RVE void evolution prediction and modeling method based on CT-image finite element

Abstract

The invention provides an RVE void evolution prediction and modeling method based on a CT-image finite element. The method comprises the following steps: acquiring a CT scanning image of a large cast ingot with void defects; image processing is carried out, and noise is removed; carrying out image segmentation, and outputting a binary image; splicing the hole binary images to obtain a three-dimensional triangular patch hole image; performing preliminary grid division on the triangular patch cavity graph to obtain a triangular patch digital cavity model; performing curved surface fitting on the triangular patch digital cavity model to obtain a non-uniform rational B-spline cavity model; and generating a casting blank RVE model containing a real cavity. According to the RVE cavity evolution prediction and modeling method based on the CT-image finite element, the microscopic original appearance of the cavity can be truly reflected, and grids can be accurately divided.

Description

Technical Field

[0001] The present invention relates to the technical field of numerical simulation of void defects in steel materials, and specifically to a RVE void evolution prediction and modeling method based on CT-image finite element. The method is particularly suitable for predicting the evolution, mechanical properties, and temperature field of steel materials containing voids with different pore sizes, different types, and other geometric characteristics during the forging process. Background Art

[0002] Large forgings, as core components of large-scale wind turbine equipment, can enhance the efficiency of these devices. However, as the cross-sectional size of the ingot increases, defects such as porosity and shrinkage cavities appear more and more in the core. Subsequent thermal deformation can lead to void deformation, resulting in stress concentration and crack induction, which negatively impacts the mechanical properties and service life of the steel. However, due to the wide range of defects and ingot sizes and the complex processing process, direct observation of the evolution of voids within the ingot is difficult. Currently, finite element simulations primarily use simplified void models pre-built on the macroscopic material structure. This model suffers from mesh distortion caused by the large scale span, resulting in deviated calculation results. Therefore, a method that can accurately reflect the impact of voids on the matrix material is urgently needed. The representative volume element (RVE) model is a key link in establishing a quantitative relationship between pore closure evolution and the macroscopic stress state. By adding appropriate boundary conditions and cutting a section of the large forging to match the void size, more accurate calculation results can be obtained.

[0003] In RVE finite element numerical simulations, the characteristic morphology of the voids in the RVE has a significant impact on the final stress distribution results of the volume unit. Currently, the main methods used at home and abroad to construct void structures are based on statistical feature extraction of experimental data, the use of random algorithms, periodic models, and the use of machine learning generative adversarial networks, variational autoencoders and other technologies to generate new samples. However, none of the above methods of constructing voids can accurately reflect the true appearance of the voids during deformation, and their simulation calculation results are biased. Therefore, it is very necessary to provide an RVE void evolution prediction and modeling method based on CT-image finite elements, which can establish an RVE void model with real morphological characteristics, improve the accuracy of simulation results, and obtain the evolution law of voids during real thermal deformation. Summary of the Invention

[0004] The purpose of the present invention is to provide a CT-image finite element-based RVE void evolution prediction and modeling method that can truly reflect the microscopic original appearance of the void and accurately divide the grid.

[0005] In order to achieve the above technical objectives, the technical solution adopted by the present invention is:

[0006] A method for predicting and modeling RVE cavity evolution based on CT-image finite element analysis, comprising the following steps:

[0007] S1, obtain a cross-sectional image of a large ingot containing void defects;

[0008] S2, preprocessing the tomographic images and performing preliminary image denoising;

[0009] S3, performing image segmentation on the denoised tomographic image to separate the cavity and the matrix to obtain a cavity binary image;

[0010] S4, stitching the binary hole images to form a stitched image, and obtaining a three-dimensional triangular face hole image;

[0011] S5, performing preliminary meshing on the triangular facet cavity map to obtain a triangular facet digital cavity model, and converting it into an STL file that can be recognized by Geomagic software;

[0012] S6, using Geomagic software to perform surface fitting on the triangular facet digital cavity model to obtain the optimized three-dimensional non-uniform rational B-spline cavity model;

[0013] S7, importing the optimized three-dimensional non-uniform rational B-spline void model into Abaqus software to generate a RVE model of the ingot containing real voids.

[0014] Furthermore, in step S1, X-ray computed tomography is used to scan the large ingot containing the void defect to directly obtain the corresponding tomographic image.

[0015] Furthermore, in step S2, the process of performing image preprocessing on the tomographic image includes the following steps:

[0016] Perform voxel correction on the tomographic images to eliminate gain differences caused by different scanning devices and repair bad voxels;

[0017] The random noise is eliminated by using the self-similarity of the image, and the directional noise is identified by the polar coordinate transformation of the voxel points to eliminate the periodic artifacts of the tomographic image; specifically, the opening operation is first used First erode and then dilate to eliminate small noise, and then use the closing operation First dilate and then erode to fill the holes, where B represents the original image and K represents the structural element in the original image. represents the corrosion operation, Represents the dilation operation; the opening and closing operations are performed alternately until the image denoising is completed.

[0018] Furthermore, in step S3, the process of performing image segmentation on the denoised tomographic image includes the following steps:

[0019] (1) Using the threshold binary method, the voxels in the tomographic image with a grayscale greater than the set grayscale threshold are assigned a grayscale of 255, and the voxels with a grayscale less than the set grayscale threshold are assigned a grayscale of 0. The tomographic image is converted into a corresponding grayscale image. The selection of the grayscale threshold is based on the OSTU segmentation algorithm optimized by the sparrow algorithm.

[0020] (2) For the areas where the threshold binary method fails to segment, the grayscale image is divided into the cavity area and the matrix area by the region growing segmentation method, and the cavity binary map is obtained. Specifically, starting from the pre-defined voxel point, the adjacent voxels are iteratively merged and the average grayscale value of the region is recalculated. The merged voxel points that meet the grayscale similarity condition are incorporated into the segmentation area, and the area is gradually expanded until no new points are added; where μ represents the average grayscale value of the area, N represents the number of voxels, and V(i′) represents the observation value of the i′th voxel point.

[0021] Furthermore, in step S3, the OSTU segmentation algorithm optimized based on the sparrow algorithm uses the maximum inter-class variance and the image grayscale histogram to calculate the optimal set of threshold combinations.

[0022] Furthermore, in step S4, each processed binary hole image is stacked, and the stacked binary hole images are interpolated and fused to obtain a three-dimensional triangular face hole image; the splicing rules are as follows: using the MarchingCubes algorithm, traverse each voxel cube to obtain the voxel cube vertex coordinates P and the corresponding field value n, and confirm the connection method of the triangular face; determine the position of the equivalent point according to the linear interpolation method Connect the equal-value points into triangles to form an equal-value surface; where P intersect Represents the interpolation result, T represents the threshold of the isosurface, P1 and P2 represent the coordinate values ​​of two vertices on an edge of the voxel cube, and n1 and n2 represent the field values ​​at these two vertices.

[0023] Furthermore, in step S5, the triangular face hole model is preliminarily meshed using the QEM algorithm to obtain a triangular face digitized hole model, and is converted into a corresponding STL file.

[0024] Furthermore, in step S6, the STL file of the triangular face digitized cavity model is imported into Geomagic software, and the surface of the triangular face digitized cavity model is optimized using the NURBS fitting algorithm and the B-spline fitting algorithm in the Geomagic software to obtain a three-dimensional non-uniform rational B-spline cavity model, and a corresponding IGES file is generated;

[0025] Step S7 further includes:

[0026] S71, importing the 3D non-uniform rational B-spline void model and the large ingot model without voids into Abaqus software, using Boolean operations to embed the 3D non-uniform rational B-spline void model into the ingot and remove redundant parts to generate a volume unit model;

[0027] S72, importing the volume element model into Hypermesh software and performing meshing using the frontier advancing algorithm in Hypermesh software;

[0028] S73, import the meshed volume element model into Abaqus software, and apply periodic boundary conditions to the volume element model; the criteria for applying periodic boundary conditions are as follows: in and They represent the displacement components in the i direction relative to two points on the periodic boundary, ε ij is the symmetric macroscopic strain tensor, L j is the size of the periodic unit in the j direction, ω ik is the antisymmetric rigid rotation tensor, Δx k is the local coordinate difference between the two points in the k direction. i, j, and k can take values ​​of 1, 2, and 3, corresponding to the x, y, and z directions respectively.

[0029] S74. Apply material properties and loads to the volume unit model and perform calculations to generate a RVE model of the ingot containing real voids.

[0030] Furthermore, in step S72, the grid division process is as follows:

[0031] S721, manually generate the frontier region of the volume element model, check the curvature of each edge of the frontier region, and calculate the new node position based on the element size and curvature constraints;

[0032] S722, connect the new nodes to the front edge to form a tetrahedral unit mesh.

[0033] Compared with the prior art, the present invention has the following beneficial effects:

[0034] First, the present method for predicting and modeling RVE void evolution based on CT-image finite elements uses 3D CT images of real large-scale ingot voids to segment and splice the images. Using the maximum inter-class variance OSTU segmentation algorithm optimized by the Sparrow algorithm and the region growing segmentation principle, the accuracy of binarization of each segmented image is ensured, ensuring that the reassembled 3D voids are more consistent with the CT image content, thereby improving the authenticity of the RVE model.

[0035] Second, the RVE cavity evolution prediction and modeling method based on CT-image finite element of the present invention, with the help of non-uniform rational B-spline algorithm (NURBS), reconstructs the triangular facet digital model into a three-dimensional non-uniform rational B-spline model, better controls the curvature of the cavity surface, makes the digital model surface smoother and has smaller errors, ensures that the image finite element model has a high-quality volume mesh, and greatly improves the accuracy of subsequent RVE model simulation results.

[0036] Third, the RVE void evolution prediction and modeling method based on CT-image finite element of the present invention adopts the representative volume element model (RVE) to solve the size span problem between large ingots and tiny voids, and at the same time overcomes the challenge of difficult observation of void defects in large ingots. It greatly improves the accuracy of finite element simulation in predicting the thermal deformation law of voids, and provides important support for accurately characterizing the internal defect evolution mechanism of large ingots. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 The overall flow chart of the RVE cavity evolution prediction and modeling method based on CT-image finite element;

[0038] Figure 2 The figures are the cross-sectional images and SEM images of the casting blank, where (a) is the cross-sectional image obtained by 3D CT scanning, and (b) is the SEM image of the casting blank;

[0039] Figure 3 Schematic diagram of the process of processing void tomographic images, where (a) is the original tomographic image, (b) is the tomographic image after denoising, (c) is the void binary image, and (d) is the triangular patch void model.

[0040] Figure 4 Schematic diagram of the steps of the RVE cavity evolution prediction and modeling method based on CT-image finite element;

[0041] Figure 5 The simulation calculation results of the RVE void model established based on the RVE void evolution prediction and modeling method based on CT-image finite element, where (a) is the RVE stress cloud map under 5% deformation, and (b) is the RVE stress cloud map under 10% deformation. DETAILED DESCRIPTION

[0042] The embodiments of the present invention are described in further detail below with reference to the accompanying drawings.

[0043] See also Figure 1 The present invention discloses a method for predicting and modeling RVE cavity evolution based on CT-image finite element, which comprises the following steps:

[0044] S1, a cross-sectional image of a large ingot containing a void defect;

[0045] S2, preprocessing the tomographic images and performing preliminary image denoising;

[0046] S3, performing image segmentation on the denoised tomographic image to separate the cavity and the matrix to obtain a cavity binary image;

[0047] S4, stitching the binary hole images to form a stitched image, and obtaining a three-dimensional triangular face hole image;

[0048] S5, performing preliminary meshing on the triangular facet cavity map to obtain a triangular facet digital cavity model, and converting it into an STL file that can be recognized by Geomagic software;

[0049] S6, using Geomagic software to perform surface fitting on the triangular facet digital cavity model to obtain the optimized three-dimensional non-uniform rational B-spline cavity model;

[0050] S7, importing the optimized three-dimensional non-uniform rational B-spline void model into Abaqus software to generate a RVE model of the ingot containing real voids.

[0051] See Figure 4 The embodiment of the present invention provides a method for predicting and modeling RVE cavity evolution based on CT-image finite element, the method comprising the following steps:

[0052] Step 1: Obtain a cross-sectional image of a large ingot with a cavity defect in the center.

[0053] like Figure 2 As shown, Figure 2 (a) is a CT scan of a large ingot core section. The gray part is the matrix material and the black part is the cavity defect. Figure 2 (b) is a SEM image obtained by electron microscopy. The voids are distributed randomly across the crystal and vary in size. The tomographic images in this example were obtained using X-ray computed tomography. Void sizes are typically 10-100 μm, so a micrometer CT instrument is required. The resolution can be set arbitrarily, but the higher the resolution, the smaller the void size detected, and the more accurate the resulting true void model.

[0054] Step 2: Preprocess the tomographic image and perform preliminary image denoising. This includes the following steps:

[0055] The voxel points are corrected to eliminate the gain differences caused by different CT devices and repair bad voxel points at the same time; then the image self-similarity is used to eliminate random noise, and the directional noise is identified by voxel point polar coordinate transformation to eliminate periodic artifacts.

[0056] The process of image denoising is as follows: first use the open operation First erode and then dilate to eliminate small noise, and then use the closing operation First dilate and then erode to fill the holes; the opening and closing operations are performed alternately until the image denoising is completed.

[0057] Step 3: Segment the denoised tomographic image to separate the voids and the matrix to obtain a void binary image.

[0058] Image segmentation combines threshold segmentation and region growing to improve the accuracy of matrix and defect segmentation. Threshold segmentation employs a threshold dichotomy method, assigning a grayscale of 255 to voxels above a set grayscale threshold and a grayscale of 0 to voxels below the set grayscale threshold. The grayscale threshold is selected using the OSTU segmentation algorithm optimized by the Sparrow algorithm. Images processed using the threshold dichotomy method are rendered in black and white. Preferably, the OSTU segmentation algorithm optimized by the Sparrow algorithm uses a maximum inter-class variance algorithm, utilizing the image grayscale histogram to determine the optimal threshold combination.

[0059] For the areas that fail to be segmented by the threshold binary method, the grayscale image is divided into the cavity area and the matrix area by the region growing segmentation method, and the cavity binary map is obtained. The region growing segmentation adopts voxel similarity. Specifically, starting from the pre-defined voxel point, the adjacent voxels are iteratively merged and the regional average grayscale value is recalculated. The voxel points that meet the grayscale similarity condition are merged into the segmentation area, and the area is gradually expanded until no new points are added.

[0060] Step 4: stitch the binary hole images to form a stitched image, and obtain a three-dimensional triangular face hole image.

[0061] Each processed binary hole image is aligned and stacked, and the stacked binary hole images are interpolated, fused, and spliced ​​to obtain a three-dimensional triangular face hole image. The splicing rules are as follows: 1) Use the Marching Cubes algorithm to traverse each voxel cube to obtain the voxel point cube vertex coordinates P and the corresponding field value n, and confirm the connection method of the triangular face; 2) Use the linear interpolation method to determine the position of the equivalent point Connect the equal-value points into triangles to form an equal-value surface. The schematic diagram of the 3D cavity image processing process of the large ingot core completed by the above rules is shown in the figure. Figure 3 shown.

[0062] Step 5: Perform preliminary meshing on the triangular facet cavity map to obtain a triangular facet digitized cavity model, which is then converted into an STL file that can be recognized by Geomagic software. Specifically, the QEM algorithm is used for preliminary meshing.

[0063] Step 6: Use Geomagic software to perform surface fitting on the triangular facet digital cavity model to obtain a three-dimensional non-uniform rational B-spline cavity model.

[0064] Specifically, the STL file of the triangular face digitized cavity model is imported into Geomagic software, and the surface of the triangular face digitized cavity model is optimized using the NURBS fitting algorithm and B-spline fitting algorithm in Geomagic software to obtain a three-dimensional non-uniform rational B-spline cavity model, and generate the corresponding IGES file.

[0065] Step 7: Import the optimized non-uniform rational B-spline void model into Abaqus to generate a RVE model of the ingot containing real voids.

[0066] Step 7 specifically includes the following steps:

[0067] Step 7.1. Import the IGES file of the three-dimensional non-uniform rational B-spline void model and the large ingot model without voids into the Abaqus software. Use Boolean operations to embed the three-dimensional non-uniform rational B-spline void model into the ingot and remove redundant parts to generate a volume unit model.

[0068] Step 7.2: Import the volume element model into Hypermesh software and use the frontier advancing algorithm in Hypermesh to remesh the mesh. The remeshing rules are as follows: 1) Manually generate the frontier region of the target volume element model, check the curvature of each edge in the frontier region, and calculate the new node positions based on the element size and curvature constraints; 2) Connect the new nodes to the frontier edges to form a tetrahedral element mesh.

[0069] Step 7.3: Import the meshed volume element model into Abaqus and apply periodic boundary conditions to the volume element model. The criteria for applying periodic boundary conditions are as follows:

[0070] Step 7.4: Apply material properties and loads to the representative volume element and calculate, and finally obtain the corresponding calculation results.

[0071] Specifically, the material properties of the large ingot were added according to the simulation steps of Abaqus software to generate an RVE model of the ingot containing real voids to realize the calculation. The simulation results obtained by the RVE void evolution prediction and modeling method based on CT-image finite element are as follows: Figure 5 shown.

[0072] Although the preferred embodiments of the present application have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present application.

[0073] Obviously, those skilled in the art may make various changes and modifications to this application without departing from the spirit and scope of this application. Thus, if these modifications and variations of this application fall within the scope of the claims of this application and their equivalents, this application is intended to include these modifications and variations.

Claims

1. A method for predicting and modeling RVE cavity evolution based on CT-image finite element, characterized by: The method comprises the following steps: S1, obtain a cross-sectional image of a large ingot containing void defects; S2, preprocessing the tomographic images and performing preliminary image denoising; S3, performing image segmentation on the denoised tomographic image to separate the cavity and the matrix to obtain a cavity binary image; S4, stitching the binary hole images to form a stitched image, and obtaining a three-dimensional triangular face hole image; S5, performing preliminary meshing on the triangular facet cavity map to obtain a triangular facet digital cavity model, and converting it into an STL file that can be recognized by Geomagic software; S6, using Geomagic software to perform surface fitting on the triangular facet digital cavity model to obtain the optimized three-dimensional non-uniform rational B-spline cavity model; S7, importing the optimized three-dimensional non-uniform rational B-spline void model into Abaqus software to generate a RVE model of the ingot containing real voids.

2. The RVE cavity evolution prediction and modeling method based on CT-image finite element according to claim 1, characterized in that: In step S1, a large ingot containing void defects is scanned by X-ray computed tomography to directly obtain a corresponding tomographic image.

3. The RVE cavity evolution prediction and modeling method based on CT-image finite element according to claim 1, characterized in that: In step S2, the process of performing image preprocessing on the tomographic image includes the following steps: Perform voxel correction on the tomographic images to eliminate gain differences caused by different scanning devices and repair bad voxels; The random noise is eliminated by using the self-similarity of the image, and the directional noise is identified by the polar coordinate transformation of the voxel points to eliminate the periodic artifacts of the tomographic image; specifically, the opening operation is first used First erode and then dilate to eliminate small noise, and then use the closing operation First dilate and then erode to fill the holes, where B represents the original image and K represents the structural element in the original image. represents the corrosion operation, Represents the dilation operation; the opening and closing operations are performed alternately until the image denoising is completed.

4. The RVE cavity evolution prediction and modeling method based on CT-image finite element according to claim 1, characterized in that: In step S3, the process of performing image segmentation on the denoised tomographic image includes the following steps: (1) Using the threshold binary method, the voxels in the tomographic image with a grayscale greater than the set grayscale threshold are assigned a grayscale of 255, and the voxels with a grayscale less than the set grayscale threshold are assigned a grayscale of 0. The tomographic image is converted into a corresponding grayscale image. The selection of the grayscale threshold is based on the OSTU segmentation algorithm optimized by the sparrow algorithm. (2) For the areas where the threshold binary method fails to segment, the grayscale image is divided into the cavity area and the matrix area by the region growing segmentation method, and the cavity binary map is obtained. Specifically, starting from the pre-defined voxel point, the adjacent voxels are iteratively merged and the average grayscale value of the region is recalculated. The merged voxel points that meet the grayscale similarity condition are incorporated into the segmentation area, and the area is gradually expanded until no new points are added; where μ represents the average grayscale value of the area, N represents the number of voxels, and V(i′) represents the observation value of the i′th voxel point.

5. The RVE cavity evolution prediction and modeling method based on CT-image finite element according to claim 4, characterized in that: In step S3, the OSTU segmentation algorithm optimized based on the sparrow algorithm uses the maximum inter-class variance and the image grayscale histogram to calculate the optimal set of threshold combinations.

6. The RVE cavity evolution prediction and modeling method based on CT-image finite element according to claim 1, characterized in that: In step S4, each processed binary hole image is stacked, and the stacked binary hole images are interpolated and fused to obtain a three-dimensional triangular face hole image. The splicing rules are as follows: Use the Marching Cubes algorithm to traverse each voxel cube, obtain the vertex coordinates P and the corresponding field value n of the voxel cube, and confirm the connection method of the triangular face; use the linear interpolation method to determine the position of the equivalent point Connect equal-value points into triangles to form equal-value surfaces; Among them, P intersect Represents the interpolation result, T represents the threshold of the isosurface, P1 and P2 represent the coordinate values ​​of two vertices on an edge of the voxel cube, and n1 and n2 represent the field values ​​at these two vertices.

7. The RVE cavity evolution prediction and modeling method based on CT-image finite element according to claim 1, characterized in that: In step S5, the QEM algorithm is used to preliminarily divide the mesh of the triangular face hole model to obtain a triangular face digitized hole model, and the model is converted into a corresponding STL file.

8. The RVE cavity evolution prediction and modeling method based on CT-image finite element according to claim 1, characterized in that: In step S6, the STL file of the triangular face digitized cavity model is imported into Geomagic software, and the surface of the triangular face digitized cavity model is optimized using the NURBS fitting algorithm and the B-spline fitting algorithm in the Geomagic software to obtain a three-dimensional non-uniform rational B-spline cavity model, and generate a corresponding IGES file.

9. The RVE cavity evolution prediction and modeling method based on CT-image finite element according to claim 1, characterized in that: Step S7 further includes: S71, importing the 3D non-uniform rational B-spline void model and the large ingot model without voids into Abaqus software, using Boolean operations to embed the 3D non-uniform rational B-spline void model into the ingot and remove redundant parts to generate a volume unit model; S72, importing the volume element model into Hypermesh software and performing meshing using the frontier advancing algorithm in Hypermesh software; S73, import the meshed volume element model into Abaqus software, and apply periodic boundary conditions to the volume element model; the criteria for applying periodic boundary conditions are as follows: in and They represent the displacement components in the i direction relative to two points on the periodic boundary, ε ij is the symmetric macroscopic strain tensor, L j is the size of the periodic unit in the j direction, ω ik is the antisymmetric rigid rotation tensor, Δx k is the local coordinate difference between the two points in the k direction. i, j, and k can take values ​​of 1, 2, and 3, corresponding to the x, y, and z directions respectively. S74. Apply material properties and loads to the volume unit model and perform calculations to generate a RVE model of the ingot containing real voids.

10. The RVE cavity evolution prediction and modeling method based on CT-image finite element according to claim 9, characterized in that: In step S72, the grid division process is as follows: S721, manually generate the frontier region of the volume element model, check the curvature of each edge of the frontier region, and calculate the new node position based on the element size and curvature constraints; S722, connect the new nodes to the front edge to form a tetrahedral unit mesh.

Images