Method and system for generating XCT-image-based two-dimensional finite element model of composite material

Abstract

The present invention belongs to the technical field of finite element analysis. Disclosed are a method and system for generating an XCT-image-based two-dimensional finite element model of a composite material. The method comprises: collecting an XCT image of a composite material in an initial state, and preprocessing same; performing statistical analysis on two-dimensional mesoscopic structure parameters; on the basis of a Pix2PixHD style transfer network, generating a virtual XCT image; on the basis of a Swin-Transformer network, implementing accurate segmentation of component regions of the virtual XCT image; on the basis of a mapping relationship between pixels and nodes, completing contour feature extraction and smoothing of segmented images; constructing a spatial coordinate system, and performing discrete differentiation on all yarn center points in each segmented image, so as to construct a feature database; on the basis of contour feature position information of each component, constructing feature details and performing mesh generation; inserting cohesive elements; and customizing material properties, assigning local material orientations, and setting boundary conditions.

Description

Method and system for generating two-dimensional finite element model of composite material based on XCT image TECHNICAL FIELD

[0001] The present application belongs to the technical field of finite element analysis, and particularly relates to a method and system for generating a two-dimensional finite element model of a composite material based on an XCT image. BACKGROUND

[0002] Basalt fiber reinforced composite material (BFRP) is a composite product manufactured by basalt fiber as a reinforcing body and resin transfer molding (RTM) process. Basalt fiber reinforced composite material has excellent functional structural characteristics such as light weight and designable performance, and has gradually replaced part of metal and alloy materials, and has been widely used in the fields of transportation, aerospace and military defense, which has great significance for realizing the upgrading of China's manufacturing industry and the light-weight transformation of industry.

[0003] XCT (X-ray Computed Tomography) scanning is a non-destructive testing technology based on X-ray, which can provide high-resolution internal structure information of materials. Through XCT scanning, we can obtain the microstructure of basalt fiber composite material, and use XCT scanning technology to obtain the internal microstructure characterization of the composite material under non-destructive conditions. This provides an important image data acquisition means for further characterization of the internal microstructure of the composite material.

[0004] However, due to the obvious anisotropy and inhomogeneity of the composite material, the macroscopic mechanical property test often needs more experimental samples, complex test equipment, high test cost and long test period. Therefore, for the deformation and damage problem of the composite material under external load, the traditional test method only studies the average mechanical properties of the macroscopic level, ignores the mechanical properties and spatial distribution characteristics of the internal components of the composite material, and fails to establish the cross-scale relationship between the micro-macro mechanical properties.

[0005] Therefore, in recent years, the finite element numerical simulation method has become a common means for studying the mechanical properties of composite materials. By qualitatively and quantitatively characterizing the microstructure parameters, a more realistic finite element model is constructed, the material properties and boundary conditions are set, and the material deformation and damage evolution under real experimental conditions are simulated, so that the surface and internal fracture mechanism of the material can be observed at a lower cost. However, most of the basalt fiber reinforced composite material finite element models often consider three-dimensional whole, ignore the internal relationship between two-dimensional finite element model and three-dimensional finite element model, and it is difficult to accurately express the mechanical response characteristics by only considering three-dimensional finite element model. SUMMARY

[0006] In view of the deficiencies of the prior art, the application provides an XCT image-based composite material two-dimensional finite element model generation method and system, which can be used for quickly and batchedly establishing an XCT image-based basalt composite material two-dimensional finite element model and researching a mesostructure damage evolution process and a rupture state thereof.

[0007] To achieve the above object, the application provides the following scheme.

[0008] A two-dimensional batch finite element model generation and numerical simulation method based on XCT digital images of composite materials, comprising the following steps:

[0009] S1: collecting XCT images of composite materials of basalt fibers in an initial state and performing preprocessing;

[0010] S2: performing parameterized statistical analysis on the preprocessed XCT images of the composite materials, quantifying mesostructure parameters of yarns inside the materials and performing two-dimensional parameterized random modeling to obtain a parameterized random modeling image;

[0011] S3: learning the parameterized random modeling image based on an adversarial learning style transfer network Pix2PixHD to generate a virtual XCT image;

[0012] S4: realizing accurate segmentation of component regions of the virtual XCT image based on a Swin-Tranformer network;

[0013] S5: based on the segmented virtual XCT image, extracting contour mesostructure features and smoothing local contours to construct a yarn and pore feature database;

[0014] S6: based on the yarn and pore feature database, extracting contour features of each yarn of each segmented image, calculating coordinate values of yarn contour centers, establishing a spatial coordinate system, based on the spatial coordinate system, discretely deriving yarn contour centers of each segmented image, and constructing a feature database;

[0015] S7: based on the feature database, extracting feature data information of a single segmented image, classifying component regions, and dividing grids;

[0016] S8: based on the divided grids, inserting cohesive force units and defining material properties of component region features, and distributing local material directions;

[0017] S9: based on the distributed local material directions, setting boundary conditions, and performing numerical simulation calculation;

[0018] S10: repeating S7-S9 to realize rapid generation and numerical simulation of a two-dimensional finite element model based on XCT digital images of composite materials.

[0019] Preferably, in the S1, the method for collecting the initial state basalt fiber composite material XCT image and preprocessing includes:

[0020] S11: Obtain complete XCT scan source image data, and perform cropping and filtering on the XCT scan source image data;

[0021] S12: Based on the cropped and filtered XCT scan source image data, equally spaced extraction of a quantitative number of segmentation images, manual labeling of each component in each segmentation image;

[0022] S13: Feature extraction is performed on the manually labeled segmentation images to obtain a real label set.

[0023] Preferably, in the S2, the method for performing parameterized statistical analysis on the preprocessed composite material XCT image, quantifying the mesostructure parameters of the internal yarns and performing two-dimensional parameterized random modeling includes:

[0024] S21: Based on the mesostructure features of the manually labeled data set, extract the mesostructure parameters of each yarn in each segmentation image;

[0025] S22: Based on the mesostructure parameters of each yarn, qualitatively and quantitatively express the mesostructure features, and apply statistical methods for parameterized statistical analysis;

[0026] S23: Based on the results of parameterized statistical analysis, the data set is expanded through parameterized random modeling to generate a virtual data set, i.e., a parameterized random modeling image.

[0027] Preferably, in the S3, the method for generating a virtual XCT image based on the adversarial learning style transfer network Pix2PixHD includes:

[0028] S31: Generate a warp yarn random model according to the warp yarn parameterized statistical analysis;

[0029] S32: Generate a weft yarn random model according to the weft yarn parameterized statistical analysis;

[0030] S33: Based on the warp yarn random model and the weft yarn random model, generate an artificially synthesized label image;

[0031] S34: Based on the artificially synthesized label image, apply an improved Pix2PixHD network to generate a basalt fiber composite material image with the same real mesostructure, and construct a hybrid data set, i.e., a virtual XCT image.

[0032] Preferably, in the S4, the method for realizing accurate segmentation of each component region of the virtual XCT image based on the Swin-Tranformer network includes:

[0033] S41: input the mixed data set into the Swin-Tranformer network to obtain a training image;

[0034] S42: divide the training image into different small blocks according to pixels through a block division module, and then convert the training image into a one-dimensional feature vector graph;

[0035] S43: accurately segment each component region in the one-dimensional feature vector graph to obtain a semantic segmentation result graph;

[0036] S44: based on the semantic segmentation result graph, use a watershed segmentation algorithm to complete the region segmentation of each yarn.

[0037] Preferably, in S5, the method for extracting profile micro features and locally smoothing the profile based on the segmented virtual XCT image comprises:

[0038] S51: select a resolution that meets the preset requirements to downsample the segmented virtual XCT image;

[0039] S52: perform feature screening on the downsampled virtual XCT image;

[0040] S53: based on the virtual XCT image after feature screening, extract the image profile features of pores and yarns, and perform coordinate value conversion according to the positional relationship between pixels and nodes to obtain profile coordinate values;

[0041] S54: set a convolution kernel that meets the preset requirements, and use the operation between the convolution kernel and the profile coordinate values to realize profile smoothing;

[0042] S55: process all segmented images according to S51-S54 to construct a yarn and pore feature database, and store the profile coordinate values of each component after coordinate system conversion.

[0043] Preferably, in S6, based on the yarn and pore feature database, the method for extracting profile features of each yarn of each segmented image, calculating the coordinate values of the profile centers of the yarns, establishing a spatial coordinate system, and using discrete derivation on the profile centers of each yarn of each segmented image based on the spatial coordinate system comprises:

[0044] S61: extract the coordinate values of the center points of the profiles of each yarn of each segmented image in the yarn and pore feature database, and establish a spatial coordinate system in the order of the segmented images;

[0045] S62: based on the spatial coordinate system, use a first-order intermediate differential derivation rule to perform discrete derivation on the center points of the profiles of each yarn of each segmented image to obtain the tangent direction of the center points, i.e. the 1 direction of the yarn.

[0046] S63: obtaining the 2nd direction from the calculated 1st direction.

[0047] Preferably, in the S7, the method for extracting single segmented image feature data information, classifying each component area, and dividing the grid based on the feature database comprises:

[0048] S71: extracting all yarn contour and pore contour coordinate value data in the single segmented image in the feature database;

[0049] S72: sequentially calling the underlying function function of the Abaqus 2021 software based on the python language;

[0050] S73: based on all yarn contour and pore contour coordinate value data and the underlying function function, dividing the restored each component area according to the position characteristics of the internal microstructure of the material, setting the grid size for grid division.

[0051] Preferably, in the S8, the method for inserting cohesive elements and defining material properties of each component area based on the divided grid and assigning local material direction comprises:

[0052] S81: inserting cohesive elements at the mutual contact interface of yarn and yarn and yarn and matrix from the completed grid;

[0053] S82: gradually assigning each unit of each yarn with the calculated local material direction vector;

[0054] S83: defining material properties and mechanical response characteristics of each component area.

[0055] The application also provides a two-dimensional batch finite element model generation and numerical simulation system based on composite material XCT digital images, comprising: an acquisition module, a modeling module, a learning module, a precise segmentation module, a first database construction module, a second database construction module, a division module, an assignment module, a calculation module, and a model generation module.

[0056] The acquisition module is used to acquire and pretreat the composite material XCT image of basalt fiber in the initial state.

[0057] The modeling module is used to perform parameterized statistical analysis on the pretreated composite material XCT image, quantify the microstructure parameters of the internal yarn of the material, and perform two-dimensional parameterized random modeling to obtain a parameterized random modeling image.

[0058] The learning module is used to learn the parameterized random modeling image based on the adversarial learning style transfer network Pix2PixHD to generate a virtual XCT image.

[0059] The precise segmentation module is used for realizing precise segmentation of each component region of the virtual XCT image based on a Swin-Tranformer network.

[0060] The first database construction module is used for extracting profile microstructure features and locally smoothing the profile based on the segmented virtual XCT image, and constructing a yarn and pore feature database.

[0061] The second database construction module is used for extracting profile features of each yarn of each segmented image, calculating coordinate values of yarn profile centers, establishing a spatial coordinate system, discretely deriving the yarn profile centers of each segmented image based on the spatial coordinate system, and constructing a feature database.

[0062] The division module is used for extracting feature data information of a single segmented image, classifying each component region, and dividing a grid based on the feature database.

[0063] The distribution module is used for inserting cohesive units and defining material properties of each component region feature based on the divided grid, and distributing local material directions.

[0064] The calculation module is used for setting boundary conditions and numerical simulation calculation based on the distributed local material directions.

[0065] The model generation module is used for repeating the division module and the calculation module to realize rapid generation and numerical simulation of a two-dimensional finite element model based on the XCT digital image of the composite material.

[0066] Compared with the prior art, the method and system have the following beneficial effects:

[0067] The application discloses a two-dimensional finite element model generation method and system based on an XCT image of a composite material. An XCT image of the composite material in an initial state is collected and preprocessed. Two-dimensional microstructure parameters are statistically analyzed. A virtual XCT image-virtual label image pair is generated based on a Pix2PixHD style transfer network to realize data enhancement. Each component in the composite material is precisely segmented based on a Swin Transformer network. Profile feature extraction and smoothing of a segmented image are completed based on a mapping relationship between pixels and nodes. A spatial coordinate system is constructed, and all yarn center points of each segmented image are discretely derived to establish a feature database. Feature regions are constructed based on component profile feature coordinate information, and a grid is divided. Cohesive units are inserted, material properties are self-defined, local material directions are distributed, boundary conditions are set, and numerical simulation calculation is completed. The application can be used for rapidly and batch-wise establishing a two-dimensional finite element model of a composite material based on an XCT image, researching a microstructure damage evolution process and a rupture state, and helping to reveal the internal relationship between the two-dimensional finite element model and the three-dimensional whole. BRIEF DESCRIPTION OF DRAWINGS

[0068] In order to more clearly illustrate the technical solutions of the present application, the drawings required to be used in the embodiments will be briefly introduced as follows. Obviously, the drawings described in the following embodiments are only some of the embodiments of the present application, and all other embodiments obtained by those of ordinary skill in the art without creative labor based on these drawings also belong to the protection scope of the present application.

[0069] Fig. 1 is a schematic diagram of the algorithm flow of generating and numerical simulation of two-dimensional batch finite element model based on composite material XCT digital image according to an embodiment of the present application;

[0070] Fig. 2 is a schematic diagram of XCT image segmentation result based on deep learning according to an embodiment of the present application;

[0071] Fig. 3 is a schematic diagram of pre-processing based on segmented XCT image according to an embodiment of the present application; wherein, Fig. 3(a) is a segmented XCT image, and Fig. 3(b) is an XCT image after using down-sampling and screening non-interest region;

[0072] Fig. 4 is a schematic diagram of local material direction distribution algorithm according to an embodiment of the present application; wherein, Fig. 4(a) is the distribution state of single yarn feature in space and center line extraction, Fig. 4(b) is the discrete derivation result of single yarn at the center point on XCT image (only the result of two points is selected as an example), Fig. 4(c) is the distance from the centroid of the calculation unit to all XCT image center points, and Fig. 4(d) is the discrete derivation result of the unit based on the shortest distance;

[0073] Fig. 5 is a schematic diagram of two-dimensional finite element model generation according to an embodiment of the present application; wherein, Fig. 5(a) is a finite element model with cohesive force element, and Fig. 5(b) is the boundary condition setting of the finite element model;

[0074] Fig. 6 is a schematic diagram of two-dimensional finite element model simulation result according to an embodiment of the present application. DETAILED DESCRIPTION

[0075] The technical solutions in the embodiments of the present application will be described clearly and completely in combination with the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, and not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor also belong to the protection scope of the present application.

[0076] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the present application will be further described in detail in combination with the drawings and specific embodiments.

[0077] Embodiment one

[0078] As shown in FIG. 1, FIG. 2, FIG. 3, FIG. 4, FIG. 5 and FIG. 6, the embodiment provides a two-dimensional batch finite element model generation and numerical simulation algorithm flowchart based on composite material XCT digital image, including the following steps:

[0079] S1, collecting an initial state composite material XCT image, based on the XCT image, manually segmenting the image according to the classification of material internal components;

[0080] Specifically, the following steps are included:

[0081] Further preferably, S1 includes:

[0082] S11, obtaining complete XCT scanning source image data, obtaining 1450 source images, appropriately cropping and filtering the source image data, and using the cropping tool to remove the segmented image with blurred edge details;

[0083] S12, extracting one segmented image every 10 segmented images, manually labeling each component in each extracted segmented image, and completely segmenting to obtain four components of resin matrix, warp yarn, weft yarn and pore;

[0084] S13, feature extraction is performed on the labeled image set to obtain a real label set, and a total of 139 real label images are obtained.

[0085] S2, performing parameterized statistical analysis on the manually segmented image, quantifying the mesostructure parameters of the internal yarn of the material and performing two-dimensional parameterized random modeling;

[0086] Further preferably, S2 includes:

[0087] S21, based on the mesostructure features of the manually labeled data set, extracting the yarn length, average width and other mesostructure parameters of each yarn in each segmented image;

[0088] S22, qualitatively and quantitatively expressing the mesostructure features, and applying statistical methods to analyze the statistical results;

[0089] S23, based on the results of parameterized statistical analysis, the data set is expanded through parameterized random modeling to generate a virtual data set;

[0090] The statistical analysis includes warp yarn geometric parameter statistical analysis and weft yarn geometric parameter statistical analysis.

[0091] The method of warp yarn geometric parameter statistical analysis is to statistically analyze the average width, width standard deviation, length and other information of the labeled warp yarn area, calculate the centroid of each warp yarn, and read the distance between the centroids of each warp yarn to generate multiple warp yarns.

[0092] The method for statistical analysis of weft geometric parameters is as follows: the ellipse fitted by the fitellipse function in the OpenCV library is used to calculate the weft inclination angle, the farthest distance of each weft in the image is manually labeled as the major axis, the two intersection points of the straight line perpendicular to the major axis are calculated to obtain the length of the minor axis, and the ratio of the major axis to the minor axis is obtained. The data of each weft collected is substituted into the elliptical equation with the same center, the ratio of the actual radius to the ideal radius at each angle is calculated, and the irregularity of the ellipse is realized.

[0093] Statistical analysis of warp geometric parameters and statistical analysis of weft geometric parameters: using the fitter library to directly and automatically search for the distribution supported in the scipy library for fitting, automatically fitting the mean, standard deviation, length, etc. of the warp, and the long diameter, the ratio of long diameter to short diameter, the inclination angle, etc. of the weft.

[0094] S3, the adversarial learning style transfer network Pix2PixHD learns to generate virtual XCT images from parameterized random modeling images;

[0095] Further preferably, S3 comprises:

[0096] S31, generating a warp random model according to the warp parameterized statistical analysis;

[0097] S32, generating a weft random model according to the weft parameterized statistical analysis;

[0098] S33, generating an artificially synthesized label image based on the warp random model and the weft random model;

[0099] S34, applying the improved Pix2PixHD network to generate basalt fiber composite material images identical to the real microstructure, and constructing a mixed data set;

[0100] The specific implementation method of S31 is as follows:

[0101] First, a straight line is generated as the center line, and then the width of the center line is filled. The center line is divided into two parts by selecting 4 to 6 nodes in a straight line, including the upper and lower end points, and randomly offsetting the nodes in the middle part. The maximum offset is selected by experience. Several points are fitted into a curve by spline interpolation, which is used as the center line. According to the length of the generated center line, the corresponding number of widths is generated, and 1-3 nodes are randomly selected. The generated width is sorted according to the nodes, and the ascending and descending order is staggered to approach the shape of the uneven width of the real warp. At the same time, this method is used to generate the required sharp end of other categories. The generated irregular rectangle is randomly deformed. Random elastic deformation is achieved by randomly generating two matrices with the same size as the image, representing the horizontal and vertical position offset of each pixel point. Use Gaussian filter to smooth it, so that the matrix is continuous, forming a coordinate offset matrix in two directions. The original coordinates of the image are applied to the two different direction offsets, and the pixel points after offset are obtained by interpolation to get the gray value on the original image coordinates, and a new image is reconstructed to achieve the effect of random elastic deformation. In addition, the median filter method is used to remove the burrs generated in the width sorting process. Different types of warp are arranged at certain intervals.

[0102] The specific implementation method of S32 is:

[0103] A basic elliptical shape is generated with the fitted short axis length and the long-short axis ratio as the two shape parameters. Nodes are set on the elliptical circumference with equal length. Then, offset is generated for all nodes according to the fitted weft offset, which is applied to the length of all nodes. Here, 720 nodes are collected during generation, and a special smoothing method is used to take the midpoint of adjacent nodes of the ellipse. After multiple smoothing, the ellipse is subjected to median filtering to eliminate burrs. Finally, the ellipse is tilted according to the fitted tilt angle.

[0104] The specific implementation method of S33 is:

[0105] The original image size is used as the generated image size, and the upper and lower boundaries of the resin area are artificially limited. Warp is generated first, and the warp is arranged in position through the generated spacing until the remaining width in the image cannot accommodate the next generated spacing and the width of the warp. Then, the positions of the two ends of the warp are selected as the horizontal boundaries of the resin area. Then, weft is generated, and the weft is randomly selected for placement. The placement area is expanded upward and downward, and this method is used to generate the shape of the truncated weft on the boundary of the original image. Then, the overlap rate with the fibers already placed is calculated to determine whether to place. Finally, the weft placement fails for a certain number of times or the total area of the fibers in the image reaches a certain level, and the placement is stopped.

[0106] The specific implementation method of S34 is:

[0107] The Pix2PixHD network is a conditional generative adversarial network (cGAN) that uses supervised learning to make the generated image more accurate in terms of gray scale distribution with the input target image. The main purpose is to train a generator G to obtain a fake image G(x) from the original image x through the generator network G, and to train a discriminator D to input the generated fake image G(x) and the real image into the discriminator D. Through training, the features of the original image are similar to those of the generated image, so that the discriminator D cannot distinguish whether the image is fake or not. By modifying the generation part of the Pix2PixHD network, the generation part trains two generator models of different scales at the same time, and the features of the two generator models of different scales are fused. Based on the generator model after feature fusion, the fiber composite material image with the same microstructure as the real one is generated. 120 pictures with better synthesis effect are manually selected from a set of 200 synthesized data to form a synthesized data set, and together with an original data set of 139 pictures, an final data set is formed. We randomly select 15 pictures from the 139 original data sets as a validation set, and the remaining 124 original data sets and 120 synthesized data sets together form a training set.

[0108] S4, based on the Swin-Tranformer network, the XCT image is accurately segmented, and the outline features between the components are clear;

[0109] Further preferably, S4 comprises:

[0110] S41, inputting the mixed data set into the Swin-Tranformer network;

[0111] S42, processing by a block division module, dividing the image into different small blocks according to pixels, and then converting into a one-dimensional feature vector;

[0112] S43, network model training, segmenting the component regions in the image;

[0113] S44, performing regional segmentation on each yarn based on a watershed segmentation algorithm;

[0114] The specific implementation method of S41 is:

[0115] The size of the picture is 1119x1431, which is randomly cropped to 448x448, and the picture is subjected to data enhancement of horizontal flipping, vertical flipping and rotation. Among them, the picture is randomly cropped because the original picture size is large, in order to reduce the calculation amount and adapt to the Swin Transformer, and speed up the training. Effective data enhancement can enhance the robustness of the model and prevent overfitting. Finally, the obtained picture is input into the network for training.

[0116] S42 is specifically implemented as follows:

[0117] The training image in the mixed data set is input into the network, divided into 7*7 pixel blocks according to the size, and then the image is divided into multiple small blocks. Then, the pixels of the small blocks in all channels are stretched into one-dimensional feature vectors. Finally, all one-dimensional feature vectors are combined into a feature vector graph.

[0118] S43 is specifically implemented as follows:

[0119] The feature vector graph is processed through four stages to obtain four feature vector graphs at different scales. Each stage is composed of block merging and Swin-Transformer Block. Swin-Transformer Block includes layer normalization layer, window multi-head self-attention layer, and multi-layer perception layer. Finally, the normalized length-2 class prediction probability is obtained through the softmax function, which corresponds to the pores, resin base, weft and warp in the 4 channels respectively. The channel with the highest prediction probability is defined as the corresponding class, so as to obtain the semantic segmentation result graph of the whole input image, as shown in FIG. 1.

[0120] S44 is specifically implemented as follows:

[0121] Based on the ResUNet+ model semantic segmentation yarn, the XCT image segmentation result is extracted along the warp and weft directions respectively, and the yarn is binarized. The binarization result is used as a label image. Secondly, the distance transformation operation is performed on the binarized image to obtain a distance transformation mapping graph. Finally, based on the mapping graph, the watershed segmentation is performed, the distance transformation maximum value is used as the water injection point, and the binary image of the yarn is used as a mask to limit the segmentation area. Considering that the watershed segmentation effect of the long strip yarn is not ideal, the binary image of the weft yarn needs to be extracted and processed along the weft laying direction. After the segmentation is completed, a small amount of misclassified areas are manually corrected to obtain the instance segmentation result.

[0122] S5, as shown in FIG. 3, based on the segmented XCT image, as shown in FIG. 3(a), the contour microstructure features are extracted and the local contour is smoothed;

[0123] Further preferably, S5 comprises:

[0124] S51, the image is down-sampled by selecting a suitable resolution, so as to retain the microstructure detail features as much as possible and reduce the calculation resource cost and calculation amount;

[0125] S52, the features of the image after down-sampling are screened, a small part of non-interest areas are filtered out, and essential feature areas such as pores and yarns are retained, as shown in FIG. 3(b);

[0126] S53, extract the image contour features of the pores and yarns, and establish a coordinate value conversion based on the positional relationship between the pixels and nodes.

[0127] S54, set an appropriate convolution kernel, and use the operation between the convolution kernel and the extracted contour coordinate values to realize contour smoothing, avoiding stress concentration and error interruption during finite element numerical calculation.

[0128] S55, process all segmented images according to the steps S51-S54, and construct a yarn and pore feature database to store the contour coordinate values of each component after coordinate system conversion.

[0129] S51 is specifically implemented as follows:

[0130] Based on the images of the completed component region segmentation task, the non-interesting area around the periphery is removed by secondary cropping to obtain an image with a length of 1020 pixels and a width of 1350 pixels and a resolution of 4.9819 um. Under the premise of preserving as much microstructure detail as possible, the image is down-sampled along the length and width directions by taking one pixel every 10 pixels, and a 102x135 pixel image with a resolution of 49.819 um is obtained. This reduces the number of pixels that can be processed and reduces the cost of computing resources.

[0131] S52 is specifically implemented as follows:

[0132] Due to the geometric characteristics of the yarn itself, the segmented image inevitably has a high degree of dispersion of multiple fibers in the yarn itself, and the features of a single yarn in a single image are distributed in multiple regions. Therefore, the contourArea function in the OpenCV library is used to filter small areas composed of multiple fibers by setting the threshold for screening to 5. Areas smaller than the fixed threshold will not be considered, and only the yarn area features with larger area will be retained, as shown in FIG. 2.

[0133] S53 is specifically implemented as follows:

[0134] The findContours function in the OpenCV library is used to extract the contour pixel coordinate value information of each yarn and pore, and the conversion is based on the positional relationship between the pixels and nodes. In addition, the coordinate system in the OpenCV library and the coordinate system in Abaqus 2021 have some differences in direction, and the node coordinate values based on the pixel contour coordinate values need to be converted to the coordinate system to ensure that the segmented image and the constructed two-dimensional finite element model have high consistency in microstructure.

[0135] S54 is specifically implemented as follows:

[0136] When considering finite element numerical simulation, the stress concentration at the interface of each component can easily cause the operation to be interrupted. Therefore, the extracted profile feature information of each component needs to be processed again, a suitable convolution kernel is set, and when the numerical value of the local profile coordinate value changes greatly, the profile is smoothed by creating additional nodes.

[0137] The specific implementation method of S55 is:

[0138] The profile coordinate value information of each yarn and pore in each segmented image is obtained along the X-axis direction and the Y-axis direction respectively, the operations of S51-S54 are repeated, a corresponding json file is established for each segmented image to store information in the form of a dictionary in python language, and a feature database is constructed.

[0139] S6, extracting the profile features of each yarn of each segmented image, calculating the coordinate values of the profile centers of the yarns, establishing a spatial coordinate system, and discretely deriving the profile centers of each yarn of each segmented image;

[0140] Further preferably, S6 comprises:

[0141] S61, extracting the coordinate values of the center points of the profiles of each yarn of each segmented image in the feature database, and establishing a spatial coordinate system in the order of the segmented images.

[0142] S62, using a first-order intermediate differential derivation rule to discretely derive the center points of the profiles of each yarn of each segmented image, to obtain the tangent direction of the center points, i.e. the 1 direction of the yarn.

[0143] S63, the 2 direction can be obtained from the calculated 1 direction.

[0144] The specific implementation method of S61 is:

[0145] Considering that the segmented image is converted into a finite element model and a single yarn is transversely isotropic, a local material orientation assignment algorithm is proposed to consider the influence of fiber orientation on mechanical properties. The profile feature coordinate value information of each yarn is extracted from the feature database along the X-axis direction and the Y-axis direction, a spatial coordinate system is established in the order of the segmented images, the distance between adjacent segmented images is assumed to be i*4.9819um, and the profile feature coordinate value information of all yarns is mapped to the constructed spatial coordinate system, as shown in FIG. 4(a).

[0146] The specific implementation method of S62 is:

[0147] Statistical calculation of the center point of the cross section of the i-th yarn Statistical analysis of the warp or weft center point of each cross-sectional image in different directions, and then the discrete derivative of each center point is solved to obtain the 2 direction of each cross-sectional Or The direction is shown in Figs. 4(b) and 4(c).

[0148] where i is the number of the segmented image, v represents the resolution of the XCT image, T is the length of the pixel on the Y-axis or Z-axis, p it is the corresponding gray value of the t pixel on the i-th yarn cross section, and is the corresponding Y-axis and Z-axis coordinates of the t pixel on the i-th yarn cross section, is the corresponding the partial derivative solution on the yarn cross section, and h is the corresponding solution step size.

[0149] S63 is specifically implemented as follows:

[0150] Since the yarn is transversely anisotropic, the principal directions are 1 direction, 2 direction and 3 direction, as long as they are perpendicular to each other, that is, the sum of the products of the corresponding components between each direction is equal to 0. Therefore, the component of the 2 direction on the X-axis is defined as 0, and the 2 direction component can be calculated. At the same time, the discrete derivative value results obtained by calculation are stored in the feature database, as shown in Fig. 4(d).

[0151] S7, extracting feature database data, constructing feature database data of each component region;

[0152] Further preferably, S7 comprises:

[0153] S71, extracting all yarn contour and pore contour coordinate value data in a single segmented image in the feature database.

[0154] S72, sequentially calling the underlying function functions of Abaqus 2021 software based on the python language.

[0155] S73, according to the position characteristics of the internal mesostructure of the material, the reduced component regions are divided, and the grid size is set for grid division.

[0156] S7 is specifically implemented as follows:

[0157] Considering that the composite material only contains resin matrix, warp yarn, weft yarn and pores, a fixed-length and width rectangle is considered as the initial model, and the length and width of the rectangle can be obtained by multiplying the length and width pixel length of the image respectively. In the rectangular area, each area is segmented, the pore area is removed, and the remaining area is classified as the resin matrix area. Taking the 330th segmented image as an example, the yarn contour coordinate value information of the image is created by calling the feature data, the Line function of Abaqus 2021 is used to draw each yarn area and the pore area, then the PartitionFaceBySketch function is used to segment the area, and finally the Cut function is used to remove the pore area. The unit types of the resin matrix and the yarn are set to CPS3 and CPS4R respectively, and the seedPart function of Abaqus 2021 is called to reasonably set the grid seed. The size of the grid density greatly affects the numerical simulation calculation cost and calculation efficiency.

[0158] S8, insert cohesive elements and define material properties for each component region feature, assign local material direction;

[0159] Further preferably, S8 comprises:

[0160] S81, the mesh model divided by S7, insert cohesive elements at the mutual contact interface of yarn and yarn and yarn and matrix, as shown in Fig. 5(a).

[0161] S82, the local material direction vector calculated by S6 is gradually assigned to each element of each yarn.

[0162] S83, define material properties and mechanical response characteristics for each component region.

[0163] S8 is specifically implemented as follows:

[0164] Insert cohesive elements with no thickness at the mutual contact interface of yarn and yarn and yarn and matrix. Accurately define material properties, such as the density, modulus and Poisson's ratio of each component material, and select appropriate material damage constitutive. Then, as shown in Fig. 4, for each yarn in the model, calculate the distance from the centroid of each element to all discrete points on the segmented images according to the corresponding discrete derivative results from the feature database, and select the directional derivative component of the discrete point with the shortest distance as the local material direction attribute of the element, as shown in Fig. 4(c).

[0165] where x ijk , y ijk , z ijk correspond to the X-axis, Y-axis and Z-axis coordinate values of the kth node in the jth element on the ith yarn cross-section, respectively. Corresponding to the X-axis, Y-axis and Z-axis coordinate values of the jth unit centroid on the ith yarn cross section.

[0166] S9, set boundary conditions, numerical simulation calculation, as shown in Figure 5 (b).

[0167] S9 is specifically implemented as follows:

[0168] Select two reference points A and B to be rigidly coupled with the upper movable platform and the lower fixed platform. By applying displacement conditions U1=0, U2=-1 and UR3=0 to reference point A, the sample is simulated to be continuously loaded in the vertical direction. At the same time, set U1=0, U2=0 and UR3=0 on the lower fixed platform to ensure that the sample is subjected to compression effect, as shown in Figure 5.

[0169] S10, repeat steps S7-S9, to realize two-dimensional batch finite element model rapid generation and numerical simulation, as shown in Figure 6.

[0170] Example two

[0171] The application also provides a two-dimensional batch finite element model generation and numerical simulation system based on composite material XCT digital images, comprising: an acquisition module, a modeling module, a learning module, a precise segmentation module, a first database construction module, a second database construction module, a division module, an assignment module, a calculation module and a model generation module.

[0172] The acquisition module is used to acquire and pretreat the XCT images of the composite material in the initial state basalt fiber.

[0173] The modeling module is used to perform parameterized statistical analysis on the pretreated XCT images of the composite material, quantify the mesostructure parameters of the yarn inside the material, and perform two-dimensional parameterized random modeling to obtain a parameterized random modeling image.

[0174] The learning module is used to learn the parameterized random modeling image based on the adversarial learning style transfer network Pix2PixHD to generate a virtual XCT image.

[0175] The precise segmentation module is used to realize precise segmentation of each component region of the virtual XCT image based on the Swin-Tranformer network.

[0176] The first database construction module is used to extract contour microstructure features and locally smooth contours based on the segmented virtual XCT image, and construct a yarn and pore feature database.

[0177] The second database construction module is used for extracting profile features of each yarn of each segmented image based on the yarn and pore feature database, calculating coordinate values of yarn profile centers, establishing a spatial coordinate system, performing discrete derivation on the yarn profile centers of each segmented image based on the spatial coordinate system, and constructing a feature database;

[0178] The division module is used for extracting feature data information of a single segmented image, classifying each component area, and dividing a grid based on the feature database;

[0179] The distribution module is used for inserting cohesive units and defining material properties of each component area feature, and distributing local material directions based on the divided grid.

[0180] The calculation module is used for setting boundary conditions and performing numerical simulation calculation based on the distributed local material directions.

[0181] The model generation module is used for repeatedly performing the division module and the calculation module, and realizing fast generation and numerical simulation of a two-dimensional finite element model based on the composite material XCT digital image.

[0182] The above-described embodiments are only used to describe the preferred modes of the present application, and do not limit the scope of the present application. Without departing from the design spirit of the present application, various modifications and improvements of the technical solutions of the present application made by those skilled in the art shall fall within the protection scope determined by the claims of the present application.

Claims

1. A method for generating and numerically simulating a two-dimensional bulk finite element model based on composite material XCT digital images, characterized in that, The method comprises the following steps: S1: collecting an initial state basalt fiber composite material XCT image and preprocessing; S2: performing parameterized statistical analysis on the preprocessed composite material XCT image, quantifying the microstructure parameters of the yarns inside the material, and performing two-dimensional parameterized random modeling to obtain a parameterized random modeling image; S3: learning the parameterized random modeling image based on the adversarial learning style transfer network Pix2PixHD to generate a virtual XCT image; S4: realizing accurate segmentation of each component region of the virtual XCT image based on the Swin-Tranformer network; S5: based on the segmented virtual XCT image, extracting contour microstructure features and smoothing the local contour to construct a yarn and pore feature database; S6: based on the yarn and pore feature database, extracting the contour feature of each yarn of each segmented image, calculating the coordinate value of the yarn contour center, establishing a spatial coordinate system, and based on the spatial coordinate system, discretely deriving the yarn contour center of each segmented image to construct a feature database; S7: based on the feature database, extracting feature data information of a single segmented image, classifying each component region, and dividing the grid; S8: based on the divided grid, inserting cohesive elements and defining material properties of each component region feature, and assigning local material direction; S9: based on the assigned local material direction, setting boundary conditions and performing numerical simulation calculation; S10: repeating S7-S9 to realize rapid generation and numerical simulation of a two-dimensional finite element model based on the composite material XCT digital image.

2. The method for generating and numerically simulating a two-dimensional bulk finite element model based on composite material XCT digital images according to claim 1, characterized in that, In S1, the method for collecting an initial state basalt fiber composite material XCT image and preprocessing comprises: S11: obtaining complete XCT scanning source image data, and performing cropping and filtering on the XCT scanning source image data; S12: based on the cropped and filtered XCT scanning source image data, extracting a quantitative number of segmented images at equal intervals, and manually labeling each component in each segmented image; S13: performing feature extraction on the manually labeled segmented image to obtain a true label set.

3. The method for generating and numerically simulating a two-dimensional bulk finite element model based on composite material XCT digital images according to claim 2, characterized in that, In S2, the method for performing parameterized statistical analysis on the preprocessed composite material XCT image, quantifying the microstructure parameters of the yarns inside the material, and performing two-dimensional parameterized random modeling comprises: S21: based on the microstructure features of the manually labeled data set, extracting the microstructure parameters of each yarn in each segmented image; S22: based on the microstructure parameters of each yarn, qualitatively and quantitatively expressing the microstructure features, and applying statistical methods for parameterized statistical analysis; S23: based on the results of parameterized statistical analysis, expanding the data set through parameterized random modeling to generate a virtual data set, i.e., a parameterized random modeling image.

4. The method for generating and numerically simulating a two-dimensional bulk finite element model based on composite material XCT digital images according to claim 3, characterized in that, In S3, the method for learning the parameterized random modeling image based on the adversarial learning style transfer network Pix2PixHD to generate a virtual XCT image comprises: S31: generating a warp yarn random model according to warp yarn parameterized statistical analysis; S32: generating a weft yarn random model according to weft yarn parameterized statistical analysis; S33: generating an artificially synthesized label image based on the warp yarn random model and the weft yarn random model; S34: based on the artificially synthesized label image, applying an improved Pix2PixHD network to generate basalt fiber composite material images identical to the real microstructure, and constructing a hybrid dataset, i.e., a virtual XCT image.

5. The method for generating and numerically simulating a two-dimensional bulk finite element model based on composite material XCT digital images according to claim 4, characterized in that, In the S4, the method for accurately segmenting each component region of the virtual XCT image based on the Swin-Tranformer network comprises: S41: inputting the hybrid dataset into the Swin-Tranformer network to obtain a training image; S42: processing the training image by a block division module to divide the training image into different small blocks according to pixels, and then converting the training image into a one-dimensional feature vector graph; S43: accurately segmenting each component region in the one-dimensional feature vector graph to obtain a semantic segmentation result graph; S44: based on the semantic segmentation result graph, using a watershed segmentation algorithm to complete the region segmentation of each yarn.

6. The method for generating and numerically simulating a two-dimensional bulk finite element model based on composite material XCT digital images according to claim 1, characterized in that, In the S5, the method for extracting contour microstructure features and smoothing local contours based on the segmented virtual XCT image comprises: S51: selecting a resolution satisfying a preset requirement to downsample the segmented virtual XCT image; S52: performing feature screening on the downsampled virtual XCT image; S53: based on the virtual XCT image after feature screening, extracting image contour features of pores and yarns, and performing coordinate value conversion according to the positional relationship between pixels and nodes to obtain contour coordinate values; S54: setting a convolution kernel satisfying a preset requirement, and using the operation between the convolution kernel and the contour coordinate values to realize contour smoothing; S55: processing all segmented images according to S51-S54 to construct a yarn and pore feature database, and storing the contour coordinate values of each component after coordinate system conversion.

7. The method for generating and numerically simulating a two-dimensional bulk finite element model based on composite material XCT digital images according to claim 6, characterized in that, In the S6, based on the yarn and pore feature database, the method for extracting contour features of each yarn of each segmented image, calculating the coordinate values of the contour centers of the yarns, establishing a spatial coordinate system, and performing discrete differentiation on the contour centers of each yarn of each segmented image based on the spatial coordinate system comprises: S61: extracting the coordinate values of the center points of the contours of each yarn of each segmented image in the yarn and pore feature database to establish a spatial coordinate system in the order of the segmented images; S62: based on the spatial coordinate system, using a first-order intermediate differential differentiation rule to perform discrete differentiation on the center points of the contours of each yarn of each segmented image to obtain the tangent direction of the center points, i.e., the 1 direction of the yarns; S63: obtaining the 2 direction from the calculated 1 direction.

8. The method for generating and numerically simulating a two-dimensional bulk finite element model based on composite material XCT digital images according to claim 7, characterized in that, In the S7, based on the feature database, the method for extracting feature data information of a single segmented image, classifying each component region, and dividing a grid comprises: S71: extracting all yarn contour and pore contour coordinate value data in a single segmented image in the feature database; S72: sequentially calling the underlying function functions of the Abaqus 2021 software based on the python language; S73: Based on all yarn profile and pore profile coordinate value data and the bottom layer function function, the reduced component area is divided according to the position characteristics of the internal microstructure of the material, and the grid size is set for grid division.

9. The method for generating and numerically simulating a two-dimensional bulk finite element model based on composite material XCT digital images according to claim 8, characterized in that, The method for inserting cohesive elements and defining material properties for each component area based on the divided grid in S8 includes: S81: Insert cohesive elements at the mutual contact interface of yarn and yarn and yarn and matrix from the completed grid division; S82: The local material direction vector calculated is gradually distributed to each element of each yarn; S83: Define material properties and mechanical response characteristics for each component area.

10. A two-dimensional bulk finite element model generation and numerical simulation system based on composite material XCT digital images, characterized in that, It includes: The acquisition module, the modeling module, the learning module, the accurate segmentation module, the first database construction module, the second database construction module, the division module, the distribution module, the calculation module and the model generation module; The acquisition module is used for collecting the XCT image of the initial state basalt fiber composite material and pre-processing; The modeling module is used for parameterized statistical analysis of the pre-processed composite material XCT image, quantification of the internal yarn microstructure parameters of the material and two-dimensional parameterized random modeling, and obtaining of a parameterized random modeling image; The learning module is used for learning the parameterized random modeling image based on the adversarial learning style transfer network Pix2PixHD to generate a virtual XCT image; The accurate segmentation module is used for realizing accurate segmentation of each component area of the virtual XCT image based on the Swin-Tranformer network; The first database construction module is used for extracting profile microstructure features and smoothing local profiles based on the segmented virtual XCT image, and constructing a yarn and pore feature database; The second database construction module is used for extracting profile features of each yarn of each segmented image based on the yarn and pore feature database, calculating the coordinate values of the yarn profile center, establishing a spatial coordinate system, and based on the spatial coordinate system, discretely deriving the yarn profile center of each segmented image to construct a feature database; The division module is used for extracting feature data information of a single segmented image based on the feature database, classifying each component area, and dividing the grid; The distribution module is used for inserting cohesive elements and defining material properties for each component area based on the divided grid, and distributing local material directions; The calculation module is used for setting boundary conditions based on the distributed local material directions and numerical simulation calculation; The model generation module is used for repeating the division module and the calculation module to realize the rapid generation and numerical simulation of a two-dimensional finite element model based on the composite material XCT digital image.

Images