A method for optimizing the flow distribution of lubricating oil circuits in high-efficiency gas turbine equipment
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-05-21
- Publication Date
- 2026-08-14
AI Technical Summary
[0006]本发明的目的是提供一种面向高效燃机装备的润滑油路流量分配优化设计方法，解决现有技术存在的解析计算方法精度不足、无法适配多支路耦合场景，CFD仿真方法依赖人工试算、设计效率低、成本高，难以同时兼顾设计精度与优化效率，缺乏多支路流量快速预测与系统化优化手段的问题
（1）提高流量分配设计的准确性：本发明通过批量CFD构建高维数据集，并采用神经网络建立多输入多输出映射模型，能够充分考虑多支路之间的耦合非线性关系，提高流量预测精度，避免传统解析模型因简化假设带来的误差；
Smart Images
Figure CN122572159A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of lubrication system design technology for gas turbine equipment, and in particular to an optimized design method for lubrication oil flow distribution in high-efficiency gas turbine equipment. Background Technology
[0002] In highly integrated equipment such as high-efficiency gas turbines and high-speed gearboxes, multiple branch lubrication circuits are used to provide lubrication and cooling for critical friction pairs such as bearings and gears. The accuracy of flow distribution directly affects the reliability of equipment operation. Currently, the design of lubrication circuit flow distribution mainly adopts two methods: analytical calculation and CFD numerical simulation.
[0003] Traditional analytical methods, when dealing with a large number of branches, exhibit significant coupling relationships between the orifices. Changing the orifice size of a particular branch not only affects the flow rate of that branch but also influences the flow rates of other branches through redistribution. Furthermore, analytical methods typically assume a constant flow coefficient and linear superposition of local losses; however, in real-world highly integrated systems, three-dimensional local flows are complex, local flow resistance varies with operating conditions, and pressure drops between branches are coupled. Therefore, analytical models struggle to accurately reflect actual flow distribution, and calculation errors are significantly amplified with the increase in the number of branches.
[0004] When using CFD-based simulation methods for design, it is typically necessary to manually modify the orifice size, update the geometric model, re-mesh, recalculate, analyze the results, and readjust the design. When there are n orifices, and each orifice has multiple optional sizes, the design space grows exponentially. Manual trial and error calculations are insufficient to cover the entire design space. Furthermore, CFD calculations are time-consuming, and repeated iterations during the design phase consume significant time and manpower.
[0005] In the engineering design phase, it is often necessary to quickly assess the impact of different throttling schemes on flow distribution. Existing technologies struggle to simultaneously achieve both accuracy and efficiency, lacking rapid prediction and systematic optimization methods for multi-branch flow distribution. Therefore, there is an urgent need for a high-precision, high-efficiency, and applicable method for optimizing the flow distribution of lubricating oil circuits in multi-branch coupled systems. Summary of the Invention
[0006] The purpose of this invention is to provide an optimized design method for lubricating oil circuit flow distribution in high-efficiency gas turbine equipment, which solves the problems of insufficient accuracy of analytical calculation methods, inability to adapt to multi-branch coupling scenarios, reliance on manual calculations, low design efficiency and high cost of CFD simulation methods, difficulty in simultaneously balancing design accuracy and optimization efficiency, and lack of rapid prediction and systematic optimization methods for multi-branch flow.
[0007] To achieve the above objectives, this invention provides a method for optimizing the flow distribution of lubricating oil circuits in high-efficiency gas turbine equipment, comprising the following steps: S1. Determine the throttle orifice and outlet parameters of the lubricating oil circuit; use the throttle orifice diameter as the input variable and the outlet flow rate as the output variable; S2. Multiple sets of orifice diameter combination samples are generated using Latin hypercube sampling; S3. Perform parametric geometric modeling of the lubrication circuit so that the orifice diameter can drive the model to update automatically; S4. Construct a unified CFD flow field calculation model, number them sequentially, and set the outlet flow rate as the output parameter; S5. Based on the parametric model, automatically complete batch CFD simulation of multiple samples and construct a dataset corresponding to the orifice diameter and outlet flow rate; S6. Use the dataset as the basis to train a multi-input multi-output surrogate model and establish a nonlinear mapping relationship between the orifice diameter and the outlet flow rate. S7. Use a surrogate model to perform feature contribution analysis and quantify the degree and trend of the influence of each throttling orifice on the flow rate of each outlet. S8. Quickly predict flow rate and select the optimal throttling orifice scheme through the surrogate model, and then complete the verification closed loop through CFD simulation.
[0008] Preferably, the determination of the throttle orifice and outlet parameters of the lubricating oil circuit in step S1 is specifically as follows: S11. Determine the number of throttling orifices in the lubrication system. The export quantity is ; S12, Definition of the The orifice diameter is the input parameter. , And determine the range of values for each aperture. ; Define the first The flow rate of each outlet is a response output parameter. , ; S13. Transform the flow allocation problem into a multi-input multi-output mapping relationship; where the input vector is... The output vector is .
[0009] Preferably, step S2 specifically involves: determining the range of values for the diameter of each throttling orifice. Inner Latin hypercube sampling (LHS) method generates A set of orifice combination schemes was developed to form a sample set. ,in, , The total number of samples, For sample index, For the first The input vector of the group of samples, For the first The orifice diameter value of the first throttling orifice in the group scheme. For the first The orifice diameter value of the second throttling orifice in the group scheme, For the first The first group of schemes The orifice diameter value of each throttling orifice.
[0010] Preferably, the parametric geometric modeling of the lubrication circuit in step S3 specifically involves: in the parametric modeling platform, determining the orifice diameters of all throttling orifices... Defined as automatically updatable parameter variables, a lubrication circuit geometric model capable of automatically updating with the parameters is established; Import the parameter set for the group of throttling orifice combination schemes to form a batch calculation task list.
[0011] Preferably, the construction of a unified CFD flow field calculation model in step S4 specifically involves: establishing a unified calculation configuration in the CFD solver, including fixing lubricating oil material parameters, turbulence model, boundary conditions and convergence criteria, to ensure that the calculation conditions of all schemes are consistent.
[0012] Preferably, step S5 specifically involves: initiating batch solving, for The combined throttling orifice scheme sequentially performs geometry update, mesh update, CFD solution, and output extraction to obtain the flow rates at each outlet: in, For export volume, For the first The flow rate of the first outlet under the group scheme. For the first The flow rate of the second outlet under the group scheme, For the first Group Scheme No. The flow rate of each outlet; Finally, the dataset was obtained. .
[0013] Preferably, step S6, training the multi-input multi-output agent model, specifically includes the following steps: S61. Based on the dataset obtained in step S5, combine the orifice diameters. As input to the model, the corresponding outflow flow As model output; S62. Standardize the input and output data; S63. Divide the dataset into training set, validation set and test set according to the ratio of 70%, 15% and 15% respectively; S64. Construct a multi-input multi-output neural network model with the following structure: the number of nodes in the input layer equals the number of orifices. The hidden layer uses a single hidden layer with 18 neurons, and the number of nodes in the output layer equals the number of output nodes. ; S65. Establish model mapping relationships: in, For the export flow predicted by the model, For neural network proxy models, For neural network model parameters; S66. Use mean squared error as the loss function, use the training set to train the model, monitor the training process through the validation set, and use early stopping or regularization to suppress overfitting. S67. The accuracy of the model is verified using the test set, resulting in a multi-input multi-output proxy model for fast traffic prediction.
[0014] Preferably, the feature contribution analysis in step S7 specifically involves: using Shapley's additivity interpretation (SHAP) analysis to calculate the contribution, thereby obtaining the marginal contribution of each input variable to the output predicted value. Based on this, the sensitivity of each throttling orifice to the outlet flow rate, the trend of the effect of increasing or decreasing the throttling orifice on the outlet flow rate, and the conflict between different outlets can be determined.
[0015] Preferably, step S8 specifically involves: pre-determining the target flow rate for each outlet. ,by To constrain design, a surrogate model is used to quickly calculate the predicted output. , to make the predicted output With target traffic To minimize the error between the two, the optimal combination of throttling orifice diameters that meets the design requirements is selected. The optimal combination is then substituted into the CFD model for verification. When the error exceeds the allowable range, additional samples are added and the surrogate model is iteratively updated to form a design closed loop.
[0016] Therefore, the present invention employs the above-mentioned method for optimizing the flow distribution of lubricating oil circuits for high-efficiency gas turbine equipment, which has the following beneficial effects: (1) Improve the accuracy of traffic allocation design: This invention constructs a high-dimensional dataset through batch CFD and establishes a multi-input multi-output mapping model using neural networks, which can fully consider the coupling nonlinear relationship between multiple branches, improve the accuracy of traffic prediction, and avoid the error caused by the simplification assumptions of traditional analytical models; (2) Significantly improve design efficiency: Through parametric modeling and automatic batch calculation mechanism, the process of manual repetitive modeling and successive trial calculation is transformed into programmatic processing. After the surrogate model is trained, the flow prediction of multiple throttling schemes can be completed in a very short time, which significantly shortens the design cycle compared with successive CFD calculation; (3) Reduce manpower and computing costs: Only one CFD program debugging is required in the early stage, and then batch calculations can be performed to establish a global mapping model; subsequent design adjustments do not require large-scale numerical simulations, thereby reducing manpower input and computing resource consumption in the engineering implementation stage. (4) Applicable to high-dimensional, multi-branch complex systems: The method of this invention does not depend on a specific number of branches. When the number of throttling orifices increases, high-dimensional system design can still be achieved by expanding the number of samples and model training. It has good scalability and engineering applicability.
[0017] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description
[0018] Figure 1 This is a flowchart illustrating the method according to an embodiment of the present invention; Figure 2 This is a diagram illustrating the highly integrated lubrication circuit system and the numbering of each throttling orifice and flow outlet according to an embodiment of the present invention. Figure 3 These are 400 design sample diagrams in the 8-dimensional space of this invention embodiment; Figure 4 This is a diagram showing the learning results of 12 output flows in an embodiment of the present invention; Figure 5 This is a SHAP analysis diagram of an embodiment of the present invention. Detailed Implementation
[0019] The following detailed description of embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.
[0020] Example 1: Optimization Design of Flow Distribution in a Highly Integrated Lubrication Circuit System This embodiment focuses on a highly integrated lubrication circuit system, which includes one main oil circuit, eight branch throttling orifices, and twelve outlets (corresponding to the lubrication positions of different bearings or gears). The inlet oil supply conditions are constant. The embodiment describes an optimized design method for lubrication circuit flow distribution in high-efficiency gas turbine equipment. Figure 1 As shown, the specific steps include: Step 1: Determine the structure and parameterized objects, such as Figure 2 As shown In the 3D software (Solidworks is selected for modeling in this example, and ANSYS Workbench is selected for calculation), the number of throttling orifices is defined. Export volume The orifice diameters of the eight throttling orifices are defined as input variables: The range of values for the orifice diameter of each throttling orifice is shown in Table 2, and is determined based on the feasibility of processing and the diameter of the pipeline to which it belongs. Table 2 Orifice Diameter Range
[0021] In the ANSYS Workbench parameter module, select all eight apertures mentioned above as parameter variables.
[0022] Step 2: Latin hypercube sampling method for generating samples The Latin hypercube sampling method was used to generate 400 orifice combination schemes, forming a sample set: ,in, , For sample index, For the first The input vector of the group of samples, For the first The orifice diameter value of the first throttling orifice in the group scheme. For the first The orifice diameter value of the second throttling orifice in the group scheme, For the first The orifice diameter value of the 8th throttling orifice in the group scheme.
[0023] The normalized distribution of the sampling space is as follows Figure 3 As shown, each variable is evenly distributed across its range of values.
[0024] Step 3: Establish a geometric parametric model of the pipeline and import the sample scheme. A three-dimensional geometric model of the lubrication pipeline is created in a three-dimensional modeling software (SolidWorks is used in this embodiment) and then imported into a parametric modeling platform (ANSYS Workbench is used in this embodiment).
[0025] Parametric Definition: The eight orifice diameters in the pipeline geometry are parametrically named and defined as parameter variables that can be assigned values externally. Specifically, each orifice diameter is labeled as a parameter. This allows the geometric model to be automatically updated based on changes in parameter values.
[0026] Sample import: The 400 orifice combination schemes generated by Latin hypercube sampling in step S2 are imported. The parameters are input into the parameter set of the parameterization platform to form a batch task list containing 400 computational tasks.
[0027] Step 4: Construct the CFD solution model and define the output parameters A unified fluid simulation solution model is established in the CFD solver (ANSYS Fluent is used in this embodiment), and the specific settings are as follows: Boundary conditions: velocity inlet, pressure outlet; Fluid properties: The density and viscosity of the lubricating oil are set according to the actual oil application design; Turbulence model: SST was used Turbulence model; Output parameter definition: Number each of the 12 outlets and define the flow rate of each outlet. , , , , , , , , , , , Defined as an output parameter; Step 5: Automatically batch calculate and generate datasets Initiate a batch solution task and sequentially execute the following automated process on the 400 orifice schemes generated in step 2: geometry update, mesh update, CFD solution, and output extraction to obtain the outlet flow rates: in, For the first The flow rate of the first outlet under the group scheme. For the first The flow rate of the second outlet under the group scheme, For the first The flow rate of the 12th outlet under the group scheme; After calculating all 400 schemes, the complete dataset is obtained: .
[0028] Step 6: Construct a neural network proxy model Construct a fully connected neural network with a single hidden layer, as follows: Input layer: 8 nodes; Hidden layer: 18 neurons; Output layer: 12 nodes.
[0029] The dataset was divided into a training set (70%), a validation set (15%), and a test set (15%).
[0030] Establish model mapping relationships: in, For the export flow predicted by the model, For neural network proxy models, For neural network model parameters; The mean squared error is used as the loss function, the training set is used for model training, the training process is monitored by the validation set, and early stopping or regularization is used to suppress overfitting. The accuracy of the model was verified using a test set, resulting in a multi-input multi-output proxy model for fast traffic prediction.
[0031] Training results as follows Figure 4 As shown, Figure 4 This figure compares the predicted values of the neural network surrogate model for 12 outflow traffic volumes with the actual values calculated by CFD. The figure contains 12 subgraphs, each corresponding to one outflow traffic volume. to In each subplot: the horizontal axis represents the actual flow rate calculated by CFD; the vertical axis represents the flow rate predicted by the neural network model; blue scatter points represent sample points in the training set; red scatter points represent sample points in the validation set; yellow scatter points represent sample points in the test set; and the red diagonal line represents the ideal reference line where the predicted value equals the actual value. The distribution of scatter points in each subplot shows that the vast majority of blue points are closely distributed near the red diagonal line, indicating that the neural network model's predictions of the flow rates at each outlet have a high degree of consistency with the actual CFD values. The model has learned the nonlinear mapping relationship between the orifice parameters and the outlet flow rate, demonstrating good prediction accuracy.
[0032] Step 7: Feature Contribution Analysis (SHAP Analysis) SHAP analysis was performed on each outflow, and the results are as follows: Figure 5 As shown, Figure 5 The results of SHAP feature contribution analysis on the neural network surrogate model are presented; in the figure: the horizontal axis represents the SHAP value, indicating the contribution of a certain orifice parameter to the predicted outlet flow rate (positive values indicate that increasing the parameter will increase the outlet flow rate, and negative values indicate that increasing the parameter will decrease the outlet flow rate); the vertical axis represents the values of each orifice parameter (…). to Each subgraph corresponds to an outflow rate (); to In the color bar, red indicates a larger value for the orifice parameter, and blue indicates a smaller value.
[0033] right Figure 5 Analysis yields the following results: Influence intensity ranking: For a certain outlet flow rate, the orifice with a larger absolute value of SHAP in the longitudinal direction indicates that the orifice has a higher degree of influence on the outlet flow rate; Impact trend: A positive SHAP value indicates that increasing the orifice diameter will increase the outlet flow rate, while a negative value indicates that it will decrease it; pass Figure 5 Designers can quantitatively understand the direction and degree of influence of each throttling orifice on the flow rate of each outlet, thereby making directional adjustments in subsequent optimizations and avoiding blind trial and error.
[0034] Step 8: Rapid Prediction and Solution Optimization Based on a trained neural network, rapid prediction of flow rates at 12 outlets can be achieved, reducing the time for a single CFD calculation (approximately 20 minutes) to the instantaneous response of a surrogate model prediction (approximately 1 second). The orifice diameters of the eight flow throttling orifices can be initially defined as maximum values, and then directional adjustments can be made based on the predicted values combined with SHAP analysis until the error between the predicted results and the target values meets the requirements.
[0035] Therefore, the present invention adopts the above-mentioned optimization design method for lubricating oil circuit flow distribution for high-efficiency gas turbine equipment, which effectively solves the core problems of insufficient analytical calculation accuracy, inability to adapt to multi-branch coupling scenarios, and CFD simulation relying on manual trial calculation, long design cycle and low efficiency in the prior art. Through a complete closed loop of parameter definition, sample generation, automated simulation, surrogate model training, interpretability analysis and optimization verification, it takes into account both high accuracy of flow calculation and high efficiency of design optimization, and realizes automated and systematic optimization of multi-branch lubricating oil circuit flow.
[0036] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.
Claims
1. A method for optimizing the flow distribution of lubricating oil circuits in high-efficiency gas turbine equipment, characterized in that, Includes the following steps: S1. Determine the throttle orifice and outlet parameters of the lubricating oil circuit; Use the orifice diameter as the input variable and the outlet flow rate as the output variable; S2. Multiple sets of orifice diameter combination samples are generated using Latin hypercube sampling; S3. Perform parametric geometric modeling of the lubrication circuit so that the orifice diameter can drive the model to update automatically; S4. Construct a unified CFD flow field calculation model, number them sequentially, and set the outlet flow rate as the output parameter; S5. Based on the parametric model, automatically complete batch CFD simulation of multiple samples and construct a dataset corresponding to the orifice diameter and outlet flow rate; S6. Use the dataset as the basis to train a multi-input multi-output surrogate model and establish a nonlinear mapping relationship between the orifice diameter and the outlet flow rate. S7. Use a surrogate model to perform feature contribution analysis and quantify the degree and trend of the influence of each throttling orifice on the flow rate of each outlet. S8. Quickly predict flow rate and select the optimal throttling orifice scheme through the surrogate model, and then complete the verification closed loop through CFD simulation.
2. The method for optimizing the flow distribution of lubricating oil circuits for high-efficiency gas turbine equipment according to claim 1, characterized in that, In step S1, the parameters for determining the throttle orifice and outlet of the lubricating oil circuit are as follows: S11. Determine the number of throttling orifices in the lubrication system. The export quantity is ; S12, Definition of the The orifice diameter is the input parameter. , And determine the range of values for each aperture. ; Define the first The flow rate of each outlet is a response output parameter. , ; S13. Transform the flow allocation problem into a multi-input multi-output mapping relationship; where the input vector is... The output vector is .
3. The method for optimizing the flow distribution of lubricating oil circuits for high-efficiency gas turbine equipment according to claim 2, characterized in that, Step S2 specifically involves determining the range of values for the diameter of each throttling orifice. Inner Latin hypercube sampling (LHS) method generates A set of orifice combination schemes was developed to form a sample set. ,in, , The total number of samples, For sample index, For the first The input vector of the group of samples, For the first The orifice diameter value of the first throttling orifice in the group scheme. For the first The orifice diameter value of the second throttling orifice in the group scheme, For the first The first group of schemes The orifice diameter value of each throttling orifice.
4. The method for optimizing the flow distribution of lubricating oil circuits for high-efficiency gas turbine equipment according to claim 3, characterized in that, Step S3, specifically the parametric geometric modeling of the lubrication circuit, involves: in the parametric modeling platform, determining the orifice diameters of all throttling orifices. Defined as automatically updatable parameter variables, a lubrication circuit geometric model capable of automatically updating with the parameters is established; Import the parameter set for the group of throttling orifice combination schemes to form a batch calculation task list.
5. The method for optimizing the flow distribution of lubricating oil circuits for high-efficiency gas turbine equipment according to claim 4, characterized in that, The construction of a unified CFD flow field calculation model in step S4 specifically involves establishing a unified calculation configuration in the CFD solver, including fixing lubricating oil material parameters, turbulence model, boundary conditions and convergence criteria, to ensure that the calculation conditions of all schemes are consistent.
6. The method for optimizing the flow distribution of lubricating oil circuits for high-efficiency gas turbine equipment according to claim 5, characterized in that, Step S5 specifically involves: initiating batch solving, for The combined throttling orifice scheme sequentially performs geometry update, mesh update, CFD solution, and output extraction to obtain the flow rates at each outlet: in, For export volume, For the first The flow rate of the first outlet under the group scheme. For the first The flow rate of the second outlet under the group scheme, For the first Group Scheme No. The flow rate of each outlet; Finally, the dataset was obtained. .
7. The method for optimizing the flow distribution of lubricating oil circuits for high-efficiency gas turbine equipment according to claim 6, characterized in that, Step S6, training the multi-input multi-output agent model, specifically includes the following steps: S61. Based on the dataset obtained in step S5, combine the orifice diameters. As input to the model, the corresponding outflow flow As model output; S62. Standardize the input and output data; S63. Divide the dataset into training set, validation set and test set according to the ratio of 70%, 15% and 15% respectively; S64. Construct a multi-input multi-output neural network model with the following structure: the number of nodes in the input layer equals the number of orifices. The hidden layer uses a single hidden layer with 18 neurons, and the number of nodes in the output layer equals the number of output nodes. ; S65. Establish model mapping relationships: in, For the export flow predicted by the model, For neural network proxy models, For neural network model parameters; S66. Use mean squared error as the loss function, use the training set to train the model, monitor the training process through the validation set, and use early stopping or regularization to suppress overfitting. S67. The accuracy of the model is verified using the test set, resulting in a multi-input multi-output proxy model for fast traffic prediction.
8. The method for optimizing the flow distribution of lubricating oil circuits for high-efficiency gas turbine equipment according to claim 7, characterized in that, The feature contribution analysis in step S7 specifically involves using Shapley's additivity interpretation (SHAP) analysis to calculate the contribution of each input variable to the output predicted value. Based on this, the sensitivity of each throttling orifice to the outlet flow rate, the trend of the effect of increasing or decreasing the throttling orifice on the outlet flow rate, and the conflict between different outlets can be determined.
9. The method for optimizing the flow distribution of lubricating oil circuits for high-efficiency gas turbine equipment according to claim 7, characterized in that, Step S8 specifically involves: pre-determining the target flow rate for each outlet. ,by To constrain design, a surrogate model is used to quickly calculate the predicted output. , to make the predicted output With target traffic To minimize the error between the two, the optimal combination of throttling orifice diameters that meets the design requirements is selected. The optimal combination is then substituted into the CFD model for verification. When the error exceeds the allowable range, additional samples are added and the surrogate model is iteratively updated to form a design closed loop.