Machine learning and regulation method for topology optimization lattice structure of anisotropic features for additive manufacturing

Abstract

The application discloses a kind of machine learning and regulation methods for topology optimization lattice structure of anisotropic feature of additive manufacturing, comprising: step one, construct data set and divide into training set and test set;Step two, construct lattice stretching structure data-driven model: construct lattice stretching structure data-driven model based on Gaussian process regression algorithm, and the data set constructed in step one is used to train lattice stretching structure data-driven model;Step three, obtain the anisotropic data of target macrostructure: based on the trained lattice stretching structure data-driven model, the anisotropic data of each lattice structure included in target macrostructure is calculated;Step four, optimize the additive manufacturing process forming parameters and relative density of target lattice structure.It can be seen from this that the application can be aimed at anisotropic feature of additive manufacturing, and machine learning method is combined under a small amount of experiment to provide support for topology optimization design and forming parameter determination of lattice structure under additive manufacturing forming constraint.

Description

TECHNICAL FIELD

[0001] The application relates to a machine learning and regulation method for topology optimization lattice structure of anisotropic characteristics of additive manufacturing, and belongs to the technical field of additive manufacturing. BACKGROUND

[0002] The laser additive manufacturing technology has developed rapidly in the fields of material processing and material repair due to its ability to meet the needs of shaping and high performance. It uses a high-energy laser beam as a heat source to achieve the shaping of three-dimensional solid parts through a layer-by-layer manufacturing process. However, the anisotropy of materials during the additive manufacturing process has always been a key factor restricting the performance improvement. Anisotropy describes the phenomenon that materials exhibit different physical and chemical properties in different directions. This property can cause significant differences in the performance of materials in different directions, especially during the shaping process. Specifically, materials may exhibit excellent performance in the shaping direction, such as higher strength and stiffness, while the performance in the direction perpendicular to the shaping direction is relatively weak. This imbalance in performance can cause problems in actual applications, especially when subjected to complex loads, the material may not be able to uniformly distribute stress, resulting in unexpected deformation or even structural damage. There is currently a lack of a topology optimization strategy that fully considers the anisotropy of materials during the additive manufacturing process.

[0003] In addition, the existing regression problem algorithm for predicting the anisotropy of additive manufacturing lattice structure still has many defects. For example, using a random forest regression algorithm to predict anisotropy, since the random forest algorithm has multiple hyperparameters, such as the number of trees and the number of features considered when splitting each node, it needs to be adjusted through cross-validation and other methods, which may require a large number of experiments and computational resources. Using a multi-scale topology optimization machine learning algorithm requires a deep understanding of physical phenomena at different scales, and the machine learning model may not fully capture these multi-scale effects, thereby reducing the accuracy of the algorithm. SUMMARY

[0004] The application aims to provide a machine learning and regulation method for topology optimization lattice structure of anisotropic characteristics of additive manufacturing to realize the topology optimization design of anisotropy in the additive manufacturing process.

[0005] To achieve the above technical purpose, the application will adopt the following technical solution:

[0006] A machine learning and regulation method for topology optimization lattice structure of anisotropic characteristics of additive manufacturing, comprising the following steps:

[0007] Step one, construct a data set and divide it into a training set and a test set:

[0008] The data set includes several elements, each of which includes the corresponding additive manufacturing process molding parameters, the relative density of the lattice structure, and the anisotropy data of the lattice structure;

[0009] Step 2: Construct a data-driven model for lattice tensile structure:

[0010] A lattice stretching structure data-driven model is constructed based on the Gaussian process regression algorithm, and the lattice stretching structure data-driven model is trained using the data set constructed in step 1;

[0011] The input of the lattice tensile structure data-driven model is the additive manufacturing process molding parameters in the training set constructed in step 1 and the relative density of the corresponding lattice structure, and the output is the anisotropic data prediction value of the corresponding lattice structure;

[0012] During the training process, a K-fold cross-validation algorithm was used to obtain the mean square error of the predicted anisotropic data of all lattice structures in the training set relative to the anisotropic data of the lattice structure in the dataset. Minimizing the mean square error was then used as the optimization goal. A Bayesian search algorithm was then used to obtain the optimal hyperparameters of the lattice stretching structure data-driven model to complete model training.

[0013] Step 3: Obtain anisotropic data of the target macrostructure:

[0014] Preset the initial values ​​of the additive manufacturing process molding parameters and the initial values ​​of the relative density for the target macrostructure;

[0015] Inputting the preset initial values ​​of the additive manufacturing process molding parameters and the initial values ​​of the relative density into the trained lattice tensile structure data-driven model, calculating the anisotropic data of each lattice structure included in the target macrostructure, and obtaining the anisotropic elastic matrix of each lattice structure included in the target macrostructure;

[0016] Step 4: Optimize the additive manufacturing process parameters and relative density of the target lattice structure:

[0017] The anisotropic data of each lattice structure included in the target macrostructure calculated in step 3 is used as the input of the lattice topology optimization algorithm. The load is applied to obtain the topological optimization configuration of the target macrostructure, and the Bayesian method is used to reversely search the corresponding additive manufacturing process molding parameters and relative density to determine the optimized values ​​of the additive manufacturing process molding parameters and relative density of the target macrostructure.

[0018] Preferably, in step 1, the construction of the data set specifically includes the following steps:

[0019] Step 1.1, obtaining a series of additive manufacturing process molding parameters, relative density of the lattice structure, and anisotropy data of the lattice structure that are mapped to each other;

[0020] Step 1.2, normalizing the additive manufacturing process molding parameters, the relative density of the lattice structure, and the anisotropy data of the lattice structure obtained in step 1;

[0021] Step 1.3: Divide the normalized additive manufacturing process parameters, the relative density of the lattice structure, and the anisotropy data of the lattice structure into two data sets, one of which is the algorithm input data set X = [x1, x2, ..., x i ,…,x n ] m×n , the other is the algorithm output data set Y=[y1,y2,…,y i ,…,y m ] 1×m ; where: x i represents the i-th element of the algorithm input data set X, which is composed of the additive manufacturing process molding parameters and the relative density of the lattice structure; y i Represents the i-th element of the algorithm output data set Y, which is composed of anisotropic data of the lattice structure; i ranges from 1 to m, where m represents the number of samples in the training set.

[0022] Preferably, in step 4, the optimized values ​​of the additive manufacturing process forming parameters of the target macrostructure and the optimized values ​​of the relative density of the lattice structure included in the target macrostructure determined by the lattice topology optimization algorithm are specifically:

[0023]

[0024] Where: n is the number of finite elements in the macrostructure design domain, ρ i represents the relative density of the i-th lattice structure in the macrostructure; U is the node displacement matrix, F is the external load vector, and v i is the volume of the i-th lattice structure, V is the total volume of the macrostructure design domain; is the average volume of the macrostructure design domain, ρ min is the minimum relative density, ρ max Relatively dense

[0025] Maximum value of degree;

[0026] K a represents the overall stiffness matrix of the macrostructure, B is the elastic displacement matrix, Ω i represents the equivalent unit of the i-th lattice structure, is the elastic matrix of the i-th lattice structure;

[0027] The elastic matrix of any lattice structure calculated by the optimized lattice tensile structure data-driven model is expressed as:

[0028]

[0029] wherein E1, E2, E3 are the elastic modulus of the corresponding lattice structure, respectively, and v 12 , v 13 , v 23 , v 31 , v 32 , v 21 are the Poisson's ratios of the corresponding lattice structure, respectively; the subscripts of the elastic modulus, shear modulus and Poisson's ratio include the numbers 1, 2, 3 corresponding to the three normal directions of the corresponding lattice structure.

[0030] Preferably, in step two, a square exponential kernel function and automatic relevance determination are used to improve the generalization ability of the lattice stretch structure data-driven model; the kernel function is expressed as:

[0031]

[0032] wherein y represents the anisotropy data of the lattice structure in the training set, y' represents the anisotropy data prediction value of the corresponding lattice structure output by the lattice stretch structure data-driven model, and σ f is taken as 1, reflects the size of white noise in the training data; Θ is a diagonal matrix, and the diagonal elements thereof are composed of the characteristic length of each structure and process parameter; and δ(y, y') is a Kronecker function.

[0033] Preferably, in step one, the ratio of the data set divided into the training set and the test set is 8:2.

[0034] Another technical purpose of the present application is to provide a computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-mentioned machine learning method for lattice topology optimization of anisotropic features for additive manufacturing.

[0035] Still another technical purpose of the present application is to provide a computer-readable storage medium having a computer program stored thereon, wherein the computer program is executed by a processor to implement the steps of the above-mentioned machine learning method for lattice topology optimization of anisotropic features for additive manufacturing.

[0036] Based on the above technical purposes, compared with the prior art, the present application has the following advantages:

[0037] ①By employing the Bayesian search algorithm, we can efficiently explore the parameter space of the Gaussian Process Regression (SVR) algorithm, identifying the optimal combination of hyperparameters. The Bayesian approach, with its probabilistic reasoning nature, quickly converges to the optimal solution with fewer evaluations when faced with complex search spaces. This process not only improves the efficiency of model training but also significantly enhances the accuracy of prediction results by precisely adjusting key hyperparameters such as kernel functions and regularization parameters;

[0038] ②In traditional machine learning model training, a large amount of data is usually required to ensure the generalization ability of the model. However, through the optimization of the SVR algorithm by the Bayesian search algorithm, even with only a small amount of data, a high prediction accuracy can be achieved. This is because the optimized model can more effectively capture the potential patterns in the data, reducing the dependence on large-scale data sets, and thus making accurate predictions even in the case of data scarcity;

[0039] ③Compared with the Random Forest Regression algorithm, the Gaussian Process Regression algorithm optimized by hyperparameters is more economical in terms of data requirements. The Random Forest algorithm often requires a large training data set for iterative calculation to improve the prediction accuracy of the model. In contrast, the SVR algorithm, with the help of Bayesian search, can achieve comparable or even higher prediction accuracy than the Random Forest algorithm even with less data. BRIEF DESCRIPTION OF DRAWINGS

[0040] Figure 1 is a flowchart of the lattice topology optimization machine learning method for anisotropic features in additive manufacturing according to the present invention;

[0041] Figure 2 is a schematic diagram of a structural member obtained by different forming direction printing experiments using the lattice topology optimization machine learning method according to the present invention;

[0042] Figure 3 is the transformation principle of the anisotropic elastic matrix in the lattice topology optimization machine learning method according to the present invention;

[0043] Figure 4 is the modulus-relative density mapping diagram in the lattice topology optimization machine learning method according to the present invention;

[0044] Figure 5 is the convergence curve of the topology optimization process in the lattice topology optimization machine learning method according to the present invention; DETAILED DESCRIPTION

[0045] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. The description of the at least one example embodiment is actually only illustrative, but not as any limitation on the present application and its application or use. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work are within the scope of protection of the present application. Unless otherwise specified, the relative arrangement, expression and numerical value of the components and steps set forth in these embodiments do not limit the scope of the present application. The technology, method and equipment known to those of ordinary skill in the related art can not be discussed in detail, but should be considered as part of the specification when appropriate. In all examples shown and discussed here, any specific value should be interpreted as merely exemplary, not as a limitation. Therefore, other examples of the example embodiments can have different values.

[0046] For ease of description, spatial relative terms such as "over", "above", "upper surface", "upper", etc. can be used herein to describe the spatial positional relationship of one device or feature with respect to other devices or features as shown in the drawings. It should be understood that the spatial relative terms are intended to include different orientations in use or operation in addition to the orientation of the device described in the drawings. For example, if the device in the drawing is inverted, the device described as "above" or "over" other devices or structures will be positioned "below" or "under" the other devices or structures. Thus, the example term "above" can include both "above" and "below" orientations. The device can also be positioned in other different ways (rotated 90 degrees or in other orientations).

[0047] As shown in Figure 1 The present application provides a machine learning strategy for topology optimization lattice structure of anisotropic features for additive manufacturing, which specifically comprises the following steps:

[0048] Step one, construct a data set and divide it into a training set and a test set:

[0049] Consider the additive manufacturing process parameters (laser power, scanning speed) and dot matrix structure parameters (i.e. change the relative density of the dot matrix structure) as input variables, by collecting the anisotropy data of the additive manufacturing dot matrix structure, including the elastic modulus, shear modulus and Poisson's ratio in three directions, a data set is constructed and divided into training set and test set. The data set is divided into training set and test set according to the proportion of 8:2. In other words, the data set includes a plurality of elements, and each element includes corresponding additive manufacturing process forming parameters, the relative density of the dot matrix structure and the anisotropy data of the dot matrix structure.

[0050] Step two, constructing a dot matrix tensile structure data driven model:

[0051] Based on the Gaussian process regression algorithm, a dot matrix tensile structure data driven model is constructed, and the data set constructed in step one is used to train the dot matrix tensile structure data driven model;

[0052] The input of the dot matrix tensile structure data driven model is the additive manufacturing process forming parameters in the training set constructed in step one and the corresponding relative density of the dot matrix structure, and the output is the predicted value of the anisotropy data of the corresponding dot matrix structure;

[0053] During the training process, the mean square error of the predicted value of the anisotropy data of all dot matrix structures in the training set relative to the anisotropy data of the dot matrix structure in the data set is obtained by using K-fold cross-validation algorithm, and the minimum mean square error is taken as the optimization target, and the optimal hyperparameter of the dot matrix tensile structure data driven model is obtained by using Bayesian search algorithm, and the model training is completed;

[0054] Specifically, a dot matrix tensile structure data driven model is constructed for AlSi10Mg structure, which is based on the Gaussian process regression algorithm, trained by the training set constructed in step one, and the input is the process parameters and the relative density of the dot matrix structure in the training set constructed in step one, and the output is the anisotropy data of the dot matrix structure. Specifically, the following steps are included:

[0055] (1) In order to eliminate the dimensional influence between parameters, make the prediction result more accurate. First, the sample data in the training set is normalized.

[0056] (2) The process parameters (laser power, scanning speed) of the additive manufacturing dot matrix structure and the dot matrix structure parameters (i.e. change the relative density of the dot matrix structure) are taken as the input of the algorithm, and the normalized anisotropy data (including the elastic modulus, shear modulus and Poisson's ratio in three directions) are taken as the output of the algorithm.

[0057] The basic single-task Gaussian process regression defines a joint probability distribution from the original space to the high-dimensional feature space by a transformation φ(·), and uses the prior distribution in the feature space as the Gaussian process. The corresponding covariance can be expressed in the form of kernel function:

[0058]

[0059] (3) Based on experience and understanding of our own data set, we use the square exponential (SE) kernel function and automatic relevance determination (ARD) to improve the generalization ability of the model:

[0060]

[0061] where y represents the anisotropy data of the point array structure in the training set, y' represents the predicted value of the anisotropy data of the corresponding point array structure output by the point array stretching structure data-driven model, σ f is taken as 1, reflects the size of white noise in the training data; Θ is a diagonal matrix, whose diagonal elements consist of the characteristic length of each structure and process parameter; δ(y, y') is the Kronecker function.

[0062] (4) For a set with l groups (where x i ∈R n is an n-dimensional feature vector) and label vector y = [y1, y2,..., y n ] T , the learning sample matrix X = {x1, x2,..., x n} is represented, given a new input X * and its corresponding output y * , y and y * satisfy the joint normal distribution:

[0063]

[0064] where y represents the output mechanical property of the training set, y + represents the predicted mechanical property of the training set, and x represents the input structure and process parameters of the training set.

[0065] The marginal distribution of y * can be obtained from the joint distribution, and the prediction form of Gaussian process regression is obtained:

[0066]

[0067] cov(y * ) = κ(X * , X * ) - Σ(X * , X) K -1 Σ(X, X* )

[0068] Multi-task Gaussian process regression defines a joint probability distribution by considering the correlation between multiple targets. D represents the number of target outputs, represents the input of multi-task, Y = {y1, y2,..., y D} represents the output of multi-task. The multi-task covariance matrix is defined as follows:

[0069]

[0070] Step three, obtain the anisotropy data of the target macrostructure:

[0071] The preset additive manufacturing process forming parameter initial value and the relative density initial value of the target macrostructure are obtained.

[0072] The preset additive manufacturing process forming parameter initial value and the relative density initial value are input into the trained point lattice stretching structure data driven model to calculate the anisotropy data of each point lattice structure included in the target macrostructure, and the anisotropic elastic matrix of each point lattice structure included in the target macrostructure is obtained, which is expressed as:

[0073]

[0074] Wherein, E1, E2, E3 are the elastic modulus of the corresponding point lattice structure, G 23 , G 13 , G 12 are the shear modulus of the corresponding point lattice structure, v 12 , v 13 , v 23 , v 31 , v 32 , v 21 are the Poisson's ratio of the corresponding point lattice structure; The numbers 1, 2 and 3 included in the subscripts of each elastic modulus, shear modulus and Poisson's ratio correspond to the three normal directions of the corresponding point lattice structure.

[0075] Step four, optimize the additive manufacturing process forming parameters and relative density of the target point lattice structure:

[0076] The anisotropy data of each point lattice structure included in the target macrostructure calculated in step three is input into the point lattice topology optimization algorithm, a load is applied to obtain the topology optimization configuration of the target macrostructure, and the corresponding additive manufacturing process forming parameters and relative density are searched in reverse by using the Bayesian method, so as to determine the optimized value of the additive manufacturing process forming parameters and the optimized value of the relative density of the target macrostructure.

[0077] The optimized values ​​of the additive manufacturing process molding parameters of the target macrostructure and the optimized values ​​of the relative density of the lattice structure included in the target macrostructure determined by the lattice topology optimization algorithm are specifically:

[0078]

[0079] Where: n is the number of finite elements in the macrostructure design domain, ρ i represents the relative density of the i-th lattice structure in the macrostructure; U is the node displacement matrix, F is the external load vector, and v i is the volume of the i-th lattice structure, V is the total volume of the macrostructure design domain; is the average volume of the macrostructure design domain, ρ min is the minimum relative density, ρ max Relatively dense

[0080] Maximum value of degree;

[0081] K a represents the overall stiffness matrix of the macrostructure, B is the elastic displacement matrix, Ω i represents the equivalent unit of the i-th lattice structure, D i H (ρ i ) is the elastic matrix of the i-th lattice structure.

[0082] like Figure 3 As shown in the figure, by collecting the elastic modulus, shear modulus and Poisson's ratio in three directions, the anisotropic elastic matrix of the lattice structure is obtained, and then the effective elastic properties are calculated based on these data. The results show that the BCC lattice structure exhibits the highest elastic modulus in the diagonal direction and the lowest elastic modulus in the X, Y and Z axis directions. Figure 4 As shown, the optimized Gaussian process regression algorithm is used as a proxy model for calculating the relative density-anisotropic mechanical properties of the lattice structure. The increase in relative density significantly improves the overall Young's modulus. In summary, the method of the present invention reversely searches for the optimal anisotropic mechanical properties under different relative densities through the Bayesian method, and derives the anisotropic elastic matrix and its corresponding process and structural parameters. The method of the present invention is different from the neural network prediction. With a small number of experiments, it combines the machine learning method to provide support for the topological optimization design and forming parameter determination of the lattice structure considering the constraints of additive manufacturing. At the same time, the method of the present invention only requires a small training set to obtain accurate prediction results.

Claims

1. A machine learning and control method for topological optimization lattice structure based on anisotropic characteristics of additive manufacturing, characterized by: The steps include: Step 1: Build a data set and divide it into training set and test set: The data set includes several elements, each of which includes the corresponding additive manufacturing process parameters, the relative density of the lattice structure, and the anisotropy data of the lattice structure; Step 2: Construct a data-driven model for lattice tensile structure: A lattice stretching structure data-driven model is constructed based on the Gaussian process regression algorithm, and the lattice stretching structure data-driven model is trained using the data set constructed in step 1; The input of the lattice tensile structure data-driven model is the additive manufacturing process molding parameters in the training set constructed in step 1 and the relative density of the corresponding lattice structure, and the output is the anisotropic data prediction value of the corresponding lattice structure; During the training process, a K-fold cross-validation algorithm was used to obtain the mean square error of the predicted anisotropic data of all lattice structures in the training set relative to the anisotropic data of the lattice structure in the dataset. Minimizing the mean square error was then used as the optimization goal. A Bayesian search algorithm was then used to obtain the optimal hyperparameters of the lattice stretching structure data-driven model to complete model training. Step 3: Obtain anisotropic data of the target macrostructure: Preset the initial values ​​of the additive manufacturing process molding parameters and the initial values ​​of the relative density for the target macrostructure; Inputting the preset initial values ​​of the additive manufacturing process molding parameters and the initial values ​​of the relative density into the trained lattice tensile structure data-driven model, calculating the anisotropic data of each lattice structure included in the target macrostructure, and obtaining the anisotropic elastic matrix of each lattice structure included in the target macrostructure; Step 4: Optimize the additive manufacturing process parameters and relative density of the target lattice structure: The anisotropic data of each lattice structure included in the target macrostructure calculated in step 3 is used as the input of the lattice topology optimization algorithm. The load is applied to obtain the topological optimization configuration of the target macrostructure, and the Bayesian method is used to reversely search the corresponding additive manufacturing process molding parameters and relative density to determine the optimized values ​​of the additive manufacturing process molding parameters and relative density of the target macrostructure.

2. The method for topological optimization lattice structure machine learning and control based on anisotropic characteristics of additive manufacturing according to claim 1, characterized in that: In step 1, the construction of the dataset includes the following steps: Step 1.1, obtaining a series of additive manufacturing process molding parameters, relative density of the lattice structure, and anisotropy data of the lattice structure that are mapped to each other; Step 1.2, normalizing the additive manufacturing process molding parameters, the relative density of the lattice structure, and the anisotropy data of the lattice structure obtained in step 1; Step 1.3: Divide the normalized additive manufacturing process parameters, the relative density of the lattice structure, and the anisotropy data of the lattice structure into two data sets, one of which is the algorithm input data set X = [x1, x2, ..., x i ,…,x n ] m×n , the other is the algorithm output data set Y=[y1,y2,…,y i ,…,y m ] 1×m ; Where: x i represents the i-th element of the algorithm input data set X, which is composed of the additive manufacturing process molding parameters and the relative density of the lattice structure; y i Represents the i-th element of the algorithm output data set Y, which is composed of anisotropic data of the lattice structure; i ranges from 1 to m, where m represents the number of samples in the training set.

3. The method for topological optimization lattice structure machine learning and control based on anisotropic characteristics of additive manufacturing according to claim 1, characterized in that: In step 4, the optimized values ​​of the additive manufacturing process molding parameters of the target macrostructure and the optimized values ​​of the relative density of the lattice structure included in the target macrostructure determined by the lattice topology optimization algorithm are specifically: Where: n is the number of finite elements in the macrostructure design domain, ρ i represents the relative density of the i-th lattice structure in the macrostructure; U is the node displacement matrix, F is the external load vector, and v i is the volume of the i-th lattice structure, V is the total volume of the macrostructure design domain, is the average volume of the design domain of the structure, ρ min is the minimum relative density, ρ max is the maximum relative density; K a represents the overall stiffness matrix of the macrostructure, B is the elastic displacement matrix, Ω i represents the equivalent unit of the i-th lattice structure, is the elastic matrix of the i-th lattice structure; The elastic matrix of any lattice structure calculated by the optimized lattice tensile structure data-driven model is expressed as: Among them, E1, E2, and E3 are the elastic moduli of the corresponding lattice structures, G 23 , G 13 , G 12 are the shear modulus of the corresponding lattice structure, ν 12 、ν 13 、ν 23 、v 31 、v 32 、v 21 are the Poisson's ratios of the corresponding lattice structures; the numbers 1, 2, and 3 included in the subscripts of the elastic modulus, shear modulus, and Poisson's ratio correspond to the three normal directions of the corresponding lattice structure.

4. The method for topological optimization lattice structure machine learning and control based on anisotropic characteristics of additive manufacturing according to claim 1, characterized in that: In step 2, the square exponential kernel function and automatic association determination are used to improve the generalization ability of the lattice tensile structure data-driven model; the kernel function is expressed as: Where: y represents the anisotropic data of the lattice structure in the training set, y' represents the anisotropic data prediction value of the corresponding lattice structure output by the lattice tensile structure data driving model, σ f Take 1, Reflects the size of the white noise in the training data; Θ is a diagonal matrix whose diagonal elements consist of the characteristic length of each structural and process parameter; δ(y,y') is the Kronecker function.

5. The method for topological optimization lattice structure machine learning and control based on anisotropic characteristics of additive manufacturing according to claim 1, characterized in that: In step 1, the data set is divided into a training set and a test set with a ratio of 8:

2.

6. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the steps of the machine learning and control method for topological optimization lattice structure with anisotropic features for additive manufacturing as described in any one of claims 1 to 5 are implemented.

7. 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 machine learning and control method for topological optimization lattice structure with anisotropic features for additive manufacturing as described in any one of claims 1 to 5 are implemented.

Images