Method and apparatus for predicting thermal conductivity of composites

Abstract

The application is suitable for the cross technical field of material science and computer science, and provides a composite material thermal conductivity prediction method and device. The method comprises the following steps: obtaining data structure features of a microstructure of a composite material to be predicted; inputting the data structure features into a pre-trained machine learning prediction model, and outputting a predicted macroscopic physical performance value by the machine learning prediction model; wherein the machine learning prediction model is obtained by training by using a structure-performance database, and the structure-performance database comprises a plurality of data pairs, each data pair comprising one data structure feature representing a microstructure of a composite material, and one pre-determined macroscopic physical performance value corresponding to the data structure feature. Therefore, when the performance of the composite material is predicted, the calculation efficiency and the prediction accuracy can be considered.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of material science and computer science, and particularly relates to a composite material thermal conductivity prediction method and device. BACKGROUND

[0002] The macroscopic physical properties, such as thermal conductivity, electrical conductivity and mechanical strength, of a composite material, for example, a nickel-yttria-stabilized zirconia anode for a solid oxide fuel cell, are determined by its complex internal microstructure. Accurate evaluation of these properties is crucial for the design of new materials, performance optimization and service reliability evaluation.

[0003] Currently, the technical means for evaluating the physical properties of composite materials mainly include experimental measurement method and numerical simulation method. Although the experimental measurement method has direct results, it is difficult to perform in-situ and accurate measurement under real working conditions for materials that need to work under extreme conditions such as high temperature and chemical reaction, and the experimental cost is high and the experimental period is long. Although the numerical simulation method, such as the multi-scale simulation method combining molecular dynamics and finite element analysis, can provide high calculation accuracy, the calculation process is extremely complex, which requires huge calculation resources and takes a very long time, for example, it may take several days or even weeks to predict the thermal conductivity of a single sample. This low efficiency seriously restricts the speed of new material research and development and structural optimization iteration.

[0004] Therefore, there is an urgent need in the prior art for a new composite material performance prediction method that can balance calculation efficiency and prediction accuracy. SUMMARY

[0005] The embodiments of the application provide a composite material thermal conductivity prediction method and device, which can balance calculation efficiency and prediction accuracy.

[0006] In a first aspect, the embodiments of the application provide a composite material thermal conductivity prediction method, comprising:

[0007] obtaining a data structure feature of a microstructure of a composite material to be predicted;

[0008] inputting the data structure feature into a pre-trained machine learning prediction model, and outputting a predicted macroscopic physical property value by the machine learning prediction model;

[0009] The machine learning prediction model is obtained by training a structure-property database, and the structure-property database contains a plurality of data pairs, each data pair including a data structure feature representing a microstructure of a composite material and a pre-determined macroscopic physical property value corresponding to the data structure feature.

[0010] In a possible implementation manner of the first aspect, the construction process of the structure-property database comprises the following steps.

[0011] The composite material as a sample is scanned and three-dimensionally reconstructed to obtain data-based structure features;

[0012] Material parameters of each component of the composite material in a preset working environment are calculated through molecular dynamics simulation, wherein the material parameters comprise interfacial thermal resistance between the components;

[0013] Finite element method heat conduction simulation is performed on the data-based structure features based on the material parameters to obtain corresponding macroscopic physical property values;

[0014] The data-based structure features are paired with the corresponding macroscopic physical property values to form the structure-property database.

[0015] In a possible implementation manner of the first aspect, the composite material as a sample is scanned and three-dimensionally reconstructed to obtain data-based structure features, and the method comprises the following steps.

[0016] The composite material is scanned and three-dimensionally reconstructed using a focused ion beam scanning electron microscope imaging technology to generate a digital three-dimensional model containing spatial distribution of each phase;

[0017] The digital three-dimensional model is converted into the data-based structure features.

[0018] In a possible implementation manner of the first aspect, the loss function of the machine learning prediction model comprises a data fitting term and a physical law constraint term;

[0019] The data fitting term is used to measure the difference between the macroscopic physical property values predicted by the machine learning prediction model and the macroscopic physical property values in the structure-property database;

[0020] The physical law constraint term is used to promote the machine learning prediction model to generate a physical field distribution that satisfies a preset physical control equation when predicting according to the data-based structure features.

[0021] In a possible implementation manner of the first aspect, the composite material is a nickel-yttria-stabilized zirconia anode for a solid oxide fuel cell, the macroscopic physical property value is thermal conductivity, and the physical control equation is a steady-state heat conduction equation.

[0022] In a possible implementation manner of the first aspect, the finite element method heat conduction simulation is performed on the data-based structure features based on the material parameters to obtain corresponding macroscopic physical property values, and the method comprises the following steps.

[0023] The material parameters are assigned to a phase region corresponding to the data-based structure feature, a steady-state heat conduction equation is solved based on a finite element method, and a corresponding macroscopic physical property value is obtained.

[0024] In a possible implementation of the first aspect, the physical field distribution includes a temperature field distribution.

[0025] The steady-state heat conduction equation is solved based on a finite element method, and a corresponding macroscopic physical property value is obtained, including:

[0026] A constant heat flux density boundary condition is applied to a heat flow input surface of the digital three-dimensional model, and a fixed temperature boundary condition is applied to a heat dissipation surface of the digital three-dimensional model.

[0027] The steady-state heat conduction equation is iteratively solved until the temperature field distribution reaches a steady state, and a temperature gradient is calculated according to the steady-state temperature field distribution.

[0028] According to the preset input heat flux density and the temperature gradient, the thermal conductivity is calculated.

[0029] In a second aspect, an embodiment of the present application provides a composite material thermal conductivity prediction device, including:

[0030] An acquisition module is configured to acquire a data-based structure feature of a microstructure of a composite material to be predicted.

[0031] A prediction output module is configured to input the data-based structure feature into a pre-trained machine learning prediction model, and output a predicted macroscopic physical property value from the machine learning prediction model.

[0032] The machine learning prediction model is obtained by training a structure-property database, and the structure-property database includes a plurality of data pairs, each data pair including a data-based structure feature representing a microstructure of a composite material and a pre-determined macroscopic physical property value corresponding to the data-based structure feature.

[0033] In a third aspect, an embodiment of the present application provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the composite material thermal conductivity prediction method of any one of the first aspect when executing the computer program.

[0034] In a fourth aspect, an embodiment of the present application provides a computer readable storage medium, and the computer readable storage medium stores a computer program, and the computer program is executed by a processor to implement the composite material thermal conductivity prediction method of any one of the first aspect.

[0035] In a fifth aspect, an embodiment of the present application provides a computer program product, which, when running on a computer device, causes the computer device to perform the composite material thermal conductivity prediction method in any one of the first aspect.

[0036] Compared with the prior art, the embodiment of the present application has the beneficial effects that:

[0037] Based on the technical solution of micro data structure feature-machine learning prediction of the present application, only the data structure feature of the micro structure of the composite material needs to be obtained, and the trained machine learning prediction model is input, so that the macro physical property value can be quickly and accurately output. The complex analysis process of modeling and calculation of the composite material is skipped, time-consuming and high-cost finite element simulation is avoided for each new structure, the calculation cost is greatly reduced, and the evaluation efficiency is significantly improved.

[0038] It can be understood that the beneficial effects of the above-mentioned second aspect to the fifth aspect can be referred to the related description in the first aspect, which will not be repeated here. BRIEF DESCRIPTION OF DRAWINGS

[0039] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the embodiments or the prior art description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0040] Figure 1 is an offline training phase flowchart provided by an embodiment of the present application for generating a structure-property database;

[0041] Figure 2 is an application flowchart of the composite material thermal conductivity prediction method provided by an embodiment of the present application;

[0042] Figure 3 is a technical principle schematic diagram of the composite material thermal conductivity prediction method provided by an embodiment of the present application;

[0043] Figure 4 is a machine learning prediction model principle and data processing flowchart of the thermal conductivity prediction provided by an embodiment of the present application;

[0044] Figure 5 is a structural schematic diagram of the composite material thermal conductivity prediction device provided by an embodiment of the present application;

[0045] Figure 6 is a structural schematic diagram of the computer device provided by an embodiment of the present application. DETAILED DESCRIPTION

[0046] In the following description, for purposes of explanation and not limitation, specific details are set forth such as particular architectures, techniques, etc. in order to provide a thorough understanding of the embodiments of the present application. However, it will be apparent to those skilled in the art that the present application can be practiced in other embodiments that depart from these specific details. In other instances, detailed descriptions of well-known methods, devices, circuits, and

[0047] It is to be understood that the terminology "includes", "has", "holds", "contains" and / or "comprising", "including", "having" and their conjugates, as used herein, means "including but not limited to", and not to the exclusion of any other term or aspect.

[0048] It is also to be understood that the terminology "and / or" as used herein refers to and encompasses any and all possible combinations of one or more of the associated listed items, and that the term "at least one of A and B" is equivalent to "A, B, or A and B".

[0049] As used in the description of the application and the appended claims, the term "if' can be interpreted to mean "when" or "upon" or "in response to determining" or "in response to detecting" depending on the context. Similarly, the phrase "if it is determined" or "if [a described condition or event] is detected" can be interpreted to mean "upon determining" or "in response to determining" or "upon [the described condition or event] being detected" or "in response to [the described condition or event] being detected", depending on the context.

[0050] In addition, the terms "first", "second", "third", etc. as used in the description of the application and the appended claims are not used to denote or imply relative importance but are used to distinguish one element from another.

[0051] Reference throughout this specification to "one embodiment" or "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment of the application. Thus, the appearances of the phrases "in one embodiment" or "in an embodiment" or "in other embodiments" or "in still other embodiments" in various places throughout this specification are not necessarily all referring to the same embodiment, unless otherwise specified. Furthermore, the terms "comprises", "comprising", "includes", "including", "has", "having", and the like are intended to be open-ended terms that do not exclude additional, unrecited elements or methods. Thus, such terms are used in the specification to indicate that additional, unrecited elements or methods are optional and do not alter the basic understanding of the application.

[0052] The embodiment of the present application provides a complete implementation process of a composite material thermal conductivity prediction method. Specifically, taking the nickel-yttria stabilized zirconia porous anode, a key component in a solid oxide fuel cell, as an application object, how to construct and apply a high-precision machine learning prediction model to realize fast and accurate prediction of the macroscopic physical performance value (thermal conductivity) of the nickel-yttria stabilized zirconia porous anode in a real working environment is described in detail. The composite material in the technical solution can be a nickel-yttria stabilized zirconia anode for a solid oxide fuel cell, and the macroscopic physical performance value is thermal conductivity. The overall process of the technical solution can be divided into two main stages: one is an offline training stage, and the core task of the offline training stage is to construct a high-fidelity structure-performance database that can accurately reflect the physical reality, and to train a machine learning prediction model with data fitting ability and physical consistency by using the database; the other is an online application stage, and the core task of the online application stage is to use the machine learning prediction model obtained in the first stage to quickly predict the thermal conductivity of any new composite material microstructure provided.

[0053] Referring to Figure 1 , which shows an offline training stage process for generating a structure-performance database in the embodiment of the present application, and the process constitutes the basis for ensuring the accuracy of the final machine learning prediction model.

[0054] First, the accurate three-dimensional microstructure of the composite material needs to be obtained. In the embodiment, as shown in Figure 1 , 101 scans and three-dimensionally reconstructs the composite material as a sample to obtain data structure features. Specifically, a focused ion beam scanning electron microscope imaging technology is used to scan and three-dimensionally reconstruct the composite material to generate a digital three-dimensional model containing the spatial distribution of each phase, and the digital three-dimensional model is converted into the data structure features.

[0055] It can be understood that the technology combines the precise cutting ability of the focused ion beam with the high-resolution imaging ability of the scanning electron microscope. In operation, a prepared composite material (such as a nickel-yttria-stabilized zirconia) sample is first placed in the vacuum chamber of the microscope, and then the surface of the sample is cut layer by layer by a focused gallium ion beam. After cutting a very thin layer (such as 10-20 nanometers thick), the scanning electron microscope images the newly exposed cross section. Through continuous cutting-imaging cycles, a series of two-dimensional cross-sectional images can be obtained. Subsequently, through special image processing software, the two-dimensional cross-sectional images are aligned, segmented, and three-dimensionally reconstructed, and finally a plurality of digital three-dimensional models 102 of the three-dimensional microstructure capable of accurately reflecting the spatial distribution, morphology, and connectivity of the pores, nickel phase, and yttria-stabilized zirconia phase of the material are generated. The digital three-dimensional model 102 is then converted into the data-based structural features, which are stored in the form of a three-dimensional voxel matrix, and each voxel is assigned a label to identify the phase it belongs to (such as nickel, yttria-stabilized zirconia, or pores). The reconstructed digital three-dimensional model 102 is converted into data-based structural features that can be processed by the machine learning prediction model. In this embodiment, a convolutional neural network or the like can be selected as the machine learning prediction model, so the microstructure needs to be converted into a three-dimensional data matrix. Specifically, the digital three-dimensional model 102 can be discretized into a three-dimensional grid of a predetermined size (such as 256x256x256 voxels), and for each voxel in the grid, a specific integer value is assigned according to the phase it belongs to, for example, if the voxel is in the pore phase, it is assigned a value of 0; if it is in the nickel phase, it is assigned a value of 1; if it is in the yttria-stabilized zirconia phase, it is assigned a value of 2. In this way, the complex microstructure is converted into a three-dimensional data matrix containing a large number of numerical values, which constitutes the data-based structural features described in this application.

[0056] At the same time, in order to make the subsequent macroscopic performance simulation close to the real physical working condition, it is necessary to calculate the material parameters of each component of the composite material under the preset working environment.

[0057] This embodiment introduces the molecular dynamics simulation shown in 103. Through molecular dynamics simulation, the material parameters 104 of each component of the composite material under the preset working environment are calculated; wherein the material parameters include the interfacial thermal resistance 105 between the components.

[0058] Specifically, since the thermal conductivity depends on the intrinsic thermal conductivity of each component of the material and the heat transfer ability of the interface between them, it is necessary to first accurately calculate these parameters at the atomic scale. In this embodiment, the composite material studied is nickel-yttria-stabilized zirconia, and the working temperature is usually between 800 degrees Celsius and 1000 degrees Celsius. The molecular dynamics method can be used to simulate pure nickel crystals and pure yttria-stabilized zirconia crystals respectively at the set working temperature, and their intrinsic thermal conductivities and Meanwhile, an atomic model of the interface between nickel and yttria-stabilized zirconia is constructed, and the interface thermal resistance (also referred to as thermal resistance) at the interface between nickel and yttria-stabilized zirconia is calculated by simulating the heat flow and temperature drop on both sides of the interface.

[0059] The interface thermal resistance 105 is a key micro parameter affecting the macro thermal conductivity of the composite material, but it is difficult to obtain accurate parameters in traditional finite element simulation. The parameters obtained by the molecular dynamics simulation in the present application constitute high-fidelity material parameters 104.

[0060] Subsequently, the results of the aforementioned two parallel steps are combined to perform a finite element method heat conduction simulation in 106, and the corresponding macro physical property values 107 are obtained by performing a finite element method heat conduction simulation on the data-based structure features based on the material parameters. Specifically, the material parameters are assigned to the phase regions corresponding to the data-based structure features, and the corresponding macro physical property values are obtained by solving the steady-state heat conduction equation based on the finite element method. Further specifically, the physical field distribution includes temperature field distribution; a constant heat flux density boundary condition is applied to the heat input surface of the digital three-dimensional model, and a fixed temperature boundary condition is applied to the heat dissipation surface of the digital three-dimensional model; the steady-state heat conduction equation is iteratively solved until the temperature field distribution reaches a steady state, and the temperature gradient is calculated according to the steady-state temperature field distribution; and the thermal conductivity is calculated according to the preset input heat flux density and the temperature gradient.

[0061] In the embodiments of the present application, for each digital three-dimensional model 102 obtained by 101, it is first imported into a finite element analysis software and meshed, and then the material parameters 104 including the interface thermal resistance 105 calculated by 103 are assigned to the corresponding phase regions (nickel phase, yttria-stabilized zirconia phase) and interface regions in the digital three-dimensional model 102. Next, a certain temperature difference is applied to the two opposite surfaces of the digital three-dimensional model 102 as a fixed temperature boundary condition (for example, one surface is set to 1073K and the opposite surface is set to 1063K), while the other sides are set as adiabatic boundary conditions, and a constant heat flux density boundary condition is applied to the heat input surface of the digital three-dimensional model. Then, by solving the steady-state heat conduction equation, the temperature field distribution inside the entire digital three-dimensional model 102 and the total heat flux through the digital three-dimensional model 102 can be calculated. According to the Fourier law, the macro physical property values 107 corresponding to the digital three-dimensional model 102 of the three-dimensional microstructure can be calculated using the applied temperature gradient and the calculated total heat flux. By repeating this process for all digital three-dimensional models 102 of the composite material, a series of structure-property data can be obtained.

[0062] Each digital three-dimensional model 102 is converted into a dataized structure feature, and the dataized structure feature is paired with a corresponding macroscopic physical property value 107 calculated by finite element heat conduction simulation 106 (as an output label) to form a data pair, and a collection of all such data pairs forms a structure-property database 108. The high-fidelity characteristics of the database are reflected in that both the input (the input has a real microstructure) and the output (the output is a simulation result based on real working condition parameters) are closest to the physical reality.

[0063] In an optional embodiment, the loss function of the machine learning prediction model includes a data fitting term and a physical law constraint term; the data fitting term is used to measure the difference between the macroscopic physical property value predicted by the machine learning prediction model and the macroscopic physical property value in the structure-property database; the physical law constraint term is used to promote the machine learning prediction model to generate a physical field distribution that satisfies a preset physical control equation when predicting according to the dataized structure feature, and the physical control equation is a steady-state heat conduction equation.

[0064] In an embodiment of the present application, considering that the input data is a dataized structure feature, a three-dimensional convolutional neural network is preferably used as the feature extraction front end of the model. The network can be stacked by multiple three-dimensional convolutional layers, activation function layers (such as linear rectifier units) and pooling layers, and can automatically learn and extract spatial structure features related to thermal conductivity from the input dataized structure feature, such as volume fraction, curvature, connectivity, etc. of each phase. At the end of the three-dimensional convolutional neural network, several fully connected layers are connected to map the extracted high-dimensional features to a single scalar output, i.e. the predicted macroscopic physical property value.

[0065] Subsequently, the core training process is entered, which aims to optimize the internal parameters (such as the weights and biases of the convolutional kernels) of the machine learning prediction model so that it can accurately make predictions. The training process uses the high-fidelity structure-property database as a data source. It should be noted that, unlike traditional neural network training, the present embodiment uses a special composite loss function to guide the training, which includes a data fitting term and a physical law constraint term.

[0066] The data fitting term is used to measure the difference between the predicted value of the model and the true value in the database. Specifically, the mean square error can be used to calculate it, and the mathematical expression is: wherein is the number of samples in the training batch, is the macroscopic physical property value predicted by the model for the th microstructure, and are the corresponding macroscopic physical property values in the database obtained by finite element method heat conduction simulation 107. The goal of minimizing the data fitting term is to make the prediction results of the model fit the high-fidelity data as much as possible.

[0067] The physical law constraint term is the key difference between the embodiments of the present application and pure data-driven models. The physical law constraint term is used to measure the degree of conformity of the physical field distribution generated internally by the machine learning prediction model in the prediction process with the preset physical control equation, so that the physical field distribution generated by the machine learning prediction model in the prediction process approximates the solution of the preset physical control equation. This loss term aims to incorporate the physical law of heat conduction as a soft constraint into the training process of the model. In the embodiments, the physical law constraint term can be constructed based on the steady-state heat conduction equation, i.e. , where is the temperature at the spatial position , and is the local thermal conductivity of the point. To calculate the loss, a large number of configuration points can be randomly sampled within each three-dimensional microstructure domain. For each configuration point, the model not only needs to predict the macroscopic thermal conductivity, but also the internal neuron activation state can be constructed to represent the local temperature field . Through automatic differentiation technology, the residual of the equation can be calculated. The physical law constraint term is the mean square value of the residuals at all configuration points: , where is the number of configuration points. The goal of minimizing the physical law constraint term is to force the prediction behavior of the model (i.e., the internally generated physical field distribution, which can also be referred to as the temperature field distribution) to comply with the basic physical law of heat conduction, even in regions without data labels.

[0068] Therefore, the total loss function is the weighted sum of the two terms: , where and are hyperparameters used to balance the importance of data fitting and physical constraints. During the training process, the model parameters are continuously adjusted to minimize by optimization algorithms such as gradient descent. When the loss function converges or reaches the preset training epoch, the training process is complete, and the machine learning prediction model described above is obtained. This model has higher prediction accuracy, stronger generalization ability, and physical reliability because it has learned both data patterns and physical laws.

[0069] For the construction of the physical law constraint term in the loss function, the above embodiments describe a scheme based on the steady-state heat conduction equation. As an alternative implementation, this embodiment will describe how to construct the physical law constraint term based on the transient heat conduction equation to deal with more complex physical scenarios. The complete form of the transient heat conduction equation is: , where is the density, is the specific heat capacity, is the time, is the internal heat source. In the absence of internal heat source, the equation simplifies to To utilize this equation, the output of the machine learning prediction model needs to be designed to be able to predict the temperature field i.e. the temperature is a function of not only space, but also time, accordingly, the input of the model also needs to increase the time dimension The way to calculate the physical loss term is to calculate the mean square value of the residual of the transient equation on the configuration points sampled in the space-time domain: The model trained in this way can not only predict the effective thermal conductivity in the steady state, but also predict the dynamic thermal response behavior of the material when it is subjected to thermal shock or periodic thermal load, which is of great significance for evaluating the thermal shock resistance and other properties of the material. Different choices of the heat conduction physical law (steady state or transient state) enable the technical solution of the present application to cope with more diversified and complex engineering problems.

[0070] In the embodiments of the present application, the physical law constraint term is introduced into the loss function of model training, and the prior knowledge of physics is integrated into the model training process. This is equivalent to providing the model with additional supervision information other than training data, forcing the prediction behavior of the model to comply with the physical law, thereby effectively preventing the model from overfitting when the training data is limited, enhancing its generalization ability to new structures, and ensuring that the prediction results are physically reasonable and self-consistent, with higher reliability. Improve the accuracy, generalization ability and physical reliability of the prediction model.

[0071] Next, turning to the online application stage, referring to Figure 2 , a flowchart of the application of a composite material thermal conductivity prediction method provided by the present application is shown.

[0072] S201, obtaining the data structure features of the microstructure of the composite material to be predicted.

[0073] In the embodiments of the present application, when the material research and development personnel or engineers want to quickly evaluate the performance of a new un- simulated nickel-yttria stabilized zirconia anode design, the composite material can be scanned and three-dimensionally reconstructed using focused ion beam scanning electron microscope imaging technology to generate a digital three-dimensional model containing the spatial distribution of each phase. After obtaining the digital three-dimensional model, it needs to be preprocessed to convert it into a standardized three-dimensional data matrix consistent with the data format used during training, i.e. data structure features. The purpose of this preprocessing step is to ensure that the input data can be correctly received and processed by the model.

[0074] S202, the digitized structural features are input into a pre-trained machine learning prediction model, which then outputs a predicted macroscopic physical performance value. The machine learning prediction model is trained using a structure-performance database, which contains multiple data pairs. Each data pair includes a digitized structural feature characterizing the microstructure of the composite material and a pre-determined macroscopic physical performance value corresponding to that digitized structural feature.

[0075] In this embodiment, the processed data structure features from the preceding steps are used as input and fed into the deployed machine learning prediction model. Upon receiving the input data, the machine learning prediction model directly outputs a scalar value—the macroscopic physical performance value—through a single, efficient forward propagation calculation (i.e., the process of data flowing from the input layer to the output layer in the network). The entire prediction process is extremely short, typically within seconds or milliseconds, a stark contrast to the hours or even days required for finite element method heat conduction simulations. This allows researchers to immediately obtain predicted values ​​for rapid evaluation of design options, thereby significantly accelerating the research and development iteration process.

[0076] For ease of understanding of the scheme in this application, the following is combined with Figure 3 The technical principle of a method for predicting the thermal conductivity of composite materials is illustrated in the schematic diagram shown.

[0077] Non-equilibrium molecular dynamics (NEMD) is a type of molecular dynamics simulation used to simulate the atomic / molecular motion and transport properties of systems in non-equilibrium states (such as those with heat flux, particle flux, etc.). For convenience, non-equilibrium molecular dynamics will be referred to as NEMD in the following text.

[0078] Ni

[100] / YSZ

[100] , Ni

[100] / YSZ

[110] , Ni

[100] / YSZ

[111] : These represent the crystal orientation matching relationship between the Ni (nickel) and YSZ (Yttria-Stabilized Zirconia) interfaces.

[100] ,

[110] , and

[111] are crystallographic orientation indices, and / represents parallel crystal orientations on both sides. Different crystal orientation matching will affect the atomic arrangement and thermal transport properties of the interface.

[0079] O (red), Zr (blue), Y (yellow), Ni (gray): These are symbols corresponding to oxygen atoms, zirconium atoms, yttrium atoms, and nickel atoms, respectively, used to identify the atomic composition of the YSZ phase (containing Zr, Y, and O) and the Ni phase (containing Ni).

[0080] Heat sink and heat source: The heat absorption end and release end set in the simulation, which generate non-equilibrium heat flow (red arrow direction in the figure) to form a temperature gradient in the system, and are used to calculate parameters such as thermal conductivity and interface thermal resistance.

[0081] FIB-SEM: Focused Ion Beam-Scanning Electron Microscopy, FIB (for convenience, described as FIB hereinafter) can mill and etch the sample in micro-nano, and SEM (for convenience, described as SEM hereinafter) is used to collect high-resolution images of the sample surface. The combination of the two realizes the reconstruction of the three-dimensional microstructure of the material.

[0082] Reference mark: Marking points on the sample for image registration, ensuring that the images of each layer can be accurately aligned when FIB mills layer by layer and SEM images layer by layer, and ensuring the accuracy of three-dimensional reconstruction.

[0083] Sample: The actual Ni-YSZ composite material sample is the object of three-dimensional reconstruction.

[0084] X1, X2, X3: Represent the spatial coordinate axis direction, which helps to describe the spatial position of sample milling and imaging, and ensures that three-dimensional reconstruction is carried out in the correct spatial coordinate system.

[0085] Three-dimensional reconstruction: The process of restoring the three-dimensional microstructure of the sample by FIB milling the sample layer by layer, SEM collecting two-dimensional images of each layer, and then image processing (registration, splicing, voxelization, etc.), and outputting a three-dimensional model containing phase distribution, pore, etc.

[0086] FEM: Finite Element Method, an engineering simulation method for numerically solving physical fields (such as thermal field and force field). The continuum is discretized into a finite number of elements, and the overall physical field distribution is obtained by solving the element equations. Here, based on the three-dimensional microstructure reconstructed by FIB-SEM, combined with the parameters obtained by NEMD (such as interface thermal resistance), the heat conduction process is simulated.

[0087] Thermal conductivity: A physical quantity representing the ability of a material to conduct heat (unit commonly W / (m・K)). In the figure, the color gradient (300K to 800K) and the three-dimensional model at different times (0μs, 4μs, 10μs, 100μs) show the evolution of the temperature field over time when heat is transmitted in the Ni-YSZ composite material, reflecting the influence of thermal conductivity on the heat transfer process - the higher the thermal conductivity, the faster the temperature diffusion and the smaller the temperature gradient.

[0088] The overall technical principle flow is as follows:

[0089] 1) NEMD part, for Ni-YSZ composite structure, build interface atomic model with different crystal orientation matching (such as [Ni100] / [YSZ100] and so on). Through NEMD simulation, heat source and heat sink are set at both ends of the model to create non-equilibrium heat flow, track atomic heat movement (vibration, collision) and energy transfer path, calculate the thermal conductivity of Ni phase and YSZ phase itself (calculation method see above), and calculate the thermal transfer resistance at the interface of different crystal orientation matching caused by the discontinuity of atomic arrangement (i.e. interface thermal resistance).

[0090] 2) FIB-SEM part, for actual Ni-YSZ composite material sample, use FIB-SEM to perform layer-by-layer milling and layer-by-layer imaging, complete three-dimensional reconstruction, obtain a three-dimensional digital model consistent with the real material microstructure, as the input of macroscopic heat simulation geometry + phase distribution.

[0091] 3) FEM part, assign the parameters obtained by NEMD to the corresponding phase regions (Ni phase, YSZ phase, phase interface) of the three-dimensional digital model reconstructed by FIB-SEM, solve the steady-state heat conduction equation using finite element method FEM, simulate the heat transfer process in the macroscopic structure of the composite material, calculate the temperature field distribution at different times (0 μs, 4 μs, etc.) and different positions, and obtain the influence law of thermal conductivity on temperature diffusion.

[0092] 4) Color gradient (300K to 800K) in thermal conductivity shows temperature field distribution, three-dimensional model at different time points (0 μs to 100 μs), presents the dynamic process of heat conduction—from initial temperature unevenness (0 μs may be close to the initial set temperature), to heat gradually diffusing, temperature gradient decreasing (such as 4 μs begins to diffuse obviously, 100 μs tends to be stable). By comparing the simulation results with experimental tests (such as laser flash method for measuring thermal conductivity), the accuracy of the model (NEMD+FIB-SEM+FEM) is verified, and finally the cross-scale analysis from micro atomic motion to macro heat conduction performance is realized.

[0093] In summary, this is a cross-scale thermal performance research scheme of micro mechanism (NEMD) → real structure (FIB-SEM) → macro performance (FEM), which uses NEMD to understand the atomic level heat transport law, uses FIB-SEM to restore the real three-dimensional morphology of the material, and then uses FEM to magnify the micro law to the macro, finally accurately analyzes and predicts the thermal conductivity and heat transfer characteristics of Ni-YSZ composite material, and provides multi-scale support for optimizing material design (such as adjusting the interface crystal orientation and controlling the microstructure).

[0094] In order to facilitate the understanding of the scheme of the present application, the principle of machine learning prediction model for predicting thermal conductivity and data processing flowchart shown in Figure 4

[0095] ​Three-dimensional feature extraction: From the microstructure (purple-blue three-dimensional microstructure in the figure), the key parameters that affect the thermal conductivity are screened out and converted into data-based structural features that the model can learn.

[0096] Thermal conductivity: the target performance to be predicted by the model, as the label value for model training (the color gradient model in the figure represents the thermal conductivity distribution at different positions).

[0097] Database: a collection of data-based structural features + macroscopic physical performance value (such as thermal conductivity) data pairs, which are functionally divided into:

[0098] Test set: after training, an independent data set used to test the generalization ability of the model and evaluate the prediction accuracy.

[0099] Validation set: used to monitor model overfitting during training and adjust hyperparameters (such as network depth, learning rate) to ensure that the model is effective for unknown data.

[0100] Training set: the core data for the model to learn the feature-thermal conductivity mapping rule, with network parameters optimized through backpropagation.

[0101] For the physical information neural network (i.e., the machine learning prediction model mentioned above), the blue node network in the figure is the feature extraction + mapping layer, which receives data-based structural features (x, y, z, t inputs, red circle nodes) and outputs predicted thermal conductivity (green circle nodes, such as u, v, w, p multi-task outputs, or single task).

[0102] MSE represents the mean square error, which is the data fitting term in the above, an indicator that measures the difference between the predicted thermal conductivity and the true thermal conductivity in the database. PDE represents the partial differential equation, which is the physical law constraint term in the above, which calculates the residual between the model prediction and the physical control equation through automatic differentiation, so that the prediction conforms to the basic laws of thermodynamics.

[0103] Loss LossD < threshold tol, which indicates whether the total loss of the model (MSE + PDE constraint loss) is less than the set threshold tol. If not No, continue to optimize the parameters through backpropagation; if Yes, training is terminated Done.

[0104] In the embodiments of the present application, the physical information neural network is based on the training set data, uses MSE loss to make the prediction conform to the true thermal conductivity, and at the same time, embeds the heat conduction PDE into the loss function to force the model to predict the temperature field / heat flow to meet the physical law (such as Fourier's law). By continuously iterating and optimizing the network parameters (such as weights and biases), the data fitting error and the physical equation residual are minimized until the total loss is less than a threshold. The trained model can quickly output the predicted thermal conductivity when inputting the data-based structural features of new composite materials (such as the data-based structural features of the untested Ni-YSZ microstructure), and then use the test set to verify the accuracy, and if it meets the standard, it can be used for material design (such as screening microstructures with low / high thermal conductivity).

[0105] The traditional pure data-driven model (such as a general neural network) is prone to inaccurate prediction due to data distribution deviation, while the physical information neural network uses PDE constraints to bind physical laws (such as heat cannot be generated / disappeared in the air), so that even with limited data, the prediction can conform to the nature of heat conduction, solving the problem of small data + high performance prediction, and it is particularly suitable for multi-scale material design (from micro-interface to macro-performance).

[0106] It can be understood that the core idea of the technical solution of the present application, i.e., constructing a structure-performance database and training a machine learning prediction model constrained by physical laws, has strong universality and scalability. Accordingly, based on the mechanical performance parameters and the electrical conductivity performance parameters, the data-based structural features are subjected to finite element method mechanical simulation and electric field simulation respectively to obtain corresponding macro mechanical performance values and electrical conductivity values; the data-based structural features and the corresponding mechanical performance values and electrical conductivity values jointly constitute a multi-physical field data pair, and the structure-performance database is expanded (i.e., the data pair constituted by the data-based structural features and the corresponding mechanical performance values, and the data pair constituted by the data-based structural features and the corresponding electrical conductivity values, are added to the structure-performance database) to be used for training the machine learning prediction model, so that the machine learning prediction model can predict the mechanical performance, electrical conductivity performance and other macro physical performance values of the composite material.

[0107] As an example, if the effective electrical conductivity of the composite material to be predicted needs to be predicted, only adaptive modification of the process in the above embodiments is required. In the data generation phase, molecular dynamics simulation will be used to calculate the intrinsic electrical conductivity of each component under a specific working condition, and finite element method simulation will no longer solve the steady-state heat conduction equation, but will solve the steady-state current continuity equation ( where is the electrical conductivity, is the electric potential). Thus, the data pairs stored in the structure-property database will be the data-ized structure features-effective conductivity pairs. During the model training stage, the physical law constraint term will also be constructed based on the current continuity equation. The final model will be able to quickly predict the effective conductivity according to the data-ized structure features of the composite microstructure to be predicted.

[0108] Similarly, if the mechanical properties of the material, such as the elastic modulus or Poisson's ratio, are to be predicted, the finite element method simulation will be replaced by a structural mechanics simulation for solving the elastic mechanics equilibrium equation. The data pairs stored in the database will be the data-ized structure features-elastic modulus pairs of the microstructure, and the physical law constraint term will be constructed based on the elastic mechanics constitutive relation and the equilibrium equation.

[0109] In summary, the methodology proposed in this application is universal and can include all macroscopic physical properties related to the microstructure of the composite material (whether thermal, electrical, mechanical properties, or fluid transport properties, etc.) in the prediction category. Only the calculation target of the molecular dynamics simulation (such as switching from thermal physical parameters to electrical, mechanical, or fluid-related parameters) and the control equation of the finite element method simulation (switching from the heat conduction equation to the current continuity equation, the elastic mechanics equilibrium equation, or the fluid mechanics equation) need to be adjusted, and the data pair type in the structure-property database and the physical law constraint term of the machine learning prediction model need to be updated simultaneously. The prediction capability can be quickly expanded without making revolutionary modifications to the core technical framework. This high degree of universality and scalability makes it widely adaptable to different material systems (such as carbon fiber reinforced resin matrix composites, ceramic matrix composites, etc.) and different performance prediction needs, providing a unified technical solution for the integrated rapid evaluation of the multi-physical properties of composites, and has broad application prospects in advanced material research and development, high-end equipment design, etc.

[0110] Moreover, the application can also not require targeted adjustment to achieve single performance prediction, but can simultaneously predict the thermal, electrical, mechanical, and fluid transport performance of composite materials and other macroscopic physical properties based on the same core technology framework. Specifically, in the structure-performance database construction phase, it is not necessary to generate single performance data pairs in batches, but through one-time multi-physical field coupling simulation, multi-dimensional performance data can be obtained simultaneously: first, through molecular dynamics simulation, the thermal, electrical, mechanical, and fluid-related parameters of each component are calculated simultaneously under the preset working environment (such as intrinsic thermal conductivity, interfacial thermal resistance, intrinsic electrical conductivity, interfacial resistance, elastic stiffness tensor, interfacial bonding strength, fluid-solid interfacial contact angle, and adsorption energy); then these multi-dimensional parameters are assigned to the digital three-dimensional model corresponding to the data-based structure feature, and through multi-physical field finite element simulation (such as thermal-electric-mechanical-fluid coupling simulation), the steady-state heat conduction equation, current continuity equation, elastic mechanics balance equation, and fluid mechanics control equation are solved simultaneously to obtain macroscopic thermal conductivity, electrical conductivity, elastic modulus, and permeability; subsequently, the data-based structure feature-multiple performance value set (including thermal, electrical, mechanical, and fluid performance values) is stored as a unified data pair in the structure-performance database, without the need to split the database or build a separate training set. In the machine learning prediction model training phase, it is also not necessary to train a model for a single performance, but to build a multi-task physical information neural network model: the model takes the data-based structure feature as the unified input, sets multiple parallel output heads, and corresponds to the prediction targets of thermal conductivity, electrical conductivity, elastic modulus, and permeability; at the same time, multiple sets of physical law constraint terms are embedded in the loss function, corresponding to the physical control equations of each performance (such as heat conduction equation residual, current continuity equation residual, elastic mechanics balance equation residual, and fluid mechanics equation residual), and the multi-task data fitting term (measuring the difference between each performance prediction value and the database true value) together form a composite loss function. During training, the model learns the common features of the microstructure through a shared feature extraction layer (such as a three-dimensional convolution layer), and then through the dedicated output head and the corresponding physical constraint term, it simultaneously masters the correlation between different performances and microstructures, and finally realizes the effect of simultaneously outputting multiple macroscopic physical performance prediction results by inputting the data-based structure feature once.The multi-performance synchronous prediction scheme can greatly improve the performance evaluation efficiency without adjusting the core technical framework (structure-performance database construction logic, physical constraint model training mechanism), and can reduce repeated microstructure scanning, simulation calculation and model training processes, and avoid resource waste. Meanwhile, since the multi-performance data comes from the same set of microstructure and coupling simulation, the internal correlation between different performances (such as the cooperative variation law of thermal conductivity and electrical conductivity) can be more truly reflected, so that the prediction result is more suitable for the multi-physical field coupling scene in the actual service of the material, and more comprehensive and efficient technical support is provided for the comprehensive performance optimization and multi-field coupling design of the composite material.

[0111] It should be understood that the size of the serial number of each step in the above embodiment does not mean the order of execution, and the execution order of each process should be determined according to its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0112] According to the composite material thermal conductivity prediction method described in the above embodiment, Figure 3 The structure block diagram of the composite material thermal conductivity prediction device provided by the embodiment of the present application is shown, and only the part related to the embodiment of the present application is shown for the convenience of description.

[0113] Referring to Figure 5 The composite material thermal conductivity prediction device comprises:

[0114] The acquisition module is configured to acquire data structure features of a microstructure of a composite material to be predicted.

[0115] The prediction output module is configured to input the data structure features into a pre-trained machine learning prediction model, and output a predicted macroscopic physical performance value from the machine learning prediction model.

[0116] The machine learning prediction model is obtained by training a structure-performance database, and the structure-performance database comprises a plurality of data pairs, each data pair comprising a data structure feature representing a microstructure of a composite material and a pre-determined macroscopic physical performance value corresponding to the data structure feature.

[0117] It should be noted that the information interaction, execution process and the like between the above modules are based on the same concept as the method embodiments of the present application, and the specific functions and technical effects brought by them can be referred to the method embodiments part, which will not be repeated here.

[0118] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the division of the above functional units and modules is exemplified, and in actual application, the above functions can be completed by different functional units and modules according to needs, that is, the internal structure of the apparatus is divided into different functional units or modules to complete all or part of the above described functions. Each functional unit or module in the embodiment can be integrated in one processing unit, or each unit can exist physically separately, or two or more units can be integrated in one unit, and the integrated unit can be realized in the form of hardware or in the form of software functional unit. In addition, the specific name of each functional unit or module is only for convenient distinction, and does not limit the protection scope of the present application. The specific working process of the units and modules in the system can refer to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0119] The embodiment of the present application further provides a computer device, which comprises at least one processor, a memory and a computer program stored in the memory and executable on the at least one processor, wherein the processor implements the steps in any of the method embodiments when executing the computer program.

[0120] The embodiment of the present application further provides a computer readable storage medium, which stores a computer program, wherein the computer program is executable by a processor to implement the steps in any of the method embodiments.

[0121] The embodiment of the present application provides a computer program product, which, when running on a computer device, enables the computer device to implement the steps in any of the method embodiments.

[0122] The integrated unit, if implemented in the form of a software function unit and sold or used as an independent product, can be stored in a computer readable storage medium. Based on such understanding, the present application can implement all or part of the processes in the above-mentioned embodiment methods through a computer program to instruct relevant hardware to complete, and the computer program can be stored in a computer readable storage medium. When the computer program is executed by a processor, the steps of each method embodiment described above can be implemented. The computer program includes computer program code, which can be in the form of source code, object code, executable files or some intermediate forms. The computer readable medium at least includes any entity or device capable of carrying the computer program code to the photographing device / terminal equipment, recording medium, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal and software distribution medium. For example, U disk, mobile hard disk, magnetic disk or optical disk, etc. In some jurisdictions, according to legislation and patent practice, the computer readable medium can not be an electrical carrier signal and a telecommunication signal.

[0123] In the above embodiments, the description of each embodiment has its own focus, and the parts not described or recorded in detail in a certain embodiment can be referred to the relevant description of other embodiments.

[0124] Those skilled in the art can appreciate that the units and algorithm steps of the examples described in combination with the embodiments disclosed herein can be realized by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized in hardware or software depends on the specific application and design constraints of the technical solution. The skilled person can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0125] In the embodiments provided in the present application, it should be understood that the disclosed apparatus / computer device and method can be implemented in other ways. For example, the apparatus / computer device embodiments described above are only schematic. The division of the modules or units is only a logical function division, and there can be another division in actual implementation. For example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the units shown or discussed can be indirect coupling or communication connection through some interface, device or unit, and can be electrical, mechanical or other forms.

[0126] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0127] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

[0128] Figure 6 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Figure 6 As shown, the computer device of this embodiment includes: at least one processor 20 ( Figure 6 (Only one is shown in the diagram), memory 21, and computer program 22 stored in said memory 21 and executable on said at least one processor 20, wherein said processor 20 executes said computer program 22 to implement the steps in any of the above embodiments of the composite material thermal conductivity prediction method.

[0129] The computer device may include, but is not limited to, a processor 20 and a memory 21. Those skilled in the art will understand that... Figure 6 The examples of computer devices are merely examples and do not constitute a limitation on computer devices. They may include more or fewer components than shown in the illustration, or combinations of certain components, or different components. For example, they may also include input / output devices, network access devices, etc.

[0130] The processor 20 may be a Central Processing Unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor.

[0131] The memory 21 can be an internal storage unit of the computer device in some embodiments, such as a hard disk or a memory of the computer device. The memory 21 can also be an external storage device of the computer device in other embodiments, such as a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) card, a Flash Card, etc. Further, the memory 21 can include both an internal storage unit and an external storage device of the computer device. The memory 21 is used to store an operating system, an application program, a BootLoader, data, and other programs, etc., such as program codes of the computer program, etc. The memory 21 can also be used to temporarily store data that has been output or will be output.

Claims

1. A method for predicting the thermal conductivity of composite materials, characterized in that, include: Obtain the data-driven structural features of the microstructure of the composite material to be predicted; The data-driven structural features are input into a pre-trained machine learning prediction model, which then outputs the predicted macroscopic physical performance values. The machine learning prediction model is obtained by training using a structure-performance database, which contains multiple data pairs. Each data pair includes a data-based structural feature characterizing the microstructure of the composite material and a predetermined macroscopic physical performance value corresponding to the data-based structural feature. The construction process of the structure-performance database includes: scanning and three-dimensional reconstruction of the composite material as a sample to obtain digitized structural features; calculating the material parameters of each component constituting the composite material under a preset working environment through molecular dynamics simulation; wherein the material parameters include the interfacial thermal resistance between components; performing finite element method thermal conduction simulation on the digitized structural features based on the material parameters to obtain the corresponding macroscopic physical performance values; and pairing the digitized structural features with the corresponding macroscopic physical performance values ​​to form the structure-performance database. The step of performing finite element method heat conduction simulation on the data-based structural features based on the material parameters to obtain the corresponding macroscopic physical performance values ​​includes: assigning the material parameters to the phase region corresponding to the data-based structural features, solving the steady-state heat conduction equation based on the finite element method, and obtaining the corresponding macroscopic physical performance values. The physical field distribution includes the temperature field distribution. The step of solving the steady-state heat conduction equation based on the finite element method to obtain the corresponding macroscopic physical performance values ​​includes: applying a constant heat flux density boundary condition to the heat input surface of the digital three-dimensional model, and applying a fixed temperature boundary condition to the heat dissipation surface of the digital three-dimensional model; iteratively solving the steady-state heat conduction equation until the temperature field distribution reaches a steady state, calculating the temperature gradient based on the steady-state temperature field distribution; and calculating the thermal conductivity based on the preset input heat flux density and the temperature gradient.

2. The method for predicting the thermal conductivity of composite materials as described in claim 1, characterized in that, The scanning and three-dimensional reconstruction of the composite material used as a sample to obtain data-driven structural features includes: The composite material was scanned and reconstructed in three dimensions using focused ion beam scanning electron microscopy to generate a digital three-dimensional model containing the spatial distribution of each phase. The digital 3D model is converted into the data-driven structural features.

3. The method for predicting the thermal conductivity of composite materials as described in claim 2, characterized in that, The loss function of the machine learning prediction model includes a data fitting term and a physical law constraint term. The data fitting term is used to measure the difference between the macroscopic physical performance values ​​predicted by the machine learning prediction model and the macroscopic physical performance values ​​in the structure-performance database. The physical law constraint term is used to ensure that the physical field distribution generated internally by the machine learning prediction model satisfies the preset physical control equation when making predictions based on the data structure features.

4. The method for predicting the thermal conductivity of composite materials as described in claim 3, characterized in that, The composite material is a nickel-yttrium oxide-stabilized zirconium oxide anode for solid oxide fuel cells; the macroscopic physical property is thermal conductivity; and the physical governing equation is a steady-state heat conduction equation.

5. A device for predicting the thermal conductivity of composite materials, characterized in that, include: The acquisition module is used to acquire the data-driven structural features of the microstructure of the composite material to be predicted; The prediction output module is used to input the data structure features into a pre-trained machine learning prediction model, and the machine learning prediction model outputs the predicted macroscopic physical performance values. The machine learning prediction model is obtained by training using a structure-performance database, which contains multiple data pairs. Each data pair includes a data-based structural feature characterizing the microstructure of the composite material and a predetermined macroscopic physical performance value corresponding to the data-based structural feature. The construction process of the structure-performance database includes: scanning and three-dimensional reconstruction of the composite material as a sample to obtain digitized structural features; calculating the material parameters of each component constituting the composite material under a preset working environment through molecular dynamics simulation; wherein the material parameters include the interfacial thermal resistance between components; performing finite element method thermal conduction simulation on the digitized structural features based on the material parameters to obtain the corresponding macroscopic physical performance values; and pairing the digitized structural features with the corresponding macroscopic physical performance values ​​to form the structure-performance database. The step of performing finite element method heat conduction simulation on the data-based structural features based on the material parameters to obtain the corresponding macroscopic physical performance values ​​includes: assigning the material parameters to the phase region corresponding to the data-based structural features, solving the steady-state heat conduction equation based on the finite element method, and obtaining the corresponding macroscopic physical performance values. The physical field distribution includes the temperature field distribution. The step of solving the steady-state heat conduction equation based on the finite element method to obtain the corresponding macroscopic physical performance values ​​includes: applying a constant heat flux density boundary condition to the heat input surface of the digital three-dimensional model, and applying a fixed temperature boundary condition to the heat dissipation surface of the digital three-dimensional model; iteratively solving the steady-state heat conduction equation until the temperature field distribution reaches a steady state, calculating the temperature gradient based on the steady-state temperature field distribution; and calculating the thermal conductivity based on the preset input heat flux density and the temperature gradient.

6. A computer device, characterized in that, The method includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the method as claimed in any one of claims 1 to 4.

7. A computer program product, characterized in that, When the computer program product is run on a computer device, it causes the computer device to perform the method as described in any one of claims 1 to 4.

Images