Horizontal shaft water turbine blade digital twinborn physical field construction method based on proxy model

Abstract

The invention provides a method for constructing a physical field of a horizontal axis water turbine blade of an agent model, and relates to the technical field of digital twinning. According to the method, rotating speed data of a water turbine under different flow velocity conditions are determined, and rotating speed values corresponding to all flow velocities are obtained; then, fluid-solid coupling simulation is carried out in combination with experimental data, and stress and deformation data of the water turbine blade under the actual operation working condition are further obtained; according to the method, the KNN algorithm is utilized to carry out order reduction processing on the obtained blade structure data, a plurality of grid points are combined into representative points, the calculation complexity is low, the processing requirement of graph rendering software on the data size in the real-time rendering process is met, a combined Kriging method is adopted to carry out digital twinning agent model training on the data after order reduction, and the real-time rendering efficiency is improved. The reconstruction of a real-time physical field of the blade is realized; iteratively optimizing the digital twin physical field proxy model according to a genetic algorithm, and solving the problem of deviation between the proxy physical field model and a real physical model; and finally, deploying the digital twin physical field agent model into a digital twin system and solving physical field distribution in real time.

Description

Technical Field

[0001] The present invention relates to the field of digital twin technology, and in particular to a method for constructing a physical field of a horizontal-axis turbine blade based on an agent model. Background Art

[0002] With the widespread application of hydraulic turbines in the energy sector, monitoring and optimizing their performance has become an important means of improving operational efficiency and extending service life. Traditional analysis of hydraulic turbine blade performance typically relies on physical experiments and numerical simulations. While these methods can provide certain reference data, they suffer from high computational complexity, poor real-time performance, and difficulty in handling large-scale, complex operating environments.

[0003] Existing fluid-structure interaction simulation technology can accurately calculate the stress and deformation of turbine blades under different operating conditions. However, due to the enormous computational effort involved, it is difficult to accurately reflect the operating status of turbine blades in real time. Furthermore, simulation results typically require processing using discrete grid points, resulting in excessively large data volumes that are difficult to display in real time within graphics rendering software. This, in turn, limits its application in real-time monitoring and optimization of turbine blade performance.

[0004] While some research based on digital twin technology has attempted to combine simulation data with actual operational data, existing digital twin physics models primarily focus on simple engineering structures. They lack accurate modeling of complex fluid dynamics and their coupling with the structure, and they also lack performance optimization models for specialized structures like turbine blades. Therefore, efficiently and in real time, building digital twin physics for turbine blades using proxy models remains a pressing technical challenge. Summary of the Invention

[0005] In response to the shortcomings of the existing technology, the present invention provides a method for constructing the physical field of a horizontal-axis turbine blade based on a proxy model. This method realizes the construction of a digital twin physical field of the turbine blade through fluid-solid coupling simulation, order reduction processing, proxy model training and optimization.

[0006] The present invention achieves the above technical objectives through the following technical means.

[0007] A method for rapidly constructing the physical field of a digital twin of a horizontal-axis turbine blade based on an agent model includes the following steps:

[0008] Step 1: Determine the turbine speed data under different flow conditions and obtain the corresponding speed values ​​at each flow rate to provide experimental data for subsequent physical field simulation;

[0009] Step 2: Combine the experimental data obtained in step 1 to perform fluid-structure coupling simulation to obtain stress data and deformation data of the turbine blades under actual operating conditions;

[0010] Step 3: Use the KNN algorithm to reduce the order of the stress and deformation data of the blade obtained in step 2, and merge multiple grid points into representative points to reduce the amount of data and reduce the computational complexity;

[0011] Step 4: Perform joint Kriging modeling on the data reduced in step 3 to construct a proxy model;

[0012] Step 5: Use genetic algorithm to optimize the agent model in step 4;

[0013] Step 6: Deploy the proxy model optimized in step 5 into the digital twin system, and use graphics rendering software to display the stress and deformation physical field distribution of the turbine blades in real time to achieve real-time monitoring and simulation.

[0014] The solution of the present invention greatly reduces the amount of calculation through KNN algorithm order reduction processing, meeting the requirements of real-time rendering. The combined Kriging method can effectively improve the accuracy of the proxy model and enhance the model's adaptability to the actual operating status of the turbine blades. The introduction of the genetic algorithm makes the optimization process of the proxy model more efficient and can quickly correct the deviation between the model and the actual physical model. Through digital twin technology, the real-time construction and visualization of the physical field of the turbine blades are realized, which has strong practical value and technological advancement.

[0015] In the above solution, in step 2, the fluid-solid coupling simulation adopts the finite element analysis method.

[0016] In the above solution, in step 3, the order reduction process of the KNN algorithm generates representative points by selecting 4 nearest grid points.

[0017] In the above scheme, in step 4, the joint kriging method uses a Gaussian process regression model to model the reduced-order data to generate a proxy physical field model.

[0018] In the above solution, in step 4, the genetic algorithm improves the prediction accuracy of the agent model by selecting a fitness function and performing crossover and mutation operations.

[0019] In the above scheme, in step six, the digital twin system includes a data acquisition module, a physical field solution module and a rendering module, which are used to monitor and visualize the physical field distribution of the turbine blades in real time.

[0020] In the above scheme, in step six, the sensors in the data material module collect the stress and deformation physical field data of the blade in real time; the physical field solution module calculates and displays the physical field distribution of the blade in real time based on the input real-time data and the calculation results of the proxy model; the rendering module renders the calculation results in real time and displays the stress and deformation distribution physical field information of the turbine blade under the current working conditions.

[0021] In the above solution, in step six, the proxy model can be calibrated and updated online by comparing it with actual blade performance data.

[0022] In the above scheme, in step 4, the mathematical formula of the proxy model is:

[0023]

[0024] Where Z(x) is the output of the proxy model, the physical field data, x i is a known training data point, k(x,x i ) is the kernel function, λ i is the weight of the training data point, and ∈(x) is the error term.

[0025] In the above scheme, in step 5, the genetic algorithm is used to further optimize the model. Through selection, crossover, and mutation operations, the optimal solution is searched to adjust the parameters of the joint kriging model so that the deviation between the output of the physical field proxy model and the actual measurement value is minimized. The effect of the model is evaluated by the fitness function in each generation of optimization. The fitness function is:

[0026]

[0027] in, is the predicted value of the surrogate model, y i is the actual measurement value, N is the number of sample points, and θ is the parameter of the proxy model.

[0028] Beneficial effects: The method of the present invention determines the speed data of the turbine under different flow conditions to obtain the corresponding speed value at each flow rate; then, in combination with the experimental data, a fluid-solid coupling simulation is performed to further obtain the stress and deformation data of the turbine blade under actual operating conditions; the obtained blade structure data is reduced in order using the KNN algorithm, and multiple grid points are merged into representative points with low computational complexity to meet the data volume processing requirements of the graphics rendering software during real-time rendering; the digital twin proxy model is trained on the reduced-order data using the joint Kriging method to achieve reconstruction of the real-time physical field of the blade; the digital twin physical field proxy model is iteratively optimized according to the genetic algorithm to solve the problem of deviation between the proxy physical field model and the real physical model; finally, the digital twin physical field proxy model is deployed into the digital twin system and the physical field distribution is solved in real time. The present invention achieves the spatial distribution of the structural field reflected in the digital twin system of the turbine, thereby realizing the physical field simulation driven by real-time operation data. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 Flowchart of the method for constructing the physical field of horizontal-axis turbine blades based on the surrogate model. DETAILED DESCRIPTION

[0030] The following describes embodiments of the present invention in detail, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present invention, and are not to be construed as limiting the present invention.

[0031] Combined with attachment Figure 1 As shown in FIG, the method for constructing the physical field of a horizontal axis turbine blade based on a proxy model includes the following steps:

[0032] 1. Based on the turbine speed data under different flow conditions, obtain the corresponding speed value at each flow rate. Obtain the speed data through experiments or simulations to provide experimental data for subsequent fluid-structure coupling simulations.

[0033] 2. Input the experimental data into the simulation software to calculate the stress and deformation of the turbine blades at different flow rates; this process is completed through numerical calculation methods using the ANSYS fluid-structure interaction simulation tool;

[0034] 3. Use the KNN algorithm to reduce the order of blade structure data in the simulation results, merge multiple grid points into a few representative points, reduce the amount of data, and reduce the computational complexity to meet the needs of real-time display;

[0035] 4. Using the joint kriging method to train a surrogate model on the reduced-order data; this surrogate model can reflect the physical field distribution of the turbine blades, including stress and deformation, in real time;

[0036] 5. Use genetic algorithms to optimize the proxy model. Through multiple generations of genetic algorithms, the parameters of the proxy model are gradually optimized to minimize the deviation between the proxy model and the real physical model, thus solving the deviation problem between the proxy physical field model and the real physical model.

[0037] 6. Deploy the optimized proxy model into the digital twin system, and use graphics rendering software to display the physical field distribution of the turbine blades, such as stress and deformation, in real time, to achieve real-time monitoring and simulation.

[0038] Example:

[0039] Step 1: Assume that the turbine speed data under different flow conditions are obtained through experiments. In the experiments, the turbine is operated at different flow rates, such as 2 m / s, 4 m / s, and 6 m / s, and the measured speeds are 1200 rpm, 1500 rpm, and 1800 rpm, respectively. Based on these experimental data, establish the relationship between speed and flow rate and obtain the corresponding speed values ​​under different flow rates.

[0040] Step 2: Using fluid-structure coupling simulation software such as ANSYS Fluent, the geometric model of the turbine blade and flow velocity data are input for meshing and fluid-structure coupling simulation. Three flow velocity conditions are selected: 2 m / s, 4 m / s, and 6 m / s. Simulation calculations are performed at each flow velocity to obtain the stress and deformation distribution of the blade at different flow velocities.

[0041] Step 3: Reduce the blade surface stress and deformation data obtained from the fluid-structure interaction simulation. Export the stress and coordinate data of each node from the ANSYS Mechnial module. Assuming the original data includes 20,000 coordinate points, use the KNN algorithm to select the five nearest neighbors of each point and merge them into a representative point. Calculate the weighted average of each group of neighboring points to obtain the representative point. This ensures data reliability while meeting the computational requirements of real-time rendering.

[0042] Step 4: Combine the real-time fluid-structure interaction simulation data and historical experimental data to build a real-time proxy model through the joint Kriging method; perform joint Kriging interpolation modeling on the reduced-order physical field data. The covariance function is used to describe the relationship between different input variables (such as node position, flow velocity, etc.) to establish a proxy model of the physical field. Its mathematical formula is:

[0043]

[0044] Where Z(x) is the output of the proxy model (physical field data), xi is the known training data point, k(x,xi) is the kernel function, λi is the weight of the training data point, and ∈(x) is the error term;

[0045] The genetic algorithm is used to further optimize the model. Through selection, crossover, and mutation operations, the optimal solution is searched to adjust the parameters of the joint kriging model so that the deviation between the output of the physical field proxy model and the actual measurement value is minimized. The effect of the model is evaluated by the fitness function in each generation of optimization. The fitness function is:

[0046]

[0047] in, is the predicted value of the surrogate model, y i is the actual measurement value, N is the number of sample points, and θ is the parameter of the proxy model;

[0048] Through this process, a high-precision, low-computational-complexity physical field proxy model is finally obtained, which can accurately reflect the physical field distribution of turbine blades under different working conditions;

[0049] Step 5: After optimization is complete, the physical field proxy model is deployed on the digital twin platform for real-time prediction and simulation of the physical field distribution. During actual turbine operation, sensors are deployed to collect real-time physical field data, such as blade stress and deformation. The sensor data, including parameters such as blade speed, flow rate, and temperature, is transmitted to the digital twin platform via serial communication. After receiving the real-time sensor data, the digital twin platform transmits the data to the optimized physical field proxy model. Specifically, the sensor data is sent to the digital twin platform via the Socket communication protocol. The platform receives the data through the Socket interface and inputs it into the physical field proxy model. The proxy model processes the input data and derives the corresponding physical field distribution results. The digital twin platform calculates and displays the blade physical field distribution in real time based on the input real-time data and the calculation results of the physical field proxy model. At this point, graphics rendering software renders the calculation results in real time, displaying physical field information such as the stress and deformation distribution of the turbine blades under the current operating conditions.

[0050] Throughout this specification, reference to terms such as "one embodiment," "some embodiments," "examples," "specific examples," or "some examples" means that a specific feature, structure, material, or characteristic described in conjunction with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, schematic representations of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

[0051] Although the embodiments of the present invention have been shown and described above, it will be understood that the above embodiments are illustrative and are not to be construed as limitations on the present invention. A person skilled in the art may change, modify, replace and modify the above embodiments within the scope of the present invention without departing from the principles and purpose of the present invention.

Claims

1. A method for rapidly constructing the digital twin physical field of a horizontal axis turbine blade based on an agent model, characterized in that: The steps include: Step 1: Determine the turbine speed data under different flow conditions and obtain the corresponding speed values ​​at each flow rate to provide experimental data for subsequent physical field simulation; Step 2: Combine the experimental data obtained in step 1 to perform fluid-structure coupling simulation to obtain stress data and deformation data of the turbine blades under actual operating conditions; Step 3: Use the KNN algorithm to reduce the order of the stress and deformation data of the blade obtained in step 2, and merge multiple grid points into representative points to reduce the amount of data and reduce the computational complexity; Step 4: Perform joint Kriging modeling on the data reduced in step 3 to construct a proxy model; Step 5: Use genetic algorithm to optimize the agent model in step 4; Step 6: Deploy the proxy model optimized in step 5 into the digital twin system, and use graphics rendering software to display the stress and deformation physical field distribution of the turbine blades in real time to achieve real-time monitoring and simulation.

2. The method for rapidly constructing the digital twin physical field of a horizontal axis turbine blade based on an agent model according to claim 1 is characterized in that: In step 2, the fluid-solid coupling simulation adopts the finite element analysis method.

3. The method for rapidly constructing the digital twin physical field of a horizontal axis turbine blade based on an agent model according to claim 1 is characterized in that: In step 3, the order reduction process of the KNN algorithm generates representative points by selecting 4 nearest grid points.

4. The method for rapidly constructing the digital twin physical field of a horizontal axis turbine blade based on an agent model according to claim 1, characterized in that: In step 4, the joint kriging method uses a Gaussian process regression model to model the reduced-order data to generate a proxy physical field model.

5. The method for rapidly constructing the digital twin physical field of a horizontal axis turbine blade based on an agent model according to claim 1, characterized in that: In step 4, the genetic algorithm selects a fitness function and performs crossover and mutation operations to improve the prediction accuracy of the proxy model.

6. The method for rapidly constructing the digital twin physical field of a horizontal axis turbine blade based on an agent model according to claim 1, characterized in that: In step six, the digital twin system includes a data acquisition module, a physical field solution module, and a rendering module, which are used to monitor and visualize the physical field distribution of the turbine blades in real time.

7. The method for rapidly constructing a digital twin physical field of a horizontal axis turbine blade based on an agent model according to claim 6, characterized in that: In step 6, the sensors in the data material module collect the stress and deformation physical field data of the blade in real time; the physical field solution module calculates and displays the physical field distribution of the blade in real time based on the input real-time data and the calculation results of the proxy model; The rendering module renders the calculation results in real time and displays the physical field information of stress and deformation distribution of turbine blades under the current working conditions.

8. The method for rapidly constructing a digital twin physical field of a horizontal axis turbine blade based on an agent model according to claim 1, characterized in that: In step six, the proxy model can be calibrated and updated online by comparing it with actual blade performance data.

9. The method for rapidly constructing a digital twin physical field of a horizontal axis turbine blade based on an agent model according to claim 1, characterized in that: In step 4, the mathematical formula of the proxy model is: Where Z(x) is the output of the proxy model, the physical field data, x i is a known training data point, k(x,x i ) is the kernel function, λ i is the weight of the training data point, and ∈(x) is the error term.

10. The method for rapidly constructing a digital twin physical field of a horizontal axis turbine blade based on an agent model according to claim 1, characterized in that: In step five, the genetic algorithm is used to further optimize the model. Through selection, crossover, and mutation operations, the optimal solution is searched to adjust the parameters of the joint kriging model so that the deviation between the output of the physical field proxy model and the actual measurement value is minimized. The effect of the model is evaluated by the fitness function in each generation of optimization. The fitness function is: in, is the predicted value of the surrogate model, y i is the actual measurement value, N is the number of sample points, and θ is the parameter of the proxy model.

Images