Rapid simulation method for silt abrasion of water turbine

Abstract

The invention relates to a rapid simulation method for sediment abrasion of a water turbine, which comprises the following steps of: discretizing a fluid region to obtain a discretized grid; performing three-dimensional simulation transient solution on the discretized grids in combination with basic flow, turbulent flow and sediment corrosion to obtain a basic simulation result and an abrasion simulation data result; obtaining a working condition parameter combination according to the basic simulation result and the abrasion simulation data result, constructing a sample matrix, and performing eigenorthogonal decomposition on the sample matrix to obtain a spatial modal matrix and a modal coefficient matrix; inputting the new working condition parameters into the adaptive radial basis function to obtain a modal coefficient under the new working condition; and performing data reconstruction according to the modal coefficient and the spatial modal matrix under the new working condition to obtain a physical field prediction result. The problems that in the prior art, sediment abrasion of the water turbine is difficult to predict, the sediment abrasion calculation speed is low, and a large number of working conditions cannot be tested can be solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of computer software, in particular to a quick simulation method for sediment abrasion of a hydraulic turbine. BACKGROUND

[0002] The water pump turbine is the most core part of the pumped storage technology, and the economic level and efficiency of a power station or power system are largely dependent on the performance of the hydraulic turbine.

[0003] Sediment abrasion of a hydraulic turbine is a serious problem common to hydropower stations in a river with a large amount of sediment. When the water flow through the hydraulic turbine contains a large amount of sediment, the flow components of the hydraulic turbine will be abraded, thereby damaging the flow components of the hydraulic turbine, leading to unstable operation and reduced efficiency of the hydraulic turbine, and affecting the safety and benefits of the unit.

[0004] Currently, the research on the abrasion of the hydraulic turbine usually uses test and numerical simulation methods. The wear test only tests a single runner blade, but the sediment abrasion test under the whole flow passage of the hydraulic turbine is costly and difficult. The numerical simulation technology can qualitatively and quantitatively evaluate the abrasion characteristics of the whole flow passage of the hydraulic turbine. The particle abrasion model accurately predicts the abrasion of the sediment when the hydraulic turbine is running, but to obtain accurate abrasion prediction simulation results, high-performance computing resources and a large amount of time are required, so the current simulation of the sediment abrasion of the hydraulic turbine has the problems of being difficult to predict and slow in sediment abrasion calculation, and cannot test a large number of working conditions. SUMMARY

[0005] The purpose of the present application is to provide a quick simulation method for sediment abrasion of a hydraulic turbine, which uses a simulation process method of sediment abrasion of a hydraulic turbine and neural network reduction, can predict the abrasion of the sediment when the hydraulic turbine is running, and can perform rapid multi-condition calculation, thereby providing guidance for later prevention and optimization design of the abrasion of the hydraulic turbine.

[0006] To achieve the above-mentioned purpose, the present application provides the following scheme:

[0007] A quick simulation method for sediment abrasion of a hydraulic turbine, comprising:

[0008] discretizing a fluid region containing the geometry of the blades of the hydraulic turbine and the geometry of the guide vanes, to obtain a discretized grid;

[0009] performing three-dimensional simulation transient solution on the discretized grid in combination with basic flow, turbulence and sediment corrosion, to obtain basic simulation results and abrasion simulation data results;

[0010] obtaining a working condition parameter combination according to the basic simulation results and the abrasion simulation data results, constructing a sample matrix, performing eigen-orthogonal decomposition on the sample matrix, to obtain a spatial modal matrix and a modal coefficient matrix;

[0011] inputting the new working condition parameter into an adaptive radial basis function to obtain a modal coefficient under the new working condition, wherein the adaptive radial basis function establishes a nonlinear mapping relationship from the working condition parameter to a denoising coefficient, and the denoising coefficient is obtained by denoising the modal coefficient matrix;

[0012] performing data reconstruction according to the modal coefficient under the new working condition and the spatial modal matrix to obtain a physical field prediction result.

[0013] Optionally, the three-dimensional simulation transient solution of the discretized grid in combination with basic flow, turbulence and silt corrosion includes:

[0014] performing water turbine moving vane fluctuation simulation according to a basic flow equation to establish a basic flow simulation model;

[0015] adding a turbulence equation to the basic flow simulation model to model a turbulence phenomenon in the operation process of the water turbine to obtain a flow model;

[0016] establishing an abrasion simulation model according to a silt abrasion equation, and inputting the abrasion simulation model and the flow model into the discretized grid to perform three-dimensional simulation transient solution.

[0017] Optionally, the silt abrasion equation is:

[0018] E=kV n f(γ);

[0019] wherein E is a dimensionless abrasion quality, k is a model constant, V is a silt collision speed, f(γ) is a dimensionless function of silt collision angle γ, and n is an index.

[0020] Optionally, the eigen-orthogonal decomposition of the sample matrix S includes:

[0021] ;

[0022] wherein, is a spatial modal matrix, is a modal coefficient matrix, is an eigenvalue.

[0023] Optionally, the denoising of the modal coefficient matrix includes denoising processing of the modal coefficient matrix by a KAN denoising method:

[0024] ;

[0025] wherein, is a denoised modal coefficient, and θ is a KAN network parameter.

[0026] Optionally, the modal coefficient under the new working condition comprises:

[0027] ;

[0028] wherein p is an input working condition parameter vector, is a modal coefficient under a new working condition, is a radial basis function, is an adaptive bandwidth parameter, and N is a sample number, is an adaptive weight parameter of the jth sample under the kth mode.

[0029] The present application has the advantages that: the present application builds a simulation modeling of sediment abrasion of a water turbine, and establishes a reduced-order model of the simulation of sediment abrasion based on a neural network, which is more than 100 times faster than traditional simulation of sediment abrasion in terms of calculation speed, and can quickly (within 10s) calculate data parameters of sediment abrasion rate and abrasion thickness. It is more suitable for rapid evaluation and prediction on site. BRIEF DESCRIPTION OF DRAWINGS

[0030] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0031] Figure 1 A flowchart of a rapid simulation method of sediment abrasion of a water turbine according to an embodiment of the present application;

[0032] Figure 2 A flowchart of a reduced-order method of simulation of sediment abrasion according to an embodiment of the present application. DETAILED DESCRIPTION

[0033] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0034] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the present application will be further described in detail below with reference to the drawings and specific embodiments.

[0035] As shown in the present embodiment, a rapid simulation method of sediment abrasion of a water turbine is provided, comprising: Figure 1

[0036] ​discretize a fluid region containing the geometry of the turbine blade and the geometry of the guide vane to obtain a discretized grid;

[0037] discretize a fluid region containing the geometry of the turbine blade and the geometry of the guide vane to obtain a discretized grid;

[0038] perform three-dimensional simulation transient solution on the discretized grid in combination with basic flow, turbulence, and silt corrosion to obtain basic simulation results and silt corrosion simulation data results;

[0039] obtain a working condition parameter combination according to the basic simulation results and the silt corrosion simulation data results, construct a sample matrix, perform intrinsic orthogonal decomposition on the sample matrix, and obtain a spatial mode matrix and a mode coefficient matrix;

[0040] input new working condition parameters into an adaptive radial basis function to obtain mode coefficients under the new working condition, wherein the adaptive radial basis function establishes a nonlinear mapping relationship from the working condition parameters to the denoising coefficients, and the denoising coefficients are obtained by denoising the mode coefficient matrix;

[0041] reconstruct data according to the mode coefficients under the new working condition and the spatial mode matrix to obtain physical field prediction results.

[0042] Further, performing three-dimensional simulation transient solution on the discretized grid in combination with basic flow, turbulence, and silt corrosion includes:

[0043] performing turbine moving guide vane fluctuation simulation according to a basic flow equation to establish a basic flow simulation model;

[0044] adding a turbulence equation to the basic flow simulation model to model turbulence phenomena in the operation process of the turbine to obtain a flow model;

[0045] establishing a silt corrosion simulation model according to a silt corrosion equation, and inputting the silt corrosion simulation model and the flow model into the discretized grid to perform three-dimensional simulation transient solution.

[0046] Specifically, the turbine silt corrosion simulation includes:

[0047] (1) discretize a fluid region containing the geometry of the turbine blade and the geometry of the guide vane to obtain a discretized grid according to the three-dimensional fluid domain geometry of the turbine moving guide vane using CFD software PERASIM Fluid. The discretized grid is used for inputting physical models and calculation parameters for simulation.

[0048] (2) perform turbine moving guide vane fluctuation simulation to establish a basic flow simulation model. The established model and parameters are input into the discretized grid for subsequent simulation calculation.

[0049] The equation is as follows: ;

[0050] Specifically, p is the density of the fluid, t is time, is the velocity vector field of the fluid, ∇⋅ is the divergence operator, ∇ is the gradient operator, p is the pressure field inside the fluid, τ is the viscous stress tensor, g is the gravity acceleration vector (or other body force), is the convection acceleration term.

[0051] (3) Add the turbulence equation to the basic flow simulation model to model the turbulence phenomenon in the operation process of the hydraulic turbine for capturing pressure pulsation in simulation. The turbulence equation is as follows:

[0052] Turbulent kinetic energy (k) equation: ;

[0053] Specific dissipation rate (ω) equation: ;

[0054] Where ∂(p) / ∂t is a non-steady term, is a convection term, P k is a turbulent kinetic energy generation term, D k is a turbulent kinetic energy dissipation term, is a turbulent kinetic energy diffusion term, P ω is a ω generation term, D ω is a ω dissipation term, is a ω diffusion term, and CD is a cross-diffusion term.

[0055] (4) Perform hydraulic turbine sediment erosion modeling:

[0056] After the flow model is completed, the erosion simulation model is modeled and coupled, and the specific steps are as follows:

[0057] a. Activate the DPM (Discrete Particle Model) model;

[0058] b. Define the sediment inlet boundary (position, flow rate, particle size distribution, etc.) and the action parameters of the sediment and the wall surface;

[0059] c. Set the sediment calculation control method;

[0060] d. Calculate the sediment motion trajectory;

[0061] e. Set the erosion boundary at the key position of the hydraulic turbine and select the erosion calculation model as follows:

[0062] Finnie erosion model equation: E=kV n f(γ);

[0063] Where: E is the dimensionless erosion mass; k is the model constant; V is the sediment collision speed; f(γ) is a dimensionless function of the sediment collision angle γ; and the value of the exponent n is in the range of 2.3-2.5. , .

[0064] (5) In the PERASIM Fluid software, combined with the basic flow equation, the turbulent flow equation, and the sediment erosion equation, the discretized grid in step (1) is simulated and solved in three dimensions using the coupling solution method of COUPLE to obtain the basic simulation results of the flow rate, pressure, and other basic simulation results of the water turbine, and the erosion simulation data results of the sediment erosion rate and the erosion amount.

[0065] Further, as shown in Figure 2 , a reduced-order modeling is performed to construct a fast calculation sediment erosion simulation reduced-order model, including:

[0066] The method of the embodiment is based on the combination of POD-KAN and ARBF algorithms, and through decomposition, denoising, and prediction of high-fidelity numerical simulation data, efficient modeling and rapid prediction of the complex sediment erosion process are realized, and the specific steps are as follows:

[0067] Step one, define the physical quantity to be studied:

[0068] According to the actual engineering requirements and the erosion mechanism analysis, the physical quantity to be studied is determined, which can include but is not limited to: pressure field distribution, wall erosion depth, flow velocity field, sediment concentration distribution, turbulent kinetic energy distribution, and other key physical quantities. These physical quantities collectively represent the fluid-particle-wall multi-field coupling behavior in the erosion process.

[0069] Step two, build a high-reliability sample data set:

[0070] The Latin hypercube sampling (LHS) method is used to determine the working condition parameter combination with good space filling and projection uniformity in the design space under the constraint of considering the correlation between parameters. Each sampling point contains key parameters such as flow velocity, sediment median particle size, particle concentration distribution, impact angle, and material hardness. For each sampling working condition, a high-fidelity CFD-DEM coupled numerical simulation verified by experiments is performed:

[0071] The numerical model verification needs to ensure grid independence through grid convergence index (GCI) analysis;

[0072] By comparing with the measurement results of classical experimental cases (such as ASME standard erosion cases) or special experimental tables, the accuracy of the numerical model is verified, and the average relative error of the key parameters is required to be less than 5%;

[0073] The finally constructed high-quality sample data set is represented as a sample matrix , where N is the number of samples, and M is the number of physical field discrete points.

[0074] Step three, POD decomposition and modal truncation criterion:

[0075] Perform the proper orthogonal decomposition (POD) on the sample matrix S:

[0076] ;

[0077] where, is the spatial mode matrix, is the modal coefficient matrix, is the eigenvalue.

[0078] Use the modal stage criterion based on energy contribution rate:

[0079] ;

[0080] where, is the eigenvalue, and η is the energy retention threshold (usually 99%-99.9%). Determine the optimal modal number K through this criterion to achieve data dimensionality reduction while ensuring accuracy.

[0081] Step four, KAN denoising processing of data modal coefficients:

[0082] Retain the spatial mode Φ obtained by POD decomposition, and implement Kolmogorov-Arnold Network (KAN) based denoising processing on the modal coefficient matrix C. KAN network learns the nonlinear noise structure of the data adaptively, and filters and smooths the coefficient matrix:

[0083] ;

[0084] where is the denoised modal coefficient, and θ is the KAN network parameter. This step effectively suppresses high-frequency noise and abnormal disturbances generated in numerical simulation or measurement.

[0085] Step five, ARBF prediction of modal coefficients:

[0086] Use the adaptive radial basis function (ARBF) method to establish a nonlinear mapping relationship from the working condition parameters to the denoised modal coefficients. ARBF dynamically adjusts the shape parameters and weights of the radial basis function to enhance the adaptability to different regions of samples:

[0087] ;

[0088] where, p is the input working condition parameter vector, is the radial basis function (such as Gaussian kernel function), is the adaptive bandwidth parameter, which is dynamically optimized through maximum likelihood estimation or cross-validation, is the adaptive weight parameter of the jth sample under the kth mode. After training, the ARBF model can quickly predict the modal coefficients under new working conditions .

[0089] Step six, data reconstruction:

[0090] The modal coefficients under the new working condition are combined with the POD space modal to reconstruct the physical field prediction result:

[0091] ;

[0092] The reconstructed is the fast prediction value of the physical quantity under the new working condition. The reduced order model greatly reduces the calculation time while maintaining the accuracy, and is suitable for multiple iterations in engineering design and real-time control.

[0093] Further, the simulation result post-processing:

[0094] The above reduced order model is run and calculated, and the post-processing of the calculated results can realize the extraction and display of point, line and surface data, has multiple flow field result visualization functions such as streamline diagram, vector diagram and curve diagram, and provides functions such as user-defined field variable calculation, result dynamic preview and animation production.

[0095] Through these functions, the internal wear law of the water turbine under different operating conditions is realized, the wear cloud diagram of the runner and guide vane blades and the wear cloud diagram of the draft tube wall are obtained, the influence of different silt contents on the water turbine is further studied, and then the structure of the water turbine is optimized, the erosion is effectively reduced, and the safety, stability and economy of the water turbine generator set operation are improved.

[0096] The above-described embodiments are only descriptions of the preferred modes of the present application and do not limit the scope of the present application. Without departing from the design spirit of the present application, various modifications and improvements to the technical solutions of the present application made by those skilled in the art shall fall within the protection scope determined by the claims of the present application.

Claims

1. A quick simulation method of sediment abrasion of a hydraulic turbine, characterized in that, The method comprises the following steps: discretize a fluid region comprising a turbine blade geometry and a guide vane geometry to obtain a discretized grid; perform three-dimensional simulation transient solution on the discretized grid in combination with basic flow, turbulence and sediment corrosion to obtain basic simulation results and abrasion simulation data results; obtain a working condition parameter combination according to the basic simulation results and the abrasion simulation data results, construct a sample matrix, perform eigenvalue orthogonal decomposition on the sample matrix to obtain a spatial mode matrix and a mode coefficient matrix; input a new working condition parameter into an adaptive radial basis function to obtain a mode coefficient under the new working condition, wherein the adaptive radial basis function establishes a nonlinear mapping relationship from the working condition parameter to a denoising coefficient, and the denoising coefficient is obtained by denoising the mode coefficient matrix; perform data reconstruction according to the mode coefficient under the new working condition and the spatial mode matrix to obtain a physical field prediction result.

2. The method of claim 1, wherein, The three-dimensional simulation transient solution on the discretized grid in combination with basic flow, turbulence and sediment corrosion comprises the following steps: perform turbine moving guide vane fluctuation simulation according to a basic flow equation to establish a basic flow simulation model; add a turbulence equation to the basic flow simulation model to model turbulence phenomena in the operation process of the turbine to obtain a flow model; establish an abrasion simulation model according to a sediment abrasion equation, and input the abrasion simulation model and the flow model into the discretized grid to perform three-dimensional simulation transient solution.

3. The method of claim 2, wherein, The sediment abrasion equation is as follows: E = kV n f(y); wherein E is a dimensionless abrasion mass; k is a model constant; V is a sediment collision speed; f(γ) is a dimensionless function of a sediment collision angle γ, and n is an index.

4. The method of claim 1, wherein, The eigenvalue orthogonal decomposition on the sample matrix S comprises the following steps: ； wherein, is the spatial modal matrix, is the modal coefficient matrix, is the eigenvalue.

5. The method of claim 4, wherein, The denoising of the mode coefficient matrix comprises denoising processing of the mode coefficient matrix by a KAN denoising method: ； wherein, are the denoised modal coefficients, and θ are the KAN network parameters.

6. The method of claim 1, wherein, The mode coefficient under the new working condition comprises: ； wherein p is an input working condition parameter vector, is a modal coefficient under a new working condition, is a radial basis function, is an adaptive bandwidth parameter, N is a sample number, is an adaptive weight parameter of the jth sample under the kth mode.

Images