Method, device and equipment for predicting microscopic burning-denudation morphology of composite material and storage medium

Abstract

The invention provides a method, a device and equipment for predicting the microscopic burning-denudation morphology of a composite material, and a storage medium, and relates to the technical field of composite material microscopic morphology evolution prediction. The method comprises the steps of obtaining a data set; determining initial conditions, boundary conditions and denudation criteria of the composite material mesoscopic burning-denudation model; constructing a physical information neural network model based on mesoscopic burning-denudation surface topography evolution; inputting the data set into a physical information neural network model for iterative training to obtain a trained physical information neural network model; and inputting the to-be-detected composite material into the trained physical information neural network model to obtain a prediction result. According to the method provided by the invention, the microscopic morphology evolution in the composite material burning-denudation process can be effectively approached, the speed and precision of the microscopic burning-denudation morphology evolution solution of the composite material are greatly improved, and the method has a huge application prospect.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of composite material micro morphology evolution prediction, and in particular to a composite material micro ablation-erosion morphology evolution fast prediction method, device, equipment and storage medium based on a physical information neural network. BACKGROUND

[0002] Carbon-carbon composites are widely used in aerospace fields, especially in thermal protection systems under high-temperature and high-speed airflow environments. However, under extreme environments, C / C composites will undergo ablation processes, leading to surface degradation, crack propagation and accelerated ablation rate, which directly affects the performance and service life of the materials. Traditional ablation morphology evolution solving methods, such as numerical simulation methods, usually perform detailed calculations based on physical equations (such as heat conduction, fluid dynamics, chemical reactions, etc.), which are complex, time-consuming and require high computing resources, and are difficult to meet the real-time and high-efficiency requirements in engineering practice. In addition, the ablation morphology evolution at the micro scale is influenced by material properties, flow field dynamics, chemical reactions and other factors, and traditional methods often cannot provide accurate and efficient solutions at the micro scale.

[0003] Therefore, there is an urgent need for a new composite material micro ablation-erosion morphology evolution fast prediction method to solve the problems encountered in the prior art. SUMMARY

[0004] The present application provides a composite material micro ablation-erosion morphology evolution fast prediction method, device, equipment and storage medium based on a physical information neural network, which is used to quickly and efficiently predict the micro morphology evolution in the ablation-erosion process of composite materials.

[0005] According to a first aspect of the present application, the present application provides a composite material micro ablation-erosion morphology prediction method, comprising: obtaining a data set; wherein the data set comprises the surface micro ablation-erosion morphology of carbon-carbon composites obtained by processing the micro ablation model of the carbon-carbon composites constructed based on the fixed volume fraction method; determining the initial conditions, boundary conditions and erosion criteria of the composite material micro ablation-erosion model; constructing a physical information neural network model based on the micro ablation-erosion surface morphology evolution of the composite material; inputting the data set into the physical information neural network model for iterative training to obtain a trained physical information neural network model; inputting the composite material to be detected into the trained physical information neural network model to obtain a composite material micro ablation-erosion morphology evolution prediction result.

[0006] The beneficial effects of the above technical solutions are: the application mainly combines deep learning technology and physical information neural network technology to solve the thermo-chemical ablation equation, can effectively approximate the micro-morphology evolution in the ablation and ablation process of the composite material under the premise of ensuring physical consistency, greatly improves the speed and accuracy of solving the micro-ablation and ablation morphology evolution of the composite material; the application technology can be widely applied in the field of aerospace, especially in the application of accurately predicting the ablation and ablation process and its influence on the structural performance, such as the outer surface of the aircraft, the engine nozzle, and the thermal protection system.

[0007] In an available implementation, the obtaining the data set comprises: constructing a micro-geometric model of the composite material; constructing a composite material ablation model based on the micro-geometric model; performing ablation recession calculation on each element in the geometric model according to the micro-ablation model of the composite material based on the solid volume fraction method, obtaining a data set of the solid volume fraction of each element, and forming a composite material surface micro-ablation and ablation morphology.

[0008] In an available implementation, the constructing a micro-geometric model of the composite material comprises: establishing a carbon-carbon composite material micro-geometric model including carbon fibers and carbon matrix based on TexGen, and converting the micro-geometric model into a discrete model of cubic elements; assigning material properties to the carbon fibers and carbon matrix, the material properties including fiber volume fraction, density, thermal conductivity, porosity, and specific heat capacity.

[0009] In an available implementation, the constructing a composite material ablation model based on the micro-geometric model comprises: constructing a spatial surface function of the composite material in three-dimensional coordinates; when the composite material ablates and recedes, each point on the surface recedes a certain distance along the normal vector direction within a certain time interval to form a new surface function after ablation; and a gas diffusion equation of the carbon matrix of the composite material when it ablates and recedes; the spatial surface function of the composite material in three-dimensional coordinates is: when the composite material ablates and recedes, each point on the surface recedes a certain distance along the normal vector direction within a certain time interval to form a new surface function after ablation, which is:

[0010]

[0011]

[0012] wherein, It is a material surface function, present at the gas-solid or liquid-solid interface. ; These are spatial coordinates; It is time; It is the reaction concentration; It refers to the speed at which the interface goes back; k It is a heterogeneous reaction rate; It is the molar volume of the solid phase; It is the outward normal of the solid phase interface; It is a gradient operator; It is the magnitude of the vector; The gas diffusion equation for the carbon matrix of the composite material during ablation retreat is as follows:

[0013] in, It is the reaction concentration; It is time; It is the diffusion rate; It is the gradient operator.

[0014] In one feasible implementation, the solid volume fraction-based method involves performing ablation regression calculations on each element of the geometric model according to a microscopic ablation model of the composite material to obtain a dataset of the solid volume fraction for each element, including: Define the solid volume fraction of each unit as:

[0015] The oxygen diffusion equation and the new surface function after ablation, rewritten using the solid volume fraction, are as follows:

[0016]

[0017] in, It is the volume of the solid in the unit cell; It is the unit volume; Indicates a gas phase unit. Represents a solid-state unit; 0 < <1 represents a hybrid unit; It is the reaction concentration; It is time; It is the diffusion coefficient; It is the outward normal of the solid phase interface; k It is a heterogeneous reaction rate; It is a gradient operator; It is the absolute value of the screen back speed; The finite volume method is used to solve the oxygen diffusion equation and the new curved surface function after ablation to obtain the data set of the solid volume fraction.

[0018] In an available embodiment, the initial conditions, boundary conditions and ablation criterion of the micro-ablation model of the composite material are determined, comprising: Setting the initial conditions of the calculation domain includes temperature field, gas concentration and material morphology; the temperature field is set as the initial temperature of the composite material and the environment ; the gas concentration is set as the environmental gas concentration ; the material morphology is set as the initial surface shape and micro-morphology distribution of the material ; Setting the boundary conditions of the calculation domain includes the top boundary and the lateral boundary; the top boundary adopts Dirichlet condition, (wherein represents the gas concentration in the initial state); the lateral boundary is symmetrical, i.e. the normal gradient of the temperature and the gas concentration is 0; The ablation criterion is the mechanical ablation criterion, specifically:

[0019]

[0020]

[0021]

[0022] wherein, is the height of the n-th layer in the composite material model, ; is the maximum height of the surface yarn in the composite material model; is the average height of the surface matrix in the composite material model; is the mechanical ablation coefficient related to the material and the environment; is the width of the weft yarn in the composite material model; is the number of warp yarns covering the weft yarn; is the length of the weft yarn; is the height of the n-th layer weft yarn; warp is the warp yarn in the composite material model; welf is the weft yarn in the composite material model; if n is warp, s is welf; if n is welf, s is warp; The mechanical erosion amount thereof in the micro scale is:

[0023] wherein, is the average height of the surface matrix;​​ is the height of the surface yarn in the composite material model; is the mechanical erosion amount.

[0024] In an implementable embodiment, the construction is based on a physical information neural network model of the composite material meso-ablation-erosion surface morphology evolution, comprising: Based on the determined initial conditions, boundary conditions and erosion criteria of the composite material meso-ablation-erosion model, residual points, boundary constraint points and initial constraint points for training are selected in the calculation domain; Construct a self-defined multi-layer perception neural network structure; Based on the self-defined improved multi-layer perception neural network structure, a neural network is constructed, input training points are trained, and a physical information neural network model is established; According to the physical information neural network model, a loss function is obtained, training is performed, and a final physical information neural network model based on the composite material meso-ablation-erosion surface morphology evolution is obtained.

[0025] In an implementable embodiment, the selection of residual points, boundary constraint points and initial constraint points for training in the calculation domain comprises: Residual points are selected in the composite material calculation domain by Latin hypercube random sampling; Boundary constraint points are set on the top boundary of the composite material, and boundary constraint points are set on the lateral boundary by applying symmetry conditions; Initial constraint points of temperature and gas concentration are set on the surface and inside of the composite material.

[0026] In an implementable embodiment, the construction of a self-defined multi-layer perception neural network structure comprises: an input layer, an encoder layer, a fusion layer, a hidden layer, an automatic differentiation layer and an output layer; The encoder layer comprises an ablation encoder and an erosion encoder; the ablation encoder is used for feature encoding processing of input data of the composite material ablation process, and outputs an ablation process feature vector; the erosion encoder is used for feature encoding processing of input data of the composite material erosion process, and outputs an erosion process feature vector; The input data of the composite material ablation process comprises surface temperature and gas concentration in the composite material ablation process; The input data of the composite material erosion process comprises airflow velocity and surface morphology in the composite material erosion process.

[0027] In an implementable embodiment, the loss function obtained according to the physical information neural network model comprises: The physical equation constraint is introduced into the physical information neural network model, the automatic differentiation technology is used to calculate the physical constraint loss, the boundary condition loss, the initial condition loss and the mechanical erosion condition loss, and the final loss function is obtained in the form of weighted sum.

[0028] In an available implementation, the physical constraint loss, the boundary condition loss, the initial condition loss and the mechanical erosion condition loss adopt the following form: The physical constraint loss, i.e. the loss function of the gas diffusion equation, is:

[0029] wherein, is the reaction concentration; is time; is the diffusion coefficient; represents the gas phase unit, represents the solid phase unit; 0 <1 represents the mixed unit; is the gradient operator; is the physical constraint loss function; The loss function of the boundary condition and the initial condition is:

[0030]

[0031] wherein, is the reaction concentration; is the gas concentration in the initial state; is the environmental gas concentration; is the weight coefficient, used for balancing the relative importance of different losses; is the temperature of the composite material; is the initial temperature of the environment; is the boundary condition loss function; is the initial condition loss function; The loss function of the mechanical erosion condition is:

[0032] wherein, is the average height of the surface matrix; is the height of the surface yarn in the composite material model; is the mechanical erosion amount. is the mechanical erosion constraint loss function; The final loss function is the weighted sum of the physical constraint loss, the boundary condition loss, the initial condition loss and the mechanical erosion constraint condition loss: .

[0033] wherein, are weight coefficients for balancing the relative importance between different sub-loss functions; is the final loss function; according to the second aspect of the present application, the present application provides a composite material meso-ablation and spallation morphology evolution rapid prediction device based on a physical information neural network, comprising: a data acquisition module for acquiring a data set; wherein the data set comprises a carbon-carbon composite material surface meso-ablation and spallation morphology obtained by processing a constructed meso-ablation model of carbon-carbon composite material based on a fixed volume fraction method; a model condition confirmation module for determining initial conditions, boundary conditions and spallation criteria of the composite material meso-ablation and spallation model; a model construction module for constructing a physical information neural network model based on the evolution of the composite material meso-ablation and spallation surface morphology; a model training module for inputting the data set into the physical information neural network model for iterative training to obtain a trained physical information neural network model; a meso-ablation and spallation morphology evolution rapid prediction module for inputting a composite material to be detected into the trained physical information neural network model to obtain a composite material meso-ablation and spallation morphology evolution prediction result.

[0034] According to the third aspect of the present application, an electronic device is provided, comprising a processor and a memory connected in communication with the processor; The memory stores computer execution instructions; The processor executes the computer execution instructions stored in the memory to implement the method of any one of the first aspect.

[0035] According to the fourth aspect of the present application, a computer readable storage medium is provided, the computer readable storage medium stores computer execution instructions, and the computer execution instructions are executed by the processor to implement the method of any one of the first aspect.

[0036] According to the fifth aspect of the present application, a computer program product is provided, comprising a computer program, and the computer program is executed by the processor to implement the method of any one of the first aspect.

[0037] Compared with the prior art, the present application has the following beneficial effects: (1) The physical information neural network-based composite material meso-ablation-erosion morphology evolution rapid prediction method provided by the application solves the problems of traditional finite volume method in grid division, difference format dimension disaster, multi-scale processing and the like, has significant computing efficiency and flexibility advantages, and in particular in the fields of complex boundary conditions, high-dimensional problems and data and physical coupling solving, the PINN (physical information neural network) provides a new solution for rapid simulation and prediction of ablation-erosion problems.

[0038] (2) Compared with the full connection neural network structure of the baseline PINN model, the self-defined multi-layer perception neural network structure provided by the application has significant advantages in expression ability, physical constraint processing ability and computing efficiency through specific optimization for ablation and erosion problems, can better handle complex boundaries, multi-scale problems and generalization ability, and improves the precision and training stability of the model.

[0039] (3) The PINN method provided by the application can directly start from the physical equation through the neural network to quickly predict the morphology evolution of the entire composite material meso-ablation-erosion process, without layer-by-layer calculation and surface updating, the network can solve the entire process at one time, and directly give the overall ablation-erosion morphology under time evolution, greatly improving the computing efficiency and reducing the time cost of layer-by-layer calculation. BRIEF DESCRIPTION OF DRAWINGS

[0040] The accompanying drawings, which are incorporated into and form part of the specification, illustrate embodiments consistent with the present application and, together with the specification, serve to explain the principles of the application.

[0041] Figure 1 A flowchart of the composite material meso-ablation-erosion morphology evolution rapid prediction method provided by the embodiment of the application is shown. Figure 2 A schematic diagram of the composite material meso-geometric model provided by the embodiment of the application is shown. Figure 3 A schematic diagram of the composite material geometric model provided by the embodiment of the application is shown. Figure 4 A schematic diagram of the self-defined improved multi-layer perception neural network structure provided by the embodiment of the application is shown.

[0042] Figure 5 A schematic diagram of the structure of the composite material meso-ablation-erosion morphology evolution rapid prediction device provided by the embodiment of the application is shown. Figure 6 A schematic diagram of the structure of the electronic device provided by the embodiment of the application is shown.

[0043] The specific embodiments of the present application have been shown by the above drawings, and will be described in more detail hereinafter. These drawings and the written description are not intended to restrict the scope of the present application concept in any way, but to illustrate the present application concept to those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION

[0044] The exemplary embodiments will be described in detail herein below with reference to the drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The implementations described in the following exemplary embodiments are not meant to represent all implementations consistent with the present application.

[0045] In order to improve the solving speed and accuracy of the composite material burn-ablation morphology evolution, the present application proposes a mesoscopic burn-ablation morphology evolution fast prediction method based on physical information neural network (PINN); the present application combines deep learning technology and physical law constraint, and trains the neural network to enable it to effectively approximate the mesoscopic morphology evolution in the complex burn-ablation process of the composite material, and provides an efficient and accurate solving and prediction method.

[0046] The technical solutions of the composite material mesoscopic burn-ablation morphology evolution fast prediction method based on physical information neural network provided by the present application will be described in detail below through specific embodiments. It should be noted that the following embodiments can exist alone or in combination, and for the same or similar content, which may not be repeated in different embodiments.

[0047] Figure 1 A flowchart of a composite material mesoscopic burn-ablation morphology evolution fast prediction method based on physical information neural network provided by an embodiment of the present application is shown in Figure 1 In some embodiments, the flow of the composite material mesoscopic burn-ablation morphology evolution fast prediction method includes the following steps: S101, obtaining a data set; wherein the data set includes the surface mesoscopic burn-ablation morphology of carbon-carbon composite material obtained by processing the mesoscopic ablation model of the carbon-carbon composite material constructed based on the fixed volume fraction method; S102, determining the initial condition, boundary condition and ablation criterion of the composite material mesoscopic burn-ablation model; S103, constructing a physical information neural network model based on the evolution of the surface morphology of the composite material mesoscopic burn-ablation; S104, inputting the data set into the physical information neural network model for iterative training to obtain a trained physical information neural network model; S105, inputting the composite material to be detected into the trained physical information neural network model to obtain the prediction result of the composite material mesoscopic burn-ablation morphology evolution.

[0048] In step S101, the specific steps for obtaining the data set are as follows: A meso-geometric model of the composite material is constructed, in this embodiment, a meso-geometric model of the C / C composite material is constructed, which is composed of carbon fibers and carbon matrix; first, a meso-geometric model is established by TexGen, and the meso-geometric model is converted into a discrete model of cubic elements, as shown in FIG. 1, and then material properties of the carbon fibers and the matrix, such as fiber volume fraction, density, thermal conductivity, porosity and specific heat capacity, are assigned. Figure 2

[0049] Based on the above-mentioned meso-geometric model, an ablation model of the composite material is constructed; in this embodiment, a spatial surface function of the carbon-carbon composite material in three-dimensional coordinates is constructed; when the composite material retreats after ablation, each point on the surface retreats a certain distance along the normal vector direction within a certain time interval to form a new surface function after ablation; and a gas diffusion equation of the carbon matrix of the composite material when it retreats after ablation; specifically: The thermo-chemical ablation of the carbonized composite material mainly includes oxidation, nitridation and sublimation of carbon; the carbonized composite material will mainly undergo the thermo-chemical ablation phenomenon caused by oxidation; in this embodiment, only the oxygen diffusion and thermo-chemical reaction at the ablation surface of the C / C composite material are considered, and a meso-ablation model of the C / C composite material is established; during the thermo-chemical ablation process, the heat and mass transfer of the carbon oxidation reaction leads to the recession of the gas-solid interface of the C / C composite material, and the recession of the gas-solid interface changes the distribution of the oxygen concentration in the boundary layer, which in turn affects the reaction rate of carbon and the recession rate of the gas-solid interface; therefore, the carbon reaction rate and the distribution of the oxygen concentration in the boundary layer are important parameters for establishing an oxidation kinetic model representing the recession of the surface of the C / C composite material. First, a spatial surface function of the C / C composite material in three-dimensional coordinates is constructed to represent the gas / solid phase interface of the C / C composite material; when the composite material retreats during the ablation process, each point on the surface retreats a certain distance along the normal vector direction within a time interval to form a new surface function after ablation; wherein the spatial surface function of the composite material in three-dimensional coordinates is:

[0050] When the composite material retreats after ablation, each point on the surface retreats a certain distance along the normal vector direction within a certain time interval to form a new surface function after ablation:

[0051]

[0052]

[0053] wherein, ​is the material surface function, which is on the gas-solid phase or liquid-solid phase interface ; is the spatial coordinate; is the time; is the reaction concentration; is the interface recession velocity; k is the heterogeneous reaction rate; is the molar volume of the solid phase; is the outer normal of the solid phase interface; is the gradient operator; is the modulus of the vector; When the carbon matrix of the C / C composite material recedes after ablation, according to the law of conservation of mass, the gas diffusion equation is:

[0054] wherein, is the reaction concentration; is the time; is the diffusion rate; is the gradient operator; Based on the solid volume fraction method, the ablation recession calculation of each unit in the discrete model is carried out according to the meso-ablation model of the composite material, and the data set of the solid volume fraction of each unit is obtained to form the surface meso-ablation and ablation morphology of the composite material; Specifically, it includes: Based on the solid volume fraction method, the solid volume fraction of each unit in the discrete model is defined as:

[0055] wherein, is the solid volume in the unit; is the unit volume; The above oxygen diffusion equation and the new surface function after ablation are rewritten as:

[0056]

[0057] wherein, is the solid volume in the unit; is the unit volume; indicates a gas phase unit, indicates a solid phase unit; 0 <1 represents a mixed unit; is the reaction concentration; is the time; is the diffusion coefficient; is the outer normal of the solid phase interface; k is the heterogeneous reaction rate; is the gradient operator; is the absolute value of the interface recession velocity; The finite volume method is used to solve the oxygen diffusion equation and the new surface function after ablation to obtain the data set of solid volume fraction.

[0058] In step S102, the initial conditions, boundary conditions and ablation criteria of the composite material meso-ablation model are determined, and the specific steps are: The initial conditions of the calculation domain include temperature field, gas concentration and material morphology; the temperature field is set as the initial temperature of the composite material and the environment ; the gas concentration is set as the environmental gas concentration ; and the material morphology is set as the initial surface shape and meso-morphology distribution of the material ; A unit cell model is established, which includes warp yarns, yarns and matrix, and the boundary conditions of the calculation domain include the top boundary and the lateral boundary, as shown in Figure 2 ; the top boundary adopts Dirichlet condition, (wherein represents the initial state gas concentration); the lateral boundary is symmetrical, i.e. the normal gradient of temperature and gas concentration is 0; Specifically, in order to accurately simulate the ablation morphology of C / C composite material, the setting of initial conditions is very important, which mainly includes temperature field, gas concentration and material morphology. These conditions determine the initial state of simulation, thereby affecting the prediction results of ablation behavior; the temperature field refers to the initial temperature distribution of the material inside and surface, which is usually set as the environmental temperature, i.e. ; the material morphology refers to the material surface shape and meso-structure characteristics, i.e. ; the gas concentration refers to the initial concentration of ablation reaction products, which is usually set as the environmental concentration, i.e. ; wherein represents the initial temperature of the material and the environment, represents the environmental gas concentration, represents the initial surface height or morphology distribution; the boundary conditions of the model domain mainly include the top boundary and the lateral boundary, the top boundary adopts Dirichlet condition, i.e. ; the lateral boundary is assumed to be symmetric, i.e. the normal gradient of temperature and gas concentration is zero; during the thermo-chemical ablation process, the constraint of the carbon matrix on the exposed fibers gradually decreases when most of the fibers are above the carbon matrix surface, the mass loss and the morphology change caused by non-thermo-chemical ablation such as mechanical erosion caused by airflow shear, aerodynamic blowing, etc. are considered; therefore, in the high-speed airflow environment, the shear force of the ablated parallel fibers far exceeds the critical force required for mechanical erosion to occur, after the mechanical erosion of the parallel yarns, the remaining material will continue to undergo thermo-chemical ablation until the matrix surface is again below the yarn surface height; during this process, thermo-chemical ablation and mechanical erosion will alternate until all the C / C composites are completely consumed; in this cyclic process, mechanical erosion will change the conditions of subsequent thermo-chemical ablation, such as the position of the gas / solid phase reaction interface and the distribution of gas concentration, at the same time, mechanical erosion will also affect the ablation performance of C / C composites, such as the surface degradation and surface roughness change of the material; according to the mechanical erosion behavior of the parallel yarns, the erosion criterion of the C / C composite is as follows:

[0059]

[0060]

[0061]

[0062] wherein, is the height of the nth layer in the composite model, ; is the maximum height of the surface yarns in the composite model; is the average height of the surface matrix in the composite model; is the mechanical erosion coefficient related to the material and the environment; is the width of the weft yarn in the composite model; is the number of warp yarns covering the weft yarn; is the length of the weft yarn; is the height of the nth layer weft yarn; warp is the warp yarn in the composite model; welf is the weft yarn in the composite model; if n is warp, s is welf; if n is welf, s is warp; When the mechanical erosion criterion of the above formula is met, the part of the exposed yarn height greater than will be peeled off; therefore, the amount of mechanical erosion at the mesoscale is:

[0063] wherein, is the average height of the surface matrix;​​ is the height of the surface yarn in the composite material model; is the mechanical erosion amount.

[0064] At step S103, a physical information neural network model based on the evolution of the microscale burn-off erosion surface morphology of the composite material is constructed, and the specific steps are as follows: Based on the determined initial conditions, boundary conditions and erosion criteria of the microscale burn-off erosion model of the composite material, residual points, boundary constraint points and initial constraint points for training are selected in the calculation domain; specifically: In order to construct a physical information neural network that conforms to the evolution of the burn-off erosion surface morphology, residual points, boundary constraint points and initial constraint points need to be reasonably selected in the calculation domain. These points are crucial for the training of the neural network, as they ensure that the model can effectively meet the physical constraint conditions and provide accurate predictions when solving the thermo-chemical ablation process; among them, the residual points are points used to minimize the residual of the physical equation in the optimization process of the neural network. For the thermo-chemical ablation simulation of C / C composite materials, in this embodiment, the residual points are distributed in the internal region and near the surface of the material, especially in the region where the gas flow contacts the material surface, and Latin hypercube sampling is used for random sampling in the calculation domain to ensure better coverage of the region; the boundary constraint points are used to apply physical boundary conditions to ensure that the solution of the neural network is consistent with the boundary conditions of the problem. For the thermo-chemical ablation simulation of C / C composite materials, this embodiment sets boundary constraint points on the surface of the material (top boundary) to specify the surface temperature as the ambient temperature, and applies symmetry conditions on the lateral boundary; the initial constraint points are used to specify the initial conditions of the simulation to ensure that the neural network correctly reflects the initial state of the material when starting training; for the thermo-chemical ablation simulation of C / C composite materials, this embodiment sets the initial conditions of temperature and gas concentration on the surface and inside.

[0065] A self-defined multi-layer perception neural network structure is constructed, and the structure of the improved multi-layer perception neural network is as shown in Figure 4 The neural network is constructed based on the self-defined improved multi-layer perception neural network structure, the training points are input for training, and the physical information neural network model is established; specifically: In order to construct a burn-ablation surface morphology evolution physical information neural network, first, a basic multilayer perception neural network is defined, which can effectively learn complex patterns and relationships mainly through multiple layers of weighted connections and nonlinear activation, and its main advantages are simple structure, easy to implement, and strong expression ability; based on this framework, in order to enable the neural network to simultaneously learn and process physical information and features related to the ablation process and the ablation process, the embodiment adds a burn encoder and an ablation encoder to the standard fully connected neural network structure, which extracts and transforms important features related to ablation and ablation by encoding or mapping the original input, the self-defined neural network structure can handle complex nonlinear relationships, and adapt to high-dimensional and dynamic ablation and ablation processes; first, the input layer includes time, spatial position (coordinates) and other information, then the burn encoder and the ablation encoder are added, the purpose of the burn encoder is to encode the input data related to the burn, such as gas concentration and surface temperature, and the output will be a high-dimensional feature vector, which contains important physical information in the burn process, such as linear burn recession rate and mass burn rate, and the high-dimensional feature vector is denoted as ; the purpose of the ablation encoder is to extract and encode input data related to the material surface ablation process, such as gas flow velocity, and the output is a feature vector containing ablation process information, i.e. mechanical erosion amount, and the feature vector is denoted as ; the outputs of the above two encoders and are fused through a feature fusion layer, and the fused feature vector is input to the subsequent neural network layer for further nonlinear transformation; then the fused feature vector is transmitted to multiple hidden layers, and the fused features are processed through a fully connected layer, that is:

[0066] wherein, is an activation function (such as ReLU, tanh, etc.), is the weight matrix of the first layer in the neural network, is the bias vector, indicates the fused feature vector; is the hidden state of the lth layer in the neural network; According to the physical information neural network model, a loss function is obtained, and the final physical information neural network model based on the micro ablation-ablation surface morphology evolution of the composite material is obtained by training; specifically: In the physical information neural network model, physical equation constraints are introduced, and automatic differentiation technology is used to calculate the physical constraint loss, boundary condition loss, initial condition loss and mechanical ablation condition loss, and the final loss function is obtained in the form of weighted sum; the specific process is: According to the description above, the loss function is obtained by the burn-off erosion physical information neural network, and the embodiment introduces the constraint of the physical equation in the training process of the neural network, that is, the automatic differentiation is added between the last layer of the full connection layer and the output layer, the purpose is to enable the network to calculate the gradient of the physical equation such as the gas diffusion equation and the new surface function after ablation, and optimize the parameters of the model according to the gradient; the automatic differentiation automatically calculates the gradient of the loss function to the network weight and input by tracking each operation in the network calculation process; wherein, the physical constraint loss, the boundary condition loss, the initial condition loss and the mechanical erosion condition loss adopt the following form: The physical constraint loss, that is, the loss function of the gas diffusion equation, is as follows:

[0067] wherein, is the reaction concentration; is the time; is the diffusion coefficient; represents the gas phase unit, represents the solid phase unit;0 <1 represents the mixed unit; is the gradient operator; is the physical constraint loss function; The loss function of the boundary condition and the initial condition is as follows:

[0068]

[0069] wherein, is the reaction concentration; is the initial state gas concentration; is the environmental gas concentration; is the weight coefficient, used to balance the relative importance of different losses; is the temperature of the composite material; is the initial temperature of the environment; is the boundary condition loss function; is the initial condition loss function; The loss function of the mechanical erosion condition is as follows:

[0070] wherein, is the average height of the surface matrix; is the height of the surface yarn in the composite material model; is the mechanical erosion amount; is the mechanical erosion constraint loss function; The embodiment further combines the above loss functions in a weighted sum form by setting a weight factor, and the weight factor determines the relative importance of the above three losses in the final loss function. The final loss function can be represented as:

[0071] wherein, is a weight coefficient for balancing the relative importance between different sub-loss functions; is the final loss function; The present application builds an ablation-erosion model based on a physical information neural network (PINN) through the above steps, and calculates the physical constraint loss through automatic differentiation, combines it with the loss of boundary conditions and initial conditions for training; During the training process, the optimizer (such as Adam) can be used to update the network parameters, and the weight factor is set to balance the physical loss and the data loss; After training, the effect of the model is tested by evaluation indicators and visualization means.

[0072] Finally, according to the trained physical information neural network, the present application can solve the surface morphology evolution of the composite material mesoscopic ablation-erosion model under different working conditions. The specific steps are: (1) load the trained model: after training, save and load the trained model to be used for predicting the surface morphology evolution under different working conditions; (2) input different working condition data: the working condition data can be different physical conditions such as gas diffusion concentration, pressure field, time, etc. By inputting these data into the trained model, the physical information neural network will predict the corresponding surface morphology change according to the physical law it has learned; (3) predict the mesoscopic surface morphology: by inputting different working conditions such as different spatial positions, time points, etc. into the network, the corresponding surface morphology such as ablation depth and mechanical erosion amount can be obtained.

[0073] The PINN method proposed by the present application can directly start from the physical equation through the neural network to quickly predict the morphology evolution of the whole ablation-erosion process. Without layer-by-layer calculation and surface update, the network can solve the whole process at one time and directly give the overall ablation-erosion morphology under time evolution, greatly improving the calculation efficiency and having great application prospect.

[0074] Figure 5 is a structural schematic diagram of a composite material mesoscopic ablation-erosion morphology evolution rapid prediction device provided by an embodiment of the present application, referring to Figure 5 The mesoscopic ablation-erosion morphology evolution rapid prediction device includes various functional modules for realizing the above mesoscopic ablation-erosion morphology evolution rapid prediction method. Any functional module can be realized by software and / or hardware.

[0075] In some embodiments, the composite material meso-ablation and spallation morphology evolution rapid prediction device 1000 based on the physical information neural network comprises a data acquisition module 1001, a model condition confirmation module 1002, a model construction module 1003, a model training module 1004, and a meso-ablation and spallation morphology evolution rapid prediction module 1005; wherein: The data acquisition module 1001 is configured to acquire a data set, wherein the data set comprises a surface meso-ablation and spallation morphology of a carbon-carbon composite material obtained by processing a constructed meso-ablation model of the carbon-carbon composite material based on a fixed volume fraction method; The model condition confirmation module 1002 is configured to determine initial conditions, boundary conditions, and spallation criteria of the composite material meso-ablation and spallation model; The model construction module 1003 is configured to construct a physical information neural network model based on the evolution of the surface morphology of the composite material meso-ablation and spallation; The model training module 1004 is configured to input the data set into the physical information neural network model for iterative training to obtain a trained physical information neural network model; The meso-ablation and spallation morphology evolution rapid prediction module 1005 is configured to input a composite material to be detected into the trained physical information neural network model to obtain a prediction result of the meso-ablation and spallation morphology evolution of the composite material.

[0076] The composite material meso-ablation and spallation morphology evolution rapid prediction device 1000 provided by the embodiments of the present application is used to execute the technical solutions provided by the foregoing composite material meso-ablation and spallation morphology evolution rapid prediction method embodiments, and has similar implementation principles and technical effects to the foregoing method embodiments, which will not be described here again.

[0077] It should be noted that the division of each module of the above apparatus is only a logical function division, and all or part of them can be integrated into a physical entity or physically separated when actually implemented. Moreover, all of these modules can be implemented in the form of software invoked by a processing element, or all of them can be implemented in the form of hardware, or part of the modules can be implemented in the form of software invoked by a processing element, and part of the modules can be implemented in the form of hardware. For example, the data acquisition module 1001 can be a separately arranged processing element, or can be implemented in a chip of the above apparatus, in addition, it can also be stored in the form of program code in the memory of the above apparatus, and the function of the above data acquisition module 1001 is invoked and executed by a processing element of the above apparatus. The implementation of other modules is similar. In addition, all or part of these modules can be integrated together, or can be independently implemented. The processing element here can be an integrated circuit with signal processing capability. In the implementation process, each step of the above method or each module can be completed by the integrated logic circuit of hardware or the instruction of software in the processing element.

[0078] Figure 6 A structural schematic diagram of an electronic device provided by an embodiment of the present application is shown in FIG. 11. Figure 6 The electronic device 1100 includes a processor 1101 and a memory 1102 connected with the processor 1101 in communication; The memory 1102 stores computer execution instructions; The processor 1101 executes the computer execution instructions stored in the memory 1102 to realize the technical solutions of the above-mentioned composite material mesoscopic burn-erosion morphology evolution rapid prediction method.

[0079] In the electronic device 1100 described above, the memory 1102 and the processor 1101 are directly or indirectly electrically connected to realize the transmission or interaction of data. For example, these elements can be electrically connected to each other through one or more communication buses or signal lines, such as through bus connection. 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 only one type of bus. The memory 1102 stores computer execution instructions for implementing the aforementioned thermal protection material surface roughness prediction method, including at least one software function module stored in the memory 1102 in the form of software or firmware, and the processor 1101 executes various function applications and data processing by running the software program and module stored in the memory 1102.

[0080] The memory 1102 at least includes one type of readable storage medium, not limited to a Random Access Memory (RAM), a Read Only Memory (ROM), a Programmable Read-Only Memory (PROM), an Erasable Programmable Read-Only Memory (EPROM), an Electric Erasable Programmable Read-Only Memory (EEPROM), etc. Among them, the memory 1102 is used to store programs, and the processor 1101 executes the programs after receiving execution instructions. Further, the software program and module in the memory 1102 described above can also include an operating system, which can include various software components and / or drivers for managing system tasks (such as memory management, storage device control, power management, etc.), and can communicate with various hardware or software components to provide an operating environment for other software components.

[0081] The processor 1101 can be an integrated circuit chip having a processing capability of signals. The processor 1101 described above can be a general processor, including a central processing unit (CPU), a network processor (NP), a digital signal processor (DSP), an application specific integrated circuit (ASIC), etc. The processor can implement or execute the disclosed methods, steps and logic block diagrams in the embodiments of the present application. The general processor can be a microprocessor, or the processor 1101 can also be any conventional processor or the like.

[0082] The electronic device 1100 is configured to implement the technical solutions provided by the aforementioned composite material microscopic burn-erosion morphology evolution fast prediction method embodiments, and has similar implementation principles and technical effects to the aforementioned method embodiments, which will not be described here again.

[0083] The present application also provides a computer readable storage medium, which stores computer execution instructions, and when the processor executes the computer execution instructions, the technical solutions of the aforementioned thermal protection material surface roughness prediction method are implemented.

[0084] The aforementioned computer readable storage medium can be implemented by any type of volatile or non-volatile storage device or their combination, 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 storage, 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.

[0085] An exemplary readable storage medium is coupled to the processor, so that the processor can read information from the readable storage medium, and can 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 (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.

[0086] The embodiment of the present application further provides a computer program product comprising a computer program which, when executed by a processor, is configured to implement the technical solution of the aforementioned thermal protection material surface roughness prediction method.

[0087] In the above embodiments, those skilled in the art can understand that the implementation of the above-mentioned method embodiments can be realized by software, hardware, firmware or any combination thereof, in whole or in part. When implemented by software, it can be realized in whole or in part in the form of a computer program product. The computer program product comprises one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the flow or function according to the embodiment of the present application 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 devices. The computer instructions can be stored in a computer readable storage medium or transmitted from one computer readable storage medium to another, for example, the computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center through wired (for example, coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (for example, infrared, wireless network, microwave, etc.) mode. 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, data center, etc. integrated with one or more available media sets. The available media can be a magnetic medium (for example, a floppy disk, a hard disk, a magnetic tape), an optical medium (for example, a DVD), or a semiconductor medium (for example, a solid state disk (SSD)) and the like.

[0088] It should be understood that the present application is not limited to the precise construction that has been described above and illustrated in the accompanying drawings, and that various modifications and changes can be made without departing from the scope thereof.

Claims

1. A method for predicting the microscopic burn-erosion morphology of composite materials, characterized in that, include: Obtain the dataset; wherein the dataset includes the micro-ablation morphology of the carbon-carbon composite surface obtained by processing the micro-ablation model of the constructed carbon-carbon composite material based on the fixed volume fraction method. Determine the initial conditions, boundary conditions, and erosion criteria for the microscopic burn-erosion model of composite materials; Construct a physical information neural network model based on the evolution of microscopic sintered-eroded surface morphology of composite materials; The dataset is input into the physical information neural network model for iterative training to obtain the trained physical information neural network model. The composite material to be tested is input into the trained physical information neural network model to obtain the prediction results of the microscopic burn-erosion morphology evolution of the composite material.

2. The method according to claim 1, characterized in that, The acquisition of the dataset includes: Construct a microscopic geometric model of the composite material; Based on the aforementioned microscopic geometric model, a composite material ablation model is constructed; Based on the solid volume fraction method, the ablation retreat calculation is performed on each unit in the geometric model according to the microscopic ablation model of the composite material to obtain the solid volume fraction dataset of each unit, so as to constitute the microscopic ablation-exfoliation morphology of the composite material surface.

3. The method according to claim 2, characterized in that, The microscopic geometric model for constructing the composite material includes: A microscopic geometric model of carbon-carbon composite materials, including carbon fibers and a carbon matrix, was established based on TexGen, and the microscopic geometric model was transformed into a discrete model of cubic elements. The carbon fibers and carbon matrix are endowed with material properties including fiber volume fraction, density, thermal conductivity, porosity, and specific heat capacity.

4. The method according to claim 3, characterized in that, The construction of the composite material ablation model based on the microscopic geometric model includes: Construct the spatial surface function of the composite material in three-dimensional coordinates; when the composite material is ablated and retreats, each point on the surface retreats a certain distance along the normal vector direction within a certain time interval to form a new surface function after ablation; and the gas diffusion equation of the carbon matrix of the composite material during ablation and retreat. The spatial surface function of the composite material in three-dimensional coordinates is: As the composite material ablates and recedes, each point on the surface recedes a certain distance along the normal vector direction within a certain time interval, forming a new surface function after ablation. in, It is a material surface function, present at the gas-solid or liquid-solid interface. ; These are spatial coordinates; It is time; It is the reaction concentration; It refers to the speed at which the interface goes back; k It is a heterogeneous reaction rate; It is the molar volume of the solid phase; It is the outward normal of the solid phase interface; It is a gradient operator; It is the magnitude of the vector; The gas diffusion equation for the carbon matrix of the composite material during ablation retreat is as follows: in, It is the reaction concentration; It is time; It is the diffusion rate; It is the gradient operator.

5. The method according to claim 4, characterized in that, The solid volume fraction method involves performing ablation regression calculations on each element of the geometric model based on the microscopic ablation model of the composite material to obtain a dataset of the solid volume fraction for each element, including: Define the solid volume fraction of each unit as: in, It is the volume of the solid in the unit cell. It is the unit volume; The oxygen diffusion equation and the new surface function after ablation, rewritten using the solid volume fraction, are as follows: in, It is the volume of the solid in the unit cell; It is the unit volume; Indicates a gas phase unit. Represents a solid-state unit; 0 < <1 represents a hybrid unit; It is the reaction concentration; It is time; It is the diffusion coefficient; It is the outward normal of the solid phase interface; k It is a heterogeneous reaction rate; It is a gradient operator; It is the absolute value of the screen back speed; The finite volume method was used to solve the oxygen diffusion equation and the new surface function after ablation, and a dataset of solid volume fractions was obtained.

6. The method according to claim 1, characterized in that, The initial conditions, boundary conditions, and erosion criteria for determining the microscopic burn-erosion model of composite materials include: The initial conditions for defining the computational domain include the temperature field, gas concentration, and material morphology; the temperature field is defined as the initial temperature of the composite material and the environment. The gas concentration is set to the ambient gas concentration. The material morphology is defined as the initial surface shape and microstructure distribution of the material. ; The boundary conditions for the computational domain include a top boundary and lateral boundaries; the top boundary uses Dirichlet conditions. ,in (This represents the gas concentration in the initial state); the lateral boundary is symmetrical, meaning the normal gradient between temperature and gas concentration is 0. The erosion criterion is a mechanical erosion criterion, specifically: in, It is the first in the composite material model The height of the floor, ; It is the maximum height of the surface yarn in the composite material model; It is the average height of the surface matrix in the composite material model; It is the mechanical erosion coefficient related to materials and the environment; It is the width of the weft yarn in the composite material model; It refers to the number of warp yarns covering the weft yarns; The length of the weft yarn; It is the first Height of the weft yarns; Its mechanical erosion at the microscale is as follows: in, It is the average height of the surface substrate; It is the height of the surface yarn in the composite material model; It is the amount of mechanical erosion.

7. The method according to claim 1 or 6, characterized in that, The construction of the physical information neural network model based on the evolution of the microscopic burn-eradication surface morphology of composite materials includes: Based on the determined initial conditions, boundary conditions, and erosion criteria of the composite material microstructure burn-erosion model, residual points, boundary constraint points, and initial constraint points for training are selected within the computational domain. Construct a custom multilayer perceptron neural network structure; A neural network is constructed based on a custom improved multilayer perceptron structure, and training points are input for training to establish a physical information neural network model. The loss function is obtained from the physical information neural network model, and the model is trained to obtain the final physical information neural network model based on the evolution of the microscopic burn-eradication surface morphology of composite materials.

8. The method according to claim 7, characterized in that, The selection of residual points, boundary constraint points, and initial constraint points for training within the computational domain includes: Latin hypercube random sampling is used to select residual points in the computational domain of composite materials; Set boundary constraint points at the top boundary of the composite material, and set boundary constraint points at the lateral boundary using symmetry conditions; Initial constraint points for temperature and gas concentration are set on the surface and inside the composite material.

9. The method according to claim 8, characterized in that, The custom multilayer perceptron neural network structure is constructed, which includes: an input layer, an encoder layer, a fusion layer, a hidden layer, an automatic micro-layer, and an output layer. The encoder layer includes an ablation encoder and an ablation encoder; the ablation encoder is used to perform feature encoding processing on the input data of the composite material ablation process and output an ablation process feature vector; the ablation encoder is used to perform feature encoding processing on the input data of the composite material ablation process and output an ablation process feature vector. The input data for the composite material ablation process includes the surface temperature and gas concentration during the composite material ablation process; The input data for the composite material ablation process includes airflow velocity and surface morphology during the ablation process.

10. The method according to claim 9, characterized in that, The loss function obtained based on the physical information neural network model includes: Physical equation constraints are introduced into the physical information neural network model. The physical constraint loss, boundary condition loss, initial condition loss and mechanical erosion condition loss are calculated using automatic differentiation technology. The final loss function is obtained by weighted summation.

11. The method according to claim 10, characterized in that: The physical constraint loss, boundary condition loss, initial condition loss, and mechanical erosion condition loss take the following forms: The physical constraint loss, i.e., the loss function of the gas diffusion equation, is: in, It is the reaction concentration; It is time; It is the diffusion coefficient; Indicates a gas phase unit. Represents a solid-state unit; 0 < <1 represents a hybrid unit; It is a gradient operator; It is the physical constraint loss function; The loss function for the boundary conditions and initial conditions is: in, It is the reaction concentration; It represents the gas concentration in the initial state; It refers to the concentration of ambient gases; These are weighting coefficients used to balance the relative importance of different losses; It is the temperature of the composite material; It is the initial temperature of the environment; It is the boundary condition loss function; It is the initial condition loss function; The loss function for the mechanical erosion condition is: in, It is the average height of the surface substrate; It is the height of the surface yarn in the composite material model; It is the amount of mechanical erosion; It is the mechanical erosion constraint loss function; The final loss function is a weighted sum of the physical constraint loss, boundary condition loss, initial condition loss, and mechanical erosion constraint loss: ； in, These are weighting coefficients used to balance the relative importance of different sub-loss functions; It is the final loss function.

12. A device for predicting the microscopic burn-erosion morphology of composite materials, characterized in that, include: The data acquisition module is used to acquire a dataset; wherein, the dataset includes the micro-ablation morphology of the carbon-carbon composite surface obtained by processing the micro-ablation model of the constructed carbon-carbon composite material based on the fixed volume fraction method. The model condition verification module is used to determine the initial conditions, boundary conditions, and erosion criteria of the composite material microstructure burn-erosion model. The model building module is used to construct a physical information neural network model based on the evolution of the microscopic sintered-eroded surface morphology of composite materials. The model training module is used to input the dataset into the physical information neural network model for iterative training to obtain the trained physical information neural network model. The rapid prediction module for the evolution of microscopic sintering-erosion morphology is used to input the composite material to be tested into the trained physical information neural network model to obtain the prediction results of the microscopic sintering-erosion morphology evolution of the composite material.

13. An electronic device, characterized in that, include: A processor, and a memory communicatively connected to the processor; The memory stores computer-executed instructions; The processor executes computer execution instructions stored in the memory to implement the method as claimed in any one of claims 1 to 10.

14. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the method as described in any one of claims 1 to 10.

Images