Micro-pneumatic turbine performance multi-objective optimization method
By designing experiments and using Latin hypercube sampling to screen key parameters, and combining fluid simulation and neural network models, the problems of large computational load and long cycle in the performance optimization of micro aero turbines were solved, and efficient multi-objective optimization under different gas supply pressure conditions was achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- CHINA JILIANG UNIV
- Filing Date
- 2026-04-02
- Publication Date
- 2026-06-26
AI Technical Summary
In the current process of optimizing the performance of micro aero turbines, the simulation calculation is large and the optimization cycle is long. Furthermore, the generalization ability of the model is insufficient under different air supply pressure conditions, making it difficult to simultaneously take into account performance indicators such as speed and output power.
Key structural parameters were screened using experimental design, a sample set was generated using Latin hypercube sampling, a dataset was constructed using computational fluid dynamics simulation, a radial basis function neural network model was established, and the model was transferred to different gas supply pressure conditions through transfer learning. Multi-objective optimization was performed using a non-dominated sorting genetic algorithm.
It significantly reduces the requirements for simulation samples and computational costs, improves the efficiency of performance prediction and optimization of micro aerodynamic turbines, and achieves multi-objective optimization under different gas supply pressure conditions.
Smart Images
Figure CN122287362A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of micro pneumatic turbine performance optimization technology, specifically to a multi-objective optimization method for the performance of micro pneumatic turbines under different air supply pressure conditions based on fluid simulation calculation, radial basis neural network, non-dominated sorting genetic algorithm and transfer learning. Background Technology
[0002] Miniature machine tools are widely used in aerospace, biomedicine, precision manufacturing, and other fields. As a core component, the miniature spindle's rotational speed, output power, and operational stability directly affect the efficiency and quality of micromachining. Miniature pneumatic turbine spindles, due to their advantages of high speed, high precision, and compact structure, are considered an important drive mechanism for miniature machine tools. Current research on the structural design, performance prediction, and structural parameter optimization of miniature pneumatic turbines mainly employs theoretical calculations and computational fluid dynamics (CFD) simulations. Theoretical calculations struggle to accurately reflect the impact of turbine geometry and subtle structural changes on performance, and also fail to reveal the intrinsic relationship between microscopic flow fields and macroscopic performance. While CFD simulations can analyze the internal flow field characteristics of turbines relatively accurately, their fluid modeling, mesh generation, and numerical solution processes are computationally intensive, time-consuming, and have low optimization efficiency.
[0003] Currently, using neural network models to replace CFD simulations for performance prediction can reduce computational costs and improve optimization efficiency to some extent. However, existing methods still have the following shortcomings: First, model training usually relies on a large number of simulation samples, resulting in high dataset construction costs and long cycles; second, the applicability and generalization ability of the models under different gas supply pressure conditions are limited; and third, the methods for multi-objective optimization of the structural parameters and performance indicators such as speed and output power of micro aerodynamic turbines are still not perfect. Therefore, a method is needed that can reduce the need for simulation samples while ensuring prediction accuracy, be applicable to different gas supply pressure conditions, and achieve multi-objective performance optimization of micro aerodynamic turbines, so as to improve the efficiency and engineering application value of micro aerodynamic turbine design and optimization. Summary of the Invention
[0004] The purpose of this invention is to provide a multi-objective optimization method for the performance of micro aerodynamic turbines, addressing the problems of large computational load, long time consumption, and low optimization efficiency in existing micro aerodynamic turbine performance prediction and optimization methods.
[0005] To achieve the objectives of this invention, the following technical solution is proposed, comprising the following steps:
[0006] Step 1: Determine the range of structural parameters for the micro aerodynamic turbine;
[0007] Step 2: Calculate the performance of micro aerodynamic turbines with different structural parameters through fluid simulation;
[0008] Step 3: Use Design of Experiments (DOE) to determine the key structural parameters that affect the performance of the micro aero turbine;
[0009] Step 4: Generate samples of key structural parameters for micro aerodynamic turbines using Latin hypercube sampling (LHS);
[0010] Step 5: Under a specific air supply pressure, calculate the performance of micro aerodynamic turbine samples with different key structural parameters through fluid simulation, and construct a dataset of key structural parameters of different micro aerodynamic turbines and their corresponding micro aerodynamic turbine performance.
[0011] Step 6: Train a radial basis function neural network (RBFNN) using the dataset to obtain a predictive model of the performance of a micro aerodynamic turbine under a specific supply pressure;
[0012] Step 7: Determine the performance optimization target of the micro aero turbine and use the non-dominated sorting genetic algorithm (NSGA II) to obtain the optimal key structural parameters of the micro aero turbine;
[0013] Step 8: Construct a new dataset of micro aerodynamic turbine performance under different key structural parameters at other air supply pressures through fluid simulation calculations;
[0014] Step 9: Obtain the RBFNN model for predicting the performance of micro aero turbines under other air supply pressures through transfer learning (TL), and use NSGA II to obtain the optimal key structural parameters of micro aero turbines under other air supply pressures.
[0015] In step 1, the range of structural parameters of the micro aerodynamic turbine is determined according to engineering needs, including: the range of values for nozzle diameter, nozzle length, nozzle incident angle, blade tip clearance, axial clearance, number of blades, blade thickness, hub diameter, and turbine diameter.
[0016] In step 2, fluid simulation is used to calculate the performance of micro aerodynamic turbines with different structural parameters, including: turbine speed, output power, output torque, efficiency and gas flow rate.
[0017] In step 6, the input of the RBFNN model is the key structural parameters of the micro aerodynamic turbine under a specific air supply pressure, and the output of the RBFNN model is the performance of the micro aerodynamic turbine.
[0018] In step 7, the performance optimization target of the micro aero turbine is determined according to the engineering needs. The NSGA II is used to perform multi-objective iterative search on the trained RBFNN prediction model to obtain the optimal combination of structural parameters of the micro aero turbine.
[0019] In step 8, a new dataset, no less than one-fifth the size of the dataset under the specific gas supply pressure, is constructed through fluid simulation calculations under other gas supply pressures.
[0020] In step 9, the RBFNN trained under a specific air supply pressure is used as the original model, and its hidden layer parameters are transferred to the target model under other air supply pressures. The target model is trained using the new dataset, the target model parameters are updated, and the performance of the micro aerodynamic turbine under other air supply pressures is optimized using NSGA II based on the transferred RBFNN model.
[0021] This invention proposes a multi-objective optimization method for the performance of micro aero turbines. Addressing the problems of high computational cost, low optimization efficiency, and poor applicability to various operating conditions in existing optimization methods, this method uses Design of Elements (DOE) to screen key structural parameters, employs Lowest Harmony Scheme (LHS) to generate samples and combines them with simulation calculations to construct a dataset, utilizes Restricted Baseline Variable Array (RBFNN) to establish a micro aero turbine performance prediction model, and integrates NSGAII to achieve multi-objective optimization of structural parameters. Furthermore, a Time Transfer Model (TL) is introduced to transfer the established prediction model to different supply pressure conditions, thereby reducing the dataset requirements and computational costs under new operating conditions. This method improves the efficiency of micro aero turbine performance prediction and structural optimization, achieving multi-objective optimization of micro aero turbine performance under different operating conditions. Attached Figure Description
[0022] Figure 1 This is a flowchart of the present invention;
[0023] Figure 2 The flowchart for RBFNN;
[0024] Figure 3 NSGA II flowchart;
[0025] Figure 4 TL flowchart; Detailed Implementation
[0026] The present invention will be further described below with reference to the embodiments. It should be noted that the following embodiments are only used to further illustrate the present invention and should not be construed as limiting the scope of protection of the present invention. Some non-essential improvements and adjustments made by those skilled in the art based on the above-described invention are still within the scope of protection of the present invention.
[0027] This invention proposes a multi-objective optimization method for the performance of micro aerodynamic turbines, aiming to address the problems existing in current micro aerodynamic turbine performance optimization processes, such as large simulation computational load, long optimization cycle, insufficient model generalization ability under different supply pressure conditions, and difficulty in simultaneously considering rotational speed and output power. This method selects key structural parameters through experimental design, constructs a sample set using Latin hypercube sampling, obtains performance data through computational fluid dynamics simulation, establishes a mapping model between micro aerodynamic turbine structural parameters and performance indicators, and achieves multi-objective optimization under different supply pressure conditions based on this model.
[0028] In this embodiment, the value range of each structural parameter is first determined based on engineering experience and processing conditions. An experimental design method is used to analyze the main parameters affecting the performance of the micro aero turbine. Further, Latin hypercube sampling is used to generate several sets of samples within each parameter value range to construct a micro aero turbine performance dataset. A radial basis function neural network is trained using this dataset to obtain a micro aero turbine performance prediction model under a specific supply pressure. A non-dominated sorting genetic algorithm is then used to obtain the optimal micro aero turbine structural parameters. Finally, transfer learning is used to obtain an RBFNN model for predicting micro aero turbine performance under different supply pressures. NSGA II is used to optimize the micro aero turbine structural parameters under different supply pressures. (Refer to...) Figure 1 A multi-objective optimization method for the performance of a micro aerodynamic turbine includes the following steps:
[0029] Step 1: Determine the range of structural parameters for the micro aerodynamic turbine;
[0030] In step 1, the range of structural parameters of the micro pneumatic turbine is determined according to engineering needs, including: the range of values for nozzle diameter, nozzle length, nozzle incident angle, blade tip clearance, axial clearance, number of blades, blade thickness, hub diameter, and turbine diameter.
[0031] Step 2: Calculate the performance of micro aerodynamic turbines with different structural parameters through fluid simulation;
[0032] In step 2, fluid simulation is used to calculate the performance of micro aerodynamic turbines with different structural parameters, including turbine speed, output power, output torque, efficiency and gas flow rate.
[0033] Step 3: Use Design of Experiments (DOE) to determine the key structural parameters that affect the performance of the micro aero turbine;
[0034] Step 4: Generate samples of key structural parameters for micro aerodynamic turbines using Latin hypercube sampling (LHS);
[0035] Step 5: Under a specific air supply pressure, calculate the performance of micro aerodynamic turbine samples with different key structural parameters through fluid simulation, and construct a dataset of micro aerodynamic turbines with different key structural parameters and their corresponding performance.
[0036] Step 6: Train a radial basis function neural network (RBFNN) using the dataset to obtain a predictive model of the performance of a micro aerodynamic turbine under a specific supply pressure;
[0037] In step 6, the input to the RBFNN model is the key structural parameters of the micro aero turbine under a specific air supply pressure, and the output of the RBFNN model is the performance of the micro aero turbine.
[0038] Step 7: Determine the performance optimization target of the micro aero turbine and use the non-dominated sorting genetic algorithm (NSGA II) to obtain the optimal key structural parameters of the micro aero turbine;
[0039] In step 7, the performance optimization target of the micro aero turbine is determined according to the engineering needs. The NSGA II is used to perform multi-objective iterative search on the trained RBFNN prediction model to obtain the optimal combination of structural parameters of the micro aero turbine.
[0040] Step 8: Construct a new dataset of micro aerodynamic turbine performance under different key structural parameters at other air supply pressures through fluid simulation calculations;
[0041] In step 8, a new dataset, no less than one-fifth the size of the dataset under the specific gas supply pressure, is constructed through fluid simulation calculations under other gas supply pressures.
[0042] Step 9: Obtain the RBFNN model for predicting the performance of micro aero turbines under other air supply pressures through transfer learning (TL), and use NSGA II to obtain the optimal key structural parameters of micro aero turbines under other air supply pressures;
[0043] In step 9, the RBFNN trained under a specific gas supply pressure is used as the original model, and its hidden layer parameters are transferred to the target model under other gas supply pressures. The target model is trained using the new dataset, the parameters of the target model are updated, and the performance of the micro aerodynamic turbine under other gas supply pressures is optimized using NSGA II based on the transferred RBFNN model.
[0044] Through the above steps, this invention utilizes RBFNN to replace a large amount of direct CFD simulation for performance prediction, leverages NSGAII to achieve coordinated optimization of the speed and output power of micro aerodynamic turbines, and uses TL to achieve rapid prediction and optimization of a small dataset under different supply pressure conditions. This method not only significantly reduces the cost of sample construction and optimization time, but also improves the efficiency of performance optimization of micro aerodynamic turbines under multiple operating conditions, demonstrating good engineering application value.
Claims
1. A multi-objective optimization method for the performance of a micro aerodynamic turbine, characterized in that, Includes the following steps: Step 1: Determine the range of structural parameters for the micro aerodynamic turbine; Step 2: Calculate the performance of micro aerodynamic turbines with different structural parameters through fluid simulation; Step 3: Use Design of Experiments (DOE) to determine the key structural parameters that affect the performance of the micro aero turbine; Step 4: Generate samples of key structural parameters for micro aerodynamic turbines using Latin hypercube sampling (LHS); Step 5: Under a specific air supply pressure, calculate the performance of micro aerodynamic turbine samples with different key structural parameters through fluid simulation, and construct a dataset of micro aerodynamic turbines with different key structural parameters and their corresponding performance. Step 6: Train a radial basis function neural network (RBFNN) using the dataset to obtain a predictive model of the performance of a micro aerodynamic turbine under a specific supply pressure; Step 7: Determine the performance optimization target of the micro aero turbine and use the non-dominated sorting genetic algorithm (NSGA II) to obtain the optimal key structural parameters of the micro aero turbine; Step 8: Construct a new dataset of micro aerodynamic turbine performance under different key structural parameters at other air supply pressures through fluid simulation calculations; Step 9: Obtain the RBFNN model for predicting the performance of micro aero turbines under other air supply pressures through transfer learning (TL), and use NSGA II to obtain the optimal key structural parameters of micro aero turbines under other air supply pressures.
2. The multi-objective optimization method for the performance of a micro aerodynamic turbine according to claim 1, characterized in that: In step 1, the range of structural parameters of the micro pneumatic turbine is determined according to engineering needs, including: the range of values for nozzle diameter, nozzle length, nozzle incident angle, blade tip clearance, axial clearance, number of blades, blade thickness, hub diameter, and turbine diameter.
3. The multi-objective optimization method for the performance of a micro aerodynamic turbine according to claim 1, characterized in that: In step 2, fluid simulation is used to calculate the performance of micro aerodynamic turbines with different structural parameters, including turbine speed, output power, output torque, efficiency and gas flow rate.
4. The multi-objective optimization method for the performance of a micro aerodynamic turbine according to claim 1, characterized in that: In step 6, the input to the RBFNN model is the key structural parameters of the micro aero turbine under a specific air supply pressure, and the output of the RBFNN model is the performance of the micro aero turbine.
5. The multi-objective optimization method for the performance of a micro aerodynamic turbine according to claim 1, characterized in that: In step 7, the performance optimization target of the micro aero turbine is determined according to the engineering needs. The NSGA II is used to perform multi-objective iterative search on the trained RBFNN prediction model to obtain the optimal combination of structural parameters of the micro aero turbine.
6. The multi-objective optimization method for the performance of a micro aerodynamic turbine according to claim 1, characterized in that: In step 8, a new dataset, no less than one-fifth the size of the dataset under the specific gas supply pressure, is constructed through fluid simulation calculations under other gas supply pressures.
7. The multi-objective optimization method for the performance of a micro aerodynamic turbine according to claim 1, characterized in that: In step 9, the RBFNN trained under a specific gas supply pressure is used as the original model, and its hidden layer parameters are transferred to the target model under other gas supply pressures. The target model is trained using the new dataset, the parameters of the target model are updated, and the performance of the micro aerodynamic turbine under other gas supply pressures is optimized using NSGAII based on the transferred RBFNN model.