CN118569145A - Reynolds stress prediction method, device, equipment and medium - Google Patents
Reynolds stress prediction method, device, equipment and medium Download PDFInfo
- Publication number
- CN118569145A CN118569145A CN202411047962.2A CN202411047962A CN118569145A CN 118569145 A CN118569145 A CN 118569145A CN 202411047962 A CN202411047962 A CN 202411047962A CN 118569145 A CN118569145 A CN 118569145A
- Authority
- CN
- China
- Prior art keywords
- turbulence
- target
- tensor
- data
- model
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 68
- 238000012549 training Methods 0.000 claims abstract description 102
- 230000006870 function Effects 0.000 claims abstract description 91
- 238000012545 processing Methods 0.000 claims abstract description 34
- 238000013528 artificial neural network Methods 0.000 claims abstract description 31
- 238000013507 mapping Methods 0.000 claims abstract description 17
- 238000013527 convolutional neural network Methods 0.000 claims description 25
- 238000004590 computer program Methods 0.000 claims description 16
- 239000013598 vector Substances 0.000 claims description 14
- 238000000605 extraction Methods 0.000 claims description 9
- 238000004364 calculation method Methods 0.000 claims description 6
- 230000004913 activation Effects 0.000 claims description 4
- 230000008859 change Effects 0.000 claims description 4
- 239000012530 fluid Substances 0.000 abstract description 9
- 230000008569 process Effects 0.000 description 13
- 238000011160 research Methods 0.000 description 8
- 238000005259 measurement Methods 0.000 description 7
- 238000010586 diagram Methods 0.000 description 6
- 238000005516 engineering process Methods 0.000 description 6
- 238000004891 communication Methods 0.000 description 5
- 238000013459 approach Methods 0.000 description 4
- 238000013473 artificial intelligence Methods 0.000 description 4
- 230000009471 action Effects 0.000 description 3
- 230000003595 spectral effect Effects 0.000 description 3
- NAWXUBYGYWOOIX-SFHVURJKSA-N (2s)-2-[[4-[2-(2,4-diaminoquinazolin-6-yl)ethyl]benzoyl]amino]-4-methylidenepentanedioic acid Chemical compound C1=CC2=NC(N)=NC(N)=C2C=C1CCC1=CC=C(C(=O)N[C@@H](CC(=C)C(O)=O)C(O)=O)C=C1 NAWXUBYGYWOOIX-SFHVURJKSA-N 0.000 description 2
- 238000012935 Averaging Methods 0.000 description 2
- 230000000903 blocking effect Effects 0.000 description 2
- 238000004422 calculation algorithm Methods 0.000 description 2
- 238000012937 correction Methods 0.000 description 2
- 238000004880 explosion Methods 0.000 description 2
- 230000010349 pulsation Effects 0.000 description 2
- 230000003321 amplification Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000012938 design process Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000003199 nucleic acid amplification method Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 239000002245 particle Substances 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 238000009877 rendering Methods 0.000 description 1
- 238000001228 spectrum Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 1
Classifications
-
- 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
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The application discloses a Reynolds stress prediction method, a device, equipment and a medium, which relate to the technical field of fluid mechanics and comprise the following steps: calculating turbulence data in a turbulent flow scenario; taking the original speed characteristic data as first target training data, and carrying out model training on a speed field characteristic and turbulence energy mapping relation of an implicit predicted turbulence energy model by utilizing the first target training data to obtain a target implicit predicted turbulence energy model; characteristic stitching is carried out on tensor invariant of the turbulence data, speed components and position information to obtain second target training data, the second target training data is input into a preset depth neural network, and scalar basis functions are obtained through the preset depth neural network; and inputting target turbulence data of a target turbulence flow scene into a Reynolds stress prediction model constructed based on a target implicit predicted turbulence energy model and a scalar basis function, so as to perform fitting processing on the output isotropic tensor and anisotropic tensor through the Reynolds stress prediction model, and obtain the Reynolds stress.
Description
Technical Field
The invention relates to the technical field of fluid mechanics, in particular to a Reynolds stress prediction method, a device, equipment and a medium.
Background
The Navier-Stokes (NS) equation is the core equation describing fluid motion in fluid dynamics, while the Reynolds average Navier-Stokes (RANS) equation is the primary method currently applied to engineering problems. The RANS equation is obtained by time-averaging the NS equation, which introduces an additional stress term, namely reynolds stress, which plays a key role in accurately describing the average characteristics of turbulent flow. In order to close the RANS equation and express the reynolds stress term, the most common approach is to use the Boussinesq vortex-induced assumption that simplifies the handling of the turbulence model by linearizing the relationship between reynolds stress and average velocity gradient. However, the Boussinesq vortex bonding assumption cannot deal with the anisotropy of the flow field, and advances in artificial intelligence technology have prompted an increasing interest in turbulence blocking problem studies on the RANS equation using deep neural networks. However, this data-driven approach, while fitting, lacks a profound understanding of the physical process, resulting in a lack of physical meaning in its predictions, i.e., inaccurate predictions of reynolds stresses. Therefore, there is an urgent need to incorporate physical principles into models to enhance the accuracy and interpretation of their predictions. However, in the prior reynolds stress research, most of the reynolds stress research is based on measurement technology, numerical calculation and a pure data driving method, and more measurement experimental data turbulence energy and turbulence dissipation rate are needed.
In summary, how to utilize less experimental data to provide an interpretable Reynolds stress anisotropic tensor double-model prediction technique for Reynolds stress of each turbulent flow scene, so as to realize accurate prediction of Reynolds stress in different turbulent flow scenes.
Disclosure of Invention
In view of the above, the present invention aims to provide a method, a device, an apparatus and a medium for predicting reynolds stress, which can provide an interpretable reynolds stress anisotropic tensor double-model prediction technique for reynolds stress of each turbulent flow scene by using less experimental data, so as to realize accurate prediction of reynolds stress in different turbulent flow scenes. The specific scheme is as follows:
In a first aspect, the application discloses a reynolds stress prediction method, which comprises the following steps:
Calculating turbulence data in a turbulence flow scene by using a preset effective viscosity assumption method;
taking the original speed characteristic data as first target training data, and carrying out model training on a speed field characteristic and turbulence energy mapping relation of an implicit predicted turbulence energy model by utilizing the first target training data so as to obtain a target implicit predicted turbulence energy model for predicting turbulence energy;
Characteristic splicing is carried out on tensor invariant, speed components and position information in the turbulence data to obtain second target training data, the second target training data is input into a preset depth neural network, and a scalar basis function is obtained through the preset depth neural network;
And inputting target turbulence data of a target turbulence flow scene into a Reynolds stress prediction model constructed based on the target implicit prediction turbulence energy model and the scalar basis function, and performing fitting processing on an isotropic tensor output based on the target implicit prediction turbulence energy model and an anisotropic tensor output based on the scalar basis function through the Reynolds stress prediction model to obtain corresponding Reynolds stress.
Optionally, the calculating turbulence data in a turbulence flow scene by using a preset effective viscosity assumption method includes:
calculating historical turbulence energy in a turbulence flow scene through an RANS equation and historical turbulence flow field information;
Calculating a strain rate tensor and a rotation rate tensor using a velocity field of the historical turbulence flow field information to construct and calculate a base tensor and a tensor invariant based on the strain rate tensor and the rotation rate tensor;
and performing variable scaling processing on the historical turbulence energy, the strain rate tensor, the rotation rate tensor, the base tensor and the tensor invariant to obtain amplified turbulence data.
Optionally, the training the implicit predicted turbulent energy model by using the original speed feature data as first target training data and using the first target training data to perform model training of a mapping relationship between speed field features and turbulent energy on the implicit predicted turbulent energy model to obtain a target implicit predicted turbulent energy model for predicting turbulent energy, including:
Extracting original speed characteristic data from the original speed field and the speed gradient data, and performing characteristic splicing on the original speed characteristic data according to the sequence of speed components and speed gradients to obtain first target training data;
Inputting the first target training data into an implicit predictive turbulence energy model, and performing coding processing on the first target training data through a fully-connected coding layer of the implicit predictive turbulence energy model to obtain a coded training feature vector;
Inputting the encoded training feature vector to a feature extraction layer of the implicit predictive turbulence energy model to extract flow field velocity features through the feature extraction layer;
And inputting the flow field speed characteristics into a preset transducer encoder so that the preset transducer encoder learns and acquires target flow field speed characteristics about the basic state and dynamic change of the flow field, and captures the mapping relation between the speed field characteristics and the turbulence energy to obtain a target implicit predicted turbulence energy model for predicting the turbulence energy.
Optionally, the inputting the second target training data into a preset depth neural network to obtain a scalar basis function through the preset depth neural network includes:
The second target training data is input into a transducer-CNN network to fit scalar basis functions for expressing anisotropic tensors of reynolds stress tensors through the transducer-CNN network.
Optionally, the inputting the second target training data into a transducer-CNN network to fit a scalar basis function for expressing an anisotropic tensor of a reynolds stress tensor through the transducer-CNN network includes:
Inputting the second target training data into a transducer-CNN network so as to obtain a scalar function through the transducer-CNN network;
scaling the base tensor calculated based on the average strain rate tensor and the average rotation rate tensor through the linear layer and the activation function to obtain a target base tensor meeting the preset numerical interval condition;
Fitting the target basis tensor and the scalar function through the transducer-CNN network to obtain a scalar basis function for expressing an anisotropic tensor of the reynolds stress tensor.
Optionally, the inputting the target turbulence data of the target turbulence flow scene into a reynolds stress prediction model constructed based on the target implicit prediction turbulence energy model and the scalar basis function, so as to perform fitting processing on an isotropic tensor output based on the target implicit prediction turbulence energy model and an anisotropic tensor output based on the scalar basis function through the reynolds stress prediction model, so as to obtain a corresponding reynolds stress, including:
inputting target turbulence data of a target turbulence flow scene into a Reynolds stress prediction model constructed based on the target implicit prediction turbulence energy model and the scalar basis function, so as to predict target turbulence energy corresponding to the target turbulence data through the target implicit prediction turbulence energy model, and calculating an isotropy tensor based on the target turbulence energy;
Determining an anisotropic tensor from the scalar basis function and a target turbulence energy;
And fitting the isotropic tensor and the anisotropic tensor through the Reynolds stress prediction model to output corresponding Reynolds stress.
Optionally, the fitting the isotropic tensor and the anisotropic tensor by the reynolds stress prediction model to output the corresponding reynolds stress includes:
By a Reynolds stress prediction model and according to Fitting the isotropic tensor and the anisotropic tensor to output a predicted reynolds stress, wherein,The isotropic tensor is represented by a graph,The anisotropic tensor is represented by a graph,Representing predicted reynolds stresses.
In a second aspect, the application discloses a reynolds stress prediction apparatus, comprising:
the data calculation module is used for calculating turbulence data in a turbulence flow scene by using a preset effective viscosity assumption method;
The first model training module is used for taking the original speed characteristic data as first target training data, and carrying out model training on the speed field characteristic and turbulence energy mapping relation of the implicit predicted turbulence energy model by utilizing the first target training data so as to obtain a target implicit predicted turbulence energy model for predicting turbulence energy;
The second model training module is used for carrying out characteristic splicing on tensor invariant, speed components and position information in the turbulence data to obtain second target training data, and inputting the second target training data into a preset depth neural network so as to obtain a scalar basis function through the preset depth neural network;
And the prediction module is used for inputting target turbulence data of a target turbulence flow scene into a Reynolds stress prediction model constructed based on the target implicit prediction turbulence energy model and the scalar basis function, and carrying out fitting processing on an isotropy tensor output based on the target implicit prediction turbulence energy model and an anisotropy tensor output based on the scalar basis function through the Reynolds stress prediction model so as to obtain corresponding Reynolds stress.
In a third aspect, the present application discloses an electronic device, comprising:
a memory for storing a computer program;
and a processor for executing the computer program to implement the steps of the reynolds stress prediction method disclosed above.
In a fourth aspect, the present application discloses a computer-readable storage medium for storing a computer program; wherein the computer program when executed by a processor implements the steps of the reynolds stress prediction method disclosed above.
The application discloses a method for calculating turbulence data in a turbulence flow scene by using a preset effective viscosity assumption method; taking the original speed characteristic data as first target training data, and carrying out model training on a speed field characteristic and turbulence energy mapping relation of an implicit predicted turbulence energy model by utilizing the first target training data so as to obtain a target implicit predicted turbulence energy model for predicting turbulence energy; characteristic splicing is carried out on tensor invariant, speed components and position information in the turbulence data to obtain second target training data, the second target training data is input into a preset depth neural network, and a scalar basis function is obtained through the preset depth neural network; and inputting target turbulence data of a target turbulence flow scene into a Reynolds stress prediction model constructed based on the target implicit prediction turbulence energy model and the scalar basis function, and performing fitting processing on an isotropic tensor output based on the target implicit prediction turbulence energy model and an anisotropic tensor output based on the scalar basis function through the Reynolds stress prediction model to obtain corresponding Reynolds stress. Therefore, the dependence on experimental data is remarkably reduced by implicitly predicting turbulence energy instead of directly using measurement experimental data, and particularly under the condition of limited experimental resources, the cost is saved, the feasibility and the efficiency of research are improved, and in addition, compared with a fully-connected network in the prior art, the Reynolds stress prediction model has more excellent performance in the aspects of processing complex data structures and extracting deep features. The complex model architecture enables the relation of the Reynolds stress to be mapped more accurately, and the accuracy and the efficiency of the Reynolds stress prediction are improved remarkably.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings that are required to be used in the embodiments or the description of the prior art will be briefly described below, and it is obvious that the drawings in the following description are only embodiments of the present invention, and that other drawings can be obtained according to the provided drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flow chart of a method for predicting Reynolds stress disclosed in the present application;
FIG. 2 is a diagram of a transducer-CNN architecture according to the present disclosure;
FIG. 3 is a diagram of a dual-model global framework for Reynolds stress in accordance with the present disclosure;
FIG. 4 is a flow chart of a method for fitting scalar functions based on tensor basis vectors in accordance with the present disclosure;
FIG. 5 is a schematic diagram of a structure of a Reynolds stress prediction apparatus according to the present disclosure;
Fig. 6 is a block diagram of an electronic device according to the present disclosure.
Detailed Description
The following description of the embodiments of the present application will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present application, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the application without making any inventive effort, are intended to be within the scope of the application.
The Navier-Stokes (NS) equation is the core equation describing fluid motion in fluid dynamics, while the Reynolds average Navier-Stokes (RANS) equation is the primary method currently applied to engineering problems. The RANS equation is obtained by time-averaging the NS equation, which introduces an additional stress term, namely reynolds stress, which plays a key role in accurately describing the average characteristics of turbulent flow. In order to close the RANS equation and express the reynolds stress term, the most common approach is to use the Boussinesq vortex-induced assumption that simplifies the handling of the turbulence model by linearizing the relationship between reynolds stress and average velocity gradient. However, the Boussinesq vortex bonding assumption cannot deal with the anisotropy of the flow field, and advances in artificial intelligence technology have prompted an increasing interest in turbulence blocking problem studies on the RANS equation using deep neural networks. However, this data-driven approach, while fitting, lacks a profound understanding of the physical process, resulting in a lack of physical meaning in its predictions, i.e., inaccurate predictions of reynolds stresses. Therefore, there is an urgent need to incorporate physical principles into models to enhance the accuracy and interpretation of their predictions. However, in the prior reynolds stress research, most of the reynolds stress research is based on measurement technology, numerical calculation and a pure data driving method, and more measurement experimental data turbulence energy and turbulence dissipation rate are needed.
Therefore, the invention provides a Reynolds stress prediction scheme, which can provide an interpretable Reynolds stress anisotropic tensor double-model prediction technology for Reynolds stress of each turbulent flow scene by using less experimental data, and can realize accurate prediction of the Reynolds stress under different turbulent flow scenes.
Referring to fig. 1, an embodiment of the invention discloses a reynolds stress prediction method, which includes:
step S11: and calculating turbulence data in a turbulence flow scene by using a preset effective viscosity assumption method.
In this embodiment, the turbulent flow scenario specifically includes: in-duct turbulent flow scenarios, atmospheric turbulent flow scenarios, water flow turbulent flow scenarios, turbulent flow scenarios of fluid-mechanical design processes, flow scenarios of airflows around the aircraft, thus, in the face of in-duct turbulent flow scenarios, the turbulent flow data comprises: the fluid velocity and direction at different positions, including time-average velocity, pulsation velocity component, etc.; pressure data: pressure values of each point in the pipeline, pressure fluctuation conditions and the like; temperature data: if heat exchange and the like are involved, temperature distribution data exists; turbulence energy data: an energy level reflecting turbulence; turbulence intensity data: indicating the intensity of the turbulence; reynolds stress data: describing interactions between speed pulsations in different directions in turbulence; spectral data: spectral information of parameters such as speed, pressure, etc. can be used to analyze the frequency characteristics of turbulence; turbulence scale data: including integrating scale, taylor microscale, etc., for describing the spatial characteristics of turbulence. Facing an atmospheric turbulent flow scenario, the turbulence data includes: wind speed data: the wind speed and direction of different positions, including average wind speed, fluctuating wind speed and the like; wind direction data: a change in wind direction; temperature data: distribution and variation of atmospheric temperature; humidity data: information of atmospheric humidity; pressure data: the value of the atmospheric pressure; turbulence intensity data: indicating the intensity of the turbulence; turbulence scale data: such as integral scale, taylor microscale, etc., for describing the spatial characteristics of turbulence; spectral data: the frequency spectrum information of parameters such as wind speed, temperature and the like can be used for analyzing the frequency characteristic of turbulence. In this way, the turbulence data in other turbulence flow scenes are the corresponding data under the actual different scenes, and will not be described in detail.
The manner in which the turbulence data is obtained is specifically as follows:
Calculating historical turbulence energy in a turbulence flow scene through an RANS equation and historical turbulence flow field information; calculating a strain rate tensor and a rotation rate tensor using a velocity field of the historical turbulence flow field information to construct and calculate a base tensor and a tensor invariant based on the strain rate tensor and the rotation rate tensor; and performing variable scaling processing on the historical turbulence energy, the strain rate tensor, the rotation rate tensor, the base tensor and the tensor invariant to obtain amplified turbulence data. It will be appreciated that the historical turbulent kinetic energy is calculated by means of an N-S equation (RANS equation) in the form of a time average, the RANS equation being shown in the following equation (1):
; formula (1)
Wherein,Is the density of the particles, which is the density,AndIs a time-averaged version of the pressure and velocity component (U, V, W) in each direction.The reynolds stress is a symmetrical second-order tensor, the diagonal components UU, VV, WW are called positive stresses, and the off-diagonal components are called tangential stresses.
The reynolds stress tensor can be decomposed into an isotropic part and an anisotropic part. The isotropic part is related to the turbulent kinetic energy of the fluid, while the anisotropic part describes the stress distribution characteristics other than the isotropic kinetic energy distribution, as shown in equation (2):
; formula (2)
Wherein,As a result of the isotropic tensor,In order to be an anisotropic tensor,As a function of the kronecker Delta,Representing the Reynolds stress to be decomposed whenWhen the value is 1, otherwise, the value is 0, k is turbulence energy, namely, the value is defined as half of the Reynolds stress tensor trace, and the obtaining formula of the turbulence energy is shown as a formula (3):
; formula (3)
Calculating historical turbulence energy in a turbulent flow scene through the formula (3) and the historical turbulence flow field information;
Wherein UU, VV, WW represent each tangential Reynolds stress component of x, y, z in the historical turbulence flow field information respectively, in addition, the historical turbulence flow field information further includes: the position information x, y, z and velocity components U, V, W in the x, y, z directions and the velocity gradients u_x, u_y, u_z, v_x, v_y, v_z, w_x, w_y, w_z, also UV, UW, VW in the final predicted target 6 reynolds stress components τ= (UU, UV, UW, VV, WV, WW).
Then, a strain rate tensor and a rotation rate tensor are calculated, wherein the strain rate tensor rotation rate tensor represents a deformation rate and a rotation rate of the flow, respectively, based on a velocity field of the flow field in the historical turbulence flow field information. The velocity field vector of the fluid isWhere u, v, w are velocity components in the x, y, z directions. Then, the strain rate tensor component can be calculated by the formula (4), and the rotation rate tensor component can be calculated by the formula (5), so that the formulas (4) and (5) are as follows:
; formula (4)
; Formula (5)
Subsequently, the basis tensors are calculated, and in the Pope corrected effective viscosity assumption, by using a specific set of basis tensors, it can be ensured that the turbulence model satisfies a physical invariance such as Galileo invariance. The base tensor is defined by a set of an average strain rate tensor S and an average rotation rate tensor R, that is, the average strain rate tensor S and the average rotation rate tensor R determined by the strain rate tensor and the rotation rate tensor obtained by the formula (4) and the formula (5), respectively, to construct the base tensorAs shown in formula (6):
; formula (6)
Calculating tensor invariant: tensor invariants can capture key physical characteristics in a flow field, such as vorticity, strain rate, turbulence intensity and the like, and a set of scalar quantities consisting of characteristic values of S and R and combinations thereof, and the scalar functions are kept unchanged under coordinate transformation as shown in a formula (7).
; Formula (7)
Then, in order to improve the numerical stability of model training and speed up training, firstly, carrying out maximum and minimum normalization processing on all data variables, enabling the data to be scaled between 0 and 1, and eliminating the dimensional influence among indexes, as shown in a formula (8). To avoid that the normalized data is at a minimum (very close to 0), the normalized data is further expanded 100 times as shown in equation (9), so that the model data and weight updates are within a more "comfortable" range of values.
; Formula (8)
; Formula (9)
Wherein,Data normalized by historical turbulence energy, strain rate tensor, rotation rate tensor, base tensor and tensor invariant are represented,The amplified turbulence data, which is 100-fold amplified from the normalized data, is shown.
It can be seen that the aforementioned historical turbulence energy, strain rate tensor, rotation rate tensor, basis tensor and tensor invariant numerical range obtained by the formulas (1) to (7) are obtained by the formulas (8) and (9), and the amplified turbulence data after the rational processing is obtained.
Step S12: and taking the original speed characteristic data as first target training data, and training a model of a speed field characteristic and turbulence energy mapping relation of the implicit predicted turbulence energy model by utilizing the first target training data so as to obtain a target implicit predicted turbulence energy model for predicting turbulence energy.
In the embodiment, original speed characteristic data is extracted from an original speed field and speed gradient data, and the original speed characteristic data is spliced according to sequential characteristics of a speed component and a speed gradient to obtain first target training data; inputting the first target training data into an implicit predictive turbulence energy model, and performing coding processing on the first target training data through a fully-connected coding layer of the implicit predictive turbulence energy model to obtain a coded training feature vector; inputting the encoded training feature vector to a feature extraction layer of the implicit predictive turbulence energy model to extract flow field velocity features through the feature extraction layer; and inputting the flow field speed characteristics into a preset transducer encoder so that the preset transducer encoder learns and acquires target flow field speed characteristics about the basic state and dynamic change of the flow field, and captures the mapping relation between the speed field characteristics and the turbulence energy to obtain a target implicit predicted turbulence energy model for predicting the turbulence energy. It can be understood that the characteristic representation with the most information quantity for flow field analysis and modeling is extracted from the original speed field and speed gradient data in the historical turbulent flow scene as the original speed characteristic data, the original speed characteristic data is subjected to characteristic splicing according to the sequence of speed components and speed gradients to obtain first target training data speed_x, and it is noted that the original speed field and the speed gradient data belong to amplified turbulent flow data, wherein the characteristic splicing result of the first target training data is shown in formula (10):
; formula (10)
As shown in fig. 2, the vector is thenThe full-connection coding layer sent into the implicit predictive turbulence energy model is used for coding to obtain the coded characteristic vectorI.e. the encoded training feature vector, the encoding process is shown in equation (11), whereinRepresenting the coding function of the fully concatenated coding layer. It should be noted that the implicit predictive turbulence energy model is a transducer-CNN architecture model.
; Formula (11)
Then, the post-coding training feature vectorExtracting flow field speed features by a feature extractor of an implicit predictive turbulence energy prediction modelThat is, flow field velocity features are extracted by the feature extraction layer as shown in equation (12), in whichA function representing a feature extraction network consisting of three one-dimensional convolution layers.
; Formula (12)
It will be appreciated that the feature extraction network, i.e. the feature extractor, is built up by three one-dimensional convolution layers.
Then, referring to FIG. 3, the flow field velocity characteristics are shownThe flow field speed characteristics and turbulence energy are mapped by the preset transducer encoder constructed by the three-layer transducer encoder, namely, the flow field speed characteristics are further processed by the three-layer transducer encoderObtaining a deeper flow field representation using a self-attention mechanismWherein, the method comprises the steps of, wherein,The flow field velocity characteristic processing process by the function formed by the three layers of the transducer encoders is shown, and then the flow field velocity characteristic is converted and output into a turbulence energy predicted value by the full-connection relation mapper, so that the mapping from the velocity field characteristic to the turbulence energy is completed, and the process is shown as a formula (13) and a formula (14).
; Formula (13)
; Formula (14)
Referring to fig. 3, implicit predictive turbulence energy prediction model training is performed based on a mean square error function, as shown in equation (15).
; Formula (15)
Wherein k is the historical turbulence energy obtained by the formula (3),The turbulence energy predicted value calculated by the formula (14) is obtained, so that the difference value between the predicted turbulence energy predicted value and the historical turbulence energy is continuously optimized to reach the optimal target with the minimum difference value, the super-parameters of the model are continuously updated in the optimizing process, the turbulence energy prediction accuracy is higher, and the current super-parameters are used as the model super-parameters of the target implicit predicted turbulence energy model after the minimized optimal target is reached.
Step S13: and performing feature stitching on tensor invariant, speed components and position information in the turbulence data to obtain second target training data, and inputting the second target training data into a preset depth neural network so as to obtain a scalar basis function through the preset depth neural network.
In the present embodiment, referring to fig. 3, the input feature (second target training data) is obtained by concatenating the lambda 1,⋯,λ5 variable, the three-directional velocity component U, V, W, and the position information x, y, z, and the second target training data is obtainedAs shown in equation (16):
; formula (16)
In this embodiment, referring to fig. 3, the acquired second target training data is input to a preset depth neural network, so that a scalar basis function expressing an anisotropic tensor of the reynolds stress tensor is fitted through the preset depth neural network. And performing fitting training on the preset deep neural network through the second target training data to obtain a scalar basis function.
Step S14: and inputting target turbulence data of a target turbulence flow scene into a Reynolds stress prediction model constructed based on the target implicit prediction turbulence energy model and the scalar basis function, and performing fitting processing on an isotropic tensor output based on the target implicit prediction turbulence energy model and an anisotropic tensor output based on the scalar basis function through the Reynolds stress prediction model to obtain corresponding Reynolds stress.
In this embodiment, when there is a reynolds stress prediction requirement of a target turbulent flow scene, target turbulence data of the target turbulent flow scene is input to a reynolds stress prediction model constructed based on the target implicit prediction turbulence energy model and the scalar basis function so as to predict target turbulence energy corresponding to the target turbulence data through the target implicit prediction turbulence energy model, and an isotropic tensor is calculated based on the target turbulence energy; determining an anisotropic tensor from the scalar basis function and a target turbulence energy; and fitting the isotropic tensor and the anisotropic tensor through the Reynolds stress prediction model to output corresponding Reynolds stress. It can be understood that after the target turbulence data is input into the reynolds stress prediction model constructed based on the target implicit prediction turbulence energy model and the scalar basis function, the target implicit prediction turbulence energy model predicts the corresponding target turbulence energy based on the input target turbulence data, and then the isotropy tensor of the reynolds stress tensor is directly related to the turbulence kinetic energy k, as shown in the formula 17:
; formula (17)
Where k is the target turbulence energy predicted based on the target implicit predicted turbulence energy model.Is a Croneck symbol whenAnd 1 if not, and 0 if not.
The anisotropic tensor is then determined jointly with the target turbulence energy by a scalar basis function, specifically Pope derives a general form of constitutive relation between the normalized anisotropic tensor b and the target basis tensor based on Caley-Hamilton theory, as shown:
; formula (18)
Wherein,Is the target base tensor after nonlinear correction of the value range. Lambda 1,…,λ5 is the tensor invariant in equation (7),Is a scalar basis function of the fitting result. b is normalized by the target turbulence energy k. The anisotropy vector calculation formula is therefore as follows:
; formula (19)
Wherein,Is calculated by the formula (18), and k is the calculated target turbulence energy. After the isotropic tensor and the anisotropic tensor are determined, fitting the isotropic tensor and the anisotropic tensor by using a Reynolds stress prediction model to output predicted Reynolds stress, wherein the fitting process is as follows:
In this embodiment, the fitting, by the reynolds stress prediction model, the isotropic tensor and the anisotropic tensor to output the corresponding reynolds stress includes: by a Reynolds stress prediction model and according to Fitting the isotropic tensor and the anisotropic tensor to output corresponding reynolds stresses, wherein,The isotropic tensor is represented by a graph,The anisotropic tensor is represented by a graph,Representing predicted reynolds stresses. It is understood that the Reynolds stress tensor can be decomposed into an isotropic part and an anisotropic part under the effective viscosity assumption based on Pope correction by combining isotropic and anisotropic fit Reynolds stresses. The isotropic tensor is calculated by equation (17)The anisotropic tensor can be calculated from the above formula (19)Thus fitting the reynolds stress as obtained in equation (20).
; Formula (20)
By implicitly predicting the turbulence energy rather than directly using the measured experimental data, the dependence on experimental resources is significantly reduced, which not only saves cost, but also improves the feasibility and efficiency of research, especially in the case of limited resources. In addition, compared with the traditional fully-connected network, the converter-CNN model architecture adopted by the invention has more excellent performance in the aspects of processing complex data structures and extracting deep features. The complex model architecture enables the method to map the relation of Reynolds stresses more accurately, and improves the accuracy of prediction.
In order to more accurately predict the Reynolds stress, a weighted loss function is designed, the characteristics of Mean Square Error (MSE) and Mean Absolute Error (MAE) are comprehensively considered, and a determination coefficient (R2) is introduced as an evaluation index to strengthen the consideration of the correlation between a model predicted value and a target value. As shown in formula (21):
; formula (21)
Wherein,As a mean absolute error loss function.Is a mean square error loss function.To determine the coefficient loss function.Represent the firstThe true reynolds stress is that,Represent the firstThe predicted reynolds stress is a function of the predicted reynolds stress,The average value of each real Reynolds stress is represented, alpha, beta and gamma are weight coefficients, and the balance of different types of prediction errors and the improvement of the overall performance are realized by carefully selecting the weight coefficients.
Therefore, the cost function of the Reynolds stress anisotropy tensor double model (Reynolds stress prediction model)Wherein, the method comprises the steps of, wherein,A weighted loss function representing predicted reynolds stresses,A weighted loss function representing true reynolds stress.
The weighted loss function designed by the invention can more comprehensively consider various prediction errors in the model training process, and compared with the traditional root mean square error loss function, the weighted loss function can optimize the overall prediction performance of the model and improve the precision and reliability of a prediction result.
The application discloses a method for calculating turbulence data in a turbulence flow scene by using a preset effective viscosity assumption method; taking the original speed characteristic data as first target training data, and carrying out model training on a speed field characteristic and turbulence energy mapping relation of an implicit predicted turbulence energy model by utilizing the first target training data so as to obtain a target implicit predicted turbulence energy model for predicting turbulence energy; characteristic splicing is carried out on tensor invariant, speed components and position information in the turbulence data to obtain second target training data, the second target training data is input into a preset depth neural network, and a scalar basis function is obtained through the preset depth neural network; and inputting target turbulence data of a target turbulence flow scene into a Reynolds stress prediction model constructed based on the target implicit prediction turbulence energy model and the scalar basis function, and performing fitting processing on an isotropic tensor output based on the target implicit prediction turbulence energy model and an anisotropic tensor output based on the scalar basis function through the Reynolds stress prediction model to obtain corresponding Reynolds stress. Therefore, the dependence on experimental data is remarkably reduced by implicitly predicting turbulence energy instead of directly using measurement experimental data, and particularly under the condition of limited experimental resources, the cost is saved, the feasibility and the efficiency of research are improved, and in addition, compared with a fully-connected network in the prior art, the Reynolds stress prediction model has more excellent performance in the aspects of processing complex data structures and extracting deep features. The complex model architecture enables the relation of the Reynolds stress to be mapped more accurately, and the accuracy and the efficiency of the Reynolds stress prediction are improved remarkably.
Referring to fig. 4, the invention discloses a scalar function fitting method based on tensor basis vectors, which comprises the following steps:
Step S21: inputting the second target training data into a transducer-CNN network to obtain a scalar function through the transducer-CNN network.
In this embodiment, the second target training data is input into a transducer-CNN network to fit scalar basis functions for expressing anisotropic tensors of reynolds stress tensors through the transducer-CNN network. It will be appreciated that in the Pope corrected effective viscosity hypothesis, the scalar basis function g (n) is a functional expression for the lambda 1,⋯,λ5 variable in equation (7), introduced to express the anisotropic portion of the reynolds stress tensor, and to ensure that the turbulence model satisfies the physical invariance. The scalar basis function expression g (n) is fitted through a deep neural network technology, and the preset deep neural network of the scalar basis function expression g (n) is identical to a transducer-CNN framework of an implicit predictive turbulence energy prediction model.
Step S22: and scaling the base tensor calculated based on the average strain rate tensor and the average rotation rate tensor through the linear layer and the activation function to obtain the target base tensor meeting the preset numerical interval condition.
In this embodiment, in order to avoid the numerical explosion of the model fitting process caused by the overlarge value range of the base tensor, the value range of the base tensor is corrected in a nonlinear manner, and the value range of the base tensor is easily caused to be numerical explosion of the model training process due to the fact that the matrix of the average strain rate tensor and the average rotation rate tensor subjected to feature amplification is transformed. The base tensor is thus passed through a linear layer and activation function (Tanh function) in turn, scaling the base tensor to a smaller interval of values as shown in equation (22):
; formula (22)
Wherein,Is the base tensor in equation (6), so that the corrected base tensor is taken as the target base tensor。
Step S23: fitting the target basis tensor and the scalar function through the transducer-CNN network to obtain a scalar basis function for expressing an anisotropic tensor of the reynolds stress tensor.
In this embodiment, for the target base tensorAnd fitting the scalar function lambda n, wherein the converter-CNN network can learn the mapping relation between the scalar function lambda n and the anisotropic tensor so as to enable the fitted scalar basis function to accurately express the anisotropic tensor of the Reynolds stress.
It can be seen that the adopted transducer-CNN model architecture has more excellent performance in terms of processing complex data structures and extracting deep features than the traditional fully connected network. The complex model architecture enables the method to map the relation of Reynolds stresses more accurately, and improves the accuracy of prediction. And the anisotropic tensor can be modified in a nonlinear manner based on the predicted turbulence energy and the fitting scalar function, so that the numerical stability problem is solved, and the capability of the model for processing complex data is enhanced. An effective solution is provided for the common numerical problem in the turbulence prediction model, and the robustness of the model in a complex application scene is ensured.
Referring to fig. 5, the invention also discloses a reynolds stress prediction device, which comprises:
A data calculation module 11, configured to calculate turbulence data in a turbulence flow scene by using a preset effective viscosity assumption method;
the first model training module 12 is configured to use the original speed feature data as first target training data, and perform model training of a speed field feature and turbulence energy mapping relationship on the implicit predicted turbulence energy model by using the first target training data, so as to obtain a target implicit predicted turbulence energy model for predicting turbulence energy;
The second model training module 13 is configured to perform feature stitching on tensor invariant, speed component and position information in the turbulence data to obtain second target training data, and input the second target training data into a preset depth neural network, so as to obtain a scalar basis function through the preset depth neural network;
The prediction module 14 is configured to input target turbulence data of a target turbulence flow scene to a reynolds stress prediction model constructed based on the target implicit predicted turbulence energy model and the scalar basis function, so as to perform fitting processing on an isotropic tensor output based on the target implicit predicted turbulence energy model and an anisotropic tensor output based on the scalar basis function through the reynolds stress prediction model, so as to obtain a corresponding reynolds stress.
The application discloses a method for calculating turbulence data in a turbulence flow scene by using a preset effective viscosity assumption method; taking the original speed characteristic data as first target training data, and carrying out model training on a speed field characteristic and turbulence energy mapping relation of an implicit predicted turbulence energy model by utilizing the first target training data so as to obtain a target implicit predicted turbulence energy model for predicting turbulence energy; characteristic splicing is carried out on tensor invariant, speed components and position information in the turbulence data to obtain second target training data, the second target training data is input into a preset depth neural network, and a scalar basis function is obtained through the preset depth neural network; and inputting target turbulence data of a target turbulence flow scene into a Reynolds stress prediction model constructed based on the target implicit prediction turbulence energy model and the scalar basis function, and performing fitting processing on an isotropic tensor output based on the target implicit prediction turbulence energy model and an anisotropic tensor output based on the scalar basis function through the Reynolds stress prediction model to obtain corresponding Reynolds stress. Therefore, the dependence on experimental data is remarkably reduced by implicitly predicting turbulence energy instead of directly using measurement experimental data, and particularly under the condition of limited experimental resources, the cost is saved, the feasibility and the efficiency of research are improved, and in addition, compared with a fully-connected network in the prior art, the Reynolds stress prediction model has more excellent performance in the aspects of processing complex data structures and extracting deep features. The complex model architecture enables the relation of the Reynolds stress to be mapped more accurately, and the accuracy and the efficiency of the Reynolds stress prediction are improved remarkably.
Further, the embodiment of the present application further discloses an electronic device, and fig. 6 is a block diagram of an electronic device 20 according to an exemplary embodiment, where the content of the figure is not to be considered as any limitation on the scope of use of the present application.
Fig. 6 is a schematic structural diagram of an electronic device 20 according to an embodiment of the present application. The electronic device 20 may specifically include: at least one processor 21, at least one memory 22, a power supply 23, a communication interface 24, an input output interface 25, and a communication bus 26. Wherein the memory 22 is configured to store a computer program that is loaded and executed by the processor 21 to implement the relevant steps of the reynolds stress prediction method disclosed in any of the foregoing embodiments. In addition, the electronic device 20 in the present embodiment may be specifically an electronic computer.
In this embodiment, the power supply 23 is configured to provide an operating voltage for each hardware device on the electronic device 20; the communication interface 24 can create a data transmission channel between the electronic device 20 and an external device, and the communication protocol to be followed is any communication protocol applicable to the technical solution of the present application, which is not specifically limited herein; the input/output interface 25 is used for acquiring external input data or outputting external output data, and the specific interface type thereof may be selected according to the specific application requirement, which is not limited herein.
Processor 21 may include one or more processing cores, such as a 4-core processor, an 8-core processor, etc. The processor 21 may be implemented in at least one hardware form of DSP (DIGITAL SIGNAL Processing), FPGA (Field-Programmable gate array), PLA (Programmable Logic Array ). The processor 21 may also include a main processor, which is a processor for processing data in an awake state, also called a CPU (Central Processing Unit ), and a coprocessor; a coprocessor is a low-power processor for processing data in a standby state. In some embodiments, the processor 21 may integrate a GPU (Graphics Processing Unit, image processor) for rendering and drawing of content required to be displayed by the display screen. In some embodiments, the processor 21 may also include an AI (ARTIFICIAL INTELLIGENCE ) processor for processing computing operations related to machine learning.
The memory 22 may be a carrier for storing resources, such as a read-only memory, a random access memory, a magnetic disk, or an optical disk, and the resources stored thereon may include an operating system 221, a computer program 222, and the like, and the storage may be temporary storage or permanent storage.
The operating system 221 is used for managing and controlling various hardware devices on the electronic device 20 and the computer program 222, so as to implement the operation and processing of the processor 21 on the mass data 223 in the memory 22, which may be Windows Server, netware, unix, linux, etc. The computer program 222 may further include a computer program that can be used to perform other specific tasks in addition to the computer program that can be used to perform the reynolds stress prediction method performed by the electronic device 20 as disclosed in any of the foregoing embodiments. The data 223 may include, in addition to data received by the electronic device and transmitted by the external device, data collected by the input/output interface 25 itself, and so on.
Further, the application also discloses a computer readable storage medium for storing a computer program; wherein the computer program when executed by a processor implements the reynolds stress prediction method of the foregoing disclosure. For specific steps of the method, reference may be made to the corresponding contents disclosed in the foregoing embodiments, and no further description is given here.
In this specification, each embodiment is described in a progressive manner, and each embodiment is mainly described in a different point from other embodiments, so that the same or similar parts between the embodiments are referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant points refer to the description of the method section.
Those of skill would further appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, computer software, or combinations of both, and that the various illustrative elements and steps are described above generally in terms of functionality in order to clearly illustrate the interchangeability of hardware and software. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the solution. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present application. The steps of a method or algorithm described in connection with the embodiments disclosed herein may be embodied directly in hardware, in a software module executed by a processor, or in a combination of the two. The software modules may be disposed in random access Memory RAM (Random Access Memory), memory, read-Only Memory ROM (Read Only Memory), electrically programmable EPROM (Electrically Programmable Read Only Memory), electrically erasable programmable EEPROM (Electric Erasable Programmable Read Only Memory), registers, hard disk, a removable disk, a CD-ROM (Compact Disc-Read Only Memory), or any other form of storage medium known in the art.
Finally, it is further noted that relational terms such as first and second, and the like are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.
The foregoing has outlined rather broadly the more detailed description of the invention in order that the detailed description of the principles and embodiments of the invention may be better understood, and in order that the present invention may be better understood; meanwhile, as those skilled in the art will have variations in the specific embodiments and application scope in accordance with the ideas of the present invention, the present description should not be construed as limiting the present invention in view of the above.
Claims (10)
1. A method for reynolds stress prediction, comprising:
Calculating turbulence data in a turbulence flow scene by using a preset effective viscosity assumption method;
taking the original speed characteristic data as first target training data, and carrying out model training on a speed field characteristic and turbulence energy mapping relation of an implicit predicted turbulence energy model by utilizing the first target training data so as to obtain a target implicit predicted turbulence energy model for predicting turbulence energy;
Characteristic splicing is carried out on tensor invariant, speed components and position information in the turbulence data to obtain second target training data, the second target training data is input into a preset depth neural network, and a scalar basis function is obtained through the preset depth neural network;
And inputting target turbulence data of a target turbulence flow scene into a Reynolds stress prediction model constructed based on the target implicit prediction turbulence energy model and the scalar basis function, and performing fitting processing on an isotropic tensor output based on the target implicit prediction turbulence energy model and an anisotropic tensor output based on the scalar basis function through the Reynolds stress prediction model to obtain corresponding Reynolds stress.
2. The method of claim 1, wherein calculating turbulence data in a turbulence flow scenario using a preset effective viscosity hypothesis method comprises:
calculating historical turbulence energy in a turbulence flow scene through an RANS equation and historical turbulence flow field information;
Calculating a strain rate tensor and a rotation rate tensor using a velocity field of the historical turbulence flow field information to construct and calculate a base tensor and a tensor invariant based on the strain rate tensor and the rotation rate tensor;
and performing variable scaling processing on the historical turbulence energy, the strain rate tensor, the rotation rate tensor, the base tensor and the tensor invariant to obtain amplified turbulence data.
3. The method of claim 2, wherein the training the implicit predicted turbulent energy model using the raw velocity feature data as first target training data for model training of a velocity field feature and turbulent energy mapping relationship of the implicit predicted turbulent energy model to obtain the target implicit predicted turbulent energy model for predicting turbulent energy comprises:
Extracting original speed characteristic data from the original speed field and the speed gradient data, and performing characteristic splicing on the original speed characteristic data according to the sequence of speed components and speed gradients to obtain first target training data;
Inputting the first target training data into an implicit predictive turbulence energy model, and performing coding processing on the first target training data through a fully-connected coding layer of the implicit predictive turbulence energy model to obtain a coded training feature vector;
Inputting the encoded training feature vector to a feature extraction layer of the implicit predictive turbulence energy model to extract flow field velocity features through the feature extraction layer;
And inputting the flow field speed characteristics into a preset transducer encoder so that the preset transducer encoder learns and acquires target flow field speed characteristics about the basic state and dynamic change of the flow field, and captures the mapping relation between the speed field characteristics and the turbulence energy to obtain a target implicit predicted turbulence energy model for predicting the turbulence energy.
4. The reynolds stress prediction method of claim 1, wherein the inputting the second target training data into a preset depth neural network to obtain a scalar basis function by the preset depth neural network includes:
The second target training data is input into a transducer-CNN network to fit scalar basis functions for expressing anisotropic tensors of reynolds stress tensors through the transducer-CNN network.
5. The method of claim 4, wherein said inputting the second target training data into a transducer-CNN network to fit a scalar basis function for expressing an anisotropic tensor of a reynolds stress tensor via the transducer-CNN network comprises:
Inputting the second target training data into a transducer-CNN network so as to obtain a scalar function through the transducer-CNN network;
scaling the base tensor calculated based on the average strain rate tensor and the average rotation rate tensor through the linear layer and the activation function to obtain a target base tensor meeting the preset numerical interval condition;
Fitting the target basis tensor and the scalar function through the transducer-CNN network to obtain a scalar basis function for expressing an anisotropic tensor of the reynolds stress tensor.
6. The method according to claim 1, wherein the inputting the target turbulence data of the target turbulence flow scene into the reynolds stress prediction model constructed based on the target implicit predicted turbulence energy model and the scalar basis function so as to fit, by the reynolds stress prediction model, an isotropic tensor output based on the target implicit predicted turbulence energy model and an anisotropic tensor output based on the scalar basis function to obtain the corresponding reynolds stress includes:
inputting target turbulence data of a target turbulence flow scene into a Reynolds stress prediction model constructed based on the target implicit prediction turbulence energy model and the scalar basis function, so as to predict target turbulence energy corresponding to the target turbulence data through the target implicit prediction turbulence energy model, and calculating an isotropy tensor based on the target turbulence energy;
Determining an anisotropic tensor from the scalar basis function and a target turbulence energy;
And fitting the isotropic tensor and the anisotropic tensor through the Reynolds stress prediction model to output corresponding Reynolds stress.
7. The method of claim 6, wherein the fitting the isotropic tensor and the anisotropic tensor by the reynolds stress prediction model to output the corresponding reynolds stresses comprises:
By a Reynolds stress prediction model and according to Fitting the isotropic tensor and the anisotropic tensor to output corresponding reynolds stresses, wherein,The isotropic tensor is represented by a graph,The anisotropic tensor is represented by a graph,Representing predicted reynolds stresses.
8. A reynolds stress prediction apparatus, comprising:
the data calculation module is used for calculating turbulence data in a turbulence flow scene by using a preset effective viscosity assumption method;
The first model training module is used for taking the original speed characteristic data as first target training data, and carrying out model training on the speed field characteristic and turbulence energy mapping relation of the implicit predicted turbulence energy model by utilizing the first target training data so as to obtain a target implicit predicted turbulence energy model for predicting turbulence energy;
The second model training module is used for carrying out characteristic splicing on tensor invariant, speed components and position information in the turbulence data to obtain second target training data, and inputting the second target training data into a preset depth neural network so as to obtain a scalar basis function through the preset depth neural network;
And the prediction module is used for inputting target turbulence data of a target turbulence flow scene into a Reynolds stress prediction model constructed based on the target implicit prediction turbulence energy model and the scalar basis function, and carrying out fitting processing on an isotropy tensor output based on the target implicit prediction turbulence energy model and an anisotropy tensor output based on the scalar basis function through the Reynolds stress prediction model so as to obtain corresponding Reynolds stress.
9. An electronic device, comprising:
a memory for storing a computer program;
a processor for executing the computer program to implement the steps of the reynolds stress prediction method of any of claims 1 to 7.
10. A computer-readable storage medium storing a computer program; wherein the computer program when executed by a processor implements the steps of the reynolds stress prediction method of any of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202411047962.2A CN118569145B (en) | 2024-08-01 | 2024-08-01 | Reynolds stress prediction method, device, equipment and medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202411047962.2A CN118569145B (en) | 2024-08-01 | 2024-08-01 | Reynolds stress prediction method, device, equipment and medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN118569145A true CN118569145A (en) | 2024-08-30 |
| CN118569145B CN118569145B (en) | 2024-10-18 |
Family
ID=92464008
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202411047962.2A Active CN118569145B (en) | 2024-08-01 | 2024-08-01 | Reynolds stress prediction method, device, equipment and medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN118569145B (en) |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111324993A (en) * | 2020-02-21 | 2020-06-23 | 苏州浪潮智能科技有限公司 | Turbulent flow field updating method, device and related equipment |
| CN117807914A (en) * | 2024-02-26 | 2024-04-02 | 深圳市金众工程检验检测有限公司 | Real-time bridge stress detection method and system |
-
2024
- 2024-08-01 CN CN202411047962.2A patent/CN118569145B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111324993A (en) * | 2020-02-21 | 2020-06-23 | 苏州浪潮智能科技有限公司 | Turbulent flow field updating method, device and related equipment |
| CN117807914A (en) * | 2024-02-26 | 2024-04-02 | 深圳市金众工程检验检测有限公司 | Real-time bridge stress detection method and system |
Non-Patent Citations (3)
| Title |
|---|
| 刘骁飞;韦安阳;罗坤;樊建人;: "粗糙壁面上湍流边界层的直接数值模拟", 工程热物理学报, no. 12, 15 December 2014 (2014-12-15), pages 105 - 108 * |
| 朱志斌;刘强;白鹏;: "低雷诺数翼型层流分离现象大涡模拟方法", 空气动力学学报, no. 06, 15 December 2019 (2019-12-15), pages 51 - 59 * |
| 赵轲;高正红;黄江涛;李静;: "基于分区拼接网格技术高升力装置流场数值模拟", 应用力学学报, 31 December 2012 (2012-12-31), pages 1 - 3 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN118569145B (en) | 2024-10-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| KR102602195B1 (en) | Quantization of trained long-short-term memory neural networks | |
| US11741373B2 (en) | Turbulence field update method and apparatus, and related device thereof | |
| CN104463381B (en) | Building energy consumption Forecasting Methodology based on KPCA and WLSSVM | |
| CN113627093B (en) | Underwater mechanism trans-scale flow field characteristic prediction method based on improved Unet network | |
| Vlassis et al. | Component-based machine learning paradigm for discovering rate-dependent and pressure-sensitive level-set plasticity models | |
| CN115307780B (en) | Sea surface temperature prediction method, system and application based on time-space information interaction fusion | |
| CN113705809B (en) | Data prediction model training method, industrial index prediction method and device | |
| CN117407771B (en) | Bearing health state assessment method and device based on digital twin and related equipment | |
| CN111339642A (en) | Simulation model calibration method, system, readable medium and device | |
| CN117077526A (en) | Digital twinning-based radar TR module temperature prediction method | |
| CN114580710B (en) | Environmental monitoring method based on transducer time sequence prediction | |
| CN115983148A (en) | CFD simulation cloud picture prediction method, system, electronic device and medium | |
| Melchers | Simulation in time-invariant and time-variant reliability problems | |
| CN118569145B (en) | Reynolds stress prediction method, device, equipment and medium | |
| Zhou et al. | Multi-fidelity physics constrained neural networks for dynamical systems | |
| KR20220159056A (en) | System and method for predicting tidal residual using bidirectional lstm neural network | |
| Morand et al. | Deep learning of first-order nonlinear hyperbolic conservation law solvers | |
| CN117669249A (en) | Equipment environment adaptability simulation and evaluation method | |
| CN117370524A (en) | Training method of reply generation model, reply sentence generation method and device | |
| CN114792097B (en) | Method and device for determining prompt vector of pre-training model and electronic equipment | |
| CN115906863B (en) | Emotion analysis method, device, equipment and storage medium based on contrast learning | |
| CN114692529A (en) | CFD high-dimensional response uncertainty quantification method and device, and computer equipment | |
| CN116933676A (en) | Flow numerical simulation method, device, equipment and medium | |
| CN114385619B (en) | Multi-channel ocean observation time sequence scalar data missing value prediction method and system | |
| Soni et al. | Capturing functional relations in fluid–structure interaction via machine learning |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |