Integrated large complex thin-wall die-casting component structure performance collaborative design method

Abstract

The invention discloses an integrated large complex thin-wall die-casting component structure performance collaborative design method, and relates to the technical field of artificial intelligence. And full-automatic optimization of the complex thin-wall part from conceptual design to performance verification is realized. According to the system, the mechanical property reliability and the multi-target cooperation capability of the design are remarkably improved while the calculation efficiency is guaranteed, and finally a light-weight and high-crashworthiness structure scheme and a traceable integrity performance report are output.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a method for collaborative design of structural performance of an integrated large-scale complex thin-walled die-cast component. Background Art

[0002] In order to reduce carbon emissions, advanced die-casting manufacturing processes should be used. Lightweight design of automobiles can save fuel. Large thin-walled castings such as the body-in-white and front engine compartment used in automobiles can be manufactured using integrated die-casting technology, but the size and complexity of large castings increase the difficulty of producing high-quality die-castings. High-strength materials are important for automobile safety, but in addition to strength requirements, automotive parts have many other requirements to meet, such as vehicle dynamics static stiffness, NVH, driving comfort standards and passive safety, etc.

[0003] For each requirement (such as NVH, statics, durability and collision), there is a separate finite element model and different calculations are performed. Since each design requirement involves different commercial solvers and physical equations, most of them use independent models to derive structural changes using structural optimization (topology optimization) technology to improve design accuracy. However, the individually designed models cannot be aligned, so the derived lightweight design parameters are not the optimal solution, and multiple iterations are required between multiple design requirements, which increases R&D costs. Summary of the invention

[0004] The purpose of this application is to provide a method for collaborative design of structural performance of integrated large-scale complex thin-walled die-cast components.

[0005] To achieve the above objectives, this application provides the following solutions: In a first aspect, the present application provides a method for collaborative design of structural performance of an integrated large-scale complex thin-walled die-casting component, comprising: Input the design parameters of complex thin-walled parts into the multi-model topology optimization framework to generate the initial load path optimization solution; Building a response surface prediction model based on the initial load path optimization scheme, and generating a training data set through experimental design; Inputting the training data set into a machine learning enhancement module to generate a design space classifier and cluster the simulation results; An optimal design group satisfying preset constraints is selected according to the classifier, and plastic deformation and damage failure simulation is performed using a material constitutive model specified by a preset standard; Conducting multi-dimensional experimental tests on the samples in the optimal design group to obtain measured data to calibrate the failure parameters of the simulation model; Perform head-on collision tests based on the calibrated simulation model and output the final design solution that meets multiple performance constraints; The optimization results of each stage in the final design solution are integrated to generate an integrated structural performance report of complex thin-walled parts.

[0006] Optionally, the step of inputting the design parameters of the complex thin-walled part into the multi-model topology optimization framework to generate an initial load path optimization solution comprises: Adjust material distribution density through density-based topology optimization algorithm to determine the optimal load path that meets statics, NVH and crash performance; Establishing a multidisciplinary coupling optimization model, wherein the multidisciplinary coupling optimization model balances the stiffness matrices of different finite element models through weight factors; The design variables under multiple load conditions are mapped to a unified parameter space, and a set of candidate solutions containing structural efficiency and performance indicators is output.

[0007] Optionally, the step of constructing a response surface prediction model based on the initial load path optimization scheme includes: Extracting design variables and response variables from the candidate solution set and constructing a second-order polynomial response surface function; Generate a training data set by a Latin hypercube sampling method specified in a preset standard, wherein the training data set covers compression, tension and shear stress states; The response surface prediction model is fitted using the training data set to output predicted values ​​of stress, deformation and damage factor.

[0008] Optionally, the step of inputting the training data set into the machine learning enhancement module comprises: Performing feature extraction on the visualization data of the simulation results to mark key stress concentration areas in the design space; Using a K-means clustering algorithm to group the training data set to generate a classifier containing failure mode labels; The classifier is coupled with the response surface prediction model to establish a probability-driven design screening rule.

[0009] Optionally, the step of using a material constitutive model specified in a preset standard to perform plastic deformation and damage failure simulation includes: Call the MAT_PLASTICITY_COMPRESSION_TENSION_MAT_124 keyword to simulate the compression-tension bimodal response of the material; Calculate the dynamic yield stress under high strain rate conditions based on the Cowper-Symonds strain rate model; The MAT_ADD_DAMAGE_DIEM keyword is used to evolve the damage factor, which is iteratively updated by the equivalent plastic displacement increment.

[0010] Optionally, the step of performing multi-dimensional experimental testing on the samples in the optimal design group includes: Perform metallographic testing and CT scanning at preset locations on the sample to obtain microstructure and porosity distribution data; Performing quasi-static tensile tests and dynamic impact tests to collect stress-strain curves and fracture toughness parameters of the samples; Through X-ray nondestructive testing and scanning electron microscope fracture analysis, the mapping relationship between defect type and simulated damage results is established.

[0011] Optionally, the step of performing a head-on collision test verification based on the calibrated simulation model includes: Constructing a whole vehicle collision finite element model, and embedding the complex thin-walled component as a key component into the whole vehicle collision finite element model; Setting the initial collision velocity and boundary conditions specified by the preset standard, and calculating the deformation energy absorption of the complex thin-walled part; The simulated deformation pattern is compared with the measured data to verify the stiffness matching and optimize the process feasibility of the final design.

[0012] In a second aspect, the present application provides an integrated large-scale complex thin-wall die-casting component structure performance collaborative design device, comprising: A processing module is used to input the design parameters of complex thin-walled parts into the multi-model topology optimization framework to generate an initial load path optimization solution; Building a response surface prediction model based on the initial load path optimization scheme, and generating a training data set through experimental design; Inputting the training data set into a machine learning enhancement module to generate a design space classifier and cluster the simulation results; An optimal design group satisfying preset constraints is selected according to the classifier, and plastic deformation and damage failure simulation is performed using a material constitutive model specified by a preset standard; Conducting multi-dimensional experimental tests on the samples in the optimal design group to obtain measured data to calibrate the failure parameters of the simulation model; The output module is used to perform head-on collision test verification based on the calibrated simulation model and output the final design solution that meets multiple performance constraints; The optimization results of each stage in the final design solution are integrated to generate an integrated structural performance report of complex thin-walled parts.

[0013] Optionally, the processing module is further used to: Adjust material distribution density through density-based topology optimization algorithm to determine the optimal load path that meets statics, NVH and crash performance; Establishing a multidisciplinary coupling optimization model, wherein the multidisciplinary coupling optimization model balances the stiffness matrices of different finite element models through weight factors; The design variables under multiple load conditions are mapped to a unified parameter space, and a set of candidate solutions containing structural efficiency and performance indicators is output.

[0014] Optionally, the processing module is further used to: Extracting design variables and response variables from the candidate solution set and constructing a second-order polynomial response surface function; Generate a training data set by a Latin hypercube sampling method specified in a preset standard, wherein the training data set covers compression, tension and shear stress states; The response surface prediction model is fitted using the training data set to output predicted values ​​of stress, deformation and damage factor.

[0015] Optionally, the processing module is further used to: Performing feature extraction on the visualization data of the simulation results to mark key stress concentration areas in the design space; Using a K-means clustering algorithm to group the training data set to generate a classifier containing failure mode labels; The classifier is coupled with the response surface prediction model to establish a probability-driven design screening rule.

[0016] Optionally, the processing module is further used to: Call the MAT_PLASTICITY_COMPRESSION_TENSION_MAT_124 keyword to simulate the compression-tension bimodal response of the material; Calculate the dynamic yield stress under high strain rate conditions based on the Cowper-Symonds strain rate model; The MAT_ADD_DAMAGE_DIEM keyword is used to evolve the damage factor, which is iteratively updated by the equivalent plastic displacement increment.

[0017] Optionally, the processing module is further used to: Perform metallographic testing and CT scanning at preset locations on the sample to obtain microstructure and porosity distribution data; Performing quasi-static tensile tests and dynamic impact tests to collect stress-strain curves and fracture toughness parameters of the samples; Through X-ray nondestructive testing and scanning electron microscope fracture analysis, the mapping relationship between defect type and simulated damage results is established.

[0018] Optionally, the output module is further used to: Constructing a whole vehicle collision finite element model, and embedding the complex thin-walled component as a key component into the whole vehicle collision finite element model; Setting the initial collision velocity and boundary conditions specified by the preset standard, and calculating the deformation energy absorption of the complex thin-walled part; The simulated deformation pattern is compared with the measured data to verify the stiffness matching and optimize the process feasibility of the final design.

[0019] In a third aspect, the present application provides a computer device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the method for collaborative design of structural performance of an integrated large-scale complex thin-walled die-casting component as described in any one of the above.

[0020] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any of the above-described methods for collaboratively designing the structural performance of an integrated large-scale complex thin-walled die-casting component.

[0021] In a fifth aspect, the present application provides a computer program product, including a computer program, which, when executed by a processor, implements the steps of any of the above-mentioned methods for collaborative design of structural performance of one-piece large-scale complex thin-walled die-cast components.

[0022] According to the specific embodiments provided in this application, this application discloses the following technical effects: This application provides a method for collaborative design of the structural performance of large-scale, complex thin-walled die-cast components. Through the collaborative process of multi-model topology optimization, machine learning-assisted classification, and experimental-simulation closed-loop verification, the fully automated optimization of complex thin-walled parts from conceptual design to performance verification is achieved. While ensuring computational efficiency, the system significantly improves the mechanical performance reliability and multi-objective collaborative capabilities of the design, and ultimately outputs a lightweight, highly crashworthy structural solution and a traceable complete performance report. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0024] Figure 1 A schematic diagram of a process flow of a method for collaboratively designing the structural performance of an integrated large-scale complex thin-walled die-casting component provided in one embodiment of the present application; Figure 2 One of the principle schematic diagrams of a method for collaborative design of structural performance of an integrated large-scale complex thin-walled die-casting component provided in one embodiment of the present application; Figure 3 A second schematic diagram of the principle of a method for collaboratively designing the structural performance of an integrated large-scale complex thin-walled die-casting component provided in one embodiment of the present application; Figure 4 A third schematic diagram of the principle of a method for collaborative design of structural performance of an integrated large-scale complex thin-walled die-casting component provided in one embodiment of the present application; Figure 5 A schematic diagram of functional modules of an integrated large-scale complex thin-wall die-casting component structure performance collaborative design device provided in one embodiment of the present application; Figure 6 A schematic diagram of the structure of a computer device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0025] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0026] like Figure 1 As shown, some embodiments of the present application provide a method for collaborative design of structural performance of an integrated large-scale complex thin-walled die-casting component, which includes: Step 101 , input the design parameters of the complex thin-walled part into a multi-model topology optimization framework to generate an initial load path optimization solution.

[0027] In the embodiments of the present application, complex thin-walled parts refer to engineering components with complex geometric shapes and thin walls, which are usually used to reduce weight and meet mechanical performance requirements. Design parameters include input variables such as geometric dimensions, material properties, and load conditions. The multi-model topology optimization framework is a computing platform that integrates multiple optimization algorithms and is used to automatically generate structural design solutions under given constraints.

[0028] The computer terminal inputs the design parameters of the complex thin-walled parts defined by the user (such as boundary conditions, objective functions, and constraints) into the multi-model topology optimization framework. The framework generates an initial load path optimization solution through iterative calculations (such as variable density method or level set method), which is presented in the form of structural material distribution to ensure the initial balance between mechanical properties and lightweight goals.

[0029] Specifically, topology optimization and response surface-based optimization are combined, and the response surface-based optimization is enhanced by expert simulation and cluster classification using machine learning. Figure 2 shown.

[0030] The overall optimization problem can be expressed as a multi-model optimization problem, and formulas (1) and (2) are as follows: (1) in, The results and (m) Each load case is mapped to its corresponding finite element model m. Here, Define the objective function, is an inequality constraint, and They are The lower and upper bounds of the design variables.

[0031] (2) Among them, the weight factor is introduced The aim is to balance different load cases and derive various generative design recommendations. Refers to the load; = 1, …, represents different load vectors, where is the number of load cases. Since the ligand quality conditions for each discipline are different, different finite element models may also need to be used, each with a specific stiffness matrix represents, m = 1, ..., , The number of finite element models.

[0032] It can be seen that in this multi-model optimization framework, the design variables are the coupling parameters between each load case, model or procedure. The best candidate design solution is obtained through this multi-model optimization framework, such as Figure 3 shown.

[0033] Step 102: construct a response surface prediction model based on the initial load path optimization scheme, and generate a training data set through experimental design.

[0034] In the present application, the response surface prediction model is a proxy model that approximates the relationship between input parameters and output responses through mathematical functions. Experimental design refers to a strategy for planning training data points through sampling methods (such as Latin hypercube or full factorial design).

[0035] The computer constructs a response surface prediction model such as a polynomial or radial basis function based on the input-output relationship in the initial load path optimization scheme. The system generates uniformly distributed sample points in the parameter space through the experimental design method, calls the simulation module to calculate the corresponding mechanical response, and forms a training data set to support model training.

[0036] Step 103: input the training data set into a machine learning enhancement module to generate a design space classifier and cluster the simulation results.

[0037] In the embodiment of the present application, the machine learning enhancement module refers to a software component that integrates a classification or regression algorithm to improve the efficiency of design space exploration. The design space classifier is a binary or multi-class discriminant model constructed through supervised learning. Clustering is the grouping operation of similar simulation results in unsupervised learning.

[0038] The computer inputs the training data set into the machine learning enhancement module, and uses the support vector machine or random forest algorithm to train the design space classifier to distinguish between feasible and infeasible designs. The system also performs K-means or hierarchical clustering on the simulation results, groups the design solutions according to the similarity of mechanical properties, and forms an optimization candidate set.

[0039] Specifically, machine learning is added to response surface optimization (RSM), which uses expert simulation to visually inspect and label each run, and uses cluster classification to classify runs with similar results, recorded as r(x), which will produce The result groups are as follows. Figure 4 As shown, all runs are marked and a classifier is generated, which can be expressed as: , ) is a scalar, Is r(x) belongs to probability.

[0040] The classifier is used in response surface-based optimization to determine the optimal design for a particular group, expressed as formula (3): (3) where the synthetic field is determined using the RSM itself, Indicates that the result belongs to The minimum probability required for a group.

[0041] Step 104 , selecting the optimal design group that meets the preset constraint conditions according to the classifier, and performing plastic deformation and damage failure simulation using the material constitutive model specified by the preset standard.

[0042] In the embodiments of the present application, the material constitutive model is a mathematical expression that describes the stress-strain relationship of the material. Plastic deformation refers to the irreversible deformation behavior of the material after it exceeds the elastic limit. Damage failure simulation is a numerical method that simulates the entire process of the material from damage initiation to fracture.

[0043] The computer selects the optimal design group that meets the constraints based on the classifier, calls the finite element analysis module to load the material constitutive model specified by the preset standard (such as the Johnson-Cook or Gurson model), and performs nonlinear dynamic simulation to evaluate the plastic deformation and damage failure behavior, and quantify the structural crashworthiness indicators.

[0044] Step 105 , performing multi-dimensional experimental tests on the samples in the optimal design group to obtain measured data to calibrate the failure parameters of the simulation model.

[0045] In the embodiments of the present application, the multi-dimensional experimental test includes physical tests under different load conditions such as static compression and dynamic impact. Measured data refers to physical quantities such as displacement and strain collected by sensors. Failure parameter calibration is the process of matching simulation results with experimental data through reverse optimization.

[0046] The computer-controlled testing machine conducts multi-condition mechanical tests on the samples in the optimal design group, and collects load-displacement curves and failure modes through force sensors and high-speed cameras. The system adjusts the failure parameters (such as fracture strain) in the simulation model based on the least squares method or genetic algorithm to ensure that the error between the numerical prediction and the actual measurement is within the threshold.

[0047] Step 106 , performing a head-on collision test verification based on the calibrated simulation model, and outputting a final design solution that meets multiple performance constraints.

[0048] In the embodiment of the present application, the head-on collision test verification is a dynamic response test of a simulated structure under an axial impact load. The multiple performance constraints include comprehensive index requirements such as stiffness, energy absorption, and peak force.

[0049] The computer uses the calibrated simulation model to perform numerical simulation of head-on collision conditions and extract key indicators such as energy absorption and collision force waveform. The system uses a multi-objective optimization algorithm (such as NSGA-II) to select the final design solution that meets the constraints of lightweight and crashworthiness, and outputs a set of geometric and performance parameters.

[0050] Step 107, integrating the optimization results of each stage in the final design solution to generate an integrated structural performance report of the complex thin-walled component.

[0051] In the embodiment of the present application, the integrated structural performance report is a standardized document summarizing the design iteration process, simulation and experimental comparison data.

[0052] The computer automatically integrates topology optimization results, machine learning classification boundaries, simulation and experimental comparison curves and other data to generate a PDF report containing convergence analysis and robustness evaluation to support design decision-making and archiving.

[0053] The embodiment of this application realizes the fully automated optimization of complex thin-walled parts from conceptual design to performance verification through the collaborative process of multi-model topology optimization, machine learning-assisted classification, and experimental-simulation closed-loop verification. While ensuring computational efficiency, the system significantly improves the mechanical performance reliability and multi-objective coordination capabilities of the design, and ultimately outputs a lightweight, highly crashworthy structural solution and a traceable complete performance report.

[0054] Optionally, the step 101 includes: Step 1011, adjusting the material distribution density through a density-based topology optimization algorithm to determine the optimal load path that meets statics, NVH and collision performance.

[0055] In the embodiment of the present application, the density-based topology optimization algorithm is a numerical method for optimizing structural performance by adjusting the material distribution density (0-1 continuous variable), where a density close to 1 indicates solid material and a density close to 0 indicates voids. Static performance refers to the mechanical properties of a structure such as stiffness and strength under static loads. NVH performance is a comprehensive evaluation index of noise, vibration, and harshness, reflecting the dynamic response characteristics of the structure. Collision performance refers to crashworthiness indicators such as energy absorption, peak force, and deformation mode of a structure under impact loads. The optimal load path is the best distribution of materials in the structure to maximize target performance (such as stiffness and lightweight).

[0056] The computer uses density-based topology optimization algorithms (such as SIMP method) to iteratively adjust the material density distribution on the finite element model with the goal of minimizing structural flexibility or maximizing natural frequency. The system calculates the impact of design variables (element density) on statics, NVH and crash performance through sensitivity analysis, generates the optimal load path that meets multi-objective constraints, and outputs the optimized geometric configuration in STL or mesh format.

[0057] Step 1012: establishing a multidisciplinary coupling optimization model, wherein the multidisciplinary coupling optimization model balances the stiffness matrices of different finite element models through weight factors.

[0058] In the embodiment of the present application, the multidisciplinary coupling optimization model is a joint optimization framework that integrates multiple disciplines (such as structural mechanics and acoustics), and realizes interdisciplinary variable interaction through coupling equations. The weight factor is a coefficient used to adjust the contribution ratio of the objective function of different disciplines to ensure that the optimization direction meets the global requirements. The stiffness matrix of the finite element model is a matrix that describes the stiffness characteristics of the structural unit and is used to calculate mechanical responses such as displacement and stress.

[0059] Based on the topology optimization results, the computer establishes a multidisciplinary coupling optimization model including statics, NVH and collision analysis. The system unifies the objective functions of each discipline (such as stiffness matrix eigenvalues ​​and modal frequencies) into a comprehensive goal through the weighted summation method, and uses a gradient optimization algorithm (such as MMA) to adjust the weight factor to balance the contributions of the stiffness matrices of different disciplines, and outputs a design variable update plan that takes into account multiple performance requirements.

[0060] Step 1013, mapping the design variables under multiple load conditions to a unified parameter space, and outputting a set of candidate solutions including structural efficiency and performance indicators.

[0061] In the embodiments of the present application, multiple load conditions are multiple load combinations (such as bending, torsion, and impact) that the structure bears in actual working conditions. Design variables are adjustable parameters (such as thickness and material density) during the optimization process. Unified parameter space is to normalize the design variables of different disciplines to the same mathematical space for interdisciplinary optimization. Structural efficiency is a performance indicator under unit mass (such as specific stiffness and specific strength). Performance indicators are parameters that quantify structural characteristics (such as displacement, stress, and natural frequency).

[0062] The computer maps the design variables (such as unit density and thickness) under statics, NVH and collision conditions to a unified parameter space, and uses principal component analysis (PCA) or radial basis function (RBF) interpolation methods to achieve data dimension reduction and association. The system evaluates the structural efficiency and performance indicators of each candidate solution through the Pareto frontier screening algorithm, and outputs a set of optimization solutions that meet multi-objective trade-offs for subsequent decision-making.

[0063] The embodiment of the present application determines the optimal material distribution through a density-based topology optimization algorithm, combines a multidisciplinary coupling model to balance stiffness, NVH and collision performance, and ultimately generates candidate solutions with high structural efficiency in a unified parameter space.

[0064] Optionally, the step 102 includes: Step 1021, extracting the design variables and response variables in the candidate solution set, and constructing a second-order polynomial response surface function.

[0065] In the embodiments of the present application, the design variables are independent parameters that can be adjusted during the optimization process, such as material thickness, geometric dimensions, or topological optimization density distribution parameters. The response variables are output performance indicators determined by the design variables, such as stress, strain, displacement, or natural frequency. The second-order polynomial response surface function is a proxy model based on quadratic polynomial regression, which is used to approximately describe the nonlinear relationship between the design variables and the response variables.

[0066] The computer extracts design variables (such as unit density and thickness distribution) and corresponding response variables (such as maximum stress and modal frequency) from the candidate solution set and constructs a second-order polynomial response surface function. The system uses the least squares method to fit the polynomial coefficients and establish a mapping relationship from the design space to the performance space.

[0067] Step 1022, generating a training data set by a Latin hypercube sampling method specified in a preset standard, wherein the training data set covers compression, tension, and shear stress states.

[0068] In the embodiments of the present application, Latin hypercube sampling method: a stratified random sampling strategy that ensures that the sample points are evenly distributed in the design space and there are no repeated projections. The training data set is a data set containing input design variables and output response variables, which is used for model training and verification. Compression, tension and shear stress states are the classification of mechanical behaviors of materials under unidirectional compression, unidirectional tension and shear loads.

[0069] The computer generates uniformly distributed sample points in the design variable space according to the Latin hypercube sampling method specified in the preset standard (such as ISO 18404). The system calls the finite element analysis module to apply compression, tension and shear load conditions to each sample point, calculates the corresponding stress, strain and damage factor, and forms a training data set covering multiple stress states. The data set is stored as a structured array, which contains input variables (such as load size, geometric parameters) and output variables (such as Von Mises stress, plastic strain).

[0070] Step 1023, using the training data set to fit the response surface prediction model, and outputting predicted values ​​of stress, deformation and damage factor.

[0071] In the embodiments of the present application, the response surface prediction model is a proxy model constructed by mathematical fitting, which is used to quickly predict the performance response of unsampled points. The stress prediction value is an estimate of the internal stress distribution of the structure under load output by the model. The deformation prediction value is an estimate of the structural displacement or shape change output by the model. The damage factor prediction value is an indicator that quantifies the risk of material failure, such as the Johnson-Cook damage parameter or equivalent plastic strain.

[0072] The computer uses the training data set to fit the second-order polynomial response surface model and uses the cross-validation method to evaluate the model accuracy (such as R² ≥ 0.95). The system solves the polynomial coefficient matrix to achieve rapid mapping of input design variables to output responses (stress, deformation, damage factor). For new design points, the model directly outputs the predicted value, replacing the time-consuming high-fidelity simulation, and supporting subsequent optimization iterations.

[0073] The embodiment of the present application achieves fast and high-precision prediction of complex mechanical responses by constructing a second-order polynomial response surface model, combining Latin hypercube sampling with multi-stress state training data. This technical solution significantly reduces the computational cost, avoids repeated calls to finite element analysis, and ensures the reliability of the proxy model under typical working conditions such as compression, tension and shear, providing efficient data-driven support for multi-objective optimization.

[0074] Optionally, the step 103 includes: Step 1031 , extracting features from the visualization data of the simulation results, and marking key stress concentration areas in the design space.

[0075] In the embodiments of the present application, the visualization data of the simulation results are graphical output data such as stress cloud maps and deformation animations generated by post-processing of finite element analysis. Feature extraction is a computational process that identifies and quantifies key geometric or mechanical features from visualization data. Critical stress concentration areas are local areas in the structure where the stress level is significantly higher than the average, which are usually the potential failure starting points.

[0076] The computer analyzes the visualized stress cloud map of the simulation results through image processing algorithms (such as edge detection and region growing method) to automatically identify the stress gradient mutation area. The system uses the threshold segmentation method to mark the stress concentration area that exceeds 80% of the material yield strength, and records its spatial coordinates, stress peak value and distribution range. These feature data are stored in a structured manner as input variables for subsequent cluster analysis.

[0077] Step 1032: Use K-means clustering algorithm to group the training data set to generate a classifier containing failure mode labels.

[0078] In the embodiment of the present application, the K-means clustering algorithm is an unsupervised machine learning method based on distance metric, which divides the data into K mutually exclusive subsets. The failure mode label is a classification mark that identifies the typical failure form of the structure (such as buckling, tearing, and fracture). The classifier is a mathematical model that can automatically assign category labels to new data.

[0079] The computer uses the characteristic parameters such as stress peak and deformation energy density in the training data set as the input vector of the K-means algorithm. The system presets the number of clusters K=3 (corresponding to three typical failure modes) and completes the data grouping by iteratively optimizing the centroid position. Each cluster center is assigned a failure mode label with clear physical meaning to form a three-classifier with a decision boundary. The classifier can automatically mark the simulation results of the new design scheme.

[0080] Step 1033, coupling the classifier with the response surface prediction model to establish probability-driven design screening rules.

[0081] In the embodiment of the present application, the probability-driven design screening rule is a scheme optimization criterion established based on statistical probability distribution. Model coupling is the process of data exchange between different mathematical models through interface protocols.

[0082] The computer couples the classifier with the response surface prediction model through the data bus interface to establish a joint evaluation framework. The system calculates the failure probability distribution of each design point in the response surface model prediction results, and generates a multidimensional screening matrix in combination with the failure mode labels output by the classifier. For candidate design solutions, when the probability of the main failure mode exceeds the preset threshold (such as buckling probability>30%), the system automatically excludes it from the optimization queue.

[0083] The embodiment of the present application realizes intelligent classification and risk quantification assessment of design solutions through visual feature extraction, failure mode clustering and probabilistic coupling analysis. This technical solution establishes a closed-loop link from simulation data to design decision-making, which significantly improves the recognition accuracy of potential failure modes while ensuring computational efficiency, and provides data support for structural reliability optimization. The probabilistic screening rules output by the system can effectively reduce the number of iterations in the later verification stage.

[0084] Optionally, the step 104 includes: Step 1041, call the MAT_PLASTICITY_COMPRESSION_TENSION_MAT_124 keyword to simulate the compression-tension dual-mode response of the material.

[0085] In the embodiment of the present application, the MAT_PLASTICITY_COMPRESSION_TENSION_MAT_124 keyword is a dual-mode elastic-plastic constitutive model in the LS-DYNA material model, which is used to distinguish the mechanical behavior of the material under compression and tension. The compression-tension dual-mode response is that the material exhibits asymmetric yield strength and hardening characteristics under different stress states (compression / tension).

[0086] The computer calls the material model interface of the finite element solver and activates the MAT_124 constitutive relationship. The system inputs the elastic modulus, Poisson's ratio, yield strength and hardening curve parameters of the material in compression and tension respectively. During the calculation process, the program automatically switches to compression or tension mode according to the current stress state of the unit, and uses different plastic flow laws to update the stress. This model can accurately simulate the asymmetric plastic behavior of metal materials under complex loads.

[0087] Step 1042, calculating the dynamic yield stress under high strain rate conditions based on the Cowper-Symonds strain rate model.

[0088] In the embodiments of the present application, the Cowper-Symonds strain rate model is an empirical formula that describes the change of material yield strength with strain rate. The high strain rate condition is a dynamic loading condition with a strain rate exceeding 10^2 s^-1. The dynamic yield stress is the critical stress value at which the material begins to undergo plastic deformation under dynamic load.

[0089] At each incremental step, the computer first obtains the current strain rate ε' through the central difference method. The system calls the Cowper-Symonds model parameters (material constants C and P) and combines the static yield stress σy to calculate the dynamic yield stress correction value in real time. This correction value is substituted into the J2 plastic flow criterion to update the equivalent stress of the current step. For strain rate sensitive materials (such as high-strength steel), this process can accurately reflect its dynamic strengthening effect.

[0090] Specifically, in order to accurately simulate the behavior of materials under different stress conditions, the Cowper-Symonds strain rate model and two load curves that scale the yield stress values ​​in compression and tension respectively are used. Through the Cowper-Symonds strain rate model, the strain rate effect of the material is taken into account, thereby accurately simulating the response of the material under high strain rate conditions.

[0091] Step 1043, using the MAT_ADD_DAMAGE_DIEM keyword to evolve the damage factor, the damage factor is iteratively updated by the equivalent plastic displacement increment.

[0092] In the embodiment of the present application, MAT_ADD_DAMAGE_DIEM keyword: material extension option used in LS-DYNA to define damage accumulation and failure criteria. Equivalent plastic displacement increment: damage evolution control parameter based on plastic strain and characteristic length. Damage factor: scalar field variable that quantifies the degree of material damage (0 means no damage, 1 means complete failure).

[0093] The computer activates the DIEM damage model in the material properties and sets parameters such as the initial damage threshold and critical displacement. The system calculates the equivalent plastic displacement increment at each time step and adds it to the damage variable D. When D reaches the critical value D_crit, the program automatically reduces the unit stiffness until the failed unit is deleted. This process is implemented through an explicit iterative algorithm to ensure that the damage evolution and dynamic response are updated synchronously.

[0094] Specifically, the MAT_ADD_DAMAGE_DIEM keyword in LSDY is used to characterize material properties and evolve the damage failure behavior of materials. The DIEM fracture failure model is divided into shear criterion and forward criterion. The shear criterion formula is as follows (4): (4) In the formula, is the equivalent plastic strain when unstable deformation occurs in biaxial tension, is the equivalent plastic strain when unstable deformation occurs under biaxial compression, is the normalized shear stress, is the shear stress parameter in biaxial tension, is the shear stress parameter in biaxial compression, is the maximum shear stress, is the equivalent stress, is the equivalent plastic strain when the material undergoes unstable deformation under different stress states, is the model parameter, and h represents a custom parameter.

[0095] The positive criterion is as follows: (5) In the formula, β is the normalized principal stress, is the first principal stress, is the equivalent stress, is the equivalent plastic strain when the material undergoes unstable deformation under different stress states, d, q, is the model parameter.

[0096] In the DIEM fracture failure model, the damage factor D is calculated using plastic displacement as shown in the following formula (6): (6) Where ΔD is the damage factor increment, is the plastic strain increment, is the failure plastic displacement of materials in different stress states.

[0097] The embodiment of the present application realizes high-precision simulation of materials under complex dynamic loads through the synergistic effect of compression-tension dual-mode constitutive model, strain rate sensitive strengthening and progressive damage model. This technical solution can simultaneously reflect the stress state dependence, strain rate effect and damage accumulation process of the material, providing a reliable theoretical framework for crashworthiness analysis and failure prediction.

[0098] Optionally, the step 105 includes: Step 1051, perform metallographic inspection and CT scanning at a preset position of the sample to obtain microstructure and porosity distribution data.

[0099] In the embodiments of the present application, metallographic testing is an experimental method for observing the microscopic structural characteristics of materials through an optical microscope or an electron microscope, which is used to analyze parameters such as grain size and phase composition. CT scanning is a computer tomography technology that obtains three-dimensional structural information inside the material through X-ray transmission imaging, and can quantify the distribution characteristics of defects such as porosity and cracks. The microstructure is the structural characteristics of the material such as grain morphology, phase distribution and interface characteristics at a microscopic scale. The porosity distribution data is a quantitative description of the proportion of the pore volume inside the material to the total volume and its spatial distribution.

[0100] The computer controls the metallographic microscope to automatically focus and capture images at preset positions on the sample, and uses image processing algorithms to identify microscopic features such as grain boundaries and precipitation phases. The system synchronously operates the industrial CT equipment to perform three-dimensional scanning at a resolution of 10μm, uses a threshold segmentation algorithm to extract the pore structure, and outputs a distribution cloud map of porosity changing with position. The test data is stored in a structured format to provide basic input for material performance analysis.

[0101] Step 1052, performing a quasi-static tensile test and a dynamic impact test, and collecting the stress-strain curve and fracture toughness parameters of the sample.

[0102] In the embodiments of the present application, the quasi-static tensile test is a uniaxial tensile test with a strain rate lower than 10^-3 s^-1, which is used to obtain basic mechanical parameters such as elastic modulus and yield strength of the material. The dynamic impact test is a high strain rate (10^2~10^4 s^-1) loading test implemented by a Hopkinson bar or a drop hammer device, which is used to evaluate the dynamic response characteristics of the material. The stress-strain curve is a graphical expression of the relationship between stress and strain of a material under the action of an external force, reflecting its elasticity, plasticity and fracture behavior. The fracture toughness parameter is an indicator that characterizes the material's ability to resist crack propagation, such as the critical stress intensity factor KIC or the fracture energy Gf.

[0103] The computer-controlled electronic universal testing machine performs quasi-static stretching at a strain rate of 0.001s^-1, and synchronously collects load-displacement data through force sensors and extensometers to automatically generate engineering stress-strain curves. The system then starts the split Hopkinson pressure bar device to perform dynamic impact tests at a strain rate of 500s^-1, and calculates the dynamic stress-strain response using strain gauge signals and high-speed photography data. All test data are filtered and stored in the material database.

[0104] Step 1053, establishing a mapping relationship between defect types and simulated damage results through X-ray nondestructive testing and scanning electron microscope fracture analysis.

[0105] In the embodiments of the present application, X-ray nondestructive testing is a nondestructive testing technology that uses the principle of X-ray transmission to detect internal defects in materials. Scanning electron microscope fracture analysis is to observe the morphological characteristics of the fracture surface of the material through a scanning electron microscope to determine the fracture mechanism. The defect type is a classification description of discontinuous structures such as pores, inclusions, and microcracks in the material. The simulated damage results are the material damage distribution and failure mode predicted by finite element analysis.

[0106] The computer-controlled X-ray flaw detector performs a panoramic scan of the test specimens, and uses a convolutional neural network to automatically identify the defect types (pores, slag inclusions, etc.) and their spatial distribution. The system then collects the fracture morphology at a magnification of 5000 times under a scanning electron microscope, and determines the fracture mode (dimples, cleavage, etc.) through a feature matching algorithm. These measured defect data are spatially registered with the simulated damage parameters (such as equivalent plastic strain distribution) to establish a quantitative mapping matrix of defect type-damage evolution.

[0107] The embodiment of this application realizes the full chain verification from microstructure to macroscopic mechanical response through the systematic comparison of multi-scale experimental detection and simulation data. This technical solution establishes a quantitative correlation model between material defect characteristics and damage evolution behavior, which significantly improves the accuracy of simulation prediction.

[0108] Optionally, the step 106 includes: Step 1061 , constructing a whole vehicle collision finite element model, and embedding the complex thin-walled component as a key component into the whole vehicle collision finite element model.

[0109] In the embodiment of the present application, the whole vehicle collision finite element model is a digital vehicle model composed of hundreds of thousands to millions of units, including the geometry and material property data of all key components such as the body and chassis.

[0110] The computer imports the parametric vehicle geometry model and generates a finite element mesh based on quadrilateral shell elements through an automatic meshing algorithm. The system embeds the optimized complex thin-walled component model into the specified position of the front cabin of the vehicle in the form of a component, and uses the multi-point constraint (MPC) method to achieve connection with adjacent components. All contact interfaces are defined as automatic surface-to-surface contact with a friction coefficient of 0.2 to ensure the continuity of the mechanical transmission path. The total degree of freedom of the model is controlled within 2 million to ensure calculation efficiency.

[0111] Step 1062, setting the initial collision velocity and boundary conditions specified by the preset standard, and calculating the deformation energy absorption of the complex thin-walled part.

[0112] In the embodiment of the present application, the initial collision velocity is the initial movement speed of the vehicle at the moment of collision, which is usually set according to regulatory requirements (such as 50km / h for frontal collision). Boundary conditions are boundary settings that constrain the freedom of motion of the model, including fixed constraints, symmetric constraints, etc. Deformation energy absorption is the mechanical energy dissipated during the plastic deformation of the structure, and is the core indicator for measuring crashworthiness.

[0113] The computer sets the initial velocity condition of 56km / h according to GB 11551 standard, and gives the initial kinetic energy to the vehicle mass unit through the *INITIAL_VELOCITY keyword. The system uses the explicit central difference method to solve the dynamic equation, and the time step is automatically adjusted to ensure numerical stability. During the calculation process, the internal energy change curve of the complex thin-walled parts is monitored in real time, and the deformation energy absorption is calculated by integrating the plastic strain energy density. The data sampling frequency is set to 10kHz to ensure the accuracy of the results.

[0114] Step 1063 , comparing the simulated deformation mode with the measured data, verifying the stiffness matching and optimizing the process feasibility of the final design solution.

[0115] In the embodiments of the present application, the simulated deformation mode is the deformation form such as structural buckling and folding predicted by finite element analysis. The measured data is the result of physical collision test collected by high-speed camera or optical measurement system. The stiffness matching degree is a quantitative evaluation index of the degree of coincidence between the simulated and measured load-displacement curves. The process feasibility is the feasibility assessment of the design scheme in the manufacturing process such as stamping and welding.

[0116] The computer performs spatiotemporal registration of the deformation sequence of the thin-walled part output by the simulation with the measured image collected by the high-speed camera, and uses the image correlation coefficient (CC) algorithm to quantify the similarity of the deformation mode. The system automatically extracts the force-displacement data of the key measuring points and calculates the stiffness matching index (R²≥0.9 is qualified). For areas that do not meet the standards, the optimization module is triggered to adjust the material thickness distribution or the rib layout, and the stamping forming limit diagram (FLD) is checked to ensure the feasibility of the process, and it is iterated until all acceptance criteria are met.

[0117] In the embodiment of the present application, a systematic evaluation of the performance of complex thin-walled parts is achieved through closed-loop verification of vehicle-level collision simulation and test data. This technical solution ensures that the component design is accurately matched with the safety requirements of the vehicle, while taking into account the manufacturing process constraints, forming a reliable transformation path from digital models to physical products.

[0118] Based on the same inventive concept, the embodiment of the present application also provides an integrated large-scale complex thin-walled die-casting component structure performance collaborative design device for realizing the integrated large-scale complex thin-walled die-casting component structure performance collaborative design method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more integrated large-scale complex thin-walled die-casting component structure performance collaborative design device embodiments provided below can refer to the limitations of the integrated large-scale complex thin-walled die-casting component structure performance collaborative design method above, and will not be repeated here.

[0119] In an exemplary embodiment, Figure 5 As shown, an integrated large-scale complex thin-wall die-casting component structure performance collaborative design device 20 is provided, comprising: A processing module 201 is used to input the design parameters of the complex thin-walled part into a multi-model topology optimization framework to generate an initial load path optimization solution; Building a response surface prediction model based on the initial load path optimization scheme, and generating a training data set through experimental design; Inputting the training data set into a machine learning enhancement module to generate a design space classifier and cluster the simulation results; An optimal design group satisfying preset constraints is selected according to the classifier, and plastic deformation and damage failure simulation is performed using a material constitutive model specified by a preset standard; Conducting multi-dimensional experimental tests on the samples in the optimal design group to obtain measured data to calibrate the failure parameters of the simulation model; An output module 202 is used to perform a head-on collision test verification based on the calibrated simulation model and output a final design solution that meets multiple performance constraints; The optimization results of each stage in the final design solution are integrated to generate an integrated structural performance report of complex thin-walled parts.

[0120] Optionally, the processing module 201 is further configured to: Adjust material distribution density through density-based topology optimization algorithm to determine the optimal load path that meets statics, NVH and crash performance; Establishing a multidisciplinary coupling optimization model, wherein the multidisciplinary coupling optimization model balances the stiffness matrices of different finite element models through weight factors; The design variables under multiple load conditions are mapped to a unified parameter space, and a set of candidate solutions containing structural efficiency and performance indicators is output.

[0121] Optionally, the processing module 201 is further configured to: Extracting design variables and response variables from the candidate solution set and constructing a second-order polynomial response surface function; Generate a training data set by a Latin hypercube sampling method specified in a preset standard, wherein the training data set covers compression, tension and shear stress states; The response surface prediction model is fitted using the training data set to output predicted values ​​of stress, deformation and damage factor.

[0122] Optionally, the processing module 201 is further configured to: Performing feature extraction on the visualization data of the simulation results to mark key stress concentration areas in the design space; Using a K-means clustering algorithm to group the training data set to generate a classifier containing failure mode labels; The classifier is coupled with the response surface prediction model to establish a probability-driven design screening rule.

[0123] Optionally, the processing module 201 is further configured to: Call the MAT_PLASTICITY_COMPRESSION_TENSION_MAT_124 keyword to simulate the compression-tension bimodal response of the material; Calculate the dynamic yield stress under high strain rate conditions based on the Cowper-Symonds strain rate model; The MAT_ADD_DAMAGE_DIEM keyword is used to evolve the damage factor, which is iteratively updated by the equivalent plastic displacement increment.

[0124] Optionally, the processing module 201 is further configured to: Perform metallographic testing and CT scanning at preset locations on the sample to obtain microstructure and porosity distribution data; Performing quasi-static tensile tests and dynamic impact tests to collect stress-strain curves and fracture toughness parameters of the samples; Through X-ray nondestructive testing and scanning electron microscope fracture analysis, the mapping relationship between defect type and simulated damage results is established.

[0125] Optionally, the output module 201 is further used for: Constructing a whole vehicle collision finite element model, and embedding the complex thin-walled component as a key component into the whole vehicle collision finite element model; Setting the initial collision velocity and boundary conditions specified by the preset standard, and calculating the deformation energy absorption of the complex thin-walled part; The simulated deformation pattern is compared with the measured data to verify the stiffness matching and optimize the process feasibility of the final design.

[0126] The embodiment of this application realizes the fully automated optimization of complex thin-walled parts from conceptual design to performance verification through the collaborative process of multi-model topology optimization, machine learning-assisted classification, and experimental-simulation closed-loop verification. While ensuring computational efficiency, the system significantly improves the mechanical performance reliability and multi-objective coordination capabilities of the design, and ultimately outputs a lightweight, highly crashworthy structural solution and a traceable complete performance report.

[0127] In an exemplary embodiment, a computer device is provided. The computer device may be a server or a terminal. The internal structure diagram thereof may be as follows: Figure 6 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. Among them, the processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store the structural performance collaborative design data of an integrated large-scale complex thin-walled die-casting component. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a method for collaborative design of the structural performance of an integrated large-scale complex thin-walled die-casting component is implemented.

[0128] Those skilled in the art will understand that Figure 6 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0129] In an exemplary embodiment, a computer device is further provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.

[0130] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0131] In an exemplary embodiment, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0132] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0133] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0134] The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. The non-relational database may include a distributed database based on blockchain, etc., but is not limited thereto. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but is not limited thereto.

[0135] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0136] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A method for collaborative design of structural performance of large-scale, complex, thin-walled die-casting components, characterized in that: The structural performance collaborative design method of the integrated large-scale complex thin-walled die-casting component includes: Input the design parameters of complex thin-walled parts into the multi-model topology optimization framework to generate the initial load path optimization solution; Building a response surface prediction model based on the initial load path optimization scheme, and generating a training data set through experimental design; Inputting the training data set into a machine learning enhancement module to generate a design space classifier and cluster the simulation results; An optimal design group satisfying preset constraints is selected according to the classifier, and plastic deformation and damage failure simulation is performed using a material constitutive model specified by a preset standard; Conducting multi-dimensional experimental tests on the samples in the optimal design group to obtain measured data to calibrate the failure parameters of the simulation model; Perform head-on collision tests based on the calibrated simulation model and output the final design solution that meets multiple performance constraints; The optimization results of each stage in the final design solution are integrated to generate an integrated structural performance report of complex thin-walled parts.

2. The method for collaborative design of structural performance of large-scale complex thin-walled die-casting components according to claim 1 is characterized in that: The step of inputting the design parameters of the complex thin-walled part into the multi-model topology optimization framework to generate an initial load path optimization solution includes: Adjust material distribution density through density-based topology optimization algorithm to determine the optimal load path that meets statics, NVH and crash performance; Establishing a multidisciplinary coupling optimization model, wherein the multidisciplinary coupling optimization model balances the stiffness matrices of different finite element models through weight factors; The design variables under multiple load conditions are mapped to a unified parameter space, and a set of candidate solutions containing structural efficiency and performance indicators is output.

3. The method for collaborative design of structural performance of large-scale complex thin-walled die-casting components according to claim 1 is characterized in that: The step of constructing a response surface prediction model based on the initial load path optimization scheme comprises: Extracting design variables and response variables from the candidate solution set and constructing a second-order polynomial response surface function; Generate a training data set by a Latin hypercube sampling method specified in a preset standard, wherein the training data set covers compression, tension and shear stress states; The response surface prediction model is fitted using the training data set to output predicted values ​​of stress, deformation and damage factor.

4. The method for collaborative design of structural performance of large-scale complex thin-walled die-casting components according to claim 1, characterized in that: The step of inputting the training data set into the machine learning enhancement module comprises: Performing feature extraction on the visualization data of the simulation results to mark key stress concentration areas in the design space; Using a K-means clustering algorithm to group the training data set to generate a classifier containing failure mode labels; The classifier is coupled with the response surface prediction model to establish a probability-driven design screening rule.

5. The method for collaborative design of structural performance of large-scale complex thin-walled die-casting components according to claim 1, characterized in that: The step of using the material constitutive model specified in the preset standard to perform plastic deformation and damage failure simulation includes: Call the MAT_PLASTICITY_COMPRESSION_TENSION_MAT_124 keyword to simulate the compression-tension bimodal response of the material; Calculate the dynamic yield stress under high strain rate conditions based on the Cowper-Symonds strain rate model; The MAT_ADD_DAMAGE_DIEM keyword is used to evolve the damage factor, which is iteratively updated by the equivalent plastic displacement increment.

6. The method for collaborative design of structural performance of large-scale complex thin-walled die-casting components according to claim 1, characterized in that: The step of performing multi-dimensional experimental testing on the samples in the optimal design group includes: Perform metallographic testing and CT scanning at preset locations on the sample to obtain microstructure and porosity distribution data; Performing quasi-static tensile tests and dynamic impact tests to collect stress-strain curves and fracture toughness parameters of the samples; Through X-ray nondestructive testing and scanning electron microscope fracture analysis, the mapping relationship between defect type and simulated damage results is established.

7. The method for collaborative design of structural performance of large-scale complex thin-walled die-casting components according to claim 1, characterized in that: The step of performing a head-on collision test verification based on the calibrated simulation model includes: Constructing a whole vehicle collision finite element model, and embedding the complex thin-walled component as a key component into the whole vehicle collision finite element model; Setting the initial collision velocity and boundary conditions specified by the preset standard, and calculating the deformation energy absorption of the complex thin-walled part; The simulated deformation pattern is compared with the measured data to verify the stiffness matching and optimize the process feasibility of the final design.

8. An integrated large-scale complex thin-wall die-casting component structural performance collaborative design device, characterized in that: The integrated large-scale complex thin-wall die-casting component structure performance collaborative design device comprises: A processing module is used to input the design parameters of complex thin-walled parts into the multi-model topology optimization framework to generate an initial load path optimization solution; Building a response surface prediction model based on the initial load path optimization scheme, and generating a training data set through experimental design; Inputting the training data set into a machine learning enhancement module to generate a design space classifier and cluster the simulation results; An optimal design group satisfying preset constraints is selected according to the classifier, and plastic deformation and damage failure simulation is performed using a material constitutive model specified by a preset standard; Conducting multi-dimensional experimental tests on the samples in the optimal design group to obtain measured data to calibrate the failure parameters of the simulation model; The output module is used to perform head-on collision test verification based on the calibrated simulation model and output the final design solution that meets multiple performance constraints; The optimization results of each stage in the final design solution are integrated to generate an integrated structural performance report of complex thin-walled parts.

9. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the method for collaborative design of structural performance of an integrated large-scale complex thin-walled die-casting component as described in any one of claims 1-7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for collaborative design of structural performance of an integrated large-scale complex thin-walled die-casting component described in any one of claims 1-7 are implemented.

Images