Methods, systems, equipment and media for optimizing the design of turbine guide vane structures

Abstract

This invention relates to the field of hydro-turbine design and simulation technology, specifically to a method, system, equipment, and medium for optimizing the design of hydro-turbine guide vane structures. The method includes: constructing and solving a multi-physics coupled simulation model of the guide vane's working environment to obtain multi-condition performance data; based on the multi-condition performance data, combining adaptive experimental design and uncertainty quantification analysis, constructing a hybrid surrogate model to predict the statistical characteristics of guide vane performance; using the hybrid surrogate model to perform multi-objective collaborative optimization of the design parameter set of the guide vane and sealing system, generating a robust optimization solution set; based on the robust optimization solution set and the guide vane dynamic characteristics extracted from the multi-condition performance data, performing dynamic modeling and parameter calibration of the sealing system; performing simulation verification and sensitivity analysis on the robust optimization solution set, and outputting parametric design results. The aim is to improve the simulation fidelity, optimization efficiency, and robustness of the optimized design of hydro-turbine guide vane structures.

Description

Technical Field

[0001] This invention relates to the field of water turbine design and simulation technology, specifically to water turbine guide vane structure optimization design methods, systems, equipment, and media. Background Technology

[0002] As a core component of the guide vane mechanism, the performance of the turbine's movable guide vane directly affects the overall efficiency, operational stability, and reliability of the turbine. Medium- and high-head mixed-flow turbines operate in harsh environments, with guide vanes subjected to the strong coupling effects of multiple physical fields, including high transient water pressure, cavitation erosion, sediment erosion, and thermal loads. Their design faces multiple challenges, including structural strength, dynamic sealing, cavitation resistance, and wear resistance. Currently, the optimization design of guide vanes mainly relies on Computer-Aided Engineering (CAE) technology, employing a sequential, unidirectional coupling simulation strategy. For example, pressure loads obtained from steady-state or quasi-steady-state flow field analysis are applied to a static structural model for verification. However, this method severs the inherent real-time interaction and bidirectional feedback mechanism between the fluid, structure, and thermal physical fields during modeling. This leads to significant deviations between the simulation results and the actual mechanical behavior, clearance dynamics, and failure mechanisms of the guide vane under complex transient conditions, resulting in insufficient fidelity of the simulation model upon which the design is based.

[0003] Existing technologies employ sophisticated simulation methods such as transient fluid-structure interaction to achieve higher fidelity. However, these methods are computationally expensive per run, making global exploration and optimization within a multi-variable, multi-constraint, and multi-objective design space difficult to achieve within engineering timeframes. The optimization process heavily relies on the designer's experience and local optimization. On the other hand, conventional optimization is usually performed under deterministic assumptions, failing to systematically account for the inherent fluctuations and uncertainties of key operating parameters such as head, sediment content, and inlet turbulence intensity. This results in the optimal design scheme potentially experiencing drastic performance degradation or even failure when actual operating conditions deviate from the design point, lacking the robustness required for practical engineering applications. Summary of the Invention

[0004] To improve the simulation fidelity, optimization efficiency, and robustness of the optimized design of turbine guide vane structures, this invention provides a method, system, equipment, and medium for the optimized design of turbine guide vane structures. The specific technical solution adopted is as follows:

[0005] The first aspect of the present invention provides a method for optimizing the design of a turbine guide vane structure, the method comprising:

[0006] A multi-physics coupled simulation model of the guide vane's working environment was constructed and solved to obtain performance data of the guide vane under multiple working conditions.

[0007] Based on multi-condition performance data, combined with adaptive experimental design and uncertainty quantification analysis, a hybrid surrogate model for predicting the statistical characteristics of guide vane performance is constructed.

[0008] A hybrid surrogate model is used to perform multi-objective collaborative optimization of the design parameter set of the guide vane and sealing system, generating a robust optimization solution set;

[0009] Based on the robust optimization solution set and the guide vane dynamic characteristics extracted from the multi-condition performance data, the sealing system is dynamically modeled and its parameters are calibrated.

[0010] The robust optimization solution set is verified by simulation and subjected to sensitivity analysis, and the parametric design results are output.

[0011] Furthermore, a multiphysics coupled simulation model of the guide vane's working environment is constructed and solved to obtain multi-condition performance data of the guide vane, including:

[0012] Establish a set of multiphysics coupled control equations that include fluid dynamics control equations, structural dynamics control equations, heat conduction equations, and erosion rate equations;

[0013] Based on the Jacobian matrix, which characterizes the coupling relationship between the fluid field, structural field, thermal field, and erosion field, a relaxation iterative algorithm is used to solve the multiphysics field coupled control equations simultaneously.

[0014] Extract multi-condition performance data of the guide vane, including guide vane stress distribution, deformation field, end face leakage velocity field, and cavitation risk distribution, from the solution results.

[0015] Furthermore, by combining adaptive experimental design and uncertainty quantification analysis, a hybrid surrogate model for predicting the statistical characteristics of guide vane performance is constructed, including:

[0016] Define a joint input space that includes guide vane body design variables, sealing system design variables, and operating condition uncertainty parameters;

[0017] An initial sample point set is generated within the joint input space using the optimal Latin hypercube sampling method.

[0018] For the initial sample point set, the steps of constructing and solving the multiphysics coupling simulation model of the guide vane working environment are performed to obtain the corresponding performance response data, including multiple guide vane performance indicators.

[0019] A hybrid surrogate model is trained based on performance response data to predict the statistical characteristics of guide vane performance.

[0020] Furthermore, based on performance response data, a hybrid surrogate model is trained to predict the statistical characteristics of guide vane performance, including:

[0021] Based on the performance response data, a hybrid surrogate model that integrates the Kriging model and the multinomial chaotic expansion is trained; wherein the Kriging model is used to establish a nonlinear mapping relationship between the performance response and all input variables, and the multinomial chaotic expansion is used to analyze the influence of the uncertainty parameters of the operating conditions on the statistical distribution of the performance response.

[0022] The covariance function parameters of the Kriging model were calibrated using the maximum likelihood estimation method;

[0023] The coefficients of the polynomial chaotic expansion are solved using a regression method.

[0024] Furthermore, a hybrid surrogate model is used to perform multi-objective collaborative optimization of the design parameter set of the guide vane and sealing system, generating a robust optimization solution set, including:

[0025] Based on the hybrid agent model, the predicted mean and variance of multiple performance indicators under a given design are calculated.

[0026] Based on the predicted mean and variance, construct an optimization objective function for each performance index;

[0027] A multi-objective evolutionary algorithm is employed to search within the design space comprised of guide vane body geometric design variables and sealing system stiffness design variables, thereby simultaneously optimizing multiple objective functions and obtaining the robust optimization solution set composed of non-dominated solutions.

[0028] Furthermore, based on the robust optimization solution set and the dynamic characteristics of the guide vane, the sealing system is dynamically modeled and its parameters are calibrated, including:

[0029] A dynamic equation describing the motion of the sealing element is established, which includes a nonlinear contact force term determined by the relative displacement and relative velocity between the sealing element and the guide vane end face.

[0030] Using the guide vane end face displacement time history in the dynamic characteristics of the guide vane as the input of the dynamic equation, the following error between the motion response of the sealing element and the guide vane end face displacement time history is minimized by adjusting the stiffness parameter and pre-compression parameter of the sealing system.

[0031] Furthermore, the robust optimization solution set is verified by simulation and subjected to sensitivity analysis, and the parametric design results are output, including:

[0032] At least one candidate design scheme is selected from the robust optimization solution set;

[0033] For the candidate design scheme, perform the steps of constructing a multiphysics coupled simulation model of the guide vane working environment and solving it to obtain performance data for verification;

[0034] Based on the polynomial chaotic expansion part in the hybrid agent model, the contribution of each operating condition uncertainty parameter to the variance of key performance indicators is calculated and used as a global sensitivity index.

[0035] The final design parameters determined through the simulation verification and sensitivity analysis are input into the parametric geometric modeling script to generate a three-dimensional model of the guide vane and sealing device.

[0036] The output includes a design document containing the final design parameters, the calibrated dynamic parameters of the sealing system, and the predicted range of key performance indicators.

[0037] The second aspect of the present invention provides a turbine guide vane structure optimization design system for executing the turbine guide vane structure optimization design method described in the first aspect of the present invention, the system comprising:

[0038] The simulation module is configured to build and solve a multi-physics coupled simulation model of the guide vane's working environment to obtain multi-condition performance data of the guide vane.

[0039] The hybrid proxy module is configured to construct a hybrid proxy model for predicting the statistical characteristics of guide vane performance based on multi-condition performance data, combined with adaptive experimental design and uncertainty quantification analysis.

[0040] The collaborative optimization module is configured to use a hybrid surrogate model to perform multi-objective collaborative optimization of the design parameter set of the guide vane and sealing system, and generate a robust optimization solution set.

[0041] The parameter calibration module is configured to perform dynamic modeling and parameter calibration of the sealing system based on the robust optimization solution set and the dynamic characteristics of the guide vane.

[0042] The output module is configured to perform simulation verification and sensitivity analysis on the robustness optimization solution set, and output the parametric design results.

[0043] The third aspect of the present invention provides an electronic device, the electronic device comprising: a processor and a memory communicatively connected to the processor; wherein the memory stores instructions executable by the processor, the instructions being executed by the processor to enable the processor to perform the steps of the turbine guide vane structure optimization design method described in the first aspect of the present invention.

[0044] The fourth aspect of the present invention provides a computer-readable storage medium storing a program for implementing a method for optimizing the design of a turbine guide vane structure. The program for implementing the method for optimizing the design of a turbine guide vane structure is executed by a processor to implement the steps of the method for optimizing the design of a turbine guide vane structure as described in the first aspect of the present invention.

[0045] The present invention has the following beneficial effects:

[0046] The turbine guide vane structure optimization design method provided by this invention constructs and solves a multi-physics coupled simulation model of the guide vane's working environment. This captures the real-time interactions between multiple physics fields, avoiding the distortion problems caused by traditional sequential unidirectional coupling analysis, thereby improving simulation accuracy and fidelity. Furthermore, a hybrid surrogate model built based on high-fidelity simulation data integrates adaptive experimental design and uncertainty quantification analysis, enabling efficient prediction of performance statistical characteristics. This transforms complex multi-physics simulations into rapidly evaluable surrogate models, solving the technical problems of high computational cost and low efficiency when directly calling high-fidelity simulations for optimization. Based on this, the hybrid surrogate model is used to perform multi-objective collaborative optimization of the guide vane and sealing system design parameter set. The design stage systematically considers operational uncertainties, generating a robust optimization solution set, ensuring that the design scheme maintains stable performance even under fluctuations in actual operating parameters. Simultaneously, based on the optimization solution set and the dynamic characteristics of the guide vane, integrated dynamic modeling and parameter calibration of the sealing system are performed, enabling the sealing design to actively adapt to the dynamic deformation of the guide vane, thus suppressing end-face leakage from a mechanistic perspective. Finally, through simulation verification and sensitivity analysis of the optimized solution set, and output of parametric design results, a method for optimizing the design of turbine guide vane structures was constructed, which ranges from high-precision analysis, efficient optimization, robust design to reliable verification. This method achieves a synergistic improvement in simulation fidelity, optimization efficiency and robustness of guide vane structure optimization design. Attached Figure Description

[0047] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0048] Figure 1 This is a flowchart of a method for optimizing the design of a turbine guide vane structure according to an embodiment of the present invention;

[0049] Figure 2 This is a schematic diagram of the structure of a turbine guide vane structure optimization design system provided in one embodiment of the present invention. Detailed Implementation

[0050] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation methods, structures, features, and effects of a turbine guide vane structure optimization design method, system, equipment, and medium proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0051] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0052] The following description, in conjunction with the accompanying drawings, details the specific scheme of the method, system, equipment, and medium for optimizing the design of a turbine guide vane structure provided by this invention.

[0053] Please see Figure 1 The diagram illustrates a flowchart of a method for optimizing the design of a turbine guide vane structure according to an embodiment of the present invention. The method includes:

[0054] Step S100: Construct a multi-physics coupled simulation model of the guide vane's working environment and solve it to obtain multi-condition performance data of the guide vane;

[0055] Step S100 specifically includes:

[0056] Step S110: Establish a set of multiphysics coupled control equations, including fluid dynamics control equations, structural dynamics control equations, heat conduction equations, and erosion rate equations.

[0057] In some embodiments, the Navier-Stokes equations based on large eddy simulation filtering are employed and coupled with the Zwart cavitation model to predict cavitation phase transitions. The hydrodynamic governing equations can be expressed as follows:

[0058]

[0059] In the formula, and They represent in the Cartesian coordinate system and Filtered fluid velocity components in the direction of motion. ,correspond direction; Indicates time; This indicates the fluid pressure after filtering; Indicates fluid density; Indicates the kinematic viscosity of a fluid; The components of the subgrid-scale stress tensor represent the influence of small-scale turbulent motions that are not directly analyzed by the grid due to filtering operations on large-scale flows. The cavitation source term, defined by the Zwart cavitation model, describes the vaporization and condensation process of liquid caused by local pressure changes, i.e., the rate of change of the volume fraction of cavitation per unit time.

[0060] In some embodiments, the deformation and vibration of the guide vane are described using an elastic dynamics equation that considers geometric nonlinearity, and the structural dynamics governing equation can be expressed as:

[0061]

[0062] In the formula, Indicates the density of the guide vane material; The displacement vector field representing the guide vane structure is specifically a function of space and time, describing the displacement of each point on the guide vane in three directions; This represents the Cauchy stress tensor, which is used to describe the stress state at any point inside the guide vane; This represents the external volume force vector, specifically the centrifugal force generated by the rotation of the guide vane;

[0063] In some embodiments, a transient heat conduction equation considering convective heat transfer and internal heat sources is adopted, and the heat conduction equation can be expressed as:

[0064]

[0065] In the formula, This indicates the specific heat capacity of the guide vane material; Represents the temperature field; This indicates the thermal conductivity of the guide vane material; This indicates the intensity of the volumetric heat source generated by frictional heat between the guide vane shaft diameter and the bearing. The heat generated by the viscosity dissipation of water flow within the boundary layer of the guide vane wall is added to the equation as an equivalent volumetric heat source.

[0066] In some embodiments, a discrete phase model is used to track sediment particles, and the Finnie erosion model is used to calculate the material loss rate. The erosion rate equation can be expressed as:

[0067]

[0068] In the formula, It represents the local material erosion rate, that is, the mass of material eroded per unit area per unit time; Indicates a single sediment particle; and It is an empirical coefficient related to the guide vane material and the properties of erosion particles, and can be calibrated experimentally. This indicates the mass of a single sediment particle; This represents the normal velocity component when sediment particles impact the wall surface. Represents the impact angle function, where It is the angle between the particle velocity vector and the tangent of the wall;

[0069] Step S120: Based on the Jacobian matrix characterizing the coupling relationship between the fluid field, structural field, thermal field and erosion field, the relaxation iterative algorithm is used to solve the multiphysics field coupling control equations simultaneously.

[0070] This step involves numerically solving the nonlinear strongly coupled equations established in step S110. The implementation process is as follows:

[0071] Step S121: Discrete the governing equations of the four physical fields—fluid, structure, heat, and erosion—in time step... Combined, forming a system with the state variables of the entire field. Nonlinear residual equations with unknowns ;in, These represent the sets of state variables for fluid, structure, heat, and erosion fields, respectively.

[0072] Step S122: Iteratively solve using the quasi-Newton method; in the... In the next iteration, the current iteration point is constructed. Approximate Jacobian matrix at ... This matrix is ​​a block matrix that contains not only the derivatives of the internal variables of each physical field with respect to their own residuals, but also inter-field coupling terms, i.e., off-diagonal blocks, such as... This describes the sensitivity of the flow field residuals to changes in structural displacement, i.e., the fluid-structure interaction strength. These coupling terms can be obtained through automatic differentiation or finite difference perturbation methods.

[0073] Step S123: Solve the system of linear equations Obtain the increment of the state variable Introduce a relaxation factor , Update the state variable as follows:

[0074]

[0075] The relaxation factor is used to stabilize the iterative process of strongly nonlinear problems and prevent divergence. This iterative process continues until the full-field residuals are obtained. The norm is less than the set convergence tolerance, thus achieving synchronous and strongly coupled solution of the four physical field states.

[0076] Step S130: Extract multi-condition performance data of the guide vane, including stress distribution, deformation field, end-face leakage velocity field, and cavitation risk distribution, from the solution results. Specifically, extract stress distribution and deformation field data of the guide vane from the structural field solution results to reflect the structural mechanical characteristics of the guide vane under multiple conditions; extract end-face leakage velocity field data and cavitation risk distribution data of the guide vane from the fluid field solution results to characterize the hydraulic performance and cavitation hazards of the guide vane; combine the erosion field solution results to extract the erosion rate data of the guide vane surface to supplement the relevant information on the wear resistance performance of the guide vane. In this embodiment, all extracted data correspond to different head and load conditions, forming a multi-condition performance dataset covering typical operating conditions of the guide vane, providing basic data support for subsequent hybrid proxy model construction and optimization design.

[0077] Step S100 establishes and solves a simulation model involving strong coupling of four fields: fluid, structure, heat, and erosion. This abandons the simplified approach of traditional sequential unidirectional analysis, enabling real-time capture of the entire interaction process in numerical simulation: the deformation of the guide vane under water pressure, the deformation's influence on the flow field and pressure distribution, friction and viscous heat generation, and dynamic damage from sediment erosion. This step produces high-fidelity simulation results, providing a reliable data foundation that accurately reflects the complex physical reality for subsequent design optimization, and resolving the simulation distortion problem caused by neglecting multi-field coupling effects.

[0078] Step S200: Based on multi-condition performance data, combined with adaptive test design and uncertainty quantification analysis, construct a hybrid surrogate model to predict the statistical characteristics of guide vane performance;

[0079] Step S200 specifically includes:

[0080] Step S210: Define a joint input space containing guide vane body design variables, sealing system design variables, and operating condition uncertainty parameters. The joint input space consists of three parts of variables: guide vane body design variables, such as spline curve control point coordinates describing the airfoil shape, blade thickness, and other geometric parameters; sealing system design variables, including dynamic parameters such as the equivalent stiffness, damping, and pre-compression of the sealing strip; and operating condition uncertainty parameters, which refer to physical quantities that naturally fluctuate during actual operation, such as operating head, water flow sediment content, and inlet turbulence intensity. These variables are integrated into a unified joint input space, where the design variables have feasible upper and lower limits, and the uncertainty parameters are described by their respective probability distribution characteristics.

[0081] Step S220: Generate an initial sample point set within the joint input space using the optimal Latin hypercube sampling method. This embodiment designs experiments within the joint input space and uses the optimal Latin hypercube sampling method to generate the initial sample point set. The aim is to ensure that the value range of each input variable is uniformly divided, and only one sample value is extracted within each small interval, thereby achieving uniform coverage in a single variable dimension. Subsequently, the positions of these sample points in the multidimensional space are adjusted through an optimization algorithm to maximize the minimum distance between any two sample points, thus obtaining a sample set that is distributed as uniformly and dispersedly as possible within the overall design space.

[0082] Step S230: For the initial sample point set, execute the step of constructing and solving the multiphysics coupled simulation model of the guide vane working environment to obtain the corresponding performance response data, including multiple guide vane performance indicators; for each sample point in step S220, drive and execute the high-fidelity full-physics coupled simulation program established in step S100 for the specific design variables and operating parameter combinations. After each simulation calculation is completed, a set of predefined key performance indicator data is extracted to form the performance response of that sample point. These performance indicators are directly related to the core functions and reliability of the guide vane, such as turbine efficiency, guide vane end face leakage flow, guide vane body maximum equivalent stress, fatigue life prediction value, and cavitation or erosion rate in key areas. Finally, a performance response dataset corresponding one-to-one with the initial sample point set is obtained.

[0083] Step S240: Based on performance response data, train a hybrid surrogate model to predict the statistical characteristics of guide vane performance, including:

[0084] Step S241: Based on the performance response data, train a hybrid surrogate model that integrates the Kriging model and the multinomial chaotic expansion; wherein, the Kriging model is used to establish a nonlinear mapping relationship between the performance response and all input variables, and the multinomial chaotic expansion is used to analyze the influence of the uncertainty parameters of the operating conditions on the statistical distribution of the performance response; in some embodiments, for the first... Performance metrics Construct the following hybrid prediction model:

[0085]

[0086] In the formula, Indicates the hybrid agent model for the first Predicted values ​​for each performance indicator; This represents the vector of design variables for the guide vane body. Represents the design variable vector of the sealing system; Represents a vector of parameters indicating uncertainties in the operating conditions; The general Kriging model predicts a linear combination of known sample response values, with weights determined by spatial correlation; its covariance function uses the Matérn5 / 2 kernel function. Let represent a polynomial chaotic expansion, where is the number of parameters with respect to the uncertainty. The sum of weighted orthogonal polynomials; , is a multi-exponential expression that represents the powers of the variables in a polynomial; This represents the multivariate orthogonal polynomial basis function selected based on the probability distribution of the uncertainty parameter. For example, Hermite polynomials are used when the uncertainty parameter follows a Gaussian distribution, and Legendre polynomials are used when it follows a uniform distribution. Indicates the first Individual performance indicators and multiple indices The corresponding PCE coefficient is a constant to be determined, used to analyze the degree of influence of the uncertainty parameter on the statistical distribution of the performance response; The selected multi-index set is used. The formula first fits the global nonlinear trend between the joint input and the performance response through the Kriging model, and then separates and quantifies the contribution of the uncertainty parameter to the performance fluctuation through PCE. The two work together to achieve synchronous prediction of the mean and variance of the performance response.

[0087] Step S242: Use the maximum likelihood estimation method to calibrate the covariance function parameters of the Kriging model; specifically, use the maximum likelihood estimation method to calibrate the covariance function parameters of the Kriging model. The parameters to be calibrated include the process variance and the correlation length of each dimension variable.

[0088] In some embodiments, based on the initially constructed joint input-performance response dataset, a likelihood function for the Kriging model is constructed to reflect the probability of observing the current performance response data under given parameter conditions. The logarithm of the likelihood function is taken to obtain the log-likelihood function, and the maximum point of this log-likelihood function is found using a numerical optimization algorithm. The parameter corresponding to this maximum point is the optimal calibration parameter for the covariance function. After calibration, the Kriging model can accurately capture the nonlinear relationship between the joint input variables and the performance response, providing a reliable global trend fitting basis for the hybrid surrogate model.

[0089] Step S243: Solve for the coefficients of the polynomial chaotic expansion using a regression method. Specifically, in this embodiment, least squares regression is preferred. First, based on the uncertainty parameters in the initial sample points, calculate the orthogonal polynomial basis function values ​​corresponding to each sample point, and construct a complete basis function matrix based on the basis function values. Then, construct the performance response vector, calculate the predicted values ​​of the Kriging model at each sample point, and obtain the residual vector through the difference between the performance response vector and the predicted values ​​of the Kriging model. Finally, construct and solve the optimization problem by minimizing the error of the product of the residual vector and the basis function matrix to obtain the coefficients of the polynomial chaotic expansion.

[0090] In summary, step S200 constructed a hybrid surrogate model integrating the Kriging model and multinomial chaotic expansion. First, a high-dimensional input space encompassing design variables and uncertainty parameters was precisely defined, and optimal Latin hypercube sampling was employed to obtain the most informative training samples with minimal high-fidelity simulation cost. Then, using these sample data, maximum likelihood estimation and regression analysis were used to calibrate the Kriging model parameters that capture global nonlinear trends and the PCE coefficients that analytically represent the impact of uncertainty, respectively, thereby training a surrogate model capable of simultaneously and accurately predicting both the mean and variance of performance. This model fundamentally solves the bottleneck problems of excessively high computational cost and low optimization efficiency that exist when directly relying on high-fidelity fully coupled simulation in step S100 for optimization. Furthermore, it embeds the inherent volatility of actual operating conditions into the model in a quantifiable manner, laying the computational and analytical foundation for subsequent efficient and robust multi-objective collaborative optimization.

[0091] Step S300: Use a hybrid surrogate model to perform multi-objective collaborative optimization on the design parameter set of the guide vane and sealing system to generate a robust optimization solution set;

[0092] Step S300 specifically includes:

[0093] Step S310: Based on the hybrid proxy model, calculate the predicted mean and variance of multiple performance indicators under a given design; for any given design scheme, the corresponding guide vane body and sealing system design parameters are input into the model;

[0094] In some embodiments, for a specific design point defined by a vector of design variables and a vector of uncertainty parameters, its first... Performance metrics The predicted value is a random variable. The hybrid surrogate model, consisting of the Kriging model and multinomial chaos expansion, can analytically derive the statistical moments of the predicted distribution of this performance index.

[0095] The predicted mean characterizes the average or expected performance of the design under uncertain operating conditions, and can be expressed as:

[0096]

[0097] In the formula, Indicates the first Performance metrics in design variables The predicted mean reflects the average performance level; Indicates the first The constant term coefficients corresponding to each performance index; No. The mean of the Kriging model predictions for each performance metric is determined by the trend function of the Kriging model.

[0098] The predictor quantifies the dispersion of performance indicators caused by random fluctuations in operating parameters, which can be expressed as:

[0099]

[0100] In the formula, Indicates the first Performance metrics in design variables The predicted variance under the given conditions characterizes the degree to which performance is affected by fluctuations in uncertain parameters; For multiple indicators The corresponding polynomial chaotic expansion coefficients; Indicators and multiple indicators The corresponding multivariate orthogonal polynomial basis functions; Describing basis functions The inner product of the basis functions is the normalization constant of the basis functions, which is determined by the type of basis functions and the distribution of uncertainty parameters, ensuring the dimensional consistency of variance calculation.

[0101] Step S320: Based on the predicted mean and variance, construct an optimization objective function for each performance index; specifically, based on the predicted mean and standard deviation of each performance index calculated in step S310, construct a robust optimization objective function for each index.

[0102] In some embodiments, the optimization objective function does not simply pursue the average performance level, but rather combines the performance mean with the standard deviation representing volatility through a weighting factor with a preset risk aversion coefficient, forming a comprehensive scalar objective. The preset risk aversion coefficient can be set by the designer based on engineering experience and requirements for different performance stability. For indicators that are desired to be as small as possible, such as leakage, the objective is to minimize the comprehensive function of the weighted sum of the performance mean and variance; for indicators that are desired to be as large as possible, such as fatigue life, maximization is achieved by minimizing the negative value of its comprehensive function. In this way, the optimization process is guided to simultaneously pursue a better performance center value and a smaller performance dispersion, thereby enabling the final design to inherently adapt to changes in actual operating conditions.

[0103] Step S330: A multi-objective evolutionary algorithm is employed to search within the design space comprised of the guide vane body geometry design variables and the sealing system stiffness design variables, simultaneously optimizing multiple objective functions to obtain the robust optimal solution set composed of non-dominated solutions. This embodiment uses a third-generation non-dominated sorting genetic algorithm. First, the algorithm randomly generates an initial population containing numerous potential design schemes within the allowed range of variables. Subsequently, for each design scheme in the population, a hybrid surrogate model is used to quickly evaluate all its robust objective function values. Based on the robust objective function values, the algorithm performs hierarchical sorting of the population individuals according to Pareto dominance, identifying those Pareto optimal solutions that are not dominated by other schemes in all objectives, and placing them at the first frontier. Simultaneously, the algorithm calculates the local crowding of each solution in its objective space to maintain the diversity of the solution set. Then, through genetic operators such as selection, crossover, and mutation, new offspring design schemes are generated by simulating the biological evolution process. After multiple rounds of iterative evolution, the algorithm finally converges and outputs a robust Pareto optimal solution set. This solution set contains a series of design options, none of which can improve one objective without compromising others. Together, they form a set of candidate solutions that achieve the best robustness trade-offs under multiple requirements such as efficiency, leakage, strength, and lifespan, which decision-makers can choose based on their final engineering preferences.

[0104] This embodiment utilizes a surrogate model to quickly and accurately quantify the mean and variance of the performance of any design scheme, thereby constructing a robust optimization objective that simultaneously pursues superior performance and minimal fluctuations. This directly internalizes the uncertainties of actual operating conditions as a driving factor in the optimization process. Subsequently, within the multidimensional design space, advanced algorithms such as NSGA-III are employed for global parallel search. This solves the problem of traditional deterministic optimization ignoring operating condition fluctuations, which leads to design failure, and also overcomes the computational infeasibility of directly optimizing based on high-fidelity simulations.

[0105] Step S400: Based on the robust optimization solution set and the guide vane dynamic characteristics extracted from the multi-condition performance data, perform dynamic modeling and parameter calibration of the sealing system;

[0106] Step S400 specifically includes:

[0107] Step S410: Establish a dynamic equation describing the motion of the sealing element, the dynamic equation including a nonlinear contact force term determined by the relative displacement and relative velocity between the sealing element and the guide vane end face; specifically, establish an accurate dynamic model for the adaptive floating sealing device to mathematically describe the response behavior of the sealing element under the dynamic displacement excitation of the guide vane.

[0108] In some embodiments, the model simplifies the sealing system into a single-degree-of-freedom mass-spring-damped system, whose core dynamic equations express the relationship between the forces and motion of the system. The established motion control equations for the sealing system are as follows:

[0109]

[0110] In the formula, The equivalent mass of the sealing element is obtained by equivalent conversion between the actual mass of the sealing copper strip and the distributed mass of the elastic element. This indicates the displacement of the sealing element relative to its mounting base; The dynamic displacement time history of the guide vane end face can be extracted from the high-fidelity fully coupled simulation results of step S100. It represents the equivalent damping coefficient of the sealing system, characterizing the energy dissipation characteristics of the system during motion. The damping force is proportional to the relative velocity between the seal and the guide vane. The equivalent stiffness of the sealing system is mainly determined by the stiffness of the elastic element supporting the sealing strip. Indicates the initial pre-compression of the sealing element; It represents the nonlinear contact force between the sealing element and the top cover or bottom ring anti-wear plate, specifically a piecewise defined function whose value depends on the real-time contact state between the sealing element and the anti-wear plate;

[0111] Nonlinear contact force It can be represented as:

[0112]

[0113] In the formula, The real-time dynamic clearance is defined as the difference between the displacement of the sealing element and the displacement of the guide vane end face. , is the key variable for determining the contact state; contact state determination: when When the contact force is zero, it indicates that the sealing element is separated from the wear-resistant plate and there is no contact. ;when When the two are in contact, a compressive force is generated, and the contact force is calculated by a nonlinear expression; The contact stiffness coefficient represents the contact interface and describes the relationship between contact pressure and compression. Nonlinear relationship between them; The nonlinear exponent of the contact force is a dimensionless empirical constant greater than 1, used to characterize the nonlinear growth characteristics of the force when materials are in contact. The contact damping coefficient represents the energy loss caused by micro-slippage between contact surfaces during the contact process, and the resulting damping force is related to the rate of change of the gap. Proportional; Indicates dynamic gap The first derivative with respect to time, i.e., the rate of change of the gap;

[0114] Step S420: Using the guide vane end face displacement time history from the guide vane dynamic characteristics as the input to the dynamic equation, the stiffness parameters and precompression parameters of the sealing system are adjusted to minimize the following error between the motion response of the sealing element and the guide vane end face displacement time history. First, the guide vane end face displacement time history under the corresponding design scheme of the robust optimization solution set is extracted from the high-fidelity fully coupled simulation results of step S100, serving as the target motion trajectory that the sealing system must track. Simultaneously, the initial design values ​​of the equivalent stiffness and initial precompression of the sealing system are obtained from the optimization solution set obtained in step S300. Using the guide vane displacement time history as the input excitation, the dynamic equation established in step S410 is solved to simulate the motion response of the sealing strip. By calculating the root mean square value of the difference between the sealing strip displacement and the guide vane displacement, the following error of the seal can be quantitatively evaluated. Finally, with the goal of minimizing this following error, the equivalent stiffness and precompression parameters of the seal are automatically adjusted within a reasonable engineering range. This is an iterative optimization process until the optimal parameter combination that best matches the movement of the sealing strip with the movement of the guide vane end face is found, thus completing the final dynamic design of the sealing subsystem.

[0115] This embodiment elevates the seal from a static, passive component to an active subsystem with clearly defined dynamic characteristics and quantifiable dynamic tracking performance. It ensures that the sealing element can automatically and in real-time compensate for radial displacement of the guide vane end face caused by changes in water pressure and temperature during actual unit operation, thereby achieving source control of dynamic leakage gaps from a physical mechanism perspective. This not only directly solves the leakage problems that lead to water energy waste and unit creep, but also achieves coordinated optimization of the guide vane body and sealing mechanism at the dynamic operation level by indexing the dynamic performance of the sealing subsystem and incorporating it into the overall design closed loop.

[0116] Step S500: Perform simulation verification and sensitivity analysis on the robust optimization solution set, and output the parametric design results;

[0117] Step S500 specifically includes:

[0118] Step S510: Select at least one candidate design scheme from the robust optimization solution set; specifically, select typical design schemes for final verification and analysis from the non-dominated solution set generated in step S300.

[0119] In some embodiments, the solution that performs best in the core objective, such as minimizing leakage or maximizing fatigue life, is directly selected.

[0120] In some embodiments, a clustering algorithm can be used to select several schemes from the Pareto front that are uniformly distributed in the target space and can represent different performance trade-offs; each selected candidate design scheme corresponds to a set of defined guide vane body geometric design variable values. and sealing system design variable values .

[0121] Step S520: For the candidate design scheme, execute the step of constructing and solving the multiphysics coupled simulation model of the guide vane working environment to obtain performance data for verification; this step is the final verification of the surrogate model's prediction accuracy. For each candidate design scheme, based on its determined design variables... As input, the global Jacobian relaxation iterative full-physics coupling simulation established in step S100 is re-executed; at this time, the uncertainty parameter... Set to its typical value, high-cost but high-fidelity simulations are used to obtain the true high-precision solutions for various performance indicators under this scheme, denoted as the verification performance data vector; simultaneously, the same design variable... Input the hybrid agent model constructed in step S200 and obtain its performance prediction mean vector. The prediction accuracy of the agent model is systematically evaluated by calculating the relative error. If the errors of all key performance indicators are less than a preset threshold, the hybrid agent model and its optimization results are considered reliable; otherwise, it is necessary to return to step S200 to supplement sample points to correct the model.

[0122] Step S530: Based on the polynomial chaotic expansion part of the hybrid surrogate model, calculate the contribution of each operating condition uncertainty parameter to the variance of key performance indicators, using it as a global sensitivity index. This step utilizes the polynomial chaotic expansion part of the constructed hybrid surrogate model to quantitatively analyze the impact of each uncertainty parameter on performance fluctuations, providing a quantitative basis for tolerance control and operational monitoring. For the selected design scheme... , its first Variance of each performance metric The model provides the following: PCE decomposes the total variance into the sum of contributions from each uncertainty parameter and their interactions; the calculation of the first... Uncertainty parameters Performance Total sensitivity index , can be represented as:

[0123]

[0124] In the formula, Represents the uncertainty parameter The overall sensitivity index, with values ​​between 0 and 1, reflects the parameter Individually, and with all possible interactions of all other uncertain parameters, the performance is affected. The proportion of the total variance contributed by the variance; This indicates that it corresponds to multiple indicators. The PCE coefficients; the numerator of this equation for the set Calculate and sum all terms in the formula, and the result is obtained by considering the parameters. The variance component explained; divide this component by the total variance. This yields the overall sensitivity index. A higher index indicates a greater impact of the uncertainty parameter on performance fluctuations. By calculating this index for all uncertainty parameters, the key sources of fluctuation with the greatest impact on performance robustness can be identified. Based on the ranking of the global sensitivity indices, parameters with significant impacts on performance fluctuations are included in the key parameter control recommendations section of the design document to guide manufacturing tolerance setting and operational monitoring priorities.

[0125] Step S540: Input the final design parameters determined through simulation verification and sensitivity analysis into the parametric geometric modeling script to generate a 3D model of the guide vane and sealing device; import the verified and analyzed final design parameter set into a pre-developed parametric computer-aided design script, which automatically drives the kernel of the 3D modeling software to accurately update all relevant geometric features and dimensional constraints according to the input parameters, generating a 3D solid model of the guide vane body and a detailed assembly model of the adaptive floating sealing device that are completely corresponding to them.

[0126] Step S550: Output a design document containing the final design parameters, the calibrated dynamic parameters of the sealing system, and the prediction range of key performance indicators. The core content of the design document includes: a list of all finalized design parameters; recommendations for manufacturing and assembly tolerances of key dimensions based on sensitivity analysis conclusions; a table of dynamic parameters of the calibrated sealing system, such as equivalent stiffness and pre-compression; the mean values ​​of key performance indicators predicted based on the hybrid surrogate model and their reasonable prediction ranges when considering uncertainty fluctuations; and a brief summary and error analysis report of the high-fidelity verification results.

[0127] This embodiment utilizes a polynomial chaotic expansion model to conduct in-depth global sensitivity analysis, quantitatively identifying the uncertainty factors that have the greatest impact on guide vane performance fluctuations. This provides a direct and scientific basis for the formulation of key manufacturing tolerance control and operation monitoring strategies. Finally, by automatically converting the verified final design parameters into a three-dimensional geometric model and compiling them into a structured design document, the connection between digital optimization results and engineering production guidance is realized.

[0128] In summary, the turbine guide vane structure optimization design method provided by this invention, by constructing and solving a multi-physics coupled simulation model of the guide vane's working environment, can capture the real-time interactions between multiple physics fields, avoiding the distortion problems caused by traditional sequential unidirectional coupling analysis, thereby improving simulation accuracy and fidelity. Furthermore, the hybrid surrogate model constructed based on high-fidelity simulation data integrates adaptive experimental design and uncertainty quantification analysis, enabling efficient prediction of performance statistical characteristics. It transforms complex multi-physics simulations into rapidly evaluable surrogate models, solving the technical problems of high computational cost and low efficiency when directly calling high-fidelity simulations for optimization. Based on this, the hybrid surrogate model is used to perform multi-objective collaborative optimization of the guide vane and sealing system design parameter set. The design stage systematically considers the uncertainties of operating conditions, generating a robust optimization solution set, ensuring that the design scheme maintains stable performance even under fluctuations in actual operating parameters. Simultaneously, based on the optimization solution set and the dynamic characteristics of the guide vane, integrated dynamic modeling and parameter calibration of the sealing system are performed, enabling the sealing design to actively adapt to the dynamic deformation of the guide vane, thus suppressing end-face leakage from a mechanistic perspective. Finally, through simulation verification and sensitivity analysis of the optimized solution set, and output of parametric design results, a complete scheme for the optimization design of turbine guide vane structure was formed, from high-precision analysis, efficient optimization, robust design to reliable verification. This achieved a synergistic improvement in simulation fidelity, optimization efficiency and robustness of guide vane structure optimization design.

[0129] Please see Figure 2 The diagram illustrates a structural schematic of a turbine guide vane structure optimization design system provided in an embodiment of the present invention. The system includes:

[0130] The simulation module is configured to build and solve a multi-physics coupled simulation model of the guide vane's working environment to obtain multi-condition performance data of the guide vane.

[0131] The hybrid proxy module is configured to construct a hybrid proxy model for predicting the statistical characteristics of guide vane performance based on multi-condition performance data, combined with adaptive experimental design and uncertainty quantification analysis.

[0132] The collaborative optimization module is configured to use a hybrid surrogate model to perform multi-objective collaborative optimization of the design parameter set of the guide vane and sealing system, and generate a robust optimization solution set.

[0133] The parameter calibration module is configured to perform dynamic modeling and parameter calibration of the sealing system based on the robust optimization solution set and the dynamic characteristics of the guide vane.

[0134] The output module is configured to perform simulation verification and sensitivity analysis on the robustness optimization solution set, and output the parametric design results.

[0135] The third aspect of the present invention provides an electronic device, the electronic device comprising: a processor and a memory communicatively connected to the processor; wherein the memory stores instructions executable by the processor, the instructions being executed by the processor to enable the processor to perform the steps of the turbine guide vane structure optimization design method described in the first aspect of the present invention.

[0136] The fourth aspect of the present invention provides a computer-readable storage medium storing a program for implementing a method for optimizing the design of a turbine guide vane structure. The program for implementing the method for optimizing the design of a turbine guide vane structure is executed by a processor to implement the steps of the method for optimizing the design of a turbine guide vane structure as described in the first aspect of the present invention.

[0137] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0138] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

Claims

1. A method for optimal design of guide vane structure of a hydraulic turbine, characterized by, The method comprises: a multi-physics field coupling simulation model of the guide vane working environment is constructed and solved to obtain guide vane multi-working condition performance data, including guide vane stress distribution, deformation field, end face leakage flow velocity field and cavitation risk distribution; based on the multi-working condition performance data, combined with adaptive experimental design and uncertainty quantification analysis, a hybrid proxy model for predicting guide vane performance statistical characteristics is constructed, including: a joint input space is defined, including guide vane body design variables, sealing system design variables and working condition uncertainty parameters; an initial sample point set is generated in the joint input space by using an optimal Latin hypercube sampling method; for the initial sample point set, the step of constructing and solving the multi-physics field coupling simulation model of the guide vane working environment is performed to obtain corresponding performance response data, including multiple guide vane performance indicators; based on the performance response data, the hybrid proxy model for predicting guide vane performance statistical characteristics is trained; the design parameter set of the guide vane and the sealing system is optimized by using the hybrid proxy model, and a robust optimization solution set is generated; based on the robust optimization solution set and the guide vane dynamic characteristics extracted from the multi-working condition performance data, a dynamic model of the sealing system is established and parameters are calibrated, including: a dynamic equation describing the motion of the sealing element is established, and the dynamic equation includes a nonlinear contact force term determined by the relative displacement and relative velocity of the sealing element and the guide vane end face; the guide vane end face displacement time history in the guide vane dynamic characteristics is taken as the input of the dynamic equation, and the stiffness parameters and pre-compression amount parameters of the sealing system are adjusted to minimize the following error between the motion response of the sealing element and the guide vane end face displacement time history; the robust optimization solution set is simulated, verified and analyzed for sensitivity, and parameterized design results are output, including: at least one candidate design scheme is selected from the robust optimization solution set; for the candidate design scheme, the step of constructing and solving the multi-physics field coupling simulation model of the guide vane working environment is performed to obtain performance data for verification; based on the polynomial chaos expansion part in the hybrid proxy model, the contribution of each working condition uncertainty parameter to the variance of the key performance indicator is calculated as a global sensitivity index; the final design parameters determined through the simulation verification and sensitivity analysis are input into a parameterized geometric modeling script to generate a three-dimensional model of the guide vane and the sealing device; a design document containing the final design parameters, the calibrated dynamic parameters of the sealing system and the prediction interval of the key performance indicators is output.

2. The method of optimizing the structure of guide vanes of a hydraulic turbine according to claim 1, characterized in that, a multi-physics field coupling simulation model of the guide vane working environment is constructed and solved to obtain guide vane multi-working condition performance data, including: a multi-physics field coupling control equation set is established, including fluid dynamics control equations, structural dynamics control equations, heat conduction equations and erosion rate equations; based on the Jacobian matrix representing the coupling relationship between the fluid field, the structure field, the heat field and the erosion field, the multi-physics field coupling control equation set is solved synchronously by using a relaxation iteration algorithm; guide vane multi-working condition performance data including guide vane stress distribution, deformation field, end face leakage flow velocity field and cavitation risk distribution are extracted from the solution.

3. The method of claim 1, wherein Based on the performance response data, a hybrid surrogate model for predicting the performance statistical characteristics of the guide vane is trained, including: Based on the performance response data, a hybrid surrogate model combining a Kriging model and a polynomial chaos expansion is trained; wherein the Kriging model is used to establish a nonlinear mapping relationship between the performance response and all input variables, and the polynomial chaos expansion is used to analyze the influence of the uncertainty parameters of the working conditions on the statistical distribution of the performance response; The maximum likelihood estimation method is used to calibrate the covariance function parameters of the Kriging model; The regression method is used to solve the coefficients of the polynomial chaos expansion.

4. The method of claim 1, wherein The hybrid surrogate model is used to perform multi-objective collaborative optimization on the design parameter set of the guide vane and the sealing system, and a robust optimization solution set is generated, including: Based on the hybrid surrogate model, the predicted mean and variance of the performance indicators under a given design are calculated; According to the predicted mean and variance, an optimization objective function is constructed for each performance indicator; A multi-objective evolutionary algorithm is used to search in a design space composed of guide vane body geometric design variables and sealing system stiffness design variables to simultaneously optimize multiple optimization objective functions and obtain a robust optimization solution set composed of non-dominated solutions.

5. A system for optimal design of guide vane structure of a hydraulic turbine, characterized by A system for performing the water turbine guide vane structure optimization design method of any one of claims 1 to 4, the system comprising: A simulation module configured to build a multi-physics field coupled simulation model of the guide vane working environment and solve it to obtain guide vane multi-working condition performance data; A hybrid surrogate module configured to build a hybrid surrogate model for predicting the performance statistical characteristics of the guide vane based on the multi-working condition performance data, combined with adaptive design of experiments and uncertainty quantification analysis; A collaborative optimization module configured to use the hybrid surrogate model to perform multi-objective collaborative optimization on the design parameter set of the guide vane and the sealing system to generate a robust optimization solution set; A parameter calibration module configured to perform dynamic modeling and parameter calibration of the sealing system based on the robust optimization solution set and the dynamic characteristics of the guide vane; An output module configured to perform simulation verification and sensitivity analysis on the robust optimization solution set and output the parameterized design results.

6. An electronic device, comprising: The electronic device includes a processor and a memory communicatively connected to the processor; wherein the memory stores instructions executable by the processor, and the instructions are executed by the processor to enable the processor to perform the steps of the water turbine guide vane structure optimization design method of any one of claims 1 to 4.

7. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a program for implementing the water turbine guide vane structure optimization design method, and the program for implementing the water turbine guide vane structure optimization design method is executed by the processor to implement the steps of the water turbine guide vane structure optimization design method of any one of claims 1 to 4.

Images