A method and system for rapid prediction of leakage characteristics of a composite material

Abstract

The present application relates to the technical field of material leakage characteristics prediction, and particularly relates to a composite material leakage characteristics fast prediction method and system. The method comprises the following steps: obtaining CT scan data of a composite material, constructing a plurality of composite material image models according to the CT scan data; simulating leakage of the composite material image models by using a leakage analysis model to obtain a composite material leakage rate; combining the composite material image models and the composite material leakage rate to establish a composite material leakage characteristics data set; constructing a composite material leakage characteristics prediction model, training and verifying the composite material leakage characteristics prediction model through the composite material leakage characteristics data set; and based on the trained composite material leakage characteristics prediction model, completing fast prediction of the composite material leakage characteristics. The present application establishes a composite material leakage characteristics prediction model and a training data set based on in-situ CT scan results, and realizes efficient and accurate prediction of the leakage performance of the composite material.

Description

Technical Field

[0001] This invention relates to the field of material leakage characteristic prediction technology, and specifically to a method and system for rapid prediction of the leakage characteristics of composite materials. Background Technology

[0002] Cryogenic propellant tanks are a crucial component of launch vehicles. Most mainstream spacecraft worldwide are attempting to utilize composite materials to reduce overall weight and production costs, and some developed countries have successfully applied them to cryogenic propellant tanks. Cryogenic propellants, such as liquid hydrogen and liquid oxygen, have storage temperatures of -183℃ and -252℃, respectively. These extreme temperatures significantly impact the mechanical and leakage properties of composite materials. Specifically, microcracks and interfacial delamination caused by thermal expansion mismatches easily occur within composite structures, leading to propellant leakage. This presents new challenges to the safety and reliability design of lightweight composite material structures for spacecraft equipment. Therefore, it is urgent to establish the evolution law of composite material leakage characteristics under cryogenic environments and develop efficient prediction methods to provide fundamental theoretical and key technological support for the structural design of composite material tanks.

[0003] Currently, the prediction of leakage characteristics of composite materials mainly relies on complex numerical models, such as computational fluid dynamics models, which are computationally expensive. While numerical simulation and theoretical methods can effectively reduce research costs, these methods are not suitable for predicting the leakage behavior of composite materials with complex geometric cracks, and the computational cost remains high. Therefore, finding a method applicable to composite materials with complex geometric cracks that can conveniently and quickly provide accurate predictions of their leakage performance is a pressing issue that needs to be addressed. Summary of the Invention

[0004] To address the shortcomings of existing methods and the needs of practical applications, and in order to improve the adaptability to composite materials with complex geometric cracks and solve the problem of conveniently and quickly predicting the leakage performance of composite materials, this invention provides a rapid prediction method for the leakage characteristics of composite materials, comprising the following steps:

[0005] This invention involves acquiring CT scan data of composite materials and constructing multiple composite material image models based on the CT scan data. A leakage analysis model is used to simulate leakage in the composite material image models to obtain the composite material leakage rate. A composite material leakage characteristic dataset is established by combining the composite material image models and the composite material leakage rate. A composite material leakage characteristic prediction model is constructed, and the model is trained and validated using the composite material leakage characteristic dataset. Based on the trained composite material leakage characteristic prediction model, rapid prediction of composite material leakage characteristics is achieved. This invention establishes a composite material leakage characteristic dataset based on in-situ CT scan results, then trains a composite material leakage characteristic prediction model. Through a machine learning model, it efficiently and accurately predicts the leakage performance of composite materials, improving the adaptability to composite materials with complex geometric cracks and solving the problem of convenient and rapid accurate prediction of composite material leakage performance.

[0006] Optionally, constructing multiple composite material image models based on the CT scan data includes the following steps:

[0007] Statistical analysis of the CT scan data yields crack parameters at different layup angles within the scanned area. Based on these crack parameters, multiple two-dimensional images are generated. These two-dimensional images are then stacked according to the layup sequence to obtain an image model of the composite material. This invention improves the adaptability to composite materials with complex geometric crack shapes by statistically analyzing CT scan data and generating multiple two-dimensional images.

[0008] Optionally, the step of combining the composite material image model and the composite material leakage rate to establish a composite material leakage characteristic dataset includes the following steps:

[0009] The composite material image model is standardized to obtain a standardized result; the composite material leakage rate is normalized to obtain a normalized result; and the standardized and normalized results are combined to establish a dataset of the composite material leakage characteristics. This invention improves the stability and computational efficiency of the invention by standardizing and normalizing the data.

[0010] Optionally, the standardization process of the composite material image model satisfies the following formula:

[0011] ,

[0012] in, This represents the value after standardization. This represents the data before standardization. Represents the given mean. This represents the given standard deviation;

[0013] The leakage rate of the composite material is normalized to satisfy the following formula:

[0014] ,

[0015] in, This represents the value after normalization. This represents the value before normalization. Indicates the minimum leakage rate. This indicates the maximum leakage rate.

[0016] Optionally, training and validating the composite material leakage characteristic prediction model using the composite material leakage characteristic dataset includes the following steps:

[0017] The composite material image model is convolutionally processed to obtain image features. These features are then vectorized, and positional information is added to the vectorization result to calculate an attention value. Physical information is further added to the vectorization result to obtain physical constraint features. The image features, attention value, physical constraint features, and the composite material leakage characteristic dataset are combined to train and validate the composite material leakage characteristic prediction model. This invention integrates positional information into the dataset, preserving the layup order information to obtain more representative image features. Integrating physical information into the dataset ensures that a sufficiently accurate model can be trained with limited training data.

[0018] Optionally, the image features obtained by performing a convolution operation on the composite material image model satisfy the following formula:

[0019] ,in, Representing an image The gray value after positional convolution. Indicates the height of the convolution kernel. Indicates the width of the convolution kernel. Indicates height width The weights of the convolution kernel, Representing an image The grayscale value before positional convolution. This invention utilizes convolution operations to perform row-by-row and column-by-column convolution operations on the image matrix.

[0020] Optionally, adding positional information to the vectorized result satisfies the following formula:

[0021] ,in, Indicates the first Even-numbered columns of rows Indicates the first In odd-numbered columns of a row, PE represents an encoding containing location information. This indicates the dimension of the input matrix. This invention uses fixed-position encoding based on sine and cosine functions to add positional information, which can convert information at any position into a linear combination of previous positions, thus helping to preserve the information of the layering order during training.

[0022] Optionally, the calculation of the attention value satisfies the following formula:

[0023] ,

[0024] in, Indicates the attention value. This represents a query in the Transformer model. Represents the key in the Transformer model. This represents the value in the Transformer model. Indicates transpose. express The transpose of the matrix, Indicates the length of the query. This represents the normalized exponential function. The attention value of this invention retains information about the layup sequence of the composite laminate, which is beneficial to the training accuracy of subsequent steps.

[0025] Optionally, physical information is added to the vectorization result to obtain physical constraint features, satisfying the following formula:

[0026] ,

[0027] in, Indicates the permeability parameter of porous media. Indicates porosity. Represents the Kozeny constant. This represents the specific surface area. This invention adds physical information for estimating the permeability of porous media, further ensuring that a sufficiently accurate model can be trained with limited training data.

[0028] Secondly, to efficiently execute the rapid prediction method for composite material leakage characteristics provided by this invention, this invention also provides a rapid prediction system for composite material leakage characteristics, including a processor, an input device, an output device, and a memory. The processor, input device, output device, and memory are interconnected. The memory stores a computer program containing program instructions. The processor is configured to call the program instructions to execute the rapid prediction method for composite material leakage characteristics as described in the first aspect of this invention. The rapid prediction system for composite material leakage characteristics of this invention has a compact structure and stable performance, and can stably execute the rapid prediction method for composite material leakage characteristics provided by this invention, further improving the overall applicability and practical application capability of this invention. Attached Figure Description

[0029] Figure 1 A flowchart of a rapid prediction method for leakage characteristics of composite materials provided in an embodiment of the present invention;

[0030] Figure 2 This is a schematic diagram illustrating the composite material image model construction process provided in an embodiment of the present invention;

[0031] Figure 3 A framework diagram of a rapid prediction system for leakage characteristics of composite materials provided in an embodiment of the present invention;

[0032] Figure 4 This is a schematic diagram of a device for rapid prediction of leakage characteristics of composite materials, provided in an embodiment of the present invention. Detailed Implementation

[0033] Specific embodiments of the present invention will now be described in detail. It should be noted that the embodiments described herein are for illustrative purposes only and are not intended to limit the invention. In the following description, numerous specific details are set forth in order to provide a thorough understanding of the invention. However, it will be apparent to those skilled in the art that these specific details are not necessary to practice the invention. In other instances, well-known circuits, software, or methods have not been specifically described to avoid obscuring the invention.

[0034] Throughout this specification, references to "an embodiment," "an embodiment," "an example," or "an example" mean that a particular feature, structure, or characteristic described in connection with that embodiment or example is included in at least one embodiment of the invention. Therefore, the phrases "in an embodiment," "in an embodiment," "an example," or "an example" appearing in various places throughout the specification do not necessarily refer to the same embodiment or example. Furthermore, specific features, structures, or characteristics can be combined in one or more embodiments or examples in any suitable combination and / or sub-combination. Moreover, those skilled in the art will understand that the illustrations provided herein are for illustrative purposes and are not necessarily drawn to scale.

[0035] Please see Figure 1 To improve the adaptability of composite materials with complex geometric cracks and to solve the problem of convenient and rapid accurate prediction of the leakage performance of composite materials, this invention provides a rapid prediction method for the leakage characteristics of composite materials, such as... Figure 1 As shown, the method includes the following steps:

[0036] S1. Obtain CT scan data of the composite material, and construct multiple composite material image models based on the CT scan data.

[0037] Please see Figure 2 In this embodiment, step S1, which involves constructing multiple composite material image models based on the CT scan data, includes the following steps:

[0038] S11. Perform statistical analysis on the CT scan data to obtain crack parameters at different layup angles within the scanned area.

[0039] Specifically, in-situ CT scan data of composite materials are obtained from CT equipment, and then the scan data is statistically analyzed to obtain crack parameters for different layup angles in the scanned area. Crack parameters include size, shape, volume fraction and spatial distribution.

[0040] S12. Based on the crack parameters, generate multiple two-dimensional images.

[0041] Use Matlab software to create a 2D image containing crack features, where the crack length is randomly distributed within the ply, the crack width is set to the average crack width of the layer, and the crack thickness completely penetrates the corresponding ply.

[0042] Furthermore, multiple two-dimensional images are generated by randomly combining different crack densities and layup angles.

[0043] S13. Stack the two-dimensional images in the layup order to obtain the composite material image model.

[0044] Specifically, two-dimensional images are stacked together according to the layup order to create a composite material image model.

[0045] S2. Use the leakage analysis model to simulate the leakage of the composite material image model and obtain the leakage rate of the composite material.

[0046] In this embodiment, the leakage analysis model is Avizo software. The composite material image model obtained in step S1 is imported into Avizo software, and its leakage analysis module is used to simulate leakage of the image model to obtain the absolute leakage of laminates with different layup sequences. The calculation results are converted using Darcy's law to obtain the leakage rate of the composite material.

[0047] Furthermore, the leakage rate of composite materials can be calculated in batches and obtained in Avizo software by writing Python scripts.

[0048] S3. Combine the composite material image model and the composite material leakage rate to establish a composite material leakage characteristic dataset.

[0049] In this embodiment, step S3, which combines the composite material image model and the composite material leakage rate to establish a composite material leakage characteristic dataset, includes the following steps:

[0050] S31. Standardize the composite material image model to obtain the standardization result.

[0051] Specifically, the composite material image model is standardized to satisfy the following formula:

[0052] ,

[0053] in, This represents the value after standardization. This represents the data before standardization. Represents the given mean. This represents the given standard deviation.

[0054] S32. Normalize the leakage rate of the composite material to obtain the normalization result.

[0055] Specifically, the leakage rate of the composite material is normalized to satisfy the following formula:

[0056]

[0057] in, This represents the value after normalization. This represents the value before normalization. Indicates the minimum leakage rate. This represents the maximum leakage rate. Using the Min-Max normalization method to normalize the leakage rate data of composite laminates with different ply layups allows the data to be mapped to the [0, 1] interval, which helps improve computational efficiency.

[0058] S33. Combine the standardized processing results and the normalized processing results to establish the composite material leakage characteristic dataset.

[0059] Organize the standardized and normalized data into a dataset. This dataset should include all the features used for analysis, i.e., independent variables, and the measurement results of leakage characteristics, i.e., dependent variables.

[0060] S4. Construct a composite material leakage characteristic prediction model, and train and validate the composite material leakage characteristic prediction model using the composite material leakage characteristic dataset.

[0061] In this embodiment, the composite material leakage characteristic prediction model constructed in step S4 is a machine learning model. Preferably, the composite material leakage characteristic prediction model is a two-dimensional convolutional neural network with an attention module.

[0062] The basic structure of a two-dimensional convolutional neural network includes an input layer, a convolutional layer, an activation function, a pooling layer, a fully connected layer, and an output layer. The attention module is a Transformer model, whose network structure mainly consists of an encoder and a decoder, both of which contain attention layers and feedforward neural network layers. By utilizing the attention layer and position encoding module of the Transformer model, information about the layup sequence of composite laminates can be preserved.

[0063] By utilizing the attention layer and position encoding module of the Transformer model, information about the layup sequence of composite laminates can be preserved. A two-dimensional convolutional neural network is used to learn from the image, extract crack features, and combine the position encoding function of the Transformer model to record the layup sequence, thereby establishing a predictive model for the leakage characteristics of composite materials that takes into account the layup sequence.

[0064] Further, step S4, which involves training and validating the composite material leakage characteristic prediction model using the composite material leakage characteristic dataset, includes the following steps:

[0065] S41. Perform a convolution operation on the composite material image model to obtain image features.

[0066] The composite material image model is read into PyTorch software, and its corresponding leakage rate is recorded as a label value; the read image is scaled to 224×224 pixels and then converted into a tensor that can be processed by a computer.

[0067] For an image matrix of size M×N, convolution is performed using a convolution kernel of size U×V. Further, the convolution operation is performed on the composite material image model to obtain image features that satisfy the following formula:

[0068] ,in, Representing an image The gray value after positional convolution. Indicates the height of the convolution kernel. Indicates the width of the convolution kernel. Indicates height width The weights of the convolution kernel, Representing an image Gray values ​​before positional convolution.

[0069] pass and The changes are implemented to achieve row-by-row and column-by-column convolution operations on the image matrix; secondly, the ReLU function is selected as the activation function, and the max pooling method is used to retain the maximum value in the region to obtain more representative image features.

[0070] S42. Vectorize the image features, add positional information to the vectorization result, and then calculate the attention value.

[0071] By utilizing the attention layer and position encoding module of the Transformer model, information on the layup sequence of the composite laminate is preserved. The attention calculation is based on the image features extracted by convolution, which are then unfolded into a one-dimensional vector and used as the input to the attention layer.

[0072] Furthermore, the addition of positional information to the vectorized result satisfies the following formula:

[0073] ,

[0074] in, Indicates the first Even-numbered columns of rows Indicates the first In odd-numbered columns of a row, PE represents an encoding containing location information. This indicates the dimension of the input matrix.

[0075] Understandably, this assumes the input is an n-layer sequence. For a 3D matrix, the positional encoding is a matrix of the same size added to the input matrix. The elements in the pos-th row, 2i-th column and the (2i+1)-th column can be calculated using the above formula. Simultaneously, trigonometric functions can be used with the sum-to-product formula to transform information at any position into a linear combination of previous positions. This allows the information of the layering order to be preserved during training.

[0076] Furthermore, the calculation of the attention value satisfies the following formula:

[0077] ,

[0078] in, Indicates the attention value. This represents a query in the Transformer model. Represents the key in the Transformer model. This represents the value in the Transformer model. Indicates transpose. express The transpose of the matrix, Indicates the length of the query. This represents the normalized exponential function.

[0079] S43. Add physical information to the vectorized result to obtain physical constraint features.

[0080] Specifically, step S43 involves adding physical information to the vectorized result to obtain physical constraint features, satisfying the following formula:

[0081] ,

[0082] in, Indicates the permeability parameter of porous media. Indicates porosity. Represents the Kozeny constant. It represents the specific surface area.

[0083] Physical information is added to the fully connected layer after convolution and calculated in the loss function. After feature extraction via tensor convolution in the composite material image model, it is unfolded into a one-dimensional vector, and porosity and specific surface area are added to the last part of the vector for physical information calculation.

[0084] S44. Combine the image features, the attention value, the physical constraint features, and the composite material leakage characteristic dataset to train and validate the composite material leakage characteristic prediction model.

[0085] In this embodiment, the image features, the attention value, the physical constraint features, and the composite material leakage characteristic dataset are combined to construct a data training set and a data validation set in proportion, and then the composite material leakage characteristic prediction model is trained and validated.

[0086] In this embodiment, the loss function of the composite material leakage characteristic prediction model is the root mean square error, which satisfies the following formula:

[0087] ,in, This represents the root mean square error. Indicates the number of observations. Indicates the first The true value of each observation. Indicates the first The predicted value of each observation.

[0088] Furthermore, the coefficient of determination (COD) was used to evaluate and validate the prediction model for the leakage characteristics of composite materials. The COD value ranges from 0 to 1; the closer the value is to 1, the better the model's fit. The formula is as follows:

[0089] ,

[0090] in, The coefficient of determination is represented by the coefficient of determination. This represents the average of the true values.

[0091] It should be understood that in order to avoid convergence difficulties caused by the order of magnitude problem, normalization is also required in each layer of training, i.e., batch normalization. In order to avoid overfitting of the neural network and loss of generalization ability, the dropout method is used to add some noise interference during training to improve the network model's anti-interference ability.

[0092] S5. Based on the trained composite material leakage characteristic prediction model, complete the rapid prediction of composite material leakage characteristics.

[0093] The CT data of the composite material to be predicted is obtained, and the leakage characteristics of the composite material are predicted using a trained composite material leakage characteristic prediction model, thus efficiently and accurately obtaining the leakage characteristics of the composite material to be predicted.

[0094] This invention uses statistical information from CT images to predict the leakage performance of composite laminates. It establishes a composite leakage image dataset and trains a neural network, which does not require high-precision CT scans or complex mesh generation, avoiding the problems of poor mesh quality and difficulty in convergence of traditional models. This greatly reduces the computational cost and enables rapid acquisition of the leakage performance of composite materials, opening up new prospects for rapidly estimating the leakage rate of different laminates.

[0095] Secondly, the composite material leakage characteristic prediction model established in this invention can calculate the leakage of composite laminates based on crack images of different sequences. Embedding the physical model can accelerate the convergence speed of the training process and improve accuracy. Compared with other neural network models or machine learning models, this invention can predict the leakage of composite materials faster while ensuring accuracy.

[0096] Please see Figure 3In an embodiment, to efficiently execute the rapid prediction method for composite material leakage characteristics provided by the present invention, the present invention also provides a rapid prediction system for composite material leakage characteristics, comprising: an input device, an output device, a processor, and a memory, wherein the input device, output device, processor, and memory are interconnected, and the memory contains program instructions for the steps of the rapid prediction method for composite material leakage characteristics. The rapid prediction system for composite material leakage characteristics of the present invention has a compact structure and stable performance, and can stably execute the rapid prediction method for composite material leakage characteristics of the present invention, further improving the overall applicability and practical application capability of the present invention.

[0097] In embodiments, the processor may be a Central Processing Unit (CPU), but it can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor. Input devices can be used to acquire data information. Output devices can be used to output the results obtained by storing program instructions contained in a computer program in the memory provided by this invention. The memory may include read-only memory and random access memory, and provides instructions and data to the processor. A portion of the memory may also include non-volatile random access memory.

[0098] In yet another alternative embodiment, please refer to Figure 4 To efficiently implement the rapid prediction method for the leakage characteristics of composite materials provided by this invention, this embodiment also provides a rapid prediction device for the leakage characteristics of composite materials, such as... Figure 4 As shown, it includes:

[0099] The memory 10 stores the computer program; the processor 20 executes the computer program to implement the aforementioned method for rapid prediction of leakage characteristics of composite materials. The system includes the memory 10, processor 20, communication interface 31, and communication bus 32. The memory 10, processor 20, and communication interface 31 all communicate with each other via the communication bus 32.

[0100] In this embodiment, the memory 10 is used to store one or more program instructions. The memory 10 may store program instructions for implementing the following functions:

[0101] Acquire CT scan data of composite materials, and construct multiple composite material image models based on the CT scan data; simulate leakage using a leakage analysis model to obtain the composite material leakage rate; combine the composite material image models and the composite material leakage rate to establish a composite material leakage characteristic dataset; construct a composite material leakage characteristic prediction model, train and validate the composite material leakage characteristic prediction model using the composite material leakage characteristic dataset; and perform rapid prediction of composite material leakage characteristics based on the trained composite material leakage characteristic prediction model.

[0102] In one possible implementation, memory 10 may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created during use. Furthermore, memory 10 may include read-only memory and random access memory, providing instructions and data to the processor. A portion of the memory may also include NVRAM. The memory stores the operating system and operating instructions, executable modules, or data structures, or subsets thereof, or extended sets thereof. The operating instructions may include various operation instructions for implementing various operations. The operating system may include various system programs for implementing various basic tasks and handling hardware-based tasks.

[0103] Processor 20 can be a Central Processing Unit (CPU), an application-specific integrated circuit, a digital signal processor, a field-programmable gate array, or other programmable logic device. Processor 20 can be a microprocessor or any conventional processor. Processor 20 can call programs stored in memory 10. Communication interface 31 can be an interface for a communication module, used to connect with other devices or systems.

[0104] Of course, it should be noted that, Figure 4 The structure shown does not constitute a limitation on the rapid prediction device for composite material leakage characteristics in this embodiment. In practical applications, the rapid prediction device for composite material leakage characteristics may include more advanced technologies. Figure 4 More or fewer components as shown, or combinations of certain components.

[0105] The embodiment also provides a storage medium storing a computer program, which, when executed by a processor, implements the steps of the above-described method for rapidly predicting the leakage characteristics of composite materials.

[0106] The storage medium can include various media that can store program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0107] In summary, this invention establishes a composite material leakage characteristic dataset based on in-situ CT scan results, then trains a composite material leakage characteristic prediction model. Through machine learning, it efficiently and accurately predicts the leakage performance of composite materials, improving its adaptability to composite materials with complex geometric cracks and solving the problem of conveniently and quickly predicting the leakage performance of composite materials accurately. Therefore, this invention effectively overcomes the various shortcomings of existing technologies and has high industrial application value.

[0108] 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 the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be covered within the scope of the present invention.

Claims

1. A method for rapid prediction of leakage characteristics of composite materials, characterized in that, The rapid prediction method for leakage characteristics of composite materials includes the following steps: Acquire CT scan data of composite materials, and construct multiple composite material image models based on the CT scan data; The leakage rate of the composite material was obtained by simulating the leakage of the composite material image model using a leakage analysis model. By combining the composite material image model and the composite material leakage rate, a composite material leakage characteristic dataset is established; A composite material leakage characteristic prediction model is constructed, and the composite material leakage characteristic prediction model is trained and validated using the composite material leakage characteristic dataset. Based on a trained composite material leakage characteristic prediction model, rapid prediction of composite material leakage characteristics is achieved. The process of constructing multiple composite material image models based on the CT scan data includes the following steps: Statistical analysis was performed on the CT scan data to obtain crack parameters at different layup angles within the scanned area; Based on the crack parameters, multiple two-dimensional images are generated; The two-dimensional images are stacked in a layer-layout order to obtain the composite material image model; The process of training and validating the composite material leakage characteristic prediction model using the composite material leakage characteristic dataset includes the following steps: The composite material image model is convolutionally processed to obtain image features; The image features are vectorized, and positional information is added to the vectorization result to calculate the attention value. Physical information is added to the vectorized results. By adding positional information, the information of the layup sequence of the composite laminate is preserved, and physical constraint characteristics are obtained. By combining the image features, the attention value, the physical constraint features, and the composite material leakage characteristic dataset, the composite material leakage characteristic prediction model is trained and validated. The step involves adding physical information to the vectorized result to obtain physical constraint features, satisfying the following formula: ， in, Indicates the permeability parameter of porous media. Indicates porosity. Represents the Kozeny constant. Representing specific surface area, adding physical information refers to adding porosity and specific surface area to the end of the vector for physical information calculation.

2. The rapid prediction method for leakage characteristics of composite materials according to claim 1, characterized in that, The process of establishing a composite material leakage characteristic dataset by combining the composite material image model and the composite material leakage rate includes the following steps: The composite material image model is standardized to obtain the standardized processing result; The leakage rate of the composite material was normalized to obtain the normalized result; By combining the results of the standardization process and the results of the normalization process, a dataset of leakage characteristics of the composite material is established.

3. The rapid prediction method for leakage characteristics of composite materials according to claim 2, characterized in that, The image model of the composite material is standardized to satisfy the following formula: ， in, This represents the value after standardization. This represents the data before standardization. Represents the given mean. This represents the given standard deviation; The leakage rate of the composite material is normalized to satisfy the following formula: ， in, This represents the value after normalization. This represents the value before normalization. Indicates the minimum leakage rate. This indicates the maximum leakage rate.

4. The rapid prediction method for leakage characteristics of composite materials according to claim 1, characterized in that, The image features obtained by performing a convolution operation on the composite material image model satisfy the following formula: in, Representing an image The gray value after positional convolution. Indicates the height of the convolution kernel. Indicates the width of the convolution kernel. Indicates height width The weights of the convolution kernel, Representing an image Gray values ​​before positional convolution.

5. The rapid prediction method for leakage characteristics of composite materials according to claim 1, characterized in that, The addition of positional information to the vectorized result satisfies the following formula: ， in, Indicates the first Even-numbered columns of rows Indicates the first In odd-numbered columns of a row, PE represents an encoding containing location information. This indicates the dimension of the input matrix.

6. The rapid prediction method for leakage characteristics of composite materials according to claim 1, characterized in that, The attention value is calculated according to the following formula: ， in, Indicates the attention value. This represents a query in the Transformer model. Represents the key in the Transformer model. This represents the value in the Transformer model. Indicates transpose. express The transpose of the matrix, Indicates the length of the query. This represents the normalized exponential function.

7. A rapid prediction system for the leakage characteristics of composite materials, characterized in that, The rapid prediction system for leakage characteristics of composite materials includes: an input device, an output device, a processor, and a memory, wherein the input device, the output device, the processor, and the memory are interconnected, and the memory includes program instructions for executing the rapid prediction method for leakage characteristics of composite materials according to any one of claims 1-6.

Images