Methods, apparatuses, and storage media for constructing a database of additive manufacturing dot lattice structure genes

Abstract

The application relates to a method, device and storage medium for constructing a database of genes of additive manufacturing lattice structures. The method comprises: indexing a lattice gene of a received lattice structure according to a predetermined coding rule, the lattice gene comprising a geometry gene and a process gene comprising material attributes, parameterizing modeling of the geometry gene to generate a unit cell model with multiple morphologies and obtain geometry data representing the unit cell model, performing high-throughput mechanical simulation on the unit cell model according to the material attributes and the geometry data to output corresponding phenotype performance data, and correlating and storing the lattice gene, the geometry data corresponding to the lattice gene and the phenotype performance data in a joint indexing manner to form a gene database, so that standardization and expansibility of lattice structure related data are realized, and efficient design and performance prediction of the lattice structure are supported.

Description

Technical Field

[0001] This application relates to the field of materials databases, and more specifically to methods, apparatus and storage media for constructing gene databases of additive manufacturing lattice structures. Background Technology

[0002] Lattice structures, as a typical lightweight, high-performance building unit, are widely used in aerospace, biomedical, and mechanical engineering fields due to their advantages in specific strength, specific stiffness, energy absorption, and thermal management. With the development of metal additive manufacturing technology, complex lattice structures can be fabricated with high precision, thereby promoting their application in functional components.

[0003] However, existing research and applications on additive manufacturing lattice structures still have the following shortcomings: (1) Data fragmentation: Geometric modeling, process parameters, and mechanical performance data are often scattered in different stages, lacking systematic integration and making it difficult to achieve integrated data utilization. (2) Lack of standardization and scalability: Different researchers use inconsistent geometric descriptions, process characterizations, and performance indicators, making it difficult to form scalable data resources. (3) Insufficient structure-process-performance correlation: Existing databases mostly stay at the material or process level, lacking a system / method to introduce the concept of materials genome engineering into lattice structure design.

[0004] In view of the above problems, there is an urgent need for a method to construct a gene database of additive manufacturing lattice structures in order to achieve data standardization and scalability, thereby supporting the efficient design and performance prediction of lattice structures. Summary of the Invention

[0005] The purpose of this application is to provide a method, apparatus and storage medium for constructing a gene database of additive manufacturing lattice structures, so as to achieve standardized integration and joint indexing of lattice genes, geometric data and phenotypic performance data, and to achieve rapid mapping and intelligent prediction between genotype and phenotypic performance of lattice structures by integrating high-throughput simulation, experimental calibration and machine learning prediction, thereby improving the efficiency of lattice structure design and performance evaluation and reducing the cost of experimentation and simulation.

[0006] In a first aspect, this application provides a method for constructing an additive manufacturing lattice structure gene database, comprising: establishing an index for received lattice genes of the lattice structure according to a predetermined encoding rule, wherein the lattice genes include geometric genes and process genes including at least material properties; performing parametric modeling on the geometric genes to generate unit cell models with multiple morphologies and obtaining geometric data representing the unit cell models; performing high-throughput mechanical simulation on the unit cell models according to the material properties and geometric data to output corresponding phenotypic performance data; and establishing and storing the lattice genes, the geometric data corresponding to the lattice genes, and the phenotypic performance data in a joint index manner to form a gene database.

[0007] According to a preferred embodiment of the first aspect, the geometric gene includes node coordinates, rod identification, unit cell topology type, relative density, structural scale parameters, and arrangement; the process gene also includes one or more process conditions selected from beam power, scanning speed, layer thickness, scanning spacing, and scanning path strategy.

[0008] According to a preferred embodiment of the first aspect, parametric modeling of the geometric gene includes: reading a predefined file representing node coordinates and member identifiers, the predefined file containing mapping information for mapping the relative density to member diameters; generating a node coordinate matrix of the unit cell based on the unit cell topology type, structural scale parameters, arrangement, and the mapping information; and determining the connection relationships between members based on the node coordinate matrix and member identifiers, and generating a structured node array and connection relationship array as the geometric data; wherein the parametric modeling generates unit cell models of different morphologies by adjusting the parameters in the geometric gene.

[0009] According to a preferred embodiment of the first aspect, the parametric modeling is executed by an automated modeling script, which generates the node coordinates and connection relationships of each unit cell using a loop iteration method, and calls a 3D modeling kernel to achieve batch automatic generation of the unit cell model.

[0010] According to a preferred embodiment of the first aspect, high-throughput mechanical simulation is performed on the unit cell model based on the material properties and geometric data to output corresponding phenotypic performance data, including: performing finite element simulation on the unit cell model based on the material properties and geometric data under set boundary conditions and loading modes to batch output the phenotypic performance data of the unit cell model; wherein, the material properties include material density, elastic modulus, yield strength, ultimate strength, Poisson's ratio, and stress-strain relationship; the phenotypic performance data includes elastic modulus, yield strength, and ultimate strength.

[0011] According to a preferred embodiment of the first aspect, the method further includes: acquiring measured performance data corresponding to the process gene, wherein the measured performance data is obtained by constructing multiple lattice samples using additive manufacturing equipment and performing mechanical property tests on them; calculating a correction coefficient based on the comparison results of the measured performance data and phenotypic performance data; and correcting the phenotypic performance data using the correction coefficient, and updating the mapping relationship between the corrected phenotypic performance data and the corresponding process conditions in the gene database.

[0012] According to a preferred embodiment of the first aspect, the method further includes: extracting a sample set for training from the gene database, the sample set including at least the matrix genes, geometric data corresponding to the matrix genes, and phenotypic performance data; training a prediction model for mapping the matrix genes to the phenotypic performance data using a machine learning algorithm based on the sample set; and storing the trained prediction model in the gene database as a model file for predicting the corresponding phenotypic performance data based on the input matrix genes.

[0013] According to a preferred embodiment of the first aspect, the method further includes: automatically triggering the retraining or calibration of the prediction model after writing new phenotypic performance data and / or new measured performance data into the gene database, and updating the retrained or calibrated prediction model in the gene database.

[0014] In a second aspect, this application provides an electronic device comprising: at least one processor; at least one memory; said at least one memory being coupled to said at least one processor and for storing instructions executed by said at least one processor, said instructions, when executed by said at least one processor, causing the electronic device to perform the method according to any one of the first aspects.

[0015] Thirdly, this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the method according to any one of the first aspects.

[0016] It should be understood that the above general description and the following detailed description are merely exemplary and do not limit this application. Attached Figure Description

[0017] The accompanying drawings, which are incorporated herein and form part of this specification, illustrate one or more embodiments of the present application and, together with the description, serve to explain the principles of the present application and to enable those skilled in the art to make and use the present application.

[0018] Figure 1A schematic diagram of one of the methods for constructing an additive manufacturing lattice structure gene database provided in this application;

[0019] Figure 2 Another flowchart illustrating the method for constructing an additive manufacturing lattice structure gene database provided in this application;

[0020] Figure 3 A schematic diagram of the parametric modeling process provided for this application;

[0021] Figure 4 A schematic diagram illustrating an exemplary automated modeling process provided in this application;

[0022] Figure 5 The diagram shows the results of obtaining phenotypic performance data provided in this application, where (a) is an example of a set of generated unit cell models, (b) is a scatter plot of elastic modulus and compressive strength corresponding to (a), and (c) is a stress-strain curve corresponding to (a).

[0023] Figure 6 A schematic diagram of the phenotypic performance data correction process provided in this application;

[0024] Figure 7 A schematic diagram illustrating the prediction model construction process provided in this application;

[0025] Figure 8 A schematic diagram of the electronic device provided in this application;

[0026] Figure 9 A schematic diagram of the structure of the computer-readable storage medium provided in this application. Detailed Implementation

[0027] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments may be implemented in various forms and should not be construed as limited to the examples set forth herein; rather, the description of these embodiments is intended to make this application more comprehensive and complete, and to fully convey the concept of the exemplary embodiments to those skilled in the art. The described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments. Numerous specific details are provided in the following description to provide a deeper understanding of embodiments of this application.

[0028] To facilitate understanding of the solutions provided in the embodiments of this application, the terms that may be involved in the embodiments of this application will be introduced first.

[0029] Additive manufacturing, also known as 3D (three-dimensional) printing. In the embodiments of this application, additive manufacturing refers to a manufacturing method that constructs a three-dimensional entity by depositing materials layer by layer, including but not limited to processes such as laser powder bed melting (LPBF), selective laser melting (SLM), electron beam melting (EBM), and direct energy deposition (DED). The term also includes pre- and post-processing and inspection steps in the additive manufacturing process, such as powder preparation, post-processing, and non-destructive testing.

[0030] Lattice structure: refers to a three-dimensional spatial structure composed of repeating or deformable unit cells arranged in a certain manner. A unit cell is usually composed of nodes and rods (also called supports). The mechanical and functional properties of a lattice structure are determined by parameters such as its topology, dimensions, and rod cross-sections.

[0031] Gene Database: A structured database system for storing, managing, and retrieving lattice genes and their corresponding geometric and phenotypic performance data. This gene database indexes lattice design elements using standardized codes and associates lattice genes, including geometric genes, union genes, and material genes, with their corresponding geometric and phenotypic performance data, supporting retrieval, analysis, prediction, and continuous updates.

[0032] Lattice genes: A set of data used to uniquely identify and describe a specific lattice configuration and preparation conditions, including subsets such as geometric genes, process genes, and material genes. Lattice genes should be machine-resolvable and mapped to geometric models and process parameters, serving as the primary key level information in the gene database.

[0033] Geometric Gene: A set of fields in the lattice gene used to characterize the geometric features of a unit cell or lattice, including but not limited to unit cell topology type, unit cell size, node coordinates, member identifier, member cross-sectional parameters, relative density, structural scale parameters, arrangement, and morphology correction factors. The geometric gene is used to drive parametric modeling, generate node arrays and connectivity arrays, and obtain geometric data that can be used for simulation.

[0034] Process Genes: The lattice genes contain a set of fields used to record additive manufacturing-related process parameters, primarily material properties and process conditions. Material properties include, but are not limited to, material density, elastic modulus, yield strength, ultimate strength, Poisson's ratio, and stress-strain relationship. Process conditions include, but are not limited to, beam power, scanning speed, layer thickness, scanning spacing, scanning path strategy, preheating temperature, and protective gas parameters. Beam power, for example, is laser power. Process genes are used to describe manufacturing conditions and for simulation calibration and process-performance mapping.

[0035] Materials genome: A set of fields used to record material information for lattice structures, including but not limited to material grade, chemical composition, powder preparation process, material supplier or manufacturer.

[0036] Unit cell model: A three-dimensional geometric and topological representation used to describe the smallest repeating element in a lattice structure, including parametric descriptions or discretized representations characterizing the unit cell morphology. A unit cell model includes, but is not limited to, a set of nodes, the connection relationships between members / elements, member cross-sections or reference radii, and identifying parameters associated with the unit cell. The set of nodes may include, for example, node identifiers and their three-dimensional coordinates, and the identifying parameters may include, for example, unit cell topology type, unit cell size, relative density, structural scale parameters, and morphology correction factors. A unit cell model can exist in the form of parametric expressions or in a discretized geometric representation, and can be exported as a geometric file for simulation or manufacturing, such as STL, STEP, or a mesh file.

[0037] Geometric data: A structured data set used to represent and store the unit cell model, including but not limited to node arrays, connection relationship arrays, member cross-sections or reference radii information, parametric expressions, and corresponding geometric files. Geometric data typically records its generation source, such as the geometric gene used, mapping rules, and script version, and may also record metadata such as generation time, model version number, and file path. In the embodiments of this application, geometric data and the unit cell model are mutually corresponding; that is, geometric data constitutes a machine-readable representation of the unit cell model, and the unit cell model is a semantic entity of the geometric data.

[0038] Node array: A structured array representing the nodes in a unit cell or lattice. It typically contains node numbers, 3D coordinates, and optional local node attributes, such as member diameter or reference radius and boundary identifier. It is the basic data structure for building connections and generating geometric entities.

[0039] Connection relationship array: A data structure used to describe how links or elements connect nodes. It is usually represented in the form of node index pairs or index sets to represent the endpoints or constituent nodes of each link or finite element element, and is used to construct links, mesh elements or topological relationships.

[0040] Phenotypic performance data refers to performance indicators related to lattice structural mechanics obtained through high-throughput simulations, such as elastic modulus, yield strength, ultimate strength, energy absorption, stiffness-density ratio, fatigue life, or other quantitative performance measures. Phenotypic performance data is used to establish gene-phenotype mappings and database indexes.

[0041] Measured performance data refers to the performance data obtained by conducting physical and mechanical tests or inspections on lattice samples prepared by additive manufacturing equipment. This includes, but is not limited to, elastic modulus, yield strength, ultimate strength, energy absorption, or fatigue life measured under specified test methods and conditions. Measured performance data can be raw measurement data or processed scalar indicators, and also includes corresponding metadata such as test methods and standards, sample preparation information, equipment model, test temperature and environmental conditions, and measurement time, used for result calibration and traceability.

[0042] High-throughput simulation refers to the use of parallel or batch numerical simulation strategies to continuously calculate the mechanical properties of a large number of single-cell models generated from different parameters or lattice genes, such as finite element simulation, in order to obtain a large-scale performance sample set for database filling and model training in a short time.

[0043] Parametric modeling refers to a modeling method that uses a set of adjustable parameters to describe the geometry of lattice units and lattice structures, and generates geometric data in a parameter-driven manner. Parametric modeling may include steps such as parameter resolution, parameter mapping, node coordinate generation, connection relationship determination, and geometric entity construction. These adjustable parameters can be selected from the geometric genes disclosed above.

[0044] Any omissions in the above terminology will be further explained and illustrated in the embodiments below. Next, embodiments of this application will be specifically described with reference to the accompanying drawings. It should be understood that in the description of any drawing, any component or feature mentioned, even if not shown in that drawing, shall be considered as relating to the teaching of that defined component or feature, provided that its reference numerals have been defined in other drawings herein. For the sake of brevity, reference numerals not shown will not be repeatedly indicated when describing a particular drawing in conjunction with other drawings.

[0045] Figure 1 This is a schematic diagram of one of the methods for constructing an additive manufacturing lattice structure gene database provided in this application. Figure 2 Another flowchart illustrating the method for constructing an additive manufacturing lattice structure gene database provided in this application. (Refer to...) Figure 1 and Figure 2 The method 100 provided in this application generally consists of steps 101-104 listed below. At the execution level, method 100 can be executed by system 200, which is, for example, an electronic device including at least one processor and at least one memory, such as a computer device, which is capable of implementing the various steps of method 100 by executing program instructions stored in the memory.

[0046] In step 101, an index is created for the lattice gene 111 of the received lattice structure according to a predetermined encoding rule, wherein the lattice gene 111 includes a geometric gene 121 and a process gene 131 that includes at least material properties.

[0047] In specific implementation, system 200 receives lattice gene 111 descriptions from upstream modules or external interfaces. The input of lattice gene 111 can adopt a structured format or a general tabular format. After receiving the input lattice gene 111, system 200 can perform standardization processing on key fields to convert the original input into an internal representation with unified semantics and format. Standardization processing includes, but is not limited to, unit unification, numerical precision standardization, string cleaning, and enumeration value mapping. After standardization processing, system 200 can parse and extract the necessary fields from geometric gene 121 and process gene 131 according to a predefined data structure. Geometric gene 121 should at least include elements for representing unit cell geometry, such as node coordinates, member identification, unit cell topology type, relative density, structural scale parameters, and arrangement. Process gene 131 should at least include material properties, such as material density, elastic modulus, yield strength, ultimate strength, Poisson's ratio, and stress-strain relationship. After extraction, the required fields are checked, including numerical range validation, enumeration value validity validation, and field integrity validation. For records that do not meet the requirements, the system marks them as abnormal and returns error information for correction. For records that meet the requirements, they can proceed to the encoding and indexing stage.

[0048] The encoding phase generates a unique identifier for each qualified record according to pre-defined encoding rules. This identifier is used to uniquely locate the gene record in the gene database 114 and serves as the primary key for subsequent searches. Encoding rules can employ segmented semantic encoding or a combination of semantic strings and hash values. The encoding format should clearly define the field order, delimiters, decimal truncation rules, and version number location to identify identical or updated records during transmission. Furthermore, while generating the unique identifier, a summary description and tag can be generated for the record to facilitate retrieval. After establishing the unique identifier, system 200 creates a multidimensional index for the record in gene database 114, including but not limited to a unique index based on the unique identifier, a geometric index based on unit cell topology or relative density intervals, an index based on structural scales, and an index based on material property intervals, to meet different query needs. In addition, system 200 can record metadata while writing records and creating indexes. This metadata includes, for example, data source, receipt or generation time, encoding rule version, submitter identifier, verification status, and processing logs.

[0049] In step 102, the geometric gene 121 is parametrically modeled to generate a single-cell model 112 with multiple morphologies and to obtain geometric data 122 representing the single-cell model 112.

[0050] Specifically, system 200 performs parametric modeling based on the geometric genes 121 established in step 101 to generate unit cell models 112 of various morphologies and obtain structured geometric data 122 characterizing the unit cell models 112. The output of parametric modeling is geometric data 122 that can be used for subsequent high-throughput mechanical simulations, or it can be exported as a geometric file for manufacturing for printing verification. The mapping information of the geometric genes 121 can be stored as part of the geometric genes 121 in the gene database 114, or it can be provided by the mapping module during modeling according to actual needs. Parametric modeling can be performed once for a single geometric gene 121, or it can be performed in batches for multiple parameter combinations to efficiently populate the gene database 114.

[0051] Figure 3 A schematic diagram illustrating an exemplary process for parametric modeling provided in this application. (Refer to...) Figure 3 In some embodiments, parametric modeling of the geometric gene 121 may consist of steps 1021-1023 listed below.

[0052] In step 1021, a predefined file representing node coordinates and member identifiers is read. The predefined file contains mapping information for mapping relative density to member diameter.

[0053] Specifically, System 200 reads the predefined file associated with Geometric Gene 121 and extracts the rules used to generate node coordinates and member identifiers, while also obtaining the mapping information that maps relative density to member diameter. The predefined file stores node generation templates, member identifier codes, and relative density-member diameter mapping rules in structured data format. The mapping rules can be implemented through lookup tables, analytical formulas, or fitting functions. After reading the predefined file, System 200 verifies the file format and field integrity. For files that fail verification, error information is recorded and the record is marked as an anomaly for correction. For files that pass verification, System 200 parses the predefined content into an internal data structure and caches it as direct input for subsequent modeling calls.

[0054] For example, system 200 can map relative density to rod diameter based on the principle of volume conservation. The overall mapping approach involves first calculating the total volume of solid material within a unit cell using unit cell geometry. This total volume consists of the sum of the volumes of all rods and the sum of the volumes at each node. The ratio of the solid volume to the unit cell envelope volume is the relative density. The goal of the mapping rule is to find the rod dimensions that satisfy this relative density under given topological and scale conditions. In specific implementation, a hierarchical strategy can be adopted: first, a simplified estimate is made based on the total rod length and topology to obtain an initial value; then, empirical corrections are made for node contributions; next, a lookup table or fitting function is used to quickly obtain a more accurate value; alternatively, numerical iterative solutions including a nodal volume model can be used to satisfy the volume conservation constraint. Nodal volumes can be approximated using a sphere or a union of end hemispheres, or they can be directly calculated during modeling using geometric Boolean merging to obtain a more accurate nodal volume.

[0055] In step 1022, the node coordinate matrix of the unit cell is generated based on the unit cell topology type, structural scale parameters, arrangement method and mapping information.

[0056] Specifically, system 200 generates the node coordinate matrix of the unit cell based on the unit cell topology type, structural scale parameters, arrangement, and mapping information obtained in step 1021. For example, the generation process first determines the initial node layout based on the unit cell topology type and structural scale parameters, and then adjusts the relative positions of the nodes within the unit cell according to the arrangement. If the mapping information specifies the correspondence between relative density and member diameter, system 200 first calculates the diameter of each member based on the relative density and writes the member diameter into the local attributes of the node or member. In addition to recording the three-dimensional coordinates of each node, the node coordinate matrix can also record node-level local attributes, such as member diameter and boundary markers, for use when constructing the geometric entity.

[0057] In step 1023, the connection relationships between the members are determined based on the node coordinate matrix and member identifiers, and a structured node array and connection relationship array are generated as geometric data 122.

[0058] Specifically, system 200 determines the connection relationships between members based on the node coordinate matrix and member identifiers obtained in step 1022, and generates a structured node array and connection relationship array as the core content of geometric data 122. This connection relationship records the endpoints of each member in the form of node index pairs or index sets, and associates each member with its cross-sectional shape, member diameter, or other geometric attributes. Based on the node array and connection relationship array, system 200 can directly construct the solid geometry. The construction method can be to construct cross-sectional entities on the member axes and generate connecting bodies at node positions, and then achieve the connection of entities through geometric fusion or Boolean operations. After completing the solid construction, system 200 performs post-processing operations on the unit cell model 112, such as removing overlaps, repairing small gaps, smoothing transitions, and ensuring minimum wall thickness. Geometric data 122 is finally output in the form of structured geometric data 122 and a geometric file. The geometric file can be in STL or STEP format, and system 200 can establish a mapping relationship between geometric data 122 and the corresponding lattice genes 111 in the gene database 114 for retrieval.

[0059] Each sub-step of step 102 can be executed in parallel. The predefined file reading and parsing in step 1021 can be cached, and the modeling and exporting tasks in steps 1022 and 1023 can be distributed to multi-core or distributed computing nodes for parallel processing.

[0060] It should be understood that the essence of parametric modeling lies in controlling the geometric morphology of the unit cell model 112 by adjusting the parameters recorded in the geometric gene 121. Adjustable parameters include, but are not limited to, unit cell topology type, relative density, structural scale parameters, arrangement, and several morphology correction factors. Relative density is used to limit the volume fraction target of the unit cell, and morphology correction factors are used to fine-tune the cross-sectional shape of the members, the transition morphology of the nodes, or local bending features. During parametric modeling, parameter adjustment should be controlled by several constraints, including but not limited to the minimum machinable feature size of the additive manufacturing equipment, the minimum wall thickness of the member cross-section, the minimum transition radius at the node connection, and the need to avoid excessive overhangs. Compared to traditional manual modeling methods that rely on graphical interfaces, parametric modeling can significantly shorten the modeling cycle, support the establishment of a large number of lattice structure models with different parameter combinations in a short time, and improve the construction efficiency of the gene database 114.

[0061] In some embodiments, parametric modeling can also be implemented through automated modeling scripts, i.e., automated modeling. The modeling script takes the parameter set of geometric genes 121 as input and processes parameter combinations sequentially or in batches according to a loop iteration method. For each parameter combination, the modeling script first generates discrete data representations of node coordinates and connection relationships, and then calls the 3D modeling kernel to construct the corresponding geometric entities, thereby generating unit cell models 112 in batches. The 3D modeling kernel can be any geometry engine that supports procedural geometric construction, such as the Open CASCADE modeling kernel. The modeling script can implement anomaly capture, logging, and metadata writing. When a modeling anomaly occurs, the modeling script can retry, downgrade, or mark the task as requiring manual review according to a preset strategy, and write detailed error logs and the current parameter vector to the gene database 114. Each generated item should record the script version, mapping rule version, input parameter vector, and generation time.

[0062] Figure 4 A schematic diagram illustrating an exemplary automated modeling process provided for this application. (Refer to...) Figure 4 System 200 employs a solid construction strategy based on the connecting lines of the rods. In this strategy, driven by an array of connection relationships, System 200 first constructs cylindrical geometry on the axis of each rod to represent the rod body, and then constructs spherical geometry at each node to ensure transitions and connections at the nodes. The cylinders and spheres are merged into a continuous solid through geometric Boolean operations, ultimately resulting in a unit cell geometry with a complete shape and correct topology. Figure 4 The typical intermediate forms of the automated modeling process are shown: spheres at node locations, cylinders along the axis, node connectors formed by merging sphere and cylinder, and complete unit cell model 112 formed by arraying / arranging.

[0063] Optionally, the system 200 can apply fillet treatment to key connections after Boolean fusion. The fillet treatment can automatically determine the fillet radius based on the cross-sectional dimensions of adjacent members or scale it according to a preset scaling factor, thereby improving the smoothness of stress transmission at the nodes.

[0064] In some complex topologies, multiple members converge at nodes at different angles. Directly using the union of spheres and cylinders can easily lead to excessive solid expansion. To address this, the system 200 can optionally first generate several small spheres at the nodes to approximate the ends of each member, then perform local Boolean synthesis on these small spheres and member cylinders, and perform local scaling or multi-segment fusion on the synthesis result to obtain a node geometry that is both continuous and not excessively expanded.

[0065] In step 103, based on the material properties and geometric data 122, a high-throughput mechanical simulation is performed on the unit cell model 112 to output the corresponding phenotypic performance data 113.

[0066] Specifically, system 200 invokes the high-throughput simulation module to perform batch finite element analysis on the unit cell model 112 generated in step 102. To this end, system 200 first reads the geometric data 122 and material properties corresponding to the current unit cell model 112 from the gene database 114, and generates a standardized simulation input file based on a predefined simulation template. This simulation template includes loading methods and boundary conditions, mesh generation strategy and element type, solver and numerical control parameters, convergence criteria, and output results. The simulation input file includes all information used for solving, such as geometric data 122, material model parameters, loading methods and boundary condition definitions, solver and numerical control parameters, and output control information. Material model parameters refer to the parameters required based on material properties and the selected constitutive model; loading methods and boundary condition definitions refer to the loading method, loading path, and constraint settings of the simulation template; solver and numerical control parameters refer to the solver type, nonlinear control, time step, incremental strategy, or parallel computing settings; and output control information refers to the field variables or scalar performance indices to be extracted. Material properties include at least material density, elastic modulus, yield strength, ultimate strength, Poisson's ratio, and stress-strain relationship. Optional parameters may also include temperature parameters, anisotropy parameters, or hardening model parameters.

[0067] System 200 performs mesh generation on unit cell model 112. Meshing can be performed using a geometric discretization algorithm, with mesh density and element type determined based on the geometric complexity of the unit cell, minimum feature size, and simulation accuracy requirements. For rod structures, beam elements, truss elements, or slender solid elements can be selected; for nodal transition regions, local refinement to solid elements is possible. During the simulation solution phase, System 200 performs finite element analysis on unit cell model 112 under defined boundary conditions and load modes. Boundary conditions can be determined based on the symmetry, periodicity, or loading direction of the unit cell; typical boundary conditions include constraining translational degrees of freedom in a certain direction or applying symmetrical boundaries to opposite surfaces. Load modes can be displacement loading, pressure loading, or force-controlled loading to simulate uniaxial compression, tension, shear, or bending conditions. Furthermore, System 200 can use automatic incremental loading or nonlinear iterative solving to obtain stress-strain response curves. All stress fields, strain fields, reactions, and displacement results generated during the simulation are temporarily stored for post-processing analysis.

[0068] After the simulation is completed, System 200 performs post-processing to extract performance data and calculate features from the simulation output data. Phenotypic performance data 113 can be obtained through analysis of the stress-strain curve: the elastic modulus can be calculated from the slope of the linear segment of the curve; the yield strength can be obtained using the 0.2% offset yield method; the ultimate strength is taken as the peak value of the stress-strain curve or the maximum bearing stress at which the overall structure becomes unstable in the simulation. In addition to the above basic performance indicators, System 200 can also calculate performance parameters such as energy absorption, stiffness-density ratio, deformation modes, or local stress concentration based on the simulation results.

[0069] Figure 5 This is a schematic diagram illustrating the phenotypic performance data acquisition results provided in this application. (Reference) Figure 5 , specifically Figure 5 (a) in the figure is a set of generated unit cell model examples, showing that the unit cells differ in topology, number of nodes and member layout. Figure 5 (b) in the middle is with Figure 5 The scatter plot of elastic modulus versus compressive strength corresponding to (a) shows the distribution trend and dispersion of elastic modulus and compressive strength, and can be used to observe the effect of topological or parameter changes on mechanical properties. Figure 5 (c) in the middle is related to Figure 5 The stress-strain curve corresponding to (a) in the figure reflects the nonlinear response, peak strength and instability characteristics of different unit cell models under the same or similar load conditions. Figure 5 The results shown are exemplary outputs. In actual applications, System 200 extracts and records the above indicators for all simulation items using the same rules, and archives each performance record along with its simulation metadata.

[0070] In high-throughput simulation, system 200 can batch schedule and parallelize simulation tasks for a large number of unit cell models 112. Each simulation task is automatically assigned to a different computing node by system 200, and the result data is collected uniformly after the task is completed. System 200 can write simulation metadata into the simulation record and store it together with the phenotypic performance data 113 in the gene database 114. The simulation metadata includes, for example, the simulation task number, solver type and version, computing resource allocation, runtime, convergence status, and exception records.

[0071] In step 104, the matrix gene 111, the geometric data 122 corresponding to the matrix gene 111, and the phenotypic performance data 113 are linked and stored using a joint index to form a gene database 114.

[0072] It should be understood that although step 104 specifies the storage of matrix genes 111, geometric data 122 corresponding to matrix genes 111, and phenotypic performance data 113, in practical applications, the timing of storing these data is not limited to after step 103. That is, step 104 can be a one-time data entry action after completing all simulations and post-processing, or it can be implemented as a continuous data entry and indexing process. That is, any data can be written into the gene database 114 immediately after it is generated and then a joint index can be established or updated.

[0073] In constructing the joint index, System 200 can establish multidimensional indexes according to predefined strategies, including unique identifier indexes, topological / unit cell type indexes, relative density / structural scale interval indexes, material property interval indexes, and performance interval indexes. For geometric files, System 200 can simultaneously establish indexes based on geometric topology or shape features, such as based on the number of nodes, total length of members, symmetry identifiers, or pre-calculated shape fingerprints, for shape similarity retrieval. The index implementation can combine inverted indexes, B-trees, spatial indexes, or nearest neighbor search structures to meet the needs of different retrieval modes. Furthermore, System 200 can provide multiple retrieval methods for the gene database 114, including direct retrieval by unique identifier, filtering by attribute range, retrieval by multidimensional combination conditions, retrieval by geometric / shape similarity, and sorting by performance indicators. Retrieval results can include structured entries and access links to geometric files. System 200 also provides a batch export interface, supporting formats such as JSON, CSV, or database dumps. Thus, this application will construct a searchable and scalable gene database 114.

[0074] In some embodiments, the lattice gene 111 may have a material gene in addition to the geometric gene 121 and the process gene 131. The material gene includes, but is not limited to, material grade, chemical composition, powder preparation process, material supplier or manufacturer identification, and supply batch number, so as to facilitate traceability or comparative analysis with subsequent new materials.

[0075] Example material gene data table:

[0076]

[0077] In some embodiments, the process gene 131, in addition to material properties, also includes process conditions such as beam power, scanning speed, layer thickness, scanning spacing, scanning path strategy, and preheating temperature. Setting these process conditions allows the mapping between "process and performance" to be incorporated into the gene database 114, enabling the effective tracking of phenotypic differences in the same geometric gene under different process conditions.

[0078] For example, the index relationship of the matrix genes 111 in the gene database 114 can be represented in the following combination form:

[0079] ID = h(G) mat G geo G proc );

[0080] Where: ID is a unique index number; G mat For materials genes; G geo For geometric genes; G proc is the process gene; h is the hash function or combination rule.

[0081] Each subgene involved in the concatenation should be input with a normalized text / structured field sequence, and the concatenation input should include the version number or timestamp of each subgene to distinguish the differences in records of the same content under different versions or different mapping rules. h can use a deterministic hash function (e.g., SHA series) to map the normalized concatenation result to generate a fixed-length unique identifier; for example, when h uses a hash function, system 200 can generate and use the ID according to the following process: First, for G... mat G geo G proc The key fields are concatenated into an intermediate string S in a predetermined order, and a version number is appended to each subfield; then a deterministic hash h(S) is calculated on S and the result is used as the ID; finally, the ID, S (as a readable combined description), and the structured content of each subgene are written into the gene database 114. Alternatively, h can also use a readable combinatorial key strategy, such as concatenating G in a fixed order. mat G geo G proc Key fields are connected with delimiters to form a composite primary key, or a timestamp and sequence number are added to the composite key.

[0082] For example, based on the data types described above, the indexing relationships of gene database 114 can be represented in the following combined form:

[0083] ID = h(G) mat G geo G proc D geo D perf ).

[0084] Where: D geo For geometric data, such as normalized concatenation of node arrays and connection arrays, deterministic hash values ​​of geometric files, and pre-computed shape feature vectors; D perf This refers to phenotypic performance data, such as standardized numerical sequences of elastic modulus, yield strength, and ultimate strength. Where D... geoIt may not be a byte-by-byte hash of the complete geometry file. For example, one or a combination of the following can be used: hashing a normalized sequence of structured node arrays and connection arrays; calculating a file-level hash of the exported geometry file and saving the file path and hash; or calculating and storing geometric feature vectors.

[0085] Figure 6 A schematic diagram illustrating the phenotypic performance data correction process provided in this application. (Reference) Figure 6 In some embodiments, method 100 also involves acquiring measured performance data 115 corresponding to process gene 131, wherein the measured performance data 115 is obtained by constructing multiple lattice samples 117 through additive manufacturing equipment 116 and performing mechanical property tests on them, and is archived together with test conditions, sample identification and test metadata; then, a correction coefficient is calculated based on the comparison result between measured performance data 115 and phenotypic performance data 113, and then the phenotypic performance data 113 is corrected using the correction coefficient, and the mapping relationship between the corrected phenotypic performance data 113 and the corresponding process conditions in gene database 114 is updated.

[0086] During the comparison process, system 200 pairs each measured record with the corresponding phenotypic performance data 113 in gene database 114 based on the joint index of material genes, geometric genes 121, and process genes 131. Before pairing, each data point undergoes a consistency processing step. After pairing, system 200 evaluates the deviation between simulated and measured values ​​and calculates the correction amount accordingly. For indicators exhibiting overall scale deviation, a multiplicative correction coefficient k can be used for correction, exemplarily expressed as k = measured performance data / phenotypic performance data. For a set of paired samples, k can be directly calculated for each sample, and the average or median can be used as the global coefficient. When the deviation is related to material type, topology, or process parameters, a grouping coefficient or parametric correction function (e.g., linear / polynomial fitting) can be used to obtain a correction model that varies with the parameters. After verification, the correction coefficient is applied to the corresponding phenotypic performance data 113 record in gene database 114, and the correction result is written to gene database 114 as a new version record.

[0087] In some embodiments, the measured performance data 115 also includes non-destructive testing (NDT) data. This NDT data is obtained by inspecting the additively manufactured lattice sample 117 without damaging the specimen, and is used to characterize the forming quality and micro-geometric features, providing intermediate quantitative references for performance analysis, simulation calibration, and process optimization. NDT methods may include CT scanning, laser profilometry, three-dimensional microscopy, ultrasonic testing, and surface scanning. Recorded items in the NDT data may include, for example, porosity, surface powder adhesion rate, and average rod diameter. The NDT data can be directly used as part of the measured performance data 115 to correct the phenotypic performance data 113 obtained from high-throughput simulation.

[0088] Figure 7 A schematic diagram illustrating the construction process of the prediction model provided in this application. (Reference) Figure 7 In some embodiments, method 100 also involves extracting a training sample set 118 from gene database 114, then using a machine learning algorithm to train a prediction model 119 for mapping matrix genes 111 to phenotypic performance data 113 based on the sample set 118, and then storing the trained prediction model 119 in gene database 114 as a model file for predicting the corresponding phenotypic performance data 113 based on the input matrix genes 111.

[0089] Specifically, system 200 first extracts a training sample set 118 from gene database 114. Sample set 118 includes at least lattice genes 111 and their corresponding geometric data 122 and phenotypic performance data 113. The lattice genes 111 include at least geometric genes 121 and process genes 131, and may also include material genes and / or simulation metadata. After preprocessing the extracted raw samples, system 200 divides sample set 118 proportionally into training, validation, and test sets. During the training phase, system 200 trains a prediction model 119 based on the training set using a machine learning algorithm to map the lattice genes 111 to the phenotypic performance data 113. This application does not limit the specific algorithm used; for example, deep neural networks, random forests, gradient boosting trees, Gaussian process regression, or combinations thereof may be used. The model structure and hyperparameters are selected based on the validation set, and cross-validation may be used during training to avoid overfitting. After training, system 200 uses a test set to evaluate the performance of prediction model 119, recording evaluation metrics such as mean squared error, root mean square error, and coefficient of determination, and saving the evaluation results. The validated prediction model 119 is saved as a model file in gene database 114. The model file supports version control and records change logs for important operations.

[0090] In model application, system 200 can invoke the stored prediction model 119 to perform inference based on the input lattice gene 111, returning predicted phenotypic performance data 113, such as elastic modulus, yield strength, and ultimate tensile strength. Prediction model 119 provides both real-time and batch inference interfaces. The real-time interface is used for prediction of a single lattice gene input; the batch interface is used for parallel inference of a large number of candidate genes to support high-throughput screening. All inference requests and responses are logged, including the input lattice gene identifier, the invoked model ID / version, the inference time, the output result, and the caller identifier. Through training and application of prediction model 119, system 200 can map lattice genes 111 to reliable phenotypic performance data 113, achieving rapid prediction of mechanical indicators such as elastic modulus, yield strength, and ultimate tensile strength without requiring individual simulations or experiments.

[0091] Furthermore, in some embodiments, after new phenotypic performance data 113 and / or new measured performance data 115 are written into the gene database 114, the system 200 can automatically trigger the retraining or calibration of the prediction model 119, so that the prediction model 119 is continuously updated as real data accumulates in the gene database 114. The triggering strategy for retraining or calibration can be based on a threshold of data volume, such as when the number of newly added measured or simulated samples reaches N or the proportion of newly added samples in the basic training set reaches a certain threshold, or it can be based on monitoring model performance, such as when the root mean square error, mean absolute error, or other indicators of the prediction model 119 on the latest validation set or production samples decrease by more than a set threshold. When any triggering condition is met, the system 200 will automatically enter the retraining or calibration of the prediction model 119. When the amount of new data is limited and mainly used to correct local biases of the model, fine-tuning can be used to continue training on the original model with a smaller learning rate, i.e., calibration, so as to quickly obtain an improved model update with low computational cost; when the amount of new data is large or the data distribution changes significantly, the prediction model 119 can be retrained from scratch.

[0092] Based on the above disclosure of the implementation method 100, it should be understood that this application, by constructing a lattice-structured gene database 114, achieves the integrated and associated storage of materials, geometry, processes, and performance, establishing a unified data description and indexing system. This database supports the retrieval, comparison, and reuse of lattice structures and provides a traceable data foundation for performance analysis under different topological and process conditions. Furthermore, by introducing a machine learning-based prediction model 119, a rapid mapping from lattice genes to phenotypic performance is achieved, enabling accurate prediction of performance under new structures or process conditions without the need for repeated simulations or experiments.

[0093] Figure 8 This is a schematic diagram of the electronic device provided in this application. In some embodiments, the electronic device 300 includes a processor 301 and a memory 302. The number of processors 301 and memory 302 may be one or more. The memory 302 is coupled to the processor 301 and is used to store instructions executed by the processor 301, which, when executed by the processor 301, cause the electronic device 300 to perform the method 100 described in any of the preceding descriptions.

[0094] Specifically, the processor 301 communicates with the memory 302. The method described in this application embodiment can be applied to the processor 301, or implemented by the processor 301. The processor 301 can be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices. The general-purpose processor, such as a microprocessor, can also complete instruction decoding and execution through a combination of hardware logic circuits and software modules. The memory 302 can be volatile memory or non-volatile memory. The volatile memory includes static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), etc. The non-volatile memory includes read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), flash memory, etc., or any combination of the two, used to store the operating system, application programs, data, and the execution code of the above methods.

[0095] It should be understood that the above-described configuration of the electronic device 300 is illustrative and should not be considered as a limitation on the embodiments of this application. Any processor type and storage medium capable of implementing the described processing, storage, and communication functions may be included within the scope of protection of this application.

[0096] Figure 9 This is a schematic diagram of the structure of a computer-readable storage medium provided in this application. The computer-readable storage medium 400 stores a computer program that, when executed by a processor, implements the methods of the embodiments of this application.

[0097] Computer-readable storage media include computer programs for executing computer processes on a computing device. In some embodiments, the computer-readable storage medium is provided using signal-bearing medium 401. Signal-bearing medium 401 may include one or more program instructions that, when executed by one or more processors, can provide the functions or parts thereof described above with respect to the method embodiments of this application. Therefore, for example, one or more features in the method may be embodied by one or more instructions associated with signal-bearing medium 401. Furthermore, Figure 9 The program instructions also describe example instructions. In some examples, the signal carrying medium 401 may include a computer-readable medium 402, which includes a non-volatile storage medium, such as a hard disk drive, optical disc (CD), digital video disc (DVD), digital magnetic tape, read-only memory (ROM), or flash memory chip, such as NOR / NAND type memory, whose data storage characteristics meet the requirements for long-term retention of computer programs.

[0098] In some embodiments, the signal carrying medium 401 may include a computer recordable medium 403, including a rewritable medium, specifically encompassing random access memory (RAM), rewritable optical disc (CD-RW / DVD-RW), solid-state drive (SSD), and phase-change memory (PCM), which supports dynamic updates of program instructions through a read / write controller.

[0099] In some embodiments, the signal carrying medium 401 may include a communication medium 404, such as, but not limited to, digital and / or analog communication media, for example, fiber optic cables, waveguides, wired communication links, wireless communication links, etc.

[0100] The signal carrying medium 401 can be transmitted by a wireless communication medium 404, such as a wireless communication medium conforming to the IEEE 802.11 standard or other transmission protocols. One or more program instructions can be, for example, computer-executable instructions or logical implementation instructions.

[0101] In some embodiments, this application also provides a computer program product, which includes computer-executable instructions that, when executed by processor 301, cause the computer to implement the methods described in any of the preceding claims. A computer program product is a product storing computer-executable instructions intended to implement the various steps and functions described in the methods of this application when executed by processor 301 of a computer system. The computer-executable instructions may include operating systems, application programs, embedded software, etc., to control and manage the process of the method. By using such a computer program product, a user can execute the methods of this application on a computer system.

[0102] It should be understood that the units and algorithm steps of the various examples described in the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0103] Other embodiments of this application will readily conceive of by those skilled in the art upon consideration of the specification and practice of the disclosure herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein. The specification and embodiments are to be considered exemplary only, and this application is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope.

Claims

1. A method for constructing a gene database of additive manufacturing lattice structures, characterized in that, include: An index is established for the received lattice genes of the lattice structure according to a predetermined encoding rule. The lattice genes include geometric genes and process genes that include at least material properties. The geometric genes include node coordinates, rod identifiers, unit cell topology type, relative density, structural scale parameters, and arrangement. The geometric gene is parametrically modeled to generate a single-cell model with various morphologies and to obtain geometric data representing the single-cell model; Based on the material properties and geometric data, high-throughput mechanical simulations are performed on the unit cell model to output the corresponding phenotypic performance data; The matrix genes, the corresponding geometric data and phenotypic performance data are linked and stored using a joint index to form a gene database. Parametric modeling of the geometric gene includes: Read a predefined file representing node coordinates and member identifiers, the predefined file containing mapping information for mapping the relative density to member diameters; The node coordinate matrix of the unit cell is generated based on the unit cell topology type, structural scale parameters, arrangement method, and mapping information; The connection relationships between the members are determined based on the node coordinate matrix and member identifiers, and a structured node array and connection relationship array are generated as the geometric data. The parametric modeling involves adjusting parameters in the geometric gene to generate unit cell models of different morphologies.

2. The method according to claim 1, characterized in that, The process parameters also include one or more process conditions from beam power, scanning speed, layer thickness, scanning spacing, and scanning path strategy.

3. The method according to claim 1, characterized in that, The parametric modeling is executed through an automated modeling script. The modeling script uses a loop iteration method to generate the node coordinates and connection relationships of each unit cell, and calls the 3D modeling kernel to realize the batch automatic generation of the unit cell model.

4. The method according to claim 1, characterized in that, Based on the material properties and geometric data, high-throughput mechanical simulations are performed on the unit cell model to output corresponding phenotypic performance data, including: Under the set boundary conditions and loading modes, finite element simulations are performed on the unit cell model based on the material properties and geometric data to output the phenotypic performance data of the unit cell model in batches; wherein, the material properties include material density, elastic modulus, yield strength, ultimate strength, Poisson's ratio, and stress-strain relationship; the phenotypic performance data includes elastic modulus, yield strength, and ultimate strength.

5. The method according to claim 2, characterized in that, The method further includes: Obtain measured performance data corresponding to the process gene, wherein the measured performance data is obtained by constructing multiple lattice samples using additive manufacturing equipment and testing their mechanical properties; The correction coefficient is calculated based on the comparison between the measured performance data and the phenotypic performance data; and... The phenotypic performance data is corrected using the correction coefficient, and the corrected phenotypic performance data and the corresponding process conditions are updated in the gene database.

6. The method according to any one of claims 1 to 5, characterized in that, The method further includes: Extract a training sample set from the gene database, the sample set including at least the matrix gene, the geometric data corresponding to the matrix gene, and the phenotypic performance data; Based on the sample set, a predictive model for mapping the matrix genes to the phenotypic performance data is trained using a machine learning algorithm; and... The trained prediction model is stored in the gene database as a model file for use in predicting phenotypic performance data based on the input matrix genes.

7. The method according to claim 6, characterized in that, The method further includes: After new phenotypic performance data and / or new measured performance data are written into the gene database, the retraining or calibration of the prediction model is automatically triggered, and the retrained or calibrated prediction model is updated in the gene database.

8. An electronic device, characterized in that, include: At least one processor; At least one memory; The at least one memory is coupled to the at least one processor and is used to store instructions executed by the at least one processor, which, when executed by the at least one processor, cause the electronic device to perform the method according to any one of claims 1 to 7.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the method according to any one of claims 1 to 7.

Images