SLM shoe sole mold light weight optimization system based on TPMS dot matrix

Abstract

The application discloses a TPMS dot array-based SLM shoe sole mold lightweight optimization system, which comprises a parameterized modeling module, a thermal-structure coupling simulation module, a sample construction module, an agent model training module and an optimization solving module, wherein the parameterized modeling module is used for generating a parameterized geometric model containing a TPMS dot array structure for a shoe sole mold entity model and outputting a dot array structure parameter set; the thermal-structure coupling simulation module is used for solving a thermal field based on the parameterized geometric model and obtaining a structure response index, and the application relates to the technical field of additive manufacturing and mold lightweight. The TPMS dot array-based SLM shoe sole mold lightweight optimization system defines three types of structure parameters, realizes parameterization and grid export, completes thermal-structure sequential coupling simulation in ANSYS, and successfully extracts the maximum stress, the minimum stress and the thermal stress difference through operations such as material attribute setting, thermal boundary loading, temperature field solving and thermal stress analysis.

Description

Technical Field

[0001] This invention relates to the fields of additive manufacturing and mold lightweighting technology, specifically to a SLM shoe sole mold lightweighting optimization system based on TPMS dot matrix. Background Technology

[0002] TPMS lattice is a structure defined by mathematical functions and characterized by periodic repetition in three-dimensional space. Essentially, it involves thickening and arraying smooth surfaces (such as Gyroid, Primitive, and Diamond) that satisfy the conditions of "minimum area and zero average curvature" to form a porous lattice with continuous curved surfaces. The geometric characteristics of TPMS structures can be diversified by controlling parameters such as unit cell size, thickness, and shell thickness, laying the structural foundation for lightweighting and thermal performance optimization. However, the relationship between these structural parameters and the thermal stress response of the mold is extremely complex, making it difficult to directly determine its optimal configuration using traditional methods. Selective Laser Melting (SLM) is another solution. Melting (SLM), as an advanced process system in the field of precision metal additive manufacturing, has shown significant technological advantages in the forming of complex metal components in recent years. It achieves the layer-by-layer accumulation forming of three-dimensional components through digital path planning. To realize the lightweight and thermal stress optimization design of shoe sole mold structures, a corresponding optimization system is required. The existing SLM shoe sole mold lightweight optimization system based on TPMS lattice has a large computational load and long time consumption when performing thermal analysis through numerical simulation. As the complexity of the model and mesh increases, the requirements for computing servers are high, making it difficult to use artificial intelligence algorithms to optimize and predict the results based on the existing numerical simulation results to address the above problems. Summary of the Invention

[0003] To address the shortcomings of existing technologies, this invention provides a lightweight optimization system for SLM shoe sole molds based on TPMS lattice, which solves the problems of large computational load and long processing time in the thermal analysis process.

[0004] To achieve the above objectives, this invention provides the following technical solution: a lightweight optimization system for SLM shoe sole molds based on TPMS dot matrix, comprising: a parametric modeling module, a thermal-structural coupling simulation module, a sample construction module, a surrogate model training module, and an optimization solution module, wherein:

[0005] The parametric modeling module is used to generate a parametric geometric model containing a TPMS lattice structure from the solid model of the shoe sole mold, and outputs a set of lattice structure parameters;

[0006] The thermal-structure coupled simulation module is used to solve the thermal field based on the parametric geometric model and obtain the structural response index;

[0007] The sample construction module is used to construct a training sample set consisting of a set of lattice structure parameters and structural response indices;

[0008] The surrogate model training module is used to train a regression surrogate model based on the training sample set in order to build a mapping relationship between the geometric parameters of the lattice structure and the structural response index.

[0009] The optimization solution module is used to optimize the objective function based on a regression surrogate model under manufacturing feasibility constraints, and outputs the optimal set of parameters for generating the SLM manufacturable shoe sole mold lattice structure.

[0010] Preferably, the parametric modeling module includes a model import module, an implicit volume conversion module, a surface generation module, a structural parameter setting module, a mesh processing module, and a model export module, wherein:

[0011] The model import module is used to import the solid model of the shoe sole mold as the basis for constructing the geometric boundary of the lattice structure;

[0012] The implicit volume conversion module is used to convert the solid model of the shoe sole mold into an implicit volume geometry, and uses the implicit volume geometry as support for subsequent lattice construction and parametric logic operations.

[0013] The surface generation module is used to generate TPMS lattice structures and fill them into implicit volume geometry.

[0014] The structural parameter setting module is used to set structural parameters in implicit geometry.

[0015] The mesh processing module is used to perform mesh processing on parametric geometric models containing TPMS lattice structures;

[0016] The model export module is used to export a parametric geometric model containing a TPMS lattice structure and meshed, and then import the parametric geometric model into the thermal-structural coupling simulation module.

[0017] Preferably, the thermal-structural coupling simulation module includes a steady-state thermal module, a static structural module, an engineering data module, a boundary condition setting module, and a fixed constraint module, wherein:

[0018] The steady-state thermal module is used for thermal analysis based on a parametric geometric model and is coupled with a static structural module;

[0019] The static structure module is used for coupled solution based on the thermal analysis results of the parametric geometric model, realizing the transfer from the temperature field to the structural field;

[0020] The engineering data module is used to set material properties based on a parametric geometric model;

[0021] The boundary condition setting module is used to set the ambient temperature, apply a constant temperature load to the upper surface of the mold, set the convective heat transfer on the surrounding surfaces of the mold, and set the convective heat transfer on the ground of the mold based on the parametric geometric model.

[0022] The fixed constraint module is used to simulate clamping states based on parametric geometric model structures.

[0023] Preferably, the sample construction module includes a parameter extraction module and a response index extraction module, wherein:

[0024] The parameter extraction module is used to extract design variables based on the parametric geometric model. The extracted design variables include: unit cell size in three dimensions, TPMS lattice structure wall thickness, and mold shell solid thickness.

[0025] The response index extraction module extracts performance response tags based on the results of solving the thermal field of the parametric geometric model. The extracted performance response tags include: maximum thermal stress. Minimum thermal stress thermal stress difference ,in: .

[0026] Preferably, the network structure built by the proxy model training module is as follows:

[0027] Input layer: 5 neurons;

[0028] Hidden layers: two layers, with a structure of [5,5], meaning 5 neurons per layer;

[0029] Output layer: 1 neuron, used for regression output;

[0030] The neuron output is represented in the following form:

[0031]

[0032] in, For input features, For the corresponding weights, For bias terms, For activation functions;

[0033] The optimization solution module is used to optimize the objective function based on a regression surrogate model under manufacturing feasibility constraints, and outputs the optimal set of parameters for generating the SLM manufacturable shoe sole mold lattice structure.

[0034] Preferably, during the network structure construction process of the proxy model training module, the input and output data are normalized to map all values ​​to the interval [-1, 1]. After training, the output values ​​are denormalized to restore the original physical quantities. Several rounds of complete training are performed, cross-validation is performed, and the mean square error, mean absolute error, root mean square error, and coefficient of determination are calculated respectively for subsequent prediction and optimization.

[0035] Preferably, the mean square error (MSE) is:

[0036] In the formula To output the predicted value, To output the true value;

[0037] The mean absolute error (MAE):

[0038] In the formula To output the predicted value, To output the true value;

[0039] The root mean square error (RMSE):

[0040] In the formula To output the predicted value, To output the true value;

[0041] Coefficient of determination ( ):

[0042] In the formula, The value is usually between [0,1]. =1 indicates a perfect prediction. =0 indicates that the model has no explanatory power. To output the predicted value, To output the true value, To output the mean;

[0043] Based on the assessment of neural structures: 0.7724 0.4526 0.8499 0.7191.

[0044] Preferably, the genetic algorithm optimization process includes three stages: input parameter standardization, performance index prediction through a neural network surrogate model, and output result destandardization.

[0045] Preferably, the optimization solution module performs optimization through the GA function, and the fitness function is the thermal stress difference predicted by the MLP model. During the optimization process of the GA function, the genetic algorithm generates a new combination of structural parameters in each generation and quickly calculates the performance index through the surrogate model. After optimization, the algorithm outputs the optimal parameter combination and the corresponding minimum thermal stress difference prediction result.

[0046] This invention also discloses a lightweight optimization system for SLM shoe sole molds based on TPMS dot matrix, specifically including the following steps:

[0047] S1. Obtain the solid model of the shoe sole mold and generate a parametric geometric model containing the TPMS lattice structure to obtain the lattice parametric geometry;

[0048] S2. The temperature field is obtained by solving the thermal field based on the parametric geometric model;

[0049] S3. The structural response index has been obtained by inputting the temperature field as a thermal load into the structural field solution.

[0050] S4. Construct a training sample set based on the set of lattice structural parameters and structural response indices, and train a regression surrogate model.

[0051] S5. Under the constraint of manufacturing feasibility, the objective function is optimized using a regression surrogate model, and the optimal parameter set is output.

[0052] S6. Generate a lattice structure model of shoe sole mold for SLM manufacturing based on the optimal parameter set.

[0053] Beneficial effects

[0054] This invention provides a lightweight optimization system for SLM shoe sole molds based on TPMS dot matrix. Compared with existing technologies, it has the following advantages:

[0055] (1) The SLM shoe sole mold lightweight optimization system based on TPMS dot matrix is ​​based on the shoe sole mold solid model. The system constructs a parametric geometric model containing TPMS dot matrix structure and imports the parametric geometric model containing TPMS dot matrix structure into the thermal-structure coupling simulation module for thermal analysis, coupling solution, material properties, boundary condition setting, constraint setting, and extraction of maximum stress, minimum stress and thermal stress difference, unit cell size in three dimensions, TPMS dot matrix structure wall thickness and mold shell solid thickness to construct a structure-performance dataset, which provides reliable data support and input basis for subsequent establishment of neural network prediction model and implementation of parameter optimization.

[0056] (2) The SLM shoe sole mold lightweight optimization system based on TPMS dot matrix clearly uses five structural parameters as input variables: x-direction unit cell size, y-direction unit cell size, z-direction unit cell size, dot matrix thickness and shell thickness. The thermal stress difference is given as the prediction target and output. An MLP neural network model with two hidden layers is built. The accuracy of prediction is improved by normalization preprocessing and multi-round training and screening, which provides accurate prediction support for subsequent optimization design.

[0057] (3) The SLM shoe sole mold lightweight optimization system based on TPMS dot matrix has designed five structural parameters as optimization variables, namely, three unit cell sizes (x, y, z), dot matrix thickness, and shell thickness, by explicitly setting the minimization of thermal stress difference as the optimization target in the system. After optimization by genetic algorithm, the optimal dot matrix structural parameters are obtained. Attached Figure Description

[0058] Figure 1 This is a block diagram illustrating the main principle of the present invention;

[0059] Figure 2 This is a schematic diagram of the parametric modeling module of the present invention.

[0060] Figure 3 This is a schematic diagram of the thermal-structural coupling simulation module of the present invention;

[0061] Figure 4 This is a schematic diagram of the sample construction module of the present invention;

[0062] Figure 5 This is a flowchart illustrating the operation of the proxy model training module of the present invention;

[0063] Figure 6 This is a flowchart illustrating the operation of the optimized solution module in this invention. Detailed Implementation

[0064] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0065] refer to Figures 1-6 The present invention provides the following three technical solutions:

[0066] The first implementation method is a lightweight optimization system for SLM shoe sole molds based on TPMS lattice, comprising: a parametric modeling module, a thermal-structural coupling simulation module, a sample construction module, a surrogate model training module, and an optimization solution module. The parametric modeling module generates a parametric geometric model containing the TPMS lattice structure from the shoe sole mold solid model and outputs a set of lattice structure parameters. The thermal-structural coupling simulation module solves the thermal field based on the parametric geometric model and obtains structural response indices.

[0067] The sample construction module is used to construct a training sample set consisting of a set of lattice structure parameters and structural response indices; the surrogate model training module is used to train a regression surrogate model based on the training sample set to construct the mapping relationship between the geometry of the lattice structure parameters and the structural response indices; the optimization solution module is used to optimize the objective function based on the regression surrogate model under manufacturing feasibility constraints, and output the optimal parameter set for generating the lattice structure of the SLM manufacturable shoe sole mold.

[0068] The parametric modeling module includes a model import module, an implicit volume conversion module, a surface generation module, a structural parameter setting module, a mesh processing module, and a model export module. Specifically: the model import module imports the solid model of the shoe sole mold into nTopology and constructs a parametric model of the TPMS lattice structure mold, serving as the geometric boundary basis for the lattice structure construction; the implicit volume conversion module converts the solid model of the shoe sole mold into an implicit volume geometry, which then supports subsequent lattice construction and parametric logic operations; the surface generation module generates the TPMS lattice structure and fills it into the implicit volume geometry; the structural parameter setting module sets the structural parameters in the implicit geometry, supporting user input; the mesh processing module performs mesh processing on the parametric geometric model containing the TPMS lattice structure, specifically through the Remesh Surface and Robust Tetrahedral Mesh modules; and the model export module exports the meshed parametric geometric model containing the TPMS lattice structure, exports the parametric geometric model as a cdb file, and imports the cdb file model data into the thermal-structural coupling simulation module.

[0069] The thermal-structural coupling simulation module includes a steady-state thermal module, a static structure module, an engineering data module, a boundary condition setting module, and a fixed constraint module. Specifically: the steady-state thermal module is used for thermal analysis based on the parametric geometric model and is coupled with the static structure module; the static structure module is used for coupled solution based on the thermal analysis results of the parametric geometric model, realizing the transfer of the temperature field to the structural field; the engineering data module is used for setting material properties based on the parametric geometric model, specifically adding the preset material property "304 stainless steel"; the boundary condition setting module is used to set the ambient temperature, apply a constant temperature load to the upper surface of the mold, and set convective heat transfer on the surrounding surfaces of the mold and the convective heat transfer on the ground of the mold based on the parametric geometric model, specifically setting the ambient temperature to 25 ℃; applying a constant temperature load of 200 ℃ to the upper surface of the mold; and setting convective heat transfer on the surrounding surfaces of the mold (convection coefficient...). Convection heat transfer is set on the bottom surface of the mold (convection coefficient) The fixed constraint module is used to simulate clamping states based on parametric geometric model structures. No additional mechanical loads are applied to the model; only thermal loads are considered. The steady-state thermal module, static structure module, engineering data module, boundary condition setting module, and fixed constraint module are all built in ANSYS Workbench.

[0070] The sample construction module includes a parameter extraction module and a response index extraction module. The parameter extraction module extracts design variables based on a parametric geometric model, including: unit cell dimensions in three dimensions, TPMS lattice structure wall thickness, and mold shell solid thickness. The response index extraction module extracts performance response labels based on the results of solving the thermal field of the parametric geometric model, including: maximum thermal stress. Minimum thermal stress thermal stress difference ,in: The sample set data extracted by the sample construction module is stored in .xlsx format, with a regular structure and consistent variable names;

[0071] The main difference between the second implementation method and the first implementation method is that:

[0072] The network structure built by the surrogate model training module is as follows:

[0073] Input layer: 5 neurons;

[0074] Hidden layers: two layers, with a structure of [5,5], meaning 5 neurons per layer;

[0075] Output layer: 1 neuron, used for regression output;

[0076] The neuron output is represented in the following form:

[0077]

[0078] in, For input features, For the corresponding weights, For bias terms, For activation functions;

[0079] The optimization solution module is used to optimize the objective function based on a regression surrogate model under manufacturing feasibility constraints, and outputs the optimal set of parameters for generating the SLM manufacturable shoe sole mold lattice structure;

[0080] In the process of building the network structure in the surrogate model training module, mapminmax is used to normalize the input and output data, mapping all values ​​to the interval [-1, 1]. The mathematical expression is:

[0081]

[0082] In the formula: The original data is the raw, unprocessed data (unit cell size in three dimensions, TPMS lattice structure wall thickness, mold shell solid thickness, and thermal stress difference output index). For the normalized data, The maximum value in the original dataset. The minimum value in the original dataset is used to avoid gradient inconsistency caused by different feature parameter units. The normalized data improves training speed and convergence stability.

[0083] After training, the output values ​​are denormalized using the mapminmax('reverse',...) function, with the mathematical expression as follows:

[0084]

[0085] In the formula: The normalized prediction data output by the neural network. The original physical quantity recovered after denormalization. The maximum value in the original dataset. It is the minimum value in the original dataset;

[0086] The original physical quantities were restored for evaluation. A total of 50 rounds of training were performed. After each round of training, the mean squared error, mean absolute error, root mean square error, and coefficient of determination on the validation set were recorded. The model with high accuracy in common regression indicators was selected as the optimal model, and it was saved together with the normalization parameters in the best_mlp_model.mat file for subsequent prediction and optimization. All normalized samples were input into the optimal neural network to obtain the predicted values. Then, the predicted results of thermal stress difference were obtained by inverse normalization. To verify the prediction accuracy of the model, the predicted values ​​were compared with the simulated real values, and a regression scatter plot was plotted. Cross-validation was performed, and the mean squared error, mean absolute error, root mean square error, and coefficient of determination were calculated for subsequent prediction and optimization.

[0087] Mean Square Error (MSE):

[0088] In the formula To output the predicted value, To output the true value, the smaller the MSE in the formula, the more accurate the model prediction is, and the more sensitive it is to large errors, making it suitable as a loss function;

[0089] Mean Absolute Error (MAE):

[0090] In the formula To output the predicted value, To output true values, MAE is less sensitive to outliers than MSE, providing a more intuitive sense of "average bias". The smaller the value, the smaller the overall error of the model.

[0091] Root Mean Square Error (RMSE):

[0092] In the formula To output the predicted value, To output the true value;

[0093] Coefficient of determination ( ):

[0094] In the formula, The value is usually between [0,1]. =1 indicates a perfect prediction. =0 indicates that the model has no explanatory power. To output the predicted value, To output the true value, To output the mean;

[0095] Based on the assessment of neural structures: 0.7724 0.4526 0.8499 0.7191; This indicates that the model has high accuracy in predicting the thermal stress difference, especially... A score of 0.7191 indicates that the model can explain approximately 72% of the output variation. and All values ​​are below 1 MPa. Although the model shows strong fitting ability overall, it should be noted that its performance is limited by the size and quality of the current dataset. Due to limitations in computing resources and simulation time, the number of training data samples is relatively limited, and the distribution of samples in the parameter space is not uniform enough. This may cause the model's prediction accuracy to decrease in some areas. In addition, due to the limitations of the geometric complexity of the mold and the diversity of lattice parameters, the neural network cannot completely capture all nonlinear features, causing individual samples to deviate from the prediction trend. Therefore, the current model has not reached the optimal level in terms of evaluation metrics, but the overall prediction accuracy is still within an acceptable range and has certain engineering application value. The genetic algorithm optimization process includes three stages: input parameter standardization, performance index prediction through neural network surrogate model, and output result destandardization and restoration.

[0096] The main difference between the third and second implementation methods is that:

[0097] The optimization module performs optimization through the GA function, with the following parameters: population size: 50, maximum number of iterations: 100, and fitness function: the thermal stress difference predicted by the MLP model. During the optimization process, the genetic algorithm generates new combinations of structural parameters in each generation (all satisfying integer constraints and boundary requirements), calculates the thermal stress difference of each parameter combination using the aforementioned fitness function, and selects combinations with better fitness for the next generation iteration based on the core operations of selection, crossover, and mutation of the genetic algorithm. After optimization, the algorithm outputs the optimal parameter combination and the corresponding prediction result of the minimum thermal stress difference. The optimal parameter combination is: unit cell size in the x, y, and z directions are 10 mm, 18 mm, and 28 mm, respectively; lattice thickness and shell thickness are 4 mm and 5 mm, respectively. At this point, the thermal stress difference is the minimum and is 244.235 MPa.

[0098] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A lightweight optimization system for SLM shoe sole molds based on TPMS dot matrix, characterized in that, include: The system includes a parametric modeling module, a thermal-structural coupling simulation module, a sample construction module, a surrogate model training module, and an optimization solution module, among which: The parametric modeling module is used to generate a parametric geometric model containing a TPMS lattice structure from the solid model of the shoe sole mold, and outputs a set of lattice structure parameters; The thermal-structure coupled simulation module is used to solve the thermal field based on the parametric geometric model and obtain the structural response index; The sample construction module is used to construct a training sample set consisting of a set of lattice structure parameters and structural response indices; The surrogate model training module is used to train a regression surrogate model based on the training sample set in order to build a mapping relationship between the geometric parameters of the lattice structure and the structural response index. The optimization solution module is used to optimize the objective function based on a regression surrogate model under manufacturing feasibility constraints, and outputs the optimal set of parameters for generating the SLM manufacturable shoe sole mold lattice structure.

2. The lightweight optimization system for SLM shoe sole mold based on TPMS dot matrix as described in claim 1, characterized in that: The parametric modeling module includes a model import module, an implicit volume transformation module, a surface generation module, a structural parameter setting module, a mesh processing module, and a model export module, wherein: The model import module is used to import the solid model of the shoe sole mold as the basis for constructing the geometric boundary of the lattice structure; The implicit volume conversion module is used to convert the solid model of the shoe sole mold into an implicit volume geometry, and uses the implicit volume geometry as support for subsequent lattice construction and parametric logic operations. The surface generation module is used to generate TPMS lattice structures and fill them into implicit volume geometry. The structural parameter setting module is used to set structural parameters in implicit geometry. The mesh processing module is used to perform mesh processing on parametric geometric models containing TPMS lattice structures; The model export module is used to export a parametric geometric model containing a TPMS lattice structure and meshed, and then import the parametric geometric model into the thermal-structural coupling simulation module.

3. The lightweight optimization system for SLM shoe sole mold based on TPMS dot matrix as described in claim 1, characterized in that: The thermal-structural coupling simulation module includes a steady-state thermal module, a static structural module, an engineering data module, a boundary condition setting module, and a fixed constraint module, wherein: The steady-state thermal module is used for thermal analysis based on a parametric geometric model and is coupled with a static structural module; The static structure module is used for coupled solution based on the thermal analysis results of the parametric geometric model, realizing the transfer from the temperature field to the structural field; The engineering data module is used to set material properties based on a parametric geometric model; The boundary condition setting module is used to set the ambient temperature, apply a constant temperature load to the upper surface of the mold, perform convective heat transfer on the surrounding surfaces of the mold, and perform convective heat transfer on the ground of the mold based on the parametric geometric model. The fixed constraint module is used to simulate clamping states based on parametric geometric model structures.

4. The lightweight optimization system for SLM shoe sole mold based on TPMS dot matrix as described in claim 1, characterized in that: The sample construction module includes a parameter extraction module and a response index extraction module, wherein: The parameter extraction module is used to extract design variables based on the parametric geometric model. The extracted design variables include: unit cell size in three dimensions, TPMS lattice structure wall thickness, and mold shell solid thickness. The response index extraction module extracts performance response tags based on the results of solving the thermal field of the parametric geometric model. The extracted performance response tags include: maximum thermal stress. Minimum thermal stress thermal stress difference ,in: .

5. The lightweight optimization system for SLM shoe sole mold based on TPMS dot matrix according to claim 1, characterized in that: The network structure built by the proxy model training module is as follows: Input layer: 5 neurons; Hidden layers: two layers, with a structure of [5,5], meaning 5 neurons per layer; Output layer: 1 neuron, used for regression output; The neuron output is represented in the following form: in, For input features, For the corresponding weights, For bias terms, For activation functions; The optimization solution module is used to optimize the objective function based on a regression surrogate model under manufacturing feasibility constraints, and outputs the optimal set of parameters for generating the SLM manufacturable shoe sole mold lattice structure.

6. The lightweight optimization system for SLM shoe sole mold based on TPMS dot matrix according to claim 5, characterized in that: During the network structure construction process of the proxy model training module, the input and output data are normalized to map all values ​​to the interval [-1, 1]. After training, the output values ​​are denormalized to restore the original physical quantities. Several rounds of complete training are performed, cross-validation is performed, and the mean square error, mean absolute error, root mean square error, and coefficient of determination are calculated respectively for subsequent prediction and optimization.

7. The lightweight optimization system for SLM shoe sole mold based on TPMS dot matrix according to claim 6, characterized in that: The mean square error (MSE): In the formula To output the predicted value, To output the true value; The mean absolute error (MAE): In the formula To output the predicted value, To output the true value; The root mean square error (RMSE): In the formula To output the predicted value, To output the true value; Coefficient of determination ( ): In the formula, The value is usually between [0,1]. =1 indicates a perfect prediction. =0 indicates that the model has no explanatory power. To output the predicted value, To output the true value, To output the mean; Based on the assessment of neural structures: 0.7724 0.4526 0.8499 0.7191.

8. The lightweight optimization system for SLM shoe sole mold based on TPMS dot matrix according to claim 1, characterized in that: The genetic algorithm optimization process includes three stages: input parameter standardization, performance index prediction through a neural network surrogate model, and output result destandardization.

9. The lightweight optimization system for SLM shoe sole mold based on TPMS dot matrix according to claim 1, characterized in that: The optimization solution module performs optimization through the GA function, and the fitness function is the thermal stress difference predicted by the MLP model. During the optimization process of the GA function, the genetic algorithm generates a new combination of structural parameters in each generation and quickly calculates the performance index through the surrogate model. After optimization, the algorithm outputs the optimal parameter combination and the corresponding minimum thermal stress difference prediction result.

10. A lightweight optimization method for SLM shoe sole molds based on TPMS dot matrix, characterized in that, The lightweight optimization system for SLM shoe sole molds based on TPMS dot matrix, as described in any one of claims 1-9, specifically includes the following steps: S1. Obtain the solid model of the shoe sole mold and generate a parametric geometric model containing the TPMS lattice structure to obtain the lattice parametric geometry; S2. The temperature field is obtained by solving the thermal field based on the parametric geometric model; S3. The structural response index has been obtained by inputting the temperature field as a thermal load into the structural field solution. S4. Construct a training sample set based on the set of lattice structural parameters and structural response indices, and train a regression surrogate model. S5. Under the constraint of manufacturing feasibility, the objective function is optimized using a regression surrogate model, and the optimal parameter set is output. S6. Generate a lattice structure model of shoe sole mold for SLM manufacturing based on the optimal parameter set.

Images