Woven CMCs finite element modeling and analysis method based on real microstructure

Abstract

The invention discloses a finite element modeling and analysis method for weaving CMCs based on a real microstructure, and belongs to the technical field of mechanical simulation of composite materials. The method comprises the following steps: acquiring a three-dimensional gray level image of the woven CMCs through X-ray CT (Computed Tomography) scanning; performing high-precision automatic segmentation on multiple components such as warp yarns, weft yarns, matrixes, pores and cracks in the image by using a deep learning technology; boundary smoothing filtering processing is carried out on a segmentation result, and a representative sub-volume region is selected according to a volume fraction consistency principle; based on the real geometric structure of the region, generating an optimized mesh model suitable for finite element calculation; and finally, giving material attributes and setting boundary conditions to carry out simulation analysis so as to accurately predict the mechanical properties of the material. The problems that a traditional ideal model cannot reflect a real microstructure, CT image segmentation is difficult, and the calculation scale is too large are solved, and the precision and efficiency of knitting CMCs finite element simulation are remarkably improved.

Description

Technical Field

[0001] This method belongs to the field of image processing and composite material finite element modeling. Specifically, it involves a finite element modeling and analysis method for braided CMCs based on real microstructure. Background Technology

[0002] Ceramic matrix composites (CMCs) are the preferred materials for hot-end components and thermal protection components of aero-engines due to their excellent mechanical properties. As an anisotropic material, the mechanical properties and damage behavior of CMCs depend not only on the material properties of each component but also on their microstructure. Therefore, accurately predicting the mechanical properties and remaining life of CMCs under the coupling of multiple factors is extremely challenging.

[0003] The Finite Element Method (FEM) is one of the important tools for studying the mechanical properties of CMCs. In the finite element analysis of braided CMCs, the Representative Volume Element (RVE) method is a commonly used technique, which uses a periodic element model for simulation analysis. However, the RVE method constructs an idealized model, which deviates from the actual microstructure of CMCs, leading to significantly increased errors in predicting the nonlinear deformation stage and the final failure mode. Furthermore, the idealized RVE model cannot effectively describe the local stress concentration effects caused by the manufacturing process, which are the main influencing factors for crack initiation and propagation within CMCs. In addition, in braided CMCs, the RVE element model struggles to accurately reconstruct the three-dimensional bending path of the yarn and the contact area formed by their mutual compression, directly affecting the accuracy of the finite element simulation.

[0004] In contrast, high-fidelity finite element models based on the true microstructure of CMCs can significantly improve the accuracy of simulation analysis. Furthermore, high-fidelity finite element models are built upon the true microstructure of woven CMCs, greatly enhancing the accuracy of finite element simulation analysis of woven CMCs.

[0005] The creation of high-fidelity woven CMCs requires accurate acquisition of the material's internal microstructure. However, 3D digital grayscale images of woven CMCs obtained from CT scans suffer from noise, artifacts, and grayscale overlap, further complicating the accurate identification of each component. Furthermore, the finite element model directly constructed from voxels of accurately identified woven CMCs results in an extremely high mesh count, far exceeding conventional computational capabilities. Therefore, achieving high-fidelity surface reconstruction and establishing high-fidelity finite element models of woven CMCs without compromising geometric fidelity is crucial.

[0006] Therefore, a method for establishing a high-fidelity finite element model for braided CMCs is provided. This method enables accurate calculation and simulation of the mechanical properties and failure behavior of braided CMCs, thereby providing solid support for the reliability and safety of braided CMCs in use. Summary of the Invention

[0007] This invention addresses the shortcomings of existing technologies by providing a finite element modeling and analysis method for woven CMCs based on real microstructure.

[0008] To achieve the above-mentioned technical objectives, the technical solution adopted by the present invention is as follows:

[0009] A finite element modeling and analysis method for braided CMCs based on real microstructure, characterized by the following steps:

[0010] Step S1: Obtain a three-dimensional digital grayscale image of the braided CMCs sample;

[0011] Step S2: Crops out sub-region images from the three-dimensional digital grayscale image, and manually annotates each material component in the sub-region image to form training data;

[0012] Step S3: Based on the training data, train a deep learning segmentation model to identify the material composition;

[0013] Step S4: Using the trained deep learning segmentation model, perform component segmentation on the complete three-dimensional digital grayscale image to obtain the three-dimensional spatial distribution data of each component;

[0014] Step S5: Post-process the segmented three-dimensional spatial distribution data and select a sub-volume that is representative of the material composition as the finite element modeling object;

[0015] Step S6: Based on the actual mesoscopic structure data of the sub-volume, generate the corresponding finite element mesh model;

[0016] Step S7: Import the finite element mesh model into the finite element analysis software, assign material properties, set boundary conditions, and perform simulation calculations to analyze the mechanical properties of the braided CMCs.

[0017] To optimize the above technical solution, the specific measures also include:

[0018] The specific steps of step S1 are as follows: scan the braided CMCs using X-ray computed tomography (CT) technology to obtain a three-dimensional digital grayscale image, observe the scanning results, and record the component categories contained in the three-dimensional grayscale image of the scanned braided CMCs.

[0019] The component categories include warp yarns, weft yarns, matrix, pores, and cracks.

[0020] In step S3, the deep learning segmentation model is a convolutional neural network.

[0021] In step S5, the post-processing includes: manually correcting the regions with erroneous identification in the segmentation results, and performing smoothing filtering on the boundaries of each component; the smoothing filtering uses recursive Gaussian filtering, and the filtering coefficients σ of the recursive Gaussian filtering in different directions in three-dimensional space are independently set according to the characteristics of the component boundaries.

[0022] In step S5, the selection criterion for the sub-volume is that the volume fraction of each material component contained therein is consistent with the total volume fraction of the corresponding component in the complete three-dimensional image.

[0023] In step S6, when generating the finite element mesh model, differentiated control parameters are set according to the geometric complexity of the actual microstructure. The control parameters include mesh type, minimum target mesh length, and maximum mesh length.

[0024] An electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the above-described finite element modeling and analysis method for braided CMCs based on real microstructure.

[0025] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-described finite element modeling and analysis method for braided CMCs based on real microstructure.

[0026] The present invention has the following beneficial effects:

[0027] 1. This method utilizes deep learning technology to achieve precise segmentation of the components (such as warp and weft yarns) within woven CMCs. Compared to traditional methods based on fixed thresholds, it better adapts to variations in material grayscale and image quality, thereby improving segmentation accuracy. Furthermore, compared to manual calibration, automatic segmentation using deep learning significantly reduces processing time and minimizes errors caused by human factors, making the entire recognition and segmentation process more efficient and reliable.

[0028] 2. This method ensures the realism of the finite element model and optimizes computational efficiency. Advanced edge smoothing technology is used when constructing the geometric model to appropriately simplify complex structures without sacrificing key details, reduce the size of the generated geometric model file, save computational resources, and facilitate large-scale simulation calculations.

[0029] 3. The high-fidelity finite element model obtained by this method provides strong support for the virtual evaluation of knitted CMCs. Its prediction results agree well with the actual performance of knitted CMCs. The numerical model established using this method can obtain more accurate data support in a shorter time. Attached Figure Description

[0030] Figure 1 These are cross-sections and three-dimensional views of CT scan images of woven CMCs samples in embodiments of the present invention.

[0031] Figure 2 This is a schematic diagram of data pairs used for deep learning training in an embodiment of the present invention; (a) is the cropped original sub-region Sample-image, and (b) is the corresponding manually labeled Label-image.

[0032] Figure 3 This is a graph showing the loss function and accuracy iteration curves during the training process of a deep learning model in an embodiment of the present invention.

[0033] Figure 4 The image shown is a segmentation result of a complete CT image by a deep learning model in an embodiment of the present invention; (a)-(e) are three-dimensional renderings of the warp, weft, matrix, crack, and pore components, respectively, and (f) is a three-dimensional fused view of all components.

[0034] Figure 5 This is a schematic diagram illustrating the selection of a representative sub-volume region (Fem-data) for finite element modeling from complete segmented data in an embodiment of the present invention, showing its relationship with the complete data.

[0035] Figure 6 This is a boundary comparison diagram before and after recursive Gaussian filtering of the segmented warp yarn components in an embodiment of the present invention.

[0036] Figure 7 This is a finite element geometric surface model diagram generated based on the real microstructure in an embodiment of the present invention.

[0037] Figure 8 This is a diagram of the finite element mesh model generated in an embodiment of the present invention.

[0038] Figure 9 This is a three-dimensional deformation cloud map obtained from finite element simulation calculations in an embodiment of the present invention.

[0039] Figure 10 This is a three-dimensional equivalent stress cloud diagram obtained from finite element simulation calculation in an embodiment of the present invention.

[0040] Figure 11 This is a three-dimensional equivalent strain cloud diagram obtained from finite element simulation calculation in an embodiment of the present invention. Detailed Implementation

[0041] To make the objectives, technical solutions, and advantages of this application clearer, the application is described and illustrated below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application. All other embodiments obtained by those skilled in the art based on the embodiments provided in this application without inventive effort are within the scope of protection of this application.

[0042] Obviously, the accompanying drawings described below are merely some examples or embodiments of this application. Those skilled in the art can apply this application to other similar scenarios based on these drawings without any inventive effort. Furthermore, it is understood that although the efforts made in this development process may be complex and lengthy, for those skilled in the art related to the content disclosed in this application, any changes to design, manufacturing, or production based on the technical content disclosed in this application are merely conventional technical means and should not be construed as insufficient disclosure of the content of this application.

[0043] In this application, the reference to "embodiment" means that a specific feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment that is mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described in this application may be combined with other embodiments without conflict.

[0044] This invention provides a finite element modeling and analysis method for braided CMCs based on real microstructure, comprising the following steps:

[0045] Step 1: Scan the braided CMCs using X-ray computed tomography (XCT) to obtain a three-dimensional digital grayscale image of size X*Y*Z, named Origin-image(X, Y, Z, resolution ratio), where X, Y, and Z are the three-dimensional dimensions of the image, and resolution ratio is the resolution of the scanned image. Observe the scanning results and record the component categories contained in the three-dimensional grayscale image of the braided CMCs after scanning.

[0046] Step 2: To achieve rapid and accurate segmentation of each component in the woven CMCs in Step 1, three-dimensional grayscale image data of arbitrary size x*y*z is cut and denoted as Sample-image (x, y, z, resolution ratio) for manual calibration (where x < X, y < Y, z < Z), and the labeled data is Label-image (x, y, z, resolution ratio);

[0047] Step 3: Take the Sample-image obtained in Step 2 as the input data set and Label-image as the target set; train the woven CMCs component segmentation model through deep learning, set the network type, the number of classification categories, the learning rate, the training verification ratio during the training process, and the number of iterations, etc.

[0048] Step 4: Through the segmentation model of woven CMCs obtained by deep learning, perform component recognition and segmentation on the complete three-dimensional grayscale image Origin-image of woven CMCs, and obtain the spatial distribution results of each component, providing a basis for subsequent construction of a high-fidelity finite element model;

[0049] Step 5: Check the recognition and segmentation results of each component of the woven CMCs obtained in Step 4 and correct the wrongly recognized parts; use recursive Gaussian filtering to soften the edges of the segmentation results of the woven CMCs. To improve the subsequent finite element calculation speed, according to the proportion of each component in the woven CMCs, select a sub-volume with the same volume fraction for simulation.

[0050] Step 6: Based on the true microstructure obtained in Step 5, set the relevant parameters of the finite element model, including the mesh type, the minimum target length of the mesh, the maximum mesh length, etc., generate the mesh model of the woven CMCs finite element model, and at the same time export the surface of the volume mesh to obtain the geometric structure of the finite element model;

[0051] Step 7: Import it into the finite element analysis software, set the true scale, material parameters, contact type, and boundary conditions of the model, perform finite element simulation analysis, and obtain the deformation, equivalent strain, equivalent stress, etc. of the high-fidelity finite element of the woven CMCs.

[0052] The following uses specific embodiments to explain the present invention:

[0053] Step 1: Scan the woven CMCs through XCT technology to obtain a three-dimensional digital grayscale image with a size of 2189px * 1197px * 1724px, and the image resolution is 3μm / px, denoted as Origin-image (2189, 1197, 1724, 3) (hereinafter simply referred to as Origin). After scanning, each cross-section and three-dimensional view of the woven CMCs are as Figure 1 As shown, based on the image content, they are divided into 5 types: warp yarns, weft yarns, matrix, pores, and cracks;

[0054] Step 2: To quickly and accurately segment the five types contained in the woven CMCs, this example arbitrarily cuts a representative 3D image data Sample-image (300, 300, 300, 3) of size 300px*300px from Origin (hereinafter referred to as Sample). The five types of data contained in Sample are manually labeled as Label-image (300, 300, 300, 3) (hereinafter referred to as Label). The cut Sample and the labeled Label are shown below. Figure 2 (a) and Figure 2 As shown in (b);

[0055] Step 3: Use the cropped original data Sample as input data and the manually labeled data Label as the target set; train the Braided CMCs component segmentation model through deep learning. The network type used in deep learning is ResNet18 residual neural network. The model has 5 classes, a learning rate of 0.0005, and a training-to-validation ratio of 3:1. After 100,000 iterations of training, the Braided_CMCs_SegModel for CMCs component recognition and segmentation is obtained. The iterative training process is as follows: Figure 3 As shown (only the iterative curves for weft yarn and pores are displayed), the accuracy of the specific segmentation model for each component is shown in the table below.

[0056] Identify categories warp yarn weft yarn matrix crack Pores Overall average accuracy 0.97514 0.97609 0.88731 0.89883 0.97628 0.942736

[0057] Step 4: Using the Braided_CMCs_SegModel segmentation model obtained through deep learning, the complete braided CMCs data (Origin) is used to identify and segment each component. The segmentation result is denoted as O_Label, and the spatial distribution of each component is shown below. Figure 4 (a)-4(f) show the warp yarn, weft yarn, matrix, crack, pores, and three-dimensional full view, respectively;

[0058] Step 5: Inspect the segmented braided CMCs data O_Label obtained in Step 4, correcting any erroneous identifications. To improve the subsequent finite element calculation speed, based on the proportion of each component in the braided CMCs, this example selects a sub-volume of 600px*300px*500px with the same volume fraction, denoted as Fem-data, to generate the finite element model of the braided CMCs. The relationship between its position and the complete data is shown below. Figure 5As shown, a recursive Gaussian filter is used to soften the edges of the woven CMCs segmentation results. The Gaussian σ values ​​in the XYZ directions are 2, 2, and 0. The comparison results before and after the warp filtering are shown below. Figure 6 As shown;

[0059] Step 6: Based on the Fem-data selected in Step 5, and taking the actual microstructure as a basis, set the relevant parameters of the finite element model, including the mesh type as quadrilateral mesh, the minimum target mesh length as 0.012 mm, and the maximum mesh length as 0.03 mm, etc., to generate the mesh model of the woven CMCs finite element model. At the same time, export the surface of the volume mesh to obtain the geometry of the finite element model. The result is as follows: Figure 7 and Figure 8 As shown;

[0060] Step 7: Import the geometric model and mesh into the finite element analysis software, set the actual dimensions of the woven CMCs (in mm), material parameters (custom material parameters based on experimental results), contact type, and boundary conditions, and perform finite element simulation analysis. The calculation results are as follows: Figures 9-11 As shown, the simulation results correspond to displacement, equivalent strain, and equivalent stress, respectively.

[0061] The above are merely preferred embodiments of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should be considered within the scope of protection of the present invention.

Claims

1. A finite element modeling and analysis method for braided CMCs based on real microstructure, characterized in that, Includes the following steps: Step S1: Obtain a three-dimensional digital grayscale image of the braided CMCs sample; Step S2: Crops out sub-region images from the three-dimensional digital grayscale image, and manually annotates each material component in the sub-region image to form training data; Step S3: Based on the training data, train a deep learning segmentation model to identify the material composition; Step S4: Using the trained deep learning segmentation model, perform component segmentation on the complete three-dimensional digital grayscale image to obtain the three-dimensional spatial distribution data of each component; Step S5: Post-process the segmented three-dimensional spatial distribution data and select a sub-volume that is representative of the material composition as the finite element modeling object; Step S6: Based on the actual mesoscopic structure data of the sub-volume, generate the corresponding finite element mesh model; Step S7: Import the finite element mesh model into the finite element analysis software, assign material properties, set boundary conditions, and perform simulation calculations to analyze the mechanical properties of the braided CMCs.

2. The finite element modeling and analysis method for braided CMCs based on real microstructure as described in claim 1, characterized in that, The specific steps of step S1 are as follows: scan the braided CMCs using X-ray computed tomography (CT) technology to obtain a three-dimensional digital grayscale image, observe the scanning results, and record the component categories contained in the three-dimensional grayscale image of the scanned braided CMCs.

3. The finite element modeling and analysis method for braided CMCs based on real microstructure as described in claim 2, characterized in that, The component categories include warp yarns, weft yarns, matrix, pores, and cracks.

4. The finite element modeling and analysis method for braided CMCs based on real microstructure as described in claim 1, characterized in that, In step S3, the deep learning segmentation model is a convolutional neural network.

5. The finite element modeling and analysis method for braided CMCs based on real microstructure as described in claim 1, characterized in that, In step S5, the post-processing includes: manually correcting the regions with erroneous identification in the segmentation results, and performing smoothing filtering on the boundaries of each component; the smoothing filtering uses recursive Gaussian filtering, and the filtering coefficients σ of the recursive Gaussian filtering in different directions in three-dimensional space are independently set according to the characteristics of the component boundaries.

6. The finite element modeling and analysis method for braided CMCs based on real microstructure as described in claim 1, characterized in that, In step S5, the selection criterion for the sub-volume is that the volume fraction of each material component contained therein is consistent with the total volume fraction of the corresponding component in the complete three-dimensional image.

7. The finite element modeling and analysis method for braided CMCs based on real microstructure as described in claim 1, characterized in that, In step S6, when generating the finite element mesh model, differentiated control parameters are set according to the geometric complexity of the actual microstructure. The control parameters include mesh type, minimum target mesh length, and maximum mesh length.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the finite element modeling and analysis method for braided CMCs based on real microstructure as described in any one of claims 1 to 7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by the processor, the program implements the finite element modeling and analysis method for braided CMCs based on real microstructure as described in any one of claims 1 to 7.

Images