Model-based collision energy absorption part design method and system, medium and equipment

Abstract

The invention discloses a model-based collision energy-absorbing component design method, system, medium and equipment, and the method comprises the steps: randomly generating an energy-absorbing component combination following a specific combination rule, and setting key structure characteristic parameters according to internal and external boundary constraints of spatial arrangement to form a first characteristic parameter combination; processing the parameter combinations by using parametric modeling and grid generation technologies, and creating a finite element analysis model of the energy absorption component; performing performance test on the model through a simulation algorithm to obtain a simulation result; training a neural network based on a simulation result to obtain a design model; and encoding the energy-absorbing component and the characteristic parameters thereof into individuals in the genetic algorithm, performing iterative optimization, evaluating the individual fitness by using the design model in each iteration, and outputting the optimal energy-absorbing component design when a preset condition is met. According to the method, the optimization algorithm and the artificial intelligence model are combined, efficient conversion from performance requirements to energy absorption component design is achieved, and the design efficiency of the vehicle body collision energy absorption component is improved.

Description

Technical Field

[0001] The invention belongs to the technical field of automobile engineering, and relates to a model-based collision energy absorption component design method, system, medium and equipment. Background Art

[0002] In the field of automotive engineering, the design of the collision energy absorption components of the vehicle body is key to ensuring the safety of passengers. Traditionally, the design process relies on the experience of engineers and simulation methods such as finite element analysis, combined with experimental verification for optimization. Although this method is effective, it often requires a lot of time and computing resources, especially in the early conceptual design stage, and it is difficult to quickly evaluate the energy absorption efficiency of different structures.

[0003] The existing technology has some shortcomings. On the one hand, it does not make much use of past vehicle model data, resulting in duplication of work and extended design cycles. On the other hand, it relies too much on the experience of engineers for the optimization of complex structures, which may lead to less than ideal design solutions. If an accurate proxy model is to be built, a large amount of simulation data is required, which is not only time-consuming but also costly. Especially when the project is under time pressure, it is difficult to ensure design quality and efficiency. Summary of the invention

[0004] In response to the shortcomings of the prior art, the present application provides a model-based collision energy absorption component design method, system, medium and equipment. By combining optimization algorithms with artificial intelligence models, it achieves an efficient conversion from performance requirements to energy absorption component design, thereby improving the design efficiency of vehicle body collision energy absorption components.

[0005] To achieve the above objectives, in a first aspect, the present invention provides a model-based collision energy absorbing component design method, comprising:

[0006] According to the connection relationship of each energy absorbing component of each reference vehicle model, randomly generate a combination of each energy absorbing component that meets the preset combination rules;

[0007] According to the preset inner and outer boundary constraint parameters of the energy absorbing component spatial arrangement, a number of key structural characteristic parameters are set for each energy absorbing component in each energy absorbing component combination to obtain a first characteristic parameter combination corresponding to each energy absorbing component combination;

[0008] According to a preset parameterized model generation algorithm and a mesh generation algorithm, each of the first characteristic parameter combinations is processed to obtain a finite element analysis model of each energy absorbing component, and according to a preset simulation algorithm, a preset finite element analysis model of a non-energy absorbing component and the finite element analysis model of each of the energy absorbing components are simulated to obtain various performance simulation results;

[0009] According to the performance simulation results, the first characteristic parameter combination and the energy absorbing component combination data, a preset original neural network model is trained to obtain a design model;

[0010] Each combination of energy-absorbing components and key structural characteristic parameters are represented as a number of population individuals of a genetic algorithm, and the population is iteratively optimized according to a preset genetic operation algorithm until a preset number of iterations is reached or the population individuals meet preset expected performance indicators, and the current combination of energy-absorbing components and the corresponding key structural characteristic parameters are output as a design result; and in each iterative optimization, each population individual is processed according to the design model to obtain predicted performance indicators of each population individual, and the fitness value is calculated according to the deviation between each predicted performance indicator and the expected performance indicator.

[0011] Compared with the prior art, the embodiments of the present application have the following beneficial effects: randomly generated combinations ensure the diversity of design schemes and avoid the influence of engineers' subjective experience; key structural feature parameters are set and specific parameter ranges are defined, so that subsequent simulation and analysis have clear goals and boundary conditions, reducing unnecessary waste of computing resources; the actual performance of different design schemes can be predicted in advance through simulation, reducing the cost and time consumption of physical experiments; the neural network model is trained in combination with key structural feature parameters and simulation results, and a large amount of simulation data is processed using AI technology to quickly establish a mapping relationship between input (structural features) and output (performance indicators), thereby improving the speed and accuracy of data analysis; the genetic algorithm works synergistically with the model and combines the principles of evolution to automatically find the optimal solution without human intervention, further improving the design efficiency while ensuring the quality of the design results.

[0012] In some embodiments of the first aspect of the present application, the randomly generating a combination of energy absorbing components that conforms to a preset combination rule according to the connection relationship of the energy absorbing components of each reference vehicle model includes:

[0013] Obtaining information of each energy absorbing component of each reference vehicle model;

[0014] According to a preset coding rule, the information of each energy absorbing component is converted into a numerical parameter;

[0015] According to the connection relationship of each energy absorbing component and a preset combination rule, each combination of the energy absorbing components is randomly generated.

[0016] Compared with the prior art, the above embodiment has the following beneficial effects: the energy-absorbing component information of each reference vehicle model is converted into numerical parameters, and component combinations are randomly generated based on connection relationships and combination rules, so that the energy-absorbing component designs between different vehicle models can be quickly transplanted and compared, reducing duplication of work; in addition, the use of unified coding rules simplifies the subsequent processing flow, improves the speed and accuracy of parametric modeling, thereby accelerating the entire design cycle and reducing development costs.

[0017] In some embodiments of the first aspect of the present application, according to the preset inner and outer boundary constraint parameters of the spatial arrangement of the energy absorbing components, a number of key structural characteristic parameters are set for each energy absorbing component in each energy absorbing component combination to obtain a first characteristic parameter combination corresponding to each energy absorbing component combination, including:

[0018] According to the preset inner and outer boundary constraint parameters of the energy absorbing component spatial arrangement, a number of structural characteristic parameters are set for each energy absorbing component;

[0019] According to the preset SGMW sensitivity analysis specification, analyze and calculate the importance value of each structural characteristic parameter to the preset safety performance index;

[0020] The importance values ​​are sorted from large to small, and a number of structural feature parameters corresponding to the importance values ​​ranked in front are selected as key structural feature parameters.

[0021] Compared with the prior art, the above embodiment has the following beneficial effects: the structural characteristic parameters are set according to the internal and external boundary constraint parameters, and the key structural characteristic parameters are optimized according to the SGMW sensitivity analysis specification to ensure the importance of the selected parameters to the safety performance indicators; the sorting selection mechanism ensures that the most important parameters are given priority, thereby improving the safety and reliability of the design scheme; at the same time, it helps to reduce unnecessary complexity and make the design more focused on the factors that have the greatest impact on the overall performance.

[0022] In some embodiments of the first aspect of the present application, the processing of each first characteristic parameter combination according to a preset parameterized model generation algorithm and a mesh generation algorithm to obtain a finite element analysis model of each energy absorbing component includes:

[0023] Extracting the first characteristic parameter combination according to a preset sampling algorithm to obtain a plurality of sampled first characteristic parameter combinations;

[0024] According to a preset parameterized model generation algorithm and a mesh generation algorithm, each of the sampled first characteristic parameter combinations is processed to obtain a finite element analysis model of each energy absorbing component.

[0025] Compared with the prior art, the above embodiment has the following beneficial effects: by sampling the first characteristic parameter combination and using a parametric model generation algorithm and a mesh generation algorithm to construct a finite element analysis model, the computing resource bottleneck problem that may be encountered when analyzing the full amount of data is solved. The sampling method not only retains the core characteristics of the original data, but also reduces the computing burden, making large-scale simulation possible and promoting more extensive design exploration and optimization.

[0026] In some embodiments of the first aspect of the present application, the method of training a preset original neural network model according to the performance simulation results, the first characteristic parameter combination, and the energy absorbing component combination data to obtain a design model includes:

[0027] Constructing an original mapping data set according to each of the performance simulation results and the corresponding first characteristic parameter combination and the corresponding energy absorbing component combination;

[0028] According to a preset preprocessing algorithm, the original mapping data set is processed to obtain a mapping data set; wherein the preprocessing algorithm includes any one or more combinations of the following: missing value processing, outlier value processing, duplicate value processing, renaming processing, format unification processing, normalization processing, and feature encoding processing.

[0029] Compared with the prior art, the above embodiment has the following beneficial effects: on the basis of constructing the original mapping data set, applying preprocessing algorithms (such as missing value processing, outlier processing, etc.) can significantly improve the quality of the data, eliminate noise interference, and enhance the effectiveness of model learning.

[0030] In some embodiments of the first aspect of the present application, the training of a preset original neural network model according to the performance simulation results, the first characteristic parameter combination and the energy absorbing component combination data to obtain a design model further includes:

[0031] According to the preset K-fold cross validation algorithm, dropout algorithm and the mapping data set, the original neural network model is trained to obtain the designed model;

[0032] The training adopts the preset AdamW algorithm to optimize the learning rate, and the algorithm is as follows:

[0033] m t =β1m t-1 +(1-β1)g t ; Where m t represents the first-order moment estimate of the current step t, g t represents the gradient of the current step, β1 and β2 represent the first-order and second-order momentum decay coefficients respectively, and v t represents the second-order moment estimate at time step t; where β1 and β2 are dynamically adjusted as follows: β1 = β 1,init (1-t / T); β2 = β 2,init ·(1-t / T); where β 1,init and β 2,init They represent the initial values ​​of the momentum decay coefficients of β1 and β2 respectively, and T is the total number of training steps; the gradient clipping algorithm is as follows: Where τ represents the gradient clipping threshold.

[0034] Compared with the prior art, the above embodiment has the following beneficial effects: K-fold cross validation, dropout and AdamW optimization algorithm are used to train the neural network model to avoid overfitting and ensure that the model has good generalization ability and stability; the application of dynamic adjustment of first-order and second-order momentum attenuation coefficients and gradient clipping algorithm further improves the convergence speed and stability during model training.

[0035] In some embodiments of the first aspect of the present application, the calculating the fitness value according to the deviation between each of the predicted performance indicators and the expected performance indicator includes:

[0036] The calculation formula for calculating the fitness value is as follows:

[0037] F(i)=-MSE(y(i),y target ), where F(i) represents the fitness value of individual i in the population, y(i) represents the prediction performance index of the designed model, and y target represents the expected performance index, and MSE represents minimizing the mean square error.

[0038] Compared with the prior art, the above embodiment has the following beneficial effects: the genetic algorithm and the model work together, combining the principle of evolution to automatically find the optimal solution without human intervention, further improving the design efficiency while ensuring the quality of the design results.

[0039] In a second aspect, the present invention further provides a model-based collision energy absorption component design system, comprising: a component combination module, a characteristic parameter setting module, a simulation module, a model training module and a processing module;

[0040] The component combination module is used to randomly generate a combination of energy absorbing components that meets a preset combination rule according to the connection relationship of the energy absorbing components of each reference vehicle model;

[0041] The characteristic parameter setting module is used to set a number of key structural characteristic parameters for each energy absorbing component in each energy absorbing component combination according to the preset inner and outer boundary constraint parameters of the energy absorbing component spatial arrangement, so as to obtain a first characteristic parameter combination corresponding to each energy absorbing component combination;

[0042] The simulation module is used to process each of the first characteristic parameter combinations according to a preset parameterized model generation algorithm and a mesh generation algorithm to obtain a finite element analysis model of each energy absorbing component, and simulate a preset finite element analysis model of a non-energy absorbing component and the finite element analysis model of each of the energy absorbing components according to a preset simulation algorithm to obtain various performance simulation results;

[0043] The model training module is used to train a preset original neural network model according to the performance simulation results, the first characteristic parameter combination and the energy absorbing component combination data to obtain a design model;

[0044] The processing module is used to characterize each energy absorbing component combination and key structural characteristic parameters as a number of population individuals of a genetic algorithm, and iteratively optimize the population according to a preset genetic operation algorithm until a preset number of iterations is reached or the population individuals meet preset expected performance indicators, and output the current energy absorbing component combination and the corresponding key structural characteristic parameters as a design result; and in each iterative optimization, various population individuals are processed according to the design model to obtain predicted performance indicators of various population individuals, and the fitness value is calculated according to the deviation between each predicted performance indicator and the expected performance indicator.

[0045] Compared with the prior art, the above embodiments of the present application have the following beneficial effects: randomly generated combinations ensure the diversity of design schemes and avoid the influence of engineers' subjective experience; key structural feature parameters are set and specific parameter ranges are defined, so that subsequent simulation and analysis have clear goals and boundary conditions, reducing unnecessary waste of computing resources; the actual performance of different design schemes can be predicted in advance through simulation, reducing the cost and time consumption of physical experiments; the neural network model is trained in combination with key structural feature parameters and simulation results, and a large amount of simulation data is processed using AI technology to quickly establish a mapping relationship between input (structural features) and output (performance indicators), thereby improving the speed and accuracy of data analysis; the genetic algorithm works synergistically with the model and combines the principles of evolution to automatically find the optimal solution without human intervention, further improving the design efficiency while ensuring the quality of the design results.

[0046] In a third aspect, the present invention also provides a model-based collision energy absorbing component design device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program implements the steps of the model-based collision energy absorbing component design method when loaded into the processor.

[0047] In a fourth aspect, an embodiment of the present application further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the model-based collision energy absorption component design method are implemented. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 : A schematic flow chart of a model-based collision energy absorption component design method provided in some embodiments of the present invention.

[0049] Figure 2: A schematic structural diagram of a model-based collision energy absorption component design system provided in some embodiments of the present invention.

[0050] Figure 3 : A structural diagram of a model-based collision energy absorbing component design device provided in some embodiments of the present invention. DETAILED DESCRIPTION

[0051] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0052] Embodiment 1:

[0053] Please refer to Figure 1 , a model-based collision energy absorption component design method provided by an embodiment of the present invention, comprising steps S1 to S5:

[0054] Step S1: randomly generating combinations of energy absorbing components that meet preset combination rules according to the connection relationship of the energy absorbing components of each reference vehicle model.

[0055] Preferably, in some embodiments of the present application, step S1 can be implemented by the following preferred implementations, including steps S11-S13, as follows:

[0056] S11: Acquire information of each energy absorbing component of each reference vehicle model;

[0057] S12: converting the information of each energy absorbing component into a numerical parameter according to a preset coding rule;

[0058] S13: randomly generating a combination of the energy absorbing components according to the connection relationship of the energy absorbing components and a preset combination rule.

[0059] For example, in the specific implementation, the anti-collision beam in the energy-absorbing component is numbered as A, and the corresponding anti-collision beams of different models are numbered as A1, A2, ..., An respectively; the energy absorbing box in the energy-absorbing component is numbered as B, and the corresponding energy absorbing box numbers of different models are B1, B2, ..., Bn; the front longitudinal beam is numbered as C, and the corresponding front longitudinal beam numbers are C1, C2, ..., Cn; the subframe is numbered as D, and the corresponding subframes are D1, D2, ..., Dn. The combination of different types of components is fully considered, and the possible combination scheme of vehicle body collision energy-absorbing components is determined according to the lap structure and connection method of the energy-absorbing components, such as A1B2C3D5, etc., and a new combination scheme of energy-absorbing components is obtained by combination.

[0060] In this preferred embodiment, the energy-absorbing component information of each reference vehicle model is converted into numerical parameters, and component combinations are randomly generated based on connection relationships and combination rules, so that the energy-absorbing component designs between different vehicle models can be quickly transplanted and compared, reducing duplication of work; in addition, the use of unified coding rules simplifies the subsequent processing flow, improves the speed and accuracy of parametric modeling, thereby accelerating the entire design cycle and reducing development costs.

[0061] Step S2: according to the preset inner and outer boundary constraint parameters of the energy absorbing component spatial arrangement, a number of key structural characteristic parameters are set for each energy absorbing component in each energy absorbing component combination to obtain a first characteristic parameter combination corresponding to each energy absorbing component combination.

[0062] Preferably, in some embodiments of the present application, step S2 can be implemented by the following preferred implementations, including steps S21 to S23, as follows:

[0063] S21: setting a number of structural characteristic parameters for each energy absorbing component according to the preset inner and outer boundary constraint parameters of the spatial arrangement of the energy absorbing component;

[0064] For example, in the specific implementation, the material thickness T1 of the anti-collision beam part, the material thickness T3 of the energy absorption box part, the inner and outer panels of the front longitudinal beam T3 / T4, the material thickness T5 of the reinforcement plate part, the material thickness T6 of the subframe, etc. are used as thickness characteristic parameters, and the cross-sectional dimensions of the energy absorption components: the height and width H1 / H2 of the anti-collision beam section, the width and height H3 / H4 of the energy absorption box section, the height and width H5 / H6 of the front longitudinal beam section, the width and height H7 / H8 of the subframe section, etc. are used as cross-sectional characteristic parameters, the length and width of the key ribs of the energy absorption box and the front longitudinal beam are used as structural characteristic variables L1 / L2, the material elastic modulus E1 / E2 / … / En of the energy absorption component, and the material stress and strain curve LC1 / LC2… etc. are used as material control parameters.

[0065] S22: Analyze and calculate the importance value of each of the structural characteristic parameters to the preset safety performance index according to the preset SGMW sensitivity analysis specification;

[0066] S23: sorting the importance values ​​from large to small, and selecting several structural feature parameters corresponding to the importance values ​​ranked in front as key structural feature parameters.

[0067] In specific implementation, collision sensitivity analysis can be used to output scatter plot matrices, influence bar charts, radar charts, dimensionality reduction projection charts and other visual analysis charts to study the effects of different structural characteristic parameters on safety performance indicators (such as energy absorption, collision acceleration, mass, etc.), as well as the geometric constraints of the structural characteristic parameters and the interference risks after combination, so as to determine the structural characteristic parameters and their ranges of the energy-absorbing components, screen out the structural characteristic parameters with less influence, and obtain a table of key structural characteristic parameters, such as the structural characteristic parameters that have an impact on the top 15 or 20 safety performance indicators.

[0068] In this preferred embodiment, steps S21-S23 set the structural characteristic parameters according to the internal and external boundary constraint parameters, and select the key structural characteristic parameters according to the SGMW sensitivity analysis specification to ensure the importance of the selected parameters to the safety performance indicators; the sorting selection mechanism ensures that the most important parameters are given priority, thereby improving the safety and reliability of the design scheme; at the same time, it helps to reduce unnecessary complexity and make the design more focused on the factors that have the greatest impact on the overall performance.

[0069] Step S3: According to the preset parameterized model generation algorithm and mesh generation algorithm, each of the first characteristic parameter combinations is processed to obtain a finite element analysis model of each energy absorbing component, and according to the preset simulation algorithm, the preset non-energy absorbing component finite element analysis model and the finite element analysis model of each of the energy absorbing components are simulated to obtain various performance simulation results.

[0070] Preferably, in some embodiments of the present application, the finite element analysis model of each energy absorbing component described in step S3 can be obtained by the following preferred implementation, including steps S31-S32, as follows:

[0071] S31: extracting the first characteristic parameter combination according to a preset sampling algorithm to obtain a plurality of sampled first characteristic parameter combinations;

[0072] S32: processing each of the sampled first characteristic parameter combinations according to a preset parameterized model generation algorithm and a mesh generation algorithm to obtain a finite element analysis model of each energy absorbing component.

[0073] In specific implementation, sampling can be performed using various sampling methods such as optimized Latin hypercube and Hammersley sampling. After sampling, the computer-aided parametric design software (SFE Concept) is used to apply base points, baselines and sections to sweep, map and other methods to establish parametric models of the body collision energy absorption components of the developed model and the benchmark model (anti-collision beam, energy absorption box, front longitudinal beam, subframe, etc.). That is, in the SFE Concept software, firstly, the base point (Influence Point) of the collision energy absorption component is created to locate the position and length of the component, and then the baseline (Base Line) is generated by connecting between the base points for component sweeping, and the feature creation of the energy absorption component is realized by creating a base section (Base Section). The key collision energy absorption component is created after multiple base sections are inserted by sweeping the base section along the baseline; the interface between the components is established through the parametric model, and the rapid combination of the body energy absorption components is realized by using the rigid, patch and common node methods, and the input of the structural feature parameters is realized by defining the variables in the parametric model, thus shortening the model combination and adjustment time. In the parametric model, the variable recording function is used to record the component features, base point positions, component thickness and other information in the base section as corresponding variables, and the parameter variables are changed by modifying the recording parameters. In the SFE Concept software, the parametric model of the body collision energy absorption component is converted into a finite element model for collision simulation analysis through the mesh generation function. The connection between the body collision energy absorption component and the non-energy absorption component finite element model is established by using patch, common node and rigid methods at the boundary of the parametric model of the energy absorption component, and the finite element analysis model for integrated simulation process calculation is obtained. The integrated simulation analysis process is built using DOE and simcode modules in isight. By adding SFE Concept software to the simcode module to generate a mesh, adding finite element simulation software for finite element analysis, and realizing information transmission through data transmission between simcode modules, an integrated simulation analysis process for vehicle body collision is established, and integrated simulation is performed to obtain parameters such as hourglass energy Eh, mass increase parameter Addmass, energy absorption Ei, vehicle body acceleration peak MaxA, vehicle body effective acceleration EA, vehicle crushing amount Displacement, and vehicle body mass mass in the vehicle body collision performance simulation results. Among them, it is necessary to confirm that the ratio of hourglass energy to total energy is less than 5%, and the mass increase parameter is less than the standard value of 5% to ensure the normal operation of vehicle collision simulation.

[0074] In this preferred embodiment, steps S31-S32 sample the first characteristic parameter combination and use a parametric model generation algorithm and a mesh generation algorithm to construct a finite element analysis model, thereby solving the computing resource bottleneck problem that may be encountered when analyzing the full data. The sampling method not only retains the core characteristics of the original data, but also reduces the computing burden, making large-scale simulation possible and promoting more extensive design exploration and optimization.

[0075] Step S4: training a preset original neural network model according to the performance simulation results, the first characteristic parameter combination and the energy absorbing component combination data to obtain a design model.

[0076] Preferably, in some embodiments of the present application, step S4 can be obtained by the following preferred implementation, including steps S41-S43, as follows:

[0077] S41: constructing an original mapping data set according to each of the performance simulation results and the corresponding first characteristic parameter combination and the corresponding energy absorbing component combination.

[0078] S42: Processing the original mapping data set according to a preset preprocessing algorithm to obtain a mapping data set; wherein the preprocessing algorithm includes any one or more combinations of the following: missing value processing, outlier value processing, duplicate value processing, renaming processing, format unification processing, normalization processing, and feature encoding processing.

[0079] For example, in the specific implementation, after constructing the original mapping data set, the data set is cleaned and preprocessed. For missing values, such as sample data with no results due to calculation failure, they are deleted; for outliers, such as noise data with abnormal energy and acceleration during the calculation process, clustering algorithms or density-based anomaly detection algorithms are used to identify and delete sample data containing outliers, thereby improving data quality; for duplicate values, duplicate data is identified and deleted by comparing the feature values ​​between samples or using hash algorithms; at the same time, the sample data header is adjusted to adapt to the AI ​​artificial intelligence algorithm logic; different formats in the original data are converted to a unified format to ensure data consistency and comparability; converting the data to a distribution with a mean of 0 and a standard deviation of 1 helps ensure that the data has similar weights during the training process; for category feature data, label encoding, one-hot encoding, etc. are applied so that the machine learning model can understand and process them. Label encoding is used for naturally ordered categories, one-hot encoding is used for unordered categories, and frequency encoding is used for features with high frequencies.

[0080] In this preferred embodiment, steps S41-S42, based on constructing the original mapping data set, apply preprocessing algorithms (such as missing value processing, outlier processing, etc.) to significantly improve the quality of the data, eliminate noise interference, and enhance the effectiveness of model learning.

[0081] S43: training the original neural network model according to a preset K-fold cross validation algorithm, a dropout algorithm and the mapping data set to obtain the designed model;

[0082] The training adopts the preset AdamW algorithm to optimize the learning rate, and the algorithm is as follows:

[0083] m t =β1m t-1 +(1-β1)g t ; Where m t represents the first-order moment estimate of the current step t, g t represents the gradient of the current step, β1 and β2 represent the first-order and second-order momentum decay coefficients respectively, and v t represents the second-order moment estimate at time step t; where β1 and β2 are dynamically adjusted as follows: β1 = β 1,init (1-t / T); β2 = β 2,init ·(1-t / T); where β 1,init and β 2,init They represent the initial values ​​of the momentum decay coefficients of β1 and β2 respectively, and T is the total number of training steps; the gradient clipping algorithm is as follows: Where τ represents the gradient clipping threshold.

[0084] For example, in the specific implementation, taking the deep learning model as the original model to be trained, the data features of the training set are used as the input of the model, denoted as X={x1,x2,...,xn}, where xi represents the feature vector of the i-th sample; the predicted target value (collision energy absorption performance) of the training set is used as the output of the model, denoted as Y={y1,y2,...,yn}, where yi represents the actual energy absorption performance value of the i-th sample; the input X is calculated through each layer of the neural network, and finally the predicted output of the model is obtained. In the forward propagation of the neural network, the activation output of the lth hidden layer a(l) = f (l) (W (l) z (l-1) +b (l) ), where z (l-1) is the output of the previous layer, W (l) is the weight matrix of the lth layer, b (l) is the bias vector, f (l) is the activation function; the loss function can use the mean square error MSE to measure the model prediction value The gap between the true value Y and the true value Y; In order to prevent overfitting, the Dropout strategy is adopted to randomly discard some neurons during the training process to reduce the model’s dependence on specific neurons z(l) = Dropout(a (l) ,p), where p is the retention probability; K-fold cross-validation technology can be introduced to improve the generalization ability of the model. The training set is divided into K parts (for example, K=5), one of which is taken as the validation set each time, and the remaining K-1 parts are taken as the training set. K times of model training and evaluation are performed, and finally the average of the K evaluation results is taken as the final performance indicator of the model to improve the generalization ability of the model; the AdamW algorithm is used to optimize the learning rate, and the core of the algorithm is m t =β1m t-1 +(1-β1)g t ; Where m t represents the first-order moment estimate of the current step t, g t represents the gradient of the current step, β1 and β2 represent the first-order and second-order momentum decay coefficients respectively, and v t represents the second-order moment estimate at time step t; where β1 and β2 are dynamically adjusted as follows: β1 = β 1,init (1-t / T); β2 = β 2,init ·(1-t / T); where β 1,init and β 2,init They represent the initial values ​​of the momentum decay coefficients of β1 and β2 respectively, and T is the total number of training steps; the gradient clipping algorithm is as follows: Where τ represents the clipping threshold of the gradient. In addition, you can train multiple models with different structures, or use different machine learning algorithms to train multiple models, and fuse the prediction results of these models by weighted average or voting. For example, for weighted average, the fused prediction value It can be expressed as: Where m represents the number of models, represents the predicted value of the jth model, w j represents the weight of the jth model, and Σ j＝1 m w j =1, weights can be assigned based on the performance of the model on the validation set; finally, the test set is used to evaluate the performance of the final model, and real test data is fed back to the model to form a closed-loop optimization mechanism to continuously improve model performance.

[0085] In this preferred embodiment, step S43 uses K-fold cross validation, dropout and AdamW optimization algorithms to train the neural network model to avoid overfitting and ensure that the model has good generalization ability and stability; dynamically adjust the first-order and second-order momentum attenuation coefficients and the application of the gradient clipping algorithm further improve the convergence speed and stability during the model training process.

[0086] Step S5: Representing each energy absorbing component combination and key structural characteristic parameters as a number of population individuals of a genetic algorithm, and iteratively optimizing the population according to a preset genetic operation algorithm until a preset number of iterations is reached or the population individuals meet preset expected performance indicators, and outputting the current energy absorbing component combination and the corresponding key structural characteristic parameters as a design result; and in each iterative optimization, processing each population individual according to the design model to obtain predicted performance indicators of each population individual, and calculating the fitness value according to the deviation between each predicted performance indicator and the expected performance indicator.

[0087] Preferably, in some embodiments of the present application, the step S5 of calculating the fitness value according to the deviation between each of the predicted performance indicators and the expected performance indicator, wherein the calculation formula for calculating the fitness value is as follows:

[0088] F(i)=-MSE(y(i),y target ), where F(i) represents the fitness value of individual i in the population, y(i) represents the prediction performance index of the designed model, and y target represents the expected performance index, and MSE represents minimizing the mean square error.

[0089] For example, in the specific implementation, first clarify the goal to be optimized, that is, the expected performance indicators (for example, minimizing peak acceleration, maximizing energy absorption, etc.), and represent the combination of energy-absorbing components and key structural characteristic parameters as parameters for algorithm processing. Different combinations of components are encoded using binary coding, integer coding, or Gray coding. For example, assuming that there are 4 components and each component has 3 optional options, a 12-bit binary code can be used to represent a component combination, with every 3 bits representing the choice of a component; characteristic parameters: the structural characteristic parameters of each component (such as wall thickness, cross-sectional dimensions, etc.) are represented as continuous numerical variables; secondly, a genetic algorithm is set up to represent the combination coding and structural characteristic parameters of the energy-absorbing components as population individuals of the genetic algorithm. The initial population size is set to 100, that is, 100 sets of component combinations and structural feature parameter codes are randomly generated, each individual represents a potential design scheme, the maximum number of iterations is set to 500 times, the crossover rate is set to 0.8, and the mutation rate is set to 0.02. The above settings can ensure the diversity and search efficiency of the population, while balancing the computational cost and optimization accuracy; then, the performance of each individual is predicted using the pre-trained design model, and the individual's code and feature parameters are used as the input of the model. The output of the model is the predicted performance index value, and the prediction result is used to calculate the fitness value of the genetic algorithm. The fitness is calculated based on the deviation between the prediction result and the expected performance index, which is defined as F(i)=-MSE(y(i), y target), where F(i) represents the fitness value of individual i in the population, y(i) represents the prediction performance index of the designed model, and y target Represents the expected performance index, and guides the genetic algorithm to optimize the population by minimizing the mean square error (MSE); subsequently, the genetic algorithm iteratively optimizes the population through operations such as selection, crossover, and mutation, continuously improving the fitness of individuals in the population, and repeats the fitness value calculation and genetic operation steps until the maximum number of iterations is reached or the individuals in the population meet the expected performance index. When the individuals in the population meet the expected performance index, the corresponding energy absorption component combination code and structural characteristic parameters are output, thereby completing the optimal design of the vehicle body energy absorption component. In order to verify the effectiveness of the optimization results, the optimized design scheme can be simulated and verified, and the deviation between the simulation results and the expected performance indicators can be compared.

[0090] In this preferred embodiment, step S5 uses the genetic algorithm and the model in conjunction with the principle of evolution to automatically find the optimal solution without manual intervention, thereby further improving the design efficiency and ensuring the quality of the design results.

[0091] In summary, compared with the prior art, the above embodiments of the present application have the following beneficial effects: randomly generated combinations ensure the diversity of design schemes and avoid the influence of engineers' subjective experience; key structural feature parameters are set and specific parameter ranges are defined, so that subsequent simulation and analysis have clear goals and boundary conditions, reducing unnecessary waste of computing resources; the actual performance of different design schemes can be predicted in advance through simulation, reducing the cost and time consumption of physical experiments; the neural network model is trained in combination with key structural feature parameters and simulation results, and a large amount of simulation data is processed using AI technology to quickly establish a mapping relationship between input (structural features) and output (performance indicators), thereby improving the speed and accuracy of data analysis; the genetic algorithm works synergistically with the model, and combines the principles of evolution to automatically find the optimal solution without human intervention, further improving the design efficiency while ensuring the quality of the design results.

[0092] Embodiment 2:

[0093] Please refer to Figure 2 , a model-based collision energy absorption component design system disclosed in an embodiment of the present invention includes: a component combination module M1, a characteristic parameter setting module M2, a simulation module M3, a model training module M4 and a processing module M5;

[0094] The component combination module M1 is used to randomly generate combinations of energy absorbing components that meet preset combination rules according to the connection relationship of the energy absorbing components of each reference vehicle model.

[0095] The component combination module M1 includes: an information acquisition unit, a data conversion unit and a combination unit;

[0096] Wherein, the information acquisition unit is used to acquire information of each energy absorbing component of each reference vehicle model;

[0097] The data conversion unit is used to convert the information of each energy absorbing component into a numerical parameter according to a preset coding rule;

[0098] The combination unit is used to randomly generate a combination of the energy absorbing components according to the connection relationship of the energy absorbing components and a preset combination rule.

[0099] The component combination module M1 of this embodiment converts the energy-absorbing component information of each reference vehicle model into numerical parameters, and randomly generates component combinations based on connection relationships and combination rules, so that the energy-absorbing component designs between different vehicle models can be quickly transplanted and compared, reducing duplication of work; in addition, the use of unified coding rules simplifies the subsequent processing flow, improves the speed and accuracy of parametric modeling, thereby accelerating the entire design cycle and reducing development costs.

[0100] The characteristic parameter setting module M2 is used to set a number of key structural characteristic parameters for each energy absorbing component in each energy absorbing component combination according to the preset inner and outer boundary constraint parameters of the energy absorbing component spatial arrangement, and obtain a first characteristic parameter combination corresponding to each energy absorbing component combination.

[0101] The characteristic parameter setting module M2 includes: a parameter setting unit, an importance value calculation unit and a parameter screening unit;

[0102] Wherein, the parameter setting unit is used to set a number of structural characteristic parameters for each energy absorbing component according to the preset inner and outer boundary constraint parameters of the spatial arrangement of the energy absorbing component;

[0103] The importance value calculation unit is used to analyze and calculate the importance value of each of the structural characteristic parameters to the preset safety performance index according to the preset SGMW sensitivity analysis specification;

[0104] The parameter screening unit is used to sort the importance values ​​from large to small, and select several structural feature parameters corresponding to the importance values ​​ranked in front as key structural feature parameters.

[0105] The characteristic parameter setting module M2 of this embodiment sets the structural characteristic parameters according to the internal and external boundary constraint parameters, and optimizes the key structural characteristic parameters according to the SGMW sensitivity analysis specification to ensure the importance of the selected parameters to the safety performance indicators; the sorting selection mechanism ensures that the most important parameters are given priority, thereby improving the safety and reliability of the design scheme; at the same time, it helps to reduce unnecessary complexity and make the design more focused on the factors that have the greatest impact on the overall performance.

[0106] The simulation module M3 is used to process each of the first characteristic parameter combinations according to a preset parameterized model generation algorithm and a mesh generation algorithm to obtain a finite element analysis model of each energy-absorbing component, and to simulate a preset finite element analysis model of a non-energy-absorbing component and the finite element analysis model of each of the energy-absorbing components according to a preset simulation algorithm to obtain various performance simulation results.

[0107] The simulation module M3 includes: a sampling unit and a finite element model generation unit;

[0108] The sampling unit is used to extract the first characteristic parameter combination according to a preset sampling algorithm to obtain a plurality of sampled first characteristic parameter combinations;

[0109] The finite element model generation unit is used to process each of the sampled first characteristic parameter combinations according to a preset parameterized model generation algorithm and a grid generation algorithm to obtain a finite element analysis model of each energy absorbing component.

[0110] The simulation module M3 of this embodiment samples the first characteristic parameter combination and uses a parametric model generation algorithm and a mesh generation algorithm to construct a finite element analysis model, thereby solving the computing resource bottleneck problem that may be encountered when analyzing the full data. The sampling method not only retains the core characteristics of the original data, but also reduces the computing burden, making large-scale simulation possible and promoting more extensive design exploration and optimization.

[0111] The model training module M4 is used to train a preset original neural network model according to the performance simulation results, the first characteristic parameter combination and the energy absorbing component combination data to obtain a design model.

[0112] The model training module M4 includes: a data set construction unit and a preprocessing unit;

[0113] The data set construction unit is used to construct an original mapping data set according to each of the performance simulation results and the corresponding first characteristic parameter combination and the corresponding energy absorbing component combination.

[0114] The preprocessing unit is used to process the original mapping data set according to a preset preprocessing algorithm to obtain a mapping data set; wherein the preprocessing algorithm includes any one or more combinations of the following: missing value processing, outlier value processing, duplicate value processing, renaming processing, format unification processing, normalization processing, and feature encoding processing.

[0115] The model training module M4 of this embodiment, based on constructing the original mapping data set, applies preprocessing algorithms (such as missing value processing, outlier processing, etc.) to significantly improve the quality of data, eliminate noise interference, and enhance the effectiveness of model learning.

[0116] The model training module M4 further includes: a training unit;

[0117] Wherein, the training unit is used to train the original neural network model according to the preset K-fold cross validation algorithm, dropout algorithm and the mapping data set to obtain the designed model;

[0118] The training adopts the preset AdamW algorithm to optimize the learning rate, and the algorithm is as follows:

[0119] m t =β1m t-1 +(1-β1)g t ; Where m t represents the first-order moment estimate of the current step t, g t represents the gradient of the current step, β1 and β2 represent the first-order and second-order momentum decay coefficients respectively, and v t represents the second-order moment estimate at time step t; where β1 and β2 are dynamically adjusted as follows: β1 = β 1,init (1-t / T); β2 = β 2,init ·(1-t / T); where β 1,init and β 2,init They represent the initial values ​​of the momentum decay coefficients of β1 and β2 respectively, and T is the total number of training steps; the gradient clipping algorithm is as follows: Where τ represents the gradient clipping threshold.

[0120] The model training module M4 of this embodiment uses K-fold cross validation, dropout and AdamW optimization algorithm to train the neural network model to avoid overfitting and ensure that the model has good generalization ability and stability; the dynamic adjustment of the first-order and second-order momentum attenuation coefficients and the application of the gradient clipping algorithm further improve the convergence speed and stability during the model training process.

[0121] The processing module M5 is used to characterize each energy absorbing component combination and key structural characteristic parameters as a number of population individuals of a genetic algorithm, and iteratively optimize the population according to a preset genetic operation algorithm until a preset number of iterations is reached or the population individuals meet preset expected performance indicators, and output the current energy absorbing component combination and the corresponding key structural characteristic parameters as a design result; and in each iterative optimization, various population individuals are processed according to the design model to obtain predicted performance indicators of various population individuals, and the fitness value is calculated according to the deviation between each predicted performance indicator and the expected performance indicator.

[0122] The processing module M5 includes: a fitness value calculation unit;

[0123] The fitness value calculation unit is used to calculate the fitness value, and the calculation formula is as follows:

[0124] F(i)=-MSE(y(i),y target ), where F(i) represents the fitness value of individual i in the population, y(i) represents the prediction performance index of the designed model, and y target represents the expected performance index, and MSE represents minimizing the mean square error.

[0125] The processing module M5 of this embodiment uses the genetic algorithm and the model to work together, combining the principle of evolution to automatically find the optimal solution without manual intervention, thereby further improving the design efficiency and ensuring the quality of the design results.

[0126] In summary, compared with the prior art, the above embodiments of the present application have the following beneficial effects: randomly generated combinations ensure the diversity of design schemes and avoid the influence of engineers' subjective experience; key structural feature parameters are set and specific parameter ranges are defined, so that subsequent simulation and analysis have clear goals and boundary conditions, reducing unnecessary waste of computing resources; the actual performance of different design schemes can be predicted in advance through simulation, reducing the cost and time consumption of physical experiments; the neural network model is trained in combination with key structural feature parameters and simulation results, and a large amount of simulation data is processed using AI technology to quickly establish a mapping relationship between input (structural features) and output (performance indicators), thereby improving the speed and accuracy of data analysis; the genetic algorithm works synergistically with the model, and combines the principles of evolution to automatically find the optimal solution without human intervention, further improving the design efficiency while ensuring the quality of the design results.

[0127] The division of the modules described above is only a logical function division. There may be other division methods in actual implementation. For example, multiple modules may be combined or integrated into another system.

[0128] Embodiment three:

[0129] Figure 3 The structure diagram of a model-based collision energy absorption component design device of the present application is shown. Figure 3 As shown, the model-based collision energy absorbing component design device may include: a processor N1, a memory N2, a data interface N3 and a communication bus N4.

[0130] Among them: the processor N1, the memory N2, and the data interface N3 communicate with each other through the communication bus N4; the data interface N3 is used for data communication with other devices such as input devices or output devices; the processor N1 is used to execute the program N5, which can specifically execute the relevant steps in the above-mentioned model-based collision energy absorption component design method embodiment.

[0131] Specifically, the program N5 may include program codes, and the program codes include computer executable instructions.

[0132] The processor N1 may be a central processing unit (CPU), or an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present application. The one or more processors included in the model-based collision energy absorption component design device may be processors of the same type, such as one or more CPUs, or may be processors of different types, such as one or more CPUs and one or more ASICs.

[0133] The memory N2 is used to store the program N5. The memory N2 may include a high-speed RAM memory, and may also include a non-volatile memory (non-volatile memory), such as at least one disk memory.

[0134] The algorithms or displays provided herein are not inherently related to any particular computer, virtual system or other device. In addition, the embodiments of the present application are not directed to any particular programming language.

[0135] Embodiment 4:

[0136] An embodiment of the present invention also provides a computer-readable storage medium, which stores at least one executable instruction. When the executable instruction is executed on a model-based collision energy absorbing component design device / system, the model-based collision energy absorbing component design device / system executes the model-based collision energy absorbing component design method in any of the above-mentioned method embodiments.

[0137] In the specification provided herein, a large number of specific details are described. However, it is understood that the embodiments of the present application can be practiced without these specific details. Similarly, in order to simplify the application and help understand one or more of the various inventive aspects, in the above description of the exemplary embodiments of the present application, the various features of the embodiments of the present application are sometimes grouped together into a single embodiment, figure, or description thereof. Wherein, the claims that follow the specific embodiment are thus clearly incorporated into the specific embodiment, wherein each claim itself is used as a separate embodiment of the application.

[0138] Those skilled in the art will appreciate that the modules in the devices in the embodiments may be adaptively changed and arranged in one or more devices different from the embodiments. The modules or units or components in the embodiments may be combined into one module or unit or component, and further may be divided into a plurality of submodules or subunits or subcomponents. Except that at least some of such features and / or processes or units are mutually exclusive.

Claims

1. A model-based collision energy absorption component design method, characterized in that: include: According to the connection relationship of each energy absorbing component of each reference vehicle model, randomly generate a combination of each energy absorbing component that meets the preset combination rules; According to the preset inner and outer boundary constraint parameters of the energy absorbing component spatial arrangement, a number of key structural characteristic parameters are set for each energy absorbing component in each energy absorbing component combination to obtain a first characteristic parameter combination corresponding to each energy absorbing component combination; According to a preset parameterized model generation algorithm and a mesh generation algorithm, each of the first characteristic parameter combinations is processed to obtain a finite element analysis model of each energy absorbing component, and according to a preset simulation algorithm, a preset finite element analysis model of a non-energy absorbing component and the finite element analysis model of each of the energy absorbing components are simulated to obtain various performance simulation results; According to the performance simulation results, the first characteristic parameter combination and the energy absorbing component combination data, a preset original neural network model is trained to obtain a design model; Representing each energy absorbing component combination and key structural characteristic parameters as a number of population individuals of a genetic algorithm, and iteratively optimizing the population according to a preset genetic operation algorithm until a preset number of iterations is reached or the population individuals meet preset expected performance indicators, and outputting the current energy absorbing component combination and the corresponding key structural characteristic parameters as a design result; In each iterative optimization, various population individuals are processed according to the design model to obtain the predicted performance indicators of various population individuals, and the fitness value is calculated according to the deviation between each predicted performance indicator and the expected performance indicator.

2. A model-based collision energy absorption component design method according to claim 1, characterized in that: The method of randomly generating a combination of energy absorbing components that meets a preset combination rule according to the connection relationship of the energy absorbing components of each reference vehicle model includes: Obtaining information of each energy absorbing component of each reference vehicle model; According to a preset coding rule, the information of each energy absorbing component is converted into a numerical parameter; According to the connection relationship of each energy absorbing component and a preset combination rule, each combination of the energy absorbing components is randomly generated.

3. A model-based collision energy absorption component design method as claimed in claim 2, characterized in that: According to the preset inner and outer boundary constraint parameters of the spatial arrangement of the energy absorbing components, a number of key structural characteristic parameters are set for each energy absorbing component in each energy absorbing component combination to obtain a first characteristic parameter combination corresponding to each energy absorbing component combination, including: According to the preset inner and outer boundary constraint parameters of the energy absorbing component spatial arrangement, a number of structural characteristic parameters are set for each energy absorbing component; According to the preset SGMW sensitivity analysis specification, analyze and calculate the importance value of each structural characteristic parameter to the preset safety performance index; The importance values ​​are sorted from large to small, and a number of structural feature parameters corresponding to the importance values ​​ranked in front are selected as key structural feature parameters.

4. A model-based collision energy absorption component design method as claimed in claim 3, characterized in that: The method of processing each of the first characteristic parameter combinations according to a preset parameterized model generation algorithm and a mesh generation algorithm to obtain a finite element analysis model of each energy absorbing component includes: Extracting the first characteristic parameter combination according to a preset sampling algorithm to obtain a plurality of sampled first characteristic parameter combinations; According to a preset parameterized model generation algorithm and a mesh generation algorithm, each of the sampled first characteristic parameter combinations is processed to obtain a finite element analysis model of each energy absorbing component.

5. A model-based collision energy absorption component design method according to claim 4, characterized in that: The method of training a preset original neural network model based on the performance simulation results, the first characteristic parameter combination and the energy absorbing component combination data to obtain a design model includes: Constructing an original mapping data set according to each of the performance simulation results and the corresponding first characteristic parameter combination and the corresponding energy absorbing component combination; According to a preset preprocessing algorithm, the original mapping data set is processed to obtain a mapping data set; wherein the preprocessing algorithm includes any one or more combinations of the following: missing value processing, outlier value processing, duplicate value processing, renaming processing, format unification processing, normalization processing, and feature encoding processing.

6. A model-based collision energy absorption component design method according to claim 5, characterized in that: The method of training a preset original neural network model according to the performance simulation results, the first characteristic parameter combination and the energy absorbing component combination data to obtain a design model further includes: According to the preset K-fold cross validation algorithm, dropout algorithm and the mapping data set, the original neural network model is trained to obtain the designed model; The training adopts the preset AdamW algorithm to optimize the learning rate, and the algorithm is as follows: m t =β1m t-1 +(1-β1)g t ; Where m t represents the first-order moment estimate of the current step t, g t represents the gradient of the current step, β1 and β2 represent the first-order and second-order momentum decay coefficients respectively, and v t represents the second-order moment estimate at time step t; where β1 and β2 are dynamically adjusted as follows: β1 = β 1,init (1-t / T); β2 = β 2,init ·(1-t / T); where β 1,init and β 2,init They represent the initial values ​​of the momentum decay coefficients of β1 and β2 respectively, and T is the total number of training steps; the gradient clipping algorithm is as follows: Where τ represents the gradient clipping threshold.

7. A model-based collision energy absorption component design method according to claim 6, characterized in that: The calculating of the fitness value according to the deviation between each of the predicted performance indicators and the expected performance indicator comprises: The calculation formula for calculating the fitness value is as follows: F(i)=-MSE(y(i),y target ), where F(i) represents the fitness value of individual i in the population, y(i) represents the prediction performance index of the designed model, and y target represents the expected performance index, and MSE represents minimizing the mean square error.

8. A model-based collision energy absorption component design system, characterized in that: include: Component assembly module, characteristic parameter setting module, simulation module, model training module and processing module; The component combination module is used to randomly generate a combination of energy absorbing components that meets a preset combination rule according to the connection relationship of the energy absorbing components of each reference vehicle model; The characteristic parameter setting module is used to set a number of key structural characteristic parameters for each energy absorbing component in each energy absorbing component combination according to the preset inner and outer boundary constraint parameters of the energy absorbing component spatial arrangement, so as to obtain a first characteristic parameter combination corresponding to each energy absorbing component combination; The simulation module is used to process each of the first characteristic parameter combinations according to a preset parameterized model generation algorithm and a mesh generation algorithm to obtain a finite element analysis model of each energy absorbing component, and simulate a preset finite element analysis model of a non-energy absorbing component and the finite element analysis model of each of the energy absorbing components according to a preset simulation algorithm to obtain various performance simulation results; The model training module is used to train a preset original neural network model according to the performance simulation results, the first characteristic parameter combination and the energy absorbing component combination data to obtain a design model; The processing module is used to characterize each energy absorbing component combination and key structural characteristic parameters as a number of population individuals of a genetic algorithm, and iteratively optimize the population according to a preset genetic operation algorithm until a preset number of iterations is reached or the population individuals meet preset expected performance indicators, and output the current energy absorbing component combination and the corresponding key structural characteristic parameters as a design result; and in each iterative optimization, various population individuals are processed according to the design model to obtain predicted performance indicators of various population individuals, and the fitness value is calculated according to the deviation between each predicted performance indicator and the expected performance indicator.

9. A model-based collision energy absorption component design device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the computer program is loaded into a processor, the steps of the model-based collision energy absorption component design method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the model-based collision energy absorption component design method according to any one of claims 1 to 7 are implemented.

Images