Method for generating two-dimensional microstructure with adjustable characteristic information based on clustering

Abstract

The present invention discloses a method for generating a two-dimensional microstructure with adjustable characteristic information based on clustering, which can realize the establishment of a more realistic virtual two-dimensional microstructure model, and further realize the control of the size, morphology and orientation of the primary phase, matrix phase and secondary phase. The method of the present invention can be used for generating a method for an adjustable microstructure characteristic model, providing a virtual model support for establishing the mapping relationship from different microstructure characteristics to mechanical properties in the follow-up; solving the problem that the morphology and orientation of the secondary phase are difficult to finely regulate, and on this basis, realizing the control of the three-phase distribution characteristics of the primary phase, secondary phase and matrix phase, and can effectively realize the establishment of a nearly realistic two-dimensional virtual microstructure model, providing technical support for predicting the mechanical properties of microstructures with different characteristic information through finite element simulation in the follow-up.

Description

Technical Field

[0001] The present invention relates to the technical field of numerical simulation, and particularly relates to a method for generating a two-dimensional microstructure with controllable characteristic information based on clustering. Background Art

[0002] In order to obtain high-performance titanium alloys that meet requirements, it is necessary to study the influence law of the microstructure of titanium alloys on their mechanical properties. Mastering the relationship between the microstructure and properties of titanium alloys can better predict the mechanical properties based on the microstructure, thereby revealing the intrinsic relationship of titanium alloys. During actual service, the duplex microstructure of titanium alloys has good comprehensive properties, especially high plasticity and impact toughness. Establishing the mapping relationship from the microscopic microstructure characteristics of the duplex microstructure to the macroscopic mechanical properties and then realizing the optimization design of the microstructure is of great guiding significance for actual experiments. However, in order to obtain a microstructure with good distribution characteristics, it is necessary to regulate through a large number of different heat treatment systems, and at the same time, a large amount of corresponding mechanical property data is required, which will consume a large amount of time, manpower and material costs. Therefore, it is urgent to establish a virtual representative volume element for finite element mechanical property simulation.

[0003] Establishing a mesoscopic tissue model similar to the characteristics of the mesoscopic tissue morphology, size, proportion, etc. of alloy materials is the primary condition for conducting finite element simulation research. Many scholars have constructed two-dimensional and three-dimensional virtual equiaxed VORONOI polycrystalline models for crystal plasticity finite element simulation based on the statistical law of grain shape. The method of the virtual polycrystalline model has high modeling efficiency and low cost, and has become one of the main methods for obtaining the mesoscopic tissue model of crystal structure. However, the current virtual equiaxed VORONOI polycrystalline model simplifies the grains into polygons and has achieved good results in the virtual modeling of some equiaxed grains. For titanium alloys with dual-phase characteristics, some scholars have approximated the duplex microstructure by randomly selecting some grains and dividing them into parallel lamellar regions. However, for the secondary α phase with various typical shapes, such as short rod-shaped, long strip-shaped, oval-shaped, olive-shaped, etc., the existing virtual modeling is difficult to more realistically reflect the distribution effect of the secondary α phase. Therefore, a new method for more realistic virtual modeling is needed to generate a large number of microstructures with good dispersion characteristics. Summary of the Invention

[0004] In view of this, the present invention provides a method for generating a two-dimensional microstructure with controllable characteristic information based on clustering, which can realize the establishment of a more realistic virtual two-dimensional microstructure model, and further realize the control of the size, morphology and orientation of the primary phase, matrix phase and secondary phase.

[0005] A method for generating a two-dimensional microstructure with adjustable characteristic information based on clustering, comprising: establishing a grid model for generating a virtual two-dimensional tissue model; inputting the number, size distribution, orientation distribution, and aspect ratio distribution of primary phase grains into the grid model to perform mean clustering with characteristic distances to generate two-dimensional primary phase grains; generating a matrix phase with a certain degree of dispersion and generating a secondary phase with adjustable characteristic information.

[0006] Among them, generating a matrix phase with a certain degree of dispersion includes: extracting the centroids of the allocated primary phase grains and performing classical mean clustering segmentation, and replacing the primary grains close to the cluster center after clustering with the matrix phase. The number of clustering segmentation parts is determined by the proportion of the primary phase.

[0007] Among them, generating a secondary phase with adjustable characteristic information includes: using the center point of the grid unit where the matrix phase is located as the area for secondary phase allocation, repeating the mean clustering with characteristic distances, and finally generating the final microstructure model by setting a threshold coefficient.

[0008] Among them, in the mean clustering with characteristic distances, based on the traditional mean clustering algorithm, the Euclidean distance in the distance criterion is replaced with an ellipsoidal distance formula with characteristic distance information; the ellipsoidal distance formula is an ellipsoidal distance formula that can be controlled by the major and minor axes and the rotation matrix, and the azimuth angle of the deflection of the ellipsoid is controlled by the rotation matrix.

[0009] Among them, using the center point of the grid unit where the matrix phase is located as the area for secondary phase allocation and repeatedly inputting the microstructure characteristic information of the secondary phase, including the number Ns, size, orientation, and aspect ratio distribution of the secondary phase grains, to perform mean clustering allocation with characteristic distance information to complete the initial generation of the secondary phase; then, statistically calculate the maximum characteristic distance t of each generated secondary phase, and set a coefficient m less than 1 to multiply this maximum characteristic distance mt as the threshold of the maximum characteristic distance of each generated secondary phase. If it is greater than this value, it is the matrix phase, and if it is less than this value, it is the secondary phase.

[0010] Beneficial effects:

[0011] 1. The method for generating a two-dimensional microstructure with adjustable characteristic information based on clustering in the present invention can be used to generate a method for an adjustable microstructure characteristic model, providing a virtual model support for establishing the mapping relationship between different microstructure characteristics and mechanical properties in the future; solving the problem that it is difficult to finely control the morphology and orientation of the secondary phase, and on this basis, realizing the control of the distribution characteristics of the primary phase, secondary phase, and matrix phase, and being able to effectively establish a near-real two-dimensional virtual microstructure model, providing technical support for predicting the mechanical properties of microstructures with different characteristic information through finite element simulation in the future.

[0012] 2. In the preferred embodiment of the present invention, in order to ensure good dispersion of the selected matrix phase in the original primary phase, a classical mean clustering algorithm can be performed, that is, the cluster centers of the primary phase grains assigned in the previous step are extracted for mean clustering, the number of clustering parts is set, and after the clustering assignment is completed, the primary phase grains close to the cluster center are selected as the matrix phase, where the number of clustering parts for this time is determined by the proportion of the overall primary phase.

[0013] 3. In the preferred embodiment of the present invention, in order to simulate the growth process of the secondary phase in the matrix phase, the center point of the grid unit where the matrix phase is located is used as the area for secondary phase assignment, and the microstructural characteristic information of the secondary phase is input in step 2, such as the number of secondary phase grains Ns, size, orientation, and aspect ratio distribution, and then the mean clustering assignment with distance characteristic information can be performed to complete the initial generation of the secondary phase. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 Schematic diagram of the two-dimensional grid model described in the present invention;

[0015] Figure 2 Schematic diagram of the mean clustering algorithm process with characteristic distance information described in the present invention;

[0016] Figure 3 Schematic diagram of the generation of the matrix phase without secondary phase described in the present invention;

[0017] Figure 4 Schematic diagram of the generation of the matrix and secondary phases described in the present invention;

[0018] Figure 5 Schematic diagram of the overall implementation process described in the present invention;

[0019] Figure 6 Schematic diagram of the grid model generated in the verification case of the present invention;

[0020] Figure 7 Schematic diagram of the generation of the representative volume element model with adjustable primary phase morphology in the verification case of the present invention;

[0021] Figure 8 Schematic diagram of the generation of the representative volume element model with adjustable matrix phase ratio in the verification case of the present invention;

[0022] Figure 9 Schematic diagram of the generation of the representative volume element model of the secondary phase with adjustable characteristic information in the verification case of the present invention;

[0023] Figure 10 Schematic diagram of the change in the average distance of the cluster center movement with the increase in the number of iterations during the characteristic distance clustering assignment of the primary and secondary phases in the verification case of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0024] The present invention will be described in detail below with reference to the accompanying drawings and by way of examples.

[0025] As a high-end material, titanium alloy is often used in the fields of aerospace and advanced military industry. The microstructure of titanium alloy mainly comes from two basic phase components, namely, the α solid solution based on α-Ti and the β solid solution based on β-Ti, which can also be called the α phase and the β phase. The properties of these two phases themselves, as well as their morphology, size, distribution and proportion in the alloy, determine the properties of the alloy. Generally, different microstructures can be generated according to different heat treatment systems. According to the content and morphological characteristics of the α phase, the titanium alloy microstructure can be divided into: equiaxed structure, lamellar structure, duplex structure and basket-weave structure. Although there are only four typical microstructures of titanium alloy, after fine-tuning the hot working process for the same type of microstructure, huge differences can occur in the microstructure parameters such as the phase morphology, phase distribution form and grain size, so that the mechanical properties of titanium alloy can vary within a large range. However, in order to obtain a microstructure with good distribution characteristics, it is necessary to regulate through a large number of different heat treatment systems, and at the same time, a large amount of corresponding mechanical property data is required, which will consume a large amount of time, manpower and material costs. Therefore, establishing a large number of virtual representative volume elements with good dispersion for finite element mechanical property simulation has great potential value.

[0026] Aiming at the existing problems, the present invention proposes a method for generating a two-dimensional microstructure with adjustable characteristic information based on clustering, innovatively realizing the establishment of a more realistic virtual two-dimensional microstructure model, strongly realizing the control of the size, morphology and orientation of the primary phase, matrix phase and secondary phase, and successfully providing model support for establishing the mapping relationship between the quantitative characteristic information of the microstructure and the mechanical properties and realizing the tissue optimization design by the finite element method in the follow-up. The method of the present invention includes four parts: establishing a grid model for generating a virtual two-dimensional tissue model, generating a primary phase with adjustable characteristic information, generating a matrix phase with a certain dispersion, and generating a secondary phase with adjustable characteristic information. The specific steps are as follows:

[0027] Step 1: Establish a grid model for generating a virtual two-dimensional tissue model: Set the element size and the size of the two-dimensional region, and generate a single-layer grid model with a specified region size based on the LS-Prepost software for subsequent allocation of different phase grains. Specifically, set the grid size in the LS-PrePost software to obtain a regular pure hexahedron single-layer grid model, that is, divide the specified two-dimensional space with hexahedron elements and obtain the centroid coordinates e(x, y, z) of each element, that is, generate the allocation space, as Figure 1 shown.

[0028] Step 2: Generate the initial phase of adjustable characteristic information: Input the number, size distribution, orientation distribution, and aspect ratio distribution of primary phase grains into the grid model for mean clustering with characteristic distances to achieve the generation of two-dimensional primary phase grains.

[0029] Among them, the mean clustering with characteristic distances replaces the Euclidean distance in the distance criterion with an ellipsoidal distance formula with characteristic distance information based on the traditional mean clustering algorithm. The ellipsoidal distance formula is an ellipsoidal distance formula that can be controlled by the major and minor axes and the rotation matrix. The azimuth angle of the deflection of the ellipsoid can be controlled by the rotation matrix, that is, the shape of the grain can be approximately defined by the size factor, the triaxial ratio, and the spatial Euler angles (in this embodiment, the grain is approximated by an ellipsoid). The specific algorithm flow is as Figure 2 shown. Subsequently, perform clustering assignment with characteristic distance information on the centroid positions e(x, y, z) of the hexahedral elements in the RVE space. The number of clusters is the number of grains Np, and input the information of the size, orientation, and aspect ratio of the primary phase grains. The initial clustering center position is n(x, y, z). Randomly perturb the distributed grain seeds through the clustering algorithm in a loop, and finally achieve the generation of two-dimensional grains, that is, globally optimize to assign all the units closest to the characteristic distance of the random grain seeds to the same grain to obtain the two-dimensional primary phase grain assignment. The flow of Step 2 is as Figure 2 shown.

[0030] Step 3: Generate a matrix phase with a certain degree of dispersion: Extract the centroids of the assigned primary phase grains and perform classical mean clustering segmentation, and replace the primary grains close to the cluster center after clustering with the matrix phase. The number of clustering segmentation parts is determined by the proportion of the primary phase. Specifically, randomly select the assigned primary phase in Step 2 as the matrix phase. To ensure that the selected matrix phase has good dispersion in the original primary phase, it can perform the classical mean clustering algorithm, that is, extract the cluster centers of the assigned primary phase grains in the previous step for mean clustering, set the number of clustering parts, and select the primary phase grains close to the cluster center as the matrix phase after completing the clustering assignment, where the number of clustering parts in this time is determined by the overall proportion of the primary phase. The flow of Step 3 is as Figure 3 shown.

[0031] Step 4: Generate the secondary phase of adjustable characteristic information: Use the center points of the grid units where the matrix phase is located as the area for secondary phase assignment, repeat the mean clustering with characteristic distances described in Step 2, and finally generate the final microstructure model by setting the threshold coefficient.

[0032] Among them, in order to simulate the growth process of secondary phases in the matrix phase, the center point of the grid cell where the matrix phase is located is used as the area for secondary phase distribution. By repeating the input of the microscopic tissue characteristic information of the secondary phase in step two, such as the number of secondary phase grains Ns, size, orientation, and aspect ratio distribution, the initial generation of secondary phases can be completed through mean clustering distribution with distance characteristic information. After that, the maximum value t of the characteristic distance of each generated secondary phase is statistically calculated, and a coefficient m less than 1 is set to multiply this maximum characteristic distance value mt as the threshold of the maximum characteristic distance of each generated secondary phase. If it is greater than this value, it is the matrix phase; if it is less than this value, it is the secondary phase. As Figure 4 shown, the generation of a microstructure with dual-phase characteristic information is finally achieved. The overall implementation flowchart is as Figure 5 shown.

[0033] Taking the grid model generated based on LS-PrePost as an example to verify the present invention, the specific steps are as follows:

[0034] First, set the grid cell size to 0.2 μm, with a length and width of 20 μm. The grid model generated based on LS-PrePost is as Figure 6 shown, and the establishment of the RVE allocation space is completed; then, input the number of primary phases to be allocated (such as: 20, 40, and 60), as well as the corresponding aspect ratios (such as: 3:1, 2:1, and 1:1), orientation angles (such as 30°, 60°, and 90°), and size (such as using a normal probability with a mean of 0.5 and standard deviations of 0.2 and 0.4) distributions for mean clustering distribution with characteristic distance, and a representative volume element model of the primary phase with adjustable characteristic information can be generated, as Figure 7 shown. After that, based on a primary phase model, by setting different primary phase ratios (0.2, 0.3, 0.4, 0.5), representative volume elements containing different matrix phase distributions can be generated, as Figure 8 shown. Then, for the matrix phase with a primary phase ratio of 0.2, secondary phase distribution is carried out. Input the number of secondary phases to be allocated (such as: 400, 600, and 800), as well as different aspect ratios (such as: 8:1, 4:1, and 2:1), different threshold distance parameters (such as: 0.9, 0.7, and 0.5), random orientation angles, and a uniform size distribution for mean clustering distribution with characteristic distance, and a representative volume element model of the secondary phase with adjustable characteristic information can be generated, as Figure 9 shown. Among them, during the clustering iteration process with characteristic distance for the primary and secondary phases, generally, the clustering of the migration perturbation of the clustering centers of the primary and secondary phase grains is less than the grid cell size after about 20 steps. Therefore, in order to save the overall distribution time, the number of clustering iteration steps is fixed at 20 for two times.

[0035] It can be seen that the present invention innovatively provides a brand-new idea for generating a more realistic two-dimensional virtual tissue model, solves the problem that it is difficult to finely control the morphology and orientation of secondary phases, and on this basis, realizes the control of the distribution characteristics of primary, secondary and matrix phases, and can effectively establish a nearly realistic two-dimensional virtual microstructure model, providing technical support for subsequent prediction of the mechanical properties of microstructures with different characteristic information through finite element simulation.

[0036] In summary, the above are only the preferred embodiments of the present invention and are not intended to limit the protection scope of the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for generating a two-dimensional microstructure with adjustable characteristic information based on clustering, characterized in that, it includes the following steps: Step 1, establish a grid model for generating a virtual two-dimensional tissue model: set the cell size and the size of the two-dimensional region, generate a grid model of a single layer with a specified region size based on LS-Prepost software for subsequent allocation of different phase grains, divide the two-dimensional space of the specified size with hexahedral cells, and obtain the centroid coordinates e(x, y, z) of each cell, that is, generate the allocation space; Step 2, generate adjustable characteristic information primary phase: input the number, size distribution, orientation distribution, and aspect ratio distribution of primary phase grains into the grid model for mean clustering with characteristic distance to realize the generation of two-dimensional primary phase grains; Among them, the mean clustering with characteristic distance uses an ellipsoidal distance formula with characteristic distance information. The ellipsoidal distance formula is an ellipsoidal distance formula controlled by the major and minor axes and the rotation matrix. The azimuth angle of the deflection of the ellipsoid is controlled by the rotation matrix, that is, the shape of the grain is approximately defined by the size factor, the triaxial ratio, and the spatial Euler angle. Perform mean clustering allocation with characteristic distance information on the centroid coordinates e(x, y, z) of the hexahedral cells in the RVE space. The number of clusters is the number of grains Np, and input the information of the size, orientation, and aspect ratio of the primary phase grains. The initial clustering center position is n(x, y, z). Randomly perturb the distributed grain seeds through the clustering algorithm in a loop, and finally realize the generation of two-dimensional grains, that is, globally optimize to assign all the cells closest to the characteristic distance of the random grain seeds to the same grain, and obtain the two-dimensional primary phase grain allocation; Step 3, generate a matrix phase with a certain degree of dispersion, including: extract the centroids of the allocated primary phase grains and perform mean clustering segmentation, and replace the primary grains close to the cluster center after clustering with the matrix phase. The number of clustering segmentation parts is determined by the proportion of the primary phase; Step 4, generate adjustable characteristic information secondary phase: use the center points of the grid cells where the matrix phase is located as the region for secondary phase allocation, repeat the mean clustering with characteristic distance described in Step 2, and finally generate the final microstructure model by setting a threshold coefficient.

2. The method according to claim 1, characterized in that, use the center points of the grid cells where the matrix phase is located as the region for secondary phase allocation, repeat inputting the microstructure characteristic information of the secondary phase, including the number of secondary phase grains Ns, size, orientation, and aspect ratio distribution, and perform mean clustering allocation with characteristic distance information to complete the initial generation of the secondary phase; then calculate the maximum characteristic distance t of each generated secondary phase, and set a coefficient m less than 1 to multiply this maximum characteristic distance mt as the threshold of the maximum characteristic distance of each generated secondary phase. If it is greater than this threshold, it is the matrix phase, and if it is less than this threshold, it is the secondary phase.

Images