A method for generating two-dimensional random pores in composite materials based on XCT images

Abstract

The present invention relates to a method for generating two-dimensional random pores in composite materials based on XCT images, comprising the following steps: collecting and preprocessing XCT images of the initial state of the composite material, and completing manual annotation of pores in the source images; quantitatively analyzing the pore regions of all images according to characteristic parameters; randomly determining the circumscribed rectangles and intersection points of the pores based on the results of the quantitative analysis; generating random inflection points between the circumscribed intersection points and generating equally spaced points between adjacent points to achieve the generation of a multi-arc splicing contour; after smoothing the sharp inflection points using a polynomial fitting function, filling the internal region of the pore contour. The method for generating two-dimensional random pores in composite materials based on XCT images of the present invention can randomly generate more complex and variable pore contours, and can study the influence of pore structure parameters on the mechanical properties of materials at the mesoscopic scale.

Description

Technical Field

[0001] The present invention relates to a method for generating pore profiles, specifically to a method for generating two-dimensional random pores of composite materials based on XCT images, belonging to the technical field of computer vision. Background Technique

[0002] Basalt Fiber Reinforced Polymer (BFRP) is an advanced composite material system prepared by a vacuum-assisted resin transfer molding (VARTM) process with basalt long fibers as the reinforcement phase. With its low-density characteristics and the advantage of adjustable anisotropy, the industrial application of this material system has important strategic value for promoting the innovation of advanced manufacturing technologies and realizing the lightweight upgrade of equipment.

[0003] X-ray Computed Tomography (XCT) is a non-destructive three-dimensional imaging technology based on the principle of X-ray transmission. Through the reconstruction of multi-angle projection data, it can achieve high-spatial-resolution characterization of the internal structure of materials. In the research of basalt fiber composites, XCT technology can non-destructively obtain full-scale mesoscopic structure information from the fiber bundle arrangement pattern to the pore defect distribution, breaking through the dimensional limitations of traditional destructive testing methods. Through three-dimensional volume data reconstruction and image segmentation algorithms, the fiber-matrix interface characteristics and pore network topological parameters can be quantitatively analyzed, providing a high-precision digital image basis for subsequent multi-scale structure modeling and performance correlation analysis. This technical system provides a key experimental characterization means for the research on the correlation between the mesoscopic structure and performance of composite materials.

[0004] The ideal geometric modeling method mainly constructs the morphological characteristics of yarns in a parametric form. However, the generated geometric models often show an obvious theoretical tendency, specifically manifested as the over-simplification of the deformation mode and the mechanical repetition of local topological features. This limitation stems from the mathematical constraints on the yarn deformation mechanism in the parametric modeling process - usually using ideal functions to describe the macroscopic shape of the yarn, but failing to effectively characterize the key physical mechanisms such as local compressive deformation caused by yarn-yarn / yarn-matrix interactions during the manufacturing process. To break through this theoretical bottleneck, the geometric modeling method based on real physical mechanisms has gradually become the mainstream research paradigm, and its technical path is mainly divided into two categories: one is to rely on XCT digital image processing technology to achieve the reverse reconstruction of the yarn mesoscopic structure, and the other is to forwardly predict the complex deformation behavior of the yarn during the forming process through multi-physics field coupling numerical simulation technology.

[0005] Although the existing research system has formed a relatively complete analysis framework for the characterization of the deformation mechanism of yarns, it generally lacks a systematic consideration of the morphological characteristics of pores. It should be particularly noted that the non-linear response and damage evolution path of the macroscopic mechanical properties of composites are essentially regulated by the synergistic action of the geometric morphology, volume fraction, and spatial distribution characteristics of pores. There are two limitations in the current pore characterization research: First, most of the existing results focus on the analysis of pores at the microscale, and there is a significant lack of research on the cross-scale coupling mechanism between pores at the mesoscale and the yarn / matrix interface interaction. Second, conventional modeling methods often simplify pores as regular geometric bodies (such as rectangles, triangles, or circles), and this idealized treatment cannot characterize the asymmetric morphological characteristics and random distribution characteristics of pores generated by yarn extrusion during actual processes (such as during VARTM molding). Restricted by the above deficiencies in characterization techniques, there is still a lack of systematic understanding of the quantitative mapping relationship between mesoscale pore structure parameters and the macroscopic mechanical response of materials.

[0006] Therefore, to solve the above problems, it is indeed necessary to provide an innovative two-dimensional random pore generation method for composites based on XCT images to overcome the defects in the prior art. Summary of the Invention

[0007] The purpose of the present invention is to provide a two-dimensional random pore generation method for composites based on XCT images, which can randomly generate more complex and variable pore contours and can study the influence of pore structure parameters on the mechanical properties of materials at the mesoscale.

[0008] To achieve the above purpose, the technical solution adopted by the present invention is: A two-dimensional random pore generation method for composites based on XCT images, which includes the following technological steps:

[0009] 1), Collect XCT images of the initial state of the composite material and perform preprocessing, and complete the manual annotation of pores in the source image;

[0010] 2), According to the characteristic parameters, perform quantitative analysis on the pore regions of all images;

[0011] 3), Based on the results of quantitative analysis, randomly determine the circumscribed rectangle and intersection points of the pores;

[0012] 4), Generate random inflection points between the circumscribed intersection points, and generate equally spaced points between adjacent points to achieve the generation of a multi-arc splicing contour;

[0013] 5), After smoothing the sharp inflection points using a polynomial fitting function, fill the internal region of the pore contour.

[0014] The two-dimensional random pore generation method for composites based on XCT images of the present invention is further: The composite material is specifically a biaxial non-crimp warp knitted composite material.

[0015] The two-dimensional random pore generation method of composite materials based on XCT images of the present invention is further as follows: Specifically, step 1) is as follows:

[0016] 1-1), Obtain complete XCT source image data, and crop and filter the XCT source image data.

[0017] 1-2), Based on the cropped and filtered XCT source image data, extract one image every ten images, and manually label the characteristic regions such as warp yarns, weft yarns, matrix, and pores in the image with different colors. The same characteristic connected region is smeared with the same gray value.

[0018] 1-3), Use Boolean operations in the OpenCV library to extract the manually labeled connected regions from the images with labels completed, and establish a label data set.

[0019] The two-dimensional random pore generation method of composite materials based on XCT images of the present invention is further as follows: Specifically, step 2) is as follows:

[0020] 2-1), Use the width W and height H of the circumscribed rectangle of the pore as characteristic parameters to determine specific pore characteristics.

[0021] 2-2), Traverse the pore regions in each picture, calculate its circumscribed width W and height H and record them. Use the normal distribution to fit the widths W and heights H of all pores, and finally obtain the quantitative analysis results.

[0022] The two-dimensional random pore generation method of composite materials based on XCT images of the present invention is further as follows: Specifically, step 3) is as follows:

[0023] 3-1), According to the quantitative analysis results, randomly select the width W and height H of the circumscribed rectangle to determine the geometric characteristics of the pores.

[0024] 3-2), Locate the rectangle in the Cartesian coordinate system, and randomly select four intersection points on the four sides of the circumscribed rectangle, denoted as Ip1(x i1 , y i1 ), Ip2(x i2 , y i2 ), Ip3(x i3 , y i3 ), and Ip4(x i4 , y i4 ).

[0025] The two-dimensional random pore generation method of composite materials based on XCT images of the present invention is further as follows: Specifically, step 4) is as follows:

[0026] 4-1), Four rectangular regions are formed between adjacent circumscribed intersection points.

[0027] 4 - 2), re - define the ranges of the four rectangular regions respectively;

[0028] 4 - 3), randomly generate multiple inflection points within the defined rectangular region ranges;

[0029] 4 - 4), determine the types of arcs generated between inflection points and circumscribed intersection points or between circumscribed intersection points;

[0030] 4 - 5), generate equally - spaced points between adjacent points;

[0031] 4 - 6), complete the generation of the multi - arc splicing contour: repeat steps 4 - 2) to 4 - 5) until all arc equidistant points are calculated. The arc equidistant points and circumscribed points together constitute the outer contour features of the pores.

[0032] The two - dimensional random pore generation method for composite materials based on XCT images of the present invention is further as follows: In the step 3 - 1), for two adjacent intersection points (x in , y in ) and (x in+1， y in+1 ), they form a rectangle. The length and width of the rectangle are the differences in abscissas and ordinates between adjacent intersection points. The coordinates of the randomly generated inflection points are accurately defined by the following formula:

[0033] min(x in , x in+1 )≤x≤max(x in ,x in+1 )

[0034] min(y in , y in+1 )≤y≤max(y in ,y in+1 )

[0035] Among them, the coordinate range of x is from the minimum value to the maximum value between x in to x in+1 , and the coordinate range of y is from the minimum value to the maximum value between y in to y in+1 .

[0036] The two - dimensional random pore generation method for composite materials based on XCT images of the present invention is further as follows: In the step 3 - 3), the range for randomly generating inflection points follows the following formula:

[0037] min(x in , x in+1 ) (1 - a)≤x≤max(x in ,x in+1 )(1 - b), 0 < a < 1, 0 < b < 1, a > b

[0038] min(y in , y in+1 ) (1 - c) ≤ y ≤ max(y in , y in+1 ) (1 - d), 0 < c < 1, 0 < d < 1, c > d;

[0039] Among them, a, b, c, and d are coefficients determined to prevent excessive bending mutations at the inflection points.

[0040] The two-dimensional random pore generation method for composite materials based on XCT images of the present invention is also as follows: The specific step 5) is as follows:

[0041] 5-1), Smooth the sharp inflection points at the connections between regions: Use the polynomial curve fitting algorithm to globally continuously reconstruct the discrete feature point set, eliminate the contour mutation points by establishing high-order differential continuity constraints, and finally form a smooth closed boundary that conforms to the true pore geometric characteristics.

[0042] 5-2), Fill the internal region of the pore contour: Store the coordinate values of all the pore contour points, and perform pixel-level filling operations on the topologically closed region of the pore contour geometry based on the polygon filling algorithm cv2.fillPoly of the OpenCV computer vision library.

[0043] Compared with the prior art, the present invention has the following beneficial effects:

[0044] 1. The two-dimensional random pore generation method for composite materials based on XCT images of the present invention deeply studies the mesoscopic structure of biaxial non-crimp warp knitted composite materials based on XCT technology. By statistically analyzing the fine structural characteristics of yarns and pores, the mesoscopic structure of the material can be accurately described, and the influence of factors such as pore content, characteristic morphology, and distribution position on the mechanical properties of the micro-structure can be analyzed, providing a scientific basis for further optimizing the preparation process and application of BNCFRP.

[0045] 2. Compared with the traditional pore modeling methods using regular geometric shapes (circles, ellipses, polygons, etc.), the present invention effectively improves the morphological fidelity and spatial heterogeneous distribution characteristics of the generated pores by integrating the quantitative statistical characteristics of real pores; the generated synthetic pore images can not only provide data enhancement support for the initial defect semantic segmentation model based on deep learning, but also provide a high-fidelity initial defect model for the composite material damage evolution simulation, which helps to deeply study the prediction of the propagation path of pore defects and the analysis of the material failure mechanism. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1It is a flowchart of the method for generating two-dimensional random pores of composite materials based on XCT images according to the present invention.

[0047] Figure 2 It is a schematic diagram after manual annotation of the XCT image in step 1) of the present invention.

[0048] Figure 3 It is a schematic diagram for qualitatively describing the pore characteristics of the present invention.

[0049] Figure 4 It is a schematic diagram after quantitative statistical fitting of the pores in the XCT image in step 2) of the present invention.

[0050] Figure 5 It is a flowchart of step 4) of the present invention.

[0051] Figure 6 It is a result display diagram of the pores having different EI indices (aspect ratios) in the embodiment of the present invention to verify the diversity of the pore generation results of this algorithm. Detailed implementation manners

[0052] Please refer to the attached Figure 1 As shown, the present invention is a method for generating two-dimensional random pores of composite materials based on XCT images, which includes the following process steps:

[0053] 1), Collect the initial state XCT image of the composite material and perform preprocessing, and complete the manual annotation of the pores in the source image.

[0054] In this embodiment, the composite material is specifically biaxial non-crimp fabric-reinforced polymer composites (BNCFRP).

[0055] This step is specifically as follows:

[0056] 1-1), Obtain the complete XCT source image data, and perform cropping and filtering on the XCT source image data. Specifically, use the XCT device to scan the composite material sample to obtain the complete XCT source data, and then import it into the Avizo software for edge cropping, and use the mean, median, and non-local mean filters to reduce noise and ensure the quality of manual segmentation.

[0057] 1-2), Based on the cropped and filtered XCT source image data, extract one image every ten images, and manually annotate the characteristic regions such as warp yarns, weft yarns, matrix, and pores in the image with different colors. The same characteristic connected region is smeared with the same gray value. Finally, 138 images with a pixel length of 1431×1119 are obtained, and the annotation results are as attached Figure 2as shown in (a) and (b) thereof.

[0058] 1-3), for the images with annotations completed, use Boolean operations in the OpenCV library to extract manually annotated connected regions and establish a label dataset.

[0059] 2), according to the characteristic parameters, conduct quantitative analysis on the pore regions of all images.

[0060] This step is specifically as follows:

[0061] 2-1), use the width W and height H of the circumscribed rectangle of the pore as characteristic parameters to determine specific pore characteristics.

[0062] As shown in the appendix Figure 3 As shown, the pore characteristics are composed of its contour and the enclosed pore area. The circumscribed rectangle with width W and height H can describe its main geometric characteristics. The pore contour must intersect the rectangle at four intersection points, thereby dividing the contour into four non-linear curve segments. In addition, between adjacent intersection points, there is at least one inflection point, further dividing the non-linear curve segments into shorter non-linear curve segments, and these short curve segments have two states of concave and convex. In addition, these shorter non-linear curve segments can be accurately approximated by a sufficient number of equally spaced points.

[0063] 2-2), traverse the pore regions in each picture in the label dataset, calculate its circumscribed width W and height H and record them. Use the normal distribution to fit the width W and height H of all pores, and finally obtain the quantitative analysis result. If the fitting value is higher, it indicates that the data fitting is better, indicating that the proportion of the variability in the dependent variable that can be explained by the fitting model is higher. The fitting result is as shown in (a) and (b) of the appendix Figure 4 as shown in (a) and (b) thereof.

[0064]

[0065] Among them, is the fitting function, which is used to fit the statistical data of the widths W and heights H of all pores in the XCT image of the composite material. After fitting, this function can be used to generate pseudo-pores in combination with the pore generation algorithm; e represents the natural constant, y 0, λ, μ, σ are constants, and their values are determined by the fitting results of the fitting function for the statistical data of the circumscribed width W and height H.

[0066]

[0067] 3), based on the quantitative analysis result, randomly determine the circumscribed rectangle and intersection points of the pore. The specific process is as follows:

[0068] 3-1), randomly select the width W and height H of the circumscribed rectangle according to the quantitative analysis results in step 2) to determine the geometric characteristics of the pores.

[0069] 3-2), position the rectangle in the Cartesian coordinate system, and randomly select four intersection points on the four sides of the circumscribed rectangle, denoted as Ip1(x i1 , y i1 ), Ip2(x i2 , y i2 ), Ip3(x i3 , y i3 ), and Ip4(x i4 , y i4 ), as shown in (a) of the appendix Figure 5 .

[0070] 4), generate random inflection points between the circumscribed intersection points and equally spaced points between adjacent points to realize the generation of a multi-arc splicing contour.

[0071] This step is specifically as follows:

[0072] 4-1), four rectangular regions are formed between adjacent circumscribed intersection points, that is, two adjacent intersection points (x in , y in ) and (x in+1， y in+1 ) form a rectangle, and the length and width of the rectangle are the differences in abscissa and ordinate between adjacent intersection points. The coordinates of the randomly generated inflection points can be accurately defined by the following formula:

[0073] min(x in , x in+1 ) ≤ x ≤ max(x in , x in+1 )

[0074] min(y in , y in+1 ) ≤ y ≤ max(y in , y in+1 )

[0075] Among them, the coordinate range of x is from the minimum to the maximum value between x in and x in+1 , and the coordinate range of y is from the minimum to the maximum value between y in and y in+1 .

[0076] 4-2), redefine the ranges of the four rectangular regions respectively to avoid regional interference.

[0077] Randomly generating inflection points within the rectangular area may result in interference areas. For example, the ordinates or abscissas of the intersection points and inflection points are parallel, leading to unreasonable deformation of the contour. Therefore, it is necessary to limit the range of randomly generated inflection points, following the formula:

[0078] min(x in , x in+1 ) (1 - a) ≤ x ≤ max(x in , x in+1 )(1 - b), 0 < a < 1, 0 < b < 1, a > b

[0079] min(y in , y in+1 ) (1 - c) ≤ y ≤ max(y in , y in+1 ) (1 - d), 0 < c < 1, 0 < d < 1, c > d;

[0080] Among them, a, b, c, and d are coefficients determined to prevent excessive bending mutations of the inflection points.

[0081] 4 - 3), Randomly generate multiple inflection points within the limited rectangular area range.

[0082] In step 4 - 2), the normal range where inflection points can be generated is further restricted, and a single inflection point is randomly generated within the limited rectangular range. If a more complex contour is required to be generated, more inflection points can be generated within the determined range according to the formula in step 4 - 2). In the present invention, the maximum number of inflection points randomly generated between adjacent circumscribed intersection points is 1. The generated inflection points Fp1(x f1 , y f1 ) and Fp2(x f2 , y f2 ) are as shown in (b) of the appendix Figure 5 .

[0083] 4 - 4), Determine the type of arc generated between the inflection point and the circumscribed intersection point or between the circumscribed intersection points.

[0084] As shown in (c) of the appendix Figure 5 , if a convex arc is to be generated between Ip1 and Fp1, the convex arc should be on the right side of the vector , connect the line segment Ip1Fp1, and draw a perpendicular bisector on this line segment, and calculate the angles between the line segment Ip1Fp1 and the adjacent two straight lines. To control the generated arc from exceeding the circumscribed rectangle range, select the adjacent straight line with the smallest angle to draw a perpendicular line, and the perpendicular line and the perpendicular bisector will intersect at point O1(x o1 , y o1 ). The radius of this convex arc can be calculated by the following formula:

[0085]

[0086] Similarly, if a concave circular arc is to be generated between Fp1 and Ip2, the concave circular arc should be on the left side of the vector . Connect the line segment Fp1Ip2, and draw a perpendicular bisector on this line segment. Calculate the angles between the line segment Fp1Ip2 and the adjacent two straight lines. To control the generated circular arc from exceeding the circumscribed rectangle range, select the adjacent straight line with the smallest angle to draw a perpendicular line, and the perpendicular line and the perpendicular bisector will intersect at point O2(x o2 , y o2 ). The radius of the concave circular arc can be calculated by the following formula:

[0087]

[0088] Finally, translate through O1(x o1 , y o1 ) and O2(x o2 , y o2 ) on the corresponding perpendicular bisectors to control the smoothness of the generated circular arc, as shown in (d) of the appendix Figure 5 .

[0089] 4-5), generate equally spaced points between adjacent points.

[0090] Taking a maximum of three equally divided points as an example in the present invention, as shown in (e) of the appendix Figure 5 , use the translated O1'(x o1 ', y o1 ') and O2'(x o2 ', y o2 ') to establish a local coordinate system, and respectively obtain the equally divided angles as , , and Calculate all equally divided points E spv (x esv , y esv ) of the circular arcs respectively based on the following formulas:

[0091] .

[0092] 4-6), complete the generation of the multi-circular arc splicing contour: repeat steps 4-2) to 4-5) until all equally divided points of the circular arcs are calculated. The equally divided points of the circular arcs and the circumscribed points together constitute the outer contour feature of the pore.

[0093] 5), use the polynomial fitting function to smooth the sharp inflection points, and then fill the inner region of the pore contour.

[0094] Specifically, this step is as follows:

[0095] 5-1), smooth the sharp inflection points at the junctions between regions.

[0096] In the process of characterizing the pore contour based on equally spaced discrete points, the probabilistically occurring local curvature anomalies may lead to the geometric discontinuity features of adjacent contour segments, thus forming sharp distortions that do not conform to the actual pore morphology. To solve this problem, the present invention uses a polynomial curve fitting algorithm to globally continuous reconstruct the discrete feature point set, eliminates the contour mutation points by establishing high-order differential continuity constraints, and finally forms a smooth closed boundary that conforms to the true geometric features of the pores. The morphological simulation effects are shown in (f) and (g) of the appendix. Figure 5 This method significantly improves the physical rationality of the generated contour while maintaining the original topological distribution law.

[0097] 5-2), fill the internal region of the pore contour.

[0098] Store the coordinate values of all pore contour points, and based on the polygon filling algorithm cv2.fillPoly in the OpenCV computer vision library, perform pixel-level filling operations on the topological closed region of the pore contour geometry. The implementation effect is shown in (h) of the appendix. Figure 5 as shown in (h) of the appendix.

[0099] To verify the regulation ability of the algorithm on the pore morphology characteristics, this study is based on the EI parametric modeling framework, and guides the geometric constraint conditions in the pore generation process by setting different aspect ratio thresholds. As shown in (a) to (e) of the appendix, Figure 6 the experimental results show that this method can effectively generate pore structures with multi-morphology characteristics, and the visual attributes such as the contour curvature distribution and branch complexity all show significant diversity characteristics. Through morphological verification and comparative analysis, it is confirmed that the generated pores can better reproduce the defect distribution law of real materials at both the global topological structure and local geometric detail levels, indicating that this algorithm has good engineering applicability in the modeling of pore heterogeneity characteristics.

[0100] The above specific implementation manners are only the preferred embodiments of this creation, and are not used to limit this creation. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of this creation shall be included within the protection scope of this creation.

Claims

1. A method for generating two-dimensional random pores in composite materials based on XCT images, characterized in that: The process steps include: 1) Collect the initial state XCT image of the composite material and preprocess it, and complete the manual annotation of the pores in the source image; specifically: 1-1), obtaining complete XCT source image data, and performing cropping and filtering on the XCT source image data; 1-2), based on the cropped and filtered XCT source image data, one image is extracted every ten images, and the characteristic areas such as warp, weft, matrix, and pores in the image are manually marked with different colors, and the connected areas with the same characteristics are painted with the same gray value; 1-3), use Boolean operations in the OpenCV library to extract manually annotated connected areas from the annotated images and establish a label data set; 2) According to the characteristic parameters, quantitative analysis is performed on the pore areas of all images; specifically: 2-1), using the width of the circumscribed rectangle of the aperture W and high H As characteristic parameters, specific pore characteristics are determined; 2-2), traverse the pore area in each image, calculate its circumscribed width W and height H and record them, use normal distribution to fit the width W and height H of all pores, and finally obtain the quantitative analysis results; 3) Based on the quantitative analysis results, randomly determine the pore's circumscribed rectangle and intersection points; 4) Generate random inflection points between circumscribed intersection points, and generate equally spaced points between adjacent points to achieve multi-arc splicing contour generation; specifically: 4-1), four rectangular areas are formed between adjacent circumscribed intersection points; 4-2), redefine the ranges of the four rectangular areas respectively; 4-3), randomly generate multiple inflection points within the limited rectangular area; 4-4), determine the type of arc generated between the inflection point and the circumscribed intersection point or between the circumscribed intersection points; 4-5), generate equally spaced points between adjacent points; 4-6) Complete the generation of multi-arc splicing contour: Repeat steps 4-2) to 4-5) until all arc dividing points are calculated. The arc dividing points and the external connection points together constitute the outer contour features of the pore; 5) After smoothing the sharp inflection points using a polynomial fitting function, fill the inner area of ​​the pore contour; specifically: 5-1) Smooth the sharp turning points at the connection between regions: Use the polynomial curve fitting algorithm to globally reconstruct the discrete feature point set, eliminate the contour mutation points by establishing high-order differential continuity constraints, and finally form a smooth closed boundary that conforms to the actual pore geometry characteristics; 5-2) Fill the inner area of ​​the pore contour: store the coordinate values ​​of all pore contour points, and perform pixel-level filling operations on the topological closed area of ​​the pore contour geometric structure based on the polygon filling algorithm cv2.fillPoly of the OpenCV computer vision library.

2. The method for generating two-dimensional random pores in composite materials based on XCT images according to claim 1, characterized in that: The composite material is specifically a biaxial non-crimp warp knitted composite material.

3. The method for generating two-dimensional random pores in composite materials based on XCT images according to claim 1, characterized in that: The step 3) is specifically as follows: 3-1), randomly select the width W and height H of the circumscribed rectangle from the quantitative analysis results to determine the geometric characteristics of the pores; 3-2), locate the rectangle in the Cartesian coordinate system, and randomly select four intersection points on the four sides of the circumscribed rectangle, set as Ip 1 ( x i1 ，y i1 ), Ip 2 ( x i2 ，y i2 ), Ip 3 ( x i3 ，y i3 )and Ip 4 ( x i4 ，y i4 ).

4. The method for generating two-dimensional random pores in composite materials based on XCT images according to claim 1, characterized in that: In step 4-1), two adjacent intersection points ( x in , y in )and( x in+1， y in+1 ) forms a rectangle, the length and width of the rectangle are the difference between the horizontal coordinates and the vertical coordinates of adjacent intersection points, and the coordinates of the randomly generated inflection points are accurately defined by the following formula: min ( x in , x in+1 ) ≤x≤max ( x in ,x in+1 ) min ( y in , y in+1 ) ≤y≤max ( y in ,y in+1 ) in, x The coordinate range is from x in arrive x in+1 From the minimum to the maximum value, y The coordinate range is from y in arrive y in+1 The minimum to maximum value in .

5. The method for generating two-dimensional random pores in composite materials based on XCT images according to claim 1, characterized in that: In step 4-3), the range of randomly generated inflection points follows the following formula: min ( x in , x in+1 )(1- a ) ≤x≤max ( x in ,x in+1 )(1- b ),0 <a< 1, 0 <b< 1, a > b min ( y in , y in+1 )(1- c ) ≤y≤max ( y in ,y in+1 ) (1- d ),0 <c< 1, 0 <d< 1, c > d ； in ,a, b, c, d A coefficient determined to prevent excessive bending and mutation of the inflection point.

Images