Method for predicting mechanical property of fiber reinforced polyurea composite material based on deep learning
Through deep learning-based methods, two-dimensional and three-dimensional models of fiber-reinforced polyurea composite materials are established and multi-dimensional convolutional neural networks are used for prediction, which solves the problems of low efficiency and low accuracy of mechanical properties prediction of fiber-reinforced polyurea composite materials, and achieves efficient and accurate mechanical properties prediction.
Patent Information
- Application Number
- CN202510267905.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-03-07
- Publication Date
- 2025-06-27
AI Technical Summary
The prior art is difficult to effectively predict the mechanical properties of fiber-reinforced polyurea composites, especially under high dimensionality and strong nonlinearity problems, the prediction efficiency of traditional methods is low and the accuracy is not high.
A deep learning-based method is adopted to establish two-dimensional and three-dimensional fiber-reinforced polyurea composite material models through image recognition and Lagrangian model, combined with numerical analysis and finite element simulation, a mathematical model between macrostructural parameters and mesostructural parameters is established, and a deep learning model training is used using a multi-dimensional convolutional neural network to improve the prediction efficiency and accuracy of mechanical performance.
The prediction efficiency and accuracy of the mechanical properties of fiber-reinforced polyurea composites are significantly improved, and the shortcomings of traditional methods under the problems of high dimensionality and strong nonlinearity are solved, and digital analysis and adaptive optimization design of fiber-reinforced polyurea composites are realized.
Smart Images
Figure CN120217845A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the field of materials, and particularly relates to a method for predicting the mechanical properties of fiber-reinforced polyurea composites based on deep learning. Background Art
[0002] The prediction of the mechanical properties of materials based on machine learning and deep learning methods has been applied in various material fields. The deep learning method can realize the end-to-end prediction of the mechanical properties of materials, directly establish the mapping relationship between the mesoscopic structure or design parameters and the mechanical properties of materials, and avoid the feature loss caused by the artificial screening of structural features. For the unique and random mesoscopic structure of fiber-reinforced polyurea composites, the deep learning method can autonomously recognize the structural features, correlate the mesoscopic structure and the macroscopic mechanical properties, thereby improving the prediction efficiency and prediction accuracy of the mechanical properties of fiber-reinforced polyurea composites. Summary of the Invention
[0003] The purpose of the present invention is to provide a method for predicting the mechanical properties of fiber-reinforced polyurea composites based on deep learning.
[0004] The technical solution for realizing the purpose of the present invention is: a method for predicting the mechanical properties of fiber-reinforced polyurea composites based on deep learning, comprising the following steps:
[0005] Step (1): Establish a two-dimensional mesoscopic fiber-reinforced polyurea composite model based on the image recognition method;
[0006] Step (2): Establish a three-dimensional fiber-reinforced polyurea composite model based on the Lagrangian model;
[0007] Step (3): Use the numerical analysis method to establish a mathematical model between the macroscopic structure parameters and the mesoscopic structure parameters, and realize the correlation of the macro-mesoscopic structure parameters of the fiber-reinforced polyurea composites;
[0008] Step (4): Establish a three-dimensional voxel model of the fiber-reinforced polyurea composite and its corresponding mechanical properties, and obtain the deep learning data set of the three-dimensional fiber-reinforced polyurea composite; establish a deep learning model containing a multi-dimensional convolutional neural network, and train it using the deep learning data set of the three-dimensional fiber-reinforced polyurea composite;
[0009] Step (5): Predict and obtain the mechanical properties of the fiber-reinforced polyurea composite according to the deep learning model containing a multi-dimensional convolutional neural network.
[0010] Further, in step (1), based on the Python programming language, the Retinex algorithm of the OpenCV library is called, and a program for identifying the boundaries of fiber-reinforced polyurea composites is written to extract the internal mesoscopic structure features. The detailed steps are as follows:
[0011] Step (11): Image highlight removal: Based on the brightness change function in the OpenCV library, remove the bright spots reflected on the surface structure of the fiber-reinforced polyurea composite material specimen, and separate the illumination and reflection components;
[0012] Step (12): Gaussian filtering: Call the cv2.GaussianBlur function of the OpenCV library to generate a Gaussian kernel according to σ and the kernel size, and convolve the Gaussian kernel with the scanned image of the fiber-reinforced polyurea composite material to obtain a smoothed image;
[0013] Step (13): Edge detection: Use the Sobel operator to extract the outer contour line of the fiber-reinforced polyurea composite structure, calculate the horizontal and vertical gradients respectively using two 3x3 kernels, combine to obtain the edge intensity, and only retain the thinner boundary line at the actual edge in the fiber-reinforced polyurea composite structure;
[0014] Step (14): Filling: Call the drawContours function of the OpenCV library to draw and fill the closed contour after extracting the boundary curve of the fiber-reinforced polyurea composite structure.
[0015] Furthermore, the internal microscopic structural features in step (1) include the number of fiber layers, ply angles, and polyurea volume fraction.
[0016] Furthermore, step (2) is specifically as follows:
[0017] Step (21): Generate the starting point of the two-dimensional microscopic fiber-reinforced polyurea composite material model in step (1), and write the Lagrangian model generation algorithm based on Python. The Lagrangian model algorithm process is as follows: First, define a Cartesian coordinate system in the spatial region and randomly generate 100 - 1000000 control points to calibrate the area where each fiber is located; the generated control points need to meet the requirements of the fiber diameter distance between control points; some model parameters need to be set when establishing the Lagrangian model, including the model space size, the number of control points N, and the minimum distance D between control points min , and the following formula is used for calculation:
[0018]
[0019] Among them, L is the Lagrangian function, λ is the Lagrange multiplier, represents the coordinate of the i-th control point, r j represents the coordinate of the j-th control point, which is used to measure the randomness of model generation, and its value range is 0 - 1, taking 0.2;
[0020] Step (22): Random fiber weaving algorithm: Based on Perlin noise to simulate natural fiber bending, each fiber advances along the set direction while randomly changing the direction to form a natural weaving effect; Cubic Hermite interpolation Fade Function is used for smoothing:
[0021] f(t) = 6t 5 -15t 4 +10t 3
[0022] where t is the normalized coordinate in the grid cell;
[0023] Let be the noise values of the four lattice points S 00 , S 10 , S 01 , S 11 respectively, and the interpolation formula:
[0024] N(x, y) = Lerp(Lerp(S 00 , S 10 , u), Lerp(S 01 , S 11 , u), v);
[0025] Step (23): Perform 3D fiber weaving - Matplotlib 3D visualization on the Lagrangian model to obtain a three-dimensional mesoscopic fiber-reinforced polyurea composite material model.
[0026] Furthermore, step (3) is specifically: Represent and perform feature engineering on the composite material mechanical property dataset through normalization, regularization processing, and principal component analysis, and reduce the dimension of the dataset by selecting the input features most relevant to the macroscopic mechanical properties; Based on the reduced dataset, with the mesoscopic parameters as the input and the macroscopic mechanical properties of the composite material as the output, select the gradient boosting tree model machine learning non-linear model for regression prediction to obtain the relationship between the macroscopic properties and the mesoscopic structure parameters.
[0027] Furthermore, the input mesoscopic parameters include the number of mesoscopic fiber layers, ply angle, and polyurea volume fraction.
[0028] Furthermore, step (4) is specifically:
[0029] Step (41): Generate a fiber-reinforced polyurea composite material geometric model through a three-dimensional mesoscopic fiber-reinforced polyurea composite material modeling plug-in, and transform it into a three-dimensional fiber-reinforced polyurea composite material voxel model recognizable by the neural network through point clouding and voxelization methods;
[0030] Step (42): Mechanical property calculation; various mechanical property parameters of the three-dimensional fiber-reinforced polyurea composite voxel model established in step (41) are obtained through finite element simulation calculation;
[0031] Step (43): Establish a deep learning data set: The deep learning data set consists of the fiber-reinforced polyurea composite voxel model and the mechanical properties obtained in step (2);
[0032] Step (44): Establishment and training of a deep learning model: A deep learning model including a multi-dimensional convolutional neural network is established, and the data set obtained in step (3) is used for the training and verification of the deep learning model;
[0033] Step (45): Deep learning model testing: A deep learning test data set is established, and the mechanical properties of the fiber-reinforced polyurea composite in the test data set are predicted by using the trained deep learning model to verify the effectiveness of the deep learning model.
[0034] Furthermore, the mechanical property parameters in step (42) include specific energy absorption and elongation.
[0035] A fiber-reinforced polyurea composite has the mechanical properties predicted by the above method.
[0036] The use of the above fiber-reinforced polyurea composite is for penetration protection of armored vehicles.
[0037] Compared with the prior art, the present invention has the following remarkable advantages:
[0038] The traditional design method of fiber-reinforced polyurea composites relies on single finite element simulation, with low efficiency, low digitization and automation, and often ignores the influence of the mechanical properties of mesoscopic fiber-reinforced polyurea composites on macroscopic properties. The present invention solves the high-dimensional and strong non-linear problems in the design process of fiber-reinforced polyurea composites, improves the prediction efficiency and accuracy of the mechanical properties of mesoscopic fiber-reinforced polyurea composites through deep learning methods; thereby realizing the performance prediction of fiber-reinforced polyurea composites based on deep learning, greatly improving the digital analysis efficiency of fiber-reinforced polyurea composites, and laying a foundation for the future adaptive optimization design of fiber-reinforced polyurea composites. BRIEF DESCRIPTION OF THE DRAWINGS
[0039] Figure 1 It is a schematic flow chart of the method for predicting the mechanical properties of fiber-reinforced polyurea composites based on deep learning of the present invention.
[0040] Figure 2 It is the modeling process based on the image recognition method of the present invention.
[0041] Figure 3This is the technical route for predicting the performance of the three-dimensional microstructure of the composite material of the present invention.
[0042] Figure 4 This is the mechanical property prediction framework for the fiber-reinforced polyurea composite material of the present invention. Specific embodiments
[0043] The present invention will be further described in detail below with reference to the accompanying drawings.
[0044] As Figure 1 shown, a method for predicting the mechanical properties of fiber-reinforced polyurea composite materials based on deep learning, the specific steps are as follows:
[0045] Step 1, establish a two-dimensional mesoscopic fiber-reinforced polyurea composite material model based on the image recognition method according to the structural characteristics of the fiber-reinforced polyurea composite material;
[0046] Step 2, establish a three-dimensional fiber-reinforced polyurea composite material model based on the Lagrangian model, and establish a two-dimensional mesoscopic fiber-reinforced polyurea composite material model based on the picture recognition algorithm; specifically:
[0047] Based on the Lagrangian model, a random fiber and solid model generation algorithm is added, and a modeling method for a three-dimensional mesoscopic solid fiber-reinforced polyurea composite material with random fiber weaving is proposed. By scanning the surface structure of the actual solid fiber-reinforced polyurea composite material, the structure boundary of the solid fiber-reinforced polyurea composite material is extracted by applying image recognition technology, and a method for establishing a two-dimensional mesoscopic solid fiber-reinforced polyurea composite material model is proposed.
[0048] Step 3, use the method of numerical analysis to establish a mathematical model between the macroscopic structure parameters and the mesoscopic structure parameters, and realize the correlation of the macro-mesoscopic structure parameters of the fiber-reinforced polyurea composite material;
[0049] Step 4, propose a mechanical property prediction framework for fiber-reinforced polyurea composite materials based on deep learning methods, and establish a one-way high-precision mapping relationship from the mesoscopic structure to the mechanical properties of fiber-reinforced polyurea composite materials, specifically:
[0050] Based on the three-dimensional mesoscopic fiber-reinforced polyurea composite material modeling plug-in and the finite element simulation method, the mesoscopic structure and the corresponding mechanical properties of the three-dimensional fiber-reinforced polyurea composite material are obtained, and a three-dimensional fiber-reinforced polyurea composite material deep learning data set is established; use a multi-dimensional convolutional neural network to identify and extract the unique three-dimensional mesoscopic structure characteristics of the fiber-reinforced polyurea composite material, and improve the prediction efficiency and accuracy of the mechanical properties of the mesoscopic fiber-reinforced polyurea composite material through deep learning methods.
[0051] Examples
[0052] A method for predicting the mechanical properties of fiber-reinforced polyurea composites based on deep learning, comprising the following steps:
[0053] Step 1, as Figure 2 A process of establishing a two-dimensional mesoscopic fiber-reinforced polyurea composite model for the structural characteristics of fiber-reinforced polyurea composites based on an image recognition method.
[0054] Based on the Retinex algorithm of the OpenCV library called by the Python programming language, a boundary recognition program for fiber-reinforced polyurea composites was written to extract the characteristics of the internal mesoscopic structure (number of fiber layers, ply angle, polyurea volume fraction). This program mainly includes the following key steps:
[0055] (1) Image high-light removal. Based on the brightness change function in the OpenCV library, the bright spots reflected on the surface structure of the fiber-reinforced polyurea composite sample were removed, the illumination and reflection components were separated, and the influence of high light was reduced.
[0056] (2) Gaussian filtering. The cv2.GaussianBlur function of the OpenCV library was called to generate a Gaussian kernel according to σ and the kernel size, and the Gaussian kernel was convolved with the scanned image of the fiber-reinforced polyurea composite to obtain a smoothed image.
[0057] (3) Edge detection. The Sobel operator was used to extract the outer contour line of the fiber-reinforced polyurea composite structure. Two 3x3 kernels were used to calculate the horizontal and vertical gradients respectively, and the edge intensity was combined. Only the thinner boundary lines located at the actual edges in the fiber-reinforced polyurea composite structure were retained.
[0058] (4) Filling. After the boundary curve of the fiber-reinforced polyurea composite structure was extracted, the drawContours function of the OpenCV library was called to draw and fill the closed contour.
[0059] Step 2, a three-dimensional fiber-reinforced polyurea composite model was established based on the Lagrangian model: Based on the Lagrangian model, a random fiber and solid model generation algorithm was added, and a modeling method for randomly woven three-dimensional mesoscopic solid fiber-reinforced polyurea composites was proposed. Specifically:
[0060] (1) Generating starting points. An algorithm for generating the Lagrangian model was written based on the Python programming language. The algorithm flow of the Lagrangian model was as follows: First, a Cartesian coordinate system was defined in the spatial region, and 100 - 1000000 control points (determined according to the number of fiber cross-sections recognized) were randomly generated to calibrate the area where each fiber is located; subsequently, the generated control points all need to meet the requirement of the fiber diameter distance between the control points; some model parameters need to be set when establishing the Lagrangian model, including the model space size, the number of control points N, and the minimum distance D between the control points minetc., are calculated using the following formula:
[0061]
[0062] where L is the Lagrangian function and λ is the Lagrange multiplier. represents the coordinates of the i-th control point, and r j represents the coordinates of the j-th control point, which is used to measure the randomness generated by the model. Its value ranges from 0 to 1, and 0.2 is taken.
[0063] (2) Random fiber braiding algorithm, based on Perlin noise, is used to simulate more natural fiber bending. Each fiber moves along a certain direction while randomly changing the direction to form a natural braiding effect. Cubic Hermite interpolation (Fade Function) is used for smoothing:
[0064] f(t) = 6t 5 - 15t 4 + 10t 3
[0065] where t is the normalized coordinate in the grid cell.
[0066] Let be S 00 , S 10 , S 01 , S 11 the noise values of the four lattice points respectively. The interpolation formula:
[0067] N(x, y) = Lerp(Lerp(S 00 , S 10 , u), Lerp(S 01 , S 11 , u), v)
[0068] (3) Establishment of a three-dimensional mesoscopic fiber-reinforced polyurea composite material model. The Lagrangian model is visualized in 3D fiber braiding - Matplotlib 3D, and Perlin noise is added to make the 3D fibers more natural for subsequent finite element simulation.
[0069] Step 3, as Figure 3 shown, for the performance prediction technical route of the three-dimensional microstructure of the composite material, a relationship between the macroscopic structure parameters and the mesoscopic structure parameters is established using numerical analysis methods, realizing the correlation of the macro-mesoscopic structure parameters of the fiber-reinforced polyurea composite material.
[0070] Data representation and feature engineering are performed on the mechanical property dataset of the composite material, specifically including normalization, regularization processing, and principal component analysis. Dimensionality reduction is carried out by selecting the input features most relevant to the macroscopic mechanical properties to improve the efficiency and prediction accuracy of the machine learning model. Based on the dimensionality-reduced dataset, with mesoscopic parameters such as the number of mesoscopic fiber layers, ply angles, and polyurea volume fraction as inputs and the macroscopic mechanical properties of the composite material as outputs, a gradient boosting tree model, a machine learning non-linear model, is selected for regression prediction to establish a machine learning-based composite material strength prediction model( Figure 3 ).
[0071] Step 4, as Figure 4 shown, a mechanical property prediction framework for fiber-reinforced polyurea composites is proposed based on the deep learning method, and a one-way high-precision mapping relationship from the mesoscopic structure to the mechanical properties of the fiber-reinforced polyurea composites is established, specifically as follows:
[0072] Based on a three-dimensional mesoscopic fiber-reinforced polyurea composite modeling plug-in and the finite element simulation method, the mesoscopic structure and corresponding mechanical properties of the three-dimensional fiber-reinforced polyurea composites are obtained, and a deep learning dataset for the three-dimensional fiber-reinforced polyurea composites is established; a multi-dimensional convolutional neural network is used to identify and extract the unique three-dimensional mesoscopic structure features of the fiber-reinforced polyurea composites, and the prediction efficiency and accuracy of the mechanical properties of the mesoscopic fiber-reinforced polyurea composites are improved through the deep learning method.
[0073] (1) Generate a geometric model of the fiber-reinforced polyurea composite through a three-dimensional mesoscopic fiber-reinforced polyurea composite modeling plug-in, and convert it into a three-dimensional voxel model of the fiber-reinforced polyurea composite that can be recognized by the neural network through point clouding and voxelization methods.
[0074] (2) Mechanical property calculation. Based on the finite element simulation method, use finite element simulation software to obtain various mechanical property parameters of the fiber-reinforced polyurea composite established in step (1) through simulation calculation, where the mechanical property parameters include specific energy absorption and elongation.
[0075] (3) Establishment of the deep learning dataset. The deep learning dataset consists of the voxel model of the fiber-reinforced polyurea composite and the mechanical properties obtained in step (2), and a data augmentation method is used during the establishment process to expand the capacity of the dataset.
[0076] (4) Establishment and training of the deep learning model. Establish a deep learning model containing a multi-dimensional convolutional neural network, and use the deep learning dataset of the fiber-reinforced polyurea composite established in step (3) for the training and validation of the deep learning model.
[0077] (5) Deep learning model testing. Establish a deep learning test dataset, and use the trained deep learning model to predict the mechanical properties of fiber-reinforced polyurea composites in the test dataset to verify the effectiveness of the deep learning model.
Claims
1. A method for predicting mechanical properties of fiber-reinforced polyurea composites based on deep learning, characterized in that: The steps include: Step (1): establishing a two-dimensional microscopic fiber-reinforced polyurea composite model based on an image recognition method; Step (2): establishing a three-dimensional fiber reinforced polyurea composite model based on the Lagrangian model; Step (3): using numerical analysis methods, a mathematical model between macroscopic structural parameters and microscopic structural parameters was established, and the macroscopic-microscopic structural parameter correlation of fiber reinforced polyurea composite materials was achieved; Step (4): establishing a voxel model of a three-dimensional fiber-reinforced polyurea composite material and corresponding mechanical properties, and obtaining a deep learning dataset of a three-dimensional fiber-reinforced polyurea composite material; establishing a deep learning model including a multidimensional convolutional neural network, and using the deep learning dataset of a three-dimensional fiber-reinforced polyurea composite material for training; Step (5): Predicting and obtaining the mechanical properties of the fiber-reinforced polyurea composite material based on a deep learning model including a multidimensional convolutional neural network.
2. The method according to claim 1, characterized in that Step (1) Based on the Python programming language, the Retinex algorithm of the OpenCV library is called to write a fiber-reinforced polyurea composite material boundary recognition program to extract the internal microstructure features. The detailed steps are as follows: Step (11): Image highlight removal: Based on the brightness change function in the OpenCV library, the bright spots reflected from the surface structure of the fiber-reinforced polyurea composite material sample are removed to separate the illumination and reflection components; Step (12): Gaussian filtering: calling the cv2.GaussianBlur function of the OpenCV library, generating a Gaussian kernel according to σ and the kernel size, and convolving the Gaussian kernel with the scanned image of the fiber-reinforced polyurea composite material to obtain a smoothed image; Step (13): Edge detection: Use the Sobel operator to extract the outer contour of the fiber-reinforced polyurea composite structure, use two 3x3 kernels to calculate the horizontal and vertical gradients respectively, combine them to obtain the edge strength, and only retain the thinner boundary lines located at the actual edge of the fiber-reinforced polyurea composite structure; Step (14): Filling: Call the drawContours function of the OpenCV library to draw and fill the closed contour after the boundary curve of the fiber-reinforced polyurea composite structure is extracted.
3. The method according to claim 2, characterized in that The internal microstructural characteristics in step (1) include the number of fiber layers, the ply angle and the polyurea volume fraction.
4. The method according to claim 3, characterized in that Step (2) is specifically as follows: Step (21): The two-dimensional microscopic fiber reinforced polyurea composite material model of step (1) is generated as a starting point, and a Lagrangian model generation algorithm is written based on Python. The Lagrangian model algorithm process is as follows: first, a Cartesian coordinate system is defined in the spatial region, and 100-1,000,000 control points are randomly generated to calibrate the region where each fiber is located; the generated control points must meet the requirements of the fiber diameter distance between the control points; when establishing the Lagrangian model, some model parameters need to be set, including the model space size, the number of control points N, and the minimum distance d between the control points. min , calculated using the following formula: Where L is the Lagrangian function, λ is the Lagrangian multiplier, represents the coordinates of the i-th control point, r j Represents the coordinates of the jth control point, which is used to measure the randomness of the model generation. Its value range is 0 to 1, and is 0.2; Step (22): Random fiber weaving algorithm: Perlin noise is used to simulate natural fiber bending, allowing each fiber to move along a set direction while randomly changing direction to form a natural weaving effect; cubic Hermite interpolation FadeFunction is used for smoothing: f(t)=6t 5 -15t 4 +10t 3 Where t is the normalized coordinate in the grid cell; Suppose they are 00 , S 10 , S 01 , S 11 Noise values at four grid points, interpolation formula: N(x,y)=Lerp(Lerp(S 00 ,S 10 ,u),Lerp(S 01 ,S 11 ,u),v); Step (23): Perform 3D fiber weaving-Matplotlib 3D visualization on the Lagrangian model to obtain a three-dimensional microscopic fiber reinforced polyurea composite model.
5. The method according to claim 4, characterized in that Step (3) is specifically as follows: performing data representation and feature engineering on the composite material mechanical properties data set through normalization, regularization and principal component analysis, and reducing the dimension of the data set by selecting the input features most relevant to the macroscopic mechanical properties; based on the reduced-dimensional data set, taking the microscopic parameters as input and the macroscopic mechanical properties of the composite material as output, selecting the gradient boosting tree model machine learning nonlinear model for regression prediction to obtain the relationship between the macroscopic properties and the microscopic structural parameters.
6. The method according to claim 5, characterized in that The input mesoscopic parameters include the number of mesoscopic fiber layers, ply angle, and polyurea volume fraction.
7. The method according to claim 6, characterized in that Step (4) is specifically as follows: Step (41): Generate a fiber-reinforced polyurea composite material geometric model through a three-dimensional microscopic fiber-reinforced polyurea composite material modeling plug-in, and convert it into a three-dimensional fiber-reinforced polyurea composite material voxel model that can be recognized by a neural network through point cloud and voxel methods; Step (42): Mechanical property calculation; various mechanical property parameters of the three-dimensional fiber reinforced polyurea composite voxel model established in step (41) are obtained by finite element simulation calculation; Step (43): Establishing a deep learning dataset: The deep learning dataset consists of a fiber reinforced polyurea composite voxel model and the mechanical properties obtained in step (2); Step (44): Establishing and training a deep learning model: Establishing a deep learning model including a multi-dimensional convolutional neural network, and using the data set obtained in step (3) for training and verifying the deep learning model; Step (45): Deep learning model testing: Establish a deep learning test data set, use the trained deep learning model to predict the mechanical properties of the fiber reinforced polyurea composite material in the test data set, and verify the effectiveness of the deep learning model.
8. The method according to claim 7, characterized in that The mechanical property parameters in step (42) include specific energy absorption and elongation.
9. A fiber-reinforced polyurea composite material, characterized in that: The method has the mechanical properties predicted by any one of claims 1 to 8.
10. Use of the fiber-reinforced polyurea composite material according to claim 9, characterized in that: Used for armored vehicle penetration protection.