Thermal protection material surface roughness prediction method, device, equipment and storage medium

Abstract

The present application provides a method, device, equipment and storage medium for predicting the surface roughness of thermal protection materials, which are used in the field of computer vision technology. The method includes: obtaining a data set; wherein the data set includes multiple two-dimensional surface roughness sample images constructed based on a thermochemical ablation theoretical model and a finite volume method to simulate the surface ablation morphology characteristics of a three-dimensional carbon-carbon composite material; constructing an MSCNN-BiLSTM attention mechanism prediction model; inputting the data set into the MSCNN-BiLSTM attention mechanism prediction model for iterative training to obtain a surface roughness prediction model; inputting the image to be detected into the surface roughness prediction model to obtain a surface roughness prediction result; wherein the image to be detected is obtained based on the carbon-carbon composite material to be detected. The method of the present application predicts the surface roughness of the carbon-carbon composite material through the MSCNN-BiLSTM attention mechanism prediction model, which significantly improves the accuracy and reliability of the prediction of the surface roughness of the thermal protection material.

Description

Technical Field

[0001] The present application relates to the field of computer vision technology, and in particular to a method, device, equipment and storage medium for predicting the surface roughness of a thermal protection material. Background Art

[0002] Carbon-carbon composites are widely used in thermal protection systems in aerospace and rockets due to their excellent high-temperature performance, lightweight characteristics and corrosion resistance, providing reliable protection for the safe operation of aircraft and spacecraft. When carbon-carbon composites are subjected to thermochemical ablation in a high-temperature environment, their surfaces will be impacted and oxidized by high-temperature gases, resulting in surface degradation and roughening. This change in surface roughness will affect the thermal conductivity and degree of oxidation / corrosion of the thermal protection system. Therefore, accurately predicting the change in surface roughness of carbon-carbon composites is of great significance to understanding the behavior mechanism of materials during thermochemical ablation.

[0003] At present, the thermochemical ablation mechanism of carbon-carbon composites and the surface decay caused by surface changes are mainly explored through experiments and simulations. In experiments, the ablation surface roughness of carbon-based ablators is obtained mainly by means of high enthalpy heat flow or arc jet, and then the change of surface roughness over time is quantitatively analyzed by high-resolution non-contact profilometer or optical technology. In simulation, the discrete variance or finite volume method is mainly used to simulate the change of surface roughness of three-dimensional carbon-carbon composites due to various factors. In previous studies, the experiments were costly and challenging. Numerical simulations are limited by model accuracy and computing resources and may not be able to fully and accurately simulate the actual ablation process.

[0004] Therefore, a new method for predicting the surface roughness of thermal protection materials is urgently needed to solve the problems encountered in the existing technology. Summary of the invention

[0005] The present application provides a method, device, equipment and storage medium for predicting the surface roughness of a thermal protection material, which are used to predict the problem of the surface roughness of a thermal protection material.

[0006] According to the first aspect disclosed in the present application, the present application provides a method for predicting the surface roughness of a thermal protection material, comprising:

[0007] Acquire a data set; wherein the data set includes a plurality of two-dimensional surface roughness sample images constructed based on a thermochemical ablation theoretical model and a finite volume method to simulate the surface ablation morphology characteristics of a three-dimensional carbon-carbon composite material;

[0008] Based on the MSCNN multi-scale convolutional neural network and the BiLSTM bidirectional long short-term memory network model, the MSCNN-BiLSTM attention mechanism prediction model is constructed;

[0009] Inputting the data set into the MSCNN-BiLSTM attention mechanism prediction model for iterative training to obtain a surface roughness prediction model;

[0010] The image to be detected is input into the surface roughness prediction model to obtain a surface roughness prediction result; wherein the image to be detected is obtained based on the carbon-carbon composite material to be detected.

[0011] In a feasible implementation manner, the acquiring of the data set includes:

[0012] Based on the carbon-carbon composite weaving structure and weaving process parameters, a carbon-carbon composite unit cell model is constructed;

[0013] The carbon-carbon composite unit cell model is simulated and solved by the finite volume method and the thermochemical ablation theoretical model to obtain a characteristic image of the surface ablation morphology evolution of the three-dimensional carbon-carbon composite material;

[0014] Based on the three-dimensional carbon-carbon composite material surface ablation morphology evolution characteristic image, a plurality of two-dimensional surface roughness sample images are obtained.

[0015] In a feasible implementation, based on the three-dimensional carbon-carbon composite material surface ablation morphology evolution characteristic image, a plurality of two-dimensional surface roughness sample images are obtained, including:

[0016] Extracting data from the characteristic image of the surface ablation morphology evolution of the three-dimensional carbon-carbon composite material based on a data point extraction algorithm to obtain a plurality of characteristic extraction data;

[0017] The feature extraction data is two-dimensionally transformed to obtain the two-dimensional surface roughness sample image.

[0018] In a feasible implementation, based on the MSCNN multi-scale convolutional neural network and the BiLSTM bidirectional long short-term memory network model, an MSCNN-BiLSTM attention mechanism prediction model is constructed, including:

[0019] Constructing a MSCNN convolutional neural network; wherein the MSCNN convolutional neural network includes a CNN input layer, a multi-scale feature extraction layer, a pooling layer, a fusion layer, a fully connected layer and a CNN output layer; the multi-scale feature extraction layer includes a first convolutional layer, a second convolutional layer and a third convolutional layer; the convolution kernel size of the first convolutional layer is 5*5; the convolution kernel size of the second convolutional layer is 3*3; the convolution kernel size of the third convolutional layer is 1*1;

[0020] Constructing a BiLSTM bidirectional long short-term memory network model; wherein the BiLSTM bidirectional long short-term memory network model includes an LSTM input layer, a forward LSTM, a backward LSTM, an activation function and an LSTM output layer; the forward LSTM and the backward LSTM both include a forget gate, a Tanh gate, an input gate and an output gate;

[0021] The output data of the CNN output layer is used as the input data of the LSTM input layer and subsequently processed to obtain the MSCNN-BiLSTM attention mechanism prediction model.

[0022] In a feasible implementation, the step of inputting the data set into the MSCNN-BiLSTM attention mechanism prediction model for iterative training to obtain a surface roughness prediction model includes:

[0023] Inputting the data set into the MSCNN multi-scale convolutional neural network for iterative training to obtain a CNN screening feature set;

[0024] Inputting the CNN screening feature set into the BiLSTM bidirectional long short-term memory network model for iterative training to obtain a training result;

[0025] The training result is introduced into the cross entropy loss function to obtain the test result;

[0026] The performance of the MSCNN-BiLSTM attention mechanism prediction model is evaluated, and when the test result meets the test condition, the training is terminated; otherwise, the network parameters of the MSCNN-BiLSTM attention mechanism prediction model are optimized according to the prediction result, and the data set is re-input into the MSCNN-BiLSTM attention mechanism prediction model for training until the test result meets the test condition, thereby obtaining the surface roughness prediction model.

[0027] In a feasible implementation, the performance evaluation of the MSCNN-BiLSTM attention mechanism prediction model includes:

[0028] Input the data set into the MSCNN-BiLSTM attention mechanism prediction model to obtain a test result;

[0029] Based on the test results, the mean absolute error, root mean square error and coefficient of determination were used to evaluate the MSCNN-BiLSTM attention mechanism prediction model;

[0030] If the mean absolute error and the root mean square error are smaller and the determination coefficient is closer to 1, then the test result meets the test conditions.

[0031] In a feasible implementation manner, the method further includes:

[0032] If the mean absolute error and the root mean square error are larger and the determination coefficient is closer to 0, the test result does not meet the test conditions.

[0033] According to a second aspect disclosed in the present application, the present application provides a device for predicting surface roughness of a thermal protection material, comprising:

[0034] A data acquisition module, used to acquire a data set; wherein the data set includes a plurality of two-dimensional surface roughness sample images constructed based on a thermochemical ablation theoretical model and a finite volume method to simulate the surface ablation morphology characteristics of a three-dimensional carbon-carbon composite material;

[0035] The model building module is used to build the MSCNN-BiLSTM attention mechanism prediction model based on the MSCNN multi-scale convolutional neural network and the BiLSTM bidirectional long short-term memory network model;

[0036] A model training module, inputting the data set into the MSCNN-BiLSTM attention mechanism prediction model for iterative training to obtain a surface roughness prediction model;

[0037] The surface roughness prediction module is used to input the image to be detected into the surface roughness prediction model to obtain a surface roughness prediction result; wherein the image to be detected is obtained based on the carbon-carbon composite material to be detected.

[0038] According to a third aspect disclosed in the present application, there is provided an electronic device, comprising a processor, and a memory communicatively connected to the processor;

[0039] The memory stores computer-executable instructions;

[0040] The processor executes the computer-executable instructions stored in the memory to implement the method according to any one of the first aspects.

[0041] According to a fourth aspect disclosed in the present application, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores computer-executable instructions, and when the computer-executable instructions are executed by a processor, they are used to implement any one of the methods described in the first aspect.

[0042] According to a fifth aspect disclosed in the present application, a computer program product is provided, including a computer program, wherein the computer program is used to implement any one of the methods in the first aspect when executed by a processor.

[0043] Compared with the prior art, this application has the following beneficial effects:

[0044] The present application provides a method, device, equipment and storage medium for predicting the surface roughness of thermal protection materials. First, the method uses a thermochemical ablation theory model and a finite volume method to simulate the surface ablation morphology characteristics of three-dimensional carbon-carbon composite materials to construct a plurality of two-dimensional surface roughness sample images to provide a data set. The data set provided by the method has higher accuracy, diversity and controllability, and at the same time, more physical knowledge is integrated into the generation process of the data set. On the basis of the data set, a multi-scale convolutional neural network is used to extract the ablation surface morphology of the carbon-carbon composite material at each moment, and then the BiLSTM mechanism is used to quickly and accurately learn the spatiotemporal characteristics and long-term relationships hidden in the time sequence law. The attention mechanism gives different weights to the learned features, so that the model pays more attention to the important time steps in the sequence and filters out redundant features; the method significantly improves the accuracy and reliability of the prediction of the surface roughness of thermal protection materials, while reducing the computational cost and time, and quickly predicts the evolution of the subsequent carbon-carbon composite surface morphology, more accurately evaluates the performance and reliability of the material, and provides guidance for material selection and structural design, thereby improving the reliability and safety of the thermal protection structure; at the same time, the method is also applicable to the surface roughness prediction of other different types of woven composite materials. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0046] Figure 1 A schematic diagram of a process for predicting the surface roughness of a thermal protection material provided in an embodiment of the present application;

[0047] Figure 2 A schematic flow chart of another method for predicting the surface roughness of a thermal protection material provided in an embodiment of the present application;

[0048] Figure 3 An example diagram of a two-dimensional gas-solid interface of the thermochemical ablation theory provided in an embodiment of the present application;

[0049] Figure 4 A schematic diagram of the structure of a unit cell model of a 2.5D layer-to-layer interlocking braided structure provided in an embodiment of the present application; wherein (a), (c), and (d) are cross sections in different directions; (b) is a schematic diagram of a cross section of a yarn;

[0050] Figure 5 A schematic diagram of a characteristic image of the surface ablation morphology evolution of a three-dimensional carbon-carbon composite material provided in an embodiment of the present application;

[0051] Figure 6 A schematic diagram of a sample image of a two-dimensional surface roughness of a carbon-carbon composite material provided in an embodiment of the present application;

[0052] Figure 7 A schematic diagram of the structure of a MSCNN multi-scale convolutional neural network provided in an embodiment of the present application;

[0053] Figure 8 A schematic diagram of the structure of a BiLSTM bidirectional long short-term memory network model provided in an embodiment of the present application;

[0054] Fig. 9 A schematic diagram of the structure of an LSTM memory network model provided in an embodiment of the present application;

[0055] Fig.10 A schematic diagram of the structure of a device for predicting the surface roughness of a thermal protection material provided in an embodiment of the present application;

[0056] Fig.11 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application.

[0057] The above drawings have shown clear embodiments of the present application, which will be described in more detail later. These drawings and text descriptions are not intended to limit the scope of the present application in any way, but to illustrate the concept of the present application to those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION

[0058] Here, exemplary embodiments are described in detail, and examples thereof are shown in the accompanying drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application.

[0059] Thermal protection materials have important application value in ballistic missiles, space shuttles, spacecraft reentry capsules, etc. Carbon-carbon composite materials are widely used in thermal protection systems in aerospace and rocket fields due to their excellent high temperature performance, lightweight characteristics and corrosion resistance, providing reliable protection for the safe operation of aircraft and spacecraft.

[0060] When carbon-carbon composites are subjected to thermochemical ablation in a high-temperature environment, their surfaces will be impacted and oxidized by high-temperature gases, resulting in surface degradation and roughening. This change in surface roughness will affect the thermal conductivity and degree of oxidation / corrosion of the thermal protection system. Therefore, accurately predicting the change in surface roughness of carbon-carbon composites is of great significance for understanding the behavior mechanism of materials during thermochemical ablation.

[0061] At present, the thermochemical ablation mechanism of carbon-carbon composites and the surface decay caused by surface changes are mainly explored through experiments and simulations. In experiments, the roughness of the ablated surface of the carbon-based ablator is obtained by means of high enthalpy heat flow or arc jet, and then the change of surface roughness over time is quantitatively analyzed by high-resolution non-contact profilometer or optical technology. Experiments require a lot of time and human resources, including sample preparation, experimental condition setting, data collection and analysis. In simulation, cubic discretization or finite volume method is mainly used to simulate the change of surface roughness of three-dimensional carbon-carbon composites due to various factors. At the same time, numerical simulation requires the establishment of complex mathematical models and a large number of calculations, especially for large-scale simulations, long time spans or complex situations, the time consumption of numerical simulation is more significant.

[0062] Therefore, in order to more realistically reflect the surface morphology evolution of carbon-carbon composites, deep learning technology is needed to supplement and verify the results of numerical simulation.

[0063] In response to the above technical problems, this application proposes a method for predicting the surface roughness of thermal protection materials. The surface roughness of carbon-carbon composite materials is predicted through the MSCNN-BiLSTM attention mechanism prediction model, which significantly improves the accuracy and reliability of the prediction of the surface roughness of thermal protection materials.

[0064] The following is a detailed description of the technical solution of the method for predicting the surface roughness of thermal protection materials provided by the present application through specific embodiments. It should be noted that the following embodiments may exist alone or in combination with each other, and the same or similar contents may not be repeated in different embodiments.

[0065] Figure 1 A schematic diagram of a method for predicting the surface roughness of a thermal protection material provided in an embodiment of the present application is shown in FIG. Figure 1 In some embodiments, the process of the method for predicting the surface roughness of the thermal protection material includes the following steps:

[0066] S101, acquiring a data set; wherein the data set includes a plurality of two-dimensional surface roughness sample images constructed based on a thermochemical ablation theoretical model and a finite volume method to simulate the surface ablation morphology characteristics of a three-dimensional carbon-carbon composite material;

[0067] S102, based on the MSCNN multi-scale convolutional neural network and BiLSTM bidirectional long short-term memory network model, build the MSCNN-BiLSTM attention mechanism prediction model;

[0068] S103, inputting the data set into the MSCNN-BiLSTM attention mechanism prediction model for iterative training to obtain a surface roughness prediction model;

[0069] Among them, the MSCNN-BiLSTM attention mechanism prediction model is iteratively trained using the data set to obtain a surface roughness prediction model, so as to predict the surface roughness of carbon-carbon composite materials using the surface roughness prediction model.

[0070] S104, inputting the image to be detected into the surface roughness prediction model to obtain a surface roughness prediction result; wherein the image to be detected is obtained based on the carbon-carbon composite material to be detected.

[0071] The surface roughness of the carbon-carbon composite material is predicted based on the trained surface roughness prediction model, so as to evaluate the surface roughness of the carbon-carbon composite material through the surface roughness prediction.

[0072] In this embodiment, a multi-scale convolutional neural network is used to extract the ablated surface morphology of the carbon-carbon composite material at each moment, and then the BiLSTM mechanism is used to quickly and accurately learn the spatiotemporal features and long-term relationships hidden in the timing rules. The attention mechanism assigns different weights to the learned features, so that the model pays more attention to important time steps in the sequence and filters out redundant features. This method significantly improves the accuracy and reliability of the prediction of the surface roughness of the thermal protection material, while reducing the computational cost and time, and quickly predicts the evolution of the subsequent surface morphology of the carbon-carbon composite material, more accurately evaluates the performance and reliability of the material, and provides guidance for material selection and structural design, thereby improving the reliability and safety of the thermal protection structure.

[0073] exist Figure 1 Based on the embodiment shown, the following Figure 2 , the technical solution of the above-mentioned thermal protection material surface roughness prediction method is further introduced.

[0074] Figure 2 A schematic diagram of another method for predicting the surface roughness of a thermal protection material provided in an embodiment of the present application is shown in FIG. Figure 2 In some embodiments, the process of the method for predicting the surface roughness of the thermal protection material includes the following steps:

[0075] S201, constructing a carbon-carbon composite unit cell model based on the carbon-carbon composite weaving structure and weaving process parameters;

[0076] Among them, based on the theory of thermochemical ablation of carbon-carbon composite materials and the finite volume method to solve the partial differential equations in the theory of thermochemical ablation;

[0077] Among them, the thermochemical ablation theory of carbon-carbon composite materials is:

[0078] Carbon-carbon composites experience thermochemical ablation near the ablation surface, and an airflow boundary layer appears around it. Complex heat and mass transfer processes occur in the boundary layer, which also causes the decay of the gas-solid interface. Therefore, the heterogeneous reactions occurring at this stage are:

[0079]

[0080] Among them, s represents the solid phase and g represents the gas phase;

[0081] Studies have shown that the boundary layer oxygen concentration distribution and carbon reaction rate are the main parameters that characterize the surface retreat of C / C composites. Therefore, the diffusion of gas and geometric changes can be used to describe the oxidation process of carbon fiber reinforced composites; the variable s is defined as the interface function of the composite material, and all points except the interface are non-zero values; the mathematical expression is:

[0082]

[0083] in, The variables represent the decay rates between different phases, where the subscript m represents the carbon fiber phase and the carbon matrix phase; The solid molar volume of the representative phase, is the molar rate of oxidation at the interface, is the local normal vector at the interface, c is the molar concentration of oxygen, and D is the diffusion coefficient;

[0084] Equations (5) and (9) describe the coupling relationship between gas diffusion and material surface erosion during thermochemical ablation. The dimensionless parameters A and Sh are the key factors affecting the oxidation morphology change of carbon-carbon composites. The dimensionless parameter A describing the oxidation reaction rate is defined as:

[0085]

[0086] in, k represents the oxidation reaction rate;

[0087] When the diffusion gas enters the material, surface thermochemical ablation occurs at the interface of the carbon-carbon composite material; the Sherwood number Sh, where the diffusion rate and the oxidation reaction rate compete with each other, is defined as:

[0088]

[0089] in, l A reference length that represents the size of the unit cell model.

[0090] During the thermochemical ablation process at the interface of carbon-carbon composite materials, accompanied by the transition from solid phase to gas phase, the surface morphology of the material undergoes continuous and irregular changes, which is mainly affected by the external thermal environment, material properties and mesoscopic structural characteristics; the multiphase motion occurring at the interface is similar to the gas-liquid two-phase flow phenomenon in computational fluid dynamics. In order to simulate this phenomenon in fluid, the finite volume method of interface capture is often used to solve the two-phase flow problem:

[0091] In the fluid domain, the gas phase diffusion process satisfies the equation:

[0092]

[0093] In the solid domain, the evolution process in the solid phase satisfies:

[0094]

[0095] In order to solve the coupled partial differential equations, the finite volume method is used for numerical calculation. Therefore, the above equations (12) and (13) are rewritten as

[0096] in, is the time increment, Defined as the volume occupied by carbon / carbon composite materials controlling the gas in the body, Defined as the volume occupied by solids in a carbon / carbon composite control body, represents the gas-solid interface contact area, such as Figure 3 As shown;

[0097] Based on the above-mentioned thermochemical ablation theory and the partial differential equation calculation method in the thermochemical ablation theory, according to the weaving structure and weaving process parameters of carbon-carbon composite materials (fiber volume fraction, length and width of weft and warp yarns), a unit cell model of 2.5D layer-to-layer interlocking structure carbon-carbon composite materials was established using Python programming and TexGen software; Figure 4 As shown;

[0098] S202, simulating and solving the carbon-carbon composite unit cell model by using a finite volume method and a thermochemical ablation theoretical model to obtain a characteristic image of the surface ablation morphology evolution of the three-dimensional carbon-carbon composite material;

[0099] The above unit cell model was solved by finite volume method and Matlab simulation to obtain the characteristic images of surface ablation morphology evolution of three-dimensional carbon-carbon composite materials, such as Figure 5 As shown;

[0100] Among them, it is found from the three-dimensional data of the fiber and matrix of the carbon-carbon composite material that the horizontal and vertical coordinates of each point increase by 0.04 in sequence; therefore, a program is written to traverse the entire data set, perform height detection on each data point, and record the highest height and corresponding coordinates; based on the data point extraction algorithm, data is extracted from the characteristic image of the surface ablation morphology evolution of the three-dimensional carbon-carbon composite material to obtain multiple feature extraction data; the description of the data point extraction algorithm is shown in Table 1 below; however, these simulated data will inevitably have outliers, which will affect the construction of the two-dimensional data and the prediction of machine learning. Therefore, it is necessary to delete these outliers in the process of extracting data.

[0101] Among them, the data point extraction algorithm is shown in the following table:

[0102] Table 1 Data point extraction algorithm

[0103]

[0104] S203, based on the three-dimensional carbon-carbon composite material surface ablation morphology evolution feature image, obtain multiple two-dimensional surface roughness sample images; wherein the feature extraction data is two-dimensionally transformed to obtain a two-dimensional surface roughness sample image; such as Figure 6 As shown;

[0105] S204, based on the MSCNN multi-scale convolutional neural network and the BiLSTM bidirectional long short-term memory network model, construct the MSCNN-BiLSTM attention mechanism prediction model.

[0106] Among them, a multi-scale convolutional neural network MSCNN and its convolutional layer and pooling layer are constructed, and its structure is as follows Figure 7 As shown in the figure, the MSCNN convolutional neural network includes a CNN input layer, a multi-scale feature extraction layer, a pooling layer, a fusion layer, a fully connected layer and a CNN output layer; the multi-scale feature extraction layer includes a first convolution layer, a second convolution layer and a third convolution layer; the convolution kernel size of the first convolution layer is 5*5; the convolution kernel size of the second convolution layer is 3*3; the convolution kernel size of the third convolution layer is 1*1;

[0107] The multi-scale convolutional neural network uses convolution kernels of different scales in the convolution operation to extract multi-scale features of the surface roughness image of carbon-carbon composite materials, integrate richer local information, and make full use of the implicit information inside the features.

[0108] The multi-scale convolutional neural network performs a convolution operation on the input surface roughness image dataset, and its mathematical expression is:

[0109]

[0110] in, is the ReLU activation function; is the weight tensor of the l-th layer CNN; is the input two-dimensional sequence; d is the length of the convolution kernel; b is the bias vector;

[0111] Construct a BiLSTM bidirectional long short-term memory network model; Figure 8 As shown in , the BiLSTM bidirectional long short-term memory network model includes an LSTM input layer, a forward LSTM, a backward LSTM, an activation function and an LSTM output layer; Fig. 9 As shown, both the forward LSTM and the backward LSTM include a forget gate, a Tanh gate, an input gate, and an output gate;

[0112] Specifically, in order to solve the problem of gradient disappearance and gradient explosion, LSTM can better capture long-term dependencies by introducing a gating mechanism, thereby improving the modeling ability of long sequence data; Fig. 9 As shown in the figure, it is mainly composed of a forget gate, a Tanh gate, an input gate, and an output gate; the LSTM network introduces memory cells to store long-term memory information. The cell state runs through the entire chain structure, allowing information to be transmitted over a long time span; the input gate controls the degree of update of new information to the memory cells, which determines how much new information can be added to the cell state in the current time step; the output of the input gate is a value between 0 and 1, representing the importance of each information; the forget gate determines which old memory information should be forgotten in the current time step, and it outputs a value between 0 and 1 through a sigmoid activation function, indicating the degree to which the memory information at the corresponding position is retained; the output gate determines which information in the memory cell should be output in the current time step, and it outputs a value between 0 and 1 through a sigmoid activation function, and uses the tanh activation function to determine the output value range.

[0113] The mathematical expressions of the forget gate, input gate, and output gate in the model are:

[0114]

[0115] in, It's the Gate of Forgetfulness. is the input gate, is the output gate; is the activation function, , , , They are the weight matrices of the forget gate, input gate, output gate and cell state respectively; , , , They are the bias vectors of forget gate, input gate, output gate and cell state respectively; Indicates the output status of the previous moment, Indicates the output status at the current moment; represents the memory unit of the previous moment, The memory unit representing the current moment; Represents the input at the current moment, Represents a new candidate value vector.

[0116] According to the above LSTM network model and the description of the problem of surface roughness prediction of carbon-carbon composite materials, this application adopts a special recurrent neural network, namely a bidirectional long short-term memory neural network (BiLSTM), and the structural diagram is shown in Figure 8; the basic module of BiLSTM is mainly composed of a forward LSTM and a backward LSTM, which jointly dynamically model the surface roughness of carbon-carbon composite materials and improve the prediction performance; the bidirectional LSTM processes the input sequence from the forward and backward directions respectively, so that the model can capture the forward and backward context information, and feed the sequence information into the two LSTM layers;

[0117] First, in the forward propagation process, the input sequence is input into the LSTM network in chronological order. The input data of each time step passes through the gating units such as the input gate, forget gate and output gate in turn to update the memory cell state and calculate the output. Then, in the forward propagation process, the input sequence is input into another independent LSTM network in reverse chronological order. Contrary to the forward propagation, the input data of each time step passes through the gating units such as the input gate, forget gate and output gate in turn to update the memory cell state and calculate the output. At each time step, BiLSTM combines the outputs of the forward and backward LSTM networks, usually using a simple connection method (such as splicing). After forward and backward propagation, the BiLSTM network obtains the comprehensive output of the forward and backward information of the entire input sequence.

[0118] When the surface roughness image of carbon-carbon composite material passes through MSCNN, it can effectively capture the local features and spatiotemporal features in the image, so that the same features have different values ​​for predicting the surface roughness of carbon-carbon composite material at different times; in order to select more detail information that needs attention from a large number of features and ignore irrelevant information, the attention mechanism is introduced to measure the importance of features. By assigning different weight values ​​to different features, the model's attention to the input features is changed, focusing on more discriminative, more critical and more important features, thereby improving the accuracy of the model; when the context vector extracted by the bidirectional LSTM is , the calculation formula involved is:

[0119]

[0120] in, is the weight vector corresponding to each time step, is the hidden state at time t, is the context vector after adding global attention at that moment, is a trainable weight matrix, is the output state at the current moment, U represents the matrix constructed by the eigenvector, and b is the bias.

[0121] The output data of the CNN output layer is used as the input data of the LSTM input layer and subsequently processed to obtain the MSCNN-BiLSTM attention mechanism prediction model.

[0122] S205, inputting the data set into the MSCNN multi-scale convolutional neural network for iterative training to obtain a CNN screening feature set;

[0123] S206, inputting the CNN screening feature set into the BiLSTM bidirectional long short-term memory network model for iterative training to obtain a training result;

[0124] S207, introducing the training result into a cross entropy loss function to obtain a test result;

[0125] Specifically, the cross entropy loss function is:

[0126]

[0127] in, and They represent the true probability distribution and the predicted probability distribution respectively, and log is the logarithmic function.

[0128] S208, evaluating the performance of the MSCNN-BiLSTM attention mechanism prediction model, and when the test result meets the test condition, ending the training; otherwise, optimizing the network parameters of the MSCNN-BiLSTM attention mechanism prediction model according to the prediction result, and re-inputting the data set into the MSCNN-BiLSTM attention mechanism prediction model for training until the test result meets the test condition, thereby obtaining the surface roughness prediction model.

[0129] Among them, the data set is input into the MSCNN-BiLSTM attention mechanism prediction model to obtain the test results;

[0130] Based on the test results, the mean absolute error, root mean square error and coefficient of determination were used to evaluate the MSCNN-BiLSTM attention mechanism prediction model;

[0131] If the mean absolute error and the root mean square error are smaller and the determination coefficient is closer to 1, then the test result meets the test conditions.

[0132] If the mean absolute error and the root mean square error are larger and the determination coefficient is closer to 0, the test result does not meet the test conditions.

[0133] Specifically, the mean absolute error satisfies the following formula:

[0134]

[0135] Specifically, the root mean square error satisfies the following formula:

[0136]

[0137] Specifically, the determination coefficient satisfies the following formula:

[0138]

[0139] Where N represents the number of samples. is the predicted value, is the true value, is the average of the true values.

[0140] S209, inputting the image to be detected into a surface roughness prediction model to obtain a surface roughness prediction result; wherein the image to be detected is obtained based on the carbon-carbon composite material to be detected.

[0141] In this embodiment, the surface roughness of the carbon-carbon composite material is predicted by the MSCNN-BiLSTM attention mechanism prediction model, which significantly improves the accuracy and reliability of the prediction of the surface roughness of the thermal protection material.

[0142] Fig.10 is a schematic diagram of a device for predicting the surface roughness of a thermal protection material provided in an embodiment of the present application, see Fig.10 The device for predicting the surface roughness of a thermal protection material includes various functional modules for implementing the aforementioned method for predicting the surface roughness of a thermal protection material. Any functional module can be implemented by software and / or hardware.

[0143] In some embodiments, the thermal protection material surface roughness prediction device 1100 includes a data acquisition module 1001, a model construction module 1002, a model training module 1003 and a surface roughness prediction module 1004. Among them:

[0144] The data acquisition module 1001 is used to acquire a data set, wherein the data set includes a plurality of two-dimensional surface roughness sample images constructed based on a thermochemical ablation theoretical model and a finite volume method to simulate the surface ablation morphology characteristics of a three-dimensional carbon-carbon composite material;

[0145] The model building module 1002 is used to build a MSCNN-BiLSTM attention mechanism prediction model based on the MSCNN multi-scale convolutional neural network and the BiLSTM bidirectional long short-term memory network model;

[0146] The model training module 1003 inputs the data set into the MSCNN-BiLSTM attention mechanism prediction model for iterative training to obtain a surface roughness prediction model;

[0147] The surface roughness prediction module 1004 is used to input the image to be detected into the surface roughness prediction model to obtain a surface roughness prediction result; wherein the image to be detected is obtained based on the carbon-carbon composite material to be detected.

[0148] The thermal protection material surface roughness prediction device 1000 provided in the embodiment of the present application is used to execute the technical solution provided in the aforementioned thermal protection material surface roughness prediction method embodiment. Its implementation principle and technical effect are similar to those in the aforementioned method embodiment and will not be repeated here.

[0149] It should be noted that it should be understood that the division of the various modules of the above device is only a division of a logical function. When actually implemented, it can be fully or partially integrated into a physical entity, or it can be physically separated. And these modules can all be implemented in the form of software called by a processing element, or they can all be implemented in the form of hardware. Some modules can also be implemented in the form of software called by a processing element, and some modules can be implemented in the form of hardware. For example, the data acquisition module 1001 can be a separately established processing element, or it can be integrated in a certain chip of the above device to be implemented. In addition, it can also be stored in the memory of the above device in the form of program code, and a certain processing element of the above device calls and executes the function of the above data acquisition module 1001. The implementation of other modules is similar. In addition, these modules can be fully or partially integrated together, or they can be implemented independently. The processing element here can be an integrated circuit with signal processing capabilities. In the implementation process, each step of the above method or each module above can be completed by an integrated logic circuit of hardware in a processor element or an instruction in the form of software.

[0150] Fig.11 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application, see Fig.11 , the electronic device 1100 includes: a processor 1101, and a memory 1102 communicatively connected to the processor 1101;

[0151] Memory 1102 stores computer executable instructions;

[0152] The processor 1101 executes the computer-executable instructions stored in the memory 1102 to implement the technical solution of the aforementioned method for predicting the surface roughness of the thermal protection material.

[0153] In the above-mentioned electronic device 1000, the memory 1102 and the processor 1101 are electrically connected directly or indirectly to realize data transmission or interaction. For example, these elements can be electrically connected to each other through one or more communication buses or signal lines, such as through a bus. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc., but it does not mean that there is only one bus or one type of bus. The memory 1102 stores computer execution instructions for implementing the above-mentioned thermal protection material surface roughness prediction method, including at least one software function module that can be stored in the memory 1102 in the form of software or firmware. The processor 1101 executes various functional applications and data processing by running the software programs and modules stored in the memory 1102.

[0154] The memory 1102 includes at least one type of readable storage medium, including but not limited to random access memory (RAM), read only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), and electrically erasable read-only memory (EEPROM). The memory 1102 is used to store programs, and the processor 1101 executes the programs after receiving the execution instruction. Furthermore, the software programs and modules in the above-mentioned memory 1102 may also include an operating system, which may include various software components and / or drivers for managing system tasks (such as memory management, storage device control, power management, etc.), and may communicate with various hardware or software components to provide an operating environment for other software components.

[0155] The processor 1101 may be an integrated circuit chip having signal processing capability. The above-mentioned processor 1101 may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), a digital signal processor (DSP), an application-specific integrated circuit (ASIC), etc. The disclosed methods, steps and logic block diagrams in the embodiments of the present application may be implemented or executed. The general-purpose processor may be a microprocessor, or the processor 1101 may be any conventional processor, etc.

[0156] The electronic device 1100 is used to execute the technical solution provided by the aforementioned thermal protection material surface roughness prediction method embodiment. Its implementation principle and technical effect are similar to those in the aforementioned method embodiment and will not be repeated here.

[0157] The embodiment of the present application also provides a computer-readable storage medium, in which computer-executable instructions are stored. When a processor executes the computer-executable instructions, the technical solution of the aforementioned method for predicting the surface roughness of thermal protection materials is implemented.

[0158] The computer-readable storage medium mentioned above can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk. The computer-readable storage medium can be any available medium that can be accessed by a general or special-purpose computer.

[0159] An exemplary readable storage medium is coupled to the processor so that the processor can read information from the readable storage medium and write information to the readable storage medium. Of course, the readable storage medium can also be a component of the processor. The processor and the readable storage medium can be located in an application specific integrated circuit (Application Specific Integrated Circuits, referred to as: ASIC). Of course, the processor and the readable storage medium can also exist as discrete components in the control device of the thermal protection material surface roughness prediction device.

[0160] An embodiment of the present application also provides a computer program product, including a computer program, which, when executed by a processor, is used to implement the technical solution of the aforementioned method for predicting the surface roughness of a thermal protection material.

[0161] In the above embodiments, those skilled in the art can understand that the implementation of the above method embodiments can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function according to the embodiment of the present invention is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website site, computer, server or data center to another website site, computer, server or data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless network, microwave, etc.). The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that includes one or more available media integrated. The available medium may be a magnetic medium (eg, a floppy disk, a hard disk, a magnetic tape), an optical medium (eg, a DVD), or a semiconductor medium (eg, a solid state disk (SSD)).

[0162] It will be appreciated that the present application is not limited to the exact construction that has been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof.

Claims

1. A method for predicting the surface roughness of a thermal protection material, characterized in that: include: Acquire a data set; wherein the data set includes a plurality of two-dimensional surface roughness sample images constructed based on a thermochemical ablation theoretical model and a finite volume method to simulate the surface ablation morphology characteristics of a three-dimensional carbon-carbon composite material; Based on the MSCNN multi-scale convolutional neural network and the BiLSTM bidirectional long short-term memory network model, the MSCNN-BiLSTM attention mechanism prediction model is constructed; Inputting the data set into the MSCNN-BiLSTM attention mechanism prediction model for iterative training to obtain a surface roughness prediction model; Inputting the image to be detected into the surface roughness prediction model to obtain a surface roughness prediction result; wherein the image to be detected is obtained based on the carbon-carbon composite material to be detected; The step of acquiring a data set comprises: Based on the carbon-carbon composite weaving structure and weaving process parameters, a carbon-carbon composite unit cell model is constructed; specifically including: Based on the thermochemical ablation theory and the partial differential equation calculation method in the thermochemical ablation theory, according to the carbon-carbon composite weaving structure and weaving process parameters: fiber volume fraction, length and width of weft and warp yarns, a unit cell model of 2.5D layer-to-layer interlocking structure carbon-carbon composite material was established using Python programming and TexGen software. The carbon-carbon composite unit cell model is simulated and solved by the finite volume method and the thermochemical ablation theoretical model to obtain a characteristic image of the surface ablation morphology evolution of the three-dimensional carbon-carbon composite material; Based on the three-dimensional carbon-carbon composite material surface ablation morphology evolution characteristic image, obtaining a plurality of two-dimensional surface roughness sample images; Based on the three-dimensional carbon-carbon composite material surface ablation morphology evolution characteristic image, a plurality of two-dimensional surface roughness sample images are obtained, including: Based on the data point extraction algorithm, data is extracted from the characteristic image of the surface ablation morphology evolution of the three-dimensional carbon-carbon composite material to obtain a plurality of characteristic extraction data; specifically including: The above-mentioned unit cell model is solved by finite volume method and Matlab simulation to obtain the characteristic image of surface ablation morphology evolution of three-dimensional carbon-carbon composite materials, wherein the horizontal and vertical coordinates of each point are found to increase by 0.04 in sequence from the three-dimensional data of the fiber and matrix of the carbon-carbon composite material; a program is written to traverse the entire data set, perform height detection on each data point, and record the highest height and corresponding coordinates; data is extracted from the characteristic image of surface ablation morphology evolution of three-dimensional carbon-carbon composite materials based on the data point extraction algorithm to obtain multiple feature extraction data; Performing two-dimensional transformation on the feature extraction data to obtain the two-dimensional surface roughness sample image; Based on the MSCNN multi-scale convolutional neural network and the BiLSTM bidirectional long short-term memory network model, the MSCNN-BiLSTM attention mechanism prediction model is constructed, including: Constructing a MSCNN convolutional neural network; wherein the MSCNN convolutional neural network includes a CNN input layer, a multi-scale feature extraction layer, a pooling layer, a fusion layer, a fully connected layer and a CNN output layer; the multi-scale feature extraction layer includes a first convolutional layer, a second convolutional layer and a third convolutional layer; the convolution kernel size of the first convolutional layer is 5*5; the convolution kernel size of the second convolutional layer is 3*3; the convolution kernel size of the third convolutional layer is 1*1; Constructing a BiLSTM bidirectional long short-term memory network model; wherein the BiLSTM bidirectional long short-term memory network model includes an LSTM input layer, a forward LSTM, a backward LSTM, an activation function and an LSTM output layer; the forward LSTM and the backward LSTM both include a forget gate, a Tanh gate, an input gate and an output gate; The output data of the CNN output layer is used as the input data of the LSTM input layer and subsequently processed to obtain the MSCNN-BiLSTM attention mechanism prediction model.

2. The method according to claim 1, characterized in that: The step of inputting the data set into the MSCNN-BiLSTM attention mechanism prediction model for iterative training to obtain a surface roughness prediction model includes: Inputting the data set into the MSCNN multi-scale convolutional neural network for iterative training to obtain a CNN screening feature set; Inputting the CNN screening feature set into the BiLSTM bidirectional long short-term memory network model for iterative training to obtain a training result; The training result is introduced into the cross entropy loss function to obtain the test result; The performance of the MSCNN-BiLSTM attention mechanism prediction model is evaluated, and when the test result meets the test condition, the training is terminated; otherwise, the network parameters of the MSCNN-BiLSTM attention mechanism prediction model are optimized according to the prediction result, and the data set is re-input into the MSCNN-BiLSTM attention mechanism prediction model for training until the test result meets the test condition, thereby obtaining the surface roughness prediction model.

3. The method according to claim 2, characterized in that The performance of the MSCNN-BiLSTM attention mechanism prediction model is evaluated, including: Input the data set into the MSCNN-BiLSTM attention mechanism prediction model to obtain a test result; Based on the test results, the mean absolute error, root mean square error and coefficient of determination were used to evaluate the MSCNN-BiLSTM attention mechanism prediction model; If the mean absolute error and the root mean square error are smaller and the determination coefficient is closer to 1, then the test result meets the test conditions.

4. The method according to claim 3, characterized in that The method further comprises: If the mean absolute error and the root mean square error are larger and the determination coefficient is closer to 0, the test result does not meet the test conditions.

5. A device for predicting the surface roughness of a thermal protection material, characterized in that: include: A data acquisition module is used to acquire a data set; wherein the data set includes a plurality of two-dimensional surface roughness sample images constructed based on a thermochemical ablation theoretical model and a finite volume method to simulate the surface ablation morphology characteristics of a three-dimensional carbon-carbon composite material; the acquired data set includes: Based on the carbon-carbon composite weaving structure and weaving process parameters, a carbon-carbon composite unit cell model is constructed; specifically including: Based on the thermochemical ablation theory and the partial differential equation calculation method in the thermochemical ablation theory, according to the carbon-carbon composite weaving structure and weaving process parameters: fiber volume fraction, length and width of weft and warp yarns, a unit cell model of 2.5D layer-to-layer interlocking structure carbon-carbon composite material was established using Python programming and TexGen software. The carbon-carbon composite unit cell model is simulated and solved by the finite volume method and the thermochemical ablation theoretical model to obtain a characteristic image of the surface ablation morphology evolution of the three-dimensional carbon-carbon composite material; Based on the three-dimensional carbon-carbon composite material surface ablation morphology evolution characteristic image, obtaining a plurality of two-dimensional surface roughness sample images; Based on the three-dimensional carbon-carbon composite material surface ablation morphology evolution characteristic image, a plurality of two-dimensional surface roughness sample images are obtained, including: Based on the data point extraction algorithm, data is extracted from the characteristic image of the surface ablation morphology evolution of the three-dimensional carbon-carbon composite material to obtain a plurality of characteristic extraction data; specifically including: The above-mentioned unit cell model is solved by finite volume method and Matlab simulation to obtain the characteristic image of surface ablation morphology evolution of three-dimensional carbon-carbon composite materials, wherein the horizontal and vertical coordinates of each point are found to increase by 0.04 in sequence from the three-dimensional data of the fiber and matrix of the carbon-carbon composite material; a program is written to traverse the entire data set, perform height detection on each data point, and record the highest height and corresponding coordinates; data is extracted from the characteristic image of surface ablation morphology evolution of three-dimensional carbon-carbon composite materials based on the data point extraction algorithm to obtain multiple feature extraction data; Performing two-dimensional transformation on the feature extraction data to obtain the two-dimensional surface roughness sample image; The model building module is used to build the MSCNN-BiLSTM attention mechanism prediction model based on the MSCNN multi-scale convolutional neural network and the BiLSTM bidirectional long short-term memory network model; the MSCNN-BiLSTM attention mechanism prediction model is built based on the MSCNN multi-scale convolutional neural network and the BiLSTM bidirectional long short-term memory network model, including: Constructing a MSCNN convolutional neural network; wherein the MSCNN convolutional neural network includes a CNN input layer, a multi-scale feature extraction layer, a pooling layer, a fusion layer, a fully connected layer and a CNN output layer; the multi-scale feature extraction layer includes a first convolutional layer, a second convolutional layer and a third convolutional layer; the convolution kernel size of the first convolutional layer is 5*5; the convolution kernel size of the second convolutional layer is 3*3; the convolution kernel size of the third convolutional layer is 1*1; Constructing a BiLSTM bidirectional long short-term memory network model; wherein the BiLSTM bidirectional long short-term memory network model includes an LSTM input layer, a forward LSTM, a backward LSTM, an activation function and an LSTM output layer; the forward LSTM and the backward LSTM both include a forget gate, a Tanh gate, an input gate and an output gate; The output data of the CNN output layer is used as the input data of the LSTM input layer and subsequently processed to obtain the MSCNN-BiLSTM attention mechanism prediction model. A model training module, inputting the data set into the MSCNN-BiLSTM attention mechanism prediction model for iterative training to obtain a surface roughness prediction model; The surface roughness prediction module is used to input the image to be detected into the surface roughness prediction model to obtain a surface roughness prediction result; wherein the image to be detected is obtained based on the carbon-carbon composite material to be detected.

6. An electronic device, characterized in that: include: A processor, and a memory communicatively connected to the processor; The memory stores computer-executable instructions; The processor executes the computer-executable instructions stored in the memory to implement the method according to any one of claims 1 to 4.

7. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, which are used to implement the method according to any one of claims 1 to 4 when executed by a processor.

Images