A honeycomb wall barrier modeling method and prediction method fusing physical mechanisms

Abstract

This invention proposes a cellular barrier modeling and prediction method that integrates physical mechanisms, relating to the field of automotive safety assessment technology. Addressing the low accuracy problem in existing cellular barrier modeling techniques, this invention obtains the nominal geometric parameters of the target cellular barrier; performs CT scanning of the cellular barrier sample and 3D reconstruction to obtain actual geometric correction values; combines the nominal geometric parameters and actual geometric correction values ​​to establish an initial minimum repeatable element, and inversely calibrates the material parameters to obtain a corrected minimum repeatable element; simulates and extracts strain curves corresponding to strain rates, constructing a constitutive relation library; establishes an initial homogeneous material model of the target cellular barrier; based on the strain rate under operating conditions in the simulation environment, selects a matching strain curve from the constitutive relation library and inputs it into the initial homogeneous material model to obtain a corrected homogeneous material model; defines a gas domain and its leakage characteristics within the corrected homogeneous material model to obtain a gas-solid coupled homogeneous model. The cellular barrier modeling method of this invention has high accuracy.

Description

Technical Field

[0001] This invention relates to the field of automotive safety assessment technology, and in particular to a cellular barrier modeling and prediction method that integrates physical mechanisms. Background Technology

[0002] Cellular barriers are core energy-absorbing components in automotive passive safety systems. With their high specific strength, high specific stiffness, and excellent energy absorption characteristics, they effectively buffer impact loads and absorb collision energy during a car collision, significantly reducing the risk of occupant injury. They are a key structure for ensuring vehicle collision safety. In Computer-Aided Engineering (CAE) simulations, accurate modeling of cellular barriers is a crucial step in automotive crashworthiness design and verification, directly determining the reliability of collision simulation results.

[0003] In existing technologies, modeling methods for cellular barriers mainly fall into two categories: one is the refined microscopic model, which accurately reconstructs each cellular cell using shell elements. While this achieves high-fidelity simulation, the computational load is enormous and the cost prohibitive. The other is the macroscopic homogeneous model, which models the cellular structure as a continuous homogeneous material. Although this improves computational efficiency, its modeling accuracy relies entirely on stress-volume-strain curves obtained through full-scale physical experiments, limiting its ability to describe the dynamic strengthening of materials under high strain rates. All existing homogeneous models completely ignore the influence of enclosed air within the cellular cells during the dynamic crushing process, and the modeling process's limited ability to describe the dynamic strengthening of materials under high strain rates results in severely insufficient simulation accuracy for cellular barrier models.

[0004] Therefore, developing a cellular barrier modeling and prediction method that integrates physical mechanisms is of great significance for improving the accuracy of cellular barrier modeling and the precision of prediction based on cellular barrier models. Summary of the Invention

[0005] To address the problem that existing homogeneous models completely ignore the influence of enclosed air within honeycomb cells during the dynamic crushing process, resulting in severely insufficient simulation accuracy of honeycomb barrier models, this invention proposes a honeycomb barrier modeling method that incorporates physical mechanisms, specifically including the following steps:

[0006] S1. Obtain the nominal geometric parameters of the target cellular barrier;

[0007] S2. Perform industrial computed tomography and 3D reconstruction on the honeycomb barrier sample to obtain the actual geometric correction value;

[0008] S3. Combining the nominal geometric parameters and the actual geometric correction values, establish the initial minimum repeatable unit of the target honeycomb barrier, and perform inversion calibration on the material parameters in the initial minimum repeatable unit to obtain the corrected minimum repeatable unit.

[0009] S4. Perform dynamic compression simulations of the modified minimum repeatable element at different strain rates, extract the dynamic stress-volume strain curves corresponding to each strain rate, and construct a constitutive relation library.

[0010] S5. In the whole vehicle collision simulation environment, establish the initial homogeneous material model of the target honeycomb barrier;

[0011] S6. Based on the strain rate of the working conditions in the whole vehicle collision simulation environment, select the matching dynamic stress-volume strain curve from the constitutive relation library, and input the dynamic stress-volume strain curve into the initial homogeneous material model to obtain the modified homogeneous material model.

[0012] S7. Define at least one gas domain within the modified homogeneous material model, and define the leakage characteristics of the gas domain to obtain a gas-solid coupled homogeneous model.

[0013] Furthermore, industrial computed tomography (CT) scanning and 3D reconstruction are performed on the cellular barrier sample to obtain actual geometric correction values. This includes: performing industrial CT scanning on the cellular barrier sample to obtain tomographic data; establishing a 3D digital model of the cellular barrier sample based on the tomographic data; and obtaining actual geometric correction values ​​based on the 3D digital model. The actual geometric correction values ​​include correction values ​​for single-layer wall width, double-layer wall width, and bending radius correction values ​​at cell wall junctions.

[0014] Furthermore, the material parameters in the initial minimum repeatable unit are inverted and calibrated to obtain the modified minimum repeatable unit, including: performing quasi-static compression simulation on the initial minimum repeatable unit to obtain the static stress-volume-strain simulation curve; performing quasi-static physical compression experiment on the honeycomb barrier sample to obtain the static stress-volume-strain experimental curve; constructing an objective function with the goal of minimizing the difference between the static stress-volume-strain simulation curve and the static stress-volume-strain experimental curve, and iteratively modifying the material parameters in the initial minimum repeatable unit through the objective spectral function to obtain the modified minimum repeatable unit.

[0015] Furthermore, the objective function is the root mean square error function between the static stress-volume strain simulation curve and the static stress-volume strain experimental curve.

[0016] Furthermore, in the vehicle collision simulation environment, an initial homogeneous material model of the target cellular barrier is established, including: creating a homogeneous solid element mesh of the target cellular barrier; assigning a homogeneous material model to the homogeneous solid element mesh, and establishing the initial homogeneous material model of the target cellular barrier.

[0017] Furthermore, based on the strain rate under the working conditions in the whole vehicle collision simulation environment, a matching dynamic stress-volume strain curve is selected from the constitutive relation library, including: selecting the dynamic stress-volume strain curve corresponding to the same or closest strain rate from the constitutive relation library based on the strain rate under the working conditions in the whole vehicle collision simulation environment.

[0018] Furthermore, defining at least one gas domain within the modified homogeneous material model includes: constructing a gas state equation based on the adiabatic compression law of an ideal gas to characterize the pressure evolution relationship of the gas domain from its initial state to the compression process; the gas domain covers the entire volume of the target honeycomb barrier or is partitioned according to structural characteristics.

[0019] Furthermore, defining the leakage characteristics of the gas domain includes characterizing the leakage characteristics of the gas domain by time-varying leakage area or leakage rate related to strain rate.

[0020] Furthermore, after obtaining the gas-solid coupled homogeneous model, the process also includes: performing simulations based on the gas-solid coupled homogeneous model to output a dynamic crushing response that includes gas effects.

[0021] The present invention also provides a prediction method, comprising:

[0022] Based on the above honeycomb barrier modeling method, gas-solid coupling homogeneous models corresponding to various specifications of honeycomb barriers are established as training base models.

[0023] Dynamic axial crushing simulation is performed on the training base model to generate a comprehensive training database; the comprehensive training database includes the initial finite element mesh, force-displacement response time series curves, and full-field mesh deformation time series sequence.

[0024] A voxelization strategy based on node quality is adopted to voxelize the initial finite element mesh, converting the initial finite element mesh of each training base model into a normalized three-dimensional voxel matrix.

[0025] The three-dimensional voxel matrix is ​​reduced in dimensionality and its features are extracted by using a convolutional neural network to obtain low-dimensional features.

[0026] A first long short-term memory neural network is constructed using low-dimensional features and time series as inputs and force-displacement response time series curves as outputs; a second long short-term memory neural network is constructed using low-dimensional features and time series as inputs and full-field mesh deformation time series as outputs.

[0027] Based on the comprehensive training database, the first long short-term memory neural network and the second long short-term memory neural network are trained to obtain the trained prediction network;

[0028] For a new target cellular barrier, voxelization and feature extraction are performed according to the above steps. The resulting low-dimensional features are then input into the prediction network to quickly predict the dynamic response of the cellular barrier.

[0029] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0030] This invention obtains the nominal geometric parameters of a target honeycomb barrier; performs industrial computed tomography (CT) scanning and 3D reconstruction on the honeycomb barrier sample to obtain actual geometric correction values; combines the nominal geometric parameters and actual geometric correction values ​​to establish an initial minimum repeatable element for the target honeycomb barrier, and performs inversion calibration on the material parameters in the initial minimum repeatable element to obtain a modified minimum repeatable element; performs dynamic compression simulations on the modified minimum repeatable element under different strain rates, extracts the dynamic stress-volume strain curve corresponding to each strain rate, and constructs a constitutive relation library; establishes an initial homogeneous material model of the target honeycomb barrier in a vehicle collision simulation environment; selects a matching dynamic stress-volume strain curve from the constitutive relation library according to the strain rate under the working conditions in the vehicle collision simulation environment, and inputs the dynamic stress-volume strain curve into the initial homogeneous material model to obtain a modified homogeneous material model; defines at least one gas domain inside the modified homogeneous material model, and defines the leakage characteristics of the gas domain to obtain a gas-solid coupled homogeneous model. By defining a gas domain covering the entire or a section of the honeycomb barrier, dynamic coupling between the mechanical behavior of the solid skeleton and the evolution of internal gas pressure is achieved. This design recreates the complete physical process of dynamic crushing of a honeycomb barrier from a physical perspective, systematically making up for the core defect of traditional models that ignore the internal gas effect, and achieving a leapfrog improvement in simulation accuracy under dynamic collision conditions.

[0031] Furthermore, by constructing an initial minimum repeatable element combined with geometric corrections from industrial computed tomography (CT) scans, and replacing traditional full-scale physical experiment calibration with simulation inversion calibration using this micro-minimum repeatable element, a corrected minimum repeatable element is formed. This method eliminates the need for full-scale experiments, allowing parameter calibration to be completed solely through micro-element simulations and small-scale sample experiments. This significantly reduces experimental and time costs, while avoiding errors caused by structural scale effects in full-scale experiments. It makes material parameters more closely resemble actual working conditions, providing a high-precision foundation for the subsequent construction of a wide-strain-rate constitutive relation library, thus ensuring the reliability of the entire simulation model from the outset. Attached Figure Description

[0032] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0033] Figure 1 This is a flowchart of a cellular barrier modeling method that integrates physical mechanisms, provided by an embodiment of the present invention;

[0034] Figure 2 This is a schematic diagram of the structure of the initial minimum repeatable unit provided in the embodiment of the present invention;

[0035] Figure 3 This is a flowchart of a prediction method provided in an embodiment of the present invention. Detailed Implementation

[0036] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0037] The specific embodiments of the present invention will be described below.

[0038] To address the low accuracy issue in existing honeycomb barrier modeling techniques, this invention obtains the nominal geometric parameters of the target honeycomb barrier; performs CT scanning and 3D reconstruction of the honeycomb barrier sample to obtain actual geometric correction values; combines the nominal geometric parameters and actual geometric correction values ​​to establish an initial minimum repeatable element, and inversely calibrates the material parameters to obtain a corrected minimum repeatable element; simulates and extracts the strain curves corresponding to the strain rates, constructing a constitutive relation library; establishes an initial homogeneous material model of the target honeycomb barrier; based on the strain rate under operating conditions in the simulation environment, selects a matching strain curve from the constitutive relation library and inputs it into the initial homogeneous material model to obtain a corrected homogeneous material model; defines a gas domain and its leakage characteristics within the corrected homogeneous material model to obtain a gas-solid coupled homogeneous model. The honeycomb barrier modeling method of this invention has high accuracy.

[0039] Example 1

[0040] Figure 1 This is a flowchart of a cellular barrier modeling method that integrates physical mechanisms, as provided in an embodiment of the present invention. Figure 1 As shown, the modeling method specifically includes the following steps:

[0041] S1. Obtain the nominal geometric parameters of the target cellular barrier.

[0042] A target cellular barrier refers to a specific specification of cellular aluminum structure to be modeled, simulated, or its performance predicted; it is a core energy-absorbing element in automotive passive safety systems. Nominal geometric parameters refer to the standard geometric parameters specified during the design, production, or shipment of the cellular barrier. These are the basic framework parameters for building the model and include three key indicators: the critical dimensions of the smallest repeating cell in the cellular barrier structure, the design thickness of a single cell wall, and the overall height of the cellular barrier.

[0043] The application scenario and product specifications of the cellular barrier to be modeled are determined. Technical documents such as design files and factory parameter lists for this specification are retrieved, and key indicators of the target cellular barrier are extracted to form the basic data for modeling. Nominal geometric parameters provide a unified basic framework for constructing the smallest repeatable unit, avoiding disordered geometric shapes during modeling and ensuring the repeatability and engineering adaptability of the model. Furthermore, by obtaining the nominal geometric parameters of the target cellular barrier, it is unnecessary to design the geometric structure from scratch; an initial model can be built directly based on the nominal parameters, significantly reducing the modeling cycle, lowering modeling complexity, and improving efficiency.

[0044] S2. Perform industrial computed tomography and 3D reconstruction on the honeycomb barrier sample to obtain the actual geometric correction value.

[0045] Specifically, this includes: performing industrial computed tomography (CT) scans on the cellular barrier sample to obtain tomographic data; establishing a three-dimensional digital model of the cellular barrier sample based on the tomographic data; and obtaining actual geometric correction values ​​based on the three-dimensional digital model, including correction values ​​for single-layer wall width, double-layer wall width, and bending radius at cell wall junctions.

[0046] A cellular barrier sample refers to a real cellular aluminum structural component that matches the specifications of the target cellular barrier. It serves as a physical carrier for industrial CT scanning and data acquisition. Industrial computed tomography (CT) is a non-destructive testing technology that uses X-rays to penetrate the sample and acquire multi-directional tomographic image data. It can accurately reconstruct the internal and external geometry of the sample. Industrial CT differs from medical CT in that it focuses on the structural details and dimensional accuracy measurement of industrial components. Based on the continuous tomographic data obtained from industrial CT scans, a three-dimensional digital model of the sample is constructed using computer algorithms, achieving a digital reconstruction of the real cellular structure, making its geometric features measurable and analyzable. The actual geometric correction value refers to the parameters obtained through measurement and statistics of the three-dimensional digital model. It is used to correct geometric deviations caused by manufacturing process variations in the idealized Y-cell (i.e., the smallest repeatable cell of the cellular barrier) model constructed based on nominal geometric parameters, ensuring that the model's geometric features are highly consistent with the real cellular structure. The single-wall width correction value refers to the actual width of a single cell wall in a real honeycomb. The double-wall width correction value refers to the actual width of the overlapping part of the double-wall junction in a real honeycomb. The bending radius correction value at the cell wall junction refers to the actual bending radius of the junction of cell walls in a real honeycomb.

[0047] A real honeycomb aluminum component with specifications completely identical to the target honeycomb barrier was selected as a sample to ensure that the material and manufacturing process of the sample were consistent with the target barrier, avoiding distortion of correction values ​​due to sample differences. The sample was fixed on an industrial CT inspection platform, and the scanning equipment was started. X-rays penetrated the sample and collected tomographic image data of different sections, covering the overall height and width range of the sample to ensure no geometric features were missed. The continuous tomographic data obtained from the scan was imported into professional 3D reconstruction software. Through algorithmic stitching and volume data visualization processing, a 1:1 3D digital model corresponding to the sample was constructed, completely restoring the internal and external geometry of the actual honeycomb structure. In the 3D digital model, key geometric features were accurately measured. Multiple representative sections were selected, and the values ​​of single-layer wall width, double-layer wall width, and bending radius at the connection were statistically analyzed. The average value was taken as the final actual geometric correction value to ensure the statistical reliability of the correction value. Traditional models construct idealized units based on nominal geometric parameters, ignoring details such as cell wall width deviations and bending at the connection that exist in actual production, resulting in a discrepancy between the geometry and the real structure, directly amplifying the dynamic simulation error. This embodiment obtains the geometric correction values ​​of the real structure through industrial CT scanning and 3D reconstruction, making the subsequently constructed Y-unit model closer to reality and greatly improving geometric accuracy.

[0048] S3. Combining the nominal geometric parameters and the actual geometric correction values, establish the initial minimum repeatable element of the target honeycomb barrier, and perform inversion calibration on the material parameters in the initial minimum repeatable element to obtain the corrected minimum repeatable element.

[0049] The initial minimum repeatable unit is the minimum repeating geometric unit of the honeycomb structure, such as... Figure 2 As shown, the initial minimum repeatable element (Y element) consists of three cell walls with an included angle of 120°. It is an initial high-fidelity finite element model constructed based on nominal geometric parameters and actual geometric correction values, without material parameter calibration. Material parameters refer to the constitutive model parameters of the aluminum foil material in the initial minimum repeatable element, specifically including initial yield stress, tangent modulus, Young's modulus, etc., which directly affect the simulation accuracy of the model's mechanical response. Inversion calibration is a mathematical parameter optimization method. By comparing simulation results with physical experimental results, and aiming to minimize the difference between the two, the material parameters are iteratively corrected in reverse to ensure that the model's simulation accuracy meets the preset requirements. The corrected minimum repeatable element is the Y element model whose material parameters accurately match the actual material properties after inversion calibration. It is the core carrier for subsequently constructing a wide strain rate constitutive relation library.

[0050] Specifically, the material parameters in the initial minimum repeatable element are inverted and calibrated to obtain the modified minimum repeatable element. This includes: performing quasi-static compression simulation on the initial minimum repeatable element to obtain the static stress-volume-strain simulation curve; performing quasi-static physical compression experiments on the honeycomb barrier sample to obtain the static stress-volume-strain experimental curve; constructing an objective function with the goal of minimizing the difference between the static stress-volume-strain simulation curve and the static stress-volume-strain experimental curve; and iteratively modifying the material parameters in the initial minimum repeatable element through the objective spectral function to obtain the modified minimum repeatable element. For example, the objective function can be the root mean square error function between the static stress-volume-strain simulation curve and the static stress-volume-strain experimental curve.

[0051] Quasi-static compression simulation refers to the simulation performed in finite element software at a low strain rate (e.g., 10). -3 s -1 Compression simulation is performed on the initial minimum repeatable element to recreate the mechanical behavior under slow compression conditions. The static stress-volume-strain simulation curve refers to the macroscopic mechanical response curve output by the quasi-static compression simulation, with volumetric strain on the horizontal axis and engineering stress on the vertical axis, reflecting the mechanical properties of the model. The static stress-volume-strain experimental curve refers to the stress-volume-strain relationship curve obtained by actual measurement after a quasi-static physical compression experiment on the honeycomb barrier sample; it serves as the standard for verifying the accuracy of the simulation model. The objective function is a mathematical expression used to quantify the difference between the simulation curve and the experimental curve; its core function is to provide optimization direction for iterative parameter correction.

[0052] Quasi-static compression simulation was performed on the initial Y element in finite element software, setting a low strain rate condition, and calculating and outputting the static stress-volume strain simulation curve. A quasi-static physical compression experiment was conducted using a cellular barrier sample of the same specifications as the target cellular barrier. Stress and volumetric strain data during the compression process were collected using sensors, generating the static stress-volume strain experimental curve. With minimizing the difference between the simulation curve and the experimental curve as the core objective, a root mean square error function was constructed as the objective function to quantify the degree of deviation between the two sets of curves. An optimization algorithm, such as a genetic algorithm, was used to iteratively adjust the material parameters of the initial Y element, aiming to minimize the objective function value. After each round of correction, the quasi-static simulation was performed again, and the objective function value was calculated. This iterative process was repeated until the objective function value met the preset tolerance, i.e., the simulation curve and the experimental curve were highly consistent. At this point, the minimum repeatable element with accurately corrected material parameters was output.

[0053] By replacing traditional full-scale physical experiment calibration with simulation inversion calibration using microscopic minimum repeatable elements, a modified minimum repeatable element is formed. This replaces the inefficient mode of traditional full-scale physical experiment calibration, eliminating the need for costly and time-consuming full-scale experiments. Accurate material parameters can be obtained solely through microscopic Y-element simulation and small-scale sample experiments, significantly reducing economic and time costs. Through a closed-loop process of simulation, experimental comparison, and iterative correction, the material parameters are highly matched with the actual aluminum foil material properties, avoiding parameter distortion caused by traditional empirical values ​​or single experimental calibration, and providing accurate mechanical input for subsequent modeling.

[0054] S4. Perform dynamic compression simulations on the modified minimum repeatable element under different strain rates, extract the dynamic stress-volume strain curves corresponding to each strain rate, and construct a constitutive relation library.

[0055] Strain rate is a physical quantity that describes the rate of deformation of a material or structure, and its unit is s (seconds). -1 The coverage area is quasi-static (10 - 3 s -1 ) to high strain rate (10 3 s -1 This covers typical working conditions during automotive collisions. Dynamic compression simulation refers to simulating compression conditions at different strain rates in finite element software, restoring and correcting the mechanical response of the smallest repeatable element under dynamic loads. The dynamic stress-volume strain curve is the core mechanical response curve output by the dynamic compression simulation. The horizontal axis represents volumetric strain, and the vertical axis represents engineering stress, directly reflecting the dynamic mechanical properties of honeycomb materials at different strain rates. The constitutive relation library is a database that stores dynamic stress-volume strain curves corresponding to different strain rates, indexed by strain rate. Its core function is to provide mechanical input adapted to different working conditions for subsequent macroscopic homogeneous models.

[0056] In finite element software such as LS-DYNA, a series of strain rate conditions covering quasi-static to high-speed collision scenarios are set. For example, typical conditions include 10... -3 s -1 10 0 s -1 10 2 s -1 10 3 s -1 For each strain rate condition, dynamic compression simulations are run to simulate the dynamic mechanical behavior of the corrected Y element at that deformation rate, recording stress and volumetric strain data throughout the process. Dynamic stress-volume strain curves are extracted from the simulation results for each condition, outlier data points are removed, and the curves are ensured to be smooth and cover the entire strain range. Using strain rate as an index, strain rate values ​​are associated and stored with the corresponding dynamic stress-volume strain curves, forming a wide-range strain rate constitutive relation library, facilitating rapid matching and retrieval by subsequent macroscopic models based on the condition. This embodiment accurately captures the differences in the mechanical response of honeycomb materials at different deformation rates through full-range strain rate simulation, providing data support for dynamic condition simulations and compensating for the shortcomings of traditional models in characterizing dynamic effects.

[0057] S5. In the whole vehicle collision simulation environment, establish the initial homogeneous material model of the target honeycomb barrier.

[0058] Specifically, this includes: creating a homogeneous solid cell mesh of the target cellular barrier; assigning a homogeneous material model to the homogeneous solid cell mesh, and establishing an initial homogeneous material model of the target cellular barrier.

[0059] Homogeneous solid element networks are mesh structures obtained by discretizing the overall geometry of a target honeycomb barrier. They consist of hexahedral or tetrahedral solid elements, with element sizes much larger than microscopic Y elements, eliminating the need to characterize individual cell details. Homogeneous material models are constitutive models describing the macroscopic average mechanical behavior of honeycomb materials, such as MAT_HONEYCOMB in LS-DYNA. They support inputting stress-volume-strain curves to define material responses, adapting to engineering-level simulation requirements. Vehicle collision simulation environments refer to professional finite element simulation software platforms used to simulate the entire process of a car collision. These platforms include collision condition settings, mechanical model solving, and result output, capable of reproducing the load conditions and boundary constraints of a real collision. The initial homogeneous material model is a macroscopic finite element model that equates the honeycomb structure to a continuous homogeneous material. Its core consists of a homogeneous solid element mesh and a macroscopic homogeneous material constitutive model. It does not incorporate gas effects and strain rate adaptation curves, serving as the basic framework for subsequent corrections.

[0060] In finite element method (FEM) software, the overall geometry of the target honeycomb barrier is discretized to generate a homogeneous mesh composed of hexahedral or tetrahedral elements. The mesh must cover the entire volume of the barrier to ensure structural integrity. A macroscopic homogeneous material constitutive model is specified for the created homogeneous mesh. This constitutive model is pre-configured with basic mechanical properties and has a reserved interface for inputting stress-volume-strain curves for subsequent adaptation to different strain rate conditions. The mesh quality and material model configuration are checked to ensure that the model has no geometric defects and no missing parameters, forming an initial homogeneous material model that can be directly modified later. Homogenization simplifies the complex honeycomb structure into a continuous material model, significantly reducing the number of meshes and solution time, making it suitable for engineering applications such as vehicle collision simulation.

[0061] S6. Based on the strain rate of the working conditions in the whole vehicle collision simulation environment, select the matching dynamic stress-volume strain curve from the constitutive relation library, and input the dynamic stress-volume strain curve into the initial homogeneous material model to obtain the modified homogeneous material model.

[0062] Among them, the modified homogeneous material model is a macroscopic model adapted to specific collision conditions, formed by inputting the dynamic stress-volume strain curve with matching strain rate of the working condition into the initial homogeneous material model, and has the ability to accurately respond to dynamic loads.

[0063] Specifically, based on the strain rate of the working conditions in the whole vehicle collision simulation environment, a matching dynamic stress-volume strain curve is selected from the constitutive relation library, including: selecting the dynamic stress-volume strain curve corresponding to the same or closest strain rate from the constitutive relation library based on the strain rate of the working conditions in the whole vehicle collision simulation environment.

[0064] The actual strain rate of the target honeycomb barrier under the current collision condition is obtained through simulation pre-analysis or engineering experience. For example, the strain rate under high-speed collision condition is determined to be 10 based on pre-analysis. 3 s -1 The low-speed operating condition may be 10. -1 s -1 Ensure that the strain rate values ​​are consistent with the index units (s) of the constitutive relation library. -1 ( ) Consistent. Specifically, if the constitutive relation library contains an index that is exactly the same as the strain rate of the operating condition, such as the strain rate of the operating condition being 10... 3 s -1 If the constitutive relation library contains a dynamic stress-volume strain curve corresponding to the strain rate, that curve is directly selected as the matching result. If the constitutive relation library does not contain a completely identical strain rate index, such as when the strain rate is 950 s⁻¹, then... -1 Only 10 remain in the warehouse. 2 s -1 10 3 s -1The corresponding curve is selected based on the strain rate that is closest to the strain rate value under the working condition. A value of 10 is chosen. 3 s -1 The corresponding curves are selected to ensure they best reflect the dynamic mechanical properties of the material under the working conditions. The completeness and validity of the selected dynamic stress-volume strain curves are checked, confirming that the curves have no abnormal fluctuations and that the parameter format is compatible with the subsequent macroscopic homogeneous material model to be input. This avoids affecting the accuracy of subsequent simulations due to curve data issues, and ultimately determines this curve as the target curve for the working conditions. In the finite element software, the macroscopic homogeneous material model configuration interface of the initial homogeneous material model is opened, and the selected dynamic stress-volume strain curve data is imported and set as the stress-volume strain response parameters of the material model. The completeness of the curve input and the compatibility of the parameters are checked to ensure that the material model can calculate the corresponding stress in real time based on the current volumetric strain during the simulation, ultimately forming a modified homogeneous material model adapted to the current working conditions.

[0065] By inputting the strain rate curve matching the working conditions, the model can realistically reproduce the mechanical response under different deformation rates, solving the core defect of insufficient dynamic effect representation in traditional models. Different collision conditions correspond to different strain rates. Through precise matching of strain rate and dynamic stress-volume strain curve, it is ensured that the mechanical properties of the modified homogeneous material model are highly consistent with the actual working conditions, laying a high-precision mechanical foundation for subsequent gas-solid coupling simulations.

[0066] S7. Define at least one gas domain within the modified homogeneous material model and define the leakage characteristics of the gas domain to obtain a gas-solid coupled homogeneous model.

[0067] A gas domain is a virtual physical region used to simulate trapped air within a cellular cell. It is a non-independent geometric entity, attached to a homogeneous mesh via simulation software, and typically covers the entire volume of the cellular barrier or is partitioned according to structural characteristics. The leakage process describes the pattern of air leakage within the gas domain during cellular collapse, characterized by time-varying leakage area or a leakage rate related to strain rate, reflecting the dynamic process of gas leakage after cell wall rupture. The gas-solid coupled homogeneous model is the final macroscopic model that integrates the solid-state mechanical behavior of a modified homogeneous material model with the dynamic effects of gas domain compression and leakage, simultaneously reproducing the coupling relationship between solid-state framework deformation and internal gas pressure evolution.

[0068] Specifically, defining at least one gas domain within the modified homogeneous material model includes: constructing a gas state equation based on the adiabatic compression law of an ideal gas to characterize the pressure evolution relationship of the gas domain from its initial state to the compression process; the gas domain covers the overall volume of the target honeycomb barrier or is partitioned according to structural characteristics.

[0069] The equation of state for the gas is:

[0070] ;

[0071] In the formula, P0 and V0 are the initial pressure and volume, respectively, P represents the instantaneous pressure at a certain moment, V represents the instantaneous volume at the corresponding moment, and γ is the adiabatic index.

[0072] Specifically, the leakage characteristics of the gas domain are defined by characterizing the leakage characteristics of the gas domain through time-varying leakage area or leakage rate related to strain rate, where δ represents the leakage rate, and the calculation formula is:

[0073] .

[0074] After obtaining the gas-solid coupling homogeneous model, the process also includes: performing simulations based on the gas-solid coupling homogeneous model and outputting a dynamic crushing response that includes gas effects.

[0075] Based on the vehicle collision requirements, simulation boundary conditions, load parameters, and solution parameters are set to ensure consistency between the working conditions and real-world collision scenarios. The coupled equations of solid-state mechanics and gas state are solved iteratively. During the calculation, the crushing deformation of the solid skeleton changes the gas domain volume in real time, thus affecting the gas pressure. Simultaneously, changes in gas pressure react on the solid skeleton, forming a dynamic coupling, until the simulation reaches the preset termination condition. After the simulation is completed, the software automatically outputs the dynamic crushing response, which includes gas effects. The dynamic crushing response is high-precision mechanical and deformation data output from the simulation, including force-displacement time-series curves, full-field mesh deformation field, stress distribution, etc. The data includes the influence of gas effects, realistically reflecting the actual performance of the honeycomb barrier in a collision.

[0076] This embodiment obtains the nominal geometric parameters of the target honeycomb barrier; performs industrial computed tomography and 3D reconstruction on the honeycomb barrier sample to obtain actual geometric correction values; combines the nominal geometric parameters and actual geometric correction values ​​to establish the initial minimum repeatable element of the target honeycomb barrier, and performs inversion calibration on the material parameters in the initial minimum repeatable element to obtain the modified minimum repeatable element; performs dynamic compression simulation on the modified minimum repeatable element under different strain rates, extracts the dynamic stress-volume strain curve corresponding to each strain rate, and constructs a constitutive relation library; establishes an initial homogeneous material model of the target honeycomb barrier in a vehicle collision simulation environment; selects a matching dynamic stress-volume strain curve from the constitutive relation library according to the strain rate of the working condition in the vehicle collision simulation environment, and inputs the dynamic stress-volume strain curve into the initial homogeneous material model to obtain the modified homogeneous material model; defines at least one gas domain inside the modified homogeneous material model, and defines the leakage characteristics of the gas domain to obtain a gas-solid coupled homogeneous model. By defining a gas domain covering the entire or a section of the honeycomb barrier, dynamic coupling between the mechanical behavior of the solid skeleton and the evolution of internal gas pressure is achieved. This design recreates the complete physical process of dynamic crushing of a honeycomb barrier from a physical perspective, systematically making up for the core defect of traditional models that ignore the internal gas effect, and achieving a leapfrog improvement in simulation accuracy under dynamic collision conditions.

[0077] Furthermore, by constructing an initial minimum repeatable element combined with geometric corrections from industrial computed tomography (CT) scans, and replacing traditional full-scale physical experiment calibration with simulation inversion calibration using this micro-minimum repeatable element, a corrected minimum repeatable element is formed. This method eliminates the need for full-scale experiments, allowing parameter calibration to be completed solely through micro-element simulations and small-scale sample experiments. This significantly reduces experimental and time costs, while avoiding errors caused by structural scale effects in full-scale experiments. It makes material parameters more closely resemble actual working conditions, providing a high-precision foundation for the subsequent construction of a wide-strain-rate constitutive relation library, thus ensuring the reliability of the entire simulation model from the outset.

[0078] Example 2

[0079] Figure 3 This is a flowchart of a prediction method provided in an embodiment of the present invention, such as... Figure 3 As shown, the specific steps include the following:

[0080] S11. Based on the honeycomb barrier modeling method, establish gas-solid coupling homogeneous models corresponding to various specifications of honeycomb barriers as training base models.

[0081] S12. Perform dynamic axial crushing simulation on the training base model to generate a comprehensive training database; the comprehensive training database includes the initial finite element mesh, force-displacement response time series curves, and full-field mesh deformation time series sequence.

[0082] S13. A voxelization strategy based on node quality is adopted to voxelize the initial finite element mesh, converting the initial finite element mesh of each training base model into a normalized three-dimensional voxel matrix.

[0083] S14. Dimensionality reduction and feature extraction of the three-dimensional voxel matrix are performed using a convolutional neural network to obtain low-dimensional features.

[0084] S15. Construct a first long short-term memory neural network with low-dimensional features and time series as input and force-displacement response time series curve as output; construct a second long short-term memory neural network with low-dimensional features and time series as input and full-field mesh deformation time series as output.

[0085] S16. Based on the comprehensive training database, train the first long short-term memory neural network and the second long short-term memory neural network to obtain the trained prediction network.

[0086] S17. For a new target cellular barrier, perform voxelization and feature extraction according to the above steps, and input the obtained low-dimensional features into the prediction network to quickly predict the dynamic response of the cellular barrier.

[0087] For example, based on the above embodiments, a high-precision gas-solid coupling homogeneous model was constructed. A dynamic crushing simulation database containing 2058 cells of different sizes for a cellular barrier was automatically generated using a parameterized script, allowing variations in cell size, width, height, and thickness within a predetermined range. A voxel mesh resolution of 20×20×100 was used to voxelize the initial finite element meshes of all 2058 models based on node quality, generating standardized three-dimensional voxel data. A 3D-CNN autoencoder, with a single-layer CNN filter as the encoder, compressed the voxel matrix into a 32-dimensional latent space feature vector. After unsupervised training, the reconstruction error (MSE) of this autoencoder for the voxel matrix was <5×10⁻⁶. -6 This demonstrates its feature extraction capability.

[0088] The force-displacement prediction specifically involves using a network with a single LSTM cell layer and 256 hidden states. Training was performed using only 16% of the total database (329 simulation samples). The correlation coefficient R between the prediction results and the high-fidelity simulation curves over the full strain range is calculated. 2 The accuracy is >0.95, and the average error in energy absorption characteristics is between 3% and 5%. Mesh deformation prediction specifically involves using a network with two LSTM cell layers, including 1024 hidden states. The average node position error in the prediction is less than 2 mm, accurately capturing the folding sequence and deformation patterns of the structure. After training, for a new design scheme, the prediction network can achieve a resolution of approximately 2.7 × 10⁻⁶. -4The force-displacement curve is predicted within seconds, and the complete deformation field is predicted within approximately 2.4 seconds, demonstrating the high computational efficiency of this embodiment.

[0089] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the technical solutions of the embodiments of the present invention.

Claims

1. A method for modeling honeycomb barriers that integrates physical mechanisms, characterized in that, include: Obtain the nominal geometric parameters of the target cellular barrier; Industrial computed tomography and 3D reconstruction were performed on honeycomb barrier samples to obtain actual geometric correction values; By combining the nominal geometric parameters and the actual geometric correction values, the initial minimum repeatable unit of the target cellular barrier is established, and the material parameters in the initial minimum repeatable unit are inverted and calibrated to obtain the corrected minimum repeatable unit. Dynamic compression simulations were performed on the modified minimum repeatable element at different strain rates. The dynamic stress-volume strain curves corresponding to each strain rate were extracted, and a constitutive relation library was constructed. In a full-vehicle collision simulation environment, an initial homogeneous material model of the target honeycomb barrier is established; Based on the strain rate of the working conditions in the whole vehicle collision simulation environment, a matching dynamic stress-volume strain curve is selected from the constitutive relation library, and the dynamic stress-volume strain curve is input into the initial homogeneous material model to obtain the modified homogeneous material model. At least one gas domain is defined within the modified homogeneous material model, and the leakage characteristics of the gas domain are defined to obtain a gas-solid coupled homogeneous model.

2. The honeycomb barrier modeling method integrating physical mechanisms according to claim 1, characterized in that, Industrial computed tomography and 3D reconstruction were performed on the honeycomb barrier samples to obtain actual geometric correction values, including: Industrial computed tomography was performed on the cellular barrier sample to obtain tomographic data; A three-dimensional digital model of the cellular barrier sample was established based on the fault data; Based on the three-dimensional digital model, the actual geometric correction values ​​are obtained, including the correction values ​​for the width of a single-layer wall, the correction values ​​for the width of a double-layer wall, and the correction values ​​for the bending radius at the cell wall junction.

3. The honeycomb barrier modeling method integrating physical mechanisms according to claim 1, characterized in that, The material parameters in the initial minimum repeatable element are inverted and calibrated to obtain the modified minimum repeatable element, including: Quasi-static compression simulation was performed on the initial minimum repeatable element to obtain the static stress-volume strain simulation curve. Quasi-static physical compression experiments were conducted on the honeycomb barrier samples to obtain static stress-volume strain experimental curves; The objective function is to minimize the difference between the static stress-volume-strain simulation curve and the static stress-volume-strain experimental curve. The objective function is then used to iteratively correct the material parameters in the initial minimum repeatable element to obtain the corrected minimum repeatable element.

4. The honeycomb barrier modeling method integrating physical mechanisms according to claim 3, characterized in that, The objective function is the root mean square error function of the static stress-volume strain simulation curve and the static stress-volume strain experimental curve.

5. The honeycomb barrier modeling method integrating physical mechanisms according to claim 1, characterized in that, In a full-vehicle collision simulation environment, an initial homogeneous material model of the target honeycomb barrier is established, including: Create a homogeneous solid cell mesh for the target cellular barrier; Assign a homogeneous material model to the homogeneous solid unit mesh to establish the initial homogeneous material model of the target honeycomb barrier.

6. The honeycomb barrier modeling method integrating physical mechanisms according to claim 1, characterized in that, Based on the strain rate under the conditions of the vehicle collision simulation environment, matching dynamic stress-volume strain curves are selected from the constitutive relation library, including: Based on the strain rate under the working conditions in the whole vehicle collision simulation environment, select the dynamic stress-volume strain curve corresponding to the same or closest strain rate from the constitutive relation library.

7. The honeycomb barrier modeling method integrating physical mechanisms according to claim 1, characterized in that, Defining at least one gas domain within the modified homogeneous material model includes: A gas state equation is constructed based on the adiabatic compression law of ideal gases to characterize the pressure evolution relationship of the gas domain from its initial state to the compression process; the gas domain covers the entire volume of the target honeycomb barrier or is partitioned according to structural characteristics.

8. The honeycomb barrier modeling method integrating physical mechanisms according to claim 7, characterized in that, The leakage characteristics of the gas domain are defined as follows: The leakage characteristics of the gas domain are characterized by time-varying leakage area or leakage rate related to strain rate.

9. The honeycomb barrier modeling method integrating physical mechanisms according to claim 1, characterized in that, After obtaining the gas-solid coupled homogeneous model, the following is also included: Simulations were performed using a gas-solid coupling homogeneous model, and the dynamic crushing response, including gas effects, was output.

10. A prediction method, characterized in that, include: According to the honeycomb barrier modeling method in claim 1, a gas-solid coupling homogeneous model corresponding to honeycomb barriers of various specifications is established as the training basis model; Dynamic axial crushing simulation is performed on the training base model to generate a comprehensive training database; the comprehensive training database includes the initial finite element mesh, force-displacement response time series curves, and full-field mesh deformation time series sequence. A voxelization strategy based on node quality is adopted to voxelize the initial finite element mesh, converting the initial finite element mesh of each training base model into a normalized three-dimensional voxel matrix. The three-dimensional voxel matrix is ​​reduced in dimensionality and its features are extracted by using a convolutional neural network to obtain low-dimensional features. A first long short-term memory neural network is constructed using low-dimensional features and time series as inputs and force-displacement response time series curves as outputs; a second long short-term memory neural network is constructed using low-dimensional features and time series as inputs and full-field mesh deformation time series as outputs. Based on the comprehensive training database, the first long short-term memory neural network and the second long short-term memory neural network are trained to obtain the trained prediction network; For a new target cellular barrier, voxelization and feature extraction are performed according to the above steps. The resulting low-dimensional features are then input into the prediction network to quickly predict the dynamic response of the cellular barrier.

Images