A method and system for gas turbine air system analysis

Abstract

The application discloses a gas turbine air system analysis method and system, and relates to the technical field of gas turbine air system simulation. The method comprises the following steps: constructing a three-dimensional CFD benchmark model of a non-standard element of an air system; establishing a parameterized CFD analysis process, generating a parameter combination matrix covering the working range of the non-standard element; performing a multi-working condition three-dimensional CFD steady-state calculation to generate a data set; constructing and verifying a reduced-order model of the non-standard element based on the data set until the prediction accuracy of the reduced-order model meets preset engineering requirements; encapsulating the verified reduced-order model as a functional model unit; and replacing the original element with the functional model unit in a one-dimensional fluid network modeling environment, and calling the functional model unit to perform system-level simulation iteration. The application realizes seamless integration with the one-dimensional fluid network environment, decouples high-precision three-dimensional modeling from system-level iteration solving, significantly reduces the calculation cost, avoids the convergence risk caused by the three-dimensional CFD embedded iteration cycle, and realizes efficient and stable analysis of the overall performance of the air system.

Description

Technical Field

[0001] This application relates to the field of gas turbine air system simulation technology, specifically to a gas turbine air system analysis method and system. Background Technology

[0002] The air system of an aero-engine and gas turbine is a complex network consisting of multiple inlets, outlets, and branches, including compressor bleed air, cooling channels, sealed cavities, and buffer cavities. Its flow and heat transfer characteristics directly affect the overall efficiency and lifespan of the engine. In engineering practice, performing full 3D CFD simulation of the entire air system faces numerous challenges, including complex geometric modeling, difficult mesh generation, high computational costs, and difficulties in guaranteeing convergence. Therefore, the industry commonly uses the one-dimensional fluid network method for system-level analysis, abstracting standard flow resistances and heat transfer elements into models based on empirical formulas, thereby achieving rapid solutions.

[0003] However, air systems contain numerous non-standard components with complex geometries. Due to the lack of universal empirical correlations, traditional methods are forced to use approximate substitutions, leading to significant modeling errors and severely impacting the accuracy of the analysis results. To compensate for this deficiency, the closest existing technical solution is to employ a one-dimensional-three-dimensional coupled analysis method. This involves using a one-dimensional model for standard components and directly calling a three-dimensional CFD solver for high-fidelity simulation of key non-standard components. However, this existing solution has the following core drawbacks: First, it consumes enormous computational resources. Each high-frequency iteration of the one-dimensional network requires running a complete three-dimensional CFD, causing the cumulative cost to rise sharply and making it difficult to support engineering optimization iterations. Second, the convergence stability of the coupled solution is poor; numerical divergence in the three-dimensional CFD itself will directly lead to the failure of the entire system solution. The root cause of these drawbacks is that existing technologies directly embed high-cost three-dimensional modeling into the system-level iterative loop, failing to effectively decouple the two goals of high-precision physical simulation and efficient system solution, resulting in resource waste and a fragile process.

[0004] Existing patent CN119337780A discloses a modeling method for pulverized coal gasification furnaces based on a CFD-reduced model. This method involves constructing a flow network model and a chemical reaction model for CFD simulation, using the simulation results to build a reduced-order model, and then mixing it with a mechanistic model to achieve high-precision simulation through multi-dimensional coupling. Existing patent CN118070713A discloses a real-time simulation method for coking processes that couples numerical simulation with artificial neural networks. This method generates a simulation dataset through CFD and uses an optimized ANN model to achieve rapid prediction of the coking process.

[0005] In summary, although the aforementioned existing patents have improved computational efficiency in their respective fields through reduced-order models or neural network proxy models, none of them have proposed solutions to the problems of high computational cost and poor convergence stability in one-dimensional-three-dimensional coupled analysis of gas turbine air systems. In other words, they have failed to effectively decouple high-precision three-dimensional modeling from efficient system solving, making them difficult to directly apply to the engineering iterative design of complex air systems. Summary of the Invention

[0006] To address the problems existing in the prior art, this application proposes a gas turbine air system analysis method and system. By constructing a reduced-order model of non-standard components offline and encapsulating it into a standardized functional model unit, it achieves seamless integration with a one-dimensional fluid network environment. This decouples high-precision three-dimensional modeling from system-level iterative solution, thereby enabling efficient and stable analysis of the overall performance of the air system.

[0007] To achieve the above objectives, the first aspect of this application proposes a method for analyzing the air system of a gas turbine. The specific technical solution is as follows: A method for analyzing a gas turbine air system includes the following steps: S1. Construct a three-dimensional CFD benchmark model of non-standard components of the air system; S2. Establish a parametric CFD analysis process to generate a parameter combination matrix that covers the working range of non-standard components; S3. Use the parameter combination matrix to perform multi-condition three-dimensional CFD steady-state calculation in the three-dimensional CFD benchmark model to generate a dataset; S4. Based on the dataset, construct and verify a reduced-order model for non-standard components until its prediction accuracy meets the preset engineering requirements. S5. Encapsulate the verified reduced-order model into a functional model unit; S6. In a one-dimensional fluid network modeling environment, replace the original non-standard component model with the functional model unit, and call the functional model unit to perform system-level simulation iteration.

[0008] Furthermore, step S1 includes: extracting the fluid domain geometry of non-standard components, performing mesh generation, setting boundary conditions and turbulence models, and completing the construction of a three-dimensional CFD model.

[0009] Furthermore, step S2 also includes: defining input variables and output variables, and constructing a parameter space sampling strategy.

[0010] Furthermore, the input and output variables are defined based on the physical characteristics of the non-standard components.

[0011] Furthermore, the parameter space sampling strategy employs the equal-interval step size method, the Latin hypercube sampling method, or the full-factor sampling method.

[0012] Furthermore, in step S3, the multi-condition three-dimensional CFD steady-state calculation is based on the three-dimensional CFD benchmark model constructed in step S1, and is submitted to the CFD solver for execution in batches through automated scripts or parameterized solvers.

[0013] Furthermore, in step S4, the construction and verification of the reduced-order model of the non-standard component includes: dividing the dataset into a training set and a validation set; selecting a machine learning algorithm, using the input variables of the training set as features and the output variables as labels to train the model; using the validation set to evaluate the model's prediction accuracy, and if the prediction error exceeds a preset threshold, adjusting the model's hyperparameters or expanding the CFD sampling points, and retraining until the accuracy meets the engineering requirements.

[0014] Furthermore, the machine learning algorithm includes multilayer perceptron neural network, support vector regression, or Gaussian process regression.

[0015] Furthermore, adjusting the model hyperparameters includes adjusting the number of layers and nodes in the neural network, the learning rate, or the regularization coefficient.

[0016] Furthermore, in step S5, the encapsulation into a functional model unit refers to exporting the reduced-order model into a model exchange format or application programming interface standard.

[0017] Furthermore, the model exchange format includes the functional model unit format.

[0018] Furthermore, the functional model unit format is supported by mainstream one-dimensional system simulation software, including but not limited to Simcenter Amesim and Simcenter Flomaster.

[0019] The second aspect of this application proposes a gas turbine air system analysis system, the specific technical solution of which is as follows: A gas turbine air system analysis system, used to implement the above-mentioned gas turbine air system analysis method, includes: Offline modeling unit, used to build and encapsulate reduced-order models for non-standard components, and output functional model unit; The online analysis unit is used to integrate the functional model unit into the one-dimensional fluid network simulation environment and call the functional model unit for rapid calculation in the system-level simulation iteration.

[0020] Furthermore, the offline modeling unit includes: The benchmark model building module is used to build three-dimensional CFD benchmark models of non-standard components. The parameter analysis module is used to establish a parameterized CFD analysis process, define input and output variables, and generate a parameter combination matrix. The dataset generation module is used to perform multi-condition 3D CFD steady-state calculations and generate datasets. The reduced-order model construction and verification module is used to train and verify the reduced-order model based on the dataset until its prediction accuracy meets the preset engineering requirements. The reduced-order model encapsulation module is used to export the verified reduced-order model into a functional model unit.

[0021] Furthermore, the online analysis unit includes: The system integration analysis module is used to replace the original non-standard component models with the functional model units in a one-dimensional fluid network modeling environment, and to call the functional model units to perform system-level simulation iterations.

[0022] By applying the above-described technical solution of this application, at least the following technical effects are achieved: 1. This application separates the high-precision 3D modeling process from the system-level iterative solution process by constructing a reduced-order model of non-standard components offline and encapsulating it into standardized functional model units. The computationally expensive 3D CFD simulation is completed offline, while the online stage only calls the lightweight reduced-order model for rapid calculation, effectively solving the problem of tight coupling and mutual constraints between high-precision physical modeling and efficient system solution in existing technologies.

[0023] 2. This application employs a reduced-order model technique. In the offline phase, a training dataset is generated through high-fidelity 3D CFD simulation, ensuring that the reduced-order model inherits the physical accuracy of the 3D CFD. In the online phase, algebraic operations are performed by calling functional model units, achieving a computational speed comparable to that of standard one-dimensional components. Compared to existing technologies that require running a 3D CFD for each system iteration, this application significantly reduces computational costs while maintaining the simulation accuracy of non-standard components, effectively balancing the trade-off between accuracy and efficiency.

[0024] 3. In the system-level iterative solution process, this application directly calls the functional model unit constructed based on the reduced-order model. This unit is essentially a deterministic algebraic model and does not involve the iterative process of the 3D CFD solver. Compared with the prior art, which requires calling the easily divergent 3D CFD solver in each iteration, this application effectively avoids global solution failure caused by the non-convergence of 3D calculations, and significantly improves the convergence stability and computational reliability of one-dimensional fluid networks.

[0025] 4. This application exports the verified reduced-order model into a functional model unit format. This format is an open and standardized model exchange format that supports seamless integration with mainstream one-dimensional system simulation software. The encapsulated functional model unit can be directly reused as a standardized component in different projects and teams without repeating three-dimensional CFD modeling and simulation, greatly improving the efficiency of gas turbine air system design, analysis, and optimization, and has good engineering application value.

[0026] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0027] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings: Figure 1 A flowchart illustrating the gas turbine air system analysis method proposed in this application is presented; Figure 2 A schematic diagram of the framework of a gas turbine air system analysis system proposed in this application is shown; Figure 3 A schematic diagram of a computer system architecture required for the execution of a computer program as proposed in this application is shown; Figure 4 A schematic diagram of the hardware structure of an electronic device proposed in this application is shown. Detailed Implementation

[0028] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0029] The present application will be further described in detail below with reference to specific embodiments, which should not be construed as limiting the scope of protection claimed in the present application.

[0030] To address the problems of high cost, poor convergence, and lack of decoupling in the solution process of existing one-dimensional-three-dimensional coupled analysis techniques, this application proposes a gas turbine air system analysis method and system. This method achieves seamless integration with a one-dimensional fluid network environment by offline construction of reduced-order models of non-standard components and encapsulation into standardized functional model units. This decouples high-precision three-dimensional modeling from system-level iterative solution, significantly reducing computational costs and avoiding the convergence risks associated with embedding iterative loops in three-dimensional CFD. This enables efficient and stable analysis of the overall performance of the air system.

[0031] According to a first aspect of this application, a method for analyzing the air system of a gas turbine is provided, which will be described in detail below through specific embodiments.

[0032] Example 1

[0033] See Figure 1 As shown in the figure, this embodiment provides a gas turbine air system analysis method, which includes the following steps: S1. Obtain the fluid domain geometry of non-standard components of the air system, mesh and set boundary conditions, and construct a CFD baseline model; S2. Determine the key input parameters and output parameters to be predicted for non-standard components, set the value range and sampling strategy, and construct a parameter combination matrix covering the working range; S3. Based on the parameter combination matrix, automatically perform multi-condition CFD steady-state calculations, establish input-output mapping relationships, and generate high-precision training datasets; S4. Select machine learning algorithms for model training and accuracy verification to ensure that the prediction accuracy of the reduced-order model meets engineering requirements. S5. Export the verified reduced-order model as a standard functional model unit to achieve seamless integration with one-dimensional simulation software. S6. In a one-dimensional network, functional model units replace the original components, and the reduced-order model is directly called for fast calculation.

[0034] Step S1 above includes: Specifically, accurate three-dimensional geometric models of non-standard components in the gas turbine air system are obtained. These "non-standard components" refer to structural elements in the gas turbine air system that cannot be accurately described using empirical formulas, such as chambers with complex internal structures and irregularly shaped transition sections. This three-dimensional geometric model is typically the original design model directly imported from the CAD design system. After obtaining the original design model of the non-standard component, its complete fluid domain is extracted as the fluid domain geometric model. Necessary geometric cleanup and simplification processes are then performed on the extracted fluid domain geometric model, such as removing small features that have little impact on the main flow field, to ensure the correct geometric topology and suitability for subsequent mesh generation.

[0035] A meshing tool is used to generate a high-quality computational mesh for the fluid domain geometry model. In the mainstream region, for elements with regular geometry, a hexahedral core mesh is preferred to improve computational efficiency; for elements with complex geometry, an unstructured tetrahedral mesh is preferred to ensure geometric adaptability; in the near-wall region, a multi-layer prismatic mesh is generated to accurately capture the boundary layer flow. Preferably, the mesh quality is checked after meshing to ensure that the distortion, aspect ratio, and orthogonality meet the accuracy requirements of subsequent CFD solutions.

[0036] After obtaining the computational mesh, boundary conditions and solution parameters are further set. To simulate the working environment of non-standard components in a real air system, parameters such as total pressure, total temperature, and turbulence intensity are given at the inlet boundary; static pressure or mass flow rate conditions are given at the outlet boundary; no-slip conditions are adopted for the solid wall surface, and it is set as an adiabatic or isothermal wall surface according to the actual heat transfer requirements.

[0037] After setting the boundary conditions, a steady-state solution is performed in the CFD solver based on the computational grid with the established boundaries. Preferably, a pressure-based coupled solver is selected, a convergence residual threshold is set, and key physical quantities such as inlet and outlet flow balance are monitored in real time. After initializing the solution, iterative calculations are performed until all monitored indicators reach the convergence criteria, thereby obtaining the detailed flow field distribution of the non-standard component under given boundary conditions. The three-dimensional CFD baseline model constructed through the above process provides a reliable physical basis for subsequent multi-condition simulations.

[0038] Step S2 above includes: After completing step S1, which establishes the 3D CFD baseline model of the non-standard component, this step establishes a parametric analysis process, clarifies the key input parameters and output parameters to be predicted that affect the performance of the non-standard component, and constructs a parameter space sampling strategy that covers the working range of the non-standard component, laying the foundation for the generation of subsequent multi-condition simulation datasets.

[0039] Specifically, based on the operating environment of non-standard components in the actual air system, key inlet parameters are identified and selected as input variables. These input variables are independent boundary conditions or special operating condition parameters that affect the flow and heat transfer characteristics of the non-standard components. They directly reflect the influence of the external environment on the performance of the non-standard components, independent of their internal structure, and typically include fluid inlet state parameters and operating conditions unique to the non-standard components. The input variables can be flexibly defined according to the specific physical characteristics of the non-standard components. In this embodiment, for components whose main characteristic is fluid resistance, typical input variables include total inlet pressure and total inlet temperature; for rotating chambers, rotational speed can be added as an input variable.

[0040] Accordingly, the outlet parameters of the non-standard components that need to be predicted are determined as output variables. These output variables are the response results of the non-standard components under given input conditions, which comprehensively reflect the impact of the internal flow and heat transfer characteristics of the components on the overall system performance. They typically include the distribution or average value of key physical quantities such as mass flow rate, pressure, and temperature at the outlet cross-section of the non-standard component. The output variables can be flexibly defined according to the specific physical characteristics of the non-standard components. In this embodiment, for flow resistance and heat exchange components in an air system, typical output variables include outlet mass flow rate, outlet total pressure, and outlet total temperature; for heat exchangers, wall temperature can be added as an output variable.

[0041] It should be noted that the specific selection of input and output variables can be adapted to the actual working environment and performance analysis requirements of non-standard components. The above examples are for illustration only and do not constitute a limitation.

[0042] After determining the input and output variables, a parameter space sampling strategy is further constructed. Specifically, the value range of each input variable is set, including initial and final values, and the discretization method is determined based on the actual operating conditions of the non-standard components. Optionally, an equal-interval step size method can be used to discretize the input variables; alternatively, optimized sampling strategies such as Latin hypercube sampling or full-factor sampling can be used to generate a representative set of operating point values ​​within the input variable space, thereby achieving sufficient coverage of the component's operating range with limited CFD computing resources. Through the above parameter space sampling, multiple different operating point values ​​are generated, each corresponding to a set of input parameter combinations, forming the parameter combination matrix for subsequent CFD simulation calculations.

[0043] The parametric analysis process in this step ensures that the training dataset fully covers the operating range of the components, thus guaranteeing the generalization ability of the reduced-order model.

[0044] Step S3 above includes: After completing the construction of the parameter combination matrix in step S2, this step further performs multi-condition three-dimensional CFD steady-state calculations based on the parameter combination matrix to generate a high-precision dataset for training the reduced-order model.

[0045] Specifically, the input parameters corresponding to each working point in the parameter combination matrix generated in step S2 are used as boundary conditions, and are submitted to the CFD solver in batches using automated scripts or a parametric solver for calculation. Each working point corresponds to a set of input variable combinations. Based on the three-dimensional CFD baseline model constructed in step S1, the solver performs steady-state calculations one by one until each working point reaches the preset convergence criterion.

[0046] After the calculation for each operating condition is completed, the corresponding output parameters are extracted as output variables. The input and output parameters for each operating condition constitute a training sample, forming a precise mapping relationship between the input and output parameters, providing a data foundation for the subsequent construction of the order reduction model. Thus, the training samples corresponding to all operating conditions together constitute a high-precision dataset.

[0047] Preferably, during batch computation, necessary data quality checks can be performed on the computation results to remove abnormal operating points and ensure the reliability of the training dataset. Through the above automated batch processing flow, training samples corresponding to multiple operating points are generated. All samples together constitute a high-precision dataset, which comprehensively covers the working range of non-standard components in actual operation, providing a reliable data foundation for the subsequent construction of the reduced-order model.

[0048] Step S4 above includes: After generating the high-precision dataset in step S3, this step constructs a reduced-order model of non-standard components based on the dataset and verifies the prediction accuracy of the reduced-order model to ensure that it can accurately replace three-dimensional CFD calculations.

[0049] Specifically, the high-precision dataset generated in step S3 is divided into a training set and a validation set. Preferably, random partitioning or stratified sampling can be used to divide the dataset into a training set and a validation set proportionally, where the training set is used for training the reduced-order model and the validation set is used to evaluate the generalization ability of the reduced-order model.

[0050] Based on the data partitioning, a suitable machine learning algorithm is selected as the kernel of the order reduction model. In this embodiment, the machine learning algorithm includes, but is not limited to, Multilayer Perceptron Neural Network (MLP), Support Vector Regression (SVR), or Gaussian Process Regression (GPR). Using the input variables of the training set as features and the output variables as labels, the order reduction model is trained to learn and fit the nonlinear mapping relationship between the input and output parameters.

[0051] After training the reduced-order model, its prediction accuracy is evaluated using a validation set. The prediction error of the reduced-order model on the validation set is calculated, using metrics such as root mean square error (RMSE) or mean absolute percentage error (MAS). Preferably, a threshold accuracy required for engineering applications can be set, such as a mass flow prediction error of less than 5% and a total pressure loss prediction error of less than 3%. If the reduced-order model's prediction error exceeds the preset threshold, the process returns to adjust the model's hyperparameters, including the number of layers and nodes in the neural network, the learning rate, and the regularization coefficient, or by increasing the CFD sampling points in step S2 to expand the dataset. The reduced-order model is then retrained and validated until its accuracy meets engineering requirements.

[0052] Step S5 above includes: After constructing and verifying the qualification of the reduced-order model in step S4, this step further encapsulates the trained reduced-order model into a standardized functional model unit to achieve seamless integration with the one-dimensional fluid network simulation environment.

[0053] Specifically, the validated reduced-order model is exported as a functional model unit format. This format is an open and standardized model exchange format that supports the sharing and reuse of models across different simulation platforms. Preferably, the functional model unit format is supported by mainstream one-dimensional system simulation software, including but not limited to SimcenterAmesim and Simcenter Flomaster.

[0054] It should be noted that the functional model unit format is only one preferred implementation of this application. The encapsulation format of the reduced-order model is not limited to the functional model unit. Other common model exchange formats or application programming interface standards can also be used, as long as they can achieve seamless integration with the one-dimensional fluid network simulation environment.

[0055] By encapsulating the reduced-order model into functional model units, the high-precision physical properties of non-standard components can be embedded into a one-dimensional system analysis environment in a lightweight form.

[0056] Step S6 above includes: After completing the encapsulation and integration in step S5, this step applies the encapsulated reduced-order model to the overall one-dimensional network analysis of the air system.

[0057] Specifically, in the one-dimensional fluid network modeling environment, the original non-standard component model is replaced by the functional model unit generated in step S5. During the system-level simulation iteration process, when it is necessary to calculate the performance of the non-standard component, the simulation software directly calls the functional model unit for rapid calculation and outputs the corresponding output variables in real time according to the input parameters of the current iteration step.

[0058] Since the functional model unit is constructed based on a reduced-order model, its calculation process involves only lightweight numerical operations, without requiring any 3D CFD calculations. This approach effectively decouples high-precision 3D modeling from system-level iterative solutions, significantly reducing the computational cost of system analysis while ensuring the simulation accuracy of non-standard components. It also avoids the convergence risks that may arise from embedding 3D CFD into iterative loops, thereby achieving efficient and stable analysis of the overall performance of the gas turbine air system.

[0059] In summary, the gas turbine air system analysis method provided in the first aspect of this application effectively decouples high-precision 3D modeling from system-level iterative solutions by offline construction of reduced-order models of non-standard components and encapsulation of these models into standardized functional model units. This method first constructs a baseline model of the non-standard components based on 3D CFD, generating a high-precision dataset covering the working range through parametric sampling and batch simulation. Then, it uses machine learning algorithms to train and validate the reduced-order model, ensuring its accuracy in predicting the characteristics of non-standard components. Finally, the validated reduced-order model is encapsulated into functional model units and integrated into a one-dimensional fluid network simulation environment, enabling rapid access and accurate calculation of non-standard components. Compared to existing methods that directly embed 3D CFD into system-level iterative loops, this method significantly reduces computational costs, avoids the risk of solution failure due to non-convergence in 3D calculations, and achieves efficient and stable analysis of the overall performance of the gas turbine air system while ensuring the simulation accuracy of non-standard components, demonstrating significant engineering application value.

[0060] Example 2

[0061] This embodiment uses the rotor disk cavity of a heavy-duty gas turbine as a specific example of a non-standard component. The rotor disk cavity is an annular chamber surrounded by adjacent turbine disks and mini-discs, containing high-speed rotation, a strong centrifugal force field, and a complex secondary flow structure. Traditional one-dimensional modeling methods are insufficient to accurately describe its flow characteristics. This embodiment will elaborate on the specific application process of the method described in this application on this type of component. Specifically, it includes the following steps: S1. Construct a three-dimensional CFD benchmark model of the rotor disk cavity; Specifically, a complete three-dimensional geometric model including the rotor disk cavity and inlet and outlet ports is established, and the rotational speed N of the rotating wall is set as a key operating parameter. A pre-swirl induced draft method is adopted at the inlet, and the total inlet pressure Pin, total inlet temperature Tin, and swirl ratio β are set as boundary conditions. The outlet boundary leads to the central cavity. The geometric model is meshed, a turbulence model is set, and a three-dimensional CFD baseline model is built, providing a reliable physical basis for subsequent multi-condition simulations.

[0062] S2. Establish a parametric analysis process and define input and output variables; Specifically, based on the actual operating environment of the rotor disk cavity, the key input variables affecting its performance and the output variables to be predicted are determined. Input variables include inlet total pressure Pin, inlet total temperature Tin, rotational speed N, and swirl ratio β; output variables include sealing air mass flow rate. The seal, the average temperature of the cavity (Tcavity), and the characteristic value of the pressure distribution within the cavity (ΔP).

[0063] The range of values ​​for each input variable is set. In this embodiment, the inlet total pressure is 8-10 bar, the inlet total temperature is 650-750 K, the rotational speed is 3000-5000 rpm, and the swirl ratio is 0.5-1.0. A Latin hypercube sampling strategy is adopted to generate 50 representative operating points in the input variable space, forming a parameter combination matrix to ensure sufficient coverage of the rotor cavity's operating range.

[0064] S3. Perform multi-condition CFD calculations to generate a training dataset; Specifically, based on the 3D CFD baseline model constructed in step S1, a multi-condition steady-state rotating flow field simulation was performed using the ANSYS CFX solver. The input variables corresponding to the 50 sets of operating points generated in step S2 were used as boundary conditions, and calculations were submitted in batches via an automated script. For each operating point, a convergence residual of less than 1 × 10⁻⁶ was set. -5 The output variable for each operating condition is extracted, and the input and output variables for each operating condition constitute a training sample, forming a precise mapping relationship between the input and output variables. The training samples from all operating conditions together constitute the dataset used for training the reduced-order model.

[0065] S4. Construct and verify a reduced-order model of the rotor disk cavity; Specifically, the dataset generated in step S3 is divided into a training set and a validation set. Preferably, a random partitioning method is used, dividing the dataset into a training set and a validation set in an 8:2 ratio, where the training set contains 40 samples and the validation set contains 10 samples. A Gaussian process regression algorithm is selected as the kernel of the reduced-order model, as this algorithm has excellent generalization ability for small-sample, highly nonlinear problems. The model is trained using the input variables of the training set as features and the output variables as labels, enabling the reduced-order model to learn and fit the nonlinear mapping relationship between the input and output variables. The prediction accuracy of the reduced-order model is evaluated using the validation set, and the prediction error of the reduced-order model on the validation set is calculated. The validation results show that the maximum prediction error of the sealed air mass flow rate is 1.8%, the prediction error of the average temperature in the cavity is less than 5 K, and the goodness of fit R² of the cavity pressure distribution feature value is greater than 0.96, indicating that the accuracy of the reduced-order model meets the engineering requirements. If the prediction error of the reduced-order model exceeds a preset threshold, the process is repeated to adjust the hyperparameters of the reduced-order model or expand the CFD sampling points, and the reduced-order model is retrained and validated until the accuracy of the reduced-order model meets the engineering requirements.

[0066] S5. Encapsulate the reduced-order model into a functional model unit; Specifically, the validated reduced-order model is exported as a functional model unit format, supporting model sharing and reuse across different simulation platforms. The interface of this functional model unit is defined as follows: input variables include inlet total pressure Pin, inlet total temperature Tin, rotational speed N, and swirl ratio β; output variables include sealing air mass flow rate. The seal, average cavity temperature Tcavity, and cavity pressure distribution characteristic value ΔP are analyzed. By encapsulating the reduced-order model into functional model units, the high-precision physical characteristics of the rotor cavity can be embedded into a one-dimensional system analysis environment in a lightweight form.

[0067] S6. Perform whole-machine air system simulation based on functional model units.

[0068] Specifically, in the Simcenter Flomaster one-dimensional fluid network modeling environment, the functional model unit generated in step S5 replaces the original simplified model of "cavity + equivalent flow resistance" for multi-condition simulation analysis of the whole machine's air system. During the system-level simulation iteration process, the simulation software directly calls this functional model unit for rapid calculation and outputs the corresponding output variables in real time according to the input variables of the current iteration step. Simulation results show that the single simulation time is shortened to 1 minute, which is significantly improved in computational efficiency compared to the estimated computation time of more than 30 hours in traditional one-dimensional-three-dimensional real-time coupled analysis; the predicted sealing air mass flow rate deviates from the experimental data by less than 5%; and the system maintains stable convergence throughout the variable speed process.

[0069] By using the above method, high-precision 3D modeling and system-level iterative solution are effectively decoupled. While ensuring the accuracy of rotor disk cavity simulation, the computational cost of system analysis is significantly reduced. At the same time, the convergence risk that may be brought about by embedding 3D CFD into iterative loops is avoided, thereby realizing efficient and stable analysis of the overall performance of the gas turbine air system.

[0070] According to a second aspect of this application, a gas turbine air system analysis system is provided, which will be described in detail below through specific embodiments.

[0071] Example 3

[0072] See Figure 2 As shown, the present application provides a gas turbine air system analysis system, which includes an offline modeling unit and an online analysis unit.

[0073] The offline modeling unit includes a baseline model building module, a parameter analysis module, a dataset generation module, a reduced-order model building and verification module, and a reduced-order model encapsulation module, used to perform one-time reduced-order model building and encapsulation for non-standard components. The online analysis unit includes a system integration analysis module, used to integrate the encapsulated functional model units into a one-dimensional fluid network simulation environment, directly calling the functional model units for rapid calculation during system-level simulation iterations to achieve high-precision one-dimensional system analysis.

[0074] The benchmark model construction module is used to execute step S1, extract the accurate fluid domain geometry of non-standard components, perform high-quality mesh generation, set boundary conditions and turbulence models, and complete the construction of the three-dimensional CFD benchmark model, providing a reliable physical basis for subsequent multi-condition simulation.

[0075] The parameter analysis module is used to execute step S2, establish a parameterized CFD analysis process, identify and select key input parameters that affect the performance of non-standard components as input variables, select output variables that need to be predicted, set the value range and sampling strategy of each input variable, and generate a parameter combination matrix that covers the working range of non-standard components.

[0076] The dataset generation module is used to execute step S3, which involves submitting the parameter combination matrix to the CFD solver in batches using an automated script, performing multi-condition steady-state calculations based on the three-dimensional CFD benchmark model, extracting the output variables of each condition point, and generating a high-precision dataset containing multiple sets of training samples.

[0077] The reduced-order model construction and verification module is used to execute step S4, which divides the high-precision dataset into a training set and a verification set, selects a machine learning algorithm to train the reduced-order model, uses the verification set to evaluate the model's prediction accuracy, and iteratively optimizes the model by adjusting the model's hyperparameters or expanding the CFD sampling points until the model accuracy meets the preset engineering requirements.

[0078] The reduced-order model encapsulation module is used to execute step S5, exporting the verified reduced-order model as a functional model unit format. This format is a standardized model exchange format that supports seamless integration with mainstream one-dimensional system simulation software.

[0079] The system integration analysis module is used to execute step S6, replacing the original non-standard component model with the functional model unit in the one-dimensional fluid network modeling environment, directly calling the functional model unit for rapid calculation during the system-level simulation iteration process, and outputting the corresponding output parameters in real time according to the input parameters of the current iteration step.

[0080] Through the above two-stage architecture, this system effectively decouples high-precision 3D modeling from system-level iterative solution. The offline stage completes a one-time high-cost calculation, while the online stage only requires lightweight calculation to complete the system analysis, significantly reducing the cost of repeated calculations and avoiding the convergence risk caused by embedding 3D CFD into iterative loops, thus achieving efficient and stable analysis of the overall performance of the gas turbine air system.

[0081] According to a third aspect of this application, a computer program product is provided, comprising a computer program. When the computer program is executed by a processor, it is capable of implementing the gas turbine air system analysis method of the first aspect of this application.

[0082] Figure 3 A schematic block diagram of a computer system 300 that can be used to run the computer program is shown. This computer system 300 can serve as an example of a hardware environment for performing the methods described above.

[0083] It should be noted that, Figure 3 The computer system 300 shown is merely an example and should not be construed as limiting the functionality and scope of this application.

[0084] Figure 3As shown, the computer system 300 includes a central processing unit (CPU) 301, which can perform various appropriate actions and processes based on programs stored in read-only memory (ROM) 302 or programs loaded from storage section 308 into random access memory (RAM). The RAM 303 also stores various programs and data required for system operation. The CPU 301, ROM 302, and RAM 303 are interconnected via a bus 304. An input / output interface 305 (I / O interface) is also connected to the bus 304.

[0085] The following components are connected to the input / output interface 305: an input section 306 including a keyboard, mouse, etc.; an output section 307 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and speakers, etc.; a storage section 308 including a hard disk, etc.; and a communication section 309 including a network interface card such as a local area network card, modem, etc. The communication section 309 performs communication processing via a network such as the Internet. A drive 310 is also connected to the input / output interface 305 as needed. A removable medium 311, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on the drive 310 as needed so that computer programs read from it can be installed into the storage section 308 as needed.

[0086] Specifically, the process described in the gas turbine air system analysis method according to the first aspect can be implemented as a computer software program. For example, embodiments of this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowchart. In such embodiments, the computer program can be downloaded and installed from a network via communication section 309, and / or installed from a removable medium. When the computer program is executed by central processing unit 301, the gas turbine air system analysis method defined in this application is performed.

[0087] According to a fourth aspect of this application, an electronic device is also provided. This embodiment will be described using this electronic device as an example of a terminal device. Figure 4 As shown, the electronic device includes a memory 402 and a processor 404. The memory 402 stores a computer program, and the processor 404 is configured to execute the above-described gas turbine air system analysis method through the computer program.

[0088] Optionally, in this embodiment, the aforementioned electronic device may be located in at least one of a plurality of network devices in a computer network.

[0089] Alternatively, as those skilled in the art will understand, Figure 4 The structure shown is for illustrative purposes only. Figure 4 This does not limit the structure of the aforementioned electronic devices. For example, the electronic device may also include components that are more... Figure 4 The more or fewer components shown (such as network interfaces, etc.), or having the same Figure 4 The different configurations shown.

[0090] The memory 402 can be used to store software programs and modules, such as the corresponding program instructions / modules of the gas turbine air system analysis method described above in this application. The processor 404 executes various functional applications and data processing by running the software programs and modules stored in the memory 402. The memory 402 may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. Specifically, the memory 402 may be used, but is not limited to, to store parameter combination matrices, high-precision datasets, reduced-order model parameters, and system analysis results. As an example, such as... Figure 4 As shown, the memory 402 described above may include, but is not limited to, the functional modules of the gas turbine air system analysis system described in the second aspect of this application. In addition, it may include other module units, which will not be elaborated in this example.

[0091] Optionally, the transmission device 406 described above is used to receive or send data via a network. Specific examples of the network described above may include wired networks and wireless networks.

[0092] In addition, the aforementioned electronic device also includes: a display 408 for displaying information such as parameter combination matrix, high-precision dataset, reduced-order model accuracy evaluation results, and system analysis results; and a connection bus 410 for connecting various module components in the aforementioned electronic device.

[0093] According to a fifth aspect of this application, a computer-readable storage medium is provided, wherein a processor of an electronic device reads computer instructions from the computer-readable storage medium and executes the computer instructions, causing the electronic device to perform the gas turbine air system analysis method described in the first aspect of this application.

[0094] Optionally, in this embodiment, those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing the hardware related to the terminal device. The program can be stored in a computer-readable storage medium, which may include: flash drive, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc.

[0095] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0096] If the integrated units in the above embodiments are implemented as software functional units and sold or used as independent products, they can be stored in the aforementioned computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause one or more electronic devices to execute all or part of the steps of the methods described in the various embodiments of this application.

[0097] In the above embodiments of this application, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0098] In the several embodiments provided in this application, it should be understood that the disclosed application can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual couplings, direct couplings, or communication connections may be through some interfaces; indirect couplings or communication connections between units or modules may be electrical or other forms.

[0099] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0100] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0101] The above are merely several specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in this application, based on the technical solution and inventive concept of this application, should be included within the scope of protection of this application.

[0102] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0103] It should be noted that, in the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions 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 one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

Claims

1. A method for analyzing the air system of a gas turbine, characterized in that, Includes the following steps: S1. Construct a three-dimensional CFD benchmark model of non-standard components of the air system; S2. Establish a parametric CFD analysis process to generate a parameter combination matrix that covers the working range of non-standard components; S3. Use the parameter combination matrix to perform multi-condition three-dimensional CFD steady-state calculation in the three-dimensional CFD benchmark model to generate a dataset; S4. Based on the dataset, construct and verify a reduced-order model for non-standard components until its prediction accuracy meets the preset engineering requirements. S5. Encapsulate the verified reduced-order model into a functional model unit; S6. In a one-dimensional fluid network modeling environment, replace the original non-standard component model with the functional model unit, and call the functional model unit to perform system-level simulation iteration.

2. The gas turbine air system analysis method according to claim 1, characterized in that, Step S1 includes: extracting the fluid domain geometry of non-standard components, performing mesh generation, setting boundary conditions and turbulence models, and completing the construction of the three-dimensional CFD model.

3. The gas turbine air system analysis method according to claim 1, characterized in that, Step S2 also includes: defining input and output variables, and constructing a parameter space sampling strategy.

4. The gas turbine air system analysis method according to claim 3, characterized in that, The input and output variables are defined based on the physical characteristics of the non-standard components.

5. The gas turbine air system analysis method according to claim 3, characterized in that, The parameter space sampling strategy employs the equal-interval step size method, the Latin hypercube sampling method, or the full-factor sampling method.

6. The gas turbine air system analysis method according to claim 1, characterized in that, In step S3, the multi-condition three-dimensional CFD steady-state calculation is based on the three-dimensional CFD benchmark model constructed in step S1, and is submitted to the CFD solver for execution in batches through automated scripts or parameterized solvers.

7. The gas turbine air system analysis method according to claim 1, characterized in that, In step S4, the construction and verification of the reduced-order model of the non-standard component includes: dividing the dataset into a training set and a validation set; selecting a machine learning algorithm, using the input variables of the training set as features and the output variables as labels to train the model; using the validation set to evaluate the model's prediction accuracy, and if the prediction error exceeds a preset threshold, adjusting the model's hyperparameters or expanding the CFD sampling points, and retraining until the accuracy meets the engineering requirements.

8. The gas turbine air system analysis method according to claim 7, characterized in that, The machine learning algorithms include multilayer perceptron neural networks, support vector regression, or Gaussian process regression.

9. The gas turbine air system analysis method according to claim 7, characterized in that, The adjustment of model hyperparameters includes adjusting the number of layers and nodes in the neural network, the learning rate, or the regularization coefficient.

10. The gas turbine air system analysis method according to claim 1, characterized in that, In step S5, encapsulating into a functional model unit means exporting the reduced-order model into a model exchange format or application programming interface standard.

11. The gas turbine air system analysis method according to claim 10, characterized in that, The model exchange format includes the functional model unit format.

12. The gas turbine air system analysis method according to claim 11, characterized in that, The functional model unit format is supported by mainstream one-dimensional system simulation software, including but not limited to Simcenter Amesim and Simcenter Flomaster.

13. A gas turbine air system analysis system, used to implement the gas turbine air system analysis method as described in any one of claims 1-12, characterized in that, include: Offline modeling unit, used to build and encapsulate reduced-order models for non-standard components, and output functional model unit; The online analysis unit is used to integrate the functional model unit into the one-dimensional fluid network simulation environment and call the functional model unit for rapid calculation in the system-level simulation iteration.

14. The gas turbine air system analysis system according to claim 13, characterized in that, The offline modeling unit includes: The benchmark model building module is used to build three-dimensional CFD benchmark models of non-standard components. The parameter analysis module is used to establish a parameterized CFD analysis process, define input and output variables, and generate a parameter combination matrix. The dataset generation module is used to perform multi-condition 3D CFD steady-state calculations and generate datasets. The reduced-order model construction and verification module is used to train and verify the reduced-order model based on the dataset until its prediction accuracy meets the preset engineering requirements. The reduced-order model encapsulation module is used to export the verified reduced-order model as a functional model unit.

15. The gas turbine air system analysis system according to claim 13, characterized in that, The online analysis unit includes: The system integration analysis module is used to replace the original non-standard component models with the functional model units in a one-dimensional fluid network modeling environment, and to call the functional model units to perform system-level simulation iterations.

16. A computer program product, comprising a computer program, characterized in that, When executed by a processor, the computer program implements the gas turbine air system analysis method according to any one of claims 1 to 12.

17. An electronic device comprising a memory and a processor, characterized in that, The memory stores a computer program, and the processor is configured to execute the gas turbine air system analysis method according to any one of claims 1 to 12 through the computer program.

18. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored computer program, wherein the computer program can be executed by an electronic device to perform the gas turbine air system analysis method according to any one of claims 1 to 12.

Images