CN111651904A - Flow field data calculation method and device based on machine learning - Google Patents
Flow field data calculation method and device based on machine learning Download PDFInfo
- Publication number
- CN111651904A CN111651904A CN202010645910.0A CN202010645910A CN111651904A CN 111651904 A CN111651904 A CN 111651904A CN 202010645910 A CN202010645910 A CN 202010645910A CN 111651904 A CN111651904 A CN 111651904A
- Authority
- CN
- China
- Prior art keywords
- flow field
- field data
- calculated
- time step
- neural network
- 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
- 238000004364 calculation method Methods 0.000 title claims abstract description 67
- 238000010801 machine learning Methods 0.000 title claims abstract description 20
- 238000013528 artificial neural network Methods 0.000 claims abstract description 77
- 238000012937 correction Methods 0.000 claims abstract description 37
- 238000000034 method Methods 0.000 claims abstract description 35
- 238000004088 simulation Methods 0.000 claims abstract description 23
- 238000012549 training Methods 0.000 claims abstract description 13
- 230000002441 reversible effect Effects 0.000 claims abstract description 9
- 238000004590 computer program Methods 0.000 claims description 14
- 230000008569 process Effects 0.000 description 8
- 238000010586 diagram Methods 0.000 description 6
- 239000012530 fluid Substances 0.000 description 6
- 230000006870 function Effects 0.000 description 5
- 230000008859 change Effects 0.000 description 4
- 210000002569 neuron Anatomy 0.000 description 3
- 238000004422 calculation algorithm Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000002093 peripheral effect Effects 0.000 description 2
- 230000001360 synchronised effect Effects 0.000 description 2
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 2
- 238000013527 convolutional neural network Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000001771 impaired effect Effects 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000012827 research and development Methods 0.000 description 1
- 238000004904 shortening Methods 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000011144 upstream manufacturing Methods 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
-
- 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
- G06N3/045—Combinations of networks
-
- 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
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2113/00—Details relating to the application field
- G06F2113/08—Fluids
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Computing Systems (AREA)
- Biophysics (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Biomedical Technology (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Software Systems (AREA)
- Data Mining & Analysis (AREA)
- Geometry (AREA)
- Computer Hardware Design (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Pure & Applied Mathematics (AREA)
- Fluid Mechanics (AREA)
- Algebra (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The application relates to a flow field data calculation method and device based on machine learning. The method comprises the following steps: generating first flow field data of a simulation object at a first time step and second flow field data of a simulation object at a second time step by using an NS equation solver; inputting the first flow field data into a preset neural network, performing reverse training by using the difference value of the output value of the neural network and the second flow field data, determining the network structure parameters of the neural network, inputting the flow field data to be calculated of the current time step into the neural network corresponding to the network structure parameters when the flow field data is calculated, obtaining the flow field correction data of the flow field data to be calculated, inputting the flow field correction data into an NS equation solver, and obtaining the flow field data of the time step to be calculated corresponding to the flow field data to be calculated. By adopting the method, the calculation speed of the NS equation solver can be accelerated.
Description
Technical Field
The application relates to the technical field of computational fluid mechanics, in particular to a flow field data calculation method and device based on machine learning.
Background
The computational fluid mechanics can obtain abundant flow field data, and is an important means for researching the fluid mechanics. However, the high-precision NS equation solver has a low convergence rate and needs a large amount of computing resources. These disadvantages lead to longer research periods and higher economic costs, limiting engineering applications of computational fluid dynamics.
Generally, methods for accelerating numerical calculation of the NS equation can be classified into two methods. The first method is to shorten the calculation time of the calculation example by adopting more calculation resources, such as multithread parallel operation and GPU operation, but this method does not reduce the calculation amount for solving the NS equation, and even brings extra calculation amount. The second method is to reduce the amount of calculation by optimizing the solver algorithm, thereby shortening the calculation time, such as the multiple grid technique. But the development of the related optimization algorithm is slow and the universality is limited at present. Therefore, a method for reducing the amount of calculation of NS equation data is urgently needed.
Disclosure of Invention
In view of the above, it is necessary to provide a machine learning-based flow field data calculation method and apparatus capable of reducing the calculation amount for solving the NS equation.
A method of machine learning based flow field data computation, the method comprising:
generating first flow field data of a simulation object at a first time step and second flow field data of a simulation object at a second time step by using an NS equation solver; wherein the first time step is less than the second time step;
inputting the first flow field data into a preset neural network, and performing reverse training by using a difference value between an output value of the neural network and the second flow field data to determine a network structure parameter of the neural network;
when the flow field data is calculated, the flow field data to be calculated at the current time step is input into the neural network corresponding to the network structure parameter, and flow field correction data of the flow field data to be calculated is obtained;
and inputting the flow field correction data into the NS equation solver to obtain flow field data of the time step to be calculated corresponding to the flow field data to be calculated.
In one embodiment, the method further comprises the following steps: judging whether the residual error of the flow field data of the time step to be calculated corresponding to the flow field data to be calculated meets a preset convergence condition or not; and if so, taking the flow field data corresponding to the flow field data to be calculated as a flow field data calculation result.
In one embodiment, the method further comprises the following steps: if the residual error of the flow field data of the time step to be calculated corresponding to the flow field data to be calculated does not meet the preset convergence condition; judging whether the error of the flow field data to be calculated and the flow field data of the time step to be calculated corresponding to the flow field data to be calculated is smaller than the preset prediction error of the neural network or not; and if so, carrying out time iteration on the current time step, not using the prediction function of a neural network, and obtaining a flow field data calculation result by only using the calculation of an NS equation solver.
In one embodiment, the method further comprises the following steps: if the error of the flow field data to be calculated and the flow field data of the time step to be calculated corresponding to the flow field data to be calculated is larger than the preset prediction error of the neural network; and performing time iteration on the current time step, and inputting the flow field data to be calculated obtained by iteration into the neural network for correction.
A machine learning based flow field data calculation apparatus, the apparatus comprising:
the data generation module is used for generating first flow field data of the simulation object at a first time step and second flow field data of the simulation object at a second time step by using an NS equation solver; wherein the first time step is less than the second time step;
the model training module is used for inputting the first flow field data into a preset neural network, performing reverse training by using a difference value between an output value of the neural network and the second flow field data, and determining a network structure parameter of the neural network;
the data correction module is used for inputting the flow field data to be calculated of the current time step into the neural network corresponding to the network structure parameter when calculating the flow field data to obtain the flow field correction data of the flow field data to be calculated;
and the calculation module is used for inputting the flow field correction data into the NS equation solver to obtain the flow field data of the time step to be calculated corresponding to the flow field data to be calculated.
In one embodiment, the calculation module is further configured to determine whether a residual error of the flow field data of the time step to be calculated, which corresponds to the flow field data to be calculated, meets a preset convergence condition; and if so, taking the flow field data corresponding to the flow field data to be calculated as a flow field data calculation result.
In one embodiment, the calculation module is further configured to determine whether a residual error of the flow field data of the time step to be calculated, which corresponds to the flow field data to be calculated, does not satisfy a preset convergence condition; judging whether the error of the flow field data to be calculated and the flow field data of the time step to be calculated corresponding to the flow field data to be calculated is smaller than the preset prediction error of the neural network or not; and if so, carrying out time iteration on the current time step, and calculating by using the flow field data to be calculated obtained by iteration and an NS equation solver to obtain a flow field data calculation result.
In one embodiment, the calculation module is further configured to determine whether the error of the flow field data to be calculated and the flow field data of the time step to be calculated corresponding to the flow field data to be calculated is greater than a preset prediction error of the neural network; and performing time iteration on the current time step, and inputting the flow field data to be calculated obtained by iteration into the neural network for correction.
A computer device comprising a memory and a processor, the memory storing a computer program, the processor implementing the following steps when executing the computer program:
generating first flow field data of a simulation object at a first time step and second flow field data of a simulation object at a second time step by using an NS equation solver; wherein the first time step is less than the second time step;
inputting the first flow field data into a preset neural network, and performing reverse training by using a difference value between an output value of the neural network and the second flow field data to determine a network structure parameter of the neural network;
when the flow field data is calculated, the flow field data to be calculated at the current time step is input into the neural network corresponding to the network structure parameter, and flow field correction data of the flow field data to be calculated is obtained;
and inputting the flow field correction data into the NS equation solver to obtain flow field data of the time step to be calculated corresponding to the flow field data to be calculated.
A computer-readable storage medium, on which a computer program is stored which, when executed by a processor, carries out the steps of:
generating first flow field data of a simulation object at a first time step and second flow field data of a simulation object at a second time step by using an NS equation solver; wherein the first time step is less than the second time step;
inputting the first flow field data into a preset neural network, and performing reverse training by using a difference value between an output value of the neural network and the second flow field data to determine a network structure parameter of the neural network;
when the flow field data is calculated, the flow field data to be calculated at the current time step is input into the neural network corresponding to the network structure parameter, and flow field correction data of the flow field data to be calculated is obtained;
and inputting the flow field correction data into the NS equation solver to obtain flow field data of the time step to be calculated corresponding to the flow field data to be calculated.
According to the flow field data calculation method, the flow field data calculation device, the computer equipment and the storage medium based on machine learning, the neural network with the prediction function is trained, when the flow field data to be calculated is calculated in the NS equation solver, the flow field data to be calculated can be predicted in advance to obtain the flow field correction data, so that the correction data is used for solving in the NS equation solver, and the flow field correction data is close to the flow field data required to be calculated, so that the convergence process can be accelerated when the NS equation is used for solving, the calculated data amount is greatly reduced, the consumption of calculation resources is reduced, and the calculation time is shortened.
Drawings
FIG. 1 is a schematic flow chart diagram of a machine learning-based flow field data calculation method in one embodiment;
FIG. 2 is a schematic diagram of a neural network in one embodiment;
FIG. 3 is a velocity cloud plot for each time step in one embodiment;
FIG. 4 is a schematic representation of flow direction velocity residuals over time in one embodiment;
FIG. 5 is a block diagram of a device for calculating flow field data based on machine learning in one embodiment;
FIG. 6 is a diagram illustrating an internal structure of a computer device according to an embodiment.
Detailed Description
In order to make the objects, technical solutions and advantages of the present application more apparent, the present application is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application.
In one embodiment, as shown in fig. 1, there is provided a machine learning-based flow field data calculation method, including the following steps:
and 102, generating first flow field data of the simulation object at a first time step and second flow field data of the simulation object at a second time step by using an NS equation solver.
The simulation object refers to a part or a whole of an aircraft, or an automobile, a ship and the like, and flow field data needs to be calculated and verified in the research and development process of the aircraft and the like, generally speaking, the simulation object moves at a certain speed in a flow area or the flow area flows at a certain speed to generate a flow field, the flow area is composed of air, water and other fluids, the change data of the physical quantities (pressure, temperature and the like) of the surface and the peripheral flow field of the simulation object is obtained through calculation, the flow field simulation data problem in the flow area is accurately obtained, and the flow field data is calculated to simulate the change brought to the physical quantities (pressure, temperature and the like) of the surface and the peripheral flow field of the simulation object by the flow of the air, the water and the like, so as to provide reference data for the.
The NS equation solver may be a conventional NS equation solver, and is not limited herein. In addition, the first time step is smaller than the second time step, and the flow field data has a potential change rule along with time, namely, the change condition of the second flow field data can be predicted approximately through the first flow field data.
And 104, inputting the first flow field data into a preset neural network, performing reverse training by using the difference value of the output value of the neural network and the second flow field data, and determining the network structure parameters of the neural network.
The neural network may employ a conventional machine learning model, such as: fully connected neural networks, convolutional neural networks, etc., but not limited to the above two types of neural networks. Specifically, as shown in fig. 2, the neural network includes an input layer, a plurality of hidden layers, and an output layer. The number of neurons of the input layer is equal to the number of input parameters, the number of neurons of each hidden layer can be adjusted according to the prediction precision, and the number of neurons of the output layer is equal to the number of output parameters.
The difference can be calculated by constructing a loss function, so that the neural network model is reversely trained by the loss function.
It is worth mentioning that a large number of combinations of the first flow field data and the second flow field data need to be constructed to train the neural network, so that the network structure parameters of the neural network can be obtained.
And 106, when calculating the flow field data, inputting the flow field data to be calculated at the current time step into a neural network corresponding to the network structure parameters to obtain flow field correction data of the flow field data to be calculated.
In the step, when the real-time flow field data is calculated, the neural network is used for predicting the flow field data to be calculated.
And 108, inputting the flow field correction data into the NS equation solver to obtain flow field data of the time step to be calculated corresponding to the flow field data to be calculated.
In the flow field data calculation method based on machine learning, the neural network with the prediction function is trained, when the flow field data to be calculated is calculated in the NS equation solver, the flow field data to be calculated can be predicted in advance to obtain the flow field correction data, so that the correction data is used for solving in the NS equation solver.
In one embodiment, after the flow field data is obtained by calculation, it is further required to verify whether the calculation result meets the requirement, that is, determine whether the residual error meets the condition, specifically: and judging whether the residual error of the flow field data of the time step to be calculated corresponding to the flow field data to be calculated meets the preset convergence condition, if so, taking the flow field data corresponding to the flow field data to be calculated as the calculation result of the flow field data.
In this embodiment, due to the existence of the neural network, the time step of the computation iteration can be greatly reduced, thereby reducing the computation time.
In one embodiment, if the residual does not satisfy the preset convergence condition, the method specifically includes the following steps: if the residual error of the flow field data of the time step to be calculated corresponding to the flow field data to be calculated does not meet the preset convergence condition; judging whether the error of the flow field data to be calculated and the flow field data of the time step to be calculated corresponding to the flow field data to be calculated is smaller than the preset prediction error of the neural network or not; and if so, carrying out time iteration on the current time step, and calculating by using the flow field data to be calculated obtained by iteration and an NS equation solver to obtain a flow field data calculation result. In this embodiment, when the residual does not satisfy the convergence condition, time iteration is performed on the current time step, and the flow field data is calculated again by the NS equation solver.
Specifically, if the current time step is tnTime step, the time step to be calculated is tn+ITime step, if t isnInputting the flow field correction data of the time step into an NS equation solver, and calculating to obtain tn+IIf the residual error of the flow field data of the time step does not meet the convergence condition, the t is judgednIterate to t at time stepn+1Time step, then tn+1And inputting the flow field data corresponding to the time step into an NS equation solver for calculation. It is worth mentioning that I > 1 and n.gtoreq.0.
In one embodiment, if the error of the flow field data to be calculated and the flow field data of the to-be-calculated time step corresponding to the flow field data to be calculated is greater than the preset prediction error of the neural network, time iteration is performed on the current time step, and the flow field data to be calculated obtained by the iteration is input into the neural network for correction.
In this embodiment, if the error of the flow field data at the time step to be calculated is greater than the preset prediction error of the neural network, it is indicated that the flow field convergence can be accelerated by predicting the flow field by the neural network, and the flow field data to be calculated obtained through iteration is input to the neural network for correction.
Specifically, if the current time step is tnTime step, the time step to be calculated is tn+ITime step, if t isnInputting the flow field correction data of the time step into an NS equation solver, and calculating to obtain tn+IThe residual error of the flow field data at time step does not satisfy the convergence condition and tn+ITime step flow field data and tnIf the flow field data to be calculated at the time step is larger than the prediction error, t is calculatednIterate to t at time stepn+1Time step, then tn+1And inputting the flow field data corresponding to the time step into a neural network for correction. It is worth mentioning that I > 1 and n.gtoreq.0.
In conclusion, the neural network is embedded in the original flow field data calculation process, and the convergence calculation process of the flow field is controlled by setting the judgment condition, so that the convergence speed of the NS equation solver can be greatly accelerated.
The above-described embodiments of the present invention are described below with reference to specific examples, which are specifically as follows:
the flow field region is typically a backward step. The height of the inlet of the flow field area is 5.2mm, and the depth of the backward step is 4.9 mm. The lengths of the upstream flow channel and the downstream flow channel of the backward step are 5mm and 100mm respectively. The velocity profile of the fluid at the inlet of the computational domain is parabolic, with a maximum velocity of 1.459 m/s. The total number of grids is 20720. Calculations were performed using simpleFoam and neural network embedded simpleFoam (denoted MLsimpleFoam).
FIG. 3 is a velocity cloud plot for each time step in the calculation process. As can be easily seen from fig. 3, the flow field obtained by mlsimplefioam calculation at the same time step is closer to the final result, which not only accelerates the evolution of the flow field to the final result by the neural network. MLsimpleFoam reaches the preset residual (1e-5) at time step 780, while it takes 1515 time steps, about 19. times that of MLsimpleFoam, for simpleFoam to reach the same residual level. In addition, as can be seen from the velocity cloud charts at the 780 th time step and the 1515 th time step in fig. 3, the results of the two calculation schemes are the same, i.e., the accuracy of the final result is not impaired by the neural network. Fig. 4 is a graph of the flow direction velocity residual curve over time step and calculation time during the calculation process. Although neural networks may cause an increase in residual errors locally, the computation time of MLsimpleFoam (26.17 seconds) embedded with neural networks as a whole is approximately 40% shorter than the computation time of simpleFoam (43.26 seconds).
In addition, the MLsimpleFoam is tested by adopting the same neural network under different incoming flow speeds and different calculation domains, and the good effect is achieved. Under different calculation examples, MLsimpleFoam can shorten the calculation time by more than 30%. Therefore, the embodiment of the invention has better applicability.
It should be understood that, although the steps in the flowchart of fig. 1 are shown in order as indicated by the arrows, the steps are not necessarily performed in order as indicated by the arrows. The steps are not performed in the exact order shown and described, and may be performed in other orders, unless explicitly stated otherwise. Moreover, at least a portion of the steps in fig. 1 may include multiple sub-steps or multiple stages that are not necessarily performed at the same time, but may be performed at different times, and the order of performance of the sub-steps or stages is not necessarily sequential, but may be performed in turn or alternately with other steps or at least a portion of the sub-steps or stages of other steps.
In one embodiment, as shown in fig. 5, there is provided a machine learning-based flow field data calculation apparatus including: a data generation module 502, a model training module 504, a data modification module 506, and a calculation module 508, wherein:
a data generating module 502, configured to generate, by using an NS equation solver, first flow field data of the simulation object at a first time step and second flow field data of the simulation object at a second time step; wherein the first time step is less than the second time step;
a model training module 504, configured to input the first flow field data into a preset neural network, perform reverse training by using a difference between an output value of the neural network and the second flow field data, and determine a network structure parameter of the neural network;
a data correction module 506, configured to, during calculation of flow field data, input the flow field data to be calculated at the current time step into the neural network corresponding to the network structure parameter, so as to obtain flow field correction data of the flow field data to be calculated;
the calculating module 508 inputs the flow field correction data into the NS equation solver to obtain flow field data of a time step to be calculated corresponding to the flow field data to be calculated.
In one embodiment, the calculating module 508 is further configured to determine whether a residual error of the flow field data of the to-be-calculated time step corresponding to the flow field data to be calculated meets a preset convergence condition; and if so, taking the flow field data corresponding to the flow field data to be calculated as a flow field data calculation result.
In one embodiment, the calculating module 508 is further configured to determine whether a residual error of the flow field data of the time step to be calculated, which corresponds to the flow field data to be calculated, does not satisfy a preset convergence condition; judging whether the error of the flow field data to be calculated and the flow field data of the time step to be calculated corresponding to the flow field data to be calculated is smaller than the preset prediction error of the neural network or not; and if so, carrying out time iteration on the current time step, and calculating by using the flow field data to be calculated obtained by iteration and an NS equation solver to obtain a flow field data calculation result.
In one embodiment, the calculating module 508 is further configured to determine that an error of the flow field data to be calculated and the flow field data of the to-be-calculated time step corresponding to the flow field data to be calculated is greater than a preset prediction error of the neural network; and performing time iteration on the current time step, and inputting the flow field data to be calculated obtained by iteration into the neural network for correction.
For specific definition of the flow field data calculation device based on machine learning, reference may be made to the above definition of the flow field data calculation method based on machine learning, and details are not repeated here. The various modules in the machine learning based flow field data calculation apparatus may be implemented wholly or partially by software, hardware and a combination thereof. The modules can be embedded in a hardware form or independent from a processor in the computer device, and can also be stored in a memory in the computer device in a software form, so that the processor can call and execute operations corresponding to the modules.
In one embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as shown in fig. 6. The computer device includes a processor, a memory, a network interface, a display screen, and an input device connected by a system bus. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device comprises a nonvolatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of an operating system and computer programs in the non-volatile storage medium. The network interface of the computer device is used for communicating with an external terminal through a network connection. The computer program is executed by a processor to implement a machine learning-based flow field data calculation method. The display screen of the computer equipment can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer equipment can be a touch layer covered on the display screen, a key, a track ball or a touch pad arranged on the shell of the computer equipment, an external keyboard, a touch pad or a mouse and the like.
Those skilled in the art will appreciate that the architecture shown in fig. 6 is merely a block diagram of some of the structures associated with the disclosed aspects and is not intended to limit the computing devices to which the disclosed aspects apply, as particular computing devices may include more or less components than those shown, or may combine certain components, or have a different arrangement of components.
In an embodiment, a computer device is provided, comprising a memory storing a computer program and a processor implementing the steps of the method in the above embodiments when the processor executes the computer program.
In an embodiment, a computer-readable storage medium is provided, on which a computer program is stored, which computer program, when being executed by a processor, carries out the steps of the method in the above-mentioned embodiments.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by hardware instructions of a computer program, which can be stored in a non-volatile computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. Any reference to memory, storage, database, or other medium used in the embodiments provided herein may include non-volatile and/or volatile memory, among others. Non-volatile memory can include read-only memory (ROM), Programmable ROM (PROM), Electrically Programmable ROM (EPROM), Electrically Erasable Programmable ROM (EEPROM), or flash memory. Volatile memory can include Random Access Memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms such as Static RAM (SRAM), Dynamic RAM (DRAM), Synchronous DRAM (SDRAM), Double Data Rate SDRAM (DDRSDRAM), Enhanced SDRAM (ESDRAM), Synchronous Link DRAM (SLDRAM), Rambus Direct RAM (RDRAM), direct bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).
The technical features of the above embodiments can be arbitrarily combined, and for the sake of brevity, all possible combinations of the technical features in the above embodiments are not described, but should be considered as the scope of the present specification as long as there is no contradiction between the combinations of the technical features.
The above-mentioned embodiments only express several embodiments of the present application, and the description thereof is more specific and detailed, but not construed as limiting the scope of the invention. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the concept of the present application, which falls within the scope of protection of the present application. Therefore, the protection scope of the present patent shall be subject to the appended claims.
Claims (10)
1. A method of machine learning based flow field data computation, the method comprising:
generating first flow field data of a simulation object at a first time step and second flow field data of a simulation object at a second time step by using an NS equation solver; wherein the first time step is less than the second time step;
inputting the first flow field data into a preset neural network, and performing reverse training by using a difference value between an output value of the neural network and the second flow field data to determine a network structure parameter of the neural network;
when the flow field data is calculated, the flow field data to be calculated at the current time step is input into the neural network corresponding to the network structure parameter, and flow field correction data of the flow field data to be calculated is obtained;
and inputting the flow field correction data into the NS equation solver to obtain flow field data of the time step to be calculated corresponding to the flow field data to be calculated.
2. The method of claim 1, wherein after inputting the flow field correction data into the NS equation solver to obtain flow field data of a time step to be calculated corresponding to the flow field data to be calculated, the method further comprises:
judging whether the residual error of the flow field data of the time step to be calculated corresponding to the flow field data to be calculated meets a preset convergence condition or not;
and if so, taking the flow field data corresponding to the flow field data to be calculated as a flow field data calculation result.
3. The method of claim 2, further comprising:
if the residual error of the flow field data of the time step to be calculated corresponding to the flow field data to be calculated does not meet the preset convergence condition;
judging whether the error of the flow field data to be calculated and the flow field data of the time step to be calculated corresponding to the flow field data to be calculated is smaller than the preset prediction error of the neural network or not;
and if so, carrying out time iteration on the current time step, and calculating by using the flow field data to be calculated obtained by iteration and an NS equation solver to obtain a flow field data calculation result.
4. The method of claim 3, further comprising:
if the error of the flow field data to be calculated and the flow field data of the time step to be calculated corresponding to the flow field data to be calculated is larger than the preset prediction error of the neural network;
and performing time iteration on the current time step, and inputting the flow field data to be calculated obtained by iteration into the neural network for correction.
5. An apparatus for computing flow field data based on machine learning, the apparatus comprising:
the data generation module is used for generating first flow field data of the simulation object at a first time step and second flow field data of the simulation object at a second time step by using an NS equation solver; wherein the first time step is less than the second time step;
the model training module is used for inputting the first flow field data into a preset neural network, performing reverse training by using a difference value between an output value of the neural network and the second flow field data, and determining a network structure parameter of the neural network;
the data correction module is used for inputting the flow field data to be calculated of the current time step into the neural network corresponding to the network structure parameter when calculating the flow field data to obtain the flow field correction data of the flow field data to be calculated;
and the calculation module is used for inputting the flow field correction data into the NS equation solver to obtain the flow field data of the time step to be calculated corresponding to the flow field data to be calculated.
6. The device according to claim 5, wherein the calculation module is further configured to determine whether a residual error of the flow field data of the time step to be calculated, corresponding to the flow field data to be calculated, meets a preset convergence condition; and if so, taking the flow field data corresponding to the flow field data to be calculated as a flow field data calculation result.
7. The device according to claim 6, wherein the calculation module is further configured to determine whether the residual error of the flow field data at the time step to be calculated corresponding to the flow field data to be calculated does not satisfy a preset convergence condition; judging whether the error of the flow field data to be calculated and the flow field data of the time step to be calculated corresponding to the flow field data to be calculated is smaller than the preset prediction error of the neural network or not; and if so, carrying out time iteration on the current time step, and calculating by using the flow field data to be calculated obtained by iteration and an NS equation solver to obtain a flow field data calculation result.
8. The apparatus according to claim 7, wherein the calculation module is further configured to, if the error of the flow field data to be calculated and the flow field data of the time step to be calculated corresponding to the flow field data to be calculated is greater than a preset prediction error of the neural network; and performing time iteration on the current time step, and inputting the flow field data to be calculated obtained by iteration into the neural network for correction.
9. A computer device comprising a memory and a processor, the memory storing a computer program, wherein the processor implements the steps of the method of any one of claims 1 to 4 when executing the computer program.
10. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the steps of the method of any one of claims 1 to 4.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010645910.0A CN111651904B (en) | 2020-07-07 | Flow field data calculation method and device based on machine learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010645910.0A CN111651904B (en) | 2020-07-07 | Flow field data calculation method and device based on machine learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111651904A true CN111651904A (en) | 2020-09-11 |
| CN111651904B CN111651904B (en) | 2024-06-07 |
Family
ID=
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113158586A (en) * | 2021-05-26 | 2021-07-23 | 华能新能源股份有限公司 | Wind power plant numerical simulation method and device combined with neural network algorithm and storage medium |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104636566A (en) * | 2015-03-13 | 2015-05-20 | 中国矿业大学 | Dynamic mesh numerical solution method based on modified fluid motion equation |
| US20170032068A1 (en) * | 2015-07-31 | 2017-02-02 | Autodesk, Inc. | Techniques for warm starting finite element analyses with deep neural networks |
| CN111222241A (en) * | 2020-01-06 | 2020-06-02 | 中国人民解放军国防科技大学 | Numerical calculation method and device for flow field data under thermochemical unbalanced condition |
| CN111222240A (en) * | 2020-01-06 | 2020-06-02 | 中国人民解放军国防科技大学 | Thermochemical unbalanced flow field data calculation method and device accelerated by GPU |
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104636566A (en) * | 2015-03-13 | 2015-05-20 | 中国矿业大学 | Dynamic mesh numerical solution method based on modified fluid motion equation |
| US20170032068A1 (en) * | 2015-07-31 | 2017-02-02 | Autodesk, Inc. | Techniques for warm starting finite element analyses with deep neural networks |
| CN111222241A (en) * | 2020-01-06 | 2020-06-02 | 中国人民解放军国防科技大学 | Numerical calculation method and device for flow field data under thermochemical unbalanced condition |
| CN111222240A (en) * | 2020-01-06 | 2020-06-02 | 中国人民解放军国防科技大学 | Thermochemical unbalanced flow field data calculation method and device accelerated by GPU |
Non-Patent Citations (2)
| Title |
|---|
| 党冠麟;刘世伟;胡晓东;张鉴;李新亮;: "基于CPU/GPU异构系统架构的高超声速湍流直接数值模拟研究", 数据与计算发展前沿, no. 01, 15 February 2020 (2020-02-15) * |
| 樊鹏飞;宋文萍;: "NS方程中耦合转捩判断的风力机翼型气动力数值模拟", 航空计算技术, no. 03, 15 May 2010 (2010-05-15) * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113158586A (en) * | 2021-05-26 | 2021-07-23 | 华能新能源股份有限公司 | Wind power plant numerical simulation method and device combined with neural network algorithm and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111126668B (en) | Spark operation time prediction method and device based on graph convolution network | |
| US11574095B2 (en) | Simulation system for semiconductor process and simulation method thereof | |
| Bao et al. | Adaptive subset searching-based deep neural network method for structural reliability analysis | |
| CN113821878B (en) | Calculation method and device for improving hypersonic aerodynamic heat flow distribution abnormality | |
| CN113987691A (en) | High-precision hybrid calculation method, device, equipment and storage medium for shock wave instability | |
| CN115221801B (en) | Aircraft uncertainty propagation analysis method and device based on dynamic approximate modeling | |
| CN110990135A (en) | Spark operation time prediction method and device based on deep migration learning | |
| CN115049176A (en) | Material performance evaluation method and device and computer equipment | |
| CN113821877B (en) | Hypersonic aerodynamic thermal numerical simulation-oriented regional decomposition calculation method and hypersonic aerodynamic thermal numerical simulation-oriented regional decomposition calculation device | |
| CN114999182A (en) | Vehicle flow prediction method, device and equipment based on LSTM feedback mechanism | |
| CN111651904B (en) | Flow field data calculation method and device based on machine learning | |
| CN111930491B (en) | Global communication optimization acceleration method and device and computer equipment | |
| CN113537591A (en) | Long-term weather prediction method and device, computer equipment and storage medium | |
| CN111651904A (en) | Flow field data calculation method and device based on machine learning | |
| CN109871603B (en) | Gas flow field prediction method for complex emission technical conditions | |
| CN113987803B (en) | Frequency control method, device, equipment and storage medium based on model correction | |
| CN114722581A (en) | Mobile state monitoring method and device based on Manhattan distance and computer equipment | |
| CN115438590A (en) | Rainfall forecast correction method and system based on BP neural network | |
| CN111125809A (en) | Building steel column generation method and device, computer equipment and storage medium | |
| CN111210105B (en) | Continuous operation performance verification, apparatus, computer device, and storage medium | |
| CN114118425A (en) | Neural network accelerated reasoning method and system based on dynamic sparse convolution | |
| CN110865882B (en) | Data processing method, data processing device, computer equipment and storage medium | |
| CN106529086A (en) | Static load intensity finite element simulation precision determination method used for mechanical structural part | |
| US20200175121A1 (en) | System and method for predicting analytical abnormality in computational fluid dynamics analysis | |
| CN115630559B (en) | Fluid-solid coupling method and device based on particle grid adaptation algorithm |
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 |