A multi-fidelity fluid data fusion simulation modeling method, device and equipment

Abstract

The present invention relates to a multi-fidelity fluid data fusion simulation modeling method, device and equipment. The method comprises: obtaining the factory fluid data of equipment parts and the fluid data of equipment parts in a simulation environment stored in a multi-fidelity file. The two data are sampled respectively using the Latin hypercube algorithm, and low-fidelity wall grid data and high-fidelity wall grid data are output, and a kernel coupling matrix is ​​constructed based on the two grid data. The multi-fidelity data obtained by merging the two grid data is used as a sample data set and input into a multi-fidelity Gaussian regression simulation model constructed based on the kernel coupling matrix. The kernel coupling matrix is ​​solved using a radial basis function kernel function to obtain a sample data set to be evaluated. After the sample data set to be evaluated is verified by a preset error evaluation index, real fluid data is obtained. The use of this method can improve the production efficiency of high-performance equipment parts.

Description

Technical Field

[0001] The present invention relates to the technical field of simulation testing of high-performance equipment components, and in particular to a multi-fidelity fluid data fusion simulation modeling method, device and equipment. Background Art

[0002] Computational fluid dynamics (CFD) simulation has become a key technology for optimizing design and predicting performance in the R&D and production of high-performance equipment components. For precision components like aircraft engine blades and turbochargers for new energy vehicles, for example, fluid dynamics performance directly determines their operational efficiency and reliability, while mesh fineness has a decisive impact on the accuracy of simulation results and computing resource consumption. While fine-mesh models enable highly accurate flow field analysis, they incur significant computational costs and lengthy simulation cycles, leading to inefficient iterative component design. Coarse-mesh models, while capable of rapid result output, struggle to meet the stringent simulation accuracy requirements of high-performance equipment.

[0003] Currently, the industry generally uses traditional methods such as interpolation and response surface models to fuse low-fidelity and high-fidelity data in an attempt to balance accuracy and efficiency. However, in the complex flow field simulation of high-performance equipment components, these methods have exposed significant drawbacks: First, a single data fusion strategy cannot accurately capture the complex mapping relationship of fluid parameters at different grid scales. Especially when it comes to key physical phenomena such as turbulence and boundary layers, the accumulated errors seriously affect design optimization; second, traditional proxy modeling requires the collection of massive sample data, resulting in a surge in computing resource consumption and difficulty in adapting to the high-dimensional design space of multi-parameter coupling of components; third, existing error correction technology fails to fully explore the correlation between high-fidelity and low-fidelity data. In the multi-dimensional simulation of complex physical fields, the dynamic adaptability and prediction accuracy of the model cannot meet the rapid iteration requirements of high-performance equipment. Therefore, there is an urgent need for a technology that can break through the dual bottlenecks of production efficiency and simulation accuracy of high-performance equipment components. Summary of the Invention

[0004] Based on this, it is necessary to provide a multi-fidelity fluid data fusion simulation modeling method, device and equipment that can improve the production efficiency of high-performance equipment parts in response to the above technical problems.

[0005] A multi-fidelity fluid data fusion simulation modeling method, the method comprising:

[0006] Obtain the factory fluid data of a certain equipment component stored in a multi-fidelity file and the fluid data used by the equipment component in a simulation environment.

[0007] The Latin hypercube algorithm is used to sample the factory fluid data and the used fluid data respectively, and the low-fidelity wall grid data and the high-fidelity wall grid data are output. The kernel coupling matrix is ​​constructed according to the low-fidelity wall grid data and the high-fidelity wall grid data.

[0008] The multi-fidelity data obtained by merging the low-fidelity wall mesh data and the high-fidelity wall mesh data are used as the sample data set and input into the multi-fidelity Gaussian regression simulation model constructed according to the kernel coupling matrix. The radial basis function kernel function is used to solve the kernel coupling matrix to obtain the sample data set to be evaluated.

[0009] After the sample data set to be evaluated is verified by the preset error evaluation index, the real fluid data is obtained.

[0010] A multi-fidelity fluid data fusion simulation modeling device, comprising:

[0011] The data acquisition module is used to obtain the factory fluid data of equipment parts stored in multi-fidelity files and the fluid data of equipment parts in use under a simulation environment.

[0012] The inner coupling matrix construction module is used to sample the factory fluid data and the used fluid data respectively using the Latin hypercube algorithm, output low-fidelity wall grid data and high-fidelity wall grid data, and construct the kernel coupling matrix according to the low-fidelity wall grid data and the high-fidelity wall grid data.

[0013] The multi-fidelity data training module is used to input the multi-fidelity data obtained by merging low-fidelity wall mesh data with high-fidelity wall mesh data as a sample data set into a multi-fidelity Gaussian regression simulation model constructed based on the kernel coupling matrix, and use the radial basis function kernel function to solve the kernel coupling matrix to obtain the sample data set to be evaluated.

[0014] The evaluation module is used to obtain real fluid data after verifying the sample data set to be evaluated through preset error evaluation indicators.

[0015] A computer device includes a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0016] Obtain the factory fluid data of equipment parts stored in multi-fidelity files and the fluid data used by equipment parts in the simulation environment.

[0017] The Latin hypercube algorithm is used to sample the factory fluid data and the used fluid data respectively, and the low-fidelity wall grid data and the high-fidelity wall grid data are output. The kernel coupling matrix is ​​constructed according to the low-fidelity wall grid data and the high-fidelity wall grid data.

[0018] The multi-fidelity data obtained by merging the low-fidelity wall mesh data and the high-fidelity wall mesh data are used as the sample data set and input into the multi-fidelity Gaussian regression simulation model constructed according to the kernel coupling matrix. The radial basis function kernel function is used to solve the kernel coupling matrix to obtain the sample data set to be evaluated.

[0019] After the sample data set to be evaluated is verified by the preset error evaluation index, the real fluid data is obtained.

[0020] The multi-fidelity fluid data fusion simulation modeling method, apparatus, and equipment described above systematically address the efficiency and reliability bottlenecks of traditional Gaussian process models in complex physical field applications through multi-fidelity data fusion and adaptive modeling mechanisms. To address computational efficiency issues, a Latin hypercube sampling algorithm is used to collaboratively sample fluid data from both the factory and the user environment, optimizing the sample structure while ensuring data representation capabilities. A kernel coupling matrix constructed from multi-fidelity wall mesh data decomposes fidelity-dependent and fidelity-independent parallel computational modules, incorporating sparse approximation techniques to significantly reduce the computational complexity of the global kernel function and achieve real-time simulation response. To address uncertainty quantification errors, a fidelity-sensitive error propagation model is designed. Adaptive kernel bandwidth adjustment is used to capture the anisotropic characteristics of flow field parameters. Tensor decomposition techniques are introduced to quantify cross-scale modeling deviations. A Bayesian evidence framework is then used to dynamically optimize model parameters, improving uncertainty quantification accuracy. The error source analysis module within the multi-level verification mechanism identifies the primary contributors to model mismatch, guiding the iterative optimization of the active learning strategy to maintain the confidence of prediction results within engineering tolerances. This technology system uses a data-driven multi-fidelity fusion architecture to improve modeling efficiency while ensuring accuracy, support multidisciplinary optimization of complex aerodynamic shapes with significantly shortened cycle time, and provide an efficient and reliable new methodology for agile design and online verification of equipment components. It effectively resolves the contradiction between high-fidelity simulation and engineering practicality, and meets the dual requirements of modern manufacturing for production efficiency and quality control. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 Schematic diagram of a flow chart of a multi-fidelity fluid data fusion simulation modeling method in one embodiment;

[0022] Figure 2 A schematic diagram of a multi-fidelity modeling process flow in one embodiment;

[0023] Figure 3 It is a structural block diagram of a multi-fidelity fluid data fusion simulation modeling device in one embodiment;

[0024] Figure 4 FIG. 1 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0025] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0026] In one embodiment, Figure 2 As shown, a multi-fidelity fluid data fusion simulation modeling method is provided, comprising the following steps:

[0027] Step 102 : Obtain the factory fluid data of the equipment parts and the fluid data of the equipment parts in the simulation environment stored in the multi-fidelity file.

[0028] Step 104 : Using the Latin hypercube algorithm to sample the factory fluid data and the used fluid data, respectively, output low-fidelity wall grid data and high-fidelity wall grid data, and construct a kernel coupling matrix based on the low-fidelity wall grid data and the high-fidelity wall grid data.

[0029] In step 106, the multi-fidelity data obtained by merging the low-fidelity wall mesh data and the high-fidelity wall mesh data is used as a sample data set and input into a multi-fidelity Gaussian regression simulation model constructed according to the kernel coupling matrix. The kernel coupling matrix is ​​solved using a radial basis function kernel function to obtain a sample data set to be evaluated.

[0030] In step 108 , the sample data set to be evaluated is verified using a preset error evaluation index to obtain real fluid data.

[0031] The aforementioned multi-fidelity fluid data fusion simulation modeling method systematically addresses the efficiency and reliability bottlenecks of traditional Gaussian process models in complex physical field applications through multi-fidelity data fusion and adaptive modeling mechanisms. To address computational efficiency issues, a Latin hypercube sampling algorithm is used to co-sample fluid data from both the factory and the user environment, optimizing the sample structure while ensuring data representation capabilities. A kernel coupling matrix constructed based on multi-fidelity wall mesh data decomposes fidelity-dependent and fidelity-independent parallel computation modules. Incorporating sparse approximation techniques significantly reduces the computational complexity of the global kernel function, achieving real-time simulation response. To address uncertainty quantification errors, a fidelity-sensitive error propagation model is designed. Adaptive kernel bandwidth adjustment is used to capture the anisotropic characteristics of flow field parameters. Tensor decomposition techniques are introduced to quantify cross-scale modeling biases. A Bayesian evidence framework is then used to dynamically optimize model parameters, improving uncertainty quantification accuracy. The error source analysis module within the multi-level verification mechanism identifies the primary contributors to model mismatch, guiding the iterative optimization of the active learning strategy to maintain the confidence of prediction results within engineering tolerances. This technology system uses a data-driven multi-fidelity fusion architecture to improve modeling efficiency while ensuring accuracy, support multidisciplinary optimization of complex aerodynamic shapes, and significantly shorten the cycle time. It provides an efficient and reliable new methodology for agile design and online verification of equipment components, effectively resolving the contradiction between high-fidelity simulation and engineering practicality, and meeting the dual requirements of modern aviation manufacturing for production efficiency and quality control.

[0032] In one embodiment, Figure 2 As shown in the figure, a multi-fidelity modeling process flow diagram is provided. The specific steps are as follows:

[0033] Data preprocessing

[0034] To ensure consistent model input format, the present invention processes the low-fidelity and high-fidelity data obtained through Latin hypercube sampling. First, the obtained wall grid point coordinates and heat flux coefficients are converted into arrays to align their dimensions. The low-fidelity data are then set to 0 and the high-fidelity data to 1, and these are used as additional inputs to the model. Finally, the low-fidelity and high-fidelity data are concatenated column by column to form the final training input for the model.

[0035] Defining the covariance kernel

[0036] A multi-fidelity Gaussian process model is constructed using a radial basis function (RBF) as the basic kernel function and a kernel coupling matrix. Specifically, the RBF provides local smoothness, and the kernel coupling matrix effectively fuses data of different fidelity levels, enabling low-fidelity data to assist high-fidelity data in prediction.

[0037] Radial Basis Function Kernel

[0038] The radial basis function is one of the most commonly used kernel functions. It has the characteristics of smoothness and locality and is suitable for constructing one-dimensional or multi-dimensional continuous functions. Its core function is to measure the similarity between different input data points. Its mathematical expression is:

[0039] ;

[0040] in: and represents the grid point coordinates of the input variables, Represents the variance of the kernel, which controls the output scale of the function, Represents the length scale, which determines the correlation distance between points in the input space. Represents the Euclidean distance between input points.

[0041] In the multi-fidelity Gaussian process model, the relationship between low-fidelity data and high-fidelity data can be modeled by the RBF kernel. For each pair of low-fidelity and high-fidelity input points and , the RBF kernel can calculate the similarity between them and measure the relationship between the two points in the input space.

[0042] Kernel coupling matrix

[0043] In order to better combine data of different fidelity, the present invention uses a coupling matrix in the multi-fidelity Gaussian process to explicitly combine the relationship between low-fidelity data and high-fidelity data. The kernel coupling matrix is ​​composed of multiple covariance matrices, each of which represents the relationship between different fidelity levels. According to the present invention, these covariance matrices are the covariance matrices of the low-fidelity data , the covariance matrix of the high-fidelity data , the cross-covariance matrix between low-fidelity and high-fidelity data , the cross-covariance matrix between high-fidelity and low-fidelity data , for example, the covariance of low-fidelity and high-fidelity data can be expressed as:

[0044] ;

[0045] Then the kernel coupling matrix is:

[0046] ;

[0047] This matrix is ​​calculated using the RBF kernel, and through model optimization, low-fidelity data and high-fidelity data are effectively integrated within the same framework. In this way, low-fidelity data not only provides auxiliary information during training, but also supports the prediction process of high-fidelity data, thereby improving the accuracy of the overall model.

[0048] Constructing a multi-fidelity Gaussian regression process

[0049] In the present invention, multi-fidelity Gaussian regression process modeling is adopted, which can effectively fuse low-fidelity and high-fidelity data and improve the accuracy of model prediction.

[0050] Joint distribution of data

[0051] Assume a total low-fidelity samples and high-fidelity samples, defined as:

[0052] ;

[0053] ;

[0054] ;

[0055] ;

[0056] in, Low-fidelity wall mesh data representing low-fidelity input, High-fidelity wall mesh data representing high-fidelity input, Low-fidelity real data representing low-fidelity output, High-fidelity real data representing high-fidelity output.

[0057] Combine the low-fidelity and high-fidelity data to form joint input data:

[0058] ;

[0059] ;

[0060] And assign each data point a fidelity index, where low fidelity is 0 and high fidelity is 1:

[0061] .

[0062] Gaussian process assumption

[0063] The data needs to obey a joint Gaussian distribution:

[0064] in, and Represent the mean functions of low fidelity and high fidelity respectively, and are usually set to 0. The covariance matrix is ​​composed of the radial basis function kernel function and the kernel coupling matrix.

[0065] After obtaining the joint Gaussian process model, we can make predictions. Given a new input data point , predict the output value of the input point through the trained joint Gaussian process model .

[0066] Predictive Analytics

[0067] After completing the construction of the multi-fidelity Gaussian process regression model, the present invention uses the model to predict new input data and evaluate the accuracy and uncertainty of the prediction. To verify the accuracy of the prediction results, the present invention uses the following error evaluation indicators:

[0068] Mean Squared Error (MSE):

[0069] ;

[0070] Root mean square error (RMSE)

[0071] ;

[0072] Coefficient of determination :

[0073] ;

[0074] in, is the mean of the true values, The closer it is to 1, the better the model fitting effect is.

[0075] At the same time, the prediction results were visualized, the prediction results of the high-fidelity simulation were plotted, and the confidence interval of the predicted mean was added to the curve to visualize the uncertainty.

[0076] In one embodiment, the Latin hypercube algorithm is used to sample the factory fluid data and the used fluid data respectively, and the sampling number of the factory fluid data is set to be twice the sampling number of the used fluid data:

[0077] ;

[0078] in, are the coordinates of the wall points, For factory fluid data or used fluid data, The dimension of factory fluid data or used fluid data, is the random permutation function of the j-th dimension, is an independent uniform random number, n is the number of samples of factory fluid data or used fluid data. Output low-fidelity wall mesh data and high-fidelity wall mesh data.

[0079] In one embodiment, based on the low-fidelity wall grid data and the high-fidelity wall grid data, a covariance matrix of the factory fluid data, a covariance matrix of the used fluid data, a cross-covariance matrix between the factory fluid data and the used fluid data, and a cross-covariance matrix between the used fluid data and the factory fluid data are generated respectively to construct a kernel coupling matrix:

[0080] ;

[0081] in, is the kernel coupling matrix, is the covariance matrix of factory fluid data, is the cross covariance matrix between factory fluid data and used fluid data, is the cross covariance matrix between the used fluid data and the factory fluid data, is the covariance matrix of the fluid data used.

[0082] In one embodiment, the multi-fidelity data obtained by merging the low-fidelity wall mesh data with the high-fidelity wall mesh data includes:

[0083] ;

[0084] ;

[0085] ;

[0086] in, For multi-fidelity data, is low-fidelity wall mesh data, It is high-fidelity wall mesh data.

[0087] In one embodiment, the multi-fidelity Gaussian regression simulation model is:

[0088] ;

[0089] ;

[0090] in, For low-fidelity real data, For high-fidelity real data, is the low-fidelity mean function, is a high-fidelity mean function, is the observed value of the nth low-fidelity data point, is the observed value of the nth high-fidelity data point, is the covariance matrix of factory fluid data, is the cross covariance matrix between factory fluid data and used fluid data, is the cross covariance matrix between the used fluid data and the factory fluid data, is the covariance matrix of the fluid data used.

[0091] In one embodiment, the radial basis function kernel function is:

[0092] ;

[0093] in, is low-fidelity wall mesh data, is high-fidelity wall mesh data, is the radial basis function kernel function, is the length scale, is the variance of the kernel, is the Euclidean distance between the low-fidelity wall mesh data and the high-fidelity wall mesh data.

[0094] In one embodiment, the preset error evaluation indicators include: mean square error and root mean square error. The mean square error is:

[0095] ;

[0096] in, is the real fluid data, is a sample data set. The root mean square error is:

[0097] ;

[0098] ;

[0099] in, is the coefficient of determination, is the mean value of the real fluid data.

[0100] It is worth noting that by combining high-fidelity and low-fidelity data through multi-fidelity Gaussian process modeling, the prediction accuracy is improved, making the proxy model closer to the real physical simulation results. By adopting the Latin hypercube sampling (LHS) method, 100 groups of samples are selected from the low-fidelity data and 50 groups of samples are selected from the high-fidelity data to ensure that the data points are evenly distributed and improve data utilization efficiency. Compared with the direct use of high-fidelity simulation, this method reduces the demand for high-fidelity data through multi-fidelity proxy models, thereby reducing computing resource consumption and improving computing efficiency. Through covariance modeling and kernel coupling matrix, uncertainty quantification technology is introduced to enable the model to evaluate the uncertainty of the prediction results and provide a more reliable decision-making basis for engineering applications. Finally, this method obtains a correlation of 0.95 between low-fidelity and high-fidelity data. The value of 0.9938 indicates that low-fidelity data can be effectively combined to optimize high-fidelity predictions. This method is suitable for complex computational fluid dynamics (CFD) problems such as wall thermal flow simulation and can be extended to other multi-scale and multi-physics simulations, such as structural simulation and aerodynamic analysis, to increase the breadth of engineering applications.

[0101] It should be understood that although Figure 1-2 The steps in the flowchart are shown in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. In addition, Figure 1-2 At least part of the steps may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least part of the sub-steps or stages of other steps.

[0102] In one embodiment, Figure 3 As shown, a multi-fidelity fluid data fusion simulation modeling device is provided, including: a data acquisition module 302, an internal coupling matrix construction module 304, a multi-fidelity data training module 306 and an evaluation module 308, wherein:

[0103] The data acquisition module 302 is used to acquire the factory fluid data of the equipment parts and the fluid data of the equipment parts in the simulation environment stored in the multi-fidelity file.

[0104] The inner coupling matrix construction module 304 is used to sample the factory fluid data and the used fluid data respectively using the Latin hypercube algorithm, output low-fidelity wall grid data and high-fidelity wall grid data, and construct the inner coupling matrix according to the low-fidelity wall grid data and the high-fidelity wall grid data.

[0105] The multi-fidelity data training module 306 is used to input the multi-fidelity data obtained by merging the low-fidelity wall mesh data and the high-fidelity wall mesh data as a sample data set into a multi-fidelity Gaussian regression simulation model constructed based on the kernel coupling matrix, and use the radial basis function kernel function to solve the kernel coupling matrix to obtain the sample data set to be evaluated.

[0106] The evaluation module 308 is used to obtain real fluid data after verifying the sample data set to be evaluated through a preset error evaluation index.

[0107] Regarding the specific definition of a multi-fidelity fluid data fusion simulation modeling device, please refer to the definition of a multi-fidelity fluid data fusion simulation modeling method above, which will not be repeated here. The various modules in the above-mentioned multi-fidelity fluid data fusion simulation modeling device can be implemented in whole or in part by software, hardware, and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0108] In one embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as follows: Figure 4 As shown. The computer device includes a processor, a memory, a network interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, a multi-fidelity fluid data fusion simulation modeling method is implemented. The display screen of the computer device can be a liquid crystal display or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a button, trackball or touchpad provided on the computer device housing, or an external keyboard, touchpad or mouse.

[0109] Those skilled in the art will understand that Figure 3-4 The structure shown in the figure is merely a block diagram of a portion of the structure related to the solution of the present invention and does not constitute a limitation on the computer device to which the solution of the present invention is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0110] In one embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0111] Obtain the factory fluid data of equipment parts stored in multi-fidelity files and the fluid data used by equipment parts in the simulation environment.

[0112] The Latin hypercube algorithm is used to sample the factory fluid data and the used fluid data respectively, and the low-fidelity wall grid data and the high-fidelity wall grid data are output. The kernel coupling matrix is ​​constructed according to the low-fidelity wall grid data and the high-fidelity wall grid data.

[0113] The multi-fidelity data obtained by merging the low-fidelity wall mesh data and the high-fidelity wall mesh data are used as the sample data set and input into the multi-fidelity Gaussian regression simulation model constructed according to the kernel coupling matrix. The radial basis function kernel function is used to solve the kernel coupling matrix to obtain the sample data set to be evaluated.

[0114] After the sample data set to be evaluated is verified by the preset error evaluation index, the real fluid data is obtained.

[0115] Those skilled in the art will appreciate that all or part of the processes in the above-described method embodiments can be implemented by instructing the relevant hardware using a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the above-described method embodiments. Any reference to memory, storage, database, or other media used in the various embodiments provided herein may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct RAMbus dynamic RAM (DRDRAM), and RAMbus dynamic RAM (RDRAM).

[0116] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0117] The above-described embodiments merely illustrate several implementations of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention. It should be noted that a person skilled in the art would be able to make numerous modifications and improvements without departing from the spirit of the present invention, all of which fall within the scope of protection of the present invention. Therefore, the scope of protection of the present invention shall be determined by the appended claims.

Claims

1. A multi-fidelity fluid data fusion simulation modeling method, characterized in that: The method comprises: Obtaining factory fluid data of equipment components and in-use fluid data of the equipment components in a simulation environment stored in a multi-fidelity file; The Latin hypercube algorithm is used to sample the factory fluid data and the used fluid data respectively, and low-fidelity wall grid data and high-fidelity wall grid data are output. A kernel coupling matrix is ​​constructed based on the low-fidelity wall grid data and the high-fidelity wall grid data. The covariance matrix of the factory fluid data, the covariance matrix of the used fluid data, the cross-covariance matrix between the factory fluid data and the used fluid data, and the cross-covariance matrix between the used fluid data and the factory fluid data are generated based on the low-fidelity wall grid data and the high-fidelity wall grid data, thereby constructing a kernel coupling matrix: in, is the kernel coupling matrix, is the covariance matrix of factory fluid data, is the cross covariance matrix between factory fluid data and used fluid data, is the cross covariance matrix between the used fluid data and the factory fluid data, is the covariance matrix of the fluid data used; Multi-fidelity data obtained by merging low-fidelity wall grid data and high-fidelity wall grid data is used as a sample data set, and is input into a multi-fidelity Gaussian regression simulation model constructed according to the kernel coupling matrix, and the kernel coupling matrix is ​​solved using a radial basis function kernel function to obtain a sample data set to be evaluated; After the sample data set to be evaluated is verified by a preset error evaluation index, real fluid data is obtained.

2. The method according to claim 1, characterized in that The Latin hypercube algorithm is used to sample the factory fluid data and the used fluid data respectively, and output low-fidelity wall mesh data and high-fidelity wall mesh data, including: The Latin hypercube algorithm is used to sample the factory fluid data and the used fluid data respectively, and the sampling number of the factory fluid data is set to be twice the sampling number of the used fluid data: in, are the coordinates of the wall points, For factory fluid data or used fluid data, The dimension of factory fluid data or used fluid data, is the random permutation function of the j-th dimension, is an independent uniform random number, n is the number of samples of factory fluid data or used fluid data; Output low-fidelity wall mesh data and high-fidelity wall mesh data.

3. The method according to claim 2, characterized in that Multi-fidelity data obtained by merging low-fidelity wall mesh data with high-fidelity wall mesh data includes: in, For multi-fidelity data, is low-fidelity wall mesh data, It is high-fidelity wall mesh data.

4. The method according to any one of claims 1 to 3, characterized in that The multi-fidelity Gaussian regression simulation model is: in, For low-fidelity real data, For high-fidelity real data, is the low-fidelity mean function, is a high-fidelity mean function, is the observed value of the nth low-fidelity data point, is the observed value of the nth high-fidelity data point, is the covariance matrix of factory fluid data, is the cross covariance matrix between factory fluid data and used fluid data, is the cross covariance matrix between the used fluid data and the factory fluid data, is the covariance matrix of the fluid data used.

5. The method according to claim 4, characterized in that The radial basis function kernel function is: in, is low-fidelity wall mesh data, is high-fidelity wall mesh data, is the radial basis function kernel function, is the length scale, is the variance of the kernel, is the Euclidean distance between the low-fidelity wall mesh data and the high-fidelity wall mesh data.

6. The method according to claim 5, characterized in that The preset error evaluation indicators include: mean square error and root mean square error; The mean square error is: in, is the real fluid data, is a sample data set; The root mean square error is: in, is the coefficient of determination, is the mean value of the real fluid data.

7. A multi-fidelity fluid data fusion simulation modeling device, characterized in that: The device comprises: A data acquisition module, used to acquire factory fluid data of equipment parts and fluid data of equipment parts in a simulation environment stored in multi-fidelity files; An inner coupling matrix construction module is used to sample the factory fluid data and the used fluid data respectively using a Latin hypercube algorithm, output low-fidelity wall grid data and high-fidelity wall grid data, and construct a kernel coupling matrix based on the low-fidelity wall grid data and the high-fidelity wall grid data; and to generate a covariance matrix of the factory fluid data, a covariance matrix of the used fluid data, a cross-covariance matrix between the factory fluid data and the used fluid data, and a cross-covariance matrix between the used fluid data and the factory fluid data based on the low-fidelity wall grid data and the high-fidelity wall grid data, thereby constructing a kernel coupling matrix: in, is the kernel coupling matrix, is the covariance matrix of factory fluid data, is the cross covariance matrix between factory fluid data and used fluid data, is the cross covariance matrix between the used fluid data and the factory fluid data, is the covariance matrix of the fluid data used; a multi-fidelity data training module, configured to input multi-fidelity data obtained by merging low-fidelity wall mesh data and high-fidelity wall mesh data as a sample data set into a multi-fidelity Gaussian regression simulation model constructed according to the kernel coupling matrix, and solve the kernel coupling matrix using a radial basis function kernel function to obtain a sample data set to be evaluated; The evaluation module is used to obtain real fluid data after verifying the sample data set to be evaluated through a preset error evaluation index.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

Images