Method for predicting and partitioning simulation analysis of material performance of die castings

Abstract

This invention discloses a method for predicting and zoning simulation analysis of die-casting material properties, belonging to the field of data prediction and simulation analysis. Specifically, it involves collecting measured material property parameters from multiple calibration points on a die-casting sample, extracting corresponding process parameters from process flow analysis results, constructing a training dataset, and establishing a nonlinear prediction model between process parameters and material properties. The newly designed die-casting is then divided into mesh partitions, and the process parameters for each partition are extracted and input into the trained prediction model to obtain the predicted material property parameters for each partition. Finally, the mesh model is partitioned according to different performance parameters, and differentiated material constitutive models and evaluation criteria are configured for linear strength analysis, nonlinear abuse analysis, and fatigue analysis. This invention achieves high-precision and high-efficiency simulation analysis of die-castings, providing a reliable basis for lean structural design and lightweighting.

Description

Technical Field

[0001] This invention relates to the field of data prediction and simulation analysis methods, and more specifically, to a method for predicting the material properties and performing performance zoning simulation analysis of automotive die-casting parts, which is particularly suitable for the accurate performance evaluation of large integrated die-casting parts for automobiles. Background Technology

[0002] With the rapid development of the new energy vehicle industry, integrated die casting technology has been widely used due to its ability to significantly reduce the number of parts, lower vehicle weight, and improve production efficiency, such as for large structural components like the front towers and rear floor. These die-cast parts are typically connected to critical systems such as the chassis, and are subjected to complex stress conditions; their structural strength and durability are directly related to the safety performance of the entire vehicle.

[0003] Current engineering simulation analyses primarily rely on the finite element method for performance evaluation of these die-cast parts, encompassing linear strength analysis of the vehicle body, nonlinear strength analysis of local abuse, and fatigue life analysis. The accuracy of these analyses is highly dependent on the input material parameters, such as yield strength, tensile strength, and elongation at break. However, the material properties of large die-cast parts are not uniform. Due to the influence of the die-casting process, the flow distance, cooling rate, and solidification temperature field distribution of the molten metal within the mold cavity are uneven, leading to differences in the microstructure of different regions of the die-cast part (such as the near-gate region and the far-gate region). This, in turn, results in significant anisotropy and regional differences in its macroscopic mechanical properties.

[0004] Currently, the industry typically addresses this issue using two approaches: one is to assign a uniform set of material parameters obtained from standard sample testing to the entire die-cast part; the other is to cut and sample key areas of interest on the component and measure their material properties. The first approach is overly idealistic, ignoring objective differences in material properties, leading to significant discrepancies between simulation results and the actual physical state, making it impossible to accurately identify risk areas, and potentially causing design waste due to local performance redundancy. While the second approach offers higher accuracy, it requires destructive sampling of the physical component, resulting in high costs and a lengthy process, especially for newly designed die-cast parts where accurate material data cannot be obtained in advance during the design phase.

[0005] Therefore, how to efficiently and cost-effectively obtain accurate material performance parameters for large die-cast automotive parts during the design phase and effectively apply them to simulation analysis is a technical problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0006] In view of the above, the present invention aims to provide a method for predicting and zoning simulation analysis of die-casting material properties, so as to solve the technical problems of large deviation between simulation analysis material parameters and actual objects, high measurement costs, and low analysis efficiency in the current industry.

[0007] The technical solution adopted in this invention is as follows:

[0008] This invention provides a method for predicting and zoning simulation analysis of die-casting material properties, including:

[0009] The construction of a material performance prediction model specifically includes: selecting a die-cast sample and identifying multiple calibration points on its surface; performing material performance tests on each calibration point to obtain the measured material performance parameters at that point, wherein the measured material performance parameters include at least yield strength, tensile strength, and elongation at break; extracting multiple process parameters for each calibration point during the die-casting process through process flow analysis, wherein the process parameters include at least temperature, distance from the gate, and flow rate; using the process parameters as input and the measured material performance parameters as output, training a deep learning network model to obtain the trained material performance prediction model.

[0010] Predicting the material properties of a target die-cast part by partitioning it, wherein the target die-cast part includes at least a large integrated die-cast structural component for automobiles, specifically including: dividing the three-dimensional model of the target die-cast part into multiple mesh partitions; extracting the process parameters corresponding to each mesh partition through process flow analysis; inputting the process parameters of each mesh partition into the trained material property prediction model, and the model output is the predicted material property parameters for that partition, wherein the material property parameters include predicted yield strength, predicted tensile strength, and predicted elongation at break;

[0011] The performance zoning simulation analysis includes: marking the mesh model of the target die casting with performance zoning based on the predicted yield strength, predicted tensile strength, and predicted elongation at break of each zoning region; performing different types of simulation analysis based on the performance zoning markings; for linear strength analysis, using the predicted yield strength of each zoning region as the stress evaluation threshold for that zoning region; for nonlinear abuse strength analysis, determining the failure criterion for each zoning region based on the predicted elongation at break of break of each zoning region; and for fatigue life analysis, generating a unique SN curve for each zoning region based on the predicted tensile strength of each zoning region as the fatigue analysis input.

[0012] In at least one of the possible implementations, after meshing the three-dimensional model of the target die casting during the prediction of the partitioned material properties of the target die casting, the method further includes:

[0013] The step of merging adjacent mesh elements with the same or similar predicted material performance parameters into a single performance region simplifies the model size for subsequent simulation analysis.

[0014] In at least one of the possible implementations, when performing nonlinear abuse strength analysis, the method further includes: constructing a constitutive curve of elastoplastic stress and strain unique to each performance partition based on the predicted yield strength and predicted tensile strength of that partition, and using it as the material input model for that partition.

[0015] In at least one of the possible implementations, when performing fatigue life analysis, based on the predicted tensile strength of each zone, a zone-specific SN curve is generated by fitting empirical formulas or a material database to replace the traditional single-material SN curve.

[0016] In at least one of the possible implementations, the deep learning network model includes: a BP neural network model;

[0017] The training process of the BP neural network model includes: data normalization processing, network structure initialization, forward propagation to calculate error, and back propagation to adjust weights and thresholds in an iterative manner until the error between the network output and the measured value converges to a preset accuracy.

[0018] In at least one of the possible implementations, the process parameters extracted from the process flow analysis further include at least one of: cooling rate or cavity pressure.

[0019] In at least one possible implementation, the linear strength analysis further includes a post-processing step: displaying stress results on contour plots of different performance partitions of the model and comparing them with the predicted yield strength of each partition to accurately identify high stress risk areas or material performance redundancy areas.

[0020] Compared with existing technologies, this invention employs deep learning algorithms, which can accurately capture the nonlinear relationship between process parameters and material properties. The prediction results are far more accurate than traditional empirical formulas or single assignments. Once the model is trained, it can be applied to the design analysis of the same series of products, eliminating the need for extensive physical sampling tests for each new structure, significantly shortening the R&D cycle and reducing R&D costs. The performance partitioning simulation concept involved ensures that the analysis results closely match the actual state, accurately identifying weak areas caused by the process and avoiding potential failure risks. At the same time, it can also identify excess areas, providing a quantitative basis for structural weight reduction and material saving, truly realizing lean and lightweight structural design. In particular, the concept proposed in this invention is not only applicable to integrated die-cast parts for new energy vehicles, but can also be extended to the analysis of other metal or non-metal formed parts with uneven performance due to the process. Attached Figure Description

[0021] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described below with reference to the accompanying drawings, wherein:

[0022] Figure 1 A flowchart for performance prediction and performance zoning simulation analysis of die-casting materials provided in this embodiment of the invention;

[0023] Figure 2 A flowchart for predicting material properties provided in an embodiment of the present invention;

[0024] Figure 3 This is a flowchart of linear strength analysis after performance partitioning provided in an embodiment of the present invention;

[0025] Figure 4 A flowchart for nonlinear abuse intensity analysis after performance partitioning provided in this embodiment of the invention;

[0026] Figure 5 This is a flowchart of fatigue analysis after performance partitioning provided in an embodiment of the present invention;

[0027] Figure 6 A schematic diagram of the sampling and testing location provided in an embodiment of the present invention;

[0028] Figure 7 Temperature distribution diagram of mold flow provided for embodiments of the present invention;

[0029] Figure 8 This is a schematic diagram of tensile strength prediction and zoning provided in an embodiment of the present invention;

[0030] Figure 9 A schematic diagram of yield strength prediction and zoning provided in an embodiment of the present invention;

[0031] Figure 10 A schematic diagram of fracture elongation prediction and zoning provided in an embodiment of the present invention;

[0032] Figure 11 This is a rear floor strength cloud map after performance partitioning provided in an embodiment of the present invention;

[0033] Figure 12 This is a schematic diagram of the performance-unpartitioned fatigue analysis results provided in an embodiment of the present invention;

[0034] Figure 13 A schematic diagram of the performance partition fatigue analysis results provided in an embodiment of the present invention;

[0035] Figure 14 A schematic diagram illustrating the performance-unpartitioned abuse intensity analysis results provided in an embodiment of the present invention;

[0036] Figure 15 This is a schematic diagram illustrating the performance partition abuse analysis results provided in an embodiment of the present invention. Detailed Implementation

[0037] Embodiments of the present invention are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0038] In response to the aforementioned problems, the core concept of this invention is to deeply integrate process simulation and performance simulation by constructing a predictive model from process parameters to material properties and performing performance partitioning simulation based on the prediction results.

[0039] Specifically, this invention proposes an embodiment of a method for predicting and zoning simulation analysis of die-casting material properties. Specifically, as follows: Figures 1-5 The process shown includes three key steps:

[0040] The first step is to construct a material performance prediction model. This step begins by selecting a representative die-cast sample and scientifically and uniformly selecting a large number of calibration points on it. Then, the actual material performance parameters for each calibration point are obtained experimentally, primarily yield strength, tensile strength, and elongation at break. Simultaneously, a high-precision process flow analysis is performed on the die-casting process of the sample, extracting the process parameters corresponding to each calibration point location, such as, but not limited to, the melt temperature at that point, the linear distance from the gate or the runner distance, and the filling velocity. This forms a one-to-one input / output sample dataset. Next, a deep learning algorithm, preferably a BP neural network adept at handling nonlinear mapping relationships, is used to train the model on this dataset. Through continuous iterative optimization of the weights and thresholds within the network, a prediction model that accurately describes the complex functional relationship between process parameters and material properties is finally obtained.

[0041] The second step is the performance prediction of the target die casting within its zones. When performance analysis is required for a newly designed, similar integrated die casting, the CAD model is first meshed using finite element methods, resulting in thousands of tiny mesh elements, which form the initial zones. Subsequently, the same process flow analysis is performed on the new die casting to extract the process parameters at the center point or nodes of each mesh zone. These process parameters are then imported in batches into the prediction model trained in the first step, allowing the model to quickly calculate the yield strength, tensile strength, and elongation at break for each mesh zone. To facilitate subsequent analysis, adjacent mesh elements with similar predicted performance values ​​can be merged to form several larger performance zones.

[0042] The third step is to perform performance zoning simulation analysis, which is the final application and value demonstration of this embodiment. Based on the performance parameters of each zone obtained in the second step, different material property labels are assigned to the mesh model in the finite element preprocessing to achieve performance zoning. On this basis, differentiated settings are made for different types of simulation analysis. Three examples are given below for reference:

[0043] Firstly, for linear strength analysis, a single yield strength threshold is no longer used in post-processing evaluation. Instead, each performance zone is evaluated based on its own predicted yield strength to determine whether the stress level in that zone is safe.

[0044] Secondly, for nonlinear abuse strength analysis, a dedicated elastoplastic stress-strain constitutive curve is constructed for each performance zone using its predicted yield strength and predicted tensile strength. After the analysis, each zone is assessed for fracture failure based on its predicted elongation at break.

[0045] Thirdly, for fatigue life analysis, the predicted tensile strength of each performance zone is used to generate a unique SN curve in fatigue analysis software (such as Ncode); during fatigue analysis, different zones call their respective SN curves to calculate damage, thereby obtaining a more accurate fatigue life distribution.

[0046] Using the flowchart as an example, the process of the above method embodiment is explained in detail:

[0047] First, select an integrated post-casting floor sample that has already been die-cast. On this sample, based on its geometric characteristics and mold flow analysis results, uniformly select N calibration points (e.g., N=50) to ensure that these points cover all typical areas from near the gate to the far end, and from the thin-walled area to the thick-walled area.

[0048] Then, using methods such as wire cutting, specimens containing calibration points are cut from the sample, and tensile tests are performed according to national standards to obtain the measured yield strength ReL at each calibration point. i Measured tensile strength Rm i And measured fracture elongation A i , i=1,2,...,N.

[0049] Furthermore, the die-casting process of the rear floor sample was accurately simulated using professional die-casting mold flow analysis software. From the simulation results, process parameters corresponding to the spatial coordinates of N calibration points were extracted: the filling temperature T at each point... i The shortest runner distance D from this point to the gate. i and the filling flow rate V at that point. iTherefore, a training set containing N samples can be constructed, where the input X... i =[T i D i V i ], output Y i =[ReL i ,Rm i A i ].

[0050] In MATLAB, using its Deep Learning Toolbox, a backpropagation (BP) neural network is constructed. The network structure can be designed as follows: 3 neurons in the input layer (corresponding to T, D, V), 3 neurons in the output layer (corresponding to ReL, Rm, A), and two hidden layers. The number of neurons is determined through trial and error or empirical formulas to be [10, 5]. Next, the input and output data are normalized, and the training function is set to trainlm with a learning rate of 0.01 and a target error of 1e-5. The dataset is randomly divided into a training set, a validation set, and a test set in a ratio of 7:2:1. Training of the network then begins. Training is stopped when the validation set error no longer decreases after multiple consecutive iterations, and the trained network model is saved.

[0051] After completing the above preparations, the next step is to predict the material properties of the target die-cast parts by region. Here, we propose a task to analyze the newly designed rear floor die-cast parts of the same series. First, in finite element preprocessing software (such as HyperMesh), the CAD model of the new rear floor is meshed with high quality, generating M mesh elements, for example, M=100,000. These mesh elements are defined as the preliminary analysis regions.

[0052] Subsequently, in the mold flow analysis software for the new rear floor, the die-casting process simulation was also performed. After the simulation was completed, the process parameters corresponding to the center point coordinates of each mesh unit were exported in batches through the software's post-processing interface, such as temperature T. j Distance Dj, Flow velocity V j Let j = 1, 2, ..., M. In practice, a data interface script can be written, taking these M sets of process parameters as input, and calling the BP neural network model trained in step one to make predictions. The model can then quickly output the predicted yield strength ReL for each grid cell. j Predicting tensile strength Rm j And predicted fracture elongation A j .

[0053] In some preferred embodiments of the present invention, to facilitate engineering applications, it is proposed that scripts be written in the post-processing software to merge adjacent mesh elements with similar predicted performance values. For example, adjacent elements with predicted yield strengths within the range of ±5 MPa are merged into one performance zone. Ultimately, the entire post-floor model is divided into P (P is much smaller than M) performance zones with different material property labels, each zone having a set of independent predicted performance parameters.

[0054] Finally, based on the aforementioned performance partitioning results, at least the following three types of simulation analyses can be carried out: (1) linear strength analysis, (2) nonlinear abuse strength analysis, and (3) fatigue life analysis.

[0055] (1) In the input file of the finite element solver (such as Abaqus), define a material card for each performance zone, assign it the corresponding elastic modulus and Poisson's ratio, and use the predicted yield strength ReL of the zone as an output control parameter; after submitting the calculation, view the stress cloud map of each zone in the post-processing, write the post-processing script, and automatically compare the Mises stress of each element with the predicted yield strength of the performance zone where the element is located; if the stress / yield strength ratio of a certain region exceeds 0.9, it is marked as a high-risk zone; if the ratio is generally lower than 0.3, it is marked as a design redundancy zone, which provides direction for subsequent morphology optimization.

[0056] (2) First, for each performance zone, a true stress-plastic strain curve is fitted using a material constitutive model (such as the Swift-Voce model) based on its predicted yield strength, predicted tensile strength, and predicted elongation at break. This curve serves as the elastoplastic material parameter specific to that zone. In Abaqus, each performance zone is assigned its own elastoplastic constitutive model, and a nonlinear analysis simulating local compression is submitted. After the analysis, the equivalent plastic strain (PEEQ) of each element is extracted. Then, through post-processing, the PEEQ value of each element is compared with the predicted elongation at break of the performance zone in which the element is located. If the PEEQ value exceeds 10% of the elongation at break of the zone, the zone is considered to have a risk of fracture.

[0057] (3) In fatigue analysis software (such as nCode DesignLife), first import the unit load stress results obtained from linear strength analysis; then, based on the predicted tensile strength of each performance zone in the material library, use the empirical formula of the steel SN curve built into the software, such as SN = b * N for reference. cWhere b and c are coefficients related to tensile strength, a unique SN curve is automatically generated for each performance region. In the fatigue analysis settings, the finite element elements of each performance region are associated with the corresponding unique SN curve, and after submitting the calculation, a more accurate fatigue damage cloud map and life prediction result that takes into account the differences in material performance regions can be obtained.

[0058] Through the above steps, this embodiment achieves high-precision and high-efficiency simulation analysis of the integrated post-floor die-casting component, providing solid data support for subsequent structural optimization design.

[0059] Finally, the performance partitioning method proposed earlier in this invention will be described in detail by combining the actual implementation process of automotive rear floor die-casting parts and front tower gearboxes as examples:

[0060] (a) Take samples from the rear floor casting as specimens for testing, such as from... Figure 6 The sampling area yielded 213 sets of material parameters, and combined with... Figure 7 The schematic temperature distribution allows us to obtain the temperature at the corresponding location in the process flow analysis, as well as parameters such as distance and speed.

[0061] For a detailed example of how to build and implement the prediction model, please refer to the following example:

[0062] %% Import Data

[0063] res=readmatrix('training data.xlsx')

[0064] %%Add path

[0065] addpath('goat\')

[0066] %% Divide the dataset into training and test sets

[0067] temp = randperm(2^13); Randomly sorts 2^13 sets of data.

[0068] P_train=res(temp(1:170),4:6)'; Select 80% for training input

[0069] T_train=res(temp(1:170),1)'; 80% is selected for training output.

[0070] M = size(P_train, 2);

[0071] P_test=res(temp(171:end),4:6)'; Select 20% for test input.

[0072] T_test=res(temp(171:end),1)'; 20% is selected for test output.

[0073] N = size(P_test, 2);

[0074] %%Data Normalization

[0075] [p_train,ps_input]=mapminmax(P_train,0,1); This normalizes the input data of the training set.

[0076] `p_test=mapminmax('apply',P_test,ps_input);` performs normalization on the test set input data.

[0077] [t_train,ps_output]=mapminmax(T_train,0,1); This normalizes the output data of the training set.

[0078] `t_test=mapminmax('apply',T_test,ps_output);` performs normalization on the test set output data.

[0079] %% Create a network

[0080] net=newff(p_train,t_train,7); Creates a neural network with 1 hidden layer and 7 neurons.

[0081] %%Setting training parameters

[0082] net.trainParam.epochs=1000; %Sets the number of iterations in the training parameters.

[0083] net.trainParam.goal=1e-6; % Error threshold setting: Sets the error threshold number in the training parameters.

[0084] `net.trainParam.lr=0.01;` %Learning rate. Sets the learning rate in the training parameters.

[0085] %%Training Network

[0086] net = train(net, p_train, t_train); Train the neural network model

[0087] %%Simulation Test

[0088] t_sim1=sim(net,p_train); This performs predictions on the training set data.

[0089] t_sim2=sim(net,p_test); This performs predictions on the test set data.

[0090] %% Denormalize the predicted data

[0091] `T_sim1=mapminmax('reverse',t_sim1,ps_output);` performs inverse normalization on the training set data for prediction.

[0092] T_sim2=mapminmax('reverse',t_sim2,ps_output); Performs inverse normalization on the predicted test set data.

[0093] Based on the predicted and actual values, the system outputs relevant indicators such as RMSE (root mean square error), R2, MAE (mean absolute error), and MBE (mean deviation error) to show the deviation between the predicted and actual values ​​and to evaluate the quality of the model.

[0094] %% Root Mean Square Error (RMSE)

[0095] error1=sqrt(sum((T_sim1-T_train).^2). / M);

[0096] error2=sqrt(sum((T_sim2-T_test).^2). / N);

[0097] %R2: Depends on the degree of fit

[0098] R1=1-norm(T_train-T_sim1)^2 / norm(T_train-mean(T_train))^2;

[0099] R2=1-norm(T_test-T_sim2)^2 / norm(T_test-mean(T_test))^2;

[0100] %MAE Mean Absolute Error

[0101] mae1=sum(abs(T_sim1-T_train)). / M;

[0102] mae2=sum(abs(T_sim2-T_test)). / N;

[0103] %MBE mean deviation error

[0104] mbe1 = sum(T_sim1 - T_train). / M;

[0105] mbe2 = sum(T_sim2 - T_test). / N;

[0106] If the average absolute error of the prediction model is ≤10%, that is, the accuracy of the prediction model reaches more than 90%, it is considered to meet the accuracy requirements in engineering.

[0107] %% Save the required variables (trained network model)

[0108] savenet.matnet

[0109] saveps_input.matps_input

[0110] saveps_output.matps_output

[0111] (ii) The component often involved in linear strength and fatigue analysis is the rear floor die casting, while the component often involved in nonlinear abuse analysis is the front tower. Therefore, process data, including temperature, distance, speed, etc., are extracted from the corresponding positions of the new structure rear floor die casting and the front tower.

[0112] The process of predicting the material properties of the structure by calling a pre-trained model and using the process data of the newly acquired structure as input is as follows:

[0113] %%Read and save file

[0114] loadnet.mat

[0115] loadps_input.mat

[0116] loadps_output.mat

[0117] %%Read the data to be predicted

[0118] kes=readmatrix('data to be predicted.xlsx');

[0119] %%Data Transpose

[0120] kes=kes';

[0121] %%Data Normalization

[0122] n_test=mapminmax('apply',kes,ps_input);

[0123] %%Simulation Test

[0124] t_sim3=sim(net,n_test);

[0125] %%Data Inverse Normalization

[0126] T_sim3=mapminmax('reverse',t_sim3,ps_output);

[0127] %%Save Results

[0128] writematrix('prediction results.xlsx', T_sim3')

[0129] refer to Figures 8-10 The illustration shows the predicted partitioning results.

[0130] (III) Linear strength analysis was performed on the model of the rear floor yield strength partitioning using the NASTRAN solver. The analysis results are as follows: Figure 11 Furthermore, the yield strength of the corresponding region was used for evaluation in the post-processing, and the analysis results are listed below:

[0131]

[0132] As can be seen, after performance zoning and yield evaluation of the parts with poor performance at the far end according to the strength analysis results, the safety factor is reduced by 9% to 18%.

[0133] (iv) Linear unit load strength analysis was performed on the model of the floor tensile strength zones. Fatigue analysis was conducted using Ncode software, and the corresponding tensile strength of each region was fitted with the SN curve. The fatigue analysis results are as follows: Figure 12 and Figure 13 For illustration; and the analysis results are as follows: After the floor is divided into zones according to tensile strength, the fatigue damage change rate is 100%~265%, and the lifespan is reduced by about 1 to 2.5 times.

[0134] (v) Nonlinear strength analysis was performed on the front tower package model divided by fracture elongation. The corresponding stress-strain curves were used for the corresponding regions. The analysis results are as follows: Figure 14 and Figure 15 As shown; and the corresponding evaluation results are as follows: after the nonlinear strength partitioning of the front tower package, the strain increase rate is 1.3%~10.0%, and the residual displacement change rate is 2.0%.

[0135] As can be seen from the above examples, it is preferable to use neural network algorithms during implementation. A material performance prediction model is built based on process data and measured data. Then, the material performance of the new structure is predicted. After various simulation analysis methods are carried out by partitioning according to the predicted material performance parameters, the results are compared with those of simulation analysis without performance partitioning. There are differences in the analysis results. The simulation analysis with the new performance partitioning has higher analysis accuracy and more precise evaluation because of the material performance parameters assigned to the partitions and the material performance of the corresponding regions in the post-processing evaluation.

[0136] In this invention, when directional terms are mentioned, they are relative concepts based on the embodiments. Furthermore, "at least one" refers to one or more, and "more than one" refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent the existence of A alone, A and B simultaneously, or B alone. A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects have an "or" relationship. "At least one of the following" and similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one of a, b, and c can represent: a, b, c, a and b, a and c, b and c, or a and b and c, where a, b, and c can be single or multiple.

[0137] The above description of the structure, features, and effects of the present invention is based on the embodiments shown in the figures. However, the above are only preferred embodiments of the present invention. It should be noted that the technical features involved in the above embodiments and their preferred methods can be reasonably combined and matched by those skilled in the art to form a variety of equivalent solutions without departing from or changing the design concept and technical effects of the present invention. Therefore, the present invention is not limited to the scope of implementation shown in the figures. Any changes made in accordance with the concept of the present invention, or modifications to equivalent embodiments, that do not exceed the spirit covered by the specification and figures, should be within the protection scope of the present invention.

Claims

1. A method for predicting and zoning simulation analysis of die-casting material properties, characterized in that, include: The construction of a material performance prediction model specifically includes: selecting a die-cast sample and identifying multiple calibration points on its surface; performing material performance tests on each calibration point to obtain the measured material performance parameters at that point, wherein the measured material performance parameters include at least yield strength, tensile strength, and elongation at break; extracting multiple process parameters for each calibration point during the die-casting process through process flow analysis, wherein the process parameters include at least temperature, distance from the gate, and flow rate; using the process parameters as input and the measured material performance parameters as output, training a deep learning network model to obtain the trained material performance prediction model. Predicting the material properties of a target die-cast part by partitioning it, wherein the target die-cast part includes at least a large integrated die-cast structural component for automobiles, specifically including: dividing the three-dimensional model of the target die-cast part into multiple mesh partitions; extracting the process parameters corresponding to each mesh partition through process flow analysis; inputting the process parameters of each mesh partition into the trained material property prediction model, and the model output is the predicted material property parameters for that partition, wherein the material property parameters include predicted yield strength, predicted tensile strength, and predicted elongation at break; The performance zoning simulation analysis includes: marking the mesh model of the target die casting with performance zoning based on the predicted yield strength, predicted tensile strength, and predicted elongation at break of each zoning region; performing different types of simulation analysis based on the performance zoning markings; for linear strength analysis, using the predicted yield strength of each zoning region as the stress evaluation threshold for that zoning region; for nonlinear abuse strength analysis, determining the failure criterion for each zoning region based on the predicted elongation at break of break of each zoning region; and for fatigue life analysis, generating a unique SN curve for each zoning region based on the predicted tensile strength of each zoning region as the fatigue analysis input.

2. The method for predicting and zoning simulation analysis of die-casting material properties according to claim 1, characterized in that, In the process of predicting the partitioned material properties of the target die casting, after meshing the 3D model of the target die casting, the following steps are also included: The step of merging adjacent mesh elements with the same or similar predicted material performance parameters into a single performance region simplifies the model size for subsequent simulation analysis.

3. The method for predicting and zoning simulation analysis of die-casting material properties according to claim 1, characterized in that, When performing nonlinear abuse strength analysis, the method also includes: constructing a constitutive curve of elastic-plastic stress and strain unique to each performance zone based on the predicted yield strength and predicted tensile strength of that zone, and using it as the material input model for that zone.

4. The method for predicting and zoning simulation analysis of die-casting material properties according to claim 1, characterized in that, When performing fatigue life analysis, based on the predicted tensile strength of each zone, a dedicated SN curve for that zone is generated by fitting empirical formulas or a material database, in order to replace the traditional single-material SN curve.

5. The method for predicting and zoning simulation analysis of die-casting material properties according to claim 1, characterized in that, The deep learning network model includes: a BP neural network model; The training process of the BP neural network model includes: data normalization processing, network structure initialization, forward propagation to calculate error, and back propagation to adjust weights and thresholds in an iterative manner until the error between the network output and the measured value converges to a preset accuracy.

6. The method for predicting and zoning simulation analysis of die-casting material properties according to claim 1, characterized in that, The process parameters extracted from the process flow analysis also include at least one of the following: cooling rate or pressure inside the mold cavity.

7. The method for predicting and zoning simulation analysis of die-casting material properties according to any one of claims 1 to 6, characterized in that, The linear strength analysis also includes a post-processing step: displaying stress results on cloud maps of different performance zones of the model, and comparing them with the predicted yield strength of each zone to accurately identify high stress risk zones or material performance redundancy zones.

Images