Industrial analogue simulation system for fluid flow characteristics

Abstract

The invention relates to the technical field of electric digital data processing, and discloses a fluid flow characteristic industrial analogue simulation system, which comprises a fluid parameter initialization module used for acquiring density field, velocity field and pressure field data of fluid and generating a distribution function matrix and a target equilibrium state distribution function matrix; the moment space basic representation module is used for performing spectral decomposition on the distribution function matrix, generating a feature vector matrix and a feature value matrix, and mapping the distribution function matrix to a moment space; and the dynamic characteristic optimization module is used for dynamically adjusting the characteristic value matrix based on the characteristics of the density gradient, the strain rate tensor and the vorticity field and generating an optimized moment space conversion matrix. According to the method, the complex dynamic characteristics of the fluid are accurately captured through characteristic decomposition and weight adjustment, the technical effect of high-precision turbulence characteristic simulation is achieved, and compared with a single parameter analysis scheme of the fluid characteristics in the prior art, the defect that multi-scale characteristics cannot be processed at the same time is overcome.

Description

Technical Field

[0001] The invention relates to the technical field of electrical digital data processing, and in particular to an industrial simulation system for fluid flow characteristics. Background Art

[0002] In the current industrial field, the simulation of fluid flow characteristics is one of the important tools for solving fluid mechanics problems, and is widely used in aerospace, energy engineering, environmental science and other fields. In the existing technology, most industrial simulation systems rely on a fixed algorithm framework to simulate flow field characteristics by analyzing density field, velocity field and pressure field.

[0003] According to the search announcement number: CN115795989B, a fluid motion simulation method, simulation terminal, electronic device and medium are disclosed. This technical solution aims to solve the problems of low computational efficiency and high data processing complexity in traditional fluid motion simulation, while improving the applicability of fluid simulation in diverse scenarios.

[0004] Although the calculation of the fluid distribution function is realized through the algorithm during the specific operation process, it mainly adopts the static characteristic matrix analysis method when dealing with complex turbulent field characteristics. This method can still meet a certain accuracy when simulating simple flow fields, but for the multi-scale characteristics of turbulence and its dynamic evolution law, the existing technology is difficult to provide global optimization calculation results. Especially when the fluid characteristics are highly nonlinear and the parameters change rapidly, the limitations of traditional methods are more obvious, resulting in local deviations and overall errors in the calculation results. Summary of the invention

[0005] In order to make up for the above shortcomings, the present invention provides an industrial simulation system for fluid flow characteristics, aiming to improve the problems of single characteristic extraction and insufficient dynamic adjustment capability in the simulation of complex flow fields in the prior art.

[0006] In a first aspect, the present invention provides the following technical solution, a fluid flow characteristics industrial simulation system, comprising:

[0007] The fluid parameter initialization module is used to collect the density field, velocity field and pressure field data of the fluid, and generate the distribution function matrix and the target equilibrium state distribution function matrix;

[0008] The moment space basic representation module is used to perform spectral decomposition on the distribution function matrix, generate eigenvector matrix and eigenvalue matrix, and map the distribution function matrix to the moment space;

[0009] A dynamic feature optimization module is used to dynamically adjust the eigenvalue matrix based on the characteristics of the density gradient, strain rate tensor and vorticity field, and generate an optimized moment space transformation matrix;

[0010] A geometric adjustment module, used for calculating a dynamic adjustment path between a distribution function matrix and a target distribution function matrix based on a logarithmic mapping and an exponential mapping of a Riemann manifold;

[0011] An optimal transmission optimization module, used to optimize the moment space transformation matrix by minimizing the transmission cost between the current distribution function matrix and the target distribution function matrix;

[0012] The simulation calculation and visualization module is used to update the distribution function based on the optimized moment space conversion matrix and output the density field, velocity field and turbulence characteristic results of the fluid.

[0013] Preferably, the fluid parameter initialization module calculates the norm of the density gradient, the modulus of the strain rate tensor and the modulus of the vorticity field, and uses the calculation results as input to the dynamic feature optimization module.

[0014] Preferably, the moment space basic representation module generates the eigenvector matrix and the eigenvalue matrix in the following manner:

[0015] Perform spectral decomposition on the distribution function matrix and decompose it into the product of the eigenvector matrix and the diagonal eigenvalue matrix;

[0016] The eigenvalues ​​in the eigenvalue matrix represent the characteristic weights of the distribution function in the moment space.

[0017] Preferably, the dynamic feature optimization module adjusts the eigenvalue matrix based on the following method:

[0018] The adjustment of the eigenvalue is calculated by the norm of the density gradient, the modulus of the strain rate tensor and the modulus of the vorticity field;

[0019] The eigenvalue matrix is ​​dynamically adjusted to generate an optimized eigenvalue matrix to adapt to the non-uniform fluid characteristics in the turbulent field.

[0020] Preferably, the geometric adjustment module optimizes the dynamic adjustment path of the moment space conversion matrix by:

[0021] Calculate the tangent vector from the current distribution function matrix to the target distribution function matrix through logarithmic mapping;

[0022] A dynamic adjustment path for generating a moment space transformation matrix through exponential mapping based on the tangent vector.

[0023] Preferably, the optimal transmission optimization module realizes optimization in the following manner:

[0024] The transmission cost between the current distribution function matrix and the target distribution function matrix is ​​defined as the quadratic norm of the difference between the two matrices;

[0025] Minimize the transmission cost and generate the final optimized moment space transformation matrix by solving the Wasserstein distance between the distribution function matrix and the target distribution function matrix.

[0026] Preferably, the simulation calculation and visualization module is used to generate density field, velocity field and turbulence characteristic simulation results based on the optimized moment-space conversion matrix, and output them in a graphical interface.

[0027] In a second aspect, the present invention provides the following technical solution, a method for industrial simulation of fluid flow characteristics, comprising the following steps:

[0028] S1: Collect the density field, velocity field and pressure field data of the fluid, generate the distribution function matrix and the target equilibrium state distribution function matrix;

[0029] S2: Perform spectral decomposition on the distribution function matrix to generate eigenvector matrix and eigenvalue matrix;

[0030] S3: Dynamically adjust the eigenvalue matrix based on the norm of the density gradient, the modulus of the strain rate tensor, and the modulus of the vorticity field to generate an optimized moment space transformation matrix;

[0031] S4: embed the distribution function matrix and the target distribution function matrix into the Riemann manifold, calculate the tangent vector based on the logarithmic mapping, and generate the dynamic adjustment path of the distribution function matrix based on the exponential mapping;

[0032] S5: Based on the Wasserstein distance between the distribution function matrix and the target distribution function matrix, the transmission cost of the two is minimized to generate the final optimized moment space transformation matrix;

[0033] S6: Update the distribution function matrix based on the final optimized moment space conversion matrix to generate the simulation results of the density field, velocity field and turbulence characteristics of the fluid;

[0034] S7: Output the updated density field, velocity field and turbulence characteristics simulation results as visualization graphics.

[0035] In a third aspect, the invention provides the following technical solution: a computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned method for industrial simulation of fluid flow characteristics when executing the computer program.

[0036] In a fourth aspect, the present invention provides the following technical solution: a readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the above-mentioned industrial simulation method for fluid flow characteristics.

[0037] The present invention has the following beneficial effects:

[0038] 1. In the present invention, the complex dynamic characteristics of the fluid are accurately captured through feature decomposition and weight adjustment, achieving the technical effect of high-precision simulation of turbulent characteristics. Compared with the solution of single parameter analysis of fluid characteristics in the prior art, it solves the problem that multi-scale characteristics cannot be processed simultaneously.

[0039] 2. In the present invention, logarithmic mapping and exponential mapping are used to generate a dynamic adjustment path for the distribution function, which effectively optimizes the evolution trajectory of fluid properties. Compared with the traditional solution based on fixed path interpolation, the present invention solves the technical problem of low simulation accuracy caused by path selection limitations.

[0040] 3. In the present invention, by combining the optimal transmission theory with the Wasserstein distance optimization, an efficient transmission mechanism from the distribution function to the target state is constructed. Compared with the adjustment method based only on the minimum cost in the prior art, this scheme overcomes the defects of large local deviation of the transmission result and insufficient global optimization. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 This is a system architecture diagram of a fluid flow characteristics industrial simulation system proposed by the present invention;

[0042] Figure 2 A method flow chart of an industrial simulation method for fluid flow characteristics proposed by the present invention; DETAILED DESCRIPTION

[0043] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0044] Embodiment 1

[0045] Reference Figure 1 In a first embodiment of the present invention, the present invention provides a fluid flow characteristic industrial simulation system, comprising:

[0046] The fluid parameter initialization module is used to collect the density field, velocity field and pressure field data of the fluid, and generate the distribution function matrix and the target equilibrium state distribution function matrix;

[0047] The fluid parameter initialization module includes:

[0048] Fluid data acquisition module: used to obtain the initial data of fluid density, velocity and pressure, which can be obtained through sensor measurement or user-defined input.

[0049] Initial condition setting module: supports setting simulation boundary conditions and basic initial states of the fluid, including temperature, viscosity and fluid boundary flow.

[0050] External data docking module: provides docking function with external databases or simulation tools to obtain supplementary fluid property data.

[0051] Specifically, in this system, the fluid parameter initialization module undertakes the starting point of the entire simulation process, and its main function is to collect and initialize the macroscopic physical property data of the fluid. These data include but are not limited to density field, velocity field, and pressure field. The initialization module not only provides basic input for the subsequent moment space representation and dynamic optimization of the system, but is also responsible for processing the gradient and related tensor information of these physical property data, thereby laying the foundation for higher-precision simulation calculations.

[0052] In general, the fluid parameter initialization module needs to have the ability to parse and structure the initial data to ensure that the data can be directly applied to the various submodules of the simulation system. As an implementation method, the module can obtain the initial flow field parameters through real-time sensors, fluid simulation databases, or user input data. In some embodiments, it can also support the input and update of dynamic data so that it can be reinitialized in time when the flow field changes.

[0053] In this embodiment, the fluid parameter initialization module first collects initial data of density field ρ, velocity field u and pressure field p. These parameters can be obtained by direct measurement equipment (such as fluid sensors) or by numerical simulation methods. Specifically, the density field ρ represents the mass distribution of the fluid per unit volume; the velocity field u represents the velocity vector distribution of the fluid, which is defined as:

[0054] u=[u x ,u y ,u z ],

[0055] where u x 、u y 、u z They represent the velocity components of the fluid in the x, y, and z directions in three-dimensional space respectively; the pressure field p represents the force of the fluid per unit area.

[0056] In some embodiments, the module further calculates the gradient of the density field The strain rate tensor S and the vorticity field ω.

[0057] Specifically, the density gradient The calculation expression is:

[0058]

[0059] in They represent the rate of change of the density field in the x, y, and z directions respectively.

[0060] The strain rate tensor S is used to describe the local deformation caused by the fluid velocity gradient and is expressed as:

[0061]

[0062] At the same time, the calculation formula of the vorticity field ω is:

[0063]

[0064] in is the curl of the velocity field, which describes the intensity and direction of rotational motion in the fluid.

[0065] In one possible implementation, the module performs a density gradient The norm of the strain rate tensor S and the vorticity field ω are calculated as input data for the dynamic optimization module.

[0066] Specifically: The norm of the density gradient is defined as:

[0067]

[0068] The modulus of the strain rate tensor is calculated as the square root of the sum of the squares of all components:

[0069]

[0070] The modulus of the vorticity field is calculated as:

[0071]

[0072] where ω x ,ω y ,ω z are the vorticity field at x 、 y 、 The component in the z direction.

[0073] As an option, the module can also accept dynamic input and update the calculation results of the above parameters in real time when the external flow field changes, thereby ensuring the dynamic response capability of the simulation system. In some embodiments, the module directly transmits the calculated density gradient, strain rate tensor and vortex field data to the moment space basic representation module through a data interface to provide input for subsequent spectral decomposition and feature optimization.

[0074] In one possible implementation, in order to improve computational efficiency, the module can perform spatial discretization on the density field, velocity field, and pressure field. For example, the finite difference or finite volume method is used to divide the continuous flow field into a finite number of grid cells, and the corresponding density, velocity, and pressure values ​​are calculated for each cell. Specifically, in the grid cell, the density gradient and velocity gradient The discretized form of can be expressed as:

[0075]

[0076] Where Δx is the grid spacing, ρ i+1 , i-1 and u x,i+1 、u x,i-1 are the density values ​​and velocity components of adjacent grid points respectively.

[0077] The moment space basic representation module is used to perform spectral decomposition on the distribution function matrix, generate eigenvector matrix and eigenvalue matrix, and map the distribution function matrix to the moment space;

[0078] The moment space basic representation module includes:

[0079] Matrix decomposition module: performs matrix processing on the input fluid distribution data, such as eigendecomposition to extract eigenvalues ​​and eigenvectors.

[0080] Basic matrix processing module: constructs matrix data for calculation through eigendecomposition results, and supports matrix normalization or sparse processing.

[0081] Fluid property extraction module: Extract key characteristic parameters such as density gradient, velocity change and rotation characteristics based on the initial fluid data.

[0082] Specifically, the main function of the moment space basic representation module is to structure the distribution function matrix output by the fluid parameter initialization module. Its core task is to map the distribution function matrix to the moment space through the spectral decomposition method, generate the eigenvector matrix and eigenvalue matrix, and provide the calculation basis for the subsequent dynamic feature optimization and geometry adjustment modules. In general, the design of this module needs to fully consider the dimension and properties of the input distribution function matrix to ensure the accuracy and applicability of the moment space conversion.

[0083] As an implementation method, the moment space basic representation module simplifies the dimension and extracts the characteristics of the distribution function matrix of the fluid by decomposing the eigenvectors and eigenvalues. In some embodiments, the module can also pre-process the characteristics of the input matrix, such as normalization or sparse processing, to enhance the numerical stability of the decomposition process.

[0084] In this embodiment, the main function of the moment space basic representation module is to convert the input distribution function matrix Perform spectral decomposition to generate eigenvector matrix and the eigenvalue matrix

[0085] Specifically, the spectral decomposition formula of the distribution function matrix F is:

[0086]

[0087] in:

[0088] Q=[q1,q2,…,q N ] is the eigenvector matrix, each column q i is a feature vector;

[0089] Λ=diag(λ1,λ2,…,λ N ) is the eigenvalue matrix, where λ i is the eigenvalue of the corresponding eigenvector. In one possible implementation, the module first checks the symmetry of the input matrix.

[0090] If F is a symmetric matrix, the standard eigendecomposition method is used; if F is asymmetric, it is processed by singular value decomposition (SVD), the formula is:

[0091]

[0092] in:

[0093] and are the left singular vector matrix and the right singular vector matrix respectively;

[0094] Σ=diag(σ1,σ2,…,σ N ) is the singular value matrix, where σ i are the singular values ​​of the matrix F.

[0095] Specifically, the element λ of the eigenvalue matrix Λ i Represents the weight of the distribution function matrix along the eigenvector direction in the moment space. As an option, this module sorts the eigenvalues ​​to determine the most important direction. The sorted matrix helps the dynamic feature optimization module prioritize important feature directions and ignore unnecessary minor features.

[0096] In some embodiments, in order to improve the efficiency of numerical calculation, the module performs sparse processing on the distribution function matrix F. Sparse processing can be achieved by setting the matrix elements less than a threshold to zero, and the formula is:

[0097]

[0098] Where ε is the preset threshold.

[0099] In a possible implementation, the module may also perform normalization processing on the input matrix according to the condition number of the matrix.

[0100] The normalization method can be expressed as:

[0101]

[0102] where ∥F∥ F represents the Frobenius norm of the matrix, defined as:

[0103]

[0104] As an option, the moment space basis representation module can also process the input matrix in blocks. For example, for a high-dimensional matrix F, the module can divide it into multiple low-dimensional sub-matrices, perform spectral decomposition on each sub-matrix independently, and finally merge the decomposition results.

[0105] This method can significantly reduce the computational complexity and is suitable for large-scale data processing scenarios. In this embodiment, the eigenvector matrix Q and eigenvalue matrix Λ generated by decomposition are directly passed to the dynamic feature optimization module. Among them, the eigenvector matrix Q is used to define the basis vectors of the moment space, and the eigenvalue matrix Λ is used to describe the weights of each basis vector in the moment space.

[0106] In some embodiments, the module also supports dynamic input, that is, when the output data of the fluid parameter initialization module is updated, the decomposition result can be updated in real time. This dynamic update capability can improve the adaptability of the system in real-time simulation scenarios.

[0107] A dynamic feature optimization module is used to dynamically adjust the eigenvalue matrix based on the characteristics of the density gradient, strain rate tensor and vorticity field, and generate an optimized moment space transformation matrix;

[0108] The dynamic feature optimization module includes:

[0109] Fluid property adjustment module: Using dynamic optimization methods, the matrix data is adjusted according to the density changes, speed changes and rotation characteristics of the fluid.

[0110] Geometric mapping adjustment module: It completes the gradual transition adjustment from the current matrix to the target matrix by calculating the dynamic adjustment path.

[0111] Specifically, the dynamic feature optimization module is an important component of the system of the present invention. Its main function is to dynamically adjust the distribution characteristics of the eigenvalue matrix based on the eigenvector matrix and eigenvalue matrix output by the moment space basic representation module, combined with the density gradient, strain rate tensor and vortex field data provided by the fluid parameter initialization module. In general, this module is used to generate an optimized moment space conversion matrix to provide efficient input for the subsequent geometry adjustment module.

[0112] As a core function, the dynamic feature optimization module captures the multi-scale turbulence characteristics in complex flow fields by optimizing the distribution of eigenvalues. In some embodiments, the module can also adaptively adjust the weights of eigenvalues ​​according to the local characteristics of the flow field, thereby improving the adaptability of the simulation system in irregular turbulence scenarios.

[0113] In this embodiment, the dynamic feature optimization module first receives the eigenvalue matrix Λ=diag(λ1,λ2,…,λ N ) and the eigenvector matrix At the same time, this module obtains the density gradient from the fluid parameter initialization module Numerical data of the strain rate tensor S and the vorticity field ω are used for the dynamic adjustment of the eigenvalue matrix.

[0114] Specifically, the adjustment of the eigenvalue matrix is ​​determined by the following formula:

[0115] Λ opt =Λ+ΔΛ,

[0116] where ΔΛ=diag(δλ1,δλ2,…,δλ N ), represents the dynamic adjustment amount of each eigenvalue.

[0117] In one possible implementation, the adjustment amount δλ i The calculation formula is:

[0118]

[0119] in:

[0120] is the norm of the density gradient, which is used to describe the rate of change of the density field;

[0121] ∥S∥ is the modulus of the strain rate tensor, which reflects the intensity of local deformation of the fluid;

[0122] ∥ω∥ is the modulus of the vorticity field, which is used to describe the rotational characteristics of the fluid;

[0123] α, β, and γ are weighting coefficients, which are used to adjust the contribution ratio of density gradient, strain rate, and vorticity field in eigenvalue adjustment.

[0124] As an option, the module supports dynamic adjustment of weighting coefficients according to actual application scenarios. For example, in a scenario where vorticity dominates the flow field, the weight of γ can be appropriately increased to enhance the modeling ability of rotation characteristics.

[0125] In general, the dynamic feature optimization module needs to ensure the efficiency and numerical stability of the calculation. In some embodiments, the module uses parallel computing technology to achieve dynamic adjustment of the eigenvalue matrix through multi-threading or GPU acceleration. In addition, in high-dimensional flow field scenarios, the module can process the input matrix in blocks, adjust the eigenvalues ​​of each block separately, and then merge the adjusted results to generate the final moment-space transformation matrix.

[0126] In a possible implementation, the module also supports dynamic response function of eigenvalue adjustment. When the input data of the fluid parameter initialization module changes, the dynamic feature optimization module can update the adjustment result of the eigenvalue matrix in real time to adapt to the rapidly changing flow field environment.

[0127] A geometric adjustment module, used for calculating a dynamic adjustment path between a distribution function matrix and a target distribution function matrix based on a logarithmic mapping and an exponential mapping of a Riemann manifold;

[0128] Specifically, in general, the role of this module is to generate a dynamic path from the current distribution function matrix to the target distribution function matrix, ensuring that the optimized moment-space transformation matrix can adapt to the nonlinear characteristics of the complex flow field, while providing high-quality initial input for the optimal transmission optimization module.

[0129] As an implementation method, the geometric adjustment module uses logarithmic mapping and exponential mapping methods on Riemann manifolds to transform the characteristic adjustment problem of moment space into a geometric shortest path problem. In some embodiments, the module also corrects the adjustment path in combination with specific parameters of the flow field, such as density gradient, velocity field changes, etc., to further improve the simulation accuracy.

[0130] In this embodiment, the main input of the geometry adjustment module includes the optimization moment space conversion matrix M generated by the dynamic feature optimization module opt , current distribution function matrix F, target distribution function matrix F targct Its main task is to calculate the adjustment path from the current matrix to the target matrix through geometric mapping.

[0131] Specifically, this module first calculates the current distribution function matrix F to the target distribution function matrix F on the Riemann manifold target The tangent vector of F (F target ) is calculated:

[0132]

[0133] in:

[0134] Log F (F target ) is the distance from the current point F to the target point F target Logarithmic mapping of ;

[0135] v represents a vector in the tangent space, which is used to describe the adjustment direction and size.

[0136] In one possible implementation, the module further maps the index Exp F (t·v) calculates the adjustment path along the tangent vector v starting from the current point:

[0137] M adj =Exp F (t·Log F (F target )),

[0138] in:

[0139] t∈[0,1] represents the time parameter in the adjustment process;

[0140] M adj is the moment space transformation matrix after dynamic adjustment.

[0141] As an option, the module supports path optimization during dynamic adjustment. During the calculation of the adjustment path, the tangent vector can be corrected in combination with the local characteristics of the flow field. For example, when there is a drastic change in the density gradient or velocity field in the flow field, a weight term can be introduced in the tangent vector:

[0142]

[0143] in:

[0144] w ρ and w u are the weights of density gradient and velocity field on the adjusted path, respectively;

[0145] is the gradient of the velocity field.

[0146] Generally, the geometry adjustment module needs to ensure the numerical stability of the adjustment path. In some embodiments, in order to avoid numerical overflow problems in the exponential mapping and logarithmic mapping calculation process, the module will normalize the input matrix.

[0147] The normalized method can be expressed as:

[0148]

[0149] where ∥F∥ F is the Frobenius norm of the distribution function matrix, and the calculation formula is:

[0150]

[0151] In some embodiments, in order to improve the calculation efficiency, the module adopts a segmented adjustment method. Specifically, the adjustment path is decomposed into multiple small segments, each of which is independently calculated through logarithmic mapping and exponential mapping, and finally the results of all small segments are spliced ​​into a complete path. This method is suitable for scenarios with high matrix dimensions and can significantly reduce the calculation complexity.

[0152] Generally, the geometry adjustment module needs to control the computational complexity while ensuring accuracy. In some embodiments, the module accelerates the computational process of exponential mapping and logarithmic mapping through parallel computing technology. For example, a GPU can be used to parallelize large-scale matrices, thereby significantly improving the adjustment efficiency.

[0153] An optimal transmission optimization module, used to optimize the moment space transformation matrix by minimizing the transmission cost between the current distribution function matrix and the target distribution function matrix;

[0154] The best transmission optimization module includes:

[0155] Transmission cost estimation module: estimates the computational cost between the current state and the target state to guide the target direction of optimization.

[0156] Optimal Adjustment Optimization Module: Generates an optimized matrix by minimizing the transmission cost for accurate calculation of fluid flow.

[0157] Specifically, the module adopts an optimization method based on Wasserstein distance to express the transmission problem in the moment space as a mathematical optimization problem. In some embodiments, the module also combines dynamic weights and block calculation methods to improve the efficiency of large-scale matrix calculations and enhance the adaptability of the system in complex flow field environments.

[0158] In this embodiment, the optimal transmission optimization module receives the initial moment space conversion matrix M from the geometric adjustment module. adj , current distribution function matrix F, target distribution function matrix F target Its main task is to generate the final optimized moment space transformation matrix M by minimizing the transmission cost final .

[0159] Specifically, this module uses Wasserstein distance as the optimization target to quantify the difference between the current distribution function matrix F and the target distribution function matrix F. targctThe transmission cost between them. The mathematical definition of Wasserstein distance is:

[0160]

[0161] in:

[0162] Γ represents the set of all possible transmission plans;

[0163] π(F,F target ) is the transmission plan, indicating that from F to F target The transmission probability distribution of

[0164] ∥FF target ∥ is the distance between the current matrix and the target matrix.

[0165] In one possible implementation, the module converts the transmission problem into a linear programming problem by discretization. The discretization expression of the transmission cost is:

[0166]

[0167] in:

[0168] π ij represents the distance from the i-th point of distribution F to distribution F target The transmission amount of the jth point;

[0169] c ij is the transmission cost, usually defined as the square of the Euclidean distance between two points:

[0170] c ij =∥F i -F target,j ∥ 2 .

[0171] Generally, this module solves the above optimization problem through the alternating direction method of multipliers (ADMM) or the Sinkhorn iterative algorithm.

[0172] In some embodiments, in order to improve computational efficiency, the module introduces a regularization term and rewrites the optimization problem as follows:

[0173]

[0174] Where: ∈ is the regularization parameter, which is used to control the smoothness of the solution,

[0175] KL(π||π0) is the Kullback-Leibler divergence, which represents the difference between the current transmission plan and the reference transmission plan π0.

[0176] As an option, the module can adjust the transmission cost in combination with dynamic weights. For example, when certain areas in the flow field are more important, these areas can be assigned higher weights. The correction formula for the transmission cost is:

[0177]

[0178] where w i is a dynamic weight related to the local characteristics of the flow field (such as density gradient or velocity field change rate). In some embodiments, in order to avoid the asymmetry of the transmission plan, the module will align the initial matrix and the target matrix. For example, the normalization method can be used to ensure that F and F target The total mass is equal to:

[0179]

[0180] In one possible implementation, the module supports block calculation, decomposing a large matrix into multiple sub-matrices, performing optimization calculations on the transmission cost separately, and then concatenating the results of each sub-matrix into the final optimization matrix. This method significantly reduces the computational complexity and is suitable for high-dimensional flow field simulation scenarios.

[0181] In this embodiment, the output of the optimal transmission optimization module is the final optimized moment-space conversion matrix The calculation formula is:

[0182] M final =M adj +ΔM,

[0183] Where ΔM represents the matrix adjustment generated by optimal transmission optimization

[0184] In general, the output of the module also includes auxiliary data, such as the cost value of the transmission path, the number of optimization iterations, etc. These data can be used to monitor the convergence of the optimization process or for further debugging.

[0185] The simulation calculation and visualization module is used to update the distribution function based on the optimized moment space conversion matrix and output the density field, velocity field and turbulence characteristic results of the fluid.

[0186] The simulation and visualization modules include:

[0187] Result calculation module: Use the optimized matrix data to calculate the density distribution, velocity distribution and rotation characteristics of the fluid and generate simulation results.

[0188] Result Visualization Module: Supports the display of simulation results in graphical form, such as generating a 2D pseudo-color map of the density field or a 3D streamline map of the velocity field.

[0189] Data export module: Export the results of calculations and simulations into standard data file formats to support further analysis or sharing.

[0190] Specifically, the simulation calculation and visualization module is the final processing unit of the fluid flow characteristic industrial simulation system of the present invention. Its main function is to update the distribution function matrix of the fluid based on the final optimized moment space conversion matrix output by the optimal transmission optimization module, and calculate the corresponding fluid characteristic parameters, such as density field, velocity field and vortex field. In general, this module has the ability of data processing and visualization output at the same time, which is used to intuitively display the simulation results in a graphical form, and provide users with a clear and easy-to-interpret flow field characteristic display.

[0191] As an implementation method, the module needs to support dynamic updating of simulation results according to the time dependency of flow field data. In some embodiments, the simulation results can also be compared with external physical measurement data to calibrate the simulation model or verify the simulation accuracy.

[0192] In order to more intuitively display the simulation results, the simulation calculation and visualization module uses a graphical interface to output key characteristic parameters such as density field, velocity field and vorticity field. In some embodiments, the module uses two-dimensional or three-dimensional contour surface diagrams, streamline diagrams and vector diagrams to display fluid characteristics. For example:

[0193] The density field can be intuitively displayed through two-dimensional pseudo-color maps or three-dimensional isosurface maps;

[0194] The velocity field represents the flow direction and velocity magnitude of the fluid through vector diagrams or streamline diagrams;

[0195] The vorticity field can display the intensity distribution in the fluid rotation area through three-dimensional isosurfaces or color mapping.

[0196] In general, the visualization function of this module supports dynamic interaction, for example, users can adjust display parameters (such as color range, vector length or time step) according to their needs to observe the details of different flow field characteristics. In some embodiments, this module can also output the simulation results in the form of animation to demonstrate the flow field evolution process.

[0197] In order to improve the efficiency of simulation calculations, this module supports parallel calculations. For example, the distribution function matrix can be decomposed into multiple sub-matrices, and the values ​​of each sub-matrix can be updated in parallel under multi-threading or GPU acceleration. The updated results are then merged into a complete distribution function matrix to ensure the efficiency and accuracy of the overall calculation.

[0198] In one possible implementation, the module also supports exporting simulation results to standard data formats (such as CSV, VTK, or HDF5) so that users can conduct further data analysis or share the results with other simulation systems.

[0199] Embodiment 2:

[0200] Reference Figure 2 In a second embodiment of the present invention, the present invention provides a method for industrial simulation of fluid flow characteristics, comprising the following steps:

[0201] S1: Collect the density field, velocity field and pressure field data of the fluid, generate the distribution function matrix and the target equilibrium state distribution function matrix;

[0202] S2: Perform spectral decomposition on the distribution function matrix to generate eigenvector matrix and eigenvalue matrix;

[0203] S3: Dynamically adjust the eigenvalue matrix based on the norm of the density gradient, the modulus of the strain rate tensor, and the modulus of the vorticity field to generate an optimized moment space transformation matrix;

[0204] S4: embed the distribution function matrix and the target distribution function matrix into the Riemann manifold, calculate the tangent vector based on the logarithmic mapping, and generate the dynamic adjustment path of the distribution function matrix based on the exponential mapping;

[0205] S5: Based on the Wasserstein distance between the distribution function matrix and the target distribution function matrix, the transmission cost of the two is minimized to generate the final optimized moment space transformation matrix;

[0206] S6: Update the distribution function matrix based on the final optimized moment space conversion matrix to generate the simulation results of the density field, velocity field and turbulence characteristics of the fluid;

[0207] S7: Output the updated density field, velocity field and turbulence characteristics simulation results as visualization graphics.

[0208] Specifically, first, the density field, velocity field and pressure field data of the fluid are collected. These data can be obtained through sensors, numerical simulation or experimental means, and the distribution function matrix and the target equilibrium distribution function matrix are initialized according to the physical properties of fluid mechanics. The distribution function matrix is ​​used to characterize the current flow field state, while the target equilibrium distribution function matrix is ​​used to describe the characteristics of the fluid in the equilibrium state. Both provide basic inputs for subsequent steps.

[0209] Next, the distribution function matrix is ​​spectrally decomposed to generate the eigenvector matrix and eigenvalue matrix. The eigenvector matrix is ​​used to define the basis vectors of the moment space, and the eigenvalue matrix reflects the weight distribution of the distribution function along each basis vector. Through decomposition, the distribution function is converted into a more tractable moment space representation, laying a mathematical foundation for subsequent optimization and dynamic adjustment.

[0210] Subsequently, the eigenvalue matrix is ​​adjusted according to the dynamic characteristics of the density gradient, strain rate tensor and vorticity field to generate an optimized moment space transformation matrix. The adjustment process combines the local characteristics of the flow field and realizes the adaptive update of the eigenvalue through a dynamic optimization algorithm, thereby capturing the multi-scale characteristics of the complex turbulent field and providing reliable support for the dynamic adjustment of the matrix.

[0211] After generating the optimized moment space transformation matrix, the distribution function matrix and the target distribution function matrix are embedded in the Riemann manifold, and their tangent vectors are calculated based on logarithmic mapping, and the dynamic adjustment path of the distribution function matrix is ​​generated through exponential mapping. The generation of this path depends on the optimal adjustment strategy on the manifold geometry, ensuring that the distribution function matrix can evolve smoothly in the moment space towards the target distribution function matrix.

[0212] Furthermore, based on the Wasserstein distance between the distribution function matrix and the target distribution function matrix, the transmission cost of the two is minimized to generate the final optimized moment space conversion matrix. The transmission cost minimization process is combined with the optimal transmission theory to achieve precise adjustment of the distribution function matrix by optimizing the transmission plan, thereby effectively improving the accuracy of flow field simulation.

[0213] Subsequently, the distribution function matrix is ​​updated using the final optimized moment space conversion matrix to generate simulation results of the density field, velocity field and turbulence characteristics of the fluid. These simulation results are generated through efficient calculation methods, which can accurately describe the dynamic behavior of the fluid in a complex environment and provide data support for the research and optimization of fluid flow characteristics.

[0214] Finally, the updated density field, velocity field and turbulence characteristics simulation results are output in graphical form. The visualization output of simulation results includes but is not limited to two-dimensional pseudo-color images, three-dimensional isosurface maps and streamline maps, which can intuitively display the characteristic distribution of the fluid and its evolution law, meeting the user's needs for intuitive interpretation of flow field information.

[0215] Embodiment 3

[0216] The third embodiment of the present invention is based on the same inventive concept and proposes a computer-readable storage medium. The computer-readable storage medium stores a computer program. When the computer program is executed by a processor, the steps of an industrial simulation method for fluid flow characteristics of the above embodiment are implemented.

[0217] Embodiment 4

[0218] The fourth embodiment of the present invention is based on the same inventive concept and proposes a computer device, which includes: a processor and a memory; the processor and the memory communicate with each other; the memory is used to store instructions; the processor is used to execute the instructions in the memory to execute an industrial simulation method of fluid flow characteristics of the above embodiment.

[0219] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0220] Finally, it should be noted that the above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for those skilled in the art to modify the technical solutions described in the aforementioned embodiments or to make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A fluid flow characteristics industrial simulation system, characterized in that: include: The fluid parameter initialization module is used to collect the density field, velocity field and pressure field data of the fluid, and generate the distribution function matrix and the target equilibrium state distribution function matrix; The moment space basic representation module is used to perform spectral decomposition on the distribution function matrix, generate eigenvector matrix and eigenvalue matrix, and map the distribution function matrix to the moment space; A dynamic feature optimization module is used to dynamically adjust the eigenvalue matrix based on the characteristics of the density gradient, strain rate tensor and vorticity field, and generate an optimized moment space transformation matrix; A geometric adjustment module, used for calculating a dynamic adjustment path between a distribution function matrix and a target distribution function matrix based on a logarithmic mapping and an exponential mapping of a Riemann manifold; An optimal transmission optimization module, used to optimize the moment space transformation matrix by minimizing the transmission cost between the current distribution function matrix and the target distribution function matrix; The simulation calculation and visualization module is used to update the distribution function based on the optimized moment space conversion matrix and output the density field, velocity field and turbulence characteristic results of the fluid.

2. The fluid flow characteristics industrial simulation system according to claim 1, characterized in that: The fluid parameter initialization module calculates the norm of the density gradient, the modulus of the strain rate tensor and the modulus of the vorticity field, and uses the calculation results as inputs of the dynamic feature optimization module.

3. The fluid flow characteristics industrial simulation system according to claim 1 is characterized in that: The moment space basic representation module generates the eigenvector matrix and the eigenvalue matrix in the following way: Perform spectral decomposition on the distribution function matrix and decompose it into the product of the eigenvector matrix and the diagonal eigenvalue matrix; The eigenvalues ​​in the eigenvalue matrix represent the characteristic weights of the distribution function in the moment space.

4. The fluid flow characteristics industrial simulation system according to claim 1, characterized in that: The dynamic feature optimization module adjusts the eigenvalue matrix based on the following method: The adjustment of the eigenvalue is calculated by the norm of the density gradient, the modulus of the strain rate tensor and the modulus of the vorticity field; The eigenvalue matrix is ​​dynamically adjusted to generate an optimized eigenvalue matrix to adapt to the non-uniform fluid characteristics in the turbulent field.

5. The fluid flow characteristics industrial simulation system according to claim 1, characterized in that: The geometric adjustment module optimizes the dynamic adjustment path of the moment space conversion matrix by: Calculate the tangent vector from the current distribution function matrix to the target distribution function matrix through logarithmic mapping; A dynamic adjustment path for generating a moment space transformation matrix through exponential mapping based on the tangent vector.

6. The fluid flow characteristics industrial simulation system according to claim 1, characterized in that: The optimal transmission optimization module achieves optimization in the following ways: The transmission cost between the current distribution function matrix and the target distribution function matrix is ​​defined as the quadratic norm of the difference between the two matrices; Minimize the transmission cost and generate the final optimized moment space transformation matrix by solving the Wasserstein distance between the distribution function matrix and the target distribution function matrix.

7. The fluid flow characteristics industrial simulation system according to claim 1, characterized in that: The simulation calculation and visualization module is used to generate density field, velocity field and turbulence characteristic simulation results based on the optimized moment-space conversion matrix, and output them in a graphical interface.

8. A method for industrial simulation of fluid flow characteristics, characterized in that: A fluid flow characteristics industrial simulation system for use in any one of claims 1 to 7, The following steps are involved: S1: Collect the density field, velocity field and pressure field data of the fluid, generate the distribution function matrix and the target equilibrium state distribution function matrix; S2: Perform spectral decomposition on the distribution function matrix to generate eigenvector matrix and eigenvalue matrix; S3: Dynamically adjust the eigenvalue matrix based on the norm of the density gradient, the modulus of the strain rate tensor, and the modulus of the vorticity field to generate an optimized moment space transformation matrix; S4: embed the distribution function matrix and the target distribution function matrix into the Riemann manifold, calculate the tangent vector based on the logarithmic mapping, and generate the dynamic adjustment path of the distribution function matrix based on the exponential mapping; S5: Based on the Wasserstein distance between the distribution function matrix and the target distribution function matrix, the transmission cost of the two is minimized to generate the final optimized moment space transformation matrix; S6: Update the distribution function matrix based on the final optimized moment space conversion matrix to generate the simulation results of the density field, velocity field and turbulence characteristics of the fluid; S7: Output the updated density field, velocity field and turbulence characteristics simulation results as visualization graphics.

9. 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 computer program, the method for industrial simulation of fluid flow characteristics as described in any one of claims 1 to 7 is implemented.

10. A readable storage medium, characterized in that: The readable storage medium stores a computer program, and when the computer program is executed by a processor, the method for industrial simulation of fluid flow characteristics according to any one of claims 1 to 7 is implemented.

Images