Method for acquiring ablation morphology of C / C composite material based on physical information neural network

Abstract

The invention discloses a method for acquiring ablation morphology of a C / C composite material based on a physical information neural network, and relates to the field of aerospace. Comprising the following steps: taking coordinate data, ablation time, a dimensionless parameter Sh and a dimensionless parameter A of a C / C composite material three-dimensional space model under a Cartesian coordinate system as inputs of a pre-trained multilayer perceptron network, and outputting an oxygen molar concentration and a solid volume fraction through the pre-trained multilayer perceptron network; and solving a gas phase diffusion equation and a solid phase evolution equation through the oxygen molar concentration and the solid volume fraction output by the multi-layer perceptron network to obtain the ablation morphology of the C / C composite material. According to the method, the gas-solid coupling equation can be efficiently and accurately solved, the limitation of a traditional numerical method is overcome, the ablation morphology of the C / C composite material under the gas-solid coupling effect is accurately obtained with low calculation cost, and powerful theoretical support and technical guarantee are provided for application of the C / C composite material in the fields of aerospace and the like.

Description

Technical Field

[0001] This invention relates to the aerospace field, and specifically to a method for obtaining the ablation morphology of C / C composite materials based on a physical information neural network. Background Technology

[0002] C / C composites refer to a pure carbon multiphase structure composed of carbon fiber or its fabric as the reinforcing phase and chemically vapor-infiltrated pyrolytic carbon or liquid-phase impregnated and carbonized resin carbon or pitch carbon as the matrix. Due to its outstanding advantages such as low density, excellent ablation resistance, and good high-temperature mechanical properties, it is widely used in the aerospace field, such as in the thermal protection systems of spacecraft and the throat liners of rocket engines. In actual service, C / C composites face extremely complex gas-solid coupling environments such as high temperatures and high-speed airflows, inevitably leading to thermochemical ablation. Accurately predicting the ablation morphology of C / C composites under gas-solid coupling is crucial. Traditional experimental testing methods, such as oxy-acetylene ablation tests and plasma ablation tests, while providing direct data on ablation and surface morphology, suffer from high cost, long cycles, complex operating condition control, and difficulty in fully revealing the intrinsic physical mechanisms of the ablation process. Traditional numerical methods such as the finite element method and the finite difference method have limitations when dealing with complex geometries, highly nonlinear and multi-scale gas-solid coupling problems, including low computational efficiency, difficulty in mesh generation, and limited ability to handle complex boundary conditions.

[0003] In recent years, deep learning technology has developed rapidly. Physical Information Neural Networks (PINNs), as an emerging numerical computation method, can solve differential equations without a mesh by embedding prior physical knowledge such as physical equations and boundary conditions into the neural network. However, there is currently no mature and systematic method or system based on PINN to solve gas-solid coupling equations to obtain the ablation morphology of C / C composite materials. Therefore, developing an efficient and accurate PINN-based method or system is of great significance for a deeper understanding of the ablation mechanism of C / C composite materials, optimizing material design, and improving the reliability of related engineering applications. Summary of the Invention

[0004] To address the aforementioned shortcomings in existing technologies, the method for obtaining the ablation morphology of C / C composite materials based on physical information neural networks provided by this invention can efficiently and accurately solve the gas-solid coupling equation, overcome the limitations of traditional numerical methods, and accurately obtain the ablation morphology of C / C composite materials under gas-solid coupling with low computational cost. This provides strong theoretical support and technical guarantee for the application of C / C composite materials in aerospace and other fields.

[0005] To achieve the above-mentioned objectives, the technical solution adopted by this invention is as follows: A method for obtaining the ablation morphology of C / C composite materials based on a physical information neural network is provided, which includes the following steps: Construct gas-solid coupling equations for C / C composite materials in a microscopic ablation model; Obtain the dimensionless parameters Sh and A of the microscopic ablation behavior of C / C composite materials; Construct a pre-trained multilayer perceptron network; construct a three-dimensional spatial model of C / C composite materials; The coordinate data of the three-dimensional spatial model of C / C composite material in the Cartesian coordinate system, ablation time, dimensionless parameter Sh and dimensionless parameter A are used as inputs to a pre-trained multilayer perceptron network. The pre-trained multilayer perceptron network outputs oxygen molar concentration and solid volume fraction. Based on the gas-solid coupling equation, the gas phase diffusion equation and solid phase evolution equation of C / C composite material in the microscopic ablation model are obtained; the initial conditions and boundary conditions of the gas phase diffusion equation and solid phase evolution equation are set. By solving the gas phase diffusion equation and the solid phase evolution equation using the oxygen molar concentration and solid volume fraction output by the multilayer perceptron network, the ablation morphology of the C / C composite material is obtained.

[0006] Furthermore, the expression for the gas-solid coupling equation is:

[0007]

[0008] in Indicates the solid phase; Indicates time; Indicates the speed at which the screen scrolls backward. , Represents the molar volume of a solid phase. This indicates the molar rate of oxidation at the interface. , The oxidation reaction rate, This represents the local normal vector at the interface. , Represents the gradient operator. This indicates the gradient with respect to the solid phase. express The model; Indicates the molar concentration of oxygen; is the diffusion coefficient.

[0009] Furthermore, the expression for calculating the dimensionless parameter Sh is:

[0010] in The fiber radius of the C / C composite material; The matrix oxidation reaction rate of the C / C composite material; is the diffusion coefficient.

[0011] Furthermore, the expression for calculating the dimensionless parameter A is:

[0012] in The matrix oxidation reaction rate in the C / C composite material; The molar volume of the matrix phase; The oxidation reaction rate of carbon fibers in C / C composite materials; denoted as , where is the molar volume of the fibrous phase.

[0013] Furthermore, the method for constructing a pre-trained multilayer perceptron network includes the following steps: The number of hidden layers and the number of neurons in each hidden layer of the multilayer perceptron network are set, and nonlinear transformations are performed in the hidden layers using activation functions; the number of hidden layers and the number of neurons are determined using the open-source framework Optuna. The gas-phase diffusion loss function, solid-phase evolution loss function, and initial condition loss function are constructed, and then the total loss function of the multilayer perceptron network is obtained. The three-dimensional spatial model of the C / C composite material with known ablation morphology and its initial conditions are used as the computational domain, and points are selected within the computational domain. The point selection method is uniform sampling, random sampling, or adaptive sampling. The selected points include oxygen molar concentration and solid volume fraction. The coordinates of the selected points, ablation time, dimensionless parameter Sh, and dimensionless parameter A are used as training samples and input into the multilayer perceptron network to obtain the oxygen molar concentration and solid volume fraction predicted by the multilayer perceptron network. The oxygen molar concentration and solid volume fraction of the selected points are used as the true labels. Combined with the oxygen molar concentration and solid volume fraction predicted by the multilayer perceptron network, the total loss value is calculated through the total loss function of the multilayer perceptron network. The parameters of the multilayer perceptron network are adjusted with the goal of minimizing the total loss value until the preset training termination condition is met, thus obtaining the pre-trained multilayer perceptron network.

[0014] Furthermore, the expression for the total loss function of a multilayer perceptron network is:

[0015] in This is the total loss function; , and These are the weighting coefficients; Let be the gas phase diffusion loss function. ; Let be the solid-state evolution loss function. ; The initial conditional loss function, ; Indicates the molar concentration of oxygen; Indicates time; Represents the ladder operator; For solid volume fraction, when When represents the gas phase unit, when Time represents a solid phase unit; The diffusion coefficient is denoted as . This represents the molar concentration of the reacting gas; The square of the L2 norm; The velocity of the solid as it recedes; This represents the concentration at the outer edge of the boundary layer.

[0016] Furthermore, the specific method for constructing a three-dimensional spatial model of C / C composite materials is as follows: A three-dimensional spatial model of C / C composite material with a three-dimensional matrix encapsulating fiber monofilaments is established. The coordinate data of the three-dimensional spatial model of C / C composite material in the Cartesian coordinate system is obtained, and the cross-sectional description of the three-dimensional spatial model of C / C composite material is given.

[0017] Furthermore, the gas phase diffusion equation is as follows:

[0018] Among them, in the fluid domain and the solid domain At the interface ; is the solid volume fraction; C is the molar concentration of the reactant gas; This represents a heterogeneous reaction rate. It represents the reciprocal of the gradient magnitude.

[0019] Furthermore, the solid phase evolution equation is as follows:

[0020] in It represents the volume fraction of solids. The velocity of the solid's retreat; in the fluid and solid domains At the interface ; For solid volume fraction; for the molar concentration of the reactant gas; This represents a heterogeneous reaction rate. It represents the reciprocal of the gradient magnitude.

[0021] Furthermore, the specific method for obtaining the ablation morphology of C / C composite materials by solving the gas phase diffusion equation and the solid phase evolution equation using the oxygen molar concentration and solid volume fraction output by the multilayer perceptron network is as follows: By using the oxygen molar concentration and solid volume fraction output by the multilayer perceptron network, the evolution of the oxygen molar concentration and solid volume fraction distribution of the material at different times is tracked through the gas phase diffusion equation and the solid phase evolution equation, thereby obtaining the development process of the ablation morphology of the three-dimensional C / C composite material over time.

[0022] The beneficial effects of this invention are as follows: This method can efficiently and accurately solve the gas-solid coupling equation, overcome the limitations of traditional numerical methods, and accurately obtain the ablation morphology of C / C composite materials under gas-solid coupling with low computational cost, providing strong theoretical support and technical guarantee for the application of C / C composite materials in aerospace and other fields. Attached Figure Description

[0023] Figure 1 This is a flowchart illustrating the method. Figure 2 This is a schematic diagram of the computational domain of the three-dimensional model in the embodiment; Figure 3 This is a yz section view of the 3D model; Figure 4 A schematic diagram showing how data points in each region are represented by different colors and symbols when solving the coupled Stokes-Darcy equations in two dimensions; Figure 5 A schematic diagram of point selection in the computational domain for PINN; Figure 6 This is a schematic diagram of the placement points and cross-sectional placement points of the 3D model; where (a) represents the 3D placement points and (b) represents the cross-sectional placement points. Figure 7 The curves showing the changes in equation loss, boundary loss, and total loss after training a multilayer perceptron network 120 times; Figure 8 The evolution of the gas-solid interface at different times; Figure 9 The ablation morphology of a single fiber at different Da numbers is shown. Detailed Implementation

[0024] The specific embodiments of the present invention are described below to enable those skilled in the art to understand the present invention. However, it should be understood that the present invention is not limited to the scope of the specific embodiments. For those skilled in the art, various changes are obvious as long as they are within the spirit and scope of the present invention as defined and determined by the appended claims. All inventions utilizing the concept of the present invention are protected.

[0025] like Figure 1 As shown, the method for obtaining the ablation morphology of C / C composite materials based on physical information neural networks includes the following steps: S1. Construct the gas-solid coupling equation for C / C composite material in the microscopic ablation model; S2. Obtain the dimensionless parameters Sh and A of the microscopic ablation behavior of C / C composite materials; S3. Construct a pre-trained multilayer perceptron network; construct a three-dimensional spatial model of C / C composite materials; S4. The coordinate data of the three-dimensional spatial model of C / C composite material in the Cartesian coordinate system, the ablation time, the dimensionless parameter Sh and the dimensionless parameter A are used as inputs to a pre-trained multilayer perceptron network. The oxygen molar concentration and solid volume fraction are output through the pre-trained multilayer perceptron network. S5. Based on the gas-solid coupling equation, obtain the gas phase diffusion equation and solid phase evolution equation of C / C composite material in the microscopic ablation model; set the initial conditions and boundary conditions for the gas phase diffusion equation and solid phase evolution equation. S6. Solve the gas phase diffusion equation and solid phase evolution equation using the oxygen molar concentration and solid volume fraction output by the multilayer perceptron network to obtain the ablation morphology of the C / C composite material.

[0026] The carbonized composite material at the ablation surface mainly contains carbon elements (carbon fibers and carbon matrix), and the thermochemical ablation phenomena occurring on the surface mainly include carbon oxidation, nitridation, and sublimation reactions. Typically, the thermochemical ablation phenomenon of carbonized composite materials mainly occurs due to oxidation; therefore, we only study the oxidation reaction of carbonized composite materials. Thermochemical ablation occurs near the ablation surface of C / C composite materials, and a gas boundary layer appears around it. Complex heat and mass transfer processes occur within the boundary layer, influencing the ablation behavior. Existing research shows that at the microscale, oxygen transport in the boundary layer is controlled by a pure diffusion mechanism. Oxygen enters the interior of the boundary layer through channels at the outer edge in an approximately pure diffusion manner and reacts with the carbon elements on the material surface. This process also causes the decay of the gas-solid interface.

[0027] In this embodiment, the boundary layer oxygen molar concentration distribution and carbon reaction rate are key parameters characterizing the surface retreat of the C / C composite ablation. Gas diffusion determines how oxygen is transported to the composite surface and its distribution within the material, while geometric changes mainly involve the changes in shape and size of the composite material due to carbon consumption during oxidation. These changes not only affect the mechanical properties of the material but also, in turn, influence the gas diffusion path and boundary layer characteristics. Therefore, the oxidation process of carbon fiber reinforced composites can be described based on gas diffusion and geometric changes. Specifically, the gas-solid coupling equation is expressed as follows:

[0028]

[0029] in Indicates the solid phase; Indicates time; Indicates the speed at which the screen scrolls backward. , Represents the molar volume of a solid phase. This indicates the molar rate of oxidation at the interface. , The oxidation reaction rate, This represents the local normal vector at the interface. , Represents the gradient operator. This indicates the gradient with respect to the solid phase. express The model; Indicates the molar concentration of oxygen; is the diffusion coefficient.

[0030] In this embodiment, the expression for calculating the dimensionless parameter Sh is:

[0031] in The fiber radius of the C / C composite material; The matrix oxidation reaction rate of the C / C composite material; is the diffusion coefficient.

[0032] In this embodiment, the expression for calculating the dimensionless parameter A is:

[0033] in The matrix oxidation reaction rate in the C / C composite material; The molar volume of the matrix phase; The oxidation reaction rate of carbon fibers in C / C composite materials; denoted as , where is the molar volume of the fibrous phase.

[0034] The method for constructing the pre-trained multilayer perceptron network in step S3 includes the following steps: S3-1. Set the number of hidden layers and the number of neurons in each hidden layer of the multilayer perceptron network, and apply nonlinear transformations using activation functions in the hidden layers. The number of hidden layers and the number of neurons are determined using the open-source framework Optuna. Common activation functions include ReLU, Tanh, and Sigmoid. This embodiment selects ReLU because it has advantages such as simple computation and fast convergence speed, and can effectively introduce nonlinear factors, enabling the multilayer perceptron network to learn more complex functional relationships. Therefore, selecting ReLU can enhance the multilayer perceptron network's ability to capture various nonlinear physical phenomena in the gas-solid coupling process. S3-2. Construct the gas phase diffusion loss function, solid phase evolution loss function, and initial condition loss function, and then obtain the total loss function of the multilayer perceptron network; S3-3. The three-dimensional spatial model of the C / C composite material with known ablation morphology and its initial conditions is used as the computational domain, and points are selected within the computational domain. The point selection method is uniform sampling, random sampling, or adaptive sampling. The selected points include oxygen molar concentration and solid volume fraction. S3-4. Input the coordinates of the selected point, ablation time, dimensionless parameter Sh and dimensionless parameter A as training samples into the multilayer perceptron network to obtain the oxygen molar concentration and solid volume fraction predicted by the multilayer perceptron network. S3-5. Take the oxygen molar concentration and solid volume fraction of the selected points as the true labels, combine them with the oxygen molar concentration and solid volume fraction predicted by the multilayer perceptron network, and calculate the total loss value through the total loss function of the multilayer perceptron network. S3-6. Adjust the parameters of the multilayer perceptron network with the goal of minimizing the total loss value until the preset training termination condition is met, and obtain the pre-trained multilayer perceptron network.

[0035] In this embodiment, the x, y, and z coordinates in the Cartesian coordinate system are used as part of the input layer. These coordinate values ​​can locate every point within and around the C / C composite material in the computational region, providing a basis for subsequent calculations of the physical quantities at that point. By introducing a time variable, the multilayer perceptron network can capture the evolution of the morphology over time during the ablation process.

[0036] The expression for the total loss function of a multilayer perceptron network is:

[0037] in This is the total loss function; , and These are the weighting coefficients; Let be the gas phase diffusion loss function. ; Let be the solid-state evolution loss function. ; The initial conditional loss function, ; Indicates the molar concentration of oxygen; Indicates time; Represents the ladder operator; For solid volume fraction, when When represents the gas phase unit, when Time represents a solid phase unit; The diffusion coefficient is denoted as . This represents the molar concentration of the reacting gas; The square of the L2 norm; The velocity of the solid as it recedes; This represents the concentration at the outer edge of the boundary layer.

[0038] like Figure 2 and Figure 3 As shown, the specific method for constructing the three-dimensional spatial model of the C / C composite material in step S3 is as follows: A three-dimensional spatial model of C / C composite material with a three-dimensional matrix encapsulating fiber monofilaments is established. The coordinate data of the three-dimensional spatial model of C / C composite material in the Cartesian coordinate system is obtained, and the cross-sectional description of the three-dimensional spatial model of C / C composite material is given.

[0039] In the microscopic ablation model, the gas-solid interface undergoes ablation retreat under oxidation, with gas phase diffusion and interface retreat phenomena being coupled. The interface retreat rate is related to the gas concentration at the interface, and interface retreat alters the gas concentration distribution. During oxidation, the solid phase transforms into the gas phase, and the geometry of the solid domain undergoes complex deformations such as fusion and disappearance. Based on this, the gas phase diffusion equation in this embodiment is:

[0040] Among them, in the fluid domain and the solid domain At the interface ; is the solid volume fraction; C is the molar concentration of the reactant gas; This represents a heterogeneous reaction rate. It represents the reciprocal of the gradient magnitude.

[0041] The solid phase evolution equation is:

[0042] in It represents the volume fraction of solids. The velocity of the solid's retreat; in the fluid and solid domains At the interface ; For solid volume fraction; for the molar concentration of the reactant gas; This represents a heterogeneous reaction rate. It represents the reciprocal of the gradient magnitude.

[0043] In computational fluid dynamics, the gas / liquid two-phase flow problem shares similarities with the gas / solid problem discussed in this paper. Therefore, simulation methods for gas / liquid two-phase flow can be applied to solve the gas phase diffusion equation and the solid phase evolution equation. Fluid volume fraction is a commonly used method for solving gas / liquid two-phase flow problems; similarly, this method introduces the solid volume fraction method, resulting in governing equations of a similar form for the evolution of the fluid and solid domains. ,in This refers to the volume of solid phase contained within a unit volume V. Specifically, the method for obtaining the ablation morphology of the C / C composite material is as follows: This is achieved by solving the gas phase diffusion equation and the solid phase evolution equation using the oxygen molar concentration and solid volume fraction output from a multilayer perceptron network. By using the oxygen molar concentration and solid volume fraction output by the multilayer perceptron network, the evolution of the oxygen molar concentration and solid volume fraction distribution of the material at different times is tracked through the gas phase diffusion equation and the solid phase evolution equation, thereby obtaining the development process of the ablation morphology of the three-dimensional C / C composite material over time.

[0044] In one embodiment of the present invention, when using a physical information neural network (a pre-trained multilayer perceptron network) to solve partial differential equations, the selection of points within the computational domain, at the boundaries, and for initial conditions is crucial. This study only involves the selection of points within the computational domain and for initial conditions. First, points are selected within the computational domain for solving the three-dimensional spatial model of the C / C composite material to describe the variation of the physical field throughout the region. Common methods for selecting points within the computational domain include uniform sampling, random sampling, and adaptive sampling.

[0045] As shown in the computational domain diagram, the gas is uniformly distributed on the surface of the solid domain. Therefore, we employ uniform sampling, i.e., sampling uniformly within the computational domain according to a certain grid spacing. This ensures that the model has relatively uniform learning and fitting across all regions of the computational domain, avoiding overfitting in some regions while underfitting in others. For the two-dimensional solution of the coupled Stokes-Darcy equations, data points in each region are represented by different colors and symbols, such as... Figure 4 As shown.

[0046] The gas phase diffusion equation and the solid phase evolution equation describe the variation of the physical system throughout the entire computational domain. By taking points within the computational domain, PINN can acquire physical information at these points, such as the variation of gas-solid interface movement. The model uses this information to learn the physical processes represented by the partial differential equations, thereby establishing a mapping relationship between inputs (spatial and temporal coordinates, etc.) and outputs (physical quantities) to fit the physical phenomena throughout the entire computational domain. Then, for the problem of the evolution of the ablation morphology of C / C composite materials over time, the distribution of physical quantities at the initial moment needs to be given as the starting condition of the problem. For the initial condition, sampling can be performed on the time slice at the initial moment, and the sampling method is similar to that of taking points within the computational domain. For example, for the one-dimensional heat transfer problem, the principle diagram of PINN's point selection in the computational domain is as follows. Figure 5 As shown.

[0047] The initial condition points provide PINN with specific state information of the physical system at the initial moment. By providing accurate initial conditions, PINN can model and predict the dynamic process of the physical system from the correct starting point, ensuring that the simulation results conform to the development laws of actual physical phenomena. During the learning process, the network parameters need to be adjusted so that the model can accurately reproduce the physical quantity values ​​at these points at the initial moment. This helps guide the model to train in the correct direction, avoiding deviations from the actual physical process in the training results, thereby improving the accuracy and reliability of the model. Therefore, in this embodiment, the three-dimensional placement points and cross-sectional placement points of the three-dimensional model are as follows: Figure 6 As shown.

[0048] In the specific implementation process, a pre-trained multilayer perceptron network is applied to new test points or regions to calculate the gas-solid physical quantities and ablation morphology parameters of C / C composite materials under gas-solid coupling. The calculation results are compared with those of traditional numerical methods (such as the finite volume coupling method) to evaluate the accuracy and reliability of the solution. The results are visualized, such as plotting the ablation rate distribution cloud map and a 3D visualization image of the ablation morphology on the surface of the C / C composite material, to intuitively understand the ablation process and morphology evolution of the C / C composite material under gas-solid coupling.

[0049] In this embodiment, as Figure 7 , Figure 8 and Figure 9 As shown, by training the established model, the loss function, the gas-solid interface changes at different times, and the ablation morphology of a single fiber at different Da numbers are obtained.

[0050] In summary, this invention can efficiently and accurately solve the gas-solid coupling equation, overcome the limitations of traditional numerical methods, and accurately obtain the ablation morphology of C / C composite materials under gas-solid coupling with low computational cost, providing strong theoretical support and technical guarantee for the application of C / C composite materials in aerospace and other fields.

Claims

1. A method for obtaining the ablation morphology of C / C composite materials based on a physical information neural network, characterized in that, Includes the following steps: Construct gas-solid coupling equations for C / C composite materials in a microscopic ablation model; The dimensionless parameters Sh and A of the microscopic ablation behavior of C / C composite materials were obtained. Construct a pre-trained multilayer perceptron network; construct a three-dimensional spatial model of C / C composite materials; The coordinate data of the three-dimensional spatial model of C / C composite material in the Cartesian coordinate system, ablation time, dimensionless parameter Sh and dimensionless parameter A are used as inputs to a pre-trained multilayer perceptron network. The pre-trained multilayer perceptron network outputs oxygen molar concentration and solid volume fraction. Based on the gas-solid coupling equation, the gas phase diffusion equation and solid phase evolution equation of C / C composite material in the microscopic ablation model are obtained; the initial conditions and boundary conditions of the gas phase diffusion equation and solid phase evolution equation are set. By solving the gas phase diffusion equation and the solid phase evolution equation using the oxygen molar concentration and solid volume fraction output by the multilayer perceptron network, the ablation morphology of the C / C composite material is obtained.

2. The method for obtaining the ablation morphology of C / C composite materials based on a physical information neural network according to claim 1, characterized in that, The expression for the gas-solid coupling equation is: in Indicates the solid phase; Indicates time; Indicates the speed at which the screen scrolls backward. , Represents the solid molar volume of a phase. This indicates the molar rate of oxidation at the interface. , The oxidation reaction rate, This represents the local normal vector at the interface. , Represents the gradient operator. This indicates the gradient with respect to the solid phase. express The model; Indicates the molar concentration of oxygen; is the diffusion coefficient.

3. The method for obtaining the ablation morphology of C / C composite materials based on a physical information neural network according to claim 1, characterized in that, The expression for calculating the dimensionless parameter Sh is: in The fiber radius of the C / C composite material; The matrix oxidation reaction rate of the C / C composite material; is the diffusion coefficient.

4. The method for obtaining the ablation morphology of C / C composite materials based on a physical information neural network according to claim 1, characterized in that, The expression for calculating the dimensionless parameter A is: in The matrix oxidation reaction rate in the C / C composite material; This represents the molar volume of the matrix phase. The oxidation reaction rate of carbon fibers in C / C composite materials; denoted as , where is the molar volume of the fibrous phase.

5. The method for obtaining the ablation morphology of C / C composite materials based on a physical information neural network according to claim 1, characterized in that, The method for constructing a pre-trained multilayer perceptron network includes the following steps: The number of hidden layers and the number of neurons in each hidden layer of the multilayer perceptron network are set, and nonlinear transformations are performed in the hidden layers using activation functions; the number of hidden layers and the number of neurons are determined using the open-source framework Optuna. The gas-phase diffusion loss function, solid-phase evolution loss function, and initial condition loss function are constructed, and then the total loss function of the multilayer perceptron network is obtained. The three-dimensional spatial model of the C / C composite material with known ablation morphology and its initial conditions are used as the computational domain, and points are selected within the computational domain. The point selection method is uniform sampling, random sampling, or adaptive sampling. The selected points include oxygen molar concentration and solid volume fraction. The coordinates of the selected points, ablation time, dimensionless parameter Sh, and dimensionless parameter A are used as training samples and input into the multilayer perceptron network to obtain the oxygen molar concentration and solid volume fraction predicted by the multilayer perceptron network. The oxygen molar concentration and solid volume fraction of the selected points are used as the true labels. Combined with the oxygen molar concentration and solid volume fraction predicted by the multilayer perceptron network, the total loss value is calculated through the total loss function of the multilayer perceptron network. The parameters of the multilayer perceptron network are adjusted with the goal of minimizing the total loss value until the preset training termination condition is met, thus obtaining the pre-trained multilayer perceptron network.

6. The method for obtaining the ablation morphology of C / C composite materials based on a physical information neural network according to claim 5, characterized in that, The expression for the total loss function of a multilayer perceptron network is: in This is the total loss function; , and These are the weighting coefficients; Let be the gas phase diffusion loss function. ; Let be the solid-state evolution loss function. ; The initial conditional loss function, ; Indicates the molar concentration of oxygen; Indicates time; Represents the ladder operator; For solid volume fraction, when When represents the gas phase unit, when Time represents a solid phase unit; The diffusion coefficient is denoted as . This represents the molar concentration of the reacting gas; The square of the L2 norm; The velocity of the solid as it recedes; This represents the concentration at the outer edge of the boundary layer.

7. The method for obtaining the ablation morphology of C / C composite materials based on a physical information neural network according to claim 1, characterized in that, The specific method for constructing a three-dimensional spatial model of C / C composite materials is as follows: A three-dimensional spatial model of C / C composite material with a three-dimensional matrix encapsulating fiber monofilaments is established. The coordinate data of the three-dimensional spatial model of C / C composite material in the Cartesian coordinate system is obtained, and the cross-sectional description of the three-dimensional spatial model of C / C composite material is given.

8. The method for obtaining the ablation morphology of C / C composite materials based on a physical information neural network according to claim 2, characterized in that, The gas phase diffusion equation is: Among them, in the fluid domain and the solid domain At the interface ; is the volume fraction of solids; C is the molar concentration of the reactant gas; This represents a heterogeneous reaction rate. It represents the reciprocal of the gradient magnitude.

9. The method for obtaining the ablation morphology of C / C composite materials based on a physical information neural network according to claim 2, characterized in that, The solid phase evolution equation is: in It represents the volume fraction of solids. The velocity of the solid's retreat; in the fluid and solid domains At the interface ; Solid volume fraction; molar concentration of the reactant gas; This represents a heterogeneous reaction rate. It represents the reciprocal of the gradient magnitude.

10. The method for obtaining the ablation morphology of C / C composite materials based on a physical information neural network according to claim 1, characterized in that, The specific method for obtaining the ablation morphology of C / C composite materials by solving the gas phase diffusion equation and the solid phase evolution equation using the oxygen molar concentration and solid volume fraction output by the multilayer perceptron network is as follows: By using the oxygen molar concentration and solid volume fraction output by the multilayer perceptron network, the evolution of the oxygen molar concentration and solid volume fraction distribution of the material at different times is tracked through the gas phase diffusion equation and the solid phase evolution equation, thereby obtaining the development process of the ablation morphology of the three-dimensional C / C composite material over time.

Images