Turbine prime mover whole machine rapid high-precision prediction method based on component cooperative prediction

Abstract

The application provides a turbine prime mover whole machine rapid high-precision prediction method based on component cooperation prediction, relates to the technical field of gas turbine performance analysis, and comprises the following steps: dividing a turbine prime mover whole machine into multiple components, calculating and obtaining component import and export boundary data sets, component flow field data sets and component boundary matching data sets, constructing a component boundary condition prediction model, a component flow field prediction model and a component boundary matching model; inputting current whole machine import and export boundary conditions into the boundary condition prediction model to obtain component import and export cross section flow field data, inputting the flow field data into the component flow field prediction model to obtain component internal flow field data; inputting the internal flow field data of each adjacent component into the component boundary matching model to obtain the internal flow field of each component after boundary matching, and splicing the internal flow field of each component according to spatial relationships to obtain whole machine flow field prediction results. The application can quickly and accurately predict turbine prime mover whole machine performance.

Description

Technical Field

[0001] The present invention relates to the technical field of gas turbine performance analysis, and in particular to a method for rapid and high-precision prediction of a complete turbine prime mover based on component collaborative prediction. Background Art

[0002] In the field of gas turbine and aircraft engine design, predicting the overall performance of turbine prime movers is a key requirement for overall engine design and optimization. Among existing overall engine performance prediction technologies, full-3D computational fluid dynamics (CFD) can achieve high-precision predictions, but computational time can range from days to months, making it difficult to meet the demands of rapid design optimization. Reduced-order modeling methods based on modal decomposition can reduce computational time to minutes, but the simplification of nonlinear coupling between components leads to a significant increase in overall engine error. Currently, data-driven technology has shown breakthrough potential in engine performance prediction. Its end-to-end modeling capabilities can achieve second-level response and metastable error, improving computational efficiency by a thousand-fold compared to traditional CFD simulations and effectively capturing the effects of nonlinear coupling between components. However, its practical application faces a dilemma: if high-precision overall engine-level CFD data is used to train the model, the high data accumulation cost and training difficulty significantly limit the method's engineering practicality. If a component-by-component iterative prediction strategy is adopted, small errors in upstream components are amplified through the data flow, resulting in overall engine prediction errors far exceeding industrial standards. Therefore, how to quickly predict the performance of the entire machine while ensuring the accuracy of the prediction results is a key technical problem to be solved in the intelligent design and operation and maintenance of turbine prime movers. Summary of the Invention

[0003] In view of this, an embodiment of the present application provides a method for rapid and high-precision prediction of a turbine prime mover as a whole based on component collaborative prediction, so as to solve the problem that the prior art is difficult to rapidly and accurately predict the performance of a turbine prime mover as a whole.

[0004] The present application provides the following technical solution: a method for rapid and high-precision prediction of a turbine prime mover based on component collaborative prediction, comprising:

[0005] The turbine prime mover is divided into multiple components, and the component inlet and outlet boundary data sets, the component flow field data sets, and the component boundary matching data sets are obtained through simulation calculations.

[0006] Based on the component inlet and outlet boundary data sets, a component boundary condition prediction model is constructed; based on the flow field data sets of each component, a flow field prediction model of each component is constructed; based on the component boundary matching data sets, a component boundary matching model is constructed;

[0007] Input the current inlet and outlet boundary conditions of the entire machine into the component boundary condition prediction model to obtain the inlet and outlet cross-sectional flow field data of each component; input the inlet and outlet cross-sectional flow field data of each component as the inlet and outlet boundary conditions of each component into the flow field prediction model of each component to obtain the internal flow field data of each component;

[0008] The internal flow field data of each adjacent component is input into the component boundary matching model to obtain the internal flow field of each component after boundary matching, and the internal flow fields of each component after boundary matching are spliced ​​according to the spatial relationship to obtain the whole machine flow field prediction result.

[0009] According to one embodiment of the present application, the multiple components include: compressor blade stage components, combustion chamber components, and several turbine blade stage components; wherein the number of the compressor blade stage components is equal to the number of compressor stages, and the number of the turbine blade stage components is equal to the number of turbine stages.

[0010] According to one embodiment of the present application, component inlet and outlet boundary datasets, component flow field datasets, and component boundary matching datasets are obtained through simulation calculations, including:

[0011] Meshing the turbine prime mover, assigning a number of inlet and outlet boundary conditions to the mesh of the mesh, and performing CFD calculations on the whole machine to obtain flow field simulation data of the whole machine. The inlet and outlet boundary conditions of the mesh of the whole machine are used as raw data, and the inlet and outlet cross-sectional flow field simulation data of each component in the flow field simulation data of the whole machine are intercepted as labels to obtain an inlet and outlet boundary data set of the component.

[0012] Meshing the multiple components, assigning a number of inlet and outlet boundary conditions to each mesh of the divided components, and performing CFD calculations on each component to obtain flow field simulation data for each component, using the inlet and outlet boundary conditions of each mesh of the component as raw data and the flow field simulation data of each component as labels to obtain the flow field dataset for each component;

[0013] For every two spatially adjacent components, the inlet and outlet boundary conditions of the preceding component along the flow direction are given, the outlet boundary conditions of the preceding component are used as the inlet boundary conditions of the following component along the flow direction, and several outlet boundary conditions of the following component are given. The two components and the whole composed of the two components are meshed and CFD calculated respectively to obtain the flow field of the preceding component, the flow field of the following component and the flow field of the whole two components, and the α part of the flow field of the preceding component is extracted. , the β part of the flow field of the latter component , and the spatial correspondence of the flow field in the overall flow field and Part of the flow field and , obtaining the component boundary matching dataset.

[0014] According to an embodiment of the present application, the α portion of the flow field of the preceding component satisfies: 50%<α<90%; and the β portion of the flow field of the succeeding component satisfies: 10%<β<50%.

[0015] According to one embodiment of the present application, the component boundary condition prediction model is a prediction model constructed based on the component inlet and outlet boundary data set, with the whole machine inlet and outlet boundary conditions as input and the inlet and outlet cross-sectional flow field data of each component as output;

[0016] The flow field prediction model of each component is a prediction model constructed based on the flow field data set of each component, with the inlet and outlet boundary conditions of the component as input and the internal flow field of the component as output;

[0017] The component boundary matching model is based on the component boundary matching dataset and the flow field and As input, the flow field and The matching model constructed as output.

[0018] According to one embodiment of the present application, the method further includes: using the component inlet and outlet boundary data set to train the component boundary condition prediction model, using the flow field data set of each component to train the flow field prediction model of each component, and using the component boundary matching data set to train the component boundary matching model.

[0019] According to one embodiment of the present application, the internal flow field data of each adjacent component is input into the component boundary matching model to obtain the internal flow field of each component after boundary matching, including:

[0020] For the internal flow field data of every two adjacent components, extract the α part of the flow field of the previous component along the flow direction and the β part of the flow field of the next component along the flow direction , input into the component boundary matching model to obtain the flow field after boundary matching and , using flow field and Replace the flow field separately and , and obtain the internal flow field of each component after the boundary matching.

[0021] Compared with the prior art, the beneficial effects that can be achieved by at least one of the above-mentioned technical solutions adopted in the embodiments of this specification include at least the following: an ultra-fast and high-precision prediction method for a turbine prime mover based on component collaborative prediction proposed in the embodiments of the present invention decomposes the turbine prime mover performance prediction problem into boundary prediction-component flow field prediction-component flow field boundary matching problems, thereby reducing the modeling and computational complexity of the whole machine flow field prediction. The boundary prediction technology proposed in the present invention only needs to realize the mapping of the whole machine inlet and outlet boundary conditions to the component inlet and outlet cross-sections, which significantly reduces the training difficulty of the data-driven model and avoids the error transmission of iterative prediction. Through the boundary prediction and parallel running flow field prediction method of each component proposed in the present invention, the flow field prediction results of the turbine prime mover can be obtained quickly and efficiently, reducing the computing resources and time cost of obtaining the whole machine flow field prediction data. At the same time, the component flow field boundary matching method proposed in the present invention ensures the continuity of the flow field boundaries of each component and improves the accuracy of the whole machine flow field prediction results. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0023] Figure 1 This is a schematic diagram of a method for rapid and high-precision prediction of a turbine prime mover based on component collaborative prediction according to an embodiment of the present invention;

[0024] Figure 2 This is a schematic diagram of the steps for implementing the data set acquisition and model training part in an embodiment of the present invention;

[0025] Figure 3 This is a schematic diagram of the implementation steps of the model prediction part in an embodiment of the present invention. DETAILED DESCRIPTION

[0026] The embodiments of the present application are described in detail below with reference to the accompanying drawings.

[0027] The following describes the embodiments of the present application through specific examples, and those skilled in the art can easily understand other advantages and effects of the present application from the contents disclosed in this specification. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. The present application can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present application. It should be noted that, in the absence of conflict, the features in the following embodiments and embodiments can be combined with each other. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without making creative work are within the scope of protection of this application.

[0028] like Figure 1 As shown, an embodiment of the present invention provides a method for rapid and high-precision prediction of a turbine prime mover based on component collaborative prediction, comprising:

[0029] Step 101: Divide the turbine prime mover into multiple components, and obtain component inlet and outlet boundary data sets, component flow field data sets, and component boundary matching data sets through simulation calculations.

[0030] Step 102: Based on the component inlet and outlet boundary data sets, a component boundary condition prediction model is constructed; based on the component flow field data sets, a flow field prediction model for each component is constructed; based on the component boundary matching data sets, a component boundary matching model is constructed;

[0031] Step 103: Input the current inlet and outlet boundary conditions of the entire machine into the component boundary condition prediction model to obtain the inlet and outlet cross-sectional flow field data of each component; input the inlet and outlet cross-sectional flow field data of each component as the inlet and outlet boundary conditions of each component into the component flow field prediction model to obtain the internal flow field data of each component;

[0032] Step 104: Input the internal flow field data of each adjacent component into the component boundary matching model to obtain the internal flow field of each component after boundary matching, and splice the internal flow fields of each component after boundary matching according to the spatial relationship to obtain the whole machine flow field prediction result.

[0033] In step 101 of this embodiment, the turbine prime mover is divided into multiple components, including: compressor blade stage components, combustion chamber components, and several turbine blade stage components; wherein the number of the compressor blade stage components is equal to the number of compressor stages, and the number of the turbine blade stage components is equal to the number of turbine stages.

[0034] In some embodiments of the present invention, component inlet and outlet boundary datasets, component flow field datasets, and component boundary matching datasets are obtained through simulation calculations, including:

[0035] Meshing the turbine prime mover, assigning a number of inlet and outlet boundary conditions to the mesh of the mesh, and performing CFD calculations on the whole machine to obtain flow field simulation data of the whole machine. The inlet and outlet boundary conditions of the mesh of the whole machine are used as raw data, and the inlet and outlet cross-sectional flow field simulation data of each component in the flow field simulation data of the whole machine are intercepted as labels to obtain an inlet and outlet boundary data set of the component.

[0036] Meshing the multiple components, assigning a number of inlet and outlet boundary conditions to each mesh of the divided components, and performing CFD calculations on each component to obtain flow field simulation data for each component, using the inlet and outlet boundary conditions of each mesh of the component as raw data and the flow field simulation data of each component as labels to obtain the flow field dataset for each component;

[0037] For every two spatially adjacent components, the inlet and outlet boundary conditions of the preceding component along the flow direction are given, the outlet boundary conditions of the preceding component are used as the inlet boundary conditions of the following component along the flow direction, and several outlet boundary conditions of the following component are given. The two components and the whole composed of the two components are meshed and CFD calculated respectively to obtain the flow field of the preceding component, the flow field of the following component and the flow field of the whole two components, and the α part of the flow field of the preceding component is extracted. , the β part of the flow field of the latter component , and the spatial correspondence of the flow field in the overall flow field and Part of the flow field and , obtaining the component boundary matching dataset. Further preferably, the α portion of the flow field of the preceding component satisfies: 50%<α<90%; and the β portion of the flow field of the succeeding component satisfies: 10%<β<50%.

[0038] In step 102 of this embodiment, the component boundary condition prediction model is a prediction model constructed based on the component inlet and outlet boundary data set, with the whole machine inlet and outlet boundary conditions as input and the inlet and outlet cross-sectional flow field data of each component as output; the component flow field prediction model is a prediction model constructed based on the component flow field data set, with the component inlet and outlet boundary conditions as input and the component internal flow field as output; the component boundary matching model is a prediction model constructed based on the component boundary matching data set, with the flow field and As input, the flow field and The matching model constructed as output.

[0039] In some embodiments of the present invention, the method further includes: using the component inlet and outlet boundary data set to train the component boundary condition prediction model, using the flow field data set of each component to train the flow field prediction model of each component, and using the component boundary matching data set to train the component boundary matching model.

[0040] In step 103 of this embodiment, the current inlet and outlet boundary conditions of the entire machine are input into the component boundary condition prediction model to obtain the inlet and outlet cross-sectional flow field data of each component, and the inlet and outlet cross-sectional flow field data of each component are input into the flow field prediction model of each component as the inlet and outlet boundary conditions of each component to obtain the internal flow field data of each component.

[0041] In step 104 of this embodiment, the internal flow field data of each adjacent component is input into the component boundary matching model to obtain the internal flow field of each component after boundary matching, including: for each two adjacent internal flow field data of each component, extracting the α part of the flow field of the previous component along the flow direction and the β part of the flow field of the next component along the flow direction , input into the component boundary matching model to obtain the flow field after boundary matching and , using flow field and Replace the flow field separately and Finally, the internal flow fields of the components after boundary matching are spliced ​​according to the spatial relationship to obtain the prediction result of the flow field of the entire machine.

[0042] In one specific implementation of the present invention, the turbine prime mover comprises a one-stage axial-flow compressor (consisting of two stages, each consisting of guide vanes and rotor blades), an annular can-type combustor, and a one-stage axial-flow reaction turbine (consisting of two stages, each consisting of nozzles and rotor blades). The compressor and turbine are rigidly connected coaxially. The implementation steps of this specific implementation of the proposed method include two parts: dataset acquisition and model training, and model prediction.

[0043] like Figure 2 As shown, the data set acquisition and model training part:

[0044] The whole machine is divided into five parts: Part A is the first stage of the compressor (guide vanes and rotor blades), Part B is the second stage of the compressor (guide vanes and rotor blades), Part C is the annular can-type combustion chamber, Part D is the first stage of the turbine (nozzle and moving blades), and Part E is the second stage of the turbine (nozzle and moving blades).

[0045] The entire machine was meshed, and CFD calculations were performed given 50 sets of inlet and outlet boundary conditions to obtain the entire machine flow field simulation data. The inlet and outlet boundary conditions were used as the raw data, and the inlet and outlet cross-sectional flow fields of each component in the entire machine flow field simulation results were intercepted as labels to obtain a component inlet and outlet boundary data set with a sample size of 50. The five components were meshed, and 1000 sets of boundary conditions were given to each component. CFD calculations were performed to obtain the flow fields of each component. The boundary conditions were used as the raw data, and the flow fields of each component were used as labels to obtain five component flow field data sets with a sample size of 1000 each. For components A and B, 5 inlet boundary conditions and 10 outlet boundary conditions are given to component A, and these 10 outlet boundary conditions are used as the inlet boundary conditions of component B. 5 outlet boundary conditions are also given to component B. Components A and B are simulated to obtain flow field data. Components A and B are meshed as a whole and simulated to obtain flow field data. The last 30% of the flow field simulation results of component A and the first 30% of the flow field simulation results of component B are used as the original data, and the partial flow field part corresponding to the overall simulation result of component AB is used as the label to construct a component AB boundary matching dataset with a sample number of 250. The boundary matching datasets of components BC, CD, and DE are constructed in the same way.

[0046] Using the component inlet and outlet boundary dataset, a deep learning model for component boundary condition prediction was trained using a graph neural network. Using the component flow field dataset, five deep learning models for flow field prediction corresponding to five components were trained using a graph neural network. Using the component boundary matching dataset, four deep learning models for flow field boundary matching corresponding to four component groups were trained using a graph neural network.

[0047] like Figure 3 As shown, the model prediction part:

[0048] The current inlet and outlet boundary conditions of the entire machine are input into the trained component boundary condition prediction deep learning model to obtain the inlet and outlet cross-sectional flow fields of the five components and obtain the inlet and outlet boundary conditions of the five components.

[0049] The inlet and outlet boundary conditions of the five components are input into the trained deep learning model for flow field prediction of the five components to obtain the flow field prediction results of the five components.

[0050] The last 30% of the flow field prediction results of component A and the first 30% of the flow field prediction results of component B are input into the deep learning model for flow field boundary matching of component AB. The output results are used to replace the corresponding input flow fields to obtain the flow field boundary matching results of components A and B. The same method is used to obtain the flow field boundary matching results of components B and C, components C and D, and components D and E.

[0051] The flow fields of components A, B, C, D, and E after boundary matching are spliced ​​in sequence to obtain ultra-fast and high-precision prediction results of the flow field of the entire turbine prime mover.

[0052] The present invention proposes an ultra-fast and high-precision prediction method for a turbine prime mover as a whole based on component collaborative prediction, which decomposes the turbine prime mover performance prediction problem into boundary prediction-component flow field prediction-component flow field boundary matching problems, thereby reducing the modeling and computational complexity of the whole machine flow field prediction. The boundary prediction technology proposed in the present invention only requires the mapping of the whole machine inlet and outlet boundary conditions to the component inlet and outlet cross-sections, significantly reducing the training difficulty of the data-driven model and avoiding the error transmission of iterative prediction. Through the boundary prediction and parallel running flow field prediction methods of the present invention, the turbine prime mover flow field prediction results can be obtained quickly and efficiently, reducing the computing resources and time cost of obtaining the whole machine flow field prediction data. At the same time, the component flow field boundary matching method proposed in the present invention ensures the continuity of the flow field boundaries of each component and improves the accuracy of the whole machine flow field prediction results.

[0053] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present application should be included in the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.

Claims

1. A fast and high-precision prediction method for a turbine prime mover based on component collaborative prediction, characterized in that: include: The turbine prime mover is divided into multiple components, and the component inlet and outlet boundary data sets, the component flow field data sets, and the component boundary matching data sets are obtained through simulation calculations. Building a component boundary condition prediction model based on the component import and export boundary data set; Based on the flow field data sets of each component, constructing a flow field prediction model for each component; Based on the component boundary matching dataset, a component boundary matching model is constructed; Input the current inlet and outlet boundary conditions of the entire machine into the component boundary condition prediction model to obtain the inlet and outlet cross-sectional flow field data of each component; input the inlet and outlet cross-sectional flow field data of each component as the inlet and outlet boundary conditions of each component into the flow field prediction model of each component to obtain the internal flow field data of each component; The internal flow field data of each adjacent component is input into the component boundary matching model to obtain the internal flow field of each component after boundary matching, and the internal flow fields of each component after boundary matching are spliced ​​according to the spatial relationship to obtain the whole machine flow field prediction result.

2. The method for rapid and high-precision prediction of a complete turbine prime mover according to claim 1, characterized in that: The multiple components include: compressor blade stage components, combustion chamber components, and several turbine blade stage components; wherein the number of the compressor blade stage components is equal to the number of compressor stages, and the number of the turbine blade stage components is equal to the number of turbine stages.

3. The method for rapid and high-precision prediction of a complete turbine prime mover according to claim 1, characterized in that: Through simulation calculations, component inlet and outlet boundary data sets, component flow field data sets, and component boundary matching data sets are obtained, including: Meshing the turbine prime mover, assigning a number of inlet and outlet boundary conditions to the mesh of the mesh, and performing CFD calculations on the whole machine to obtain flow field simulation data of the whole machine. The inlet and outlet boundary conditions of the mesh of the whole machine are used as raw data, and the inlet and outlet cross-sectional flow field simulation data of each component in the flow field simulation data of the whole machine are intercepted as labels to obtain an inlet and outlet boundary data set of the component. Meshing the multiple components, assigning a number of inlet and outlet boundary conditions to each mesh of the divided components, and performing CFD calculations on each component to obtain flow field simulation data for each component, using the inlet and outlet boundary conditions of each mesh of the component as raw data and the flow field simulation data of each component as labels to obtain the flow field dataset for each component; For every two spatially adjacent components, the inlet and outlet boundary conditions of the preceding component along the flow direction are given, the outlet boundary conditions of the preceding component are used as the inlet boundary conditions of the following component along the flow direction, and several outlet boundary conditions of the following component are given. The two components and the whole composed of the two components are meshed and CFD calculated respectively to obtain the flow field of the preceding component, the flow field of the following component and the flow field of the whole two components, and the α part of the flow field of the preceding component is extracted. , the β part of the flow field of the latter component , and the spatial correspondence of the flow field in the overall flow field and Part of the flow field and , obtaining the component boundary matching dataset.

4. The method for rapid and high-precision prediction of a complete turbine prime mover according to claim 3, characterized in that: The α portion of the flow field of the preceding component satisfies: 50%<α<90%; the β portion of the flow field of the succeeding component satisfies: 10%<β<50%.

5. The method for rapid and high-precision prediction of a complete turbine prime mover according to claim 3, characterized in that: The component boundary condition prediction model is a prediction model constructed based on the component inlet and outlet boundary data set, with the whole machine inlet and outlet boundary conditions as input and the inlet and outlet cross-sectional flow field data of each component as output; The flow field prediction model of each component is a prediction model constructed based on the flow field data set of each component, with the inlet and outlet boundary conditions of the component as input and the internal flow field of the component as output; The component boundary matching model is based on the component boundary matching dataset and the flow field and As input, the flow field and The matching model constructed as output.

6. The method for rapid and high-precision prediction of a complete turbine prime mover according to claim 1, characterized in that: The method further comprises: The component inlet and outlet boundary data sets are used to train the component boundary condition prediction model, the component flow field data sets are used to train the component flow field prediction model, and the component boundary matching data sets are used to train the component boundary matching model.

7. The method for rapid and high-precision prediction of a complete turbine prime mover according to claim 1, characterized in that: Inputting the internal flow field data of each adjacent component into the component boundary matching model to obtain the internal flow field of each component after boundary matching, including: For the internal flow field data of every two adjacent components, extract the α part of the flow field of the previous component along the flow direction and the β part of the flow field of the next component along the flow direction , input into the component boundary matching model to obtain the flow field after boundary matching and , using flow field and Replace the flow field separately and , and obtain the internal flow field of each component after the boundary matching.

Images