Material microstructure three-dimensional reconstruction method

Abstract

The invention relates to the technical field of material microstructure three-dimensional characterization, and discloses a material microstructure three-dimensional reconstruction method, which comprises the following steps: acquiring multi-source data of multiple layers of material slices of a to-be-reconstructed material; the multi-source data corresponding to each layer of material slice comprises three-dimensional shape information and microscopic structure characteristic information of the material slice; for each layer of material slice, based on the multi-source data, constructing a three-dimensional point cloud picture corresponding to the material slice, and based on grain boundary feature constraints, performing interpolation processing on missing value pixel points in the three-dimensional point cloud picture; and generating a three-dimensional reconstruction image corresponding to the to-be-reconstructed material based on the three-dimensional point cloud atlas corresponding to each material slice. According to the method, the problem of reconstruction errors caused by neglecting the characteristics of the non-flat slices prepared by the continuous slicing method in the traditional three-dimensional reconstruction technology is effectively solved, the three-dimensional reconstruction precision of the non-flat continuous slices is remarkably improved, and a high-precision solution is provided for three-dimensional crystallography representation of metal materials.

Description

Technical Field

[0001] This invention relates to the field of three-dimensional characterization technology of material microstructure, and in particular to a method for three-dimensional reconstruction of material microstructure. Background Technology

[0002] In the field of three-dimensional characterization of material microstructures, the sequential slicing method is one of the mainstream techniques for achieving three-dimensional reconstruction of the internal structure of materials. This method involves slicing the material layer by layer, combining techniques such as scanning electron microscopy (SEM) and electron backscatter diffraction (EBSD) to obtain information such as the two-dimensional morphology and crystal orientation of each slice, and then achieving overall structural reconstruction through interlayer registration, data fusion, and three-dimensional modeling. However, existing techniques are generally based on the idealized assumption of "smooth slice surface," ignoring the non-smoothness of the slice surface (such as local undulations and edge deformation) caused by mechanical stress, material anisotropy, or limitations of the preparation process during actual slice preparation. Non-smooth slices introduce geometric position deviations and structural feature distortions between slice layers, leading to cumulative errors in traditional methods when performing interlayer data registration and feature correlation. Especially when dealing with samples with complex crystallographic features, such as metallic materials, the non-smoothness of the slices significantly affects the accuracy of grain boundary identification, grain orientation mapping, and the reconstruction of the three-dimensional topology. Current research on 3D reconstruction of continuous slices mainly focuses on algorithm optimization or single-modal data fusion under flat slicing conditions. It has not yet effectively solved the impact of geometric distortion caused by non-flat slices on the accuracy of 3D reconstruction, making it difficult to meet the needs of high-precision characterization of microcrystalline features of metallic materials.

[0003] Chinese patent CN 118655133 A discloses a method for three-dimensional characterization of polycrystalline superalloy grains. This method uses glow discharge sputtering to prepare two-dimensional microstructure images of polycrystalline superalloy grains of different layers layer by layer and then performs three-dimensional reconstruction. However, this method does not consider the non-uniform surface problem caused by the difference in sputtering yield of different phases during glow discharge sputtering. Traditional three-dimensional reconstruction assumes that the slice is an ideal plane, resulting in errors in the reconstruction results. A study published by Beijing Institute of Technology in the journal *Physics in Medicine & Biology*, titled "Adaptive tetrahedral interpolation for reconstruction of uneven freehand 3D ultrasound," proposed a three-dimensional ultrasound reconstruction algorithm based on probabilistic fusion tetrahedral interpolation. This algorithm addresses the problem of random discrete spatial distribution of slices acquired by freehand ultrasound, improving reconstruction accuracy. However, this method is only applicable to ultrasound slices in the medical field and cannot solve the problem of uneven reconstruction of continuous microstructure slices. Summary of the Invention

[0004] The purpose of this invention is to provide a method for three-dimensional reconstruction of material microstructure. By fusing multi-source data and performing layered interpolation, the accuracy of three-dimensional reconstruction of non-flat slices is significantly improved. This effectively solves the problem of reconstruction error caused by neglecting the characteristics of non-flat continuous slices in traditional three-dimensional reconstruction techniques, and provides a high-precision solution for three-dimensional crystallographic characterization of metallic materials.

[0005] To achieve the above objectives, the present invention provides the following solution:

[0006] A method for three-dimensional reconstruction of material microstructure, comprising the following steps:

[0007] Obtain multi-source data of multi-layer material slices of the material to be reconstructed; the multi-source data corresponding to each layer of material slice includes the three-dimensional morphology information and microstructure characteristics of the material slice;

[0008] For each material slice, a three-dimensional point cloud map corresponding to the material slice is constructed based on the multi-source data, and the missing value pixels in the three-dimensional point cloud map are interpolated based on the grain boundary feature constraints.

[0009] Based on the three-dimensional point cloud map corresponding to each of the material slices, a three-dimensional reconstruction image corresponding to the material to be reconstructed is generated.

[0010] Furthermore, the three-dimensional topography information is obtained using a white light interferometer three-dimensional profilometer;

[0011] The microscopic tissue characteristics are obtained through microscopic tissue characterization techniques, including scanning electron microscopy, metallographic microscopy, or electron backscatter diffraction.

[0012] Furthermore, the interpolation processing of missing value pixels in the 3D point cloud image based on grain boundary feature constraints specifically includes:

[0013] Based on the three-dimensional point cloud map corresponding to the material slice, the boundary interpolation region and the microstructure interpolation region of the material slice are determined; the boundary interpolation region is the interpolation region corresponding to the boundary between each microstructure in the material slice, and the microstructure interpolation region is the interpolation region corresponding to each microstructure.

[0014] Based on at least one of the boundary interpolation region and the microstructure interpolation region, the missing value pixels in the three-dimensional point cloud are interpolated.

[0015] Furthermore, determining the boundary interpolation region and microstructure interpolation region of the material slice specifically includes:

[0016] Based on a preset image processing algorithm, the tissue feature values ​​in the three-dimensional point cloud image are analyzed to identify the boundaries between different microscopic tissues.

[0017] The identified boundary and its preset neighborhood range are defined as the boundary interpolation region;

[0018] The continuous region belonging to the same microstructure in the three-dimensional point cloud image, excluding the boundary interpolation region, is defined as the microstructure interpolation region.

[0019] Furthermore, the method based on a preset image processing algorithm analyzes the tissue feature values ​​in the three-dimensional point cloud image to identify the boundaries between different microscopic tissues, specifically including...

[0020] Using edge detection algorithms or image segmentation algorithms, target pixels at the boundary between any two adjacent microscopic tissues can be located in a 3D point cloud image.

[0021] Based on preset connectivity rules or morphological operations, the target pixels are connected to form continuous boundary lines or boundary regions; wherein, for any target pixel, if preset constraints are met, it is determined whether it belongs to a boundary line or boundary region by comparing the microstructure characteristics of the target pixel and its neighborhood. The preset constraints are set according to the type of material to be reconstructed.

[0022] Furthermore, based on the boundary interpolation region, interpolation processing is performed on the missing value pixels in the 3D point cloud image, specifically including:

[0023] Based on the spatial coordinates of the missing value pixel in the boundary interpolation region, obtain the tissue feature values ​​of the non-missing pixels belonging to different microstructures in their preset neighborhood;

[0024] Based on the physical properties and boundary characteristics of the material to be reconstructed, an interpolation algorithm based on a machine learning model or an adaptive weighted image interpolation algorithm is used to perform interpolation processing and calculate the tissue feature values ​​of the missing pixel points.

[0025] Furthermore, based on the microscopic tissue interpolation region, interpolation processing is performed on the missing value pixels in the three-dimensional point cloud image, specifically including:

[0026] Based on the spatial coordinates of the missing pixel in the 3D point cloud, obtain the organizational feature values ​​of the non-missing pixels in its preset neighborhood;

[0027] Based on the physical properties of the material to be reconstructed, an image interpolation algorithm based on spatial proximity is used to perform interpolation processing and calculate the tissue feature values ​​of the missing pixel points.

[0028] Furthermore, based on the microscopic tissue interpolation region, the interpolation processing of missing value pixels in the three-dimensional point cloud image also includes:

[0029] For each missing value pixel in the microstructure interpolation region, based on the coordinates of the missing value pixel and the tissue feature values ​​of non-missing pixels in the preset neighborhood of the missing value pixel, it is determined whether the missing value pixel is an interpolation point in the microstructure interpolation region. If so, interpolation processing is performed on the missing value pixel.

[0030] Furthermore, for each material slice, constructing a 3D point cloud map corresponding to the material slice based on the multi-source data specifically includes:

[0031] Data preprocessing is performed on three-dimensional morphological information and microstructural characteristics;

[0032] A three-dimensional point cloud spatial coordinate system is established, and the preprocessed three-dimensional morphology information and microstructure feature information are mapped to the spatial coordinate system. Based on the spatial coordinates (x, y, z) in the three-dimensional morphology information, the microstructure feature information is assigned point by point to the three-dimensional point cloud spatial coordinate system.

[0033] Based on the spatial coordinate system mapping results, a three-dimensional point cloud map containing spatial coordinates and microscopic tissue feature values ​​is generated, where pixels with missing values ​​are generated for areas that were not successfully mapped.

[0034] According to specific embodiments provided by the present invention, the following technical effects are disclosed: The three-dimensional reconstruction method for the microstructure of materials provided by the present invention is a correction method based on multi-source data fusion and grain boundary constraints. It acquires the three-dimensional morphology information and microstructure characteristics of multi-layer material slices to be reconstructed. The three-dimensional morphology information reflects the material's positional information, and the microstructure characteristics reflect the crystal orientation of the material slices. For each layer of material slices, a three-dimensional point cloud map is constructed based on the three-dimensional morphology information and the microstructure characteristics. The three-dimensional point cloud map reflects the three-dimensional topological structure characteristics of the material slices, constrained by grain boundary characteristics. The missing feature values ​​are filled in by interpolation; based on the three-dimensional point cloud map corresponding to each material slice, a three-dimensional reconstructed image of the material to be reconstructed is generated; through the above method, for each material slice, combined with three-dimensional morphology information and microstructure feature information, an accurate three-dimensional point cloud map reflecting the three-dimensional topological structure features of the material slice can be obtained. By superimposing the three-dimensional point cloud maps corresponding to each material slice, a high-precision three-dimensional reconstructed image can be obtained. This effectively solves the error problem caused by ignoring the non-flatness of continuous slices in traditional three-dimensional reconstruction, significantly improves the three-dimensional reconstruction accuracy of non-flat continuous slices, and provides a high-precision solution for the three-dimensional crystallographic characterization of metallic materials. Attached Figure Description

[0035] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0036] Figure 1 A flowchart of the method for three-dimensional reconstruction of material microstructure provided by the present invention;

[0037] Figure 2 This is a schematic diagram of the three-dimensional morphology information and grain feature information corresponding to the material slices provided by the present invention, wherein (a) is the three-dimensional morphology information and (b) is the EBSD orientation information;

[0038] Figure 3 This is a grayscale diagram of a certain layer in the three-dimensional point cloud image corresponding to the material slice provided by the present invention.

[0039] Figure 4 A schematic diagram of the boundaries between various microstructures provided by the present invention;

[0040] Figure 5 This is a schematic diagram of the boundary interpolation region and the microstructure interpolation region provided by the present invention. Detailed Implementation

[0041] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0042] To facilitate understanding of the three-dimensional reconstruction method of material microstructure provided in the embodiments of the present invention, relevant knowledge of the embodiments of the present invention will be introduced first.

[0043] Glow discharge, a self-sustaining discharge phenomenon in low-pressure rare gases, achieves sample atomization and ionization through argon ion bombardment. Argon ions are randomly collided and uniformly sputtered onto the sample surface at a wide angle, resulting in low energy (approximately tens of eV) and shallow damage (a few angstroms), clearly revealing interface features such as grains and grain boundaries. However, the difference in sputtering yield between different phases during glow discharge sputtering leads to uneven sputtering pits, which accumulate with the number of slice layers. Traditional 3D reconstruction methods assume the slices are ideal planes, ignoring the unevenness issue, resulting in significant reconstruction errors.

[0044] To address the aforementioned problems, the material structure three-dimensional reconstruction method provided in this embodiment of the invention aims to solve the error problem caused by neglecting the non-flatness of glow discharge sputtering in traditional three-dimensional reconstruction methods. By fusing white light interference three-dimensional morphology data with microstructure feature information to establish a spatial coordinate system mapping relationship, and combining a three-layer constraint model of grain boundary division, orientation continuity constraint and statistical optimization interpolation, a dynamic weight allocation strategy is adopted to iteratively correct the missing values ​​of the point cloud map, which significantly improves the reconstruction accuracy of non-flat sputtered slices.

[0045] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0046] like Figures 1-5 As shown, the method for three-dimensional reconstruction of material microstructure provided by the present invention includes the following steps:

[0047] Step 101: Prepare a series of microstructure sections of the material to be reconstructed using a continuous slicing method (glow sputtering / mechanical polishing / focused ion beam / ultramicroscopic cutting, etc.) until a sufficient number of series sections are obtained.

[0048] In this embodiment of the invention, the material to be reconstructed is, for example, an alloy material, such as a nickel-based superalloy; the microstructure feature information specifically refers to the grain information obtained by EBSD; the continuous slicing method specifically refers to the glow discharge sputtering method.

[0049] In practical applications, the continuous slicing of the material to be reconstructed includes the following steps:

[0050] Step a: Set positioning markers around the reconstruction area on the surface of the material to be reconstructed for positioning during the acquisition of microscopic images of the reconstruction area.

[0051] In practical applications, laser etching combined with focused ion beam can be used to set positioning markers. Positioning markers are set around the pre-reconstruction area, for example, four markers, which are cylindrical pits with a diameter of 5-20 μm and a depth of 20-50 μm.

[0052] As the material is prepared layer by layer by glow discharge sputtering, if the positioning markers are unclear, they can be reset in their original positions. The positioning markers can be positioned at the same location on the surface of the material to be reconstructed, and can be prepared layer by layer along the depth direction of the material surface.

[0053] Step b: Under the set working conditions, perform glow discharge sputtering on the material surface to prepare microstructures.

[0054] In practical applications, the polished material to be reconstructed can be assembled into a glow discharge sputtering device, using a 12mm diameter anode cylinder, with a given discharge voltage of 800V to 1000V and a discharge current of 60 to 100mA, for glow discharge sputtering for 30s to 3min.

[0055] Step 102: Obtain multi-source data corresponding to the multi-layer material slices of the material to be reconstructed; the multi-source data corresponding to each layer of material slices includes the three-dimensional morphology information and microstructure feature information of the material slices. The three-dimensional morphology information is used to reflect the material location information, and the microstructure feature information is obtained through common techniques (such as SEM, metallographic microscope, EBSD).

[0056] In practical applications, multi-source data acquisition for the material to be reconstructed includes the following steps:

[0057] Step a: Acquire the sputtering pit morphology and microstructure images of the reconstructed region on the surface of the material after glow discharge sputtering.

[0058] In step a, the acquisition of sputtering crater morphology refers to obtaining the three-dimensional morphology information of the material slice; the acquisition of microstructure images refers to obtaining the EBSD orientation information of the material slice.

[0059] It should be noted that the acquisition of sputtering crater morphology (i.e., the acquisition of 3D morphology information) can be achieved through the following steps:

[0060] Step a-1: Adjust and set the lens magnification, z-axis scanning range, acquisition area size, and light intensity of the white light interferometer 3D profilometer;

[0061] Step a-2: Use a white light interferometer to measure the glow discharge pits under a low magnification lens. Using the edge of the glow discharge pit as a reference, measure the average depth of the bottom of each glow discharge pit. The difference between the depth measurements of two adjacent glow discharge pits is the average interlayer spacing of the two adjacent glow discharge sample surfaces. Ensure that the average interlayer spacing is the same by controlling the glow discharge conditions and time.

[0062] Step a-3: Use a white light interferometer to measure the pre-reconstructed area within the glow discharge crater under a high-magnification lens; use the positioning markers as the basis for each acquisition.

[0063] Step b, acquiring the microscopic tissue features of the reconstructed region (i.e., obtaining EBSD orientation information), can be achieved through the following steps:

[0064] Step b-1: Adjust and set the operating parameters of the scanning electron microscope, including acquisition parameters, acquisition area size, image resolution, scanning step size, etc.

[0065] Step b-2: Use a scanning electron microscope (SEM) to acquire EBSD images of the surface of each layer of material slice prepared by glow discharge sputtering; during each image acquisition, the SEM is positioned and calibrated based on the positioning markers.

[0066] Figure 2 This is a schematic diagram of the three-dimensional morphology information and EBSD orientation information corresponding to the material slices provided by the present invention.

[0067] Step 103: For each material slice, construct a three-dimensional point cloud map corresponding to the material slice based on the three-dimensional morphology information and the EBSD orientation information; the three-dimensional point cloud map is used to reflect the three-dimensional topological structure features of the material slice.

[0068] Optionally, before constructing the 3D point cloud map corresponding to the material slice, at least one of the following needs to be performed:

[0069] Step a: Export the three-dimensional topography information acquired by white light interferometry as a numerical position matrix format;

[0070] Step b: Export the EBSD orientation information (such as EBSD image information of microstructures) acquired by scanning electron microscope as a numerical position matrix format;

[0071] Step c: Preprocess the three-dimensional topography information and EBSD orientation information. Specifically, downsample the three-dimensional topography information acquired by white light interferometry. In order to match the resolution of the EBSD orientation information, perform noise reduction filtering on the grain orientation information in the EBSD orientation information.

[0072] Step d: Using the positioning markers set around the reconstructed area as a reference, rotate, translate, and crop the two-dimensional microscopic tissue images and three-dimensional morphological information of each layer.

[0073] In this embodiment, multi-source data fusion is performed based on three-dimensional topography information and EBSD orientation information to construct a three-dimensional point cloud map corresponding to the material slice. Specifically, this can be achieved in the following ways:

[0074] The 3D topographic information acquired by the white light interferometer is mapped to the EBSD orientation information in a spatial coordinate system to generate a 3D point cloud map containing missing values ​​(NaN values). Specifically, a 3D point cloud spatial coordinate system is established, and the EBSD orientation information is assigned point by point to the spatial coordinates in the 3D point cloud map according to the spatial coordinates (x, y, z) in the 3D topographic information.

[0075] Step 104: Interpolate the missing pixel values ​​in the three-dimensional point cloud to improve the point cloud data. The interpolation process specifically includes: First, based on the crystallographic orientation difference between voxels in the three-dimensional point cloud, the point cloud is divided into a boundary interpolation region and a microstructure interpolation region; then, the missing values ​​in the microstructure interpolation region are filled using the neighborhood mode interpolation method, and the missing values ​​in the boundary interpolation region are filled using a crystallographic orientation prediction method based on machine learning.

[0076] In practical applications, interpolation processing for missing pixel values ​​in a 3D point cloud image includes the following steps:

[0077] Step a: Based on the three-dimensional point cloud map, distinguish the boundary interpolation region and the microstructure interpolation region of the material slice; the boundary interpolation region is the interpolation region corresponding to the boundary between each microstructure in the material slice, and the microstructure interpolation region is the interpolation region corresponding to each microstructure, specifically:

[0078] Step a-1: For adjacent microstructures in the 3D point cloud image, relevant algorithms in the field of image processing are used to identify multiple target pixels located on the boundary between them. Specifically, by comparing the microstructure feature values ​​of the target pixel and its neighborhood, and combining this with specific constraints set according to the type of material to be reconstructed, it is determined whether the point belongs to the boundary. The specific constraints refer to: a preset crystallographic orientation difference threshold (for example, 10° for nickel-based superalloys); if the crystallographic orientation difference angle between a central voxel and at least one of its adjacent voxels is greater than this threshold, then the central voxel is determined to be a voxel constituting the boundary.

[0079] Step a-2: Using the boundary core pixel as the center, expand outward by 1-3 pixels (adjust according to the slice resolution) to form the boundary interpolation area;

[0080] Step a-3: In the 3D point cloud map, the continuous regions belonging to the same microstructure, excluding the boundary interpolation region, are defined as the microstructure interpolation region.

[0081] Step b: Perform interpolation processing on the missing pixel values ​​in the 3D point cloud image, including:

[0082] Step b-1, Microstructure interpolation region (within the grain):

[0083] For pixels with missing values ​​within this region, interpolation is calculated based on the traditional neighborhood mode interpolation method and preset constraints:

[0084] Extract the effective pixels in the neighborhood of the missing value pixel within a specific range, and use a specific algorithm (such as mode, median or average) to count the neighborhood feature values ​​to generate candidate interpolation;

[0085] Determine whether the candidate interpolation value meets the constraints of microstructure characteristics. If it does, assign the candidate interpolation value to the missing pixel.

[0086] Step b-2, Boundary interpolation region (near grain boundaries):

[0087] For the missing voxels in this region, a crystallographic orientation prediction method based on machine learning is used for filling. This method specifically includes two stages: model training and interpolation application.

[0088] Model training phase: Training samples are extracted from the non-missing value part of the current 3D point cloud. Each sample contains the orientation of a central voxel as a label and the orientation of its neighborhood as features, which are used to train a machine learning prediction model.

[0089] Interpolation prediction stage: For any missing value voxel, its neighborhood features are extracted and input into the trained model. The model predicts the crystallographic orientation of the missing value voxel and fills it with the predicted value.

[0090] Using algorithms related to image processing or machine learning, interpolation is performed based on the coordinates of missing pixels in the boundary interpolation region and their neighborhood feature values, combined with the material type.

[0091] These image processing or machine learning algorithms include, but are not limited to: using the results of the interpolation point recognition model to make a judgment; judging based on the microscopic tissue feature values ​​of the neighborhood of the missing pixel; and using specific image processing or machine learning algorithms to make a judgment. The specific judgment method will be selected based on the type and characteristics of the material to be reconstructed, which will not be elaborated here.

[0092] Figure 3 This is a grayscale schematic diagram of the three-dimensional point cloud corresponding to the material slices provided by this invention. Figure 3 The grayscale image shown includes NaN values.

[0093] Figure 4 This is a schematic diagram of the boundaries between the various microstructures provided by the present invention. See also... Figure 4 As shown, the white areas represent the boundaries between the various microscopic tissues.

[0094] Figure 5 This is a schematic diagram of the boundary interpolation region and the microstructure interpolation region provided by the present invention. (See also...) Figure 5 As shown, the light-colored area is the boundary interpolation area, and the shaded area is the microstructure interpolation area.

[0095] Step 105: Based on the three-dimensional point cloud map corresponding to each of the material slices, generate a three-dimensional reconstruction image corresponding to the material to be reconstructed.

[0096] In this embodiment of the application, each three-dimensional point cloud map can be imported into three-dimensional reconstruction software for image registration and interpolation processing to obtain the final three-dimensional reconstructed image.

[0097] The material structure three-dimensional reconstruction method provided by this invention, for each material slice, combines three-dimensional morphology information and EBSD orientation information to obtain an accurate three-dimensional point cloud map that reflects the three-dimensional topological structure characteristics of the material slice. By superimposing the three-dimensional point cloud maps corresponding to each material slice, a high-precision three-dimensional reconstruction image can be obtained. This effectively solves the error problem caused by neglecting the non-flatness of glow discharge sputtering in traditional three-dimensional reconstruction, and significantly improves the three-dimensional reconstruction accuracy of non-flat glow discharge sputtered continuous slices, providing a high-precision solution for the three-dimensional crystallographic characterization of high-temperature alloys and other metallic materials.

[0098] Example

[0099] The following detailed description, in conjunction with specific embodiments, further illustrates a method for three-dimensional reconstruction of material structures provided by the present invention.

[0100] Taking nickel-based polycrystalline superalloys as an example, the three-dimensional reconstruction of the structure of nickel-based polycrystalline superalloys specifically includes the following steps:

[0101] 1. Sample wire cutting and polishing.

[0102] The nickel-based polycrystalline superalloy sample was cut into cylindrical samples with a diameter of 30 mm and a thickness of 3 mm. The sample surface was then ground and polished to obtain a smooth and clean surface.

[0103] 2. Reconstruct the region and set positioning markers.

[0104] Four marker points were set around the reconstructed region of the nickel-based polycrystalline superalloy sample to serve as the basis for locating microstructure and three-dimensional morphology data. Marking was performed using laser processing and focused ion beam (FIB) methods. Laser processing was performed outside the sputtering domain, while a vertical line perpendicular to the test area was set inside the sample for rough positioning.

[0105] FIB etched four cylindrical pits, each 10 μm in diameter and 20 μm deep. As the layer-by-layer fabrication is carried out using glow discharge sputtering, the marked points can be reset in their original positions if they become unclear.

[0106] 3. Preparation of microstructures from glow discharge sputtered samples.

[0107] The polished nickel-based polycrystalline superalloy was assembled into a glow discharge apparatus, and the microstructure of the sample was prepared by glow discharge sputtering under the working conditions of a discharge voltage of 1000V, a discharge current of 100mA, and a sputtering time of 3min.

[0108] 4. Acquisition of three-dimensional morphology information of sample surface.

[0109] First, a preliminary scan was performed using a low-magnification combination (5× objective and 1× eyepiece) to quickly obtain an overall overview of the sputtered surface, at which point the lateral resolution was 1.376 μm. Using the edge of the glow discharge pit as a reference, the average depth of the bottom of each glow discharge pit was measured. The difference in depth measurements between two adjacent glow discharge pits was the average interlayer spacing between the surfaces of the two adjacent glow discharge sputtered sample layers. The average interlayer spacing was ensured to be the same by controlling the glow discharge sputtering conditions and time.

[0110] Then, a high-magnification combination (50× objective and 1× eyepiece) was used for scanning, which improved the lateral resolution to 0.138μm. Local and precise analysis was performed on a 600μm×60mm area to ensure that the surface morphology information of the corresponding area in the EBSD data could be fully covered and matched.

[0111] 5. Acquisition of microstructure images of sample surface.

[0112] Adjust and set the scanning electron microscope (SEM) acquisition parameters, acquisition area size, image resolution, scanning step size, etc. Select the SEM acceleration voltage as 15kV, beam current as 3nA, EBSD scanning step size as 1μm, and acquisition area as 500μm×500μm to acquire EBSD images of the surface of each sample layer prepared by glow discharge sputtering.

[0113] After the samples prepared by glow discharge sputtering are subjected to microscopic imaging of the reconstructed area of ​​the sample surface and three-dimensional morphology acquisition, they are reassembled into the glow discharge sputtering equipment.

[0114] For example, after obtaining the grain micrograph of the first layer of a nickel-based polycrystalline superalloy, the sample collected for the first layer is placed back into the glow discharge sputtering device to prepare the microstructure of the second layer in the original position.

[0115] Repeat step 3 to prepare the sample surface for the next layer until a two-dimensional microstructure image of 20 layers of nickel-based polycrystalline superalloy grains is obtained.

[0116] 6. Multi-source data collaborative mapping.

[0117] The morphological data acquired by white light interferometry and the microscopic tissue images acquired by scanning electron microscopy are exported as grayscale images and then converted into numerical position matrix format.

[0118] The topographic information and EBSD information acquired by white light interferometry were rotated, translated and cropped to obtain a reconstructed region of 200μm×200μm.

[0119] Specifically, a three-dimensional point cloud spatial coordinate system with a size of 200×200×40 is established, and the EBSD orientation information is assigned point by point to the three-dimensional point cloud spatial coordinate system according to the three-dimensional contour position information (x, y, z).

[0120] 7. Grain boundary feature constraint interpolation.

[0121] We now have 40 layers of sliced ​​data, and we are performing operations on each layer.

[0122] An improved Canny edge detection algorithm is used to delineate the grain boundary forbidden regions. M(i,j) is the EBSD Euler angle value of the microstructure of the center pixel, N8(k) is the eight-neighborhood of the center pixel coordinates, M(k) is the EBSD Euler angle value of the eight-neighborhood, and θ is a threshold set according to the specific material type; in this embodiment, it is set to 10. When there exists a k such that |M(i,j)-M(k)|>θ, this center pixel is determined to be a grain boundary; a 1-3px area around the grain boundary is designated as a forbidden region.

[0123] The area not designated as a no-interpolation zone is the grain core region. Interpolation in this region is forced to satisfy the orientation continuity constraint, which restricts the gradual change of orientation within the grain, and prohibits the use of external data. Specifically, when interpolating in this region, the orientation continuity constraint is considered, which means comparing the average value of the Euler angles of the 8-neighborhood of the NaN pixel at the center point. If the difference between the two is less than ε, the mode interpolation method is used; in this embodiment, ε is taken as 2.

[0124] Based on the probability density function fitted to the actual particle size distribution, the D10 / D50 / D90 feature parameters are extracted.

[0125] By combining convolutional neural network (CNN), neighborhood-based median, and distance-weighted mode (Distance-Weighted Mode) methods, three sets of candidate interpolations are predicted and generated.

[0126] The CNN is assigned the highest weight in the first iteration. In subsequent iterations, the weights are dynamically allocated, and the performance of each interpolation method is evaluated pixel by pixel. Through iterative correction, the optimal distribution is gradually approached to ensure interpolation accuracy.

[0127] 8. The 40 corrected slices were reconstructed in three dimensions and imported into the three-dimensional reconstruction software for image registration and interpolation to obtain the three-dimensional reconstructed structure of the grains in the target nickel-based polycrystalline superalloy microstructure.

[0128] In summary, the three-dimensional reconstruction method for material structures provided by this invention integrates sample wire cutting and polishing, marking points in the reconstruction region, glow discharge sputtering microstructure preparation, acquisition of microstructure images of the sample surface, acquisition of three-dimensional morphology information of the sample surface, multi-source data collaborative mapping, grain boundary feature constraint interpolation, series of image three-dimensional reconstructions, and qualitative and quantitative statistical characterization of the reconstruction. The embodiments of this invention effectively solve the error problem caused by neglecting the non-flatness of glow discharge sputtering in traditional three-dimensional reconstruction, and accurately obtain quantitative statistical information such as the spatial distribution, size, and volume fraction of grains, providing reliable three-dimensional structural data support for material performance analysis, interface behavior research, and process optimization.

[0129] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for three-dimensional reconstruction of the microstructure of a material, characterized in that, Includes the following steps: Obtain multi-source data of multi-layer material slices of the material to be reconstructed; the multi-source data corresponding to each layer of material slice includes the three-dimensional morphology information and microstructure characteristics of the material slice; For each material slice, a three-dimensional point cloud map corresponding to the material slice is constructed based on the multi-source data, and the missing value pixels in the three-dimensional point cloud map are interpolated based on the grain boundary feature constraints. Based on the three-dimensional point cloud map corresponding to each of the material slices, a three-dimensional reconstruction image corresponding to the material to be reconstructed is generated.

2. The method for three-dimensional reconstruction of material microstructure according to claim 1, characterized in that, The three-dimensional topography information was obtained using a white light interferometer three-dimensional profilometer. The microscopic tissue characteristics are obtained through microscopic tissue characterization techniques, including scanning electron microscopy, metallographic microscopy, or electron backscatter diffraction.

3. The method for three-dimensional reconstruction of material microstructure according to claim 1, characterized in that, The interpolation process for missing pixel values ​​in a 3D point cloud image based on grain boundary feature constraints specifically includes: Based on the three-dimensional point cloud map corresponding to the material slice, the boundary interpolation region and the microstructure interpolation region of the material slice are determined; the boundary interpolation region is the interpolation region corresponding to the boundary between each microstructure in the material slice, and the microstructure interpolation region is the interpolation region corresponding to each microstructure. Based on at least one of the boundary interpolation region and the microstructure interpolation region, the missing value pixels in the three-dimensional point cloud are interpolated.

4. The method for three-dimensional reconstruction of material microstructure according to claim 3, characterized in that, The determination of the boundary interpolation region and microstructure interpolation region of the material slice specifically includes: Based on a preset image processing algorithm, the tissue feature values ​​in the three-dimensional point cloud image are analyzed to identify the boundaries between different microscopic tissues. The identified boundary and its preset neighborhood range are defined as the boundary interpolation region; The continuous region belonging to the same microstructure in the three-dimensional point cloud image, excluding the boundary interpolation region, is defined as the microstructure interpolation region.

5. The method for three-dimensional reconstruction of material microstructure according to claim 4, characterized in that, The preset image processing algorithm analyzes the tissue feature values ​​in the three-dimensional point cloud image to identify the boundaries between different microscopic tissues, specifically including... Using edge detection algorithms or image segmentation algorithms, target pixels at the boundary between any two adjacent microscopic tissues can be located in a 3D point cloud image. Based on preset connectivity rules or morphological operations, the target pixels are connected to form continuous boundary lines or boundary regions; wherein, for any target pixel, if preset constraints are met, it is determined whether it belongs to a boundary line or boundary region by comparing the microstructure characteristics of the target pixel and its neighborhood. The preset constraints are set according to the type of material to be reconstructed.

6. The method for three-dimensional reconstruction of material microstructure according to any one of claims 3-5, characterized in that, Based on the boundary interpolation region, interpolation processing is performed on the missing value pixels in the 3D point cloud image, specifically including: Based on the spatial coordinates of the missing value pixel in the boundary interpolation region, obtain the tissue feature values ​​of the non-missing pixels belonging to different microstructures in their preset neighborhood; Based on the physical properties and boundary characteristics of the material to be reconstructed, an interpolation algorithm based on a machine learning model or an adaptive weighted image interpolation algorithm is used to perform interpolation processing and calculate the tissue feature values ​​of the missing pixel points.

7. The method for three-dimensional reconstruction of material microstructure according to any one of claims 3-5, characterized in that, Based on the microscopic tissue interpolation region, interpolation processing is performed on the missing value pixels in the three-dimensional point cloud image, specifically including: Based on the spatial coordinates of the missing pixel in the 3D point cloud, obtain the organizational feature values ​​of the non-missing pixels in its preset neighborhood; Based on the physical properties of the material to be reconstructed, an image interpolation algorithm based on spatial proximity is used to perform interpolation processing and calculate the tissue feature values ​​of the missing pixel points.

8. The method for three-dimensional reconstruction of material microstructure according to claim 7, characterized in that, Based on the microstructure interpolation region, interpolation processing is performed on the missing value pixels in the three-dimensional point cloud image, which further includes: For each missing value pixel in the microstructure interpolation region, based on the coordinates of the missing value pixel and the tissue feature values ​​of non-missing pixels in the preset neighborhood of the missing value pixel, it is determined whether the missing value pixel is an interpolation point in the microstructure interpolation region. If so, interpolation processing is performed on the missing value pixel.

9. The method for three-dimensional reconstruction of material microstructure according to claim 1, characterized in that, For each material slice, based on the multi-source data, a 3D point cloud map corresponding to the material slice is constructed, specifically including: Data preprocessing is performed on three-dimensional morphological information and microstructural characteristics; A three-dimensional point cloud spatial coordinate system is established, and the preprocessed three-dimensional morphology information and microstructure feature information are mapped to the spatial coordinate system. Based on the spatial coordinates (x, y, z) in the three-dimensional morphology information, the microstructure feature information is assigned point by point to the three-dimensional point cloud spatial coordinate system. Based on the spatial coordinate system mapping results, a three-dimensional point cloud map containing spatial coordinates and microscopic tissue feature values ​​is generated, where pixels with missing values ​​are generated for areas that were not successfully mapped.

Images