A fluid-structure coupling method for turbine dynamic processes based on machine learning

Abstract

This invention discloses a machine learning-based fluid-structure coupling method for turbine dynamic processes. The method comprises: constructing a high-precision machine learning proxy model of the turbine and its dynamic adjustment simulation model based on the turbine's variable operating condition characteristic curve and flow field information dataset; using the turbine dynamic adjustment simulation model to solve the operating condition parameters during the turbine adjustment process, using these operating conditions as input to the high-precision machine learning proxy model of the turbine to obtain the turbine blade aerodynamic load distribution and the impeller surface flow field pressure distribution, which are input as boundary conditions into the impeller finite element model, solving for the impeller deformation and equivalent stress, and completing the turbine fluid-structure coupling analysis under dynamic processes. This method can ensure the efficient and safe operation of the turbine during frequent adjustment processes and can provide a reference for the design of subsequent similar turbines.

Description

Technical Field

[0001] The present invention belongs to the field of fluid-solid coupling analysis in fluid machinery, and specifically relates to a fluid-solid coupling method for turbine dynamic processes based on machine learning. Background Art

[0002] Turbines are the main working components of various energy systems, and are mainly used in vehicle engine turbochargers, medium and low temperature waste heat power generation devices, renewable energy power generation devices, chemical process expanders, rocket engine turbopumps, etc.

[0003] Turbine structures include centripetal, centrifugal, axial, and mixed flow types. Their geometry, structural dimensions, and rotational speed are determined by overall thermodynamic design parameters. Turbines often operate under frequent adjustments and variable operating conditions, and are required to operate at high efficiency. To ensure this high efficiency, turbine fluid-structure interaction analysis is necessary. Conventional fluid-structure interaction analysis primarily considers the turbine under steady-state and variable operating conditions; it lacks consideration of the flow field, structural strength, and dynamic adjustment characteristics.

[0004] Existing technologies are also unable to construct a functional mapping relationship between turbine operating conditions and flow field structure through fluid-solid coupling analysis to provide a reference for the subsequent design of new turbines of the same type. Summary of the Invention

[0005] The purpose of the present invention is to overcome the above-mentioned shortcomings and provide a turbine dynamic process fluid-solid coupling method based on machine learning that can ensure efficient and safe operation of the turbine during frequent adjustments and provide a reference for the design of subsequent similar turbines.

[0006] The technical solutions adopted by the present invention to solve its technical problems are:

[0007] A fluid-solid coupling method for a turbine dynamic process based on machine learning of the present invention comprises the following steps:

[0008] (1) Based on the turbine variable operating characteristic curve and flow field information dataset, a high-precision machine learning agent model of the turbine and its dynamic adjustment simulation model are constructed;

[0009] (2) The turbine dynamic regulation simulation model is used to solve the operating condition parameters during the turbine regulation process. The operating condition is used as input into the turbine high-precision machine learning agent model to obtain the aerodynamic load distribution of the turbine blades and the flow field pressure distribution on the wheel surface;

[0010] (3) The results in step (2) are input into the impeller finite element model as boundary conditions, and the impeller deformation and equivalent stress are obtained to complete the turbine fluid-solid coupling analysis under the dynamic process.

[0011] Preferably, the turbine variable operating characteristic curve in step (1) is an expansion ratio-power curve and an expansion ratio-flow curve at different reduced speeds, which are used to obtain the power and flow of the turbine at different reduced speeds and expansion ratios.

[0012] Preferably, the turbine flow field information data set in step (1) is the aerodynamic force distribution on the surface of the turbine blade and the flow field pressure distribution on the surface of the turbine impeller disk.

[0013] Preferably, the turbine high-precision machine learning agent model in step (1) includes an input module, a solution module, and an output module. The input module includes the turbine operating condition input parameter expansion ratio π tt and reduced speed n cor The solution module includes a machine learning algorithm, and the output module includes turbine performance parameters and flow parameters. Performance parameters include flow rate and power, while flow parameters include the pressure distribution on the impeller blade surface and the aerodynamic load distribution on the impeller disk surface.

[0014] Preferably, the machine learning method in the turbine high-precision machine learning agent model in step (1) includes a support vector machine method, a random forest method, an artificial neural network method, etc.

[0015] Preferably, the turbine dynamic regulation simulation model in step (1) is a grid-connected constant speed and variable load dynamic simulation model.

[0016] Compared with existing technologies, the present invention offers significant advantages. As can be seen from the above technical solutions, the present invention simultaneously performs fluid-structure interaction analysis on both the flow field and structural strength. Furthermore, it leverages the turbine's dynamic adjustment characteristics (speed overshoot and dynamic adjustment time) to solve the turbine's dynamic adjustment process, thereby ensuring efficient and safe predictions during frequent adjustments. The present invention also utilizes machine learning methods to establish a high-precision proxy model for the turbine, which not only improves solution efficiency but also allows the model to be updated based on new solution results to enhance accuracy. Furthermore, the machine learning method directly constructs a functional mapping between turbine operating conditions and aerodynamic performance parameters and flow field structure, providing a reference for the design of subsequent new turbines of the same type. Therefore, the present invention offers the advantages of short solution time and high solution efficiency, enabling fluid-structure interaction analysis of large-scale turbines with multiple impeller rows and multiple stages under dynamic conditions. It can be integrated with currently established turbine simulation platforms, variable operating condition adjustment programs, and turbine design programs, resulting in a wide range of applications. As the model is used more frequently, the number of sample points increases, further improving the model's prediction accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a schematic diagram of the present invention;

[0018] Figure 2 The turbine high-precision machine learning agent model of the present invention;

[0019] Figure 3 The turbine dynamic adjustment simulation system of the present invention;

[0020] Figure 4 This is the finite element analysis model of the present invention. DETAILED DESCRIPTION

[0021] Reference Figure 1 , a fluid-structure coupling method for turbine dynamic processes based on machine learning, comprising the following steps:

[0022] (1) Based on the turbine variable operating characteristic curve and flow field information data set, a high-precision machine learning agent model of the turbine and its dynamic adjustment simulation model are constructed. The turbine variable operating characteristic curve is the expansion ratio-flow curve and the expansion ratio-power curve at different reduced speeds; the turbine flow field information data set is the aerodynamic force distribution on the turbine blade surface and the pressure distribution on the turbine disk surface.

[0023] (2) If Figure 2 As shown in the figure, the artificial neural network method is used to build a turbine machine learning agent model. The input module in the model includes the turbine expansion ratio π tt , converted speed n cor , the solution module is the artificial neural network solution method, and the output module is the aerodynamic distribution matrix of the turbine blade surface { F 1 , F 2 , F 3 ,……, F n} and the pressure distribution matrix on the turbine disk surface { P 1 , P 2 , P 3 ,……, P n}.

[0024] (3) Use matlab simulink software to build a turbine dynamic regulation simulation system, such as Figure 3As shown in the figure. Based on given parameters such as turbine speed, turbine inlet temperature, and turbine outlet total pressure, the volumetric equation module and the turbine high-precision machine learning proxy model calculate turbine flow, power, blade surface pressure distribution, and turbine disk surface pressure distribution under different flow disturbances. The power is compared with actual load changes to provide a PID control model, which is fed back to the valve intake control valve to change the opening and the volumetric inlet flow rate, achieving closed-loop control. The flow rate is provided to the volumetric equation module, which combines the volumetric inlet total temperature and volumetric inlet flow rate to obtain the turbine inlet pressure. This is combined with the turbine outlet pressure to obtain the turbine expansion ratio, which is provided to the turbine high-precision machine learning proxy model.

[0025] (4) Apply the turbine blade surface pressure distribution and turbine disk surface pressure distribution in step (3) as boundary conditions to the turbine impeller blades and disk, and apply the turbine speed in the dynamic adjustment simulation as a constraint to the turbine impeller to construct the impeller finite element model, such as Figure 4 By solving the model, the impeller deformation and equivalent stress are obtained, and the turbine fluid-structure coupling analysis under dynamic process is completed.

[0026] The above description is merely a preferred embodiment of the present invention and does not constitute any form of limitation to the present invention. Any simple modification, equivalent change and modification made to the above embodiment based on the technical essence of the present invention without departing from the content of the technical solution of the present invention shall still fall within the scope of the technical solution of the present invention.

Claims

1. A fluid-structure coupling method for turbine dynamic processes based on machine learning, comprising the following steps: (1) Based on the turbine variable operating characteristic curve and flow field information dataset, a high-precision machine learning agent model of the turbine and its dynamic adjustment simulation model are constructed; The turbine variable operating condition characteristic curves are expansion ratio-power curves and expansion ratio-flow curves at different reduced speeds, which are used to obtain the power and flow of the turbine at different reduced speeds and expansion ratios. The turbine flow field information dataset is the aerodynamic force distribution on the turbine blade surface and the flow field pressure distribution on the turbine impeller disk surface. (2) The turbine dynamic regulation simulation model is used to solve the operating condition parameters during the turbine regulation process. The operating condition is used as input into the turbine high-precision machine learning agent model to obtain the aerodynamic load distribution of the turbine blades and the flow field pressure distribution on the wheel surface; (3) The results in step (2) are input into the impeller finite element model as boundary conditions, and the impeller deformation and equivalent stress are obtained to complete the turbine fluid-solid coupling analysis under the dynamic process.

2. The method for fluid-structure coupling in a turbine dynamic process based on machine learning according to claim 1, wherein: The turbine high-precision machine learning agent model in step (1) includes an input module, a solution module, and an output module. The input module includes the turbine operating condition input parameter expansion ratio and reduced speed The solution module includes a machine learning algorithm, and the output module includes turbine performance parameters and flow parameters. The performance parameters include flow rate and power; the flow parameters include the flow field pressure distribution on the impeller blade surface and the aerodynamic load distribution on the impeller disk surface.

3. The method for fluid-structure coupling in a turbine dynamic process based on machine learning according to claim 1, wherein: The machine learning methods in the turbine high-precision machine learning agent model described in step (1) include support vector machine method, random forest method, and artificial neural network method.

4. The method for fluid-structure coupling in a turbine dynamic process based on machine learning according to claim 1, wherein: The turbine dynamic regulation simulation model in step (1) is a grid-connected constant speed and variable load dynamic simulation model.

Images