Training method and device of flow field reconstruction model, medium and program product

Abstract

The invention discloses a flow field reconstruction model training method and device, a medium and a program product. According to the method, data enhancement processing is carried out on an initial flow field according to macroscopic dimensionless characteristic parameters of the initial flow field obtained in advance, and sample data with labels are obtained; wherein the sample data is flow field boundary condition data, and the label is flow field data; dividing the sample data into training data and test data according to a preset proportion; and training a pre-constructed model based on the training data, and evaluating the trained model based on the test data to obtain a flow field reconstruction model. According to the technical scheme, model training is carried out based on the flow field data, so that the flow field is rapidly and accurately reconstructed, the calculation time is greatly shortened, the requirement for calculation equipment is lowered, and the flow field reconstruction efficiency is improved.

Description

Technical Field

[0001] The present application relates to the field of flow field reconstruction technology, and in particular to a training method, equipment, medium and program product for a flow field reconstruction model. Background Art

[0002] Flow heat transfer refers to the exchange of heat between a fluid and the surrounding environment during its flow. Flow field reconstruction models can reconstruct and reproduce the fluid's motion, providing a scientific basis for solving various fluid mechanics problems. For example, in aerospace, thermal radiation from components such as aircraft fuselage surfaces, engine systems, and fuel systems can be analyzed to optimize the aircraft's cooling system and improve its performance and safety. Another example is nuclear power generation, where flow heat transfer analysis of heat exchangers in high-temperature reactors can be performed to ensure proper operation.

[0003] Currently, solutions to flow and heat transfer problems in narrow slit channels and other complex flow and heat transfer problems are typically based on traditional fluid dynamics analysis software. Traditional fluid dynamics analysis software, based on the fundamental principles and equations of fluid mechanics, uses numerical analysis and calculation methods to simulate fluid flow. However, for complex fluid systems with large-scale and multi-scale characteristics, manual feature extraction is often required to achieve dimensionality reduction. This results in long computational cycles, low accuracy, and difficulty in addressing the effects of nonlinear coupling between features.

[0004] Therefore, how to provide a technical solution that can quickly and accurately reconstruct the flow field is a technical problem that needs to be solved urgently by those skilled in the art. Summary of the Invention

[0005] The present application provides a training method, device, medium and program product for a flow field reconstruction model. By training the model based on flow field data, the flow field can be reconstructed quickly and accurately, which greatly reduces the calculation time, reduces the requirements for computing equipment, and improves the efficiency of flow field reconstruction.

[0006] According to one aspect of the present application, a method for training a flow field reconstruction model is provided, the method comprising:

[0007] According to the macro-dimensionalless characteristic parameters of the initial flow field obtained in advance, data enhancement processing is performed on the initial flow field to obtain sample data with labels; wherein the sample data is flow field boundary condition data, and the label is flow field data;

[0008] Dividing the sample data into training data and test data according to a preset ratio;

[0009] The pre-built model is trained based on the training data, and the trained model is evaluated based on the test data to obtain a flow field reconstruction model.

[0010] According to another aspect of the present application, an electronic device is provided, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the training method for the flow field reconstruction model described in any embodiment of the present application.

[0011] According to another aspect of the present application, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the training method of the flow field reconstruction model described in any embodiment of the present application when executed.

[0012] According to another aspect of the present application, a computer program product is provided, which includes a computer program. When the computer program is executed by a processor, it implements the training method of the flow field reconstruction model described in any embodiment of the present application.

[0013] The technical solution provided by this application performs data augmentation processing on the initial flow field based on pre-acquired macro-dimensionalless characteristic parameters of the initial flow field to obtain labeled sample data; wherein the sample data is the flow field boundary condition data and the label is the flow field data; the sample data is divided into training data and test data according to a preset ratio; a pre-built model is trained based on the training data, and the trained model is evaluated based on the test data to obtain a flow field reconstruction model. By training the flow field data, this technical solution can quickly reconstruct the flow field, significantly reducing calculation time, lowering the requirements for computing equipment, and improving the efficiency of flow field reconstruction and simulation.

[0014] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present application, nor is it intended to limit the scope of the present application. Other features of the present application will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0016] Figure 1 This is a flowchart of a method for training a flow field reconstruction model provided in Example 1 of the present application.

[0017] Figure 2A schematic diagram of the training process of a generative adversarial network model provided in Example 1 of the present application.

[0018] Figure 3 A schematic diagram of the training process of a feedforward neural network model provided in Example 1 of the present application.

[0019] Figure 4 This is a flowchart of a method for training a flow field reconstruction model provided in Example 2 of the present application.

[0020] Figure 5 It is a structural diagram of a device for implementing a training method for a flow field reconstruction model in an embodiment of the present application. DETAILED DESCRIPTION

[0021] In order to enable those skilled in the art to better understand the present invention, the following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of this application.

[0022] It should be noted that the terms "initial", "training", "test", "standard", "sample", "first", "second", etc. in the specification and claims of this application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0023] Example 1

[0024] Figure 1 This is a flow chart of a method for training a flow field reconstruction model provided in the first embodiment of the present application. This embodiment is applicable to the case of training a flow field reconstruction model. The method can be executed by a training device for a flow field reconstruction model. The training device for a flow field reconstruction model can be implemented in the form of hardware and / or software. The training device for a flow field reconstruction model can be configured in a device with data processing capabilities. Figure 1 As shown, the method includes the following steps.

[0025] S110: Perform data enhancement processing on the initial flow field based on the macro-dimensionalless characteristic parameters of the initial flow field obtained in advance to obtain sample data with labels, wherein the sample data is flow field boundary condition data and the labels are flow field data.

[0026] The initial flow field may be a flow field actually measured in advance by a particle image velocimetry technique, a hot wire anemometer or other equipment, or a flow field simulated based on theoretical assumptions or empirical formulas.

[0027] The macroscopic dimensionless characteristic parameters can be parameters that represent the fluid properties in the initial flow field and are independent of specific physical units. Examples include the Reynolds number, which represents the similarity factor of the fluid's viscosity, the Nusselt number, which represents convective heat transfer capacity, the Prandtl number, which represents the relationship between the fluid's temperature boundary layer and its flow boundary layer, and the Mach number, which represents the fluid's compressibility. This embodiment of the present application is not limited to these parameters, and appropriate macroscopic dimensionless characteristic parameters can be selected based on actual needs.

[0028] In this application, data enhancement processing can be performed on the initial flow field while ensuring that the macroscopic dimensionless characteristic parameters of the initial flow field remain unchanged, that is, without changing the initial flow field characteristics. For example, new flow field data can be generated by adjusting some physical parameters of the initial flow field. Based on the basic principles of fluid mechanics, some small disturbances can be added to the initial flow field to simulate unstable factors or external interference that may exist in the actual flow field, thereby increasing the complexity and authenticity of the flow field data. The initial flow field can also be subjected to geometric transformation processing, such as rotating the velocity vector diagram of the flow field by a certain angle to simulate flow field data in different directions.

[0029] By performing data enhancement processing on the initial flow field, the diversity of flow field data can be enriched, thereby improving the generalization ability of subsequent models.

[0030] Furthermore, labels can be added to the enhanced flow field data to obtain labeled sample data for subsequent model training and analysis. The types of label data and sample data can be determined based on the applicable scenario of the trained model.

[0031] For example, in the field of nuclear energy technology, printed circuit board heat exchangers (PCBHEs) have been proposed for handling heat transfer between the coolant and the power cycle working fluid in the secondary loop of advanced high-temperature reactors. Compared to conventional shell-and-tube heat exchangers, PCBHEs achieve a very high surface area density and a small hydraulic diameter, which translates to higher heat transfer rates and a smaller form factor. This makes PCBHEs well-suited for pressurized s-CO2 Brayton cycles operating at pressures up to 1000 bar and temperatures between -150°C and 900°C. Consequently, PCBHEs are expected to become standard components for secondary heat exchangers in sodium / s-CO2 cycles for fourth-generation nuclear power reactors.

[0032] For this type of printed circuit board heat exchanger, the internal fluid flow and heat transfer problem can be viewed as a narrow channel flow and heat transfer problem. When training this type of flow field reconstruction model, the boundary condition data optionally includes at least one of inlet flow rate, inlet temperature, and heat flux; and the flow field data includes at least one of velocity, temperature, and pressure.

[0033] Among them, the inlet flow rate can be the volume or mass of the fluid passing through a certain cross-section per unit time at the entrance of the initial flow field, which determines the overall flow velocity distribution and flow rate of the fluid in the narrow slit channel; the inlet temperature can be the temperature of the fluid when entering the initial flow field, which determines the exchange intensity between the flow field and the surrounding environment or other components; the heating flux can be the heat transfer rate through a unit area per unit time, which is used to simulate the heating or cooling process of the surface of an object in the flow field.

[0034] Among them, flow field data can provide clear learning objectives and supervision information for the subsequent model training process, so that the model can learn the mapping relationship between flow field data and sample data based on label data, thereby realizing the prediction of flow field data.

[0035] Specifically, different flow field reconstruction models can be established for different flow field data types. For example, if the sample data is labeled as x-direction velocity data, a flow field reconstruction model for x-direction velocity can be trained; for another example, if the sample data is labeled as temperature data, a flow field reconstruction model for temperature can be trained.

[0036] S120: Divide the sample data into training data and test data according to a preset ratio.

[0037] In this application, the preset ratio of training data to test data can be determined based on factors such as the total amount of sample data, the complexity of the model, and computing resources. For example, if the sample data volume is large and the model is relatively simple, a ratio of 80%:20% or 90%:10% can be selected to allow more data to be used for training the model and improve the model's learning effect. If the sample data volume is small, a ratio of 70%:30% may be selected to ensure that the test data is sufficiently representative to evaluate the model's performance.

[0038] In addition, if hyperparameter verification is required, the sample data can be divided into training data, verification data, and test data in a ratio of 75%:15%:15%.

[0039] S130: Training the pre-built model based on the training data, and evaluating the trained model based on the test data to obtain a flow field reconstruction model.

[0040] Among them, the pre-built model can select the appropriate model architecture according to the characteristics of the flow field data, such as convolutional neural network, recurrent neural network, feedforward neural network, generative adversarial network, etc.

[0041] Specifically, the input layer, output layer, and hidden layer of the model can be set according to the training data, including the number of neurons, the number of hidden layers, the connection method, the loss function expression, and the optimization algorithm.

[0042] The number of neurons in the input layer matches the number of boundary conditions. The size of the hidden layer is selected based on the actual computational scale and can be expanded in both width and depth. The connection method can be fully connected, the loss function is the MSE function, and the gradient algorithm and Adam optimization algorithm can be selected.

[0043] During model training, you can set an appropriate training cycle for the model and choose between CPU or GPU hardware resources. Each training cycle outputs the prediction accuracy of both the training data and the test data. If the test data prediction accuracy decreases, training is stopped. After model training is complete, the generalization prediction accuracy of the model is tested on the test data. If the accuracy does not meet the requirements, the model is retrained.

[0044] Furthermore, models that meet the generalization accuracy can be deployed according to user needs to implement the input parameterized boundary conditions, obtain high-precision flow field data, and complete rapid simulation prediction of the flow field.

[0045] Optionally, the pre-constructed model is a generative adversarial network model, which includes a generator and a discriminator; accordingly, the pre-constructed model is trained based on the training data, including: inputting the training data into the generator to generate false data; inputting the training data and the false data into the discriminator, calculating the loss function according to the output result of the discriminator, and training the pre-constructed model based on the loss function; or, the pre-constructed model is a feedforward neural network model, which includes an input layer, a hidden layer and an output layer; accordingly, the pre-constructed model is trained based on the training data, including: inputting the training data into the input layer, and passing through each of the hidden layers and the output layer in turn to obtain a predicted output; determining the learning error of the feedforward neural network model according to the predicted output and the training data; and updating the model parameters of the feedforward neural network model according to the learning error.

[0046] The generator learns the characteristics of the training data and, under the guidance of the discriminator, tries to fit the random noise distribution to the true distribution of the training data as closely as possible, thereby generating data similar to the characteristics of the training data, i.e., fake data. The discriminator distinguishes whether the input data is real training data or fake data generated by the generator and provides feedback to the generator. The two networks are trained alternately, improving their capabilities simultaneously, until the data generated by the generator is indistinguishable from the real data and reaches a certain balance with the discriminator's capabilities.

[0047] Specifically, such as Figure 2 As shown in the figure, a batch of noise samples can be randomly generated as input to the generator; the generator generates a batch of fake data, and mixes the generated fake data with the real training data as input to the discriminator; the discriminator discriminates the input samples and outputs the discrimination results; the loss function of the generator and the discriminator is calculated according to the discrimination results; the parameters of the generator and the discriminator are updated, and the gradient is updated through back propagation; the above steps are repeated until the fake data generated by the generator reaches the expected quality.

[0048] The loss function of the generator can be to maximize the probability of the discriminator's judgment on the training data, while the loss function of the discriminator can be to maximize the probability of judgment on the training data and minimize the probability of judgment on false data.

[0049] The above technical solution enriches the diversity of data by training sample data in a generative adversarial network, improves the model's ability to learn flow field data and boundary conditions, and thus improves the prediction accuracy of the flow field reconstruction model.

[0050] In addition, the pre-built model can also be a feedforward neural network model. Specifically, Figure 3As shown in Figure 1, training data is fed into the model's input layer. Each neuron in the hidden layer then receives and outputs the data to the next layer, ultimately to the output layer. During training, a loss function can be defined to measure the difference between the model's predicted values and the actual values. Common loss functions include mean squared error (MSE) and cross-entropy loss.

[0051] Optionally, the trained model is evaluated based on the test data to obtain a flow field reconstruction model, including: inputting the test data into the trained model to perform flow field reconstruction to obtain predicted data, and determining a first data feature of the predicted data, and determining a second data feature of a label corresponding to the test data; wherein the first data feature and the second data feature are key data features, and the key feature data include at least one of the average flow rate of the cross section, the average temperature of the cross section, the maximum temperature value, the minimum temperature value, the maximum pressure value and the minimum pressure value; the model parameters of the trained model are adjusted according to the first data feature and the second data feature to obtain a flow field reconstruction model.

[0052] The key data features may be parameters that characterize the characteristics of the flow field data. In this application, the accuracy of the predicted data is evaluated by determining the differences between the test data and the predicted data in terms of the key data features.

[0053] Among them, the cross-sectional average flow rate can be the volume or mass of fluid passing through the cross-section per unit time in a given flow field cross-section. The cross-sectional average temperature can be the average value of the fluid temperature in a specific cross-section in the flow field. The maximum temperature value can be the maximum temperature measured or calculated within the flow field study area. The minimum temperature value can be the minimum temperature measured or calculated within the flow field study area. The maximum pressure value can be the maximum pressure occurring in a specific area of the flow field or within the entire study range. The minimum pressure value can be the minimum pressure occurring in a specific area of the flow field or within the entire study range.

[0054] In this application, the difference between the first data feature and the second data feature can be quantified by using distance metrics such as Euclidean distance and Manhattan distance to reflect the degree of deviation between the model prediction result and the actual situation. Based on the calculated difference, the backpropagation algorithm is used to pass the difference information back to each layer of the model, and the contribution of each model parameter to the difference, that is, the gradient, is calculated. Then, based on the gradient information, an appropriate optimization algorithm such as stochastic gradient descent and Adam is used to adjust the model parameters. The direction of adjustment is to make the change in model parameters reduce the difference between the predicted data and the actual data, thereby improving the performance of the model.

[0055] The above technical solution can further optimize the trained model based on test data, so that the model can reconstruct the flow field more accurately and improve the reliability and effectiveness of the model in practical applications.

[0056] An embodiment of the present invention provides a method for training a flow field reconstruction model. The method performs data enhancement processing on the initial flow field based on pre-acquired macro-dimensionalless characteristic parameters of the initial flow field to obtain labeled sample data. The sample data is flow field boundary condition data, and the label is flow field data. The sample data is divided into training data and test data according to a preset ratio. The pre-built model is trained based on the training data, and the trained model is evaluated based on the test data to obtain a flow field reconstruction model. This technical solution, by training the model based on flow field data, can achieve rapid and accurate reconstruction of the flow field, significantly reducing computing time, lowering the requirements for computing equipment, and improving the efficiency of flow field reconstruction.

[0057] Example 2

[0058] Figure 4 This is a flow chart of a training method for a flow field reconstruction model provided in the second embodiment of this application. This embodiment is optimized based on the above embodiment, specifically optimizing the process of data enhancement processing of the flow field. Figure 4 As shown, the method of this embodiment specifically includes the following steps.

[0059] S210 : Determine the amount of flow field data required for the initial flow field based on the macro-dimensionalless characteristic parameters of the initial flow field obtained in advance.

[0060] The amount of flow field data may be the amount of physical quantity data related to the flow field required to be acquired when analyzing flow field characteristics and behaviors, and typically 50-500 sets of data are required.

[0061] For example, the amount of flow field data required for the initial flow field can be determined based on the pre-acquired Reynolds number of the initial flow field. The Reynolds number can be used to distinguish between laminar and turbulent fluid flow. A smaller Reynolds number indicates a more significant effect of viscous forces, while a larger Reynolds number indicates a more significant effect of inertia. Therefore, a larger Reynolds number indicates a greater amount of flow field data required.

[0062] By determining the required amount of flow field data, it is possible to avoid low model prediction accuracy due to insufficient sample data during subsequent model training.

[0063] S220: performing data cleaning processing on the initial flow field to obtain standard data.

[0064] Among them, data cleaning processing can include data missing value processing, outlier processing, data format standardization processing, data consistency processing and data deduplication processing.

[0065] For the processing of missing data values, a comprehensive inspection of the initial flow field data can be performed to determine where in the data set there are missing values; according to the spatial correlation of the flow field data, linear interpolation, spline interpolation and other methods are used to fill in the data.

[0066] For outlier processing, statistical methods such as box plots and Z-score method, or machine learning algorithms such as isolation forest and One-Class SVM can be used to identify outliers in flow field data. Detected outliers can be processed according to the specific situation. If they are caused by measurement errors, the outliers can be deleted or corrected. If they are real anomalies, they can be retained and used for subsequent analysis.

[0067] Data format standardization can include unifying data types and standardizing data units. Unifying data types ensures consistency across all physical quantities in flow field data. For example, standardizing all velocity values to floating-point numbers avoids situations where some velocity values are integers while others are floating-point numbers, ensuring accurate data processing and analysis. Standardizing data units unifies the units of flow field data obtained from different sources or measurement methods. For example, standardizing velocity data to m / s and pressure data to Pa.

[0068] Data consistency can be addressed by determining the consistency of the logical relationships between the various physical quantities in the flow field data. For example, according to the Bernoulli equation, in an ideal flow field, there is a certain relationship between velocity, pressure, and altitude. If the values of these physical quantities in the flow field data do not conform to this relationship, data inconsistency may exist. The correct data can be determined by combining the physical principles of the flow field, measurement methods, and actual conditions, and any erroneous data can be adjusted.

[0069] For data deduplication, the existence of duplicate data can be determined by comparing various fields in the flow field data, such as measurement time, measurement location, and physical quantity values. For the identified duplicate data, one of them is retained and the remaining duplicate records are removed to avoid data redundancy from interfering with subsequent analysis. At the same time, the computational complexity of data processing can also be reduced.

[0070] Through the above data cleaning and processing operations, the quality of the initial flow field data can be effectively improved, and data that meets the standards can be obtained, providing a reliable data basis for subsequent flow field analysis, modeling, and prediction.

[0071] Optionally, the initial flow field is subjected to data cleaning processing to obtain standard data, including: obtaining the residual history curve of the initial flow field, and eliminating the non-converged data in the initial flow field according to the residual history curve to obtain standard data; or, obtaining the flow field data distribution map of the initial flow field, and eliminating abnormal data points in the flow field data distribution map to obtain standard data; or, based on the trilinear interpolation method, interpolating adjacent data points in the initial flow field to obtain standard data.

[0072] The residual history curve represents the variation of the residual error during the iterative calculation of the initial flow field with the number of iterations. As the number of iterations increases, the residual error gradually decreases and approaches a stable value. Therefore, a convergence threshold can be set for the residual history curve in advance. The convergence of the residual history curve can be determined based on the convergence threshold, and non-converged flow field data can be eliminated.

[0073] The flow field data distribution diagram may be visualization data of the distribution of one or more physical quantities in the flow field in space.

[0074] For example, a flow field data distribution graph can display flow field data. For example, a velocity cloud graph uses different colors to represent the velocity distribution in a flow field. In an ideal flow field, velocity changes are continuous and gradual. However, if isolated color blocks appear in the velocity cloud graph, indicating a significant velocity difference from the surrounding area, the data points in that area can be marked as abnormal and removed.

[0075] For another example, the flow field data distribution diagram may be a violin plot, and abnormal data points may be removed by analyzing the data distribution form, quartiles, and median of the violin plot.

[0076] Trilinear interpolation is a method for interpolating flow field data in three-dimensional space. Specifically, the coordinates of missing data points are first determined in the three-dimensional flow field data space. The eight closest data points to the missing data point are selected as neighboring points, with the missing data point as the center. Linear interpolation is then performed in the x-, y-, and z-directions. Finally, the estimated values calculated using trilinear interpolation are used to replace the original data points, completing the data completion.

[0077] By eliminating or interpolating the non-convergent data in the initial flow field through the above three methods, the quality of the flow field data can be effectively improved, and the deviation of the calculation results caused by non-convergent data can be avoided, thereby making the analysis and decision-making based on the flow field data more accurate and reliable.

[0078] S230 : Expand the standard data according to the amount of flow field data to obtain labeled sample data.

[0079] If the amount of flow field data does not meet the requirements, the standard data can be expanded to 3-10 times the original size. It should be noted that if the amount of flow field data meets the requirements, step S230 can be skipped and the process can jump directly to step S240.

[0080] Specifically, data augmentation can be performed based on data enhancement techniques, such as geometric transformation operations or noise addition operations. Generative models can also be used for data augmentation, such as using the generator in a generative adversarial network to generate high-quality flow field data, or generating new flow field data through a variational autoencoder.

[0081] Optionally, the standard data is expanded according to the amount of flow field data to obtain labeled sample data, including: based on the macro-dimensionless characteristic parameter similarity method, the standard data is expanded according to the amount of flow field data to obtain labeled sample data.

[0082] Among them, the macro-dimensional non-dimensional characteristic parameter similarity method uses some important dimensionless parameters in the flow field to determine similar flow field conditions, and then expands and processes the existing standard data to obtain labeled sample data.

[0083] Specifically, based on the demand for flow field data volume and the analysis of the characteristic parameters of the flow field data, the range of macro-dimensionless characteristic parameters of similar working conditions that need to be expanded can be determined, and appropriate parameter values can be selected within this range to construct new flow field working conditions.

[0084] By expanding the standard data based on the macro-dimensionalless characteristic parameter similarity method, it is possible to meet the demand for flow field data volume while ensuring that the expanded data has similar physical properties to the original data, providing a richer and more representative data basis for subsequent flow field analysis and modeling.

[0085] S240: Divide the sample data into training data and test data according to a preset ratio.

[0086] S250: Training the pre-built model based on the training data, and evaluating the trained model based on the test data to obtain a flow field reconstruction model.

[0087] An embodiment of the present invention provides a method for training a flow field reconstruction model. The method determines the amount of flow field data required for the initial flow field based on pre-acquired macro-dimensionalless characteristic parameters of the initial flow field; performs data cleaning on the initial flow field to obtain standard data; expands the standard data based on the amount of flow field data to obtain labeled sample data; divides the sample data into training data and test data according to a preset ratio; trains a pre-built model based on the training data, and evaluates the trained model based on the test data to obtain a flow field reconstruction model. This technical solution, by performing data cleaning and expansion on the initial flow field data, provides a reliable data foundation for subsequent flow field analysis, modeling, and prediction, thereby improving the prediction accuracy of the flow field reconstruction model.

[0088] Example 3

[0089] Figure 5 A schematic diagram of the structure of an apparatus 10 that can be used to implement an embodiment of the present application is shown. The apparatus is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The apparatus can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices (such as helmets, glasses, watches, etc.) and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present application described and / or required herein.

[0090] like Figure 5 As shown, device 10 includes at least one processor 11 and a memory, such as a read-only memory (ROM) 12 and a random access memory (RAM) 13, communicatively connected to the at least one processor 11. The memory stores a computer program executable by the at least one processor, and processor 11 can perform various appropriate actions and processes based on the computer program stored in the read-only memory (ROM) 12 or loaded from storage unit 18 into the random access memory (RAM) 13. RAM 13 can also store various programs and data required for the operation of device 10. Processor 11, ROM 12, and RAM 13 are interconnected via a bus 14. An input / output (I / O) interface 15 is also connected to bus 14.

[0091] Various components in device 10 are connected to I / O interface 15, including an input unit 16, such as a keyboard, mouse, etc.; an output unit 17, such as various types of displays, speakers, etc.; a storage unit 18, such as a magnetic disk, optical disk, etc.; and a communication unit 19, such as a network card, modem, wireless communication transceiver, etc. Communication unit 19 allows device 10 to exchange information / data with other devices via a computer network such as the Internet and / or various telecommunication networks.

[0092] The processor 11 can be any general-purpose and / or specialized processing component with processing and computing capabilities. Some examples of the processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various specialized artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The processor 11 executes the various methods and processes described above, such as the training method for the flow field reconstruction model.

[0093] In some embodiments, the training method for the flow field reconstruction model can be implemented as a computer program, which is tangibly contained in a computer-readable storage medium, such as the storage unit 18. In some embodiments, part or all of the computer program can be loaded and / or installed on the device 10 via the ROM 12 and / or the communication unit 19. When the computer program is loaded into the RAM 13 and executed by the processor 11, one or more steps of the training method for the flow field reconstruction model described above can be performed. Alternatively, in other embodiments, the processor 11 can be configured to execute the training method for the flow field reconstruction model by any other appropriate means (for example, by means of firmware).

[0094] Various embodiments of the systems and techniques described herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), system-on-chip systems (SOCs), programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include being implemented in one or more computer programs that are executable and / or interpreted on a programmable system that includes at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0095] Computer programs for implementing the methods of the present application may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when the computer program is executed by the processor, the functions / operations specified in the flowcharts and / or block diagrams are implemented. The computer program may be executed entirely on the machine, partially on the machine, as a stand-alone software package, partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0096] In the context of the present application, a computer-readable storage medium can be a tangible medium that can contain or store a computer program for use by an instruction execution system, device or equipment or used in combination with an instruction execution system, device or equipment. A computer-readable storage medium can include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared or semiconductor systems, devices or equipment, or any suitable combination of the foregoing. Alternatively, a computer-readable storage medium can be a machine-readable signal medium. A more specific example of a machine-readable storage medium can include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0097] To provide interaction with a user, the systems and techniques described herein can be implemented on a device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the device. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0098] The systems and techniques described herein can be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), a blockchain network, and the Internet.

[0099] A computing system may include clients and servers. The clients and servers are typically remote from each other and typically interact via a communication network. This client-server relationship arises through computer programs running on the respective computers, creating a client-server relationship. The server may be a cloud server, also known as a cloud computing server or cloud host. This server is a hosting product within the cloud computing service ecosystem that addresses the management difficulties and limited scalability of traditional physical hosting and VPS services.

[0100] It should be understood that the various forms of the processes shown above can be used to reorder, add, or delete steps. For example, the steps described in this application can be performed in parallel, sequentially, or in a different order, as long as the desired results of the technical solution of this application can be achieved. This is not limited herein.

[0101] The above specific embodiments do not constitute a limitation on the scope of protection of this application. Those skilled in the art will appreciate that various modifications, combinations, sub-combinations, and substitutions may be made based on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application shall be included within the scope of protection of this application.

Claims

1. A training method for a flow field reconstruction model, characterized in that: The method comprises: According to the macro-dimensionalless characteristic parameters of the initial flow field obtained in advance, data enhancement processing is performed on the initial flow field to obtain sample data with labels; wherein the sample data is flow field boundary condition data, and the label is flow field data; Dividing the sample data into training data and test data according to a preset ratio; The pre-built model is trained based on the training data, and the trained model is evaluated based on the test data to obtain a flow field reconstruction model.

2. The method according to claim 1, characterized in that According to the macro-dimensionalless characteristic parameters of the initial flow field obtained in advance, data enhancement processing is performed on the initial flow field to obtain sample data with labels, including: Determining the amount of flow field data required for the initial flow field based on the macro-dimensionalless characteristic parameters of the initial flow field obtained in advance; Performing data cleaning on the initial flow field to obtain standard data; According to the amount of flow field data, the standard data is expanded to obtain sample data with labels.

3. The method according to claim 2, characterized in that The initial flow field is subjected to data cleaning processing to obtain standard data, including: Obtaining a residual history curve of the initial flow field, and eliminating non-convergent data in the initial flow field according to the residual history curve to obtain standard data; Alternatively, a flow field data distribution diagram of the initial flow field is obtained, and abnormal data points in the flow field data distribution diagram are eliminated to obtain standard data; Alternatively, based on the trilinear interpolation method, interpolation processing is performed on adjacent data points in the initial flow field to obtain standard data.

4. The method according to claim 2, characterized in that According to the amount of flow field data, the standard data is expanded to obtain sample data with labels, including: Based on the macro-dimensionalless characteristic parameter similarity method, the standard data is expanded according to the amount of flow field data to obtain sample data with labels.

5. The method according to claim 1, wherein The pre-built model is a generative adversarial network model, which includes a generator and a discriminator. Accordingly, the pre-built model is trained based on the training data, including: Inputting the training data into the generator to generate fake data; Inputting the training data and the fake data into the discriminator, calculating a loss function according to an output result of the discriminator, and training a pre-built model based on the loss function; Alternatively, the pre-built model is a feedforward neural network model, which includes an input layer, a hidden layer, and an output layer; accordingly, training the pre-built model based on the training data includes: Inputting the training data into the input layer and sequentially passing through each of the hidden layers and the output layer to obtain a predicted output; Determining a learning error of the feedforward neural network model based on the predicted output and the training data; The model parameters of the feedforward neural network model are updated according to the learning error.

6. The method according to claim 1, characterized in that The trained model is evaluated based on the test data to obtain a flow field reconstruction model, including: Inputting the test data into the trained model to reconstruct the flow field, obtaining predicted data, and determining a first data feature of the predicted data, and determining a second data feature corresponding to a label of the test data; wherein the first data feature and the second data feature are key data features, and the key feature data include at least one of a cross-sectional average flow rate, a cross-sectional average temperature, a maximum temperature value, a minimum temperature value, a maximum pressure value, and a minimum pressure value; The model parameters of the trained model are adjusted according to the first data feature and the second data feature to obtain a flow field reconstruction model.

7. The method according to claim 1, characterized in that The boundary condition data includes at least one of an inlet flow rate, an inlet temperature, and a heating flux; The flow field data includes one of velocity, temperature and pressure.

8. An electronic device, characterized in that: The device comprises: At least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the training method for the flow field reconstruction model according to any one of claims 1 to 7.

9. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the training method for the flow field reconstruction model according to any one of claims 1 to 7 when executed.

10. A computer program product, characterized in that The computer program product comprises a computer program, which, when executed by a processor, implements the method for training a flow field reconstruction model according to any one of claims 1 to 7.

Images