An intelligent identification and modeling method for a preform in an aviation composite special-shaped component

Abstract

The application discloses an intelligent identification and modeling method for an aviation composite special-shaped component inner preform, and belongs to the technical field of aviation composite material microstructure digitization reconstruction and modeling. The method comprises the following steps: S1, data acquisition; S2, data preprocessing; S3, model training preparation; S4, intelligent segmentation; S5, three-dimensional reconstruction; S6, grid optimization; S7, topology processing; and S8, model output. The aviation composite special-shaped component inner preform intelligent identification and modeling method is adopted, topology errors and grid conforming problems that are prone to occur in explicit modeling are combined with grid Boolean operation and post-processing repair algorithms, grid intersections and discontinuous regions are accurately processed, grid gaps, missing surfaces and other closed defects introduced by the Boolean operation are repaired, the node conformality, topology connectivity and spatial closedness of the preform substructure grid are ensured, manual participation is reduced, and the modeling efficiency and stability are improved.

Description

Technical Field

[0001] This invention belongs to the field of digital reconstruction and modeling technology of microstructure of aerospace composite materials, specifically involving a method for intelligent identification and modeling of prefabricated bodies in aerospace composite irregular-shaped components. Background Technology

[0002] Digital reconstruction and modeling of the microstructure of aerospace composite materials refers to the process of converting the internal microstructure of aerospace composite materials, such as continuous fiber-reinforced ceramic matrix composites, into a digital form using technological means. Automatic identification of the microstructure of prefabricated components within irregularly shaped aerospace composite parts can provide a foundation for subsequent three-dimensional voxel reconstruction and refined modeling.

[0003] However, existing technologies have the following shortcomings: they are complex to implement, computationally intensive, require advanced programming and mathematical skills, and have a high threshold for engineering applications; they have limited support for Boolean operations on complex microstructures, and it is difficult to effectively extract and process multi-component coplanar regions, resulting in the fiber bundle region mesh being difficult to completely close, affecting the accuracy of subsequent simulation analysis; some processes require a large amount of manual intervention, and there is a lack of integrated and automated processing methods, which reduces modeling efficiency and stability.

[0004] Therefore, a new method is urgently needed. Summary of the Invention

[0005] The purpose of this invention is to provide an intelligent identification and modeling method for prefabricated bodies within aerospace composite irregular-shaped components. This method addresses the topological errors and mesh conformality problems that easily occur in explicit modeling. By combining mesh Boolean operations and post-processing repair algorithms, it accurately handles mesh intersections and discontinuous regions, repairs mesh gaps and missing surfaces, and other closure defects introduced by Boolean operations, ensuring the node conformality, topological connectivity, and spatial closure of the prefabricated substructure mesh. At the same time, it reduces manual intervention and improves the efficiency and stability of modeling.

[0006] To achieve the above objectives, this invention provides a method for intelligent identification and modeling of prefabricated bodies within aerospace composite irregular-shaped components, comprising:

[0007] S1. Perform CT scans on the target area of ​​the irregularly shaped aerospace composite material component to obtain the original CT digital slice data.

[0008] S2. Preprocess the original CT digital slice data in S1 to obtain the preprocessed slice sequence;

[0009] S3. Select 20% of the slice samples from the preprocessed slice sequence in S2, and manually label them to construct the training set;

[0010] S4. Based on the U-Net neural network framework built with Python, automatic instance segmentation of fiber bundles is achieved on all preprocessed slice sequences in S2 to obtain segmentation label maps;

[0011] S5. Reconstruct the 3D voxel model of the segmentation label map in S4 to obtain the STL mesh file;

[0012] S6. Import the STL mesh file from S5 into the prefabricated body surface mesh model, and automatically wrap and smooth the surface mesh of the substructure within the prefabricated body to obtain a smoothed surface mesh model.

[0013] S7. Based on the smoothed surface mesh model in S6, Boolean operations are performed on the prefabricated substructure mesh to handle mesh intersections and discontinuous regions, and a topologically connected mesh model is obtained.

[0014] S8. Based on the topologically connected mesh model in S7, extract the common interface regions of each component generated after Boolean operation, complete the spatial closure of the partition, and perform tetrahedral volume mesh subdivision on the closed triangular mesh domain to obtain the prefabricated volume model.

[0015] Preferably, S2 includes the following steps:

[0016] S201. Batch import raw CT digital slice data, set the target region's XYZ three-axis spatial range, and focus on the representative volume element region containing the preform using a clipping formula; the clipping formula is:

[0017] I crop (x,y)=I(x0+x,y0+y),x∈[0,w],y∈[[0,h];

[0018] Where (x0, y0) are the coordinates of the top left corner of the clipping window; h is the height of the window; and w is the width of the window.

[0019] An adaptive histogram equalization method is used to improve fiber bundle identification, and local contrast is enhanced through histogram equalization transformation; the transformation formula is:

[0020]

[0021] Where, r k n represents the k-th gray level; j L is the number of pixels at gray level j; L is the number of gray levels; MN is the image size.

[0022] S202. Gaussian filtering is used to smooth out random noise and high-frequency artifacts. The formula is:

[0023]

[0024] Where G(i,j) is a two-dimensional Gaussian kernel function;

[0025] S203. Morphological opening and closing operations are used to eliminate tiny isolated noise points and micropores. The formula is as follows:

[0026]

[0027] Where I is the original binary image; B is the structuring element; For expansion operation; For erosion calculation.

[0028] Preferably, S4 includes the following steps:

[0029] S401. The labeled dataset in S3 is paired with the corresponding preprocessed slice sequence in S2 and normalized. The image and label are then cropped at the center and filled at the boundary to ensure that the input size is adapted to the network structure and the network input size.

[0030] S402. A U-Net encoder-decoder structure is adopted. The encoding path extracts multi-scale features through convolution and pooling, while the decoding path restores features and spatial localization through upsampling and skip connections, outputting a multi-channel prediction map, represented as follows:

[0031] f(x)=Decoder(Encoder(x)+Skip Connections);

[0032] Among them, Encoder represents extracting high-dimensional semantic features, Decoder represents gradually restoring the spatial structure, and SkipConnections represents ensuring the fusion of features and location information;

[0033] S403 uses multi-class cross-entropy as the main loss, and introduces the Dice coefficient and Tversky loss index; the multi-class cross-entropy loss is expressed as:

[0034]

[0035] Among them, y i,k P is the true label for pixel i; i,k For predicting probabilities; L CE For multi-class cross-entropy loss;

[0036] The Dice coefficient is expressed as:

[0037]

[0038] Among them, y i For real labels; p i ∈ represents the predicted probability; ∈ represents the smoothing term;

[0039] S404. The Adam optimizer is used for training. The Dice coefficient is monitored through the validation set. The EarlyStopping and ReduceLROnPlateau mechanisms are used to prevent overfitting and adaptively adjust the learning rate. After training, the model weights are saved.

[0040] S405: Perform batch inference on all CT slices and output pixel-level segmentation label maps;

[0041] S406. Automatically predict the segmentation labels of batch CT slices and save the segmentation labels as standard image formats.

[0042] Preferably, S5 includes the following steps:

[0043] S501: Batch import 2D segmentation masks, identify materials according to RGB thresholds, and combine them into a voxel array V in three-dimensional space. (x,y,z) Each voxel is assigned a material label M. (x,y,z) in{0,1,2,3}; correspond to the background, matrix, fiber bundle A, and fiber bundle B, respectively.

[0044] S502. Isolated noise points are eliminated through three-dimensional morphological opening and closing operations, as shown below:

[0045]

[0046] Combined with connected component analysis, components with volumes smaller than the minimum connected component volume threshold S are removed. min The minute components, retaining the main structure, are represented as:

[0047]

[0048] Among them, V filtered (x,y,z) represents the output value of the voxel at position (x,y,z); S min The minimum connected component volume threshold is V(x,y,z); V(x,y,z) is the original value.

[0049] S503. Merge each material region into a single three-dimensional voxel model according to priority, and add a background boundary to the outer layer of the voxel.

[0050] S504. Traverse the voxel array. For each non-background volume (x, y, z), sequentially detect its adjacent voxels (x', y', z') in the six positive and negative axis directions. If the material numbers are different, generate an interface patch between them. The vertex coordinates of the patch are determined according to the voxel center and facing direction, and the patch is classified into the corresponding material interface.

[0051] S505. Each interface panel is generated as a quadrilateral, then split into two triangular faces; the formula for decomposing a quadrilateral into a triangle is:

[0052] Face quad =[v1,v2,v3,v4]→{[v1,v2,v3],[v1,v3,v4]};

[0053] Among them, Face quad v1, v2, v3, v4 are the four vertices of the quadrilateral face; {[v1, v2, v3], [v1, v3, v4]} are the two triangular faces obtained after splitting.

[0054] S506. After the voxel mesh coordinates are expanded by the boundary, they need to be translated as a whole to restore the original spatial position. Export each type of interface triangular mesh as a standard STL file to obtain the STL mesh file.

[0055] Preferably, S8 includes the following steps:

[0056] S801. For each fiber bundle surface mesh, using the minimum outer cube parameter of the matrix surface mesh, determine whether all vertices of its faces satisfy the following criteria:

[0057] A_min≤V≤A_max;

[0058] Where [A_min, A_max] is the minimum bounding cube of the region boundary;

[0059] If the condition is met, then the face piece is retained;

[0060] S802. For the prefabricated components after Boolean operations, extract the common interface between them to ensure the internal mesh is completely closed. Specifically:

[0061] Use a KD tree to search the vertex space and locate the common nodes between A and B; check whether the common nodes are on the same edge and identify the boundary loop edge; based on the connected component partitioning, divide A and B into several patches with the common boundary loop edge as the boundary, and extract the interface patch except for the main patch; merge all the interface patches and export them, assign the patches to the missing face component, and construct a closed internal interface mesh.

[0062] S803. For each substructure of the prefabricated body and the outer boundary of the region, perform the common interface identification and patch partitioning operation of S802, extract all common interface patches related to the boundary and merge and export them, so that each substructure of the prefabricated body is completely closed to the outside world.

[0063] S804. After all substructures within the prefabricated body are closed and topologically continuous, tetrahedral volume meshing is performed using the closed boundary as the constraint surface to generate the prefabricated body mesh model.

[0064] The present invention also provides an intelligent identification and modeling device for prefabricated bodies in aerospace composite irregular-shaped components, including: a data acquisition module for performing CT scans on the target area of ​​the irregular-shaped aerospace composite component to obtain original CT digital slice data;

[0065] A data preprocessing module, connected to the data acquisition module, is used to preprocess the raw CT digital slice data to obtain the preprocessed slice sequence.

[0066] The training set construction module, connected to the data preprocessing module, is used to select 20% of the slice samples from the preprocessed slice sequence and manually label them to construct the training set.

[0067] The intelligent segmentation module, connected to the training set construction module, is used to automatically segment fiber bundles from all preprocessed slice sequences using the U-Net neural network framework built in Python, and obtain segmentation label maps.

[0068] The 3D reconstruction module, connected to the intelligent segmentation module, is used to reconstruct the 3D voxel model of the segmentation label map to obtain an STL mesh file.

[0069] The mesh optimization module, connected to the 3D reconstruction module, is used to import the STL mesh file into the prefabricated surface mesh model, automatically cover and smooth the surface mesh of the substructure within the prefabricated body, and obtain a smoothed surface mesh model.

[0070] The topology processing module, connected to the mesh optimization module, performs Boolean operations on the prefabricated substructure mesh based on the smoothed surface mesh model, processes mesh intersections and discontinuous regions, and obtains a topologically connected mesh model.

[0071] The model output module is connected to the topology processing module. Based on the topologically connected mesh model, it extracts the common interface regions of each component generated after Boolean operation, completes the spatial closure of the partition, and performs tetrahedral volume mesh subdivision on the closed triangular mesh domain to obtain the prefabricated volume model.

[0072] Therefore, the present invention employs the above-mentioned intelligent identification and modeling method for prefabricated bodies within aerospace composite irregular-shaped components. Compared with the prior art, the technical solution of the present invention has the following beneficial effects:

[0073] (1) This invention adopts explicit geometric modeling, which overcomes the problems of implicit geometric modeling being complex, computationally intensive, and having a high threshold for engineering applications. It achieves intuitive operation, high efficiency, easy integration with subsequent simulation processes, and reduced difficulty in engineering applications.

[0074] (2) This invention improves segmentation accuracy by using U-Net neural network and combines the Boolean operation function of open source tools with Python script to overcome the problems of fiber bundles intersecting and topological structure and mesh continuity being destroyed due to the segmentation accuracy error of artificial intelligence in explicit modeling. It also achieves technical effects by using self-developed mesh post-processing algorithm to process mesh intersections and discontinuous regions and ensure the consistency and topological connectivity of substructure mesh boundary nodes.

[0075] (3) This invention integrates CT slice preprocessing, deep learning segmentation, and three-dimensional reconstruction, overcoming the problems of low modeling efficiency and stability caused by the need for a large amount of manual intervention in some processes and the lack of integrated automated processing methods in the existing technology. It achieves the technical effect of reducing manual intervention and improving modeling efficiency and stability.

[0076] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0077] Figure 1 This is a flowchart illustrating an embodiment of the intelligent identification and modeling method for prefabricated bodies within aerospace composite irregular-shaped components according to the present invention;

[0078] Figure 2 This is a schematic diagram of the microstructure obtained by CT scanning of the target area of ​​a turbine blade in an embodiment of the intelligent identification and modeling method for prefabricated bodies in aerospace composite irregular components according to the present invention;

[0079] Figure 3 This is a training diagram fed into the neural network, representing an embodiment of the intelligent recognition and modeling method for prefabricated bodies within aerospace composite irregular-shaped components according to the present invention; wherein, Figure 3 In the image, (a) represents the original image; where Figure 3 (b) in the image represents the labeled image;

[0080] Figure 4 This is a training model prediction segmentation result diagram of an embodiment of the intelligent recognition and modeling method for prefabricated bodies within aerospace composite irregular-shaped components according to the present invention; wherein, Figure 4 In the image, (a) represents the original image; where Figure 4 (b) in the image represents the segmentation result;

[0081] Figure 5 This is a schematic diagram of a three-dimensional reconstructed voxel model of a preform in an embodiment of the intelligent identification and modeling method for preforms in aerospace composite irregular-shaped components according to the present invention.

[0082] Figure 6 This is a schematic diagram of the preform surface mesh model, which is an embodiment of the intelligent identification and modeling method for preforms in aerospace composite irregular-shaped components according to the present invention.

[0083] Figure 7 This is a schematic diagram of the preform surface mesh reconstruction and smoothing model after an embodiment of the intelligent identification and modeling method for preforms in aerospace composite irregular-shaped components according to the present invention.

[0084] Figure 8 This is a schematic diagram of the intersecting area of ​​the prefabricated substructures in an embodiment of the intelligent identification and modeling method for prefabricated internal structures in aerospace composite irregular-shaped components according to the present invention.

[0085] Figure 9 This is a schematic diagram illustrating the post-processing of a prefabricated body after mesh Boolean operations, according to an embodiment of the intelligent identification and modeling method for prefabricated bodies within aerospace composite irregular-shaped components of the present invention; wherein, Figure 9 In the diagram, (a) represents the open surface of the structured Boolean operation; where Figure 9 (b) in the diagram represents the extracted red coplanar interface;

[0086] Figure 10 This is a schematic diagram of a three-dimensional mesh model of a prefabricated body, representing an embodiment of the intelligent identification and modeling method for prefabricated bodies within aerospace composite irregular-shaped components according to the present invention. Detailed Implementation

[0087] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. Unless otherwise defined, the technical or scientific terms used in the present invention should have the ordinary meaning understood by those skilled in the art.

[0088] Example 1

[0089] like Figures 1-10 As shown, the present invention provides a method for intelligent identification and modeling of prefabricated bodies within aerospace composite irregular-shaped components, comprising the following steps:

[0090] S1. Perform CT scans on the target area of ​​the irregularly shaped aerospace composite material component to obtain the original CT digital slice data, which includes the initial image information of the prefabricated structure.

[0091] S2. Preprocess the original CT digital slice data in S1 using a self-developed Python script, specifically as follows:

[0092] S201. Batch import raw CT digital slice data from the specified folder, set the target area's XYZ three-axis spatial range, and focus on the representative volume element (RVE) region containing the preform using a clipping formula to exclude irrelevant background and reduce data processing volume; the clipping formula is:

[0093] I crop (x,y)=I(x0+x,y0+y),x∈[0,w],y∈[[0,h];

[0094] Where (x0, y0) are the coordinates of the top-left corner of the clipping window; h is the height of the window; and w is the width of the window.

[0095] Adaptive histogram equalization (CLAHE) and other methods are employed to improve fiber bundle identification. Histogram equalization transform enhances local contrast and avoids artifacts caused by global over-enhancement. The transformation formula is as follows:

[0096]

[0097] Where, r k n represents the k-th gray level; j L is the number of pixels at gray level j; L is the number of gray levels; MN is the image size.

[0098] S202. Gaussian filtering is used to smooth out random noise and high-frequency artifacts. The formula is:

[0099]

[0100] Where G(i,j) is a two-dimensional Gaussian kernel function;

[0101] S203. Morphological opening and closing operations are used to eliminate tiny isolated noise points and micropores. The formula is as follows:

[0102]

[0103] Where I is the original binary image; B is the structuring element; For expansion operation; For erosion calculation;

[0104] By using region cropping, CLAHE contrast enhancement, Gaussian filtering, and morphological opening and closing operations, the saliency and segmentation accuracy of prefabricated structures in CT slices are significantly enhanced, noise and structural artifacts are effectively suppressed, and preprocessed slice sequences are obtained.

[0105] S3. Select 20% of the slice samples from the preprocessed slice sequence in S2 at equal intervals along the Z-axis, manually label the fiber bundles with different orientations, and obtain the labeled dataset.

[0106] S4. Using the U-Net neural network framework built with Python, automatic instance segmentation of fiber bundles is achieved on all preprocessed slice sequences in S2. Specifically:

[0107] S401. The labeled dataset in S3 is paired with the corresponding preprocessed slice sequence in S2 and normalized. The image and label are then cropped at the center and filled at the boundary to ensure that the input size is adapted to the network structure and the network input size.

[0108] S402. A U-Net encoder-decoder structure is adopted. The encoding path extracts multi-scale features through convolution and pooling, while the decoding path restores features and spatial localization through upsampling and skip connections, outputting a multi-channel prediction map, represented as follows:

[0109] f(x)=Decoder(Encoder(x)+Skip Connections);

[0110] Among them, Encoder represents extracting high-dimensional semantic features, Decoder represents gradually restoring the spatial structure, and SkipConnections represents ensuring the fusion of features and location information;

[0111] S403 uses sparse categorical crossentropy as the main loss, and introduces metrics such as the Dice coefficient and Tversky loss to improve the segmentation performance for regions with a small number of fiber bundles; the sparse categorical crossentropy loss is expressed as:

[0112]

[0113] Among them, y i,k P is the true label for pixel i; i,k For predicting probabilities; L CE For multi-class cross-entropy loss;

[0114] The Dice coefficient is expressed as:

[0115]

[0116] Among them, y i For real labels; p i ∈ represents the predicted probability; ∈ represents the smoothing term;

[0117] S404. The Adam optimizer is used for training. The Dice coefficient and other metrics are monitored through the validation set. Mechanisms such as EarlyStopping and ReduceLROnPlateau are used to prevent overfitting and adaptively adjust the learning rate. After training, the model weights are saved.

[0118] S405: Perform batch inference on all CT slices and output pixel-level segmentation label maps;

[0119] S406. Automatically predict segmentation labels for batch CT slices and obtain segmentation label maps; save the segmentation label maps in a standard image format.

[0120] S5. Develop a Python script to reconstruct a 3D voxel model of the segmentation label map in S4, and extract the surface triangular mesh of each component, specifically:

[0121] S501. Batch import 2D segmentation masks, identify materials (such as fiber bundles A and B, matrix) according to RGB thresholds, and combine them into a voxel array V in three-dimensional space. (x,y,z) Each voxel is assigned a material label M. (x,y,z) in{0,1,2,3}; correspond to the background, matrix, fiber bundle A (warp direction), and fiber bundle B (weft direction), respectively.

[0122] S502. Isolated noise points are eliminated through three-dimensional morphological opening and closing operations, as shown below:

[0123]

[0124] Combined with connected component analysis, components with volumes smaller than the minimum connected component volume threshold S are removed. min The minute components, retaining the main structure, are represented as:

[0125]

[0126] Among them, V filtered (x,y,z) represents the output value of the voxel at position (x,y,z); S min The minimum connected component volume threshold is V(x,y,z); V(x,y,z) is the original value.

[0127] S503. Merge each material region into a single three-dimensional voxel model according to priority, and add background boundaries to the outer layer of the voxel to avoid losing the model structure during subsequent boundary surface extraction.

[0128] S504. Traverse the voxel array. For each non-background volume (x, y, z), sequentially detect its adjacent voxels (x', y', z') in the six positive and negative axis directions. If the material numbers are different, generate an interface patch between them. The vertex coordinates of the patch are determined according to the voxel center and facing direction, and the patch is classified into the corresponding material interface.

[0129] S505. Each interface patch is generated as a quadrilateral, then split into two triangular faces. Spatial coordinate hashing is used to remove duplicate vertices and eliminate redundancy; triangular mesh optimization improves mesh quality and output efficiency, using the formula for decomposing a quadrilateral into triangles:

[0130] Face quad=[v1,v2,v3,v4]→{[v1,v2,v3],[v1,v3,v4]};

[0131] Among them, Face quad v1, v2, v3, v4 are the four vertices of the quadrilateral face; {[v1, v2, v3], [v1, v3, v4]} are the two triangular faces obtained after splitting.

[0132] S506. After the voxel mesh coordinates are expanded by the boundary, they need to be translated as a whole to restore their original spatial positions. Finally, the triangular mesh of each type of interface is exported as a standard STL file to obtain the STL mesh file; this provides an input interface for subsequent finite element mesh generation and modeling integration; thus realizing efficient and automated 3D voxel reconstruction and interface surface meshing of prefabricated bodies in irregular aerospace composite components, ensuring the continuity of the model;

[0133] S6. Import the STL mesh file from S5 into the prefabricated surface mesh model. Use a Python script to call the AlphaWrap function of the open-source tool Pymeshlab to automatically wrap and smoothly reconstruct the surface mesh of the substructure within the prefabricated body, generating a high-quality closed surface mesh. Although S6 optimizes the surface quality, there are still potential topological errors and mesh conformal defects at the intersections of complex substructures, which need to be repaired by subsequent methods.

[0134] S7. Based on the smoothed surface mesh model in S6, the MeshBoolean function of the open-source tool Pymeshlab is called through a Python script to perform Boolean operations on the prefabricated substructure mesh. During the operation, redundant nodes and cross patches caused by factor structure interference are detected and corrected in real time, and mesh intersections and discontinuous regions are handled to ensure the consistency and topological connectivity of the boundary nodes of the substructure mesh.

[0135] S8. Based on the topologically connected mesh model in S7, the common interface regions of each component generated after Boolean operations are automatically extracted using a Python script to complete the spatial closure of all partitions. Then, tetrahedral volume meshing is performed on the closed triangular mesh domain to obtain the prefabricated volume model. This includes the following steps:

[0136] S801. For each fiber bundle surface mesh, using the minimum outer cube parameters A_min and A_max of the matrix surface mesh, determine whether all vertices of its faces satisfy the following conditions:

[0137] A_min≤V≤A_max;

[0138] Where [A_min, A_max] is the minimum bounding cube of the region boundary;

[0139] If the judgment condition is met, the facet is retained to achieve the screening of the fiber bundle surface inside the preform area;

[0140] S802. For prefabricated components after Boolean operations (such as fiber bundles A and B), extract the common interface between them to make the internal mesh completely closed. Specifically: use a KD-tree to perform vertex space search to accurately locate the common nodes between A and B; check whether the common nodes are on common edges to identify boundary ring edges; based on connected components, divide A and B into several patches with the boundary ring edges as the boundary, and extract the interface patches except for the main patch; merge all the interface patches and export them, assign the patches to the missing surface components, and construct a closed internal interface mesh;

[0141] S803. For each substructure of the prefabricated body and the outer boundary of the region, perform the common interface identification and patch partitioning operation of S802, extract all common interface patches related to the boundary and merge and export them, so that each substructure of the prefabricated body is completely closed to the outside world.

[0142] S804. After all substructures within the prefabricated body are closed and topologically continuous, tetrahedral volume meshing is performed using the closed boundary as the constraint surface to generate the prefabricated body mesh model.

[0143] In this embodiment:

[0144] S1, Select C f / SiC turbine guide vane blades were used as implementation samples. CT scans were performed on selected target areas of the blade components to initially obtain CT image information containing the prefabricated structure.

[0145] S2. Scripts are developed based on Python (Ver 3.9), mainly relying on libraries such as numpy, opencv-python, pydicom, and matplotlib. The specific processing flow is as follows:

[0146] The slice data obtained from CT scans were imported and grayscale normalized to ensure that the normalized pixel values ​​were distributed in the [0,1] range. Preliminary functional unit region cropping was performed on the image to be processed, resulting in a three-axis size of 250×100×250 for the model, focusing on the key region containing the prefab. To unify the size of subsequent neural network training, each slice was padded with zeros using the pad_to_multiple function, ensuring that both its height and width were divisible by 16. The processed single-channel slice size was (112, 256, 1). After secondary slice size processing, OpenCV's CLAHE (Adaptive Histogram Equalization) function was used to enhance contrast. Gaussian blur was then used to denoise the slices, improving their quality and yielding the processed slice image.

[0147] S3. From the sliced ​​images processed in S2 (a total of 251 images, used to construct the 3D voxel model), to ensure that the samples used to train the neural network prediction model can cover the structural features of the entire target domain of the prefabricated body, approximately 20% of the slices (i.e., 50 images) are selected from the dataset using an equidistant Z-axis spacing method as the training set data. Using the open-source software ImageJ, the fiber bundles in different directions are manually labeled: red represents warp yarns (fiber bundle A), blue represents weft yarns (fiber bundle B), and black represents the background area, generating a corresponding number of labeled datasets (GroundTruth).

[0148] S4. Build the U-Net neural network based on Python (Ver 3.9), mainly relying on libraries such as numpy, opencv-python, pydicom, matplotlib, tensorflow, and scikit-learn. The process is as follows:

[0149] The preprocessed slices are saved as NumPy arrays and used with labeled images to generate a training set. The training set is divided into 80% for training and 20% for validation. Before the training set is fed into the network, data augmentation processing (such as rotation, translation, scaling, etc.) is performed to improve the richness of feature samples.

[0150] The architecture employs the classic U-Net symmetric structure, including downsampling and upsampling paths. Each layer contains convolution, BatchNorm (batch normalization), and ReLU activation functions. Downsampling is achieved through max pooling, and upsampling is achieved through transposed convolution. The backbone structure parameters are as follows: input layer size is (112, 256, 1); the downsampling stage has four levels with convolution kernel numbers of 64, 128, 256, and 512 respectively; the upsampling stage also has four levels with convolution kernel numbers set in reverse order of 512, 256, 128, and 64; the output layer is a 1×1 convolution, outputting 3 channels of logits (corresponding to the background, fiber bundle A, and fiber bundle B respectively).

[0151] The network employs SparseCategoricalCrossentropy loss (from_logits=True) and monitors comprehensive metrics such as Sparse_Categorical_Crossentropy, Dice coefficient, and Tversky Focal Loss in real time to balance segmentation accuracy and class balance. The Dice coefficient and Tversky Focal Loss are expressed as follows:

[0152]

[0153] Loss = (1 - Tversky) γ ;

[0154] Where A represents the ground truth region; B represents the segmentation region predicted by the model; TP represents the region of the predicted pair; FP represents the false positive; FN represents the false negative; α represents the weight of FP; β represents the weight of FN; and γ represents the focusing parameter.

[0155] The learning rate is automatically adjusted and an early termination strategy is adopted using Early Stopping and ReduceLROnPlateau. Key parameter settings: Adam optimizer is used, and the initial learning rate is 1e. -3 (Dynamically adjusted), batch size is 8, maximum iteration is 300 epochs;

[0156] After training, the optimal model is saved, and the full CT slice sequence is automatically predicted and segmented. The segmentation labels with the same size as the original mask are output and saved in PNG format. The segmentation results are visualized using a three-color system of background, red, and blue.

[0157] S5. Develop a Python script to process the 2D slice label image output by U-Net segmentation. The specific process is as follows:

[0158] Load the sliced ​​label image (size 251×101, slice layers 251), and automatically identify the material based on the label information and color thresholds: red (R>128, G<50, B<50) is identified as fiber bundle A (value 2), blue (B>128, R<50, G<50) is identified as fiber bundle B (value 3), and the remaining areas are automatically identified as matrix (value 1) or background (value 0); reconstruct the 3D voxel model of the prefabricated body through the custom function load_color_masks;

[0159] Three-dimensional opening and closing operations are performed on the obtained voxel models. Each group of voxels performs two opening and closing operations on 9×9×9 structural elements to remove isolated noise generated by segmentation.

[0160] The current voxel model is 3D voxel encoded. Fiber bundle A, fiber bundle B, and matrix are combined into a 3D integer voxel array. A background region is introduced outside the region to construct a four-value voxel system. The six types of interfaces of the voxel array are automatically identified (background-matrix, background-fiber bundle A / B, matrix-fiber bundle A / B, AB interface). The interfaces of dissimilar adjacent voxels are determined for each voxel and four-node rectangular patches are output. These patches are then split into triangular patches.

[0161] All interfaces are divided into triangular patches, and independent 3D STL mesh files are generated for each type of interface;

[0162] S6. Develop a Python script to import the prefabricated surface mesh model, use a loop function to traverse each substructure of the prefabricated body, and call the AlphaWrap function in the open-source tool Pymeshlab library to reconstruct the surface mesh (function parameter settings: alpha 0.002, offset 0.0005); after the surface reconstruction is completed, call the HCLaplacian Smooth function to smooth the surface mesh, generate a high-quality closed surface mesh, and output the processed prefabricated model.

[0163] S7. Since the substructures inside the prefabricated body are in contact with each other, after the surface covering mesh reconstruction and smoothing, the contact surfaces will become intersecting areas. Therefore, the MeshBoolean function of the open-source tool Pymeshlab is called through a Python script to perform Boolean operations on the prefabricated substructure mesh. During the operation, a "node consistency verification mechanism" is introduced to detect and correct redundant nodes and intersecting patches caused by factor structure interference in real time. This effectively solves the problems of "non-convergence" and "boundary misalignment" that are easy to occur in traditional Boolean operations, and finally obtains a topologically connected mesh model.

[0164] S8. Develop a script based on Python (mainly relying on Trimesh and SciPy libraries) to perform post-processing on the prefabricated model after completing Boolean operations at the mesh level. The process is as follows:

[0165] Load the prefabricated mesh STL file, use the spatial clipping algorithm, based on the region cube bounding box of the matrix, to distinguish the mesh vertices of substructures such as fiber bundles, and retain the triangular faces of substructures that are completely located within the bounding box of the matrix.

[0166] Trimesh decomposition is used to obtain all connected domains of the triangular mesh of the prefabricated fiber bundle A or B body. A fast nearest neighbor search (KDTree) is then performed on the vertices of another substructure (such as fiber bundle B or the matrix) using a spatial tolerance tol = 1e. -6For each connected component, identify triangular faces that share nodes and edges with another substructure, and divide them into patches using a boundary "cutting" method (the patch is a subset of the original structure, retaining the node index and face index of the original structure); automatically detect the area of ​​each patch, regard the main patch with the largest area as the main region, and output and merge all the secondary patches as boundary patches to ensure that all gaps are filled; assign the extracted face patches to the substructures that are missing the region for shared node processing to form an internal spatial closed domain;

[0167] Repeat the patch identification process described above with the outer box of the region boundary and the surfaces of fiber bundles A and B to find all patches at the boundary with the external region to prevent missing closed domains; output the "patch" triangular facets between all partitions and the outside to provide the necessary boundaries for the closure of the spatial domain;

[0168] The main surfaces, patches, and external interface patches of each partition are aggregated to form a closed triangular mesh of the overall space. The mesh merging and deduplication mechanism of Trimesh is used to ensure the uniqueness of nodes and the continuity of surfaces. The closed triangular surface region of the prefab is used as input, and the mesh is automatically generated by Delaunay (the target mesh size is 1mm). Finally, the prefab volume mesh model is output.

[0169] Therefore, the present invention adopts the above-mentioned intelligent identification and modeling method for prefabricated bodies in aerospace composite irregular components. This method addresses the topological errors and mesh conformance problems that are prone to occur in explicit modeling. It combines mesh Boolean operations and post-processing repair algorithms to accurately handle mesh intersections and discontinuous regions, repair mesh gaps, missing surfaces and other closure defects introduced by Boolean operations, and ensure the node conformance, topological connectivity and spatial closure of the prefabricated body substructure mesh. At the same time, it reduces manual intervention and improves the efficiency and stability of modeling.

[0170] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for intelligent identification and modeling of prefabricated bodies within aerospace composite irregular-shaped components, characterized in that, Includes the following steps: S1. Perform CT scans on the target area of ​​the irregularly shaped aerospace composite material component to obtain the original CT digital slice data. S2. Preprocess the original CT digital slice data in S1; Obtain the preprocessed slice sequence; S3. Select 20% of the slice samples from the preprocessed slice sequence in S2, and manually label them to construct the training set; S4. Based on the U-Net neural network framework built with Python, automatic instance segmentation of fiber bundles is achieved on all preprocessed slice sequences in S2 to obtain segmentation label maps; S5. Reconstruct the 3D voxel model of the segmentation label map in S4 to obtain the STL mesh file; S6. Import the STL mesh file from S5 into the prefabricated body surface mesh model, and automatically wrap and smooth the surface mesh of the substructure within the prefabricated body to obtain a smoothed surface mesh model. S7. Based on the smoothed surface mesh model in S6, Boolean operations are performed on the prefabricated substructure mesh. During the operation, redundant nodes and cross patches caused by factor structure interference are detected and corrected in real time. Mesh intersections and discontinuous regions are processed to obtain a topologically connected mesh model. S8. Based on the topologically connected mesh model in S7, extract the common interface regions of each component generated after Boolean operation, complete the spatial closure of the partition, and perform tetrahedral volume mesh subdivision on the closed triangular mesh domain to obtain the prefabricated volume model.

2. The intelligent identification and modeling method for prefabricated bodies within aerospace composite irregular-shaped components according to claim 1, characterized in that, S2 includes the following steps: S201. Batch import raw CT digital slice data, set the target region's XYZ three-axis spatial range, and focus on the representative volume element region containing the preform using a clipping formula; the clipping formula is: ； in, The coordinates of the top-left corner of the cropping window; The height of the window; The width of the window; An adaptive histogram equalization method is used to improve fiber bundle identification, and local contrast is enhanced through histogram equalization transformation; the transformation formula is: ； in, For the first grayscale levels; grayscale The number of pixels; L The number of gray levels; MN Image size; S202. Gaussian filtering is used to smooth out random noise and high-frequency artifacts. The formula is: ； in, It is a two-dimensional Gaussian kernel function; S203. Morphological opening and closing operations are used to eliminate tiny isolated noise points and micropores. The formula is as follows: ； ； in, The original binary image; B For structural elements; For expansion operation; For erosion calculation.

3. The intelligent identification and modeling method for prefabricated bodies within aerospace composite irregular-shaped components according to claim 1, characterized in that, S4 includes the following steps: S401. The labeled dataset in S3 is paired with the corresponding preprocessed slice sequence in S2 and normalized. The image and label are then cropped at the center and filled at the boundary to ensure that the input size is adapted to the network structure and the network input size. S402. A U-Net encoder-decoder structure is adopted. The encoding path extracts multi-scale features through convolution and pooling, while the decoding path restores features and spatial localization through upsampling and skip connections, outputting a multi-channel prediction map, represented as follows: ； Among them, Encoder represents extracting high-dimensional semantic features, Decoder represents gradually restoring the spatial structure, and SkipConnections represents ensuring the fusion of features and location information; S403 uses multi-class cross-entropy as the main loss, and introduces the Dice coefficient and Tversky loss index; the multi-class cross-entropy loss is expressed as: ； in, For pixels i The true label; To predict probabilities; For multi-class cross-entropy loss; The Dice coefficient is expressed as: ； in, This is a real label; To predict probabilities; For smoothing terms; S404. The Adam optimizer is used for training. The Dice coefficient is monitored through the validation set. The EarlyStopping and ReduceLROnPlateau mechanisms are used to prevent overfitting and adaptively adjust the learning rate. After training, the model weights are saved. S405: Perform batch inference on all CT slices and output pixel-level segmentation label maps; S406. Automatically predict the segmentation labels of batch CT slices and save the segmentation labels as standard image formats.

4. The intelligent identification and modeling method for prefabricated bodies within aerospace composite irregular-shaped components according to claim 2, characterized in that, S5 includes the following steps: S501: Batch import 2D segmentation masks, identify materials according to RGB thresholds, and combine them into a voxel array in three-dimensional space. Each voxel is assigned a material label. ; These correspond to the background, matrix, fiber bundle A, and fiber bundle B, respectively. S502. Isolated noise points are eliminated through three-dimensional morphological opening and closing operations, as shown below: ； ； Combined with connected component analysis, components with volumes smaller than the minimum connected component volume threshold are removed. The minute components, retaining the main structure, are represented as: ； in, For voxels in The output value of the position; The minimum connected component volume threshold; The original value; S503. Merge each material region into a single three-dimensional voxel model according to priority, and add a background boundary to the outer layer of the voxel. S504, Traverse the voxel array, for each non-background voxel... It sequentially detects adjacent voxels along the six positive and negative axes. If the material numbers are different, an interface patch is generated between them; the vertex coordinates of the patch are determined by the voxel center and the facing direction, and the patch is classified into the corresponding material interface. S505. Each interface panel is generated as a quadrilateral, then split into two triangular faces; the formula for decomposing a quadrilateral into a triangle is: ； in, It is a quadrilateral facet; These are the four vertices of the quadrilateral face. These are the two triangular facets obtained after splitting; S506. After the voxel mesh coordinates are expanded by the boundary, they need to be translated as a whole to restore the original spatial position. Export each type of interface triangular mesh as a standard STL file to obtain the STL mesh file.

5. The intelligent identification and modeling method for prefabricated bodies within aerospace composite irregular-shaped components according to claim 1, characterized in that, S8 includes the following steps: S801. For each fiber bundle surface mesh, using the minimum outer cube parameter of the matrix surface mesh, determine whether all vertices of its faces satisfy the following criteria: ； in, The minimum enclosing cube of the region boundary; If the condition is met, then the face piece is retained; S802. For the prefabricated components after Boolean operations, extract the common interface between them to ensure the internal mesh is completely closed. Specifically: Use a KD tree to search the vertex space and locate the common nodes between A and B; check whether the common nodes are on the same edge and identify the boundary loop edge; based on the connected component partitioning, divide A and B into several patches with the common boundary loop edge as the boundary, and extract the interface patch except for the main patch; merge all the interface patches and export them, assign the patches to the missing face component, and construct a closed internal interface mesh. S803. For each substructure of the prefabricated body and the outer boundary of the region, perform the common interface identification and patch partitioning operation of S802, extract all common interface patches related to the boundary and merge and export them, so that each substructure of the prefabricated body is completely closed to the outside world. S804. After all substructures within the prefabricated body are closed and topologically continuous, tetrahedral volume meshing is performed using the closed boundary as the constraint surface to generate the prefabricated body mesh model.

6. A device for intelligent identification and modeling of prefabricated bodies within aerospace composite irregular-shaped components, applied to the method for intelligent identification and modeling of prefabricated bodies within aerospace composite irregular-shaped components as described in any one of claims 1-5, characterized in that, include: The data acquisition module is used to perform CT scans on the target area of ​​irregularly shaped aerospace composite material components to obtain raw CT digital slice data. A data preprocessing module, connected to the data acquisition module, is used to preprocess the raw CT digital slice data; Obtain the preprocessed slice sequence; The training set construction module, connected to the data preprocessing module, is used to select 20% of the slice samples from the preprocessed slice sequence and manually label them to construct the training set. The intelligent segmentation module, connected to the training set construction module, is used to automatically segment fiber bundles from all preprocessed slice sequences using the U-Net neural network framework built in Python, and obtain segmentation label maps. The 3D reconstruction module, connected to the intelligent segmentation module, is used to reconstruct the 3D voxel model of the segmentation label map to obtain an STL mesh file. The mesh optimization module, connected to the 3D reconstruction module, is used to import the STL mesh file into the prefabricated surface mesh model, automatically cover and smooth the surface mesh of the substructure within the prefabricated body, and obtain a smoothed surface mesh model. The topology processing module, connected to the mesh optimization module, performs Boolean operations on the prefabricated substructure mesh based on the smoothed surface mesh model, processes mesh intersections and discontinuous regions, and obtains a topologically connected mesh model. The model output module is connected to the topology processing module. Based on the topologically connected mesh model, it extracts the common interface regions of each component generated after Boolean operation, completes the spatial closure of the partition, and performs tetrahedral volume mesh subdivision on the closed triangular mesh domain to obtain the prefabricated volume model.

Images