a grid sequence;
[0086] S220: Run finite element analysis of the three-dimensional geometric model using meshes of different densities in the mesh sequence to obtain a number of mesh analysis results;
[0087] S230: Setting an analysis result threshold based on the mesh analysis result. When the mesh density is gradually refined, if the change between the analysis results of the three-dimensional geometric model for two adjacent mesh densities is less than or equal to the analysis result threshold, then determining that the smaller of the two adjacent mesh densities is the appropriate mesh density.
[0088] In the implementation process of this embodiment, in order to achieve a balance between mesh density and computational efficiency and to improve the precision of analysis and the accuracy of calculation, a mesh convergence analysis method may be used to obtain a suitable mesh density. Specifically, first, a series of grids of different densities can be created. These grids can be generated from coarse to fine, or in an incremental manner. Then, loading conditions are clearly defined, and finite element analysis is run on each grid of different density using the loading conditions to obtain analysis results under different grid densities. Then, the grid analysis results on grids of different densities are compared to determine whether the finite element analysis converges on grids of different densities. Convergence is manifested in that the grid analysis results change very little when the grid density is gradually refined. For example, finite element analysis is run using each grid of different density. For each grid of different density, the results of finite element analysis, such as stress, strain, displacement, etc., are collected and compared. Then, the changes in the analysis results on different grid densities are compared to observe whether the finite element analysis results are getting closer and closer to a stable value as the grid density increases. If the results tend to be stable, it is determined that the finite element analysis converges on grids of different densities. The above comparison is based on two adjacent grid densities. In this embodiment, when the change between the analysis results of the three-dimensional geometric model for two adjacent grid densities is less than or equal to the analysis result threshold, the smaller of the two adjacent grid densities is taken as the appropriate grid density.
[0089] Further, define the loading conditions and boundary conditions for finite element analysis, such as Figure 3 As shown, they include:
[0090] S310: Obtaining the working condition of the fender;
[0091] S320: According to the working condition of the fender, a load combination is established, and the load on the fender is calculated, and the loading condition of the finite element analysis is defined according to the calculated load.
[0092] The above is the definition process of loading conditions, and for the definition of boundary conditions, the following steps are continued:
[0093] S330: Collect structural characteristics of the fender, determine support points and constraint conditions of the fender, and define boundary conditions of finite element analysis according to the structural characteristics, support points and constraint conditions.
[0094] Specifically, it is first necessary to obtain the actual use conditions of the fender, including static conditions, dynamic conditions, etc. According to the above conditions, multiple load combinations can be established. Each load combination contains various load conditions that the fender bears under the same condition. The size and direction of each load are calculated. The calculation result of each load combination represents the comprehensive loading conditions under a certain condition. The loading conditions under different conditions can be defined through the comprehensive loading conditions. Regarding the definition of loading conditions, in this preferred solution, it is necessary to obtain the structural characteristics of the fender, including its geometric shape, material properties, connection method, support structure, etc. The support point of the fender is determined according to the position where the fender is connected to the vehicle structure. Then, according to the support point and other structural characteristics, the constraint conditions of the fender can be determined, including the displacement of the fixed support point, the stiffness of the simulated support structure, the prohibition of certain displacements or rotations, etc. With the support point, constraint conditions and structural characteristics, the boundary conditions of the finite element analysis can be defined.
[0095] Furthermore, based on the finite element analysis results, the fender structural strength is evaluated, such as Figure 4 As shown, including:
[0096] S510: Analyze stress and strain data in the finite element analysis results according to the finite element analysis results to obtain stress and strain evaluation results;
[0097] S520: Analyze displacement data in the finite element analysis results according to the finite element analysis results to obtain a displacement evaluation result;
[0098] S530: Calculate the safety factor of the fender. If the safety factor is less than or equal to 1, the structure is determined to be unqualified. If the safety factor is greater than 1, the structure is determined to be qualified. The formula for calculating the safety factor of the fender is:
[0099] Safety factor = maximum allowable stress / maximum withstand stress;
[0100] The maximum allowable stress refers to the maximum allowable stress specified in the design or specification of the structure; the maximum bearing stress refers to the maximum stress value that the structure can withstand when subjected to external loads, calculated by calculation;
[0101] S540: Estimating fatigue life of qualified fenders to obtain fatigue life assessment results;
[0102] S550: Based on the stress-strain evaluation results, the displacement evaluation results, and the fatigue life evaluation results, the fender structural strength is evaluated to obtain a structural strength evaluation result.
[0103] In this embodiment, it is first necessary to extract stress and strain data from the finite element analysis results of the fender, and obtain stress and strain evaluation results by analyzing the stress and strain data, including focusing on key areas, such as the edge of the fender, the connection point and the supporting structure, identifying the maximum stress value and the maximum strain value of the key area, analyzing the stress and strain distribution, determining whether there are high stress concentration areas and high strain areas, and analyzing the stress data and strain data, including normal stress, shear stress, principal stress, normal strain, shear strain, etc.; then extracting displacement data from the finite element analysis results to understand the displacement of the fender under different load conditions and determine whether there are abnormal or unstable displacements. The displacement assessment result is used to determine whether the structure will be excessively deformed, especially under extreme working conditions. The safety factor of the fender is then calculated. If the safety factor is less than or equal to 1, it means that the fender structure may have safety hazards and the structure is judged to be unqualified. If the safety factor is greater than 1, it means that the fender still has a certain safety margin under stress. Fatigue life estimation is then performed to determine whether fatigue failure will occur in the long-term use of the fender. Finally, based on the evaluation results of stress, strain, displacement, safety factor and fatigue life, a comprehensive evaluation of the structural strength of the fender is performed to help determine whether the fender is strong enough to withstand loads under various working conditions.
[0104] As another preferred method, the fender structural strength is evaluated based on the finite element analysis results, including:
[0105] A510: Obtain finite element analysis results and structural parameters of the fender to form a data set. Each sample in the data set may include a finite element analysis result sequence and structural parameters related thereto.
[0106] A520: Design a neural network model. The architecture of the neural network model includes:
[0107] The input layer receives a comprehensive input sequence formed by the series connection of the finite element analysis result sequence and the structural parameters of the fender;
[0108] The gated recurrent unit layer accepts the comprehensive input sequence of the current time step and the hidden state of the previous time step, and generates the hidden state of the current time step;
[0109] Hidden layers, which include several neurons that receive the hidden states generated by the gated recurrent unit layer as input; this allows the neural network to further process features and build representations of the model in the hidden layers for more advanced feature extraction and learning tasks;
[0110] The output layer generates structural strength assessment results through a single neuron;
[0111] The loss function calculates the gap between the predicted value of the neural network model and the actual structural strength; this gap is usually expressed in numerical form, and the minimization of the loss function is achieved through the back-propagation algorithm, which helps the neural network adjust its weights and parameters to improve the performance of the model;
[0112] Activation functions, applied to neurons in the hidden and output layers to introduce nonlinear properties;
[0113] A530: Pass the synthetic input sequence to the neural network model to train the neural network model; the goal is to enable the model to predict the structural strength of the fender. According to the provided finite element analysis results and structural parameters, during the training process, the model will adjust its weights and related parameters to minimize the loss function;
[0114] A540: Validate and test the trained model. Specifically, by inputting new finite element analysis results and structural parameters into the model to check whether it can accurately predict the structural strength of the fender, validation and testing are key steps to ensure the generalization performance of the model.
[0115] In the above optimization scheme, by adopting the neural network model, structural strength assessment can be performed in a short time, thereby greatly improving the efficiency of the assessment; the neural network model can be trained on a large-scale data set, making the structural strength assessment process more automated, reducing the need for manual intervention and reducing human errors; the trained neural network model has good generalization capabilities and can cope with various fender designs and working conditions, rather than being limited to specific cases.
[0116] In step A1, the finite element analysis results and structural parameters are combined into a data set to make the data set more comprehensive and rich, so that it can not only capture the behavior of the fender, but also include the key factors that affect the behavior, such as geometry, materials and support conditions, etc.; this step adapts the evaluation model to the changes of different fenders. By changing the structural parameters, fenders of different types, different geometries or different materials can be evaluated.
[0117] In the fender structural strength assessment, finite element analysis usually produces a data sequence about time changes, which includes structural response information at different time steps. This time series can include multiple data points, each of which corresponds to a specific time point or time period. Therefore, "sequence" here refers to the change of finite element analysis results over time, or the structural response data recorded over time steps. For the architecture of the neural network model in step A2, one of the main advantages of the gated recurrent unit layer is its ability to capture time dependencies. In the neural network, the gated recurrent unit layer helps the model understand and utilize information at different time points to predict the strength of the fender structure. By processing time series data, the model can identify and learn time-related dynamic patterns, thereby more accurately simulating the behavior of the fender at different time points, which helps to improve the performance of the model, especially in the structural strength assessment task that needs to consider time factors.
[0118] In the process of fender structural strength assessment, the hidden state represents the network internal memory or information of the current time step, and the gated recurrent unit layer has a memory function, which can calculate and update the hidden state of the current time step based on the hidden state of the previous time step and the input of the current time step. The hidden state can contain information about the structural behavior learned by the model when analyzing the finite element analysis results of past time steps to help the model better predict the structural strength of future time steps, which helps the model better understand and consider time-dependent dynamic patterns.
[0119] As a preferred embodiment of the above, the calculation formula for a neuron to receive the hidden state generated by the gated recurrent unit layer as input is as follows:
[0120]
[0121] Among them, Z j is the weighted input of neuron j in the hidden layer;
[0122] W ij Enter X for the connection i and the weight of neuron j;
[0123] X i is an element from the hidden state of the gated recurrent unit layer;
[0124] b j is the bias term of neuron j, used to adjust the output;
[0125] n is the number of elements in the hidden state.
[0126] In the implementation process of the above-mentioned fender structural strength evaluation method for new energy vehicles, it is also preferred to include a process of adjusting the weights and bias items according to the gradient information of the loss function through a back propagation algorithm until the conditions for stopping training are met, such as Figure 5 As shown, including:
[0127] B100: In the initial stage of the neural network model, the weights and bias terms are randomly initialized;
[0128] B200: In each training iteration, first use the current weights and biases to perform forward propagation to obtain the prediction results, and calculate the gradients of the weights and biases respectively;
[0129] B300: Update weights and bias terms, use the gradient of the weight to update the weight, append a term related to the bias term when updating, use the gradient of the bias term to update the bias term, and append a term related to the weight when updating;
[0130] B400: Repeat the calculation of the gradient of the weight and the gradient of the bias term, and the process of updating the weight and bias term until the conditions for stopping training are met.
[0131] Adding terms related to each other when updating weights and bias terms usually involves a regularization method. In the above preferred embodiment, by introducing regularization terms related to each other when updating weights and bias terms, the generalization performance, stability and controllability of the neural network can be improved, making it more suitable for various practical applications.
[0132] In the implementation process, regularization helps reduce the overfitting of the model to the training data by constraining the range of values of weights and bias terms. Overfitting means that the model performs well on the training data, but performs poorly on unseen data. By introducing regularization terms, the complexity of the model can be reduced, making it more generalizable to new data, thereby improving the performance of the model. Regularization helps prevent parameter divergence, that is, prevents the value of weights or bias terms from becoming very large, which helps improve the stability and convergence of the model and makes training more successful; regularization can introduce prior knowledge by appending terms related to weights or bias terms to constrain the behavior of the model.
[0133] Taking the example of updating weights using their gradients, this can be accomplished by simply adding a term related to the bias term to the gradient update:
[0134] Assume that when updating the weight W i , and the value associated with the bias term is B i , B is recommended i is a scalar, and the gradient descent update rule is as follows:
[0135] New weight Wi_new = W i - Learning rate*(gradient of weights + B i )
[0136] In the above implementation, the learning rate is a hyperparameter used to control the step size of the weight update in each iteration. During the implementation, a fixed learning rate can be selected, and of course an adaptive learning rate can also be used; the gradient of the weight is calculated according to the loss function W i The partial derivative of i It is a value related to the bias term. A specific value can be selected during the implementation process. This value represents the set target value to be achieved by the neural network model during the optimization process. In the training of the neural network, the optimization goal may include minimizing the loss function or achieving a certain performance indicator, which is selected according to the specific situation.
[0137] Similarly, assuming that when updating the bias term b j , and the value associated with the weight is C, which can be taken as a scalar, and the gradient descent update rule is as follows:
[0138] New bias term b j_new = b j - Learning rate * (gradient of bias + C)
[0139] In this implementation, the learning rate is still a hyperparameter used to control the step size of the bias term update in each iteration; the gradient of the bias term is calculated based on the loss function b j The partial derivative of. C is the value associated with the weight, and B i Specific values can be selected during the implementation process. These values represent the set target values that the neural network model should achieve during the optimization process. In the training of the neural network, the optimization goal may include minimizing the loss function or achieving a certain performance indicator, which is selected according to the specific situation.
[0140] This method is similar to the way of updating weights, but when updating the bias term, a value C related to the weight is introduced, so that the update process of the weight and bias term affects each other. This can be used to control the relationship between the weight and bias term in the model to meet the needs of specific tasks.
[0141] Embodiment 2:
[0142] like Figure 6 As shown, a fender structural strength evaluation system for new energy vehicles includes:
[0143] Model building module, building a three-dimensional geometric model of the fender;
[0144] A meshing module is used to mesh the three-dimensional geometric model of the fender and obtain the meshing result of the model;
[0145] Condition definition module, which defines the loading conditions and boundary conditions for finite element analysis;
[0146] The finite element analysis module performs finite element analysis on the fender 3D geometric model based on the model meshing results and the loading and boundary conditions to obtain the finite element analysis results;
[0147] The structural strength assessment module assesses the structural strength of the fender based on the finite element analysis results.
[0148] The above-mentioned evaluation system in the present invention can effectively implement the fender structural strength evaluation method of new energy vehicles, and the technical effects that can be achieved are as described in the above-mentioned embodiments, which will not be repeated here.
[0149] Furthermore, the grid partitioning module includes:
[0150] A grid sequence creation unit creates several grids of different densities to obtain a grid sequence;
[0151] A mesh analysis unit, which runs finite element analysis of a three-dimensional geometric model using meshes of different densities in a mesh sequence to obtain several mesh analysis results;
[0152] The grid selection unit sets an analysis result threshold based on the grid analysis result. When the grid density is gradually refined, if the change between the analysis results of the three-dimensional geometric model for two adjacent grid densities is less than or equal to the analysis result threshold, the smaller of the two adjacent grid densities is determined to be the appropriate grid density.
[0153] Similarly, the above-mentioned optimization schemes for the system can also respectively achieve the optimization effects corresponding to the method in Example 1, which will not be repeated here.
[0154] Although the present application has been described in conjunction with specific features and embodiments thereof, it is obvious that various modifications and combinations may be made thereto without departing from the spirit and scope of the present application. Accordingly, this specification and the accompanying drawings are merely exemplary illustrations of the present application as defined therein, and are deemed to have covered any and all modifications, variations, combinations or equivalents within the scope of the present application. Obviously, a person skilled in the art may make various modifications and variations to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application is intended to include these modifications and variations.
Claims
1. A method for evaluating the structural strength of a fender of a new energy vehicle, characterized in that: include: Establish the 3D geometric model of the fender; Performing mesh division on the three-dimensional geometric model to obtain a model mesh division result; Define loading conditions and boundary conditions for finite element analysis; Based on the model meshing result, performing finite element analysis on the three-dimensional geometric model according to the loading conditions and boundary conditions to obtain a finite element analysis result; According to the finite element analysis result, evaluating the structural strength of the fender; Defining loading conditions for finite element analysis includes: obtaining the working conditions of the fender, including static working conditions and dynamic working conditions; establishing a load combination according to the working conditions of the fender, and calculating the load on the fender, and defining the loading conditions for finite element analysis according to the calculated load; Defining the boundary conditions of the finite element analysis includes: collecting the structural characteristics of the fender, including the geometric shape, material properties, connection mode, and support structure, determining the support points and constraint conditions of the fender, and defining the boundary conditions of the finite element analysis according to the structural characteristics, support points, and constraint conditions, specifically, determining the support points of the fender according to the position where the fender is connected to the vehicle structure, and determining the constraint conditions of the fender according to the support points and other structural characteristics, including the displacement of the fixed support points, simulating the stiffness of the support structure, and prohibiting the setting of displacement or rotation; According to the finite element analysis result, the structural strength of the fender is evaluated, including: Analyze the stress and strain data in the finite element analysis results to obtain stress and strain assessment results, including focusing on key areas, identifying the maximum stress value and the maximum strain value of the key areas, analyzing the stress and strain distribution, determining whether there are high stress concentration areas and high strain areas, and analyzing the stress data and strain data; Analyzing the displacement data in the finite element analysis results to determine whether there is abnormal or unstable deformation and obtaining a displacement evaluation result; The safety factor of the fender is calculated. If the safety factor is less than or equal to 1, the structure is determined to be unqualified. If the safety factor is greater than 1, the structure is determined to be qualified. The formula of the safety factor is: Safety factor = maximum allowable stress / maximum withstand stress; The maximum allowable stress is the maximum allowable stress specified in the design or specification of the structure; the maximum bearing stress is the maximum stress value that the structure can withstand when subjected to external loads, obtained by calculation; The fatigue life of the qualified fender is estimated, and the fender still has a safety margin under the force, so as to obtain a fatigue life assessment result; Based on the stress-strain evaluation result, the displacement evaluation result and the fatigue life evaluation result, the structural strength of the fender is evaluated to obtain a structural strength evaluation result.
2. The method for evaluating the structural strength of a fender of a new energy vehicle according to claim 1, characterized in that: After meshing the three-dimensional geometric model and before performing finite element analysis, a mesh convergence analysis is performed to determine the appropriate mesh density, including: Create several grids of different densities and obtain a grid sequence; Run finite element analysis of the three-dimensional geometric model using meshes of different densities in the mesh sequence to obtain a number of mesh analysis results; An analysis result threshold is set based on the grid analysis result. When the grid density is gradually refined, if the change between the analysis results of the three-dimensional geometric model for two adjacent grid densities is less than or equal to the analysis result threshold, the smaller of the two adjacent grid densities is determined to be the appropriate grid density.
3. The method for evaluating the structural strength of a fender of a new energy vehicle according to claim 1, characterized in that: According to the finite element analysis result, the structural strength of the fender is evaluated, including: Acquiring the finite element analysis results and the structural parameters of the fender to form a data set, wherein each sample in the data set includes a finite element analysis result sequence and structural parameters related thereto; Design a neural network model, the architecture of the neural network model includes: The input layer receives a comprehensive input sequence formed by the series connection of the finite element analysis result sequence and the structural parameters of the fender; a gated recurrent unit layer that receives the integrated input sequence at a current time step and a hidden state at a previous time step and generates a hidden state at a current time step; A hidden layer, comprising a plurality of neurons, wherein the neurons receive the hidden state generated by the gated recurrent unit layer as input; The output layer generates structural strength assessment results through a single neuron; A loss function, calculating the difference between the predicted value of the neural network model and the actual structural strength; An activation function is applied to the neurons of the hidden layer and the output layer to introduce nonlinear properties; It also includes transmitting the comprehensive input sequence to the neural network model to train the neural network model; Validate and test the trained model.
4. The method for evaluating the structural strength of a fender of a new energy vehicle according to claim 3, characterized in that: The calculation formula for the neuron to receive the hidden state generated by the gated recurrent unit layer as input is as follows: Among them, Z j is the weighted input of neuron j in the hidden layer; W ij Enter X for the connection i and the weight of neuron j; X i is an element from the hidden state of the gated recurrent unit layer; b j is the bias term of neuron j, used to adjust the output; n is the number of elements in the hidden state.
5. The method for evaluating the structural strength of a fender of a new energy vehicle according to claim 4, characterized in that: The method further includes adjusting the weight and bias items according to the gradient information of the loss function by a back propagation algorithm until the condition for stopping training is met, including: In the initial stage of the neural network model, weights and bias items are randomly initialized; In each training iteration, the current weights and biases are first used for forward propagation to obtain the prediction results, and the gradients of the weights and biases are calculated respectively; Updating weights and bias terms, updating weights using the gradient of the weights, appending a term related to the bias term when updating, updating bias terms using the gradient of the bias term, appending a term related to the weight when updating; Repeat the calculation of the gradient of the weight and the gradient of the bias term, and the process of updating the weight and bias term until the condition for stopping training is met.
6. A fender structural strength evaluation system for a new energy vehicle, using the fender structural strength evaluation method for a new energy vehicle as claimed in claim 1, characterized in that: The system comprises: Model building module, building a three-dimensional geometric model of the fender; A meshing module is used to mesh the three-dimensional geometric model to obtain a model meshing result; Condition definition module, which defines the loading conditions and boundary conditions for finite element analysis; A finite element analysis module, based on the model meshing result, performs finite element analysis on the three-dimensional geometric model of the fender according to the loading condition and the boundary condition to obtain a finite element analysis result; The structural strength evaluation module evaluates the structural strength of the fender according to the finite element analysis result.
7. A fender structural strength evaluation system for new energy vehicles according to claim 6, characterized in that: The grid division module comprises: A grid sequence creation unit creates several grids of different densities to obtain a grid sequence; A mesh analysis unit, which uses meshes of different densities in the mesh sequence to run a finite element analysis of a three-dimensional geometric model to obtain a number of mesh analysis results; The grid selection unit sets an analysis result threshold based on the grid analysis result. When the grid density is gradually refined, if the change between the analysis results of the three-dimensional geometric model for two adjacent grid densities is less than or equal to the analysis result threshold, the smaller of the two adjacent grid densities is determined to be the appropriate grid density.
