Modeling method for honeycomb barrier cells

Abstract

This invention provides a modeling method for honeycomb barrier cells that efficiently optimizes the configuration parameters of a honeycomb barrier cell, constructing a simulation model of a honeycomb structure used in the front of a vehicle for collision safety design that combines geometric detail and physical validity. [Solution] The method involves constructing a discretized Y-shaped cross-sectional cell element model based on the honeycomb structure and fitting the stress-strain curve using this model. This allows for high-precision reproduction of the deformation behavior of the honeycomb structure and efficient optimization of the constituent parameters.

Description

Technical Field

[0001] The present invention relates to the technical field of vehicles, and particularly to a method for modeling honeycomb barrier cells executed by a computer.

Background Art

[0002] In the field of automotive crash safety tests, honeycomb aluminum materials are widely used in constructing barrier models for simulating the collision energy absorption process because their mechanical properties are very suitable for the front-end structure (front vehicle structure) of vehicles. However, the conventional modeling methods have the following problems. That is, although the equivalent material model can simplify calculations, it is difficult to capture the details of non-linear deformation in the honeycomb porous structure. Also, the solid element model cannot appropriately represent local buckling behavior due to excessive homogenization processing. Furthermore, the shell element model can reproduce geometric features, but due to being affected by boundary effects such as adhesive failure and air confinement, it is difficult to accurately reproduce the progressive collapse behavior during honeycomb crushing. Particularly, when the honeycomb is subjected to a dynamic impact, in the conventional model, non-physical deformation easily occurs at the bottom, and it is impossible to appropriately reproduce the regular folding pattern that dominantly progresses from the top in the actual structure. Therefore, the reliability of the simulation results is significantly reduced.

[0003] On the other hand, the calibration accuracy of the material's constitutive parameters directly affects the effectiveness of the simulation. Since the conventional method depends on local test data and empirical formulas, it cannot fully capture the overall correspondence relationship for complex dynamic responses. As a result, the consistency between the stress-strain curve and the test results is low. Particularly, when the strain rate sensitivity and temperature effect are combined, the parameter inverse analysis process is likely to fall into a local optimum solution, and it is difficult to establish a strong correlation between the material's properties and the macroscopic mechanical behavior. Therefore, constructing a honeycomb simulation model that combines geometric detail and physical validity while ensuring computational efficiency, and efficiently optimizing its constituent parameters, has become a crucial challenge for advancing collision safety design. [Prior art documents] [Patent Documents]

[0004] [Patent Document 1] China Patent Publication CN 119416599 A [Patent Document 2] China Patent Publication CN 115497582 A [Overview of the project]

[0005] In order to solve the above technical problems, the present invention provides a method for modeling honeycomb barrier cells. According to the present invention, by constructing a discretized Y-shaped cross-sectional cell element model, the problem of air confinement can be solved and the accuracy of the stress-strain curve can be improved.

[0006] This invention provides a computer-based method for modeling honeycomb barrier cells. The method is as follows: The process involves constructing a discretized Y-shaped cross-sectional cell element model based on the geometric symmetry of the honeycomb structure, The process includes fitting the discretized Y-shaped cross-sectional cell element model to the stress-strain curve of the honeycomb material, The process of constructing a discretized Y-shaped cross-sectional cell element model based on the geometric symmetry of the honeycomb structure is as follows: A step of determining the smallest Y-shaped cell element based on the aforementioned honeycomb structure, The process includes determining the discretized Y-shaped cross-sectional cell element model based on the smallest Y-shaped cell element, The smallest Y-shaped cell element is constructed by bonding together two flange wall surfaces arranged at a 120° angle. The process of fitting the stress-strain curve of the honeycomb material using the discretized Y-shaped cross-sectional cell element model is as follows: A step of determining the parameter space of the material's constituent parameters corresponding to the smallest Y-shaped cell element, The process involves performing multiple parameter samplings within the aforementioned parameter space, executing a dynamic compression simulation, and obtaining corresponding stress-strain data. A step of training a parameter inverse analysis model based on the correspondence between the stress-strain data obtained by the dynamic compression simulation and the target test data, The process includes: performing an inverse analysis on the test curve using the parameter inverse analysis model described above to determine the optimal configuration parameters and generating a stress-strain curve based on those configuration parameters.

[0007] The present invention provides the following technical benefits.

[0008] The computer-based honeycomb barrier cell modeling method provided by the present invention effectively improves upon the problems of conventional models, such as residual air effects and non-physical deformation due to reduced strength at adhesive interfaces, from a structural standpoint by constructing discretized Y-shaped cross-sectional cell elements. Conventional solutions do not consider the airflow characteristics inside the honeycomb cell and the dynamic fracture behavior of the adhesive, resulting in abnormal buckling in the bottom region in simulations, which is clearly different from the regular crushing pattern in actual impact tests where the top preferentially buckles. This method accurately represents the load transfer path between adjacent honeycomb walls by discretized modeling of geometrically symmetric Y-shaped cell elements, and eliminates interference from deformation laws due to residual air. As a result, the simulation model conforms favorably to the actual physical characteristics in the dynamic compression process, where the top preferentially folds and the bottom remains stable, significantly improving the spatial consistency of the deformation pattern.

[0009] By differentiating the thickness of flange walls and employing various modeling techniques for adhesive connections, it is possible to strike a balance between the reproducibility of the interface's mechanical properties and computational efficiency. A hybrid configuration of single-layer and double-layer shell elements avoids the virtual displacement at the bottom due to the hourglass effect in conventional solid element models by adjusting local stiffness, and also suppresses the occurrence of unexpected strength reduction at the adhesive interface during the compression process. Furthermore, by combining the overall optimization function of the parameter inverse analysis model, the relationships between constituent parameters are analyzed multivariately using a neural network. This improves upon the conventional problem of insufficient adaptability over a wide strain rate and temperature range, and allows the morphological features of the simulation curve at each stage of elastic deformation, yielding, plastic flow, and densification to be well matched to the test results. Furthermore, by using an adaptive sampling mechanism in the parameter space, sampling is concentrated in highly sensitive parameter regions, thereby improving calibration efficiency and providing a processing procedure suitable for rapid modeling of honeycomb materials from different batches. Furthermore, regarding the modeling of adhesive connections, methods such as contact algorithms, equivalent stiffness evaluation, and definition of material properties can be appropriately combined to flexibly accommodate the desired simulation accuracy. This allows for accurate reproduction of interfacial separation behavior while reducing the computational resource load associated with iterative corrections required in conventional methods due to adhesive failure, thereby achieving both practicality and reliability in predicting the performance of honeycomb barriers under complex operating conditions. [Brief explanation of the drawing]

[0010] To further clarify the embodiments of the present invention, the drawings are briefly described below. These drawings show only some embodiments of the present invention, and those skilled in the art can conceive of other embodiments based on these drawings.

[0011] [Figure 1] This is a flowchart of a modeling method for a honeycomb barrier cell according to one embodiment of the present invention. [Figure 2] This is a schematic diagram of the structure of the smallest Y-shaped cell element according to one embodiment of the present invention. [Figure 3] It is a schematic diagram of the restraint direction according to an embodiment of the present invention. [Figure 4] It is a schematic diagram of the structure of the minimum Y-shaped cell element according to an embodiment of the present invention. [Figure 5] It is a schematic diagram of the structure of the minimum Y-shaped cell element according to another embodiment of the present invention. [Figure 6] It is a schematic diagram of the structure of the minimum Y-shaped cell element according to still another embodiment of the present invention. [Figure 7] It is a diagram showing the comparison result of the deformation pattern according to an embodiment of the present invention. [Figure 8] It is a diagram showing the comparison result of the average compressive stress according to an embodiment of the present invention. [Figure 9] It is a diagram showing the deformation pattern of the honeycomb in the actual test. [Figure 10] It is a diagram showing the simulation result by the conventional modeling. [Figure 11] It is a diagram showing the simulation result by the modeling of the present invention.

Mode for Carrying Out the Invention

[0012] In order to make the object, configuration, and effect of the present invention clearer, the embodiments of the present invention will be described below. Note that the embodiments described below are merely examples of the present invention and do not limit the present invention. Other embodiments that can be easily conceived by those skilled in the art based on the present invention are also included in the technical scope of the present invention.

[0013] In the conventional modeling method of the honeycomb barrier, there is a problem that distortion occurs in the deformation pattern, and it is difficult to achieve both calculation efficiency and accuracy. In contrast, the present invention features a cooperative optimization method based on the reconstruction of discretized cells and the inverse analysis of parameters. Specifically, by decomposing the honeycomb structure into geometrically symmetric Y-shaped cross-sectional cell elements, non-physical deformation due to residual air effects occurring in the continuum model is suppressed. Furthermore, by establishing a correspondence between the material's constituent parameters and the macroscopic mechanical response, and applying a parameter inverse analysis model using dynamic compression simulation data, the limitations of conventional calibration methods over a wide strain rate range are overcome.

[0014] Embodiments of the present invention will be described in detail below with reference to the drawings and examples. Specifically, this will include three main steps: construction of the topology of the discretized cells, modeling of multiple patterns of the adhesive interface, and inverse optimization of the configuration parameters, as well as verification results from testing.

[0015] Figure 1 is a flowchart of the modeling method for the honeycomb barrier cell of the present invention. This method, Step S1 involves constructing a discretized Y-shaped cross-sectional cell element model based on the geometric symmetry of the honeycomb structure, The process includes step S2, which involves fitting a discretized Y-shaped cross-sectional cell element model to the stress-strain curve of the honeycomb material.

[0016] The modeling method for honeycomb barrier cells is characterized by reproducing the geometric and mechanical properties of the honeycomb structure using a discretization modeling technique. Conventional modeling methods use continuum elements or homogenization processes, making it difficult to accurately represent the porous topological structure of honeycomb aluminum. As a result, deformation behavior in compression simulations deviates from experimental results. To solve this problem, this method decomposes a single honeycomb block into discretized Y-shaped cross-sectional cell elements based on the geometric symmetry of the honeycomb structure. The spatial arrangement of each cell element is determined based on the analysis of geometric symmetry, and the connection relationships between adjacent elements are defined by a local coordinate system. This suppresses non-physical deformations caused by air retention or simplification of adhesive interfaces in conventional models. Furthermore, discretization modeling allows complex honeycomb structures to be represented as a set of standardized cell elements that can be expanded by topological decomposition. This makes it possible to represent the relationships between mechanical responses between elements while preserving the actual geometric shape of the honeycomb walls, and to more closely approximate the load transfer path during compression to actual material behavior.

[0017] In this embodiment, a discretized model is constructed using a finite element analysis preprocessing tool, and a dynamic compression simulation is performed using an explicit dynamic solver. Figure 2 is a schematic diagram of the structure of the smallest Y-shaped cell element according to one embodiment of the present invention, and Figure 3 is a schematic diagram of the constraint direction according to one embodiment of the present invention. Based on the geometric symmetry of the honeycomb structure, the smallest Y-shaped cell element (UC element) shown in Figure 2 is adopted as the basic element, and a local coordinate system is set on the surface of each UC element. Then, according to the constraint direction shown in Figure 3, displacement constraints in the normal direction (x direction) and coupling constraints in the tangential direction (yz direction) are applied to the three surface nodes. This suppresses the influence of the element's edge curvature on the deformation pattern.

[0018] To fit the stress-strain curve of a honeycomb material based on a discretized model, it is necessary to establish a correspondence between the material's constituent parameters and its macroscopic mechanical response. Conventional parameter calibration relies on empirical formulas or local test data, making it difficult to simultaneously capture the interaction of effects such as strain hardening, strain rate sensitivity, and temperature softening. This method optimizes the material's constituent parameters through an inverse analysis process, combines this with dynamic compression simulations to generate stress-strain data based on multiple parameter samplings, and constructs a parameter-response database that covers a wide range of operating conditions. Therefore, the synergistic effect of discretized modeling and inverse parameter analysis allows the simulation model to not only reproduce the deformation behavior in which the upper part preferentially buckles during the dynamic compression process, but also to predict stress plateau characteristics at different strain rates, thereby improving the agreement between simulation results and physical tests.

[0019] In some embodiments, the step of constructing a discretized Y-shaped cross-sectional cell element model based on the geometric symmetry of the honeycomb structure is: A process of determining the smallest Y-shaped cell element based on the honeycomb structure, The process includes the step of constructing a discretized Y-shaped cross-sectional cell element model based on the smallest Y-shaped cell elements, Here, the smallest Y-shaped cell element is constructed by bonding two flange walls positioned at a 120° angle to each other.

[0020] Determining the smallest Y-shaped cell element is fundamental to discretized modeling. The honeycomb structure possesses periodicity and symmetry, which determine the spatial arrangement rules of the microscopic elements. Microscopic geometric analysis of honeycomb aluminum confirms that it is composed of multiple periodically arranged hexagonal cells, and that each hexagonal cell can be decomposed into six Y-shaped cross-sectional elements. Each Y-shaped cross-sectional element contains two flange walls, and the angle between these flange walls is determined by the symmetry of the honeycomb, for example, 120°. If one vertex of the hexagonal cell is taken as the center of symmetry, adjacent flange walls extend at a predetermined angle and intersect each other, forming the basic arrangement of the Y-shaped cross-section. The method of connecting flange walls directly affects the overall stiffness and deformation pattern of the honeycomb structure. In modeling, the connection between flange walls is realized through bonding using geometric constraints and contact algorithms, so that the load transfer between adjacent elements matches the behavior of actual material interfaces.

[0021] The smallest Y-shaped cell elements are expanded through duplication and spatial transformation. Based on the periodicity of the honeycomb, single Y-shaped cell elements are arranged along the expansion direction of the honeycomb block, and the connection angles and spacing between elements are adjusted to match the actual geometric structure of the honeycomb. In the discretized model, the connection relationships between elements are defined by shared nodes or contact pairs. This avoids the stiffness distortion that occurred with continuous elements in conventional one-piece modeling. Furthermore, the discretization method can suppress non-physical displacements at the bottom in compression simulations by adjusting the stiffness of local elements, thereby controlling deformation to preferentially occur in the upper region. This result is consistent with the crushing laws observed in experiments.

[0022] In some embodiments, the process of fitting a discretized Y-shaped cross-sectional cell element model to a honeycomb material stress-strain curve is performed. A step of setting the parameter space for the constituent parameters of the material corresponding to the smallest Y-shaped cell element, The process involves performing multiple parameter samplings within the parameter space, executing a dynamic compression simulation, and obtaining corresponding stress-strain data. The process involves training a parameter inverse analysis model based on the correspondence between stress-strain data obtained from simulations and target test data, and The process includes: performing an inverse analysis on the test curve using a parameterized inverse analysis model to determine the optimal configuration parameters, and generating a stress-strain curve based on those configuration parameters.

[0023] The inverse analysis of the constituent parameters of honeycomb materials is achieved through optimization combining parameter space sampling and dynamic compression simulation. Setting the parameter space requires establishing reasonable parameter ranges for key parameters such as yield stress, hardening coefficient, and strain rate sensitivity coefficient, taking into account the manufacturing process and mechanical properties of honeycomb aluminum. Latin hypercube sampling is employed for parameter sampling to ensure a uniform distribution of sampling points in multidimensional space and reduce bias in parameter combinations. Each parameter set is input into a finite element model, and stress-strain data is obtained by performing a dynamic compression simulation. Here, the simulation conditions must match the loading rate and boundary conditions of the physical test to ensure data comparability. Furthermore, the obtained simulation data is preprocessed to extract features including key indicators such as elastic modulus, yield plateau length, and densification strain, and a database showing the correspondence between parameters and responses is constructed.

[0024] The parameter-inverse analysis model is trained using a deep neural network. The input layer corresponds to features extracted from the stress-strain curve, and the output layer is the configuration parameter vector. Here, the input features include the gradient in the elastic region, the mean stress at the plastic plateau, densification strain, etc., and are extracted by piecemeal sampling. The network optimizes its weights using a backpropagation algorithm, and the loss function considers both the absolute error of the parameters and the similarity of the curve shape. The trained model allows for the estimation of the optimal combination of constituent parameters based on the test curve, enabling a departure from conventional empirical calibration methods. Furthermore, if the agreement between the fitted curve and the test data is insufficient during the inverse analysis process, the parameter sampling is extended to retrain the model, and the uncertainty in the parameter space is gradually reduced through iterative optimization. In this process, parameter updates and simulation processing are scheduled by automated scripts, significantly reducing the manual workload and providing a standardized process for rapid modeling of different batches of honeycomb material.

[0025] Referring to Figure 2, in some embodiments, the smallest Y-shaped cell element includes a first flange wall 11, a second flange wall 12, and a third flange wall 13. The first flange wall 11 and the second flange wall 12 are single-layer shell elements, and the third flange wall 13 is a double-layer shell element.

[0026] The design of flange wall thickness directly affects the mechanical properties and computational efficiency of the model. Conventional solid element models cannot adequately represent the variation in stiffness at adhesive connections, leading to problems such as reduced reproducibility of interface behavior and the hourglass effect. The first flange wall 11 and the second flange wall 12 use single-layer shell elements to represent the actual thickness of the honeycomb aluminum foil. The third flange wall 13 uses double-layer shell elements to equivalently represent the adhesive joint of adjacent aluminum foils. In other words, the shell element with twice the thickness is an approximate model for equivalently representing the mechanical properties of the adhesive interface. The double-layer shell element is constructed by stacking shell elements, and its equivalent stiffness can be matched to the stiffness of the actual adhesive interface, thus avoiding iterative convergence problems in the contact algorithm. Regions consisting of single-layer shell elements preferentially buckle, forming the initial folded region at the top. On the other hand, regions consisting of double-layer shell elements exhibit slower buckling progression due to their higher rigidity, resulting in stepwise deformation. Because of this combination of thicknesses, the overall mechanical response can be optimized by adjusting local rigidity, suppressing abnormal deformation at the bottom, reducing the number of interface contact pairs, and lowering the complexity of the calculations.

[0027] Simulation results of the combined thickness model accurately reproduce the behavior during deformation, where the upper part preferentially collapses while the bottom region remains stable due to increased rigidity. This configuration maintains high accuracy while significantly improving computational efficiency, providing a practical method for modeling large-scale honeycomb barriers.

[0028] Figure 4 is a schematic diagram showing the structure of the smallest Y-shaped cell element according to one embodiment of the present invention. In some embodiments, the third flange wall 13 uses a single-layer shell element with twice the thickness to equivalently represent the bonding connection between the two flange walls. If the single-layer thickness of the shell element is h, then in Figure 4 the thickness of the shell element is 2h.

[0029] The bonding surface of the third flange wall 13 is realized by a two-layer shell element. In this design, the bonding interface of adjacent honeycomb aluminum foils is modeled equivalently. Conventional contact algorithms have the problem of reduced computational convergence due to the need to set complex contact pairs and failure criteria. In contrast, this method, based on the concept of geometric superposition, integrates the bonding region of two single-layer shell elements as a double-layer shell element, thereby making its equivalent bending stiffness correspond to the bending resistance of the actual bonding interface. The thickness parameter of the two-layer shell element is determined based on the bonding process of the honeycomb aluminum foil, ensuring that the stiffness distribution in the interface region closely resembles the mechanical behavior of the actual material. In dynamic compression simulations, increasing the thickness in the region consisting of the two-layer shell element forms a localized stiffness-reinforced zone, suppressing the progression of unexpected deformation due to premature failure of the adhesive interface. This equivalent modeling method reduces the number of contact pairs and the load of iterative calculations, improving simulation efficiency and avoiding the burden of model reconstruction due to adhesive failure correction required in conventional contact algorithms.

[0030] The geometric continuity of the double-layer shell elements ensures the continuity of the load transfer path. During the compression process, the double-layer shell elements of the third flange wall 13 function as a mechanical transition region, uniformly distributing the stress generated by upper buckling to adjacent elements. This suppresses lattice strain caused by localized stress concentration. Simulation results show that this configuration appropriately reproduces the gradual fracture behavior at the adhesive interface of the honeycomb material, and the deformation pattern of the entire model closely matches the crushing behavior observed in physical tests.

[0031] Figure 5 is a schematic diagram showing the structure of the smallest Y-shaped cell element according to another embodiment of the present invention. In some embodiments, the third flange wall 13 represents the connection between two flange walls by adhesion through contact between two single-layer shell elements.

[0032] The adhesive connection of the third flange wall 13 is represented by a contact algorithm between two single-layer shell elements, and this method allows the actual geometric shape of the honeycomb aluminum foil to be maintained. Two single-layer shell elements are arranged parallel to each other along the adhesive interface, and the behavior of the contact surface is defined by a master node and a slave node. The master node is subjected to a normal contact force and a tangential frictional stress, and the slave node follows the motion of the master node according to constraint equations, thereby ensuring continuity of load transfer at the interface. The contact algorithm has a threshold for adhesive strength, and delamination occurs when the interfacial stress exceeds this threshold, thereby reproducing the actual delamination behavior of the adhesive layer. Such a contact-based model allows for detailed definition of contact conditions and avoids the inaccuracies in interfacial strength caused by the averaging of stiffness in conventional equivalent thickness models.

[0033] The parameters of the contact algorithm must be set to correspond to the bonding process of the honeycomb aluminum foil. The normal contact stiffness is calibrated based on the compressive modulus of the adhesive layer, and the tangential friction coefficient is determined based on interface shear test data. In dynamic compression simulations, the failure threshold of the contact pair controls the separation timing of the adhesive interface, and the stepwise buckling progression behavior during the deformation process is reproduced. This method allows for a detailed representation of the mechanical behavior of the interface and reduces the computational load by defining the contact pairs locally. This method can be applied to simulations of small to medium-sized honeycomb models where high-precision evaluation of interface failure behavior is required.

[0034] Figure 6 is a schematic diagram showing the structure of the smallest Y-shaped cell element according to another embodiment of the present invention. In some embodiments, the third flange wall 13 represents the connection of two flange walls by adhesion between two single-layer shell elements and an adhesive layer 14 placed between the two single-layer shell elements.

[0035] The adhesive portion of the third flange wall surface 13 is composed of two single-layer shell elements and an intermediate adhesive layer 14, and this configuration allows for a more accurate representation of the mechanical properties of the interface. The adhesive layer 14 is composed of thin solid elements, the thickness of which is set to correspond to the actual adhesive application thickness. The adhesive layer elements are connected to upper and lower single-layer shell elements via shared nodes, forming a "sandwich" structure. For example, a solid element with a thickness of 0.05 mm can be placed between single-layer shell elements, the mechanical behavior of the interface can be defined using an elastoplastic constitutive model of the adhesive, and the separation process can be characterized based on normal and tangential stress functions. The material properties of the adhesive layer are defined based on the tensile-shear coupled response of the adhesive, and the normal strength and tangential delamination threshold are calibrated using interface test data. By introducing these independent adhesive layer elements, it becomes possible to accurately simulate the gradual failure process of the adhesive interface under complex loading conditions, including normal delamination, shear slip, and mixed-mode fracture.

[0036] The mechanical response of the adhesive layer elements directly affects the deformation pattern of the entire model. In dynamic compression simulations, as plastic strain accumulates in the adhesive layer 14, interfacial fracture occurs, and the upper and lower shell elements gradually separate. This fracture mechanism closely matches the process from elastic deformation to fracture of the adhesive layer 14 observed in physical tests, allowing the simulation model to capture the stress relaxation characteristics in the later stages of collapse of the honeycomb material. Such precise modeling of the adhesive layer elements makes it possible to evaluate the impact of interfacial fracture on the overall mechanical behavior of the honeycomb with high accuracy, and can be applied to accurately predict the performance of honeycomb structures under extreme usage conditions in fields such as aerospace.

[0037] Figure 7 shows a comparison of deformation patterns according to embodiments of the present invention. As shown in Figure 7, the simulation results of the three models (double-thickness shell element, contact algorithm model, and adhesive layer solid element model) all agree well with the deformation behavior observed in actual physical tests, and the characteristic of preferential buckling at the top is reproduced. Model 1 is an embodiment corresponding to Figure 4, Model 2 is an embodiment corresponding to Figure 5, and Model 3 is an embodiment corresponding to Figure 6. Of these, the deformation propagation path of the shell element model with double the thickness (corresponding to Figure 4) is the most regular, and it is confirmed that asymmetric folding is effectively suppressed by strengthening the rigidity at the adhesive interface.

[0038] Figure 8 shows a comparison of average compressive stress according to embodiments of the present invention. In Figure 8, the average compressive stress and test results of three different models are compared. Model 1 is the embodiment corresponding to Figure 4, Model 2 is the embodiment corresponding to Figure 5, and Model 3 is the embodiment corresponding to Figure 6. The stress-strain curve of the shell element model with double the thickness (corresponding to Figure 4) shows an error of less than 5% from the test value in the plastic plateau region, which is clearly superior to conventional modeling methods (errors exceeding 15%). This indicates that, through the reconstruction of discretized cells and optimization of the inverse parameter analysis, the model of the present invention can appropriately represent the dynamic energy absorption characteristics of the honeycomb material.

[0039] Table 1 shows a comparison of the computation times for three different models (corresponding to Figures 4, 5, and 6). Model 1 (corresponding to Figure 4) has the shortest computation time, at approximately 10 minutes. Here, Model 1 corresponds to the model in Figure 4, Model 2 corresponds to the model in Figure 5, and Model 3 corresponds to the model in Figure 6.

[0040] Table 1: Comparison of computation times for three different models JPEG0007899492000002.jpg44170 In some embodiments, the constituent parameters of the material follow the following flow stress equation. JPEG0007899492000003.jpg100170

[0041] The constitutive equations for the material need to simultaneously represent the strain hardening, strain rate sensitivity, and temperature softening effects of honeycomb aluminum. The mathematical form of these equations appropriately reproduces the nonlinear characteristics of the stress-strain relationship in dynamic compression simulations, and its main components include strain hardening, strain rate, and temperature softening terms. The strain hardening term expresses the increase in stress due to the accumulation of plastic strain as a power function, and the hardening index controls the slope of the rise of the stress-strain curve. The strain rate term expresses the relationship between strain rate and stress increase as a logarithmic function, reflecting the increase in the material's deformation resistance under high strain rate conditions. The temperature softening term expresses the decrease in material strength with increasing temperature exponentially, and the softening index determines the degree of this decrease. Furthermore, each parameter in the equation corresponds to the microscopic mechanical behavior of the honeycomb aluminum; for example, the yield stress corresponds to the initial plastic deformation initiation condition, and the hardening coefficient reflects the resistance to dislocation motion.

[0042] Calibration of the configuration parameters must be performed by combining test data under multiple operating conditions. Strain hardening and strain rate sensitivity are obtained by consolidation tests under a constant strain rate, and the temperature softening effect is calibrated by temperature variation tests. In the simulation model, the combination of parameters must be set to ensure consistency in the shape of the stress-strain curve under different loading conditions. For example, the stress plateau under high strain rate conditions is reproduced by adjusting the strain rate sensitivity coefficient to match the steep rise characteristics of the test curve. Furthermore, the stress decrease with increasing temperature is reproduced by optimizing the softening index. By combining multiple terms in this way, the constitutive equation can accommodate a wide range of operating conditions, providing a reliable theoretical framework for coupled analysis of transient impact responses and thermodynamic effects during vehicle collisions.

[0043] In some embodiments, the parameter inverse analysis model is a neural network model that takes features extracted from stress-strain data as input and outputs a vector of constituent parameters of the material.

[0044] In designing the architecture of a neural network model, both the accuracy and computational efficiency of the inverse parameter analysis must be considered. The input layer receives pre-processed stress-strain curve features, including important geometric features such as the gradient in the elastic region, the length of the plastic plateau, and densification strain. A fully connected structure is employed in the hidden layer, and high-dimensional mapping of features is performed by a nonlinear activation function. The output layer corresponds to the configuration parameter vector, with each node relating to physical quantities such as yield stress and hardening coefficient. The network training data is generated from dynamic compression simulation results based on parameter sampling, and the dataset is configured to cover the mechanical response range under typical operating conditions of a honeycomb material. During the learning process, it is necessary to consider the balance between the absolute error of the parameters as a loss function and the similarity of the curve shapes. For example, by quantifying the shape differences between curves using the Dynamic Time Warping (DTW) algorithm, it is possible to avoid convergence to local optima caused by simple mean squared errors.

[0045] The inverse analysis capability of a neural network depends on the effective representation of features. Feature extraction from stress-strain curves is performed using piecewise sampling with a sliding window and principal component analysis, configured to retain important shape information of the curve while reducing the effects of noise. A trained network can estimate combinations of material parameters based on the variability characteristics of the test curve. For example, it can estimate the hardening index from the slope of the plastic plateau or evaluate the temperature softening effect based on the behavior of densification strain. This model reduces the reliance on empirical formulas in conventional inverse analysis methods and is particularly effective for multi-parameter coupled nonlinear problems, such as stress relaxation characteristics involving the synergistic effect of strain rate and temperature. Simulation verification results show that when using the configuration parameters inversely analyzed by a neural network, the fitted stress-strain curves agree well with test data over a wide range of operating conditions, providing reliable support for modeling honeycomb materials under complex loading conditions.

[0046] In practical implementation, initial configuration parameters can be obtained in the following two ways:

[0047] 1) Using the fitting function of the JOHNSON-COOK model in finite element analysis software, input the yield strength, tensile strength, and necking strain of the material, and automatically calculate the parameters. 2) Using a parameter optimization tool, the least squares method is applied to the stress-strain data obtained experimentally, and the optimal combination of parameters is found through iterative means. These parameters are used as the initial sampling set for training the neural network.

[0048] In some embodiments, the generation of the parameter sampling set includes dynamically adjusting the sampling density in the parameter space and optimizing the distribution of sampling points according to the parameter sensitivity by feedback based on the prediction error of the parameter inverse analysis model. The specific adjustment rules are as follows:

[0049] If the prediction error of the parameter inverse analysis model exceeds a predetermined threshold (e.g., 10%), the parameter sampling interval is reduced to 50% of the original interval and resampling is performed. On the other hand, if the model's accuracy meets the requirements but the computation time is excessively long, the sampling interval is doubled from the original interval.

[0050] As an example, when considering the calibration of the yield stress parameter A, the initial sampling range is set to 100-500 MPa, the sampling interval to 100 MPa, and the values ​​are set to 100, 200, 300, 400, and 500.

[0051] If the fitting error after the initial inverse analysis exceeds 10%, the sampling interval is reduced to 50 MPa and resampling is performed (taking values ​​of 100, 150, 200, ..., 500) to improve accuracy. On the other hand, if the error is less than 5% but the time required for a single simulation exceeds 24 hours, the sampling interval is increased to 200 MPa (taking values ​​of 100, 300, and 500) to reduce the computational load.

[0052] The inverse analysis involves an iterative optimization process. If the degree of agreement between the fitted stress-strain curve and the target test curve does not meet the specified conditions, the parameter sampling set is expanded to update the parameterized inverse analysis model and generate a new fitted curve. This process is repeated until the specified conditions are met.

[0053] Dynamic adjustment of the parameter sampling set is achieved through iterative optimization based on feedback control. The initial parameter space range is set based on material processes and existing test data, and uniformly distributed sampling points are generated using the Latin hypercube method. The degree of agreement between the simulation curve corresponding to each parameter group and the target test curve is quantified using a similarity index, and the results are fed back to the sampling control unit. The control unit identifies regions with high parameter sensitivity based on this score distribution, and in subsequent iterations, increases the sampling density in those regions while reducing redundant sampling in regions with low sensitivity. For example, if the degree of agreement fluctuates greatly with even slight changes in a part of a parameter space, that region is judged to be highly sensitive, and it is necessary to increase the number of sampling points by dividing the grid accordingly.

[0054] The iterative optimization process converges stepwise to the optimal parameter solution through closed-loop feedback. In each iteration, sampled data is added to update the parameter inverse analysis model, improving prediction accuracy, especially in sensitive regions. If the degree of agreement with the fitted curve obtained from the current inverse analysis model falls below a predetermined threshold, the parameter space is expanded. For example, the parameter range can be expanded along the direction of high sensitivity, or new parameter combinations (dimensions) can be introduced. This process is repeated until the degree of agreement meets a predetermined condition or the number of iterations reaches the upper limit. The dynamic adjustment strategy of the present invention significantly improves calibration efficiency by balancing "exploration" (sampling in unknown regions) and "utilization" (subdivision of highly sensitive regions). Furthermore, this mechanism reduces the number of simulations required for parameter calibration and ensures the overall optimization of parameters obtained by inverse analysis. As a result, a fundamental technology is provided that enables rapid consistency evaluation of multiple lots of honeycomb materials.

[0055] In practical implementation, a commercially available 3003 series aluminum honeycomb sample with dimensions of 250 mm × 250 mm × 150 mm and a cell wall thickness of 19 mm can be used for the test. The impact test is performed using a 72 kg indenter at a speed of 5.24 m / s. The compressive stress is calculated using the formula σ = F / A, where F is the impact force and A is the cross-sectional area of ​​the sample in the planar direction. Figure 9 shows the deformation pattern of the honeycomb in an actual test, Figure 10 shows the simulation results using conventional modeling, and Figure 11 shows the simulation results using the modeling of the present invention. As shown in Figures 9 to 11, the deformation pattern in the actual test is characterized by buckling progressing preferentially from the top, and the simulation results of the present invention agree well with this characteristic. On the other hand, in the conventional model, non-physical deformation occurs at the bottom.

[0056] An embodiment of this application may be a computer-readable storage medium that stores a program for causing a computer to execute each step of the honeycomb barrier cell modeling method according to this application.

[0057] Computer program instructions can be created as program code for executing the processing in the embodiments of this application using any combination of one or more programming languages. The programming language is not limited to object-oriented programming languages ​​such as Java and C++, but may also include procedural programming languages ​​such as "C". Furthermore, the above program code may be executed on the user's device, or it may be partially executed on the user's device with other parts executed on a remote computer or server.

[0058] A computer-readable storage medium may include any combination of one or more readable media. The readable media may be a computer-readable signal medium or a computer-readable storage medium. Computer-readable storage media may, but are not limited to, devices or equipment using electrical, magnetic, optical, electromagnetic, infrared, or semiconductor technologies, or any combination thereof. Specific examples of computer-readable storage media (non-exclusive list) include electrical connections with one or more wires, portable disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disc read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any combination thereof.

[0059] The terms used in this application are for the purpose of describing specific embodiments and are not intended to limit the scope of this application. In the specification and claims of this application, unless the context clearly indicates otherwise, terms such as “1,” “one,” “one kind,” and / or “the” are not limited to a singular number but may include multiple numbers. Furthermore, terms such as “include,” “contain,” or similar terms mean non-exclusive inclusion, so that a process, method, or apparatus containing certain elements may also include other elements, processes, methods, or apparatus specific to those elements, in addition to those explicitly stated. Moreover, even if it is written as “contain one…,” this does not preclude the inclusion of the same element further.

[0060] In this specification, the principles and embodiments of this application have been described using specific examples. These descriptions are for the purpose of facilitating understanding of this application and do not limit it. The embodiments described above are merely examples, and those skilled in the art can make various improvements, modifications, or changes without departing from this application. Furthermore, the above technical features may be combined as appropriate, and any such improvements, modifications, changes, or combinations thereof are all included within the scope of protection of this application. [Explanation of symbols]

[0061] 11. First flange wall 12. Second flange wall 13. Third flange wall 14 Adhesive layer

Claims

1. A computer-based method for modeling honeycomb barrier cells, The process involves constructing a discretized Y-shaped cross-sectional cell element model based on the geometric symmetry of the honeycomb structure, The process includes the step of fitting the discretized Y-shaped cross-sectional cell element model to the stress-strain curve of the honeycomb material, The process of constructing a discretized Y-shaped cross-sectional cell element model based on the geometric symmetry of the honeycomb structure is as follows: A step of determining the smallest Y-shaped cell element based on the aforementioned honeycomb structure, The process includes determining the discretized Y-shaped cross-sectional cell element model based on the smallest Y-shaped cell element, The smallest Y-shaped cell element is constructed by bonding together two flange walls arranged at a 120° angle. The process of fitting the stress-strain curve of the honeycomb material using the discretized Y-shaped cross-sectional cell element model is as follows: A step of setting the parameter space for the constituent parameters of the material corresponding to the smallest Y-shaped cell element, The process involves performing multiple parameter samplings within the parameter space and executing a dynamic compression simulation to obtain corresponding stress-strain data. The process involves training a parameter inverse analysis model based on the correspondence between the stress-strain data obtained from the simulation and the target test data, The process includes: performing an inverse analysis on the test curve using the parameter inverse analysis model described above to determine the optimal configuration parameters, and generating a stress-strain curve based on the configuration parameters, The constituent parameters of the aforementioned material are expressed by the following flow stress equation: The aforementioned parameter inverse analysis model is a neural network model that takes features extracted from stress-strain data as input and outputs a vector of constituent parameters of the material. The generation of the parameter sampling set includes a step of dynamically adjusting the sampling density in the parameter space and corresponding the distribution of sampling points to the parameter sensitivity by feedback based on the prediction error of the parameter inverse analysis model. The inverse analysis includes an iterative optimization process. If the degree of agreement between the fitted stress-strain curve and the target test curve does not meet a predetermined condition, the parameter sampling set is expanded to update the parameter inverse analysis model, a new fitted stress-strain curve is generated, and this process is repeated until the degree of agreement meets the predetermined condition. A method for modeling a honeycomb barrier cell, characterized by the following:

2. The method according to claim 1, characterized in that the smallest Y-shaped cell element includes a first flange wall, a second flange wall, and a third flange wall, wherein the first flange wall and the second flange wall are single-layer shell elements, and the third flange wall is a double-layer shell element.

3. The method according to claim 2, characterized in that the third flange wall surface is equivalently modeled by a shell element having a thickness twice that of the single-layer shell element, thereby representing the connection between the two flange wall surfaces by adhesion.

4. The method according to claim 2, characterized in that the third flange wall surface models the connection between the two flange wall surfaces by adhesion through contact between the two single-layer shell elements.

5. The method according to claim 2, characterized in that the third flange wall surface models the connection between two flange wall surfaces by adhesion using two single-layer shell elements and an adhesive layer disposed between the two single-layer shell elements.

Images