CN116306206A - Airfoil transonic buffeting flow field rapid prediction method based on deep neural network - Google Patents
Airfoil transonic buffeting flow field rapid prediction method based on deep neural network Download PDFInfo
- Publication number
- CN116306206A CN116306206A CN202211381140.9A CN202211381140A CN116306206A CN 116306206 A CN116306206 A CN 116306206A CN 202211381140 A CN202211381140 A CN 202211381140A CN 116306206 A CN116306206 A CN 116306206A
- Authority
- CN
- China
- Prior art keywords
- flow field
- neural network
- airfoil
- network model
- unsteady
- 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
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 30
- 238000000034 method Methods 0.000 title claims abstract description 28
- 238000003062 neural network model Methods 0.000 claims abstract description 29
- 238000012549 training Methods 0.000 claims abstract description 23
- 238000004364 calculation method Methods 0.000 claims abstract description 14
- 239000011159 matrix material Substances 0.000 claims description 13
- 230000006870 function Effects 0.000 claims description 10
- 238000005457 optimization Methods 0.000 claims description 9
- 238000004088 simulation Methods 0.000 claims description 8
- 238000012360 testing method Methods 0.000 claims description 8
- 238000010606 normalization Methods 0.000 claims description 6
- 238000000926 separation method Methods 0.000 claims description 6
- 238000004422 calculation algorithm Methods 0.000 claims description 5
- 238000012545 processing Methods 0.000 claims description 4
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 claims description 3
- 230000003044 adaptive effect Effects 0.000 claims description 3
- 230000004927 fusion Effects 0.000 claims description 3
- 238000013507 mapping Methods 0.000 claims description 3
- 230000001172 regenerating effect Effects 0.000 claims description 3
- 230000001131 transforming effect Effects 0.000 claims description 3
- 230000009466 transformation Effects 0.000 claims description 2
- 239000010410 layer Substances 0.000 description 19
- 238000013136 deep learning model Methods 0.000 description 7
- 238000013461 design Methods 0.000 description 6
- 238000000605 extraction Methods 0.000 description 4
- 239000012530 fluid Substances 0.000 description 4
- 230000035939 shock Effects 0.000 description 4
- 238000000354 decomposition reaction Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 238000013527 convolutional neural network Methods 0.000 description 2
- 125000004122 cyclic group Chemical group 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 230000003993 interaction Effects 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 239000000470 constituent Substances 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000008034 disappearance Effects 0.000 description 1
- 230000010365 information processing Effects 0.000 description 1
- 230000007787 long-term memory Effects 0.000 description 1
- 238000011089 mechanical engineering Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000005312 nonlinear dynamic Methods 0.000 description 1
- 230000000737 periodic effect Effects 0.000 description 1
- 238000012706 support-vector machine Methods 0.000 description 1
- 239000002344 surface layer Substances 0.000 description 1
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]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2113/00—Details relating to the application field
- G06F2113/08—Fluids
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
- G06F2119/14—Force analysis or force optimisation, e.g. static or dynamic forces
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T90/00—Enabling technologies or technologies with a potential or indirect contribution to GHG emissions mitigation
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- Computer Hardware Design (AREA)
- General Health & Medical Sciences (AREA)
- Data Mining & Analysis (AREA)
- Geometry (AREA)
- Biophysics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Pure & Applied Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Medical Informatics (AREA)
- Mathematical Optimization (AREA)
- Mathematical Analysis (AREA)
- Fluid Mechanics (AREA)
- Algebra (AREA)
- Aerodynamic Tests, Hydrodynamic Tests, Wind Tunnels, And Water Tanks (AREA)
Abstract
The invention provides a rapid prediction method of an airfoil transonic buffeting flow field based on a deep neural network, which comprises the steps of generating an unsteady flow field sample set; constructing a neural network model for rapid prediction of an unsteady flow field; training the constructed neural network model; the trained deep neural network is used for fast prediction of an unsteady flow field. According to the invention, the flow field information based on the calculation grid is used as the input of the neural network model, and the flow field information at the future moment is predicted, so that compared with the input based on the picture or the uniform grid, the flow field information has higher resolution under the standard of the same data quantity, and the model has higher prediction precision. According to the invention, the flow field parameters are interpolated onto the simplified interpolation grid according to the concerned flow field region, so that the flow field parameters are used for predicting the unsteady flow field, the number of data points and the time consumption can be reduced as much as possible while the prediction precision of the unsteady flow field is ensured, and the efficiency is improved.
Description
Technical Field
The invention relates to the field of deep learning and computational fluid mechanics, in particular to a rapid prediction method for an airfoil transonic buffeting flow field based on a deep neural network.
Background
Under transonic flow conditions, shock/boundary layer separation interactions may cause unstable motion of the entire flow field, a phenomenon known as transonic buffeting. In airfoil design, transonic buffeting is a very critical issue because buffeting occurs to limit the upper performance limit of the airfoil. The buffeting phenomenon starts from trailing edge flow separation caused by interaction of shock wave boundary layers, and the flow field is characterized by the changes of shock wave positions and intensities, the occurrence and disappearance of trailing edge shedding vortex and the post flow separation. When an aircraft flies at a transonic speed, if the aircraft enters a buffeting starting boundary, a pulsating pressure load caused by transonic buffeting not only can influence the cruising performance and the flight quality of the aircraft, but also can possibly cause aircraft structural fatigue to cause flight accidents. Therefore, it is necessary to develop an airfoil optimization design method considering the buffeting problem, so as to avoid or delay the buffeting phenomenon, thereby improving the comprehensive performance of the airfoil.
The main current airfoil design means is airfoil optimization design based on computational fluid dynamics (Computational Fluid Dynamics, CFD), the main idea is to continuously adjust an initial airfoil through CFD calculation and optimization algorithm, and finally obtain the airfoil with optimal target performance. However, the transonic buffeting flow field contains a flow structure with strong nonlinearity such as a separation vortex and a shock wave, if the flow structure is directly subjected to numerical simulation in the optimization process, a great amount of calculation time and cost are consumed, and meanwhile, the current intelligent optimization algorithm needs to call a great amount of CFD (computational fluid dynamics) process, so that the optimization efficiency is reduced.
Various aerodynamic reduced order models (Reduced Order Models, ROM) based on CFD simulation data have therefore been proposed by researchers. ROM means that linear and nonlinear models are used to express the mapping relation between input and output in a complex aerodynamic system, and the methods are widely applied to aerospace and mechanical engineering as agent models for efficient calculation of the hydrodynamic system, and typically represent eigenvalue orthogonal decomposition (ProperOthogonal Decomposition, POD), dynamic modal decomposition (Dynamic Mode Decompositon, DMD), kriging models, support vector machines, artificial neural networks, radial basis functions and the like. Compared with numerical simulation, the reduced-order model method greatly reduces the complexity of complex system solving, improves modeling and solving efficiency, and simultaneously maintains reasonable precision.
However, the traditional reduced order model method has some significant disadvantages, and most reduced order models are based on linear methods, so that nonlinear characteristics are difficult to capture. Meanwhile, most of the methods are limited to prediction of low-dimensional performance, such as aerodynamic coefficient, pressure distribution and the like, and it is difficult to describe a high-fidelity complete flow field. In addition, these methods are sensitive to model parameters and do not easily scale to high dimensional design variables. Compared with the method, the deep neural network (Deep Netural Network, DNN) can easily process high-dimensional design variables, has strong processing capacity for nonlinear problems and multi-scale problems and problems including time sequence information, is expected to make up for the inherent defects of the traditional reduced-order model, is very suitable for modeling of a nonlinear dynamics system, and constructs a transonic buffeting flow field of a predicted airfoil of the reduced-order model to reduce calculation load.
Disclosure of Invention
For different flow states and flow problems, the problems to be paid attention to are different when constructing a prediction model based on deep learning. And different neural networks are also suitable for solving the problems of different types due to different construction ideas. The convolutional neural network is very suitable for modeling of image and space data, and the cyclic neural network such as long-term memory and the like is suitable for modeling of time sequence data. The wing-shaped flow field generating transonic buffeting has strong nonlinear characteristics in time and space, and the independent convolutional neural network and the cyclic neural network have limitations to a certain extent.
Therefore, the invention combines the performance advantages of two neural networks, and constructs a novel hybrid neural network model for predicting an airfoil unsteady flow field, and the model has good time feature extraction capability and spatial feature extraction capability.
The invention provides a rapid prediction method of an airfoil transonic buffeting flow field based on a deep neural network, which can rapidly obtain flow field information at a future moment under the condition of knowing flow field information of a finite continuous equal time interval, and comprises the following steps:
step 1: generating an unsteady sample set; the method specifically comprises the following steps:
step 1.1: selecting a reference airfoil of an unsteady sample set, generating an airfoil calculation grid by adopting an elliptic partial differential equation, mapping the grid from a physical space to a calculation space of a uniform rectangular grid in a plane by coordinate transformation, solving an unsteady flow field of the reference airfoil by adopting a body-separation vortex simulation (Detached Eddy Simulation, DES) method, and outputting flow field data (comprising pressure, density and speed components) of limited moments of a fixed time interval;
step 1.2: regenerating an interpolation grid according to the flow area concerned by the prediction problem, and interpolating flow field information on the grid to serve as a sample for training and testing;
step 1.3: and (3) carrying out normalization processing on the sample data set, and linearly transforming the interval of the sample data to [0,1] by adopting a linear normalization method.
Step 2: constructing a neural network model for rapid prediction of an unsteady flow field;
the constructed neural network model for the rapid prediction of the unsteady flow field is a hybrid neural network model, and the composition modules of the neural network model comprise a Convlstm module, a convolution module and a deconvolution module; the Convlstm module comprises a Convlstm layer, the convolution module sequentially comprises a convolution layer, a BN layer and a ReLU layer, and the deconvolution module sequentially comprises a deconvolution layer, a BN layer and a ReLU layer.
The neural network model adopts four downsampling and four upsampling, the feature images of the upsampling (expanding path) and downsampling (contracting path) corresponding layers have the same shape by adjusting the neural network parameters, skip connection (skip connection) is added between the feature images with the same dimension, and data fusion adopts addition operation.
The input data of the neural network model is a four-dimensional matrix, and the meaning of the matrix is as follows: the output data of the array is a three-dimensional matrix, wherein the matrix is defined as the channel number multiplied by the height multiplied by the width, the sequence number corresponds to the number of the multi-moment flow fields, the channel number corresponds to the variable number of the multivariable flow fields, and the height and the width correspond to the size of the two-dimensional flow fields.
Step 3: training the constructed neural network model; the root mean square loss (Mean Squared Error Loss, MSE) of flow field parameters is taken as a loss function, an Adam optimizer and an adaptive learning rate adjustment algorithm are adopted, and the initial learning rate is set to be 1 multiplied by 10 -3 And training the model, wherein the optimization target is that the loss function is minimum until the loss function of the training sample data set is not reduced.
Step 4: the trained neural network model is used for fast prediction of unsteady flow fields: and inputting the flow field information with known continuous finite time intervals into a trained neural network to obtain the flow field information at the future time.
Advantageous effects
Compared with the prior art, the invention has the following advantages:
1. according to the invention, the flow field information based on the calculation grid comprises pressure intensity, density, x-direction speed component and z-direction speed component as the input of the neural network model, and the flow field information at the future moment is predicted.
2. According to the invention, the flow field parameters are interpolated onto the simplified interpolation grid according to the concerned flow field region, so that the flow field parameters are used for predicting the unsteady flow field, the number of data points and the time consumption can be reduced as much as possible while the prediction precision of the unsteady flow field is ensured, and the efficiency is improved.
3. According to the invention, convlstm and U-Net are combined to construct a hybrid neural network model, and on the basis of the powerful space-time information processing capability of Convlstm, the U-Net further enhances the extraction capability of the whole network on space-time characteristics, so that the prediction precision of an unsteady flow field can be improved.
4. The invention aims at the structure of the transonic buffeting flow field of the wing profile and training the neural network, and has high pertinence, so that the transonic buffeting flow field of the wing profile can be rapidly and accurately predicted.
Additional aspects and advantages of the invention 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 the invention.
Drawings
The foregoing and/or additional aspects and advantages of the invention will become apparent and may be better understood from the following description of embodiments taken in conjunction with the accompanying drawings in which:
FIG. 1 is a schematic diagram of a deep neural network for fast prediction of unsteady flow fields according to the present invention
FIG. 2 shows an airfoil flow-around calculation grid, (a) shows a grid overview, and (b) shows a partial enlarged view.
Fig. 3 shows an interpolation grid for neural network training, (a) shows a grid overview, and (b) shows a partial enlarged view.
Fig. 4 shows the effect of predicting the flow field at a certain moment in the flow around the airfoil where the buffeting phenomenon occurs. The method comprises the steps of (a) calculating the pressure of an airfoil bypass flow field for the CFD, (b) predicting the pressure of the airfoil bypass flow field for the deep learning model, (c) predicting the error of the pressure of the airfoil bypass flow field for the deep learning model, (d) calculating the density of the airfoil bypass flow field for the CFD, (e) predicting the density of the airfoil bypass flow field for the deep learning model, (f) predicting the error of the density of the airfoil bypass flow field for the deep learning model, (g) calculating the x-direction velocity component of the airfoil bypass flow field for the CFD, (h) predicting the x-direction velocity component of the airfoil bypass flow field for the deep learning model, (i) predicting the z-direction velocity component of the airfoil bypass flow field for the deep learning model, (j) calculating the z-direction velocity component of the airfoil bypass flow field for the CFD, (k) predicting the z-direction velocity component of the airfoil bypass flow field for the deep learning model.
Detailed Description
The following detailed description of embodiments of the invention is exemplary and intended to be illustrative of the invention and not to be construed as limiting the invention.
The airfoil transonic buffeting flow field rapid prediction method based on the deep neural network comprises the steps of generating an unsteady flow field sample set; constructing a neural network model for rapid prediction of an unsteady flow field; training the constructed neural network model; the trained deep neural network is used for fast prediction of an unsteady flow field. The method comprises the following specific steps:
step 1: generating an unsteady flow field sample set:
1) Transonic turbulent flow field of the supercritical airfoil OAT15A with the chord length of 0.23m is simulated by adopting a deshuffling vortex numerical simulation method (Detached Eddy Simulation, DES). The calculation states are as follows: re=13.04×10 6 ，Ma＝0.73，α＝3.5°，T ∞ = 487.98R, setting the time step to Δt= 5.189 ×10 -6 s. The calculation grid adopts CH-shaped structured grid, 657 wing sections are distributed circumferentially, encryption is carried out at the middle parts of the front edge, the rear edge and the upper surface of the wing sections, and the height of the first layer of the auxiliary surface layer is set to be 4.6x10 -7 m, a global and local distribution diagram of the computational grid is shown in fig. 2. The boundary condition of the object plane is an adiabatic and slip-free wall surface, and the turbulence model adopts an SST turbulence model. And (5) storing a complete flow field every 10 time steps, and extracting the complete flow field at 2000 periodic regular moments.
2) And regenerating an interpolation grid according to the flow field structure and model training requirements, and interpolating flow field information on the grid to serve as a sample for training and testing. As shown in fig. 3, which is a global and local distribution diagram of interpolation grids for neural network training, the two grid blocks have sizes of 245×89 and 61×221, respectively.
3) And (3) carrying out normalization processing on the sample data set, and linearly transforming the interval of the sample data to [0,1] by adopting a linear normalization method.
Step 2: constructing a neural network model for fast prediction of an unsteady flow field:
1) A hybrid neural network model is constructed, and the constituent modules thereof comprise a Convlstm module, a convolution module and a deconvolution module. The model uses four downsamples and four upsamples altogether. By adjusting the neural network parameters, the feature graphs of the up-sampling (expanding path) and down-sampling (contracting path) corresponding layers also have the same shape, skip connection (skip connection) is added between feature graphs with the same dimension, and data fusion adopts addition operation.
The convolution module of the model sequentially comprises a convolution layer, a BN layer and a ReLU layer, the deconvolution module sequentially comprises a deconvolution layer, a BN layer and a ReLU layer, and the Convlstm module comprises a Convlstm layer.
2) The input data of the neural network is a four-dimensional matrix, the matrix has the meaning of serial number, channel number, height and width, the output data of the neural network is a three-dimensional matrix, and the matrix has the meaning of channel number, height and width. The number of the sequences corresponds to the number of the flow fields at multiple moments, the number of the channels corresponds to the variable number of the multivariable flow fields, and the height and the width correspond to the size of the two-dimensional flow fields.
Step 3: training the constructed neural network model;
1) The sample set is divided into a training sample and a test sample, and the ratio of the training sample to the test sample is 4:1.
2) The root mean square loss (Mean Squared Error Loss, MSE) of flow field parameters (pressure, density, x-direction velocity component and z-direction velocity component) is used as a loss function, an Adam optimizer and an adaptive learning rate adjustment algorithm are adopted, and the initial learning rate is set to be 1 multiplied by 10 -3 And training the model, wherein the optimization target is that the loss function is minimum until the loss function of the training sample data set is not reduced.
Step 4: the trained deep neural network is used for fast prediction of an unsteady flow field.
And inputting the flow field information with known continuous finite time intervals into a trained neural network to obtain the flow field information at the future time. The average absolute error (Mean Absolute Error, MAE) of most samples on the training set is smaller than 0.002, the average absolute error (MAE) of most samples on the test set is smaller than 0.006, and the prediction accuracy is higher, so that the U-Convlstm constructed by us has good space-time feature extraction capability. The predicted effect of a certain moment in the transonic buffeting flow field is shown in fig. 4.
Although embodiments of the present invention have been shown and described above, it will be understood that the above embodiments are illustrative and not to be construed as limiting the invention, and that variations, modifications, alternatives, and variations may be made in the above embodiments by those skilled in the art without departing from the spirit and principles of the invention.
Claims (4)
1. A rapid prediction method of an airfoil transonic buffeting flow field based on a deep neural network is characterized by comprising the following steps of: the method comprises the following steps:
step 1: generating an unsteady sample set:
step 1.1: selecting a reference airfoil of an unsteady sample set, generating an airfoil calculation grid, mapping the grid from a physical space to a calculation space of a uniform rectangular grid in a plane through coordinate transformation, solving an unsteady flow field of the reference airfoil, and outputting flow field data of limited moments of a fixed time interval;
step 1.2: regenerating an interpolation grid according to the flow area concerned by the prediction problem, and interpolating flow field information on the grid to serve as a sample for training and testing;
step 1.3: carrying out normalization processing on the sample data set, and linearly transforming the interval of the sample data to [0,1] by adopting a linear normalization method;
step 2: constructing a neural network model for fast prediction of an unsteady flow field:
constructing a neural network model for fast prediction of an unsteady flow field by adopting a hybrid neural network model; the hybrid neural network model comprises a Convlstm module, a convolution module and a deconvolution module; the Convlstm module comprises a Convlstm layer, the convolution module sequentially comprises a convolution layer, a BN layer and a ReLU layer, and the deconvolution module sequentially comprises a deconvolution layer, a BN layer and a ReLU layer;
the hybrid neural network model adopts four downsampling and four upsampling, the feature images of the upsampling and downsampling corresponding layers have the same shape by adjusting the neural network parameters, skip connection is added between the feature images with the same dimension, and data fusion adopts addition operation;
the input data of the hybrid neural network model is a four-dimensional matrix, and the meaning of the matrix is as follows: the number of the channels is multiplied by the height is multiplied by the width, the output data of the matrix is a three-dimensional matrix, the meaning of the matrix is that the number of the channels is multiplied by the height is multiplied by the width, the number of the channels corresponds to the number of the flow fields at multiple moments, the number of the channels corresponds to the variable number of the multivariable flow fields, and the height and the width correspond to the size of the two-dimensional flow fields;
step 3: training the constructed neural network model;
step 4: the trained neural network model is used for fast prediction of unsteady flow fields: and inputting the flow field information with known continuous finite time intervals into a trained neural network model to obtain the flow field information at the future time.
2. The rapid prediction method for the wing-shaped transonic buffeting flow field based on the deep neural network, which is characterized by comprising the following steps of: in the step 1.1, an elliptic partial differential equation is adopted to generate an airfoil calculation grid, and a body-separation vortex simulation method is adopted to solve the unsteady flow field of the reference airfoil.
3. The rapid prediction method for the wing-shaped transonic buffeting flow field based on the deep neural network, which is characterized by comprising the following steps of: in step 1.1, the output flow field data includes pressure, density and velocity components.
4. A base according to claim 3A rapid prediction method for airfoil transonic buffeting flow field of a deep neural network is characterized in that: in the step 3, firstly dividing a sample data set into a training sample and a test sample, wherein the ratio of the training sample to the test sample is 4:1; the root mean square loss of flow field parameters is used as a loss function, an Adam optimizer and an adaptive adjustment learning rate algorithm are adopted, and the initial learning rate is set to be 1 multiplied by 10 -3 And training the model, wherein the optimization target is that the loss function is minimum until the loss function of the training sample data set is not reduced any more, and obtaining the trained neural network model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211381140.9A CN116306206A (en) | 2022-11-06 | 2022-11-06 | Airfoil transonic buffeting flow field rapid prediction method based on deep neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211381140.9A CN116306206A (en) | 2022-11-06 | 2022-11-06 | Airfoil transonic buffeting flow field rapid prediction method based on deep neural network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116306206A true CN116306206A (en) | 2023-06-23 |
Family
ID=86796551
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211381140.9A Pending CN116306206A (en) | 2022-11-06 | 2022-11-06 | Airfoil transonic buffeting flow field rapid prediction method based on deep neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116306206A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117150951A (en) * | 2023-08-10 | 2023-12-01 | 中国船舶集团有限公司第七一九研究所 | Pump equipment three-dimensional flow field calculation acceleration method |
-
2022
- 2022-11-06 CN CN202211381140.9A patent/CN116306206A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117150951A (en) * | 2023-08-10 | 2023-12-01 | 中国船舶集团有限公司第七一九研究所 | Pump equipment three-dimensional flow field calculation acceleration method |
| CN117150951B (en) * | 2023-08-10 | 2024-03-01 | 中国船舶集团有限公司第七一九研究所 | Pump equipment three-dimensional flow field calculation acceleration method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Umetani et al. | Learning three-dimensional flow for interactive aerodynamic design | |
| Esmaeilzadeh et al. | Meshfreeflownet: A physics-constrained deep continuous space-time super-resolution framework | |
| CN113111430B (en) | Elastic aircraft flight dynamics modeling method based on nonlinear aerodynamic order reduction | |
| Lombardi et al. | Numerical simulation of sailing boats: dynamics, FSI, and shape optimization | |
| CN112016167A (en) | Aircraft aerodynamic shape design method and system based on simulation and optimization coupling | |
| Chen et al. | FlowDNN: a physics-informed deep neural network for fast and accurate flow prediction | |
| Yamazaki et al. | Geometry parameterization and computational mesh deformation by physics-based direct manipulation approaches | |
| CN116306206A (en) | Airfoil transonic buffeting flow field rapid prediction method based on deep neural network | |
| CN114065662B (en) | Method suitable for rapidly predicting airfoil flow field with variable grid topology | |
| CN112380793A (en) | Turbulence combustion numerical simulation parallel acceleration implementation method based on GPU | |
| Jacob et al. | Deep learning for real-time aerodynamic evaluations of arbitrary vehicle shapes | |
| Alauzet et al. | Some progress on CFD high lift prediction using metric-based anisotropic mesh adaptation | |
| Lyu et al. | Multi-fidelity prediction of fluid flow based on transfer learning using Fourier neural operator | |
| Hu et al. | Flow field reconstructions with gans based on radial basis functions | |
| Deng et al. | Prediction of transonic flow over supercritical airfoils using geometric-encoding and deep-learning strategies | |
| CN111275250B (en) | Strong current region sea surface temperature forecasting method based on anisotropy | |
| CN110705183A (en) | Multilayer grid LBM evolution method with buffer area | |
| CN103218493B (en) | A kind of quick method for numerical simulation such as geometric analysis such as grade based on multi grid | |
| Scholten et al. | An uncoupled method for fluid-structure interaction analysis with application to aerostructural design | |
| CN115906421A (en) | Method for simulating space-time conservation numerical value of fish habitat flow field | |
| Shahane et al. | Convolutional neural network for flow over single and tandem elliptic cylinders of arbitrary aspect ratio and angle of attack | |
| Mackle et al. | Efficient and Flexible Methodology for the Aerodynamic Shape Optimization of Hypersonic Vehicle Concepts in a High-Dimensional Design Space | |
| CN112182909A (en) | Flow solver establishing method for industrial CAE direction | |
| CN108564658A (en) | A kind of gradual three-dimensional entity model structure system and method based on reverse-engineering | |
| Romano et al. | Convolutional-neural-network-based DES-level aerodynamic flow field generation from URANS data |
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 |