Porous medium heat and mass transport physical field prediction method, system and equipment and medium

Abstract

The invention discloses a porous medium heat and mass transport physical field prediction method, system and device and a medium. The porous medium heat and mass transportation physical field prediction method comprises the following steps: obtaining dimensioned input data of a porous medium heat and mass transportation process under a target working condition, the dimensioned input data at least comprising a pore structure, an initial boundary value condition and a physical property parameter; and taking the dimensional input data as data of an input end of the prediction model, and obtaining a prediction result of the porous medium heat and mass transportation physical field corresponding to the target working condition through the prediction model, the prediction model is trained in advance under the joint constraint of a dimensional input and output data pair and a heat and mass transport control equation obtained according to pre-obtained porous medium heat and mass transport physical field data, and the dimensional input and output data pair comprises dimensional input data and dimensional output data under corresponding conditions. By adopting the method, the prediction efficiency of the heat and mass transport physical field of the porous medium can be effectively improved, and the cost input of prediction can be reduced.

Description

Technical Field

[0001] The present application relates to the technical field of heat and mass transport in porous media, and in particular to a method, system, device and medium for predicting the physical field of heat and mass transport in porous media. Background Art

[0002] Heat and mass transport processes in porous media are widespread in nature and engineering, with important applications in fluid transport within biological tissues, energy resource development and utilization, and aerospace thermal protection. Accurately describing these processes is crucial for optimizing design and precise control in practical applications. Physical field information, including phase, velocity, pressure, and temperature fields, provides a detailed description of these processes and is crucial for characterizing the intrinsic physical processes and enabling downstream optimization.

[0003] At present, the commonly used methods for obtaining the physical field of heat and mass transport in porous media include micro-model visualization experiments and pore-scale numerical simulation methods. The micro-model visualization experiment method uses etching or additive manufacturing methods to process porous structures and encapsulate them into micro-models containing fluid inlet and outlet channels. Fluid is introduced into the micro-model, and fluid phase field and velocity field information is obtained by visual observation. The time and economic cost of micro-model processing in this method is high, and it is difficult to meet the rapid prediction needs in practical applications. The pore-scale numerical simulation method constructs a set of algebraic equations through mesh division, control equations and initial and boundary conditions, and obtains detailed inter-pore physical field information through iterative calculations. However, due to the complex solid skeleton structure of porous media and the complex interface interaction of multiphase fluids, pore-scale numerical simulation is difficult, and the calculation time cycle is several hours or even weeks, which is difficult to meet the rapid prediction needs in practical applications and difficult to support dynamic optimization design. Summary of the Invention

[0004] Based on this, it is necessary to provide a method, system, equipment and medium for predicting the physical field of heat and mass transport in porous media to address the above technical problems, which can effectively improve the efficiency of predicting the physical field of heat and mass transport in porous media and reduce the cost investment of prediction.

[0005] In a first aspect, the present application provides a method for predicting the physical field of heat and mass transport in porous media, comprising:

[0006] Obtaining dimensional input data of a heat and mass transport process in a porous medium under target working conditions, wherein the dimensional input data at least includes pore structure, initial and boundary conditions, and physical property parameters;

[0007] The dimensioned input data is used as input data of a prediction model to obtain a prediction result of a porous medium heat and mass transport physical field corresponding to the target working condition through the prediction model, wherein:

[0008] The prediction model is pre-trained based on dimensional input-output data pairs obtained from pre-acquired porous media heat and mass transport physical field data and the joint constraints of the heat and mass transport control equations. The dimensional input-output data pairs include dimensional input data and dimensional output data under conditions corresponding to the dimensional input data.

[0009] In some examples, before using the dimensioned input data as input data of a prediction model to obtain a prediction result of a porous medium heat and mass transport physical field corresponding to the target operating condition through the prediction model, the method further includes:

[0010] Obtain physical field data of heat and mass transport in porous media;

[0011] Obtaining the dimensioned input-output data pair according to the porous medium heat and mass transport physical field data;

[0012] Design the architecture of the initial prediction model based on the physics of heat and mass transport.

[0013] In some examples, after designing the initial predictive model architecture based on the physics of heat and mass transport, the following also occurs:

[0014] The initial prediction model is trained according to the joint constraints of the dimensional input-output data pair and the heat and mass transport control equation until the loss between the output of the initial prediction model and the label meets a predetermined condition, thereby obtaining the trained prediction model, wherein the label is the dimensional output data in the dimensional input-output data pair under the conditions corresponding to the dimensional input data.

[0015] In some examples, after designing the initial predictive model architecture based on the physics of heat and mass transport, the following also occurs:

[0016] Performing dimensionless conversion on the dimensional input-output data pair to determine a dimensionless input-output data pair;

[0017] The initial prediction model is trained according to the joint constraints of the dimensionless input-output data pair and the heat and mass transport control equation until the loss between the output of the initial prediction model and the label meets a predetermined condition, thereby obtaining the trained prediction model.

[0018] The label is dimensionless output data in the dimensionless input-output data pair under the condition corresponding to the dimensionless input data.

[0019] In some examples, the prediction result of the porous medium heat and mass transport physical field corresponding to the target operating condition obtained by the prediction model is a prediction result of the dimensionless physical field of the porous medium heat and mass transport corresponding to the target operating condition. After obtaining the prediction result of the dimensionless physical field of the porous medium heat and mass transport corresponding to the target operating condition by the prediction model, the method further includes:

[0020] The dimensionless physical field output by the prediction model is converted into a dimensional physical field according to the operating parameters of the target operating condition, so as to obtain a final prediction result of the physical field of heat and mass transport in the porous medium under the target operating condition.

[0021] In some examples, the dimensionless transformation of the dimensional input-output data pair to determine the dimensionless input-output data pair is implemented by the following formula:

[0022]

[0023] Where: x, y, z are spatial coordinates, t is time, u, v, w are the fluid velocity components in the x, y, and z directions respectively, p is the fluid pressure, L is the characteristic size, U is the characteristic velocity, ρ is the fluid density, and * represents a dimensionless physical quantity.

[0024] In some examples, after the dimensional input-output data pair is dimensionless and the dimensionless input-output data pair is determined, the method further includes:

[0025] The dimensionless output data is normalized using the following formula:

[0026]

[0027] Where std represents the standardized variable, the averaging operator in the denominator represents the arithmetic mean value in the data set after the porous medium volume averaging operation, and Re represents the characteristic Reynolds number.

[0028] In a second aspect, a porous media heat and mass transport physical field prediction system is provided, comprising:

[0029] An acquisition module is used to obtain dimensional input data of the heat and mass transport process of the porous medium under target working conditions, wherein the dimensional input data at least includes pore structure, initial and boundary conditions and physical property parameters;

[0030] A prediction module is used to use the dimensional input data as data at the input end of a prediction model to obtain a prediction result of the porous medium heat and mass transport physical field corresponding to the target working condition through the prediction model, wherein: the prediction model is pre-trained based on the dimensional input-output data pair obtained from the pre-acquired porous medium heat and mass transport physical field data and the heat and mass transport control equation. The dimensional input-output data pair includes dimensional input data and dimensional output data under conditions corresponding to the dimensional input data.

[0031] In a third aspect, a computer device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the steps of the method for predicting the physical field of heat and mass transport in porous media according to the first aspect and any possible implementation of the first aspect are implemented.

[0032] In a fourth aspect, a computer-readable storage medium is provided, on which a computer program is stored. When the program is executed by a processor, the steps of the method for predicting the physical field of heat and mass transport in porous media according to the first aspect and any possible implementation of the first aspect are implemented.

[0033] According to an embodiment of the present application, the dimensioned input data of the porous medium heat and mass transport process under specified working conditions is used as the input data of the prediction model, so that the physical field of the porous medium heat and mass transport under the specified working conditions can be quickly predicted by the prediction model. This prediction of the physical field of the porous medium heat and mass transport using a machine learning model replaces the acquisition of the physical field of the porous medium heat and mass transport using experiments and numerical simulations used in related technologies. Since the trained model can infer the physical field of the porous medium heat and mass transport under specified working conditions without iteration, the prediction speed of the physical field is effectively improved. In addition, the machine learning model has strong representation capabilities, thus ensuring the prediction accuracy of the physical field. In addition, the training of the machine learning model is driven by both data and the heat and mass transport control equations, which can ensure that the prediction results of the machine learning model conform to physical laws. The interpretability and generalization of the machine learning model can be enhanced by explicitly incorporating physical knowledge. In addition, high-precision pore-scale pressure and temperature fields that conform to the characteristics of experimental data can be obtained, which has the effect of indirect experimental measurement of pore-scale pressure and temperature fields. Because experimental methods in related technologies struggle to obtain pore-scale pressure and temperature fields, numerical simulations often involve simplifications, resulting in limited accuracy for the simulated pressure and temperature fields. Therefore, the embodiments of the present application, driven by both data and equations during the machine learning model training phase, can obtain pressure and temperature fields that satisfy the constraints of the experimentally measured phase field, velocity field, and the mass, momentum, and energy conservation equations. Depending on the characteristics of different physical processes, the pressure and temperature fields can be acquired separately or simultaneously. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Other features, objects and advantages of the present application will become more apparent upon reading the detailed description of non-limiting embodiments made with reference to the following drawings:

[0035] Figure 1 This is a flow chart of the method for predicting the physical field of heat and mass transport in porous media provided in an embodiment of the present application;

[0036] Figure 2 Another flow chart of the method for predicting the physical field of heat and mass transport in porous media provided in an embodiment of the present application;

[0037] Figure 3 This is another flow chart of the method for predicting the physical field of heat and mass transport in porous media provided in the embodiments of the present application;

[0038] Figure 4 This is a structural block diagram of the porous media heat and mass transport physical field prediction system provided in the embodiments of the present application;

[0039] Figure 5 This is a structural block diagram of the computer device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0040] The present application will be further described in detail below with reference to the embodiments and accompanying drawings. It should be understood that the specific embodiments described herein are intended only to illustrate the relevant application and are not intended to limit the application. It should also be noted that, for ease of description, only the portions relevant to the application are shown in the accompanying drawings.

[0041] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments in this application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0042] The following describes in detail the method, system, device and medium for predicting the physical field of heat and mass transport in porous media according to the embodiments of the present application in conjunction with the accompanying drawings.

[0043] The porous medium heat and mass transport physical field prediction method, system, device and medium of the present application solve the problems of high time and economic cost in high-precision acquisition of porous medium heat and mass transport physical field in related technologies.

[0044] Figure 1 FIG. 1 is a flow chart of a method for predicting the physical field of heat and mass transport in porous media according to an embodiment of the present application. Figure 1 As shown, the method for predicting the physical field of heat and mass transport in porous media according to an embodiment of the present application includes the following steps:

[0045] S101: Obtaining dimensional input data of a heat and mass transport process in a porous medium under target working conditions, wherein the dimensional input data at least includes pore structure, initial and boundary conditions, and physical property parameters.

[0046] The target working condition is a specified working condition, that is, the specified working condition input is first preprocessed to determine the dimensional input data of the heat and mass transport process of the porous medium under the above-mentioned specified working condition, wherein the dimensional input data includes but is not limited to pore structure, initial and boundary conditions, and physical properties.

[0047] In one embodiment of the present application, preprocessing includes but is not limited to: cleaning abnormal data, interpolating the pore structure of the porous medium to the required spatial resolution, processing the binary pore structure data into a continuous distribution such as Euclidean distance, etc.

[0048] S102: Using the dimensional input data as the data at the input end of the prediction model to obtain a prediction result of the porous medium heat and mass transport physical field corresponding to the target working condition through the prediction model, wherein: the prediction model is pre-trained based on the dimensional input-output data pair obtained from the pre-acquired porous medium heat and mass transport physical field data, and the heat and mass transport control equations. The dimensional input-output data pair includes dimensional input data and dimensional output data under conditions corresponding to the dimensional input data.

[0049] In one embodiment of the present application, before using the dimensioned input data as the data at the input end of the prediction model to obtain the prediction result of the porous medium heat and mass transport physical field corresponding to the target working condition through the prediction model, the method further includes:

[0050] Obtaining physical field data for heat and mass transport in porous media; obtaining the dimensioned input-output data pair based on the physical field data for heat and mass transport in porous media; and designing the architecture of an initial prediction model based on the physical knowledge of heat and mass transport. The design of the architecture of the initial prediction model is guided by the physical knowledge of heat and mass transport, so that the prediction model satisfies certain physical prior constraints. Therefore, the output of the prediction model is a physical field prediction result that conforms to physical laws, is interpretable, and is strongly generalized.

[0051] Specifically, obtaining high-precision data on the physical field of heat and mass transport in porous media usually involves obtaining phase and velocity field data through pore-scale experimental methods. Alternatively, existing pore-scale numerical simulation methods can be used to obtain phase, velocity, pressure, temperature, and component concentration field data.

[0052] In the above description, pore-scale experimental methods include but are not limited to bright-field observation, particle image velocimetry (PIV), confocal laser scanning microscopy, and stroboscopic X-ray microtomography. Pore-scale numerical simulation methods include but are not limited to the volume of fluid (VOF), phase field model (PFM), and lattice Boltzmann method (LBM).

[0053] Based on the physical field data of heat and mass transport in porous media, dimensional input and output data pairs are obtained. Specifically, the high-precision physical field data of heat and mass transport in porous media is preprocessed to determine the preprocessed dimensional input and output data pairs. The dimensional input data includes, but is not limited to, pore structure, initial and boundary conditions, and physical property parameter data. The dimensional output data includes the physical field data under the corresponding conditions. In this example, dimensional data preprocessing includes cleaning abnormal data, interpolating the pore structure and heat and mass transport physical field data to the required spatial resolution, and processing the binary pore structure data into a continuous distribution such as Euclidean distance.

[0054] The architecture of the initial prediction model is designed based on the physical knowledge of heat and mass transport, specifically: the initial prediction model is designed based on the physical knowledge of heat and mass transport, wherein the initial prediction model is also called the machine learning model architecture. The machine learning model architecture design includes but is not limited to: the selection of the type of neural network module, the design of the number and size of hidden layers, the design of the data flow transmission method, and the design of the activation function. The ideas for designing the machine learning model architecture based on the physical knowledge of heat and mass transport include but are not limited to: designing activation functions and neural network modules that meet physical prior properties, such as: symmetry, non-negativity, positivity, monotonicity, partial differential operators and basic principles of thermodynamics. In a specific example, for the problem of simultaneous prediction of multiple physical fields, a single or multiple decoder combinations can also be designed, corresponding to the simultaneous decoding of multiple physical fields or the decoding of each physical field by a specific decoder.

[0055] In one embodiment of the present application, after designing the architecture of the initial prediction model based on the physical knowledge of heat and mass transport, it also includes: training the initial prediction model based on the joint constraints of the dimensional input-output data pair and the heat and mass transport control equation until the loss between the output of the initial prediction model and the label meets a predetermined condition, thereby obtaining the trained prediction model, wherein the label is the dimensional output data in the dimensional input-output data pair under the conditions corresponding to the dimensional input data.

[0056] like Figure 2As shown in the figure, one implementation process of the method for predicting the physical field of heat and mass transport in porous media is: obtaining high-precision data of the physical field of heat and mass transport in porous media; preprocessing the high-precision data of the physical field of heat and mass transport in porous media, and determining the dimensional input-output data pairs after preprocessing; designing a machine learning model architecture based on the physical knowledge of heat and mass transport; then training the machine learning model according to the dimensional input-output data pairs and the constraints of the heat and mass transport control equations to determine the model parameters; finally, predicting the physical field of heat and mass transport in porous media based on the trained machine learning model.

[0057] According to the method for predicting the physical field of heat and mass transport in porous media according to the embodiment of the present application, the dimensioned input data of the heat and mass transport process in porous media under specified working conditions is used as the input data of the prediction model, so that the physical field of heat and mass transport in porous media under the specified working conditions can be quickly predicted by the prediction model. This prediction of the physical field of heat and mass transport in porous media using a machine learning model replaces the acquisition of the physical field of heat and mass transport in porous media using experiments and numerical simulations in related technologies. Since the trained model can infer the physical field of heat and mass transport in porous media under specified working conditions without iteration, the prediction speed of the physical field is effectively improved. Moreover, the machine learning model has strong representation capabilities, thus ensuring the prediction accuracy of the physical field. In addition, the training of the machine learning model is driven by both data and the heat and mass transport control equations, which can ensure that the prediction results of the machine learning model conform to physical laws. The interpretability and generalization of the machine learning model can be enhanced by explicitly incorporating physical knowledge. In addition, high-precision pore-scale pressure and temperature fields that conform to the characteristics of experimental data can be obtained, which has the effect of indirect experimental measurement of pore-scale pressure and temperature fields. Because experimental methods in related technologies struggle to obtain pore-scale pressure and temperature fields, numerical simulations often involve simplifications, resulting in limited accuracy for the simulated pressure and temperature fields. Therefore, the embodiments of the present application, driven by both data and equations during the machine learning model training phase, can obtain pressure and temperature fields that satisfy the constraints of the experimentally measured phase field, velocity field, and the mass, momentum, and energy conservation equations. Depending on the characteristics of different physical processes, the pressure and temperature fields can be acquired separately or simultaneously.

[0058] In one embodiment of the present application, data can also be organized in the form of dimensionless input and output. For physical similarity problems, only one set of training data is needed, which reduces the redundancy of training data. In the prediction stage, the specified working conditions are converted into dimensionless inputs, and the model predicts the dimensionless physical field and then converts it into the corresponding dimensional physical field. This process ensures that the model trained by data of one scale can be applied at different scales and has scale invariance. In addition, since the model introduces physical prior constraints in the design and training stages, it can further reduce the dependence on the amount of training data, and realize physical similarity and scale invariance physical field prediction. Specifically, combined with Figure 3 As shown, the method for predicting the physical field of heat and mass transport in porous media, after designing the architecture of the initial prediction model based on the physical knowledge of heat and mass transport, further includes: non-dimensionalizing the dimensional input-output data pair to determine the dimensionless input-output data pair; training the initial prediction model based on the joint constraints of the dimensionless input-output data pair and the heat and mass transport control equation until the loss between the output of the initial prediction model and the label meets a predetermined condition, thereby obtaining the trained prediction model, wherein the label is the dimensionless output data in the dimensionless input-output data pair under the conditions corresponding to the dimensionless input data.

[0059] Furthermore, the prediction result of the porous medium heat and mass transport physical field corresponding to the target working condition obtained by the prediction model is the prediction result of the dimensionless physical field of the porous medium heat and mass transport corresponding to the target working condition. After the prediction result of the dimensionless physical field of the porous medium heat and mass transport corresponding to the target working condition is obtained by the prediction model, it also includes: converting the dimensionless physical field output by the prediction model into a dimensional physical field according to the working condition parameters of the target working condition, so as to obtain the final prediction result of the porous medium heat and mass transport physical field under the target working condition.

[0060] In the above example, the dimensioned input-output data pair is dimensionless, and the dimensionless input-output data pair is determined, which is achieved by the following formula:

[0061]

[0062] Where: x, y, z are spatial coordinates, t is time, u, v, w are the fluid velocity components in the x, y, and z directions, respectively, p is the fluid pressure, L is the characteristic dimension, U is the characteristic velocity, and ρ is the fluid density. * denotes a dimensionless physical quantity.

[0063] After the dimensionless input and output data pairs are dimensioned to determine the dimensionless input and output data pairs, the method further includes: normalizing the dimensionless output data using the following formula:

[0064]

[0065] Where std represents the standardized variable, the averaging operator in the denominator represents the arithmetic mean value in the data set after the porous medium volume averaging operation, and Re represents the characteristic Reynolds number.

[0066] Specifically, the dimensional input and output data pairs are dimensionlessized to determine dimensionless input and output data pairs. The dimensionless output data is standardized to unify the distribution characteristics of data from different types of physical fields. This ensures that the influence weights of different physical fields on the model loss function and the backpropagation gradient optimization process are consistent, reducing the difficulty of simultaneous training of multiple physical fields and improving prediction accuracy. By designing an end-to-end model mapping dimensionlessly and guiding the design of dimensionless training datasets with similarity theory, physical similarity predictions can be achieved while reducing dataset redundancy.

[0067] In the above embodiment, the specific calculation method of the averaging operator is as follows:

[0068]

[0069] Among them, ψ represents a general variable that needs to be processed by the averaging operator, Represents the dimensionless volume of the porous medium fluid domain, N represents the number of data entries in the data set, and i represents the i-th data entry in the data set. The specific calculation method of the characteristic Reynolds number is:

[0070]

[0071] Where μ is the fluid dynamic viscosity. The standardized processing method of the pressure field ensures that the pressure data variation range is of the same order of magnitude as the velocity, improving the training effect of the pressure field and velocity field participating in the model training at the same time. Optionally, the dimensionless training data set design is guided by similarity theory. For the single-phase flow process in porous media, the data set is designed according to the two dimensions of dimensionless pore structure and Reynolds number. For the two-phase flow process in porous media, the data set is designed according to the six dimensions of dimensionless pore structure, Reynolds number, capillary number, viscosity ratio, density ratio and contact angle.

[0072] Based on the dimensionless input and output data, heat and mass transport control equations, and boundary value constraints, the machine learning model is trained to determine model parameters. Specifically, the model loss function driven by both data and domain knowledge is designed as follows:

[0073] Loss = (1-α-β)Loss data +αLoss equation +βLoss b.c.

[0074] Where Loss represents the loss function, α and β are weighting coefficients. Data represents data constraints, equation represents the constraints of the heat and mass transport control equation, and bc represents the boundary condition constraints. The data constraint method means comparing the model output results with the labeled data and calculating the loss function using methods such as mean absolute error or mean square error. The heat and mass transport control equation constraint means substituting the model output physical field into the control equation to determine whether the output physical field satisfies the basic conservation laws and quantitatively calculate the deviation. The heat and mass transport control equation constraint includes adding one or more of the mass conservation equation, momentum conservation equation, energy conservation equation, and component transport equation to the loss function in the form of regularization terms to constrain the model training process.

[0075] For fully connected neural networks, automatic differentiation techniques are used to calculate the derivative terms in the governing equations. For convolutional neural networks, the Sobel operator is used to calculate the derivative terms in the governing equations. Boundary value constraints involve comparing the boundary values of the model output physical fields with the boundary conditions of the actual physical process, and using methods such as mean absolute error or mean square error to calculate the loss function. For the second and third types of boundary value conditions involving derivative calculation, automatic differentiation or the Sobel operator can be used to calculate the boundary derivatives of the model output physical fields. Heat and mass transport governing equation constraints and boundary value constraints do not require labeled data, reducing the model training process's reliance on large amounts of training data. Heat and mass transport governing equation constraints ensure that the model output phase, velocity, pressure, and temperature fields satisfy fundamental conservation laws, reducing the probability of unphysical values in model predictions. When experimental measurements only provide phase and velocity field data, introducing governing equation constraints ensures that the model output pressure and temperature fields match the experimental data characteristics, enabling high-precision predictions for multiple types of physical fields.

[0076] The dimensionless physical field of heat and mass transport in porous media is predicted using a trained machine learning model. During the prediction phase, the pore structure, initial and boundary conditions, and physical properties of the specified working condition are converted into dimensionless form and input into the trained machine learning model. The model then infers the dimensionless physical field predictions.

[0077] The dimensionless physical field is then converted to a dimensional field based on the actual operating parameters, yielding the user's desired prediction results for the specified operating conditions. This dimensional field supports subsequent optimization design and dynamic control. The conversion of the dimensionless physical field to a dimensional field is the inverse of the dimensionless and normalization processes.

[0078] This approach not only improves the speed of physical field prediction, ensures that prediction results are physically consistent, interpretable, and strongly generalizable, and that high-precision pore-scale pressure and temperature fields that conform to the characteristics of experimental data are obtained, achieving the effect of indirect experimental measurement of pore-scale pressure and temperature fields, but also ensures physical similarity and scale invariance in physical field predictions, reduces reliance on the amount of training data, and organizes data in a dimensionless input-output format. For physical similarity problems, only one set of training data is required, reducing training data redundancy. During the prediction phase, the specified operating conditions are converted into dimensionless inputs, and the model predicts the dimensionless physical fields, which are then converted into the corresponding dimensional physical fields. This process ensures that a model trained on data at one scale can be applied at different scales and is scale-invariant. Furthermore, by introducing physical prior constraints during the model design and training phases, reliance on the amount of training data can be further reduced.

[0079] Figure 4 FIG. 1 is a structural block diagram of a porous media heat and mass transport physical field prediction system according to an embodiment of the present application. Figure 4 As shown, the porous media heat and mass transport physical field prediction system according to an embodiment of the present application includes: an acquisition module 410 and a prediction module 420, wherein:

[0080] An acquisition module 410 is configured to obtain dimensional input data of a porous medium heat and mass transport process under target working conditions, wherein the dimensional input data at least includes pore structure, initial and boundary conditions, and physical property parameters;

[0081] The prediction module 420 is used to use the dimensional input data as the data at the input end of the prediction model to obtain the prediction result of the porous medium heat and mass transport physical field corresponding to the target working condition through the prediction model, wherein: the prediction model is pre-trained based on the dimensional input-output data pair obtained from the pre-acquired porous medium heat and mass transport physical field data and the heat and mass transport control equation. The dimensional input-output data pair includes dimensional input data and dimensional output data under conditions corresponding to the dimensional input data.

[0082] According to the porous media heat and mass transport physical field prediction system of the embodiment of the present application, the dimensioned input data of the porous media heat and mass transport process under the specified working conditions is used as the input data of the prediction model, so that the porous media heat and mass transport physical field under the specified working conditions can be quickly predicted by the prediction model. This prediction of the porous media heat and mass transport physical field using a machine learning model replaces the acquisition of the porous media heat and mass transport physical field using experiments and numerical simulations used in related technologies. Since the trained model can infer the porous media heat and mass transport physical field under the specified working conditions without iteration, the prediction speed of the physical field is effectively improved. Moreover, the machine learning model has strong representation capabilities, thus ensuring the prediction accuracy of the physical field. In addition, the training of the machine learning model is driven by both data and the heat and mass transport control equations, which can ensure that the prediction results of the machine learning model conform to the physical laws. The interpretability and generalization of the machine learning model can be enhanced by explicitly incorporating physical knowledge. In addition, high-precision pore-scale pressure and temperature fields that conform to the characteristics of experimental data can be obtained, which has the effect of indirect experimental measurement of pore-scale pressure and temperature fields. Because experimental methods in related technologies struggle to obtain pore-scale pressure and temperature fields, numerical simulations often involve simplifications, resulting in limited accuracy for the simulated pressure and temperature fields. Therefore, the embodiments of the present application, driven by both data and equations during the machine learning model training phase, can obtain pressure and temperature fields that satisfy the constraints of the experimentally measured phase field, velocity field, and the mass, momentum, and energy conservation equations. Depending on the characteristics of different physical processes, the pressure and temperature fields can be acquired separately or simultaneously.

[0083] The specific definition of the porous media heat and mass transport physical field prediction system can be found in the definition of the porous media heat and mass transport physical field prediction method described above, and will not be repeated here. The various modules of the above-mentioned porous media heat and mass transport physical field prediction system can be implemented in whole or in part through software, hardware, or a combination thereof. The above-mentioned modules can be embedded in or independent of the processor of the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the corresponding operations of the above modules.

[0084] In one embodiment, a computer device is provided. Figure 5 This is a block diagram of the computer device provided in an embodiment of the present application. The computer device includes a memory and a processor. The memory stores a computer program. When the processor executes the computer program, it implements the aforementioned porous media heat and mass transport physical field prediction method embodiment. For example, the following steps are performed: obtaining dimensional input data of the porous media heat and mass transport process under target working conditions, wherein the dimensional input data includes at least pore structure, initial and boundary conditions, and physical property parameters;

[0085] The dimensioned input data is used as input data of a prediction model to obtain a prediction result of a porous medium heat and mass transport physical field corresponding to the target working condition through the prediction model, wherein:

[0086] The prediction model is pre-trained based on dimensional input-output data pairs obtained from pre-acquired porous media heat and mass transport physical field data and the joint constraints of the heat and mass transport control equations. The dimensional input-output data pairs include dimensional input data and dimensional output data under conditions corresponding to the dimensional input data.

[0087] The present application also provides a computer-readable storage medium storing a computer program. When a processor executes the computer program, the aforementioned method for predicting the physical field of heat and mass transport in porous media is implemented. For example, the method includes: obtaining dimensional input data of the heat and mass transport process in porous media under target working conditions, wherein the dimensional input data includes at least pore structure, initial and boundary conditions, and physical property parameters;

[0088] The dimensioned input data is used as input data of a prediction model to obtain a prediction result of a porous medium heat and mass transport physical field corresponding to the target working condition through the prediction model, wherein:

[0089] The prediction model is pre-trained based on dimensional input-output data pairs obtained from pre-acquired porous media heat and mass transport physical field data and the joint constraints of the heat and mass transport control equations. The dimensional input-output data pairs include dimensional input data and dimensional output data under conditions corresponding to the dimensional input data.

[0090] The present application embodiment provides a computer program product, which includes instructions. When the instructions are executed, the method described in the embodiment of the present application is executed. For example, you can execute Figure 1 The steps of the porous media heat and mass transport physical field prediction method shown in the figure, for example, are as follows: obtaining dimensional input data of the porous media heat and mass transport process under target working conditions, wherein the dimensional input data at least includes pore structure, initial and boundary conditions, and physical property parameters;

[0091] The dimensioned input data is used as input data of a prediction model to obtain a prediction result of a porous medium heat and mass transport physical field corresponding to the target working condition through the prediction model, wherein:

[0092] The prediction model is pre-trained based on dimensional input-output data pairs obtained from pre-acquired porous media heat and mass transport physical field data and the joint constraints of the heat and mass transport control equations. The dimensional input-output data pairs include dimensional input data and dimensional output data under conditions corresponding to the dimensional input data.

[0093] Those skilled in the art will appreciate that all or part of the processes in the methods for implementing the above embodiments can be accomplished by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include processes of the embodiments of the above methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application may include at least one of non-volatile and volatile memory. Non-volatile memory may include read-only memory (ROM), magnetic tape, floppy disk, flash memory or optical memory, etc. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0094] The technical features of the above embodiments can be combined arbitrarily. In order to make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0095] The above embodiments merely illustrate several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent application. It should be noted that a person of ordinary skill in the art may make various modifications and improvements without departing from the spirit of the present application, all of which fall within the scope of protection of the present application. Therefore, the scope of protection of the present patent application shall be determined by the appended claims.

Claims

1. A method for predicting the physical field of heat and mass transport in porous media, characterized in that: include: Obtaining dimensional input data of a heat and mass transport process in a porous medium under target working conditions, wherein the dimensional input data at least includes pore structure, initial and boundary conditions, and physical property parameters; The dimensioned input data is used as input data of a prediction model to obtain a prediction result of a porous medium heat and mass transport physical field corresponding to the target working condition through the prediction model, wherein: The prediction model is pre-trained based on dimensional input-output data pairs obtained from pre-acquired porous media heat and mass transport physical field data and the joint constraints of the heat and mass transport control equations. The dimensional input-output data pairs include dimensional input data and dimensional output data under conditions corresponding to the dimensional input data.

2. The method for predicting physical fields of heat and mass transport in porous media according to claim 1, characterized in that: Before using the dimensioned input data as data at the input end of a prediction model to obtain a prediction result of a porous medium heat and mass transport physical field corresponding to the target working condition through the prediction model, the method further includes: Obtain physical field data of heat and mass transport in porous media; Obtaining the dimensioned input-output data pair according to the porous medium heat and mass transport physical field data; Design the architecture of the initial prediction model based on the physics of heat and mass transport.

3. The method for predicting the physical field of heat and mass transport in porous media according to claim 2, characterized in that: After designing the initial predictive model architecture based on the physics of heat and mass transport, it also includes: The initial prediction model is trained according to the joint constraints of the dimensional input-output data pair and the heat and mass transport control equation until the loss between the output of the initial prediction model and the label meets a predetermined condition, thereby obtaining the trained prediction model, wherein the label is the dimensional output data in the dimensional input-output data pair under the conditions corresponding to the dimensional input data.

4. The method for predicting physical fields of heat and mass transport in porous media according to claim 2, characterized in that: After designing the initial predictive model architecture based on the physics of heat and mass transport, it also includes: Performing dimensionless conversion on the dimensional input-output data pair to determine a dimensionless input-output data pair; The initial prediction model is trained according to the joint constraints of the dimensionless input-output data pair and the heat and mass transport control equation until the loss between the output of the initial prediction model and the label meets a predetermined condition, thereby obtaining the trained prediction model. The label is dimensionless output data in the dimensionless input-output data pair under the condition corresponding to the dimensionless input data.

5. The method for predicting physical fields of heat and mass transport in porous media according to claim 4, characterized in that: in, The prediction result of the porous medium heat and mass transport physical field corresponding to the target working condition obtained by the prediction model is a prediction result of the dimensionless physical field of the porous medium heat and mass transport corresponding to the target working condition. After the prediction result of the dimensionless physical field of the porous medium heat and mass transport corresponding to the target working condition is obtained by the prediction model, the method further includes: The dimensionless physical field output by the prediction model is converted into a dimensional physical field according to the operating parameters of the target operating condition, so as to obtain a final prediction result of the physical field of heat and mass transport in the porous medium under the target operating condition.

6. The method for predicting physical fields of heat and mass transport in porous media according to claim 4 or 5, characterized in that: The dimensionless conversion of the dimensioned input-output data pair to determine the dimensionless input-output data pair is achieved by the following formula: Where: x, y, z are spatial coordinates, t is time, u, v, w are the fluid velocity components in the x, y, and z directions respectively, p is the fluid pressure, L is the characteristic size, U is the characteristic velocity, ρ is the fluid density, and * represents a dimensionless physical quantity.

7. The method for predicting physical fields of heat and mass transport in porous media according to claim 6, characterized in that: After the dimensioned input-output data pair is dimensionless and the dimensionless input-output data pair is determined, the method further includes: The dimensionless output data is normalized using the following formula: Where std represents the standardized variable, the averaging operator in the denominator represents the arithmetic mean value in the data set after the porous medium volume averaging operation, and Re represents the characteristic Reynolds number.

8. A physical field prediction system for heat and mass transport in porous media, characterized in that: include: An acquisition module is used to obtain dimensional input data of the heat and mass transport process of the porous medium under target working conditions, wherein the dimensional input data at least includes pore structure, initial and boundary conditions and physical property parameters; A prediction module is used to use the dimensional input data as data at the input end of a prediction model to obtain a prediction result of the porous medium heat and mass transport physical field corresponding to the target working condition through the prediction model, wherein: the prediction model is pre-trained based on the dimensional input-output data pair obtained from the pre-acquired porous medium heat and mass transport physical field data and the heat and mass transport control equation. The dimensional input-output data pair includes dimensional input data and dimensional output data under conditions corresponding to the dimensional input data.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the method for predicting the physical field of heat and mass transport in porous media according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium comprising a memory and a computer program stored in the memory and executable on a processor, characterized in that: When the program is executed by a processor, the method for predicting the physical field of heat and mass transport in porous media according to any one of claims 1 to 7 is implemented.

Images