A method and device for predicting performance of dot matrix structure units based on machine learning

Abstract

The application discloses a kind of lattice structure unit performance prediction method and device based on machine learning, wherein the method comprises: obtaining trained neural network model, the neural network model is input with D2 vector and is output with absorption energy;The surface point cloud analysis of the three-dimensional model of the lattice structure unit to be measured is carried out, the D2 distribution associated with the lattice structure unit to be measured is obtained, and the corresponding D2 vector to be measured is then obtained;The D2 vector to be measured is input into the neural network model for solving, and the absorption energy of the lattice structure unit to be measured is obtained.The application can quickly and efficiently predict the mechanical properties of any lattice structure unit to be measured according to the input geometric shape characteristics of the lattice structure unit to be measured by building a neural network model, which can avoid unnecessary material waste and further reduce the design time cost of the lattice structure.

Description

Technical Field

[0001] This invention relates to the field of computer technology, specifically to a method and apparatus for predicting the performance of lattice structural units based on machine learning. Background Technology

[0002] A lattice structure is a network-like geometric structure formed by connecting a series of lattice structural units. Lattice structures manufactured using additive manufacturing technology are widely used in many fields due to their adjustable mechanical properties. During the design phase of a lattice structure, it is usually necessary to prioritize the design of each individual lattice structural unit, as the performance of each unit affects the overall performance of the lattice structure. Currently, there are two main methods for predicting the performance of lattice structural units: First, multiple lattice structural unit samples with different design parameters are printed using additive manufacturing technology, and then physical experiments are conducted on these samples using a universal testing machine to select the lattice structural unit sample with the best mechanical properties. However, since the printing materials used in additive manufacturing are generally expensive, the entire process inevitably leads to material waste and time costs. Second, finite element analysis is used to simulate and analyze the lattice structural units. However, because the internal structure of a lattice structural unit contains a large number of microstructures, the workload of structural modeling and response analysis is enormous, undoubtedly extending the design cycle of the lattice structural unit. Summary of the Invention

[0003] This invention provides a method and apparatus for predicting the performance of lattice structural units based on machine learning, in order to solve one or more technical problems existing in the prior art, and at least provide a beneficial option or create conditions.

[0004] Firstly, a machine learning-based method for predicting the performance of lattice structural units is provided, the method comprising:

[0005] Obtain a trained neural network model, which takes a D2 vector as input and absorbs energy as output;

[0006] Surface point cloud analysis is performed on the three-dimensional model of the lattice structure unit to be tested to obtain the D2 distribution associated with the lattice structure unit to be tested, and then the corresponding D2 vector to be tested is obtained.

[0007] The D2 vector to be tested is input into the neural network model for solution, and the absorbed energy of the lattice structure unit to be tested is obtained.

[0008] Furthermore, the training process of the neural network model is as follows:

[0009] Obtain several three-dimensional models corresponding to several lattice structural units, wherein the shape types of the several lattice structural units are different;

[0010] Surface point cloud analysis is performed on the aforementioned three-dimensional models to obtain several D2 distributions associated with the aforementioned lattice structural units, and then several corresponding D2 vectors are obtained.

[0011] Perform uniaxial compression simulation on the several three-dimensional models according to the predetermined compression feed rate to obtain several force-displacement curves corresponding to the several lattice structural units, and then obtain several corresponding absorbed energies.

[0012] Based on the aforementioned D2 vectors and the aforementioned absorbed energy, a training dataset is constructed;

[0013] A neural network model consisting of an input layer, a hidden layer, and an output layer is constructed, and the training dataset is input into the neural network model for iterative training.

[0014] Furthermore, the predetermined compression feed rate is 1 mm.

[0015] Furthermore, each of the several absorbed energies is obtained by integrating the corresponding force-displacement curve.

[0016] Furthermore, the process of obtaining the D2 distribution to be tested is as follows:

[0017] Several point cloud data pairs are randomly obtained from the three-dimensional model of the lattice structure unit to be tested, and then the distance between two points is calculated for each point cloud data pair to obtain several corresponding distance values.

[0018] A frequency histogram is constructed based on the aforementioned distance values ​​and the predetermined number of partitions. The frequency histogram is the D2 distribution to be tested.

[0019] Furthermore, the process of obtaining the D2 vector to be measured is as follows:

[0020] The number of distance values ​​contained in each distance interval of the D2 distribution to be tested is divided by the predetermined number of partitions to obtain the D2 vector to be tested.

[0021] Furthermore, the predetermined number of partitions is 1000.

[0022] Secondly, a machine learning-based performance prediction device for lattice structural units is provided, the device comprising:

[0023] The acquisition module is used to acquire a trained neural network model, which takes a D2 vector as input and absorbed energy as output.

[0024] The analysis module is used to perform surface point cloud analysis on the three-dimensional model of the lattice structure unit under test, obtain the D2 distribution associated with the lattice structure unit under test, and then obtain the corresponding D2 vector under test.

[0025] The solution module is used to input the D2 vector to be tested into the neural network model for solution, so as to obtain the absorbed energy of the lattice structure unit to be tested.

[0026] Thirdly, a computer device is provided, including a memory and a processor, the memory storing a computer program, the processor executing the computer program to implement the machine learning-based lattice structure unit performance prediction method as described in the first aspect.

[0027] Fourthly, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the machine learning-based lattice structure unit performance prediction method as described in the first aspect.

[0028] The present invention has at least the following beneficial effects: it uses D2 vectors to represent the geometric shape features of lattice structural units, uses absorbed energy to represent the mechanical properties of lattice structural units, and builds a neural network model and pre-learns the correspondence between the mechanical properties of lattice structural units of different shape types and their geometric shape features. This allows the trained neural network model to quickly and efficiently predict the mechanical properties of any lattice structural unit under test based on its geometric shape features, without the need to print out the lattice structural unit under test through additive manufacturing technology for physical experimental analysis. This avoids unnecessary material waste and further reduces the design time cost of lattice structures. Attached Figure Description

[0029] The accompanying drawings are provided to further understand the technical solutions of the present invention and constitute a part of the specification. They are used together with the embodiments of the present invention to explain the technical solutions of the present invention, and do not constitute a limitation on the technical solutions of the present invention.

[0030] Figure 1 This is a flowchart illustrating a machine learning-based method for predicting the performance of lattice structural units in an embodiment of the present invention.

[0031] Figure 2 This is a schematic diagram of the structure of any lattice structure unit in the embodiment of the present invention before compression;

[0032] Figure 3 This is a schematic diagram of the structure of any lattice structure unit in the embodiment of the present invention after compression;

[0033] Figure 4This is a schematic diagram of the force-displacement curve corresponding to any lattice structural unit in the embodiments of the present invention;

[0034] Figure 5 This is a schematic diagram of the composition of a machine learning-based lattice structure unit performance prediction device in an embodiment of the present invention.

[0035] Figure 6 This is a schematic diagram of the hardware structure of the computer device in an embodiment of this disclosure. Detailed Implementation

[0036] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0037] It should be noted that although functional modules are divided in the system diagram and a logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than the module division in the system or the order in the flowchart. The terms "first," "second," "third," "fourth," etc., used in the specification and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units that are not explicitly listed and are inherent to these processes, methods, products, or apparatuses.

[0038] Please refer to Figure 1 , Figure 1 This is a flowchart illustrating a machine learning-based method for predicting the performance of lattice structural units, as provided in an embodiment of the present invention. The method includes the following:

[0039] Step S110: Obtain the trained neural network model, wherein the neural network model takes the D2 vector as input and the absorbed energy as output;

[0040] Step S120: Perform surface point cloud analysis on the three-dimensional model of the lattice structure unit to be tested to obtain the D2 distribution associated with the lattice structure unit to be tested, and then obtain the corresponding D2 vector to be tested.

[0041] Step S130: Input the D2 vector to be tested into the neural network model for solution to obtain the absorbed energy of the lattice structure unit to be tested.

[0042] In this embodiment of the invention, the specific training process of the neural network model mentioned in step S110 above includes the following:

[0043] Step S111: Obtain several lattice structure units of different shapes, and then use SolidWorks software to draw the three-dimensional model corresponding to each lattice structure unit, and then output several three-dimensional model files corresponding to the several lattice structure units; wherein, the different shape types include hexagonal star, snowflake, and other shape types that exhibit symmetrical characteristics.

[0044] Step S112: Combining CloudCompare software and Python software, perform surface point cloud analysis on each of the several 3D model files individually, and then output several D2 distributions corresponding to the several lattice structure units, and then extract several corresponding D2 vectors based on the several D2 distributions.

[0045] Step S113: Use Abaqus software to perform uniaxial compression simulation on each of the several three-dimensional model files individually. Before each simulation, the predetermined compression feed should be set to 1mm. Then, output several force-displacement curves corresponding to the several lattice structural units. Then, use Matlab software to perform integration processing on each of the several force-displacement curves individually to obtain several absorbed energies corresponding to the several lattice structural units when they are compressed by 1mm.

[0046] Step S114: A training dataset can be constructed based on the several absorbed energies and several D2 vectors corresponding to the several lattice structure units, and each training data in the training dataset includes the absorbed energy and D2 vector corresponding to any lattice structure unit.

[0047] Step S115: Build a neural network model using the Sklearn toolkit that comes with Python software. The neural network model basically includes an input layer, a hidden layer, and an output layer. The Sklearn toolkit is actually an open-source machine learning toolkit based on the Python language.

[0048] Step S116: Perform iterative training on the neural network model using the training dataset to obtain a trained neural network model.

[0049] It should be noted that the D2 distribution can effectively characterize the geometric shape features of its corresponding lattice structural unit, and the absorbed energy can effectively characterize the mechanical properties of its corresponding lattice structural unit. Since machine learning has a strong nonlinear fitting ability, the neural network model trained by machine learning can fit the correspondence between the mechanical properties of different lattice structural units and their geometric shape features.

[0050] In step S112 above, by way of example, one lattice structure unit is arbitrarily selected from the plurality of lattice structure units mentioned in step S111 and denoted as U1. The specific solution process for the D2 vector corresponding to the lattice structure unit U1 is then explained, including the following steps:

[0051] (1) Obtain the three-dimensional model file corresponding to the lattice structure unit U1 and denote it as FU1. Then, use CloudCompare software to extract all surface point clouds covered in the input three-dimensional model file FU1 to obtain the first surface point cloud coordinate set corresponding to the lattice structure unit U1.

[0052] (2) Perform an internal search on the first surface point cloud coordinate set and obtain several first point cloud data pairs from it by random sampling. Each first point cloud data pair contains two point cloud data with known coordinate information, and at least one point cloud data is different between every two first point cloud data pairs.

[0053] (3) In the plurality of first point cloud data pairs, the first distance value between the two point cloud data in each first point cloud data pair is calculated using the Euclidean distance formula, and then the plurality of first distance values ​​corresponding to the plurality of first point cloud data pairs can be obtained.

[0054] (4) Set the predetermined number of partitions to 1000, filter out the first maximum distance value from the several first distance values ​​and record it as Lmax1 and the first minimum distance value and record it as Lmin1, so that the width of the first interval can be determined as (Lmax1-Lmin1) / 1000.

[0055] (5) A first frequency histogram (also known as the first D2 distribution) can be constructed based on the width of the first interval and the plurality of first distance values. The first distance interval in the first frequency histogram starts with the first minimum distance value Lmin1 and the last distance interval in the first frequency histogram ends with the first maximum distance value Lmax1. The first frequency histogram is used to count the plurality of first distance values ​​to obtain the number of first distance values ​​falling within each distance interval.

[0056] (6) Starting from the first distance interval in the first frequency histogram and continuing to the last distance interval, divide the number of first distance values ​​contained in each distance interval by the predetermined number of partitions to obtain the first D2 vector (that is, the D2 vector corresponding to the lattice structure unit U1), and the dimension of the first D2 vector is the same as the predetermined number of partitions.

[0057] It should be noted that steps (2) to (6) above are performed using Python software to solve the problem; in addition, the number of the plurality of first point cloud data pairs far exceeds the predetermined number of partitions, so as to ensure that the first frequency histogram describes the geometric features of the lattice structure unit U1 more accurately and reliably.

[0058] In this embodiment of the invention, the specific implementation process of step S113 is described by way of example as follows: First, SolidWorks software is used to generate... Figure 2 The 3D model file of the lattice structure unit shown is then imported into Abaqus software and executed according to the predetermined compression feed rate. Figure 3 The uniaxial compression simulation operation shown is presented, and the force-displacement curve corresponding to the lattice structure element is output as follows. Figure 4 As shown, Matlab software was used to perform rapid point sampling and fitting analysis on the force-displacement curve to obtain the expression F(x) for the force-displacement curve. Further integration of this expression F(x) yields the energy absorbed by the lattice structure unit when compressed by 1 mm.

[0059] It should be noted that Abaqus is a powerful finite element analysis software capable of engineering simulation. It includes a rich library of elements that can simulate arbitrary geometries and a material model library that can simulate the performance characteristics of various typical engineering materials such as metals, rubber, polymers, composite materials, and reinforced concrete. The intelligent solver built into Abaqus can solve a large number of stress-displacement analysis problems in structures and can effectively overcome the non-convergence and nonlinearity problems existing in other finite element analysis software. It has the advantages of faster calculation convergence speed and more accurate and reliable simulation results.

[0060] In this embodiment of the invention, the specific implementation process of step S120 includes the following:

[0061] Step S121: Denote the lattice structure unit to be tested as U2, use SolidWorks software to build the three-dimensional model corresponding to the lattice structure unit U2 to be tested, and then output the corresponding three-dimensional model file and denote it as FU2.

[0062] Step S122: Using CloudCompare software, extract all surface point clouds contained in the input 3D model file FU2 to obtain the second surface point cloud coordinate set corresponding to the test lattice structure unit U2.

[0063] Step S123: Perform an internal search on the second surface point cloud coordinate set and obtain several pairs of second point cloud data from it by random sampling. Each pair of second point cloud data contains two point cloud data with known coordinate information, and at least one point cloud data is different between every two pairs of second point cloud data.

[0064] Step S124: In the plurality of second point cloud data pairs, the second distance value between the two point cloud data in each second point cloud data pair is calculated using the Euclidean distance formula, thereby obtaining the plurality of second distance values ​​corresponding to the plurality of second point cloud data pairs.

[0065] Step S125: Set the predetermined number of partitions to 1000, filter out the second maximum distance value from the plurality of second distance values ​​and record it as Lmax2 and the second minimum distance value and record it as Lmin2, thereby determining the width of the second interval as (Lmax2-Lmin2) / 1000.

[0066] Step S126: Based on the second interval width and the plurality of second distance values, a corresponding second frequency histogram (also known as the D2 distribution to be tested) can be constructed. The first distance interval in the second frequency histogram starts with the second minimum distance value Lmin2, and the last distance interval in the second frequency histogram ends with the second maximum distance value Lmax2. The second frequency histogram is used to count the plurality of second distance values ​​to obtain the number of second distance values ​​falling within each distance interval.

[0067] Step S127: Starting from the first distance interval in the second frequency histogram and continuing to the last distance interval, divide the number of second distance values ​​contained in each distance interval by the predetermined number of partitions to obtain the D2 vector to be tested corresponding to the lattice structure unit U2 to be tested, and the dimension of the D2 vector to be tested is the same as the predetermined number of partitions.

[0068] It should be noted that the above steps S123 to S127 are performed using Python software to solve the problem; in addition, the number of the plurality of second point cloud data pairs far exceeds the predetermined number of partitions, so as to ensure that the second frequency histogram describes the geometric features of the test lattice structure unit U2 more accurately and reliably.

[0069] It should be noted that after performing step S130 above, if the designer believes that the energy absorbed by the lattice structure unit under test still cannot meet the original design requirements, the geometric configuration of the lattice structure unit under test will be readjusted and then the above step S120 will be performed again, so that the final lattice structure unit under test has better mechanical properties.

[0070] In this embodiment of the invention, the geometric features of the lattice structure unit are represented by the D2 vector, and the mechanical properties of the lattice structure unit are represented by the absorbed energy. By building a neural network model and pre-learning the correspondence between the mechanical properties of the lattice structure unit under different shape types and its geometric features, the trained neural network model can quickly and efficiently predict the mechanical properties of any lattice structure unit under test based on the geometric features of any input lattice structure unit under test. This eliminates the need to print out the lattice structure unit under test using additive manufacturing technology for physical experimental analysis, avoiding unnecessary material waste and further reducing the design time cost of the lattice structure.

[0071] Please refer to Figure 5 , Figure 5 This is a schematic diagram of the composition of a machine learning-based lattice structure unit performance prediction device provided in an embodiment of the present invention. The device includes:

[0072] The acquisition module 210 is used to acquire a trained neural network model, wherein the neural network model uses the D2 vector as input to predict the final required absorbed energy.

[0073] Analysis module 220 is used to acquire the three-dimensional model of the lattice structure unit to be tested and perform surface point cloud analysis on it, thereby acquiring the D2 distribution associated with the lattice structure unit to be tested, and then extracting the corresponding D2 vector based on the D2 distribution to be tested.

[0074] The solution module 230 is used to solve the input D2 vector to be tested using the neural network model in order to obtain the absorbed energy of the lattice structure unit to be tested.

[0075] The content of the above method embodiments is applicable to the device embodiments. The functions implemented by the device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are the same as those of the above method embodiments, so they will not be repeated here.

[0076] Furthermore, embodiments of the present invention also provide a computer-readable storage medium storing a computer program. When executed by a processor, the computer program implements the machine learning-based lattice structure unit performance prediction method described in the above embodiments. The computer-readable storage medium includes, but is not limited to, any type of disk (including floppy disks, hard disks, optical disks, CD-ROMs, and magneto-optical disks), ROM (Read-Only Memory), RAM (Random Access Memory), EPROM (Erasable Programmable Read-Only Memory), EEPROM (Electrically Erasable Programmable Read-Only Memory), flash memory, magnetic cards, or optical cards. In other words, the storage device includes any medium on which a device (e.g., a computer, mobile phone, etc.) stores or transmits information in a readable form, and can be a read-only memory, a disk, or an optical disk, etc.

[0077] also, Figure 6 This is a schematic diagram of the hardware structure of a computer device provided in an embodiment of the present invention. The computer device includes components such as a processor 320, a memory 330, an input unit 340, and a display unit 350. Those skilled in the art will understand that... Figure 6 The illustrated device structure is not intended to limit all devices and may include more or fewer components than shown, or combine certain components. The memory 330 can be used to store the computer program 310 and various functional modules. The processor 320 runs the computer program 310 stored in the memory 330, thereby performing various functional applications and data processing of the device. The memory can be internal memory or external memory, or include both internal and external memory. Internal memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), flash memory, or random access memory. External memory may include hard disks, floppy disks, ZIP disks, USB flash drives, magnetic tapes, etc. The memory 330 disclosed in the embodiments of this invention includes, but is not limited to, these types of memory. The memory 330 disclosed in the embodiments of this invention is only an example and not a limitation.

[0078] Input unit 340 is used to receive signal input and user-input keywords. Input unit 340 may include a touch panel and other input devices. The touch panel can collect user touch operations on or near it (such as operations performed by the user using a finger, stylus, or any suitable object or accessory on or near the touch panel) and drive the corresponding connection device according to a pre-set program; other input devices may include, but are not limited to, one or more of physical keyboards, function keys (such as play control buttons, power buttons, etc.), trackballs, mice, joysticks, etc. Display unit 350 can be used to display user-input information or information provided to the user, as well as various menus of the terminal device. Display unit 350 may be in the form of a liquid crystal display, organic light-emitting diode, etc. Processor 320 is the control center of the terminal device, connecting various parts of the entire device through various interfaces and lines, performing various functions and processing data by running or executing software programs and / or modules stored in memory 320, and calling data stored in memory.

[0079] As one embodiment, the computer device includes a processor 320, a memory 330, and a computer program 310, wherein the computer program 310 is stored in the memory 330 and configured to be executed by the processor 320, and the computer program 310 is configured to perform the machine learning-based lattice structure unit performance prediction method in the above embodiment.

[0080] Although the description of this application has been quite detailed and particularly focused on several of the described embodiments, it is not intended to limit itself to any of these details or embodiments or any particular embodiment. Rather, it should be considered as effectively covering the intended scope of this application by referring to the appended claims and taking into account the prior art, which provides for a broad possible interpretation of these claims. Furthermore, the foregoing description of this application with respect to embodiments foreseeable by the inventors is intended to provide a useful description, and non-substantial modifications to this application that have not yet been foreseen may still represent equivalent modifications.

Claims

1. A method for predicting the performance of lattice structural units based on machine learning, characterized in that, The method includes: Obtain a trained neural network model, which takes a D2 vector as input and absorbs energy as output; Surface point cloud analysis is performed on the three-dimensional model of the lattice structure unit to be tested to obtain the D2 distribution associated with the lattice structure unit to be tested, and then the corresponding D2 vector to be tested is obtained. The D2 vector to be tested is input into the neural network model for solution to obtain the absorbed energy of the lattice structure unit to be tested; The training process of the neural network model is as follows: Obtain several three-dimensional models corresponding to several lattice structural units, wherein the shape types of the several lattice structural units are different; Surface point cloud analysis is performed on the aforementioned three-dimensional models to obtain several D2 distributions associated with the aforementioned lattice structural units, and then several corresponding D2 vectors are obtained. A uniaxial compression simulation is performed on the several three-dimensional models according to a predetermined compression feed rate to obtain several force-displacement curves corresponding to the several lattice structural units, and then several absorbed energies are obtained accordingly; wherein, the predetermined compression feed rate is 1 mm, and each of the several absorbed energies is obtained by integrating the corresponding force-displacement curves. A training dataset is constructed based on the aforementioned D2 vectors and the aforementioned absorbed energy. A neural network model consisting of an input layer, a hidden layer, and an output layer is constructed, and the training dataset is input into the neural network model for iterative training.

2. The machine learning-based performance prediction method for lattice structural units according to claim 1, characterized in that, The process of obtaining the D2 distribution to be tested is as follows: Several point cloud data pairs are randomly obtained from the three-dimensional model of the lattice structure unit to be tested, and then the distance between two points is calculated for each point cloud data pair to obtain several corresponding distance values. A frequency histogram is constructed based on the aforementioned distance values ​​and the predetermined number of partitions. The frequency histogram is the D2 distribution to be tested.

3. The machine learning-based performance prediction method for lattice structural units according to claim 2, characterized in that, The process of obtaining the D2 vector to be measured is as follows: The number of distance values ​​contained in each distance interval of the D2 distribution to be tested is divided by the predetermined number of partitions to obtain the D2 vector to be tested.

4. The machine learning-based performance prediction method for lattice structural units according to claim 2, characterized in that, The predetermined number of partitions is 1000.

5. A machine learning-based performance prediction device for lattice structural units, characterized in that, The device includes: The acquisition module is used to acquire a trained neural network model, which takes a D2 vector as input and absorbed energy as output. The analysis module is used to perform surface point cloud analysis on the three-dimensional model of the lattice structure unit under test, obtain the D2 distribution associated with the lattice structure unit under test, and then obtain the corresponding D2 vector under test. The solution module is used to input the D2 vector to be tested into the neural network model for solution, so as to obtain the absorbed energy of the lattice structure unit to be tested; The training process of the neural network model is as follows: Obtain several three-dimensional models corresponding to several lattice structural units, wherein the shape types of the several lattice structural units are different; Surface point cloud analysis is performed on the aforementioned three-dimensional models to obtain several D2 distributions associated with the aforementioned lattice structural units, and then several corresponding D2 vectors are obtained. A uniaxial compression simulation is performed on the several three-dimensional models according to a predetermined compression feed rate to obtain several force-displacement curves corresponding to the several lattice structural units, and then several absorbed energies are obtained accordingly; wherein, the predetermined compression feed rate is 1 mm, and each of the several absorbed energies is obtained by integrating the corresponding force-displacement curves. Based on the aforementioned D2 vectors and the aforementioned absorbed energy, a training dataset is constructed; A neural network model consisting of an input layer, a hidden layer, and an output layer is constructed, and the training dataset is input into the neural network model for iterative training.

6. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, The processor executes the computer program to implement the machine learning-based lattice structure unit performance prediction method as described in any one of claims 1 to 4.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the machine learning-based lattice structure unit performance prediction method as described in any one of claims 1 to 4.

Images