CN113051835A - Efficient prediction method for unsteady flow field of impeller machinery based on machine learning - Google Patents
Efficient prediction method for unsteady flow field of impeller machinery based on machine learning Download PDFInfo
- Publication number
- CN113051835A CN113051835A CN202110402027.3A CN202110402027A CN113051835A CN 113051835 A CN113051835 A CN 113051835A CN 202110402027 A CN202110402027 A CN 202110402027A CN 113051835 A CN113051835 A CN 113051835A
- Authority
- CN
- China
- Prior art keywords
- flow field
- impeller
- sample point
- rotating speed
- working condition
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/28—Design optimisation, verification or simulation using fluid dynamics, e.g. using Navier-Stokes equations or computational fluid dynamics [CFD]
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Geometry (AREA)
- General Engineering & Computer Science (AREA)
- Computer Hardware Design (AREA)
- Software Systems (AREA)
- Medical Informatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- Algebra (AREA)
- Computing Systems (AREA)
- Fluid Mechanics (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Mathematical Physics (AREA)
- Pure & Applied Mathematics (AREA)
- Structures Of Non-Positive Displacement Pumps (AREA)
Abstract
The invention relates to a machine learning-based efficient prediction method for a variable working condition flow field of an impeller machine, which comprises the following steps: the system comprises a sample point database, a sample point data integration module, an artificial intelligent unsteady flow field agent model, an unsteady working condition solving module and a result post-processing and visualization module. The method has the advantages of short solving time, rich flow field details, strong learning ability and strong working condition adaptability.
Description
Technical Field
The invention belongs to the technical field of machinery, and particularly relates to a machine learning-based efficient prediction method for a variable working condition flow field of an impeller machine.
Background
In recent years, with the wide application of impeller machinery in various energy systems of new energy, compressed air energy storage, waste heat utilization and the like, the operating conditions of the impeller machinery have the characteristics of multiple adjustment process types, wide working condition range and frequent adjustment, and the impeller machinery structurally has the development trend of multiple blade rows and multiple parallel blades. Under various working conditions, the accurate and rapid solution of the internal flow field of the impeller machine is the basis and key for realizing the real-time simulation, the optimization of pneumatic performance and the high-efficiency wide-working-condition adjustment of the equipment, plays a vital role in the high-efficiency and safe operation of the impeller machine, and is also a difficult problem to be solved at present.
At present, the method for solving the mechanical performance and the flow field of the impeller mainly comprises a one-dimensional aerodynamic performance prediction model, a two-dimensional flow surface model, a full three-dimensional computational fluid dynamics (3D-CFD) model and the like. The one-dimensional aerodynamic performance prediction model can quickly determine the corresponding relation between the operation working condition and the performance index according to the turbomachine external curve, can quickly determine the turbomachine external characteristic, has the advantage of short solving time, but cannot provide flow field details, so that the flow field evolution rule in the dynamic variable working condition process of the turbomachine is difficult to reveal; the two-dimensional flow surface model not only can provide an external characteristic line of the impeller machine, but also can provide partial flow field distribution parameters and flow field details, more reference information can be provided for adjustment and control of the impeller machine, but the solving time is increased, and most flow surface models cannot consider the viscosity influence of the working medium, so the solving precision is low; the current mainstream solution method is a full three-dimensional CFD model: the model can provide abundant flow field detail information by solving the three-dimensional viscous Navier-Stokes equation, has high solving precision and long solving time, and is generally more than several orders of magnitude of the solving time of the two methods.
Disclosure of Invention
The invention aims to provide a machine learning-based efficient prediction method for the variable working condition flow field of the impeller machinery, which is short in solving time, rich in flow field details, strong in learning capacity and strong in working condition adaptability, aiming at the defects.
In order to achieve the purpose, the invention adopts the following technical scheme.
The invention relates to a machine learning-based efficient prediction method for a variable working condition flow field of an impeller machine, which comprises the following steps of:
(1) sample point database: inputting the operating condition parameters of the turbine, namely the expansion ratio pi tt and the reduced rotating speed ncor, into a sample point; the output sample points are a set formed by static entropy S values at different positions (flow field position points 1 ', 2', 3 'and 120') at the impeller outlet of the turbine, and a sample point database is obtained by solving a Navier-stokes equation in three-dimensional flow field software Numeca by giving an expansion ratio and a reduced rotating speed as known conditions;
(2) a sample point data integration module: carrying out array combination and row-column transposition on the input sample points and the output sample points to form a database which can enable the artificial intelligent flow field agent model to identify;
(3) artificial intelligence neural network flow field agent model: the artificial intelligent flow field agent model completes the learning of flow field data by adopting a neural network algorithm and completes the prediction of the flow field under the operation condition except a sample point database, wherein the artificial intelligent flow field agent model is provided with 1 hidden layer, the number of neurons in each layer is 200, and an activation function is a Sigmoid function;
the unsteady state working condition solving module comprises: a matlab simulink is adopted to form a turbine unsteady state working condition solving module, the turbine outlet flow loss static entropy distribution in the variable expansion ratio process and the variable reduced rotating speed process is solved, the change data points of the expansion ratio and the reduced rotating speed along with time are obtained, and the data points are taken as input variables and are brought into an artificial intelligent flow field agent model to be solved;
(4) a result post-processing and visualization module: firstly, interpolating and fitting static entropy distribution solving values corresponding to expansion ratios at different moments to obtain a rule that the distribution of the outlet entropy of the impeller changes along with the expansion ratios; and secondly, interpolating and fitting the corresponding solution values of the static entropy distribution at different moments under the reduced rotating speed to obtain the change rule of the impeller outlet entropy distribution along with the reduced rotating speed.
Compared with the prior art, the invention has obvious beneficial effects, and the technical scheme can show that: the invention can output abundant and various flow field parameters according to the working condition requirement and expand the sample point library according to the result by combining the artificial intelligence algorithm (independent of explicit complicated nonlinear equation solving, having self-learning function and associative storage function, etc.) through the existing flow field parameter database of the turbine under different working conditions, and not under the condition of complicated three-dimensional flow field solving process after model training, thereby having the advantages of short solving time, abundant flow field details and strong learning capability, and being particularly suitable for the performance and flow field analysis of multi-row impeller and multi-stage combined large-scale impeller machinery under dynamic working conditions. Even if more sample data points are needed in the model training stage, the training can be separated from the limitation of the working condition range of the sample points after the training is finished, and then the rapid solution of the internal flow field of the impeller machine can be finished according to the actual working condition parameters. With the increase of the use times of the model, the number of sample points is increased, so that the prediction accuracy of the model is further improved. The method has the advantages of high solving precision and short solving time for solving the internal flow field of the impeller machine under the unstable working condition; the invention can be combined with the current mature impeller mechanical simulation platform, variable working condition adjusting program and impeller mechanical design program, and has the characteristic of wide application range.
Drawings
FIG. 1 is a schematic of the present invention;
FIG. 2 is a block diagram of a solution for non-steady-state operating conditions of a turbine according to the present invention;
FIG. 3 is a sample point database diagram of the present invention;
FIG. 4 is a flow field agent model of an artificial intelligence neural network of the present invention;
FIG. 5 shows the variation law of the static entropy distribution of the impeller outlet predicted by the method of the present invention with the expansion ratio;
FIG. 6 shows the variation rule of impeller outlet static entropy distribution predicted by the method of the present invention with reduced rotation speed;
FIG. 7 shows the result of comparing the prediction accuracy of the static entropy distribution at the impeller outlet by the method of the present invention.
Detailed Description
The following detailed description will be given to specific embodiments, structures, features and effects of a method for efficiently predicting a variable-condition flow field of an impeller machine based on machine learning, which is provided by the present invention, with reference to the accompanying drawings and preferred embodiments.
As shown in fig. 1, a method for efficiently predicting a variable-condition flow field of an impeller machine based on machine learning includes the following steps:
(1) sample point database: as shown in fig. 2. The input sample point is the turbine operating condition parameter expansion ratioπ tt Reduced rotational speedn cor (ii) a The output sample points are a set formed by static entropy S values at different positions (flow field position points 1 ', 2', 3 'and 120') at the impeller outlet of the turbine, and a sample point database is obtained by solving a Navier-stokes equation in three-dimensional flow field software Numeca by giving an expansion ratio and a reduced rotating speed as known conditions;
(2) a sample point data integration module: carrying out array combination and row-column transposition on the input sample points and the output sample points to form a database which can enable the artificial intelligent flow field agent model to identify;
(3) artificial intelligence neural network flow field agent model: the artificial intelligence flow field agent model completes the learning of flow field data by adopting a neural network algorithm and further completes the prediction of the flow field under the operation condition except the sample point database, as shown in fig. 3, the artificial intelligence flow field agent model has 1 hidden layer, the number of neurons in each layer is 200, and the activation function is a Sigmoid function;
the unsteady state working condition solving module comprises: a matlab simulink is adopted to form a turbine unsteady state working condition solving module, as shown in fig. 4, the turbine outlet flow loss static entropy distribution in the variable expansion ratio process and the variable reduced rotating speed process is solved, the change data points of the expansion ratio and the reduced rotating speed along with time are obtained, and the data points are taken as input variables and are introduced into an artificial intelligent flow field agent model to be solved;
(4) a result post-processing and visualization module: as shown in fig. 5, interpolation and fitting are performed on the solution values of the static entropy distribution corresponding to the expansion ratio at different moments to obtain the change rule of the impeller outlet entropy distribution along with the expansion ratio; as shown in fig. 6, interpolation and fitting are performed on the corresponding solution values of the static entropy distribution at different times under the reduced rotation speed, so that the change rule of the impeller outlet entropy distribution along with the reduced rotation speed is obtained.
The results obtained by the invention are verified and compared with a numerical simulation method. As shown in fig. 7, the prediction result of the mach number distribution of the exit of the turbomachine is substantially consistent with the solution result of the full three-dimensional CFD model, which indicates that the method can obtain a solution result with higher accuracy in a shorter time.
The invention is suitable for various impeller machines applied to different fields.
The above description is only a preferred embodiment of the present invention, and is not intended to limit the present invention in any way, and any simple modification, equivalent change and modification made to the above embodiment according to the technical spirit of the present invention are within the scope of the present invention without departing from the technical spirit of the present invention.
Claims (2)
1. A method for efficiently predicting a variable working condition flow field of an impeller machine based on machine learning comprises the following steps:
(1) sample point database: inputting the operating condition parameters of the turbine, namely the expansion ratio pi tt and the reduced rotating speed ncor, into a sample point; the output sample points are a set formed by static entropy S values at different positions (flow field position points 1 ', 2', 3 'and 120') at the impeller outlet of the turbine, and a sample point database is obtained by solving a Navier-stokes equation in three-dimensional flow field software Numeca by giving an expansion ratio and a reduced rotating speed as known conditions;
(2) a sample point data integration module: carrying out array combination and row-column transposition on the input sample points and the output sample points to form a database which can enable the artificial intelligent flow field agent model to identify;
(3) artificial intelligence neural network flow field agent model: the artificial intelligent flow field agent model completes the learning of flow field data by adopting a neural network algorithm and completes the prediction of the flow field under the operation condition except a sample point database, wherein the artificial intelligent flow field agent model is provided with 1 hidden layer, the number of neurons in each layer is 200, and an activation function is a Sigmoid function;
the unsteady state working condition solving module comprises: a matlab simulink is adopted to form a turbine unsteady state working condition solving module, the turbine outlet flow loss static entropy distribution in the variable expansion ratio process and the variable reduced rotating speed process is solved, the change data points of the expansion ratio and the reduced rotating speed along with time are obtained, and the data points are taken as input variables and are brought into an artificial intelligent flow field agent model to be solved;
(4) a result post-processing and visualization module: firstly, interpolating and fitting static entropy distribution solving values corresponding to expansion ratios at different moments to obtain a rule that the distribution of the outlet entropy of the impeller changes along with the expansion ratios; and secondly, interpolating and fitting the corresponding solution values of the static entropy distribution at different moments under the reduced rotating speed to obtain the change rule of the impeller outlet entropy distribution along with the reduced rotating speed.
2. The efficient prediction method for the variable-condition flow field of the turbomachinery based on the machine learning of claim 1, wherein: the different positions at the impeller outlet of the turbine are flow field position points 1 ', 2', 3 'and 120'.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110402027.3A CN113051835A (en) | 2021-04-14 | 2021-04-14 | Efficient prediction method for unsteady flow field of impeller machinery based on machine learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110402027.3A CN113051835A (en) | 2021-04-14 | 2021-04-14 | Efficient prediction method for unsteady flow field of impeller machinery based on machine learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113051835A true CN113051835A (en) | 2021-06-29 |
Family
ID=76519668
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110402027.3A Pending CN113051835A (en) | 2021-04-14 | 2021-04-14 | Efficient prediction method for unsteady flow field of impeller machinery based on machine learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113051835A (en) |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20120303564A1 (en) * | 2011-05-26 | 2012-11-29 | World Heart Corporation | Systems and methods for predicting characteristics of an artificial heart using an artificial neural network |
| CN106874569A (en) * | 2017-01-19 | 2017-06-20 | 中国科学院工程热物理研究所 | One kind becomes geometry split gas turbine real-time simulation modeling method |
| CN111241752A (en) * | 2020-01-16 | 2020-06-05 | 北京航空航天大学 | Centrifugal impeller comprehensive optimization method based on digital twinning and reinforcement learning |
| CN111814272A (en) * | 2020-07-07 | 2020-10-23 | 中国科学院工程热物理研究所 | Turbine pneumatic-dynamic response intelligent optimization design method based on machine learning |
| CN111859746A (en) * | 2020-07-10 | 2020-10-30 | 西安交通大学 | Method for predicting variable working condition performance of turbomachinery based on flow field reconstruction |
| CN112084727A (en) * | 2020-10-26 | 2020-12-15 | 中国人民解放军国防科技大学 | Transition prediction method based on neural network |
-
2021
- 2021-04-14 CN CN202110402027.3A patent/CN113051835A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20120303564A1 (en) * | 2011-05-26 | 2012-11-29 | World Heart Corporation | Systems and methods for predicting characteristics of an artificial heart using an artificial neural network |
| CN106874569A (en) * | 2017-01-19 | 2017-06-20 | 中国科学院工程热物理研究所 | One kind becomes geometry split gas turbine real-time simulation modeling method |
| CN111241752A (en) * | 2020-01-16 | 2020-06-05 | 北京航空航天大学 | Centrifugal impeller comprehensive optimization method based on digital twinning and reinforcement learning |
| CN111814272A (en) * | 2020-07-07 | 2020-10-23 | 中国科学院工程热物理研究所 | Turbine pneumatic-dynamic response intelligent optimization design method based on machine learning |
| CN111859746A (en) * | 2020-07-10 | 2020-10-30 | 西安交通大学 | Method for predicting variable working condition performance of turbomachinery based on flow field reconstruction |
| CN112084727A (en) * | 2020-10-26 | 2020-12-15 | 中国人民解放军国防科技大学 | Transition prediction method based on neural network |
Non-Patent Citations (2)
| Title |
|---|
| 王星: ""NACA 翼型叶顶对低展弦比向心涡轮影响特性"", 《机 械 工 程 学 报》 * |
| 高杰等: "燃气轮机涡轮叶顶间隙气热技术研究进展", 《航空学报》 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106874569B (en) | Variable-geometry split-shaft type gas turbine real-time simulation modeling method | |
| CN106227967A (en) | Industrial steam turbine low-pressure stage group vane type line optimization method | |
| CN111859746B (en) | Method for predicting variable working condition performance of turbomachinery based on flow field reconstruction | |
| CN111898212B (en) | Impeller mechanical profile design optimization method based on BezierGAN and Bayesian optimization | |
| CN110083896B (en) | Standard reaching conformity degree evaluation method for wind turbine generator operating power curve | |
| CN111190349A (en) | Method, system and medium for monitoring state and diagnosing fault of ship engine room equipment | |
| CN114254460A (en) | Turbo-machine pneumatic robustness optimization method based on graph convolution neural network | |
| CN108198268A (en) | A kind of production equipment data scaling method | |
| Kadhim et al. | Design optimization workflow and performance analysis for contoured endwalls of axial turbines | |
| CN103149840A (en) | Semanteme service combination method based on dynamic planning | |
| Wu | Dynamic performance simulation analysis method of split shaft gas turbine based on RBF neural network | |
| CN111079920B (en) | Prediction method for flow non-uniformity coefficient of turbine gas collection cavity outlet | |
| Safari et al. | A novel combination of adaptive tools for turbomachinery airfoil shape optimization using a real-coded genetic algorithm | |
| CN113051835A (en) | Efficient prediction method for unsteady flow field of impeller machinery based on machine learning | |
| CN117473890A (en) | Micro turbine flow field prediction method and device based on mechanical learning | |
| CN112464478A (en) | Control law optimization method and device for water turbine speed regulating system | |
| CN113379103B (en) | Prediction method of pump equipment internal flow field based on reduced order model | |
| CN112632728B (en) | Turbine mechanical blade profile design and performance prediction method based on deep learning | |
| CN112052621B (en) | Supercritical carbon dioxide impeller machinery supercritical region prediction and control method based on CNN | |
| CN115062560B (en) | Turbine multi-scale flow reduced-order coupling method based on machine learning | |
| CN107194509B (en) | Wind power prediction method based on time interval fuzzy operator and approximate weight integration | |
| Yu | Extrapolation for aeroengine gas path faults with SVM bases on genetic algorithm | |
| CN117454721B (en) | Wind power plant wake superposition effect evaluation method and medium based on digital simulation experiment | |
| CN115062560A (en) | Turbine multi-scale flow reduced-order coupling method based on machine learning | |
| Cerri et al. | Gas path analysis and gas turbine re-mapping |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210629 |
|
| RJ01 | Rejection of invention patent application after publication |