Nuclear power plant flow field and material field rapid prediction method and system based on machine learning
By combining POD decomposition technology, fully connected artificial neural network technology and graph convolutional neural network technology, a variety of machine learning algorithm models are built, which solves the problem of rapid prediction of flow and material fields under variable geometric conditions of nuclear power plants, and achieves efficient and fast prediction effects.
Patent Information
- Application Number
- CN202411991910.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-31
- Publication Date
- 2025-04-18
AI Technical Summary
The prior art is difficult to quickly predict the flow field and material field under the geometric changing conditions of nuclear power plants while ensuring accuracy, resulting in low computational efficiency.
POD decomposition technology, fully connected artificial neural network technology and graph convolutional neural network technology are used, and a variety of machine learning algorithms are combined to build flow field and matter field prediction models to predict different geometric changes.
It realizes rapid prediction of the flow and material fields of the nuclear power plant within 2 minutes, improves calculation efficiency and real-time performance, and ensures prediction accuracy under variable geometric conditions.
Smart Images
Figure CN119848462A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of fluid mechanics, and in particular relates to a method and system for rapid prediction of flow fields and material fields in a nuclear power plant based on machine learning. Background Art
[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.
[0003] Computational fluid dynamics (CFD) simulation technology is a method of applying computers to fluid mechanics solutions. By solving various equations related to flow (continuity equation, momentum equation, mass conservation equation, diffusion equation, etc.), it can not only accurately obtain the structural characteristics of plant structures, but also simulate their impact on the surrounding flow field and airborne radioactive nuclides, thereby simulating flow fields and concentration fields that are close to the actual situation.
[0004] However, conventional numerical simulation methods require a lot of time to advance the solution. If the grid size is huge, the calculation time is often very long. Therefore, how to improve the efficiency of numerical simulation while ensuring accuracy has become another difficult problem.
[0005] Due to the explosive development of artificial intelligence technology in image processing, pattern recognition, natural language processing and other fields, more and more researchers have begun to explore the application of machine learning technology in CFD simulation. The most common application is to use machine learning methods to train a large amount of flow field data to build a flow field prediction model, thereby replacing traditional CFD simulation and significantly reducing computing time and resource consumption in practical applications.
[0006] At the same time, the data-driven neural network model also obtains highly reliable flow field sample data based on the existing flow field simulation method, and then uses a machine learning model to learn the mapping relationship between input parameters, such as operating parameters, and the numerical solution of the flow field, replacing the traditional time-consuming discrete solution process of partial differential equations, thereby quickly and efficiently obtaining the numerical solution of the flow field.
[0007] However, although such methods are widely used in many fields such as wing airfoil flow field prediction and wind tunnel control prediction, they are mainly limited to dealing with flow field prediction problems under fixed working conditions. For nuclear power plants with variable working conditions or geometric changes, such as movement, change, or increase or decrease in quantity, there are still great challenges. It is impossible to achieve rapid prediction of flow and material fields while ensuring accuracy under variable geometric conditions, and improve the efficiency of numerical simulation calculations. Summary of the invention
[0008] In order to overcome the shortcomings of the above-mentioned prior art, the present invention provides a method and system for rapid prediction of flow field and material field of nuclear power plants based on machine learning. POD decomposition technology and fully connected artificial neural network technology are used for fixed geometry and variable operating conditions of nuclear power plants, and graph convolutional neural network technology is used for variable geometry and variable operating conditions of nuclear power plants. It combines the characteristics of variable geometry and variable operating conditions problems and uses a variety of machine learning algorithm models to realize rapid prediction of flow field and material field of nuclear power plants, which provides an effective solution to the current problem of variable geometry of nuclear power plant structures and can also improve the accuracy of environmental impact assessment and accident emergency decision-making of nuclear power plants.
[0009] To achieve the above objectives, one or more embodiments of the present invention provide the following technical solutions: The first aspect of the present invention provides a method for rapid prediction of flow field and material field in a nuclear power plant based on machine learning.
[0010] A method for rapid prediction of flow and material fields in nuclear power plants based on machine learning, comprising: Obtain the sampling space of the building geometry and operating parameters of the nuclear power plant site; Through the sampling algorithm, the flow field and material field data in the sampling space are collected and preprocessed to obtain a training data set based on the geometry and operating parameters of the nuclear power plant site; Obtain the geometric variation characteristics of buildings at nuclear power plant sites; According to different geometric change characteristics, different machine learning algorithms are used to build flow field and material field prediction models; Inputting the training data set based on the geometry and operating parameters of the nuclear power plant site into the flow field and material field prediction model to obtain the flow field and material field prediction results; The flow field and material field prediction results are visualized, and various mathematical statistics are set to compare the prediction results.
[0011] A second aspect of the present invention provides a rapid prediction system for flow fields and material fields in a nuclear power plant based on machine learning.
[0012] A rapid prediction system for flow and material fields in nuclear power plants based on machine learning, including: The data acquisition module is configured to acquire the sampling space of the building geometry and operating condition parameters of the nuclear power plant site; Through the sampling algorithm, the flow field and material field data in the sampling space are collected and preprocessed to obtain a training data set based on the geometry and operating parameters of the nuclear power plant site; The model building module is configured to obtain geometric variation characteristics of buildings at the nuclear power plant site; According to different geometric change characteristics, different machine learning algorithms are used to build flow field and material field prediction models; A fast prediction module is configured to input the training data set based on the geometry and operating parameters of the nuclear power plant site into the flow field and material field prediction model to obtain flow field and material field prediction results; The visualization module is configured to visualize the flow field and material field prediction results and set a variety of mathematical statistics for data comparison of the prediction results.
[0013] The third aspect of the present invention provides 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 steps in a method as described in the first aspect of the present invention are implemented.
[0014] A fourth aspect of the present invention provides a computer-readable storage medium having a program stored thereon, which, when executed by a processor, implements the steps in a method as described in the first aspect of the present invention.
[0015] A fifth aspect of the present invention provides a computer program product comprising instructions, which, when running on a computer, enables the computer program to implement the steps in a method as described in the first aspect of the present invention when executed by a processor.
[0016] One or more of the above technical solutions have the following beneficial effects: The present invention combines the characteristics of variable geometry and variable working condition problems, integrates three machine learning algorithms, utilizes multiple machine learning algorithm models, and simultaneously considers the influence of geometric parameters and meteorological changes, and can achieve rapid prediction of flow field and material field within 2 minutes.
[0017] The present invention integrates multiple sampling algorithms and constructs flow field and material field training data, which can reasonably reduce the sample size and realize automatic sampling of different sample sizes.
[0018] Compared with traditional numerical simulation methods, the present invention reduces the consumption of computing resources and time during online prediction, and improves the efficiency and real-time performance of prediction.
[0019] The present invention can quickly compare the calculation results of fluid mechanics software with the predicted flow field and material field data by visualizing the flow field and material field results and setting a variety of mathematical statistics for data comparison of the predicted results.
[0020] Advantages of additional aspects of the present invention will be given in part in the following description, and in part will become obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS
[0021] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.
[0022] Figure 1 A flow chart of a method for rapid prediction of flow field and material field in a nuclear power plant based on machine learning provided in Embodiment 1 of the present invention; Figure 2 A schematic diagram of sampling settings provided in Embodiment 1 of the present invention; Figure 3 This is a schematic diagram of a machine learning module-fixed geometry prediction in accordance with the first embodiment of the present invention; Figure 4 A schematic diagram of a machine learning module-variable geometry prediction provided in the first embodiment of the present invention; Figure 5 A schematic diagram comparing the CFD flow field and the flow field predicted by the graph convolutional neural network model provided in Example 1 of the present invention; Figure 6 Schematic diagram for comparing the CFD material field and the material field predicted by the graph convolutional neural network model provided in Example 1 of the present invention. DETAILED DESCRIPTION
[0023] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.
[0024] It should be noted that the terms used herein are for describing specific embodiments only and are not intended to be limiting of exemplary embodiments according to the present invention.
[0025] In the absence of conflict, the embodiments of the present invention and the features of the embodiments may be combined with each other.
[0026] Embodiment 1 This embodiment discloses a method for rapid prediction of flow field and material field of a nuclear power plant based on machine learning. Based on grid topology connection and combining the characteristics of variable geometry and variable working condition problems, a variety of machine learning algorithm models are used to achieve rapid prediction of flow field of a nuclear power plant, such as Figure 1 As shown, including: Step 1: Obtain the sampling space of the building geometry and operating parameters of the nuclear power plant site; Step 2: Using a sampling algorithm, collect the flow field and material field data in the sampling space, and perform preprocessing to obtain a training data set based on the geometry and operating parameters of the nuclear power plant site; Step 3: Obtain the geometric change characteristics of the nuclear power plant site building; according to different geometric change characteristics, use different machine learning algorithms to build flow field and material field prediction models; Step 4: Input the training data set based on the geometry and operating parameters of the nuclear power plant site into the flow field and material field prediction model to obtain the flow field and material field prediction results; Step 5: Visualize the flow field and material field prediction results, and set a variety of mathematical statistics for data comparison of the prediction results.
[0027] Numerical simulation methods have become an important method that can be mutually verified with wind tunnel tests and have been widely used in scheme selection, planning and design. Although computer performance has been greatly improved, conventional numerical simulation methods still require a lot of time to advance and solve. If the grid volume is huge, the calculation time is often very long. How to further improve the efficiency of numerical simulation calculations while ensuring accuracy is a difficult problem faced by current engineering applications. Therefore, this embodiment is a method for rapid prediction of flow fields under variable geometry and variable working conditions based on grid topology connections.
[0028] In order to more clearly illustrate this embodiment, the implementation process of the method for rapid prediction of flow field and material field in a nuclear power plant based on machine learning can be specifically described as follows: Step 1: Obtain the sampling space of the building geometry and operating parameters of the nuclear power plant site.
[0029] In this embodiment, simulation software is called in a batch processing manner through simulation optimization technology to obtain the sampling space of the building geometry and operating parameters of the nuclear power plant site for specific fluid mechanics problems, such as fixed geometry or variable geometry conditions.
[0030] Specifically, it means first obtaining the calculation boundary conditions under certain geometric conditions based on different terrains, meteorological conditions, and combinations of artificial structures.
[0031] Among them, meteorological conditions include 5 types of wind speeds, 16 azimuth wind directions, and 6 types of stability; terrain conditions should include 2 types: simple terrain (flat terrain can be used) and complex terrain (actual nuclear power plant terrain can be used); artificial structures adopt the actual nuclear power plant general layout.
[0032] Step 2: Through the sampling algorithm, the flow field and material field data in the sampling space are collected and preprocessed to obtain a training data set based on the geometry and operating parameters of the nuclear power plant site.
[0033] In this embodiment, through a variety of intelligent sampling algorithms, vector and scalar field data such as flow field and material field can be simulated under different terrains, meteorological conditions, and artificial structure combinations. Through data interface development, the training data set required for machine learning is extracted and transformed, that is, the training data set is obtained based on the geometric change characteristics of the nuclear power plant site buildings and the operating parameters.
[0034] Specific data such as Figure 2As shown, it mainly includes speed, direction and stability, among which speed and direction are continuous variables, and stability is a discrete variable.
[0035] For continuous variables such as speed and direction, we first need to determine the upper and lower limits and the number of levels of the sample variable. The upper and lower limits represent the sampling range of the variable, and the number of levels represents the number of samples of the variable.
[0036] Stability is a discrete variable. There are six types of stability: A, B, C, D, E, and F. In a prediction, you can select all stability parameters at once using the Select All button.
[0037] The various intelligent sampling algorithms used, including uniform sampling, partial factor sampling, and Latin hypercube, can reasonably reduce the sample size and realize automatic sampling of different sample sizes. The parameter range covered by the sampled data set must be as large as possible and cover the parameter space to be studied as evenly as possible to ensure the effectiveness of subsequent model training.
[0038] In this embodiment, preprocessing mainly includes checking data integrity and rationality, processing abnormal or unreasonable data points in flow field or material field data, performing data augmentation operations to improve data validity, and standardizing or normalizing flow field or material field data to ensure data comparability.
[0039] Step 3: Obtain the geometric change characteristics of the nuclear power plant site buildings; according to different geometric change characteristics, use different machine learning algorithms to construct flow field and material field prediction models.
[0040] In this embodiment, the simulation technology of steps 1 and 2 is used to obtain a training data set based on the geometry and operating parameters of the nuclear power plant site, such as Figure 3 and Figure 4 As shown in the figure, after determining the training data set, it is necessary to select different machine learning algorithms to build a prediction model based on different geometric change characteristics.
[0041] The geometric variation feature refers to whether the type of building at the nuclear power plant site to be tested is fixed geometry or variable geometry. Corresponding prediction models are constructed for different geometric situations.
[0042] In this embodiment, different machine learning algorithms are used, namely, prediction models of POD decomposition technology, fully connected neural network technology, and graph convolutional neural network technology.
[0043] (1) POD decomposition technology refers to obtaining the optimal standard orthogonal basis of the sampled flow field information (assuming the order is m, where m is much smaller than M), and then projecting each flow field information in the sampled flow field onto this set of standard orthogonal basis to obtain the projection coefficient of each flow field on this set of standard orthogonal basis.
[0044] The interpolation object of the POD decomposition technology is converted from the information on the grid points of the flow field data set to the projection coefficients of the flow field on each standard orthogonal basis, reducing the M-order interpolation to the m-order interpolation. Converting the interpolation of the flow field point set to the interpolation of the projection coefficients of the flow field on the orthogonal basis can greatly reduce the amount of calculation in the interpolation process.
[0045] (2) Fully connected artificial neural network technology refers to representing the three-dimensional physical field by a one-dimensional vector. That is, the physical field solution vector is sorted according to the grid points to obtain a one-dimensional vector. The physical field prediction task is transformed into: predicting the physical field solution vector by inputting a parameter vector.
[0046] (3) Graph convolutional neural network technology converts the above dataset into a storage structure based on a grid topology connection graph, using nodes and edges to represent spatial connection relationships, and features on nodes and edges to represent flow field information. By performing convolution operations on the nodes around each grid node, local feature information is extracted and passed to the global layer for final prediction.
[0047] In this embodiment, a machine learning algorithm corresponding to the geometric change characteristics of the nuclear power plant is used to quickly obtain the predicted flow field and material field data, as follows: POD decomposition technology and fully connected artificial neural network technology are suitable for fixed geometry prediction when the input and output dimensions are fixed, that is, when the grid is determined.
[0048] Graph convolutional neural networks are suitable for variable geometry prediction for cases with variable geometry characteristics, that is, the number and location distribution of cooling towers in the case will change.
[0049] In this embodiment, when constructing a prediction model, it is also necessary to define an appropriate loss function to measure the error between the prediction result and the actual simulation data. For general flow field or material field prediction tasks, commonly used loss functions include root mean square error and absolute value error.
[0050] The model is trained using the training set data, the loss function between the model prediction and the target data is calculated, the gradient of the network parameters is obtained through back propagation of the loss, and the parameters are updated through gradient descent, so that the prediction results gradually approach the real data.
[0051] The model parameters are optimized by minimizing the error loss function between the model's prediction and the real data to obtain an optimized machine learning model.
[0052] Step 4: Input the training data set based on the geometry and operating parameters of the nuclear power plant site into the flow field and material field prediction model to obtain the flow field and material field prediction results.
[0053] In this embodiment, the prediction results obtained include variables corresponding to the flow field and the material field, wind velocity and pollutant concentration, etc., which can also be collectively referred to as variable fields.
[0054] In this embodiment, the brief algorithm principles of the three model architectures are as follows: 1.POD decomposition technology model architecture: POD decomposition technology is used to first reduce the dimensionality of the training data set, and then a fully connected neural network is used to predict the coefficients of the POD main mode to achieve the prediction of high-dimensional variable fields.
[0055] 2. Fully connected artificial neural network technology model architecture: The training data set is converted into high-dimensional vector variables according to the spatial grid position sorting, and then the fully connected neural network is used to directly predict the high-dimensional variable field.
[0056] 3. Graph neural network model architecture: The training data set is first converted into node features on the topological graph according to the topological connection relationship of the spatial grid, and then the graph convolutional network is used to extract information such as geometry and input parameters, and finally the variable field on the grid points is predicted.
[0057] Specifically, in this embodiment, taking the graph neural network model as an example, a rapid prediction process of the flow field under variable geometry and variable working conditions is provided: Firstly, according to the variable geometry and operating condition parameters of the nuclear power plant, where the variable geometry refers to the geometric change characteristics of the nuclear power plant and the variable operating condition refers to the changes in wind direction, wind speed and stability, the sampling space is set and a large number of nuclear power plant flow field data sets are constructed based on fluid simulation software.
[0058] Secondly, based on the variable geometry characteristics of nuclear power plants, the graph neural network model algorithm is selected to build a prediction model to train the flow field dataset.
[0059] Finally, the corresponding numerical solution of the flow field is obtained, including flow velocity, flow direction, and statistical comparison between the predicted field and the calculation results of the fluid mechanics software.
[0060] like Figure 5 , 6 As shown in the figure, the rapid prediction results of the flow field and material field near the nuclear power plant building are shown. The results show that the flow field and material field predicted by the graph neural network model are very close to the flow field and material field simulated by CFD, achieving a good prediction effect.
[0061] As shown in Table 1, a comparison of different machine learning model architectures is provided, showing the model complexity, size, and training efficiency of the model architecture using different machine learning algorithms. It can also be seen that by using different machine learning algorithms to build a prediction model, the situation of variable geometry and variable working conditions can be better considered.
[0062] Table 1 Comparison of different machine learning model architectures
[0063] Step 5: Visualize the flow field and material field prediction results, and set a variety of mathematical statistics to compare the prediction results.
[0064] In this embodiment, based on Python scripts, the simulation and rapid calculation of the flow field and material field of a nuclear power plant are realized, and the visualization analysis of the prediction results is realized by creating geometric, line graphs, vector graphs and contour graphs. Compared with traditional numerical simulation methods, it can greatly reduce the consumption of computing resources and time, and improve the efficiency and real-time performance of flow field prediction.
[0065] By setting a variety of mathematical statistics, such as maximum and minimum values, root mean square error, etc., the calculation results of fluid mechanics software can be quickly compared with the predicted flow field and material field data.
[0066] For the technical methods described in steps 1-5, the various functions of the modules including simulation object input, simulation program call, data sampling, machine learning, data post-processing and related training sets, simulation results, post-processing data, scripts, batch processing, etc. are integrated.
[0067] Through simulation optimization technology, simulation software is called in batch mode. Through sampling algorithms, vector and scalar field data such as flow field and material field can be obtained through simulation under different terrains, meteorological conditions, and combinations of artificial structures. Through data interface development, the training data set required for machine learning is extracted and transformed. Based on the training data set, the corresponding machine learning algorithms are applied to the characteristics of the flow field and material field, and the machine learning model is trained and optimized. By inputting different terrains, meteorological conditions, and combinations of artificial structures, the flow field and material field data can be quickly obtained. Ultimately, the simulation calculation and data post-processing of the flow field and material field of the nuclear power plant to be tested can be realized, and the result data obtained by the machine learning model can be compared, and the solution accuracy and efficiency of the machine learning model can be verified.
[0068] Embodiment 2 The purpose of this embodiment is to provide a rapid prediction system for flow field and material field of a nuclear power plant based on machine learning, including: The data acquisition module is configured to acquire the sampling space of the building geometry and operating condition parameters of the nuclear power plant site; Through the sampling algorithm, the flow field and material field data in the sampling space are collected and preprocessed to obtain a training data set based on the geometry and operating parameters of the nuclear power plant site; The model building module is configured to obtain geometric variation characteristics of buildings at the nuclear power plant site; According to different geometric change characteristics, different machine learning algorithms are used to build flow field and material field prediction models; A fast prediction module is configured to input the training data set based on the geometry and operating parameters of the nuclear power plant site into the flow field and material field prediction model to obtain flow field and material field prediction results; The visualization module is configured to visualize the flow field and material field prediction results and set a variety of mathematical statistics for data comparison of the prediction results.
[0069] In this embodiment, various functions of modules including simulation object input, simulation program calling, data sampling, machine learning, data post-processing and related training sets, simulation results, post-processing data, scripts, batch processing, etc. are integrated.
[0070] Embodiment 3 The purpose of this embodiment is to provide a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above method when executing the program.
[0071] Embodiment 4 The purpose of this embodiment is to provide a computer-readable storage medium.
[0072] A computer-readable storage medium stores a computer program, which executes the steps of the above method when executed by a processor.
[0073] Embodiment 5 The purpose of this embodiment is to provide a computer program product containing instructions, which, when running on a computer, enables the computer to execute the methods and functions involved in any of the above embodiments. The steps involved in the apparatus of the above embodiment correspond to the method embodiment 1, and the specific implementation method can refer to the relevant description part of embodiment 1. The term "computer-readable storage medium" should be understood as a single medium or multiple media including one or more instruction sets; it should also be understood to include any medium that can store, encode or carry an instruction set for execution by a processor and enable the processor to execute any method in the present invention.
[0074] Those skilled in the art should understand that the modules or steps of the present invention described above can be implemented by a general-purpose computer device, or alternatively, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. The present invention is not limited to any specific combination of hardware and software.
[0075] Although the above describes the specific implementation mode of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art on the basis of the technical solution of the present invention without creative work are still within the scope of protection of the present invention.
Claims
1. A method for rapid prediction of flow field and material field in a nuclear power plant based on machine learning, characterized in that: include: Obtain the sampling space of the building geometry and operating parameters of the nuclear power plant site; Through the sampling algorithm, the flow field and material field data in the sampling space are collected and preprocessed to obtain a training data set based on the geometry and operating parameters of the nuclear power plant site; Obtain the geometric variation characteristics of buildings at nuclear power plant sites; According to different geometric change characteristics, different machine learning algorithms are used to build flow field and material field prediction models; Inputting the training data set based on the geometry and operating parameters of the nuclear power plant site into the flow field and material field prediction model to obtain the flow field and material field prediction results; The flow field and material field prediction results are visualized, and various mathematical statistics are set to compare the prediction results.
2. A method for rapid prediction of flow field and material field in a nuclear power plant based on machine learning as claimed in claim 1, characterized in that: The sampling space for obtaining the geometry and operating parameters of the nuclear power plant site building is specifically to call the simulation software in a batch processing manner through simulation optimization technology to obtain the calculation boundary conditions under the given geometry conditions of the nuclear power plant site building based on different terrains, meteorological conditions, and combinations of artificial structures.
3. A method for rapid prediction of flow field and material field in a nuclear power plant based on machine learning as claimed in claim 1, characterized in that: The flow field and material field data in the sampling space are collected by the sampling algorithm, specifically, by adopting a variety of intelligent sampling algorithms, the vector and scalar field data of the flow field and material field are obtained by simulation based on different terrains, meteorological conditions, and combinations of artificial structures; Vector and scalar field data specifically include: speed, direction and stability; among which speed and direction are continuous variables, and stability is a discrete variable.
4. A method for rapid prediction of flow field and material field in a nuclear power plant based on machine learning as claimed in claim 1, characterized in that: According to different geometric change characteristics, different machine learning algorithms are used to construct flow field and material field prediction models, specifically including: For the flow field under fixed geometry and working conditions, that is, when the input and output dimensions are fixed and the grid is determined, the POD decomposition technology and fully connected artificial neural network technology are used to construct the flow field and material field prediction model for prediction; The specific process of using POD decomposition technology to construct the flow field and material field prediction model for prediction is as follows: first, the training data set is reduced in dimension, and then the coefficients of the POD main mode are predicted using a fully connected neural network to achieve the prediction of the high-dimensional variable field; The specific process of using fully connected artificial neural network technology to construct a flow field and material field prediction model for prediction is: converting the training data set into high-dimensional vector variables according to the spatial grid position sorting, and then using the fully connected neural network to directly predict the high-dimensional variable field.
5. A method for rapid prediction of flow field and material field in a nuclear power plant based on machine learning as claimed in claim 1, characterized in that: The method uses different machine learning algorithms to construct flow field and material field prediction models for different geometric change characteristics, and also includes: for flow fields with variable geometry and variable working conditions, using a graph convolutional neural network model to construct flow field and material field prediction models for prediction; The specific process is as follows: first, the training data set is converted into node features on a topological graph according to the topological connection relationship of the spatial grid, then the graph convolutional network is used to extract the geometry and input parameter information, and finally the variable field on the grid points is predicted.
6. A method for rapid prediction of flow field and material field in a nuclear power plant based on machine learning as claimed in claim 1, characterized in that: The flow field and material field prediction results are visualized, specifically: by creating geometry, line graphs, vector graphs and contour graphs, to achieve visual analysis of the prediction results; The method of setting a plurality of mathematical statistics for data comparison of prediction results is specifically to quickly compare the calculation results of the fluid mechanics software with the predicted flow field and material field data by setting maximum and minimum values and root mean square errors.
7. A rapid prediction system for flow and material fields in nuclear power plants based on machine learning, characterized in that: include: The data acquisition module is configured to acquire the sampling space of the building geometry and operating condition parameters of the nuclear power plant site; Through the sampling algorithm, the flow field and material field data in the sampling space are collected and preprocessed to obtain a training data set based on the geometry and operating parameters of the nuclear power plant site; The model building module is configured to obtain geometric variation characteristics of buildings at the nuclear power plant site; According to different geometric change characteristics, different machine learning algorithms are used to build flow field and material field prediction models; A fast prediction module is configured to input the training data set based on the geometry and operating parameters of the nuclear power plant site into the flow field and material field prediction model to obtain flow field and material field prediction results; The visualization module is configured to visualize the flow field and material field prediction results and set various mathematical statistics for data comparison of the prediction results.
8. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps of the method described in any one of claims 1 to 6 are implemented.
9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method described in any one of claims 1 to 6 are performed.
10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.