A flow field detection method, device, equipment and medium

Abstract

The present invention relates to the field of flow field analysis. The present invention discloses a flow field detection method, device, equipment and medium, including obtaining experimental original data from an atmospheric turbulence simulation device; establishing a mathematical simulation model according to the experimental original data, and obtaining simulation data from the mathematical simulation model; performing data verification on the simulation data and the experimental original data to obtain training data; establishing a deep learning model; training the deep learning model according to the training data to obtain an optimized deep learning model; performing inversion calculation on the internal flow field of the atmospheric turbulence simulation device by optimizing the deep learning model to obtain inversion data; thanks to the fast inversion capability of the deep learning model, the calculation cycle of the mathematical simulation model is significantly shortened, and the simulation efficiency is improved, so that the experimenter can perform real-time regulation and control of the atmospheric turbulence simulation device based on the inversion data, respond to the experimental needs in real time, and improve the accuracy and efficiency of the experiment.

Description

Technical Field

[0001] The present invention relates to the field of flow field analysis, and in particular to a flow field detection method, device, equipment and medium. Background Art

[0002] As light travels through the atmosphere, it is significantly affected by atmospheric turbulence. In order to deeply analyze this phenomenon, the atmospheric turbulence simulation device has become an important quantitative analysis tool. However, the precise measurement of the physical field inside the device faces many challenges: not only is the measurement difficult and costly, but the traditional sensor arrangement can only capture a limited and observable data sample, making it difficult to fully reveal the true situation of the internal flow field and temperature field. Under certain extreme conditions, direct measurement even becomes unfeasible, prompting researchers to turn to building mathematical simulation models to deeply explore the flow characteristics inside the device.

[0003] Despite this, the construction and application of mathematical simulation models is not easy. It requires huge computing resources, and as the scale of the simulation device expands, the amount of calculation increases exponentially, resulting in low simulation efficiency and difficulty matching real-time needs. This means that the device cannot be regulated based on the simulation results, which affects the accuracy and efficiency of the experiment. Summary of the invention

[0004] The present invention aims to improve at least one technical problem in the background technology.

[0005] The first aspect of the present invention provides a flow field detection method, comprising the following steps: acquiring experimental raw data from an atmospheric turbulence simulation device; establishing a mathematical simulation model based on the experimental raw data, and obtaining simulation data from the mathematical simulation model; performing data verification on the simulation data and the experimental raw data to obtain training data; establishing a deep learning model; training the deep learning model based on the training data to obtain an optimized deep learning model; and inverting the internal flow field of the atmospheric turbulence simulation device by optimizing the deep learning model to obtain inversion data.

[0006] Optionally, in a first implementation method of the first aspect of the present invention, the obtaining of experimental raw data from an atmospheric turbulence simulation device includes: the atmospheric turbulence simulation device includes a plurality of sensors for detecting the inside of the device; obtaining basic data of each sensor from the atmospheric turbulence simulation device, and performing data cleaning and noise reduction on the basic data to obtain experimental observation data; obtaining experimental input data from the atmospheric turbulence simulation device; and merging the experimental observation data and the experimental input data to obtain experimental raw data.

[0007] Optionally, in a second implementation method of the first aspect of the present invention, establishing a mathematical simulation model based on the experimental original data and obtaining simulation data from the mathematical simulation model includes: performing data analysis on the experimental original data to obtain simulation model type information; performing modeling based on the simulation model type information and the experimental original data to obtain a mathematical simulation model; and using the mathematical simulation model to mathematically simulate the internal flow field of an atmospheric turbulence simulation device to obtain simulation data.

[0008] Optionally, in a third implementation method of the first aspect of the present invention, the data verification is performed based on the simulation data and the experimental original data to obtain training data, including: performing data verification on the simulation data and the experimental original data to obtain a simulation error; adjusting the mathematical simulation model according to the simulation error to obtain an optimized mathematical simulation model; performing mathematical simulation on the internal flow field of the atmospheric turbulence simulation device based on the optimized mathematical simulation model to obtain optimized simulation data; and merging the optimized simulation data and the experimental original data to obtain training data.

[0009] Optionally, in a fourth implementation method of the first aspect of the present invention, establishing a deep learning model includes: determining the type of deep learning model from a preset model library based on the experimental raw data to obtain a basic model; obtaining deep learning configuration information from the experimental raw data, and setting the network parameters of the basic model based on the deep learning configuration information to obtain a configuration model; obtaining boundary condition constraints and physical equation constraints from the experimental raw data; and imposing constraints on the configuration model based on the boundary condition constraints and physical equation constraints to obtain a deep learning model.

[0010] Optionally, in a fifth implementation manner of the first aspect of the present invention, the deep learning model is trained according to the training data to obtain an optimized deep learning model, including: extracting a training set and a validation set from the training data; training the deep learning model according to the training set to obtain multiple candidate models; evaluating the multiple candidate models through the validation set to obtain evaluation results, and obtaining an optimized deep learning model according to the evaluation results.

[0011] Optionally, in a sixth implementation method of the first aspect of the present invention, the inversion calculation of the internal flow field of the atmospheric turbulence simulation device is performed by optimizing the deep learning model to obtain inversion data, including: obtaining inversion input data from the atmospheric turbulence simulation device; inputting the inversion input data into the optimized deep learning model, and using the optimized deep learning model to perform inversion calculation on the internal flow field of the atmospheric turbulence simulation device to obtain inversion data.

[0012] The second aspect of the present invention provides a flow field detection device, including: an acquisition module, used to acquire experimental raw data from an atmospheric turbulence simulation device; a simulation module, used to establish a mathematical simulation model based on the experimental raw data, and obtain simulation data from the mathematical simulation model; a verification module, used to perform data verification on the simulation data and the experimental raw data to obtain training data; an establishment module, used to establish a deep learning model; a training module, used to train the deep learning model based on the training data to obtain an optimized deep learning model; an inversion module, used to perform inversion calculations on the internal flow field of the atmospheric turbulence simulation device by optimizing the deep learning model to obtain inversion data.

[0013] Optionally, in a first implementation method of the second aspect of the present invention, the acquisition module includes: a first acquisition unit, used to acquire basic data of each sensor from an atmospheric turbulence simulation device, and perform data cleaning and noise reduction on the basic data to obtain experimental observation data; a second acquisition unit, used to acquire experimental input data from the atmospheric turbulence simulation device; a first merging unit, used to merge the experimental observation data and the experimental input data to obtain experimental original data.

[0014] Optionally, in a second implementation method of the second aspect of the present invention, the simulation module includes: an analysis unit, used to perform data analysis on the experimental raw data to obtain simulation model type information; a first modeling unit, used to perform modeling based on the simulation model type information and the experimental raw data to obtain a mathematical simulation model; a first simulation unit, used to use the mathematical simulation model to perform mathematical simulation on the internal flow field of the atmospheric turbulence simulation device to obtain simulation data.

[0015] Optionally, in a third implementation method of the second aspect of the present invention, the verification module includes: a verification unit, which is used to perform data verification on simulation data and experimental original data to obtain a simulation error; an optimization unit, which is used to adjust the mathematical simulation model according to the simulation error to obtain an optimized mathematical simulation model; a second simulation unit, which performs mathematical simulation on the internal flow field of the atmospheric turbulence simulation device based on the optimized mathematical simulation model to obtain optimized simulation data; and a second merging unit, which is used to merge the optimized simulation data and the experimental original data to obtain training data.

[0016] Optionally, in a fourth implementation method of the second aspect of the present invention, the establishment module includes: a second modeling unit, used to determine the type of deep learning model from a preset model library according to the experimental original data to obtain a basic model; a configuration unit, used to obtain deep learning configuration information from the experimental original data, and set the network parameters of the basic model according to the deep learning configuration information to obtain a configuration model; a third acquisition unit, used to obtain boundary condition constraints and physical equation constraints from the experimental original data; a constraint unit, used to impose constraints on the configuration model according to the boundary condition constraints and physical equation constraints to obtain a deep learning model.

[0017] Optionally, in a fifth implementation of the second aspect of the present invention, the training module includes: an extraction unit, used to extract a training set and a verification set from the training data; a training unit, used to train the deep learning model according to the training set to obtain multiple candidate models; an evaluation unit, used to evaluate the multiple candidate models through the verification set to obtain evaluation results, and obtain an optimized deep learning model according to the evaluation results.

[0018] Optionally, in a fifth implementation of the second aspect of the present invention, the inversion module includes: a fourth acquisition unit, used to acquire inversion input data from an atmospheric turbulence simulation device; an inversion unit, used to input the inversion input data into an optimized deep learning model, and use the optimized deep learning model to perform inversion calculations on the internal flow field of the atmospheric turbulence simulation device to obtain inversion data.

[0019] The third aspect of the present invention provides a flow field detection device, which includes: a memory and at least one processor, wherein instructions are stored in the memory; at least one of the processors calls the instructions in the memory so that the computer device executes each step of the flow field detection method described in any one of the above items.

[0020] A fourth aspect of the present invention provides a computer-readable storage medium, on which instructions are stored, and when the instructions are executed by a processor, each step of any of the above-mentioned flow field detection methods is implemented.

[0021] In the technical solution of the present invention, a mathematical simulation model constructed by directly utilizing the original experimental data is used to provide high-quality simulation data for the deep learning model, thereby training an efficient deep learning model, and then using the deep learning model to carry out flow field inversion calculations to obtain inversion data; thanks to the fast inversion capability of the deep learning model, the calculation cycle of the mathematical simulation model is significantly shortened, and the simulation efficiency is improved, so that experimenters can perform real-time regulation of the atmospheric turbulence simulation device based on the inversion data, respond to experimental needs in real time, and improve the accuracy and efficiency of the experiment. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The above and / or additional aspects and advantages of the present invention will become apparent and easily understood from the description of the embodiments in conjunction with the following drawings, in which:

[0023] Figure 1 A first flow chart of a flow field detection method provided by an embodiment of the present invention;

[0024] Figure 2 A second flow chart of the flow field detection method provided by an embodiment of the present invention;

[0025] Figure 3 A third flow chart of the flow field detection method provided by an embodiment of the present invention;

[0026] Figure 4 A fourth flow chart of the flow field detection method provided by an embodiment of the present invention;

[0027] Figure 5 A fifth flow chart of the flow field detection method provided by an embodiment of the present invention;

[0028] Figure 6 A sixth flow chart of the flow field detection method provided by an embodiment of the present invention;

[0029] Figure 7 A seventh flow chart of the flow field detection method provided by an embodiment of the present invention;

[0030] Figure 8 A schematic diagram of the structure of a flow field detection device provided in an embodiment of the present invention;

[0031] Fig. 9 Another structural schematic diagram of a flow field detection device provided by an embodiment of the present invention;

[0032] Fig.10 A schematic diagram of the structure of a flow field detection device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0033] The present invention provides a flow field detection method, device, equipment and medium. The present invention directly uses a mathematical simulation model constructed by raw experimental data to provide high-quality simulation data for a deep learning model, thereby training an efficient deep learning model, and then using the deep learning model to perform flow field inversion calculations to obtain inversion data. Thanks to the fast inversion capability of the deep learning model, the calculation cycle of the mathematical simulation model is significantly shortened, and the simulation efficiency is improved, so that experimenters can perform real-time regulation of the atmospheric turbulence simulation device based on the inversion data, respond to experimental needs in real time, and improve the accuracy and efficiency of the experiment.

[0034] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and cannot be understood as limiting the present invention.

[0035] The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0036] For ease of understanding, the specific process of the embodiment of the present invention is described below. Figure 1 , an embodiment of the flow field detection method in the embodiment of the present invention includes:

[0037] 101. Obtaining raw experimental data from the atmospheric turbulence simulation device;

[0038] In this embodiment, the device input parameters are preset, the atmospheric turbulence simulation device is configured according to the device input parameters, and the atmospheric turbulence simulation device is started to simulate the atmospheric turbulence; the atmospheric turbulence simulation device is provided with an internal sensor, and the observation data of the internal flow field of the atmospheric turbulence simulation device can be obtained from the internal sensor; the observation data includes the temperature data, pressure data and optical data of the flow field; the device input parameters include the temperature, size, and emission parameters of the laser of the device. After collecting the observation data and the device input parameters, the experimental observation data and the experimental input data are integrated to form the experimental raw data.

[0039] 102. Establish a mathematical simulation model based on the original experimental data, and obtain simulation data from the mathematical simulation model;

[0040] In this embodiment, in order to analyze the flow field inside the atmospheric turbulence simulation device, a mathematical simulation model needs to be established; based on the physical characteristics of atmospheric turbulence and the original experimental data, a suitable simulation model type is analyzed, such as a large eddy simulation (LES) model or a Navier-Stokes (RANS) model; then, the boundary conditions of the mathematical simulation model are set according to the original experimental data to establish the mathematical simulation model. The internal flow field of the atmospheric turbulence simulation device is mathematically simulated by the mathematical simulation model to obtain simulation data.

[0041] 103. Perform data verification on the simulation data and the original experimental data to obtain training data;

[0042] In this embodiment, the simulation data is compared and analyzed with the original experimental data, and the simulation data is verified and adjusted according to the results of the comparative analysis. If there are significant differences between the simulation data and the experimental data, the mathematical simulation model is adjusted and optimized, and the difference between the simulation data and the experimental data is gradually reduced through repeated iterations and verifications, the accuracy of the simulation data and the performance of the mathematical simulation model are improved, and finally the optimized mathematical simulation model and simulation data are obtained; the input parameters and output parameters of the optimized mathematical simulation model are integrated to obtain training data for training the deep learning model.

[0043] 104. Build a deep learning model;

[0044] In this embodiment, according to the characteristics of the original experimental data, a suitable deep learning model type is selected from a preset model library, such as a deep neural network (DNN), a convolutional neural network (CNN), a recursive neural network (RNN), a long short-term memory network (LSTM), etc., to construct a basic model; boundary condition constraints and physical equation constraints are obtained according to the original experimental data; the boundary condition constraints include data such as the size, internal temperature, and internal pressure of the atmospheric turbulence simulation device; the physical equation constraints include fluid mechanics equations, such as the Navier-Stokes equations; the boundary condition constraints and the physical equation constraints are added to the basic model to complete the establishment of the deep learning model.

[0045] 105. Train the deep learning model according to the training data to obtain an optimized deep learning model;

[0046] In this embodiment, a training set and a validation set are extracted from the training data, and the deep learning model is trained using the training set to generate multiple candidate models. Then, the performance of the candidate models is evaluated using the validation set, and based on the evaluation results, the model with the best performance is selected as the baseline model. The baseline model is further trained and optimized, model parameters are adjusted, training strategies are improved, or the amount of data is increased, and finally an optimized deep learning model is obtained.

[0047] 106. The internal flow field of the atmospheric turbulence simulation device is inverted by optimizing the deep learning model to obtain inversion data;

[0048] In this embodiment, input data and output data are obtained from the atmospheric turbulence simulation device, and these data are input into the optimized deep learning model; because the optimized deep learning model has mastered the complex characteristics of the internal flow field of the atmospheric turbulence simulation device through training, the optimized deep learning model can reconstruct the internal flow field of the atmospheric turbulence simulation device and invert and calculate the inversion data; the inversion data is used to reflect the characteristics of the flow field; the atmospheric turbulence simulation device can be controlled in real time based on the inversion data; at the same time, considering the limitation of the number of sensors inside the atmospheric simulation device, the data that can be measured experimentally has limitations and may not reflect the actual situation of the internal flow field of the atmospheric simulation device; and through the inversion calculation of the internal flow field of the atmospheric simulation device by the deep learning model, more comprehensive inversion data can be output, which can make up for the shortcomings of the experimental data and provide a more accurate and reliable basis for the experiment.

[0049] In an embodiment of the present invention, the atmospheric turbulence simulation device is configured and started by presetting the device input parameters, and the temperature, pressure and laser optical data of the flow field are collected by using internal sensors, which are integrated with the device input parameters to form experimental raw data; based on the experimental raw data, a suitable mathematical simulation model is established, and the optimization model is adjusted by comparing and analyzing the simulation data with the experimental raw data until the simulation data accurately reflects the experimental situation to form training data; according to the characteristics of the experimental data, a suitable deep learning model type is selected, boundary condition constraints and physical equation constraints are added, and the model is trained and optimized using the training data to obtain an optimized deep learning model; the input and output data of the atmospheric turbulence simulation device are input into the optimized deep learning model, and the internal flow field is inverted and calculated to obtain inversion data reflecting the flow field characteristics; thanks to the fast inversion capability of the deep learning model, the calculation cycle of the mathematical simulation model is significantly shortened, and the simulation efficiency is improved, so that the experimenter can perform real-time regulation and control of the atmospheric turbulence simulation device based on the inversion data, respond to experimental needs in real time, and improve the accuracy and efficiency of the experiment.

[0050] See also Figure 2 , a second embodiment of the flow field detection method in the embodiment of the present invention includes:

[0051] 201. The atmospheric turbulence simulation device includes a plurality of sensors for detecting the interior of the device;

[0052] In this embodiment, the sensors inside the device include but are not limited to a temperature sensor for acquiring temperature data, an optical camera for acquiring optical data, a wind speed sensor for measuring the air flow speed and direction inside the simulation device, and a pressure sensor for measuring the air pressure inside the simulation device.

[0053] 202. Obtain the basic data of each sensor from the atmospheric turbulence simulation device, and perform data cleaning and noise reduction on the basic data to obtain experimental observation data;

[0054] In this embodiment, basic data of each sensor are obtained from the atmospheric turbulence simulation device, and these data include basic temperature data, basic pressure data and basic optical data; the basic temperature data reflects the temperature distribution at different positions inside the simulation device, and is used to establish a temperature field; the basic pressure data reflects the pressure distribution at different positions inside the simulation device, and is used to establish a pressure field; the basic optical data include spot centroid fluctuation parameters and light intensity fluctuation parameters, which are used to evaluate the influence of atmospheric turbulence on laser transmission characteristics; these data are then cleaned to remove abnormal values, duplicate values ​​and invalid data in the data to ensure the accuracy and reliability of the data; the noise interference in the data is then reduced to improve the signal-to-noise ratio of the data; after data cleaning and noise reduction processing, experimental observation data are obtained.

[0055] 203. Obtain experimental input data from the atmospheric turbulence simulation device;

[0056] In this embodiment, the experimental input parameters include but are not limited to: the temperature of the top plate and the bottom plate, laser parameters, and measurement interval time; the temperature of the top plate and the bottom plate is used to control the size of the temperature gradient in the atmospheric turbulence simulation device, thereby controlling the average temperature in the cabin; in this example, the central part of the atmospheric turbulence simulation device is perpendicular to the line connecting the top plate and the bottom plate. There is a through hole for transmitting and receiving the laser for testing, and by setting different laser parameters, different atmospheric turbulence coherence lengths are simulated accordingly; the measurement interval time determines the frequency of data acquisition, and a faster acquisition frequency can capture the changes in atmospheric turbulence more dynamically; the experimental input parameters are used to configure the operating parameters of the atmospheric turbulence simulation device.

[0057] 204. Merge the experimental observation data and the experimental input data to obtain the original experimental data;

[0058] In this embodiment, the experimental observation data and the experimental input data are first aligned to ensure that the experimental observation data and the experimental input data are consistent in time and space, that is, they correspond to data in the same time period and under the same experimental conditions; then the experimental observation data and the experimental input data are merged according to the corresponding fields to form an experimental original data containing all necessary information.

[0059] See also Figure 3 , a third embodiment of the flow field detection method in the embodiment of the present invention includes:

[0060] 301. Analyze the original experimental data to obtain simulation model type information;

[0061] In this embodiment, based on the original experimental data, data analysis can be performed on the internal flow field of the atmospheric turbulence device to analyze the experimental state of the laser transmitting at different powers and frequencies under different boundary conditions, and the simulation model type information is obtained based on the results of the data analysis. The simulation model type information can be a large eddy simulation (LES) model or a Navier-Stokes (RANS) model; if the data analysis shows that the turbulence scale varies greatly in the experiment and it is necessary to capture the detailed vortex structure, then the large eddy simulation (LES) model in computational fluid dynamics (CFD) is preferred; if the data analysis shows that the experiment mainly focuses on the average flow characteristics and the computing resources are limited, then the Reynolds-averaged Navier-Stokes (RANS) model may be more suitable.

[0062] 302. Modeling is performed based on the simulation model type information and the experimental original data to obtain a mathematical simulation model;

[0063] In this embodiment, after selecting the simulation model type information, first, according to the geometric dimensions of the experimental device, an accurate three-dimensional model is drawn in the software; then, the mathematical simulation model input parameters are set according to the original experimental data; the mathematical simulation model input parameters include boundary conditions and grid parameters; the boundary conditions include wind speed, temperature, wall conditions, etc.; the three-dimensional model is divided into multiple calculation grids by a grid division tool to obtain grid parameters; when the simulation model type information is a LES model, it is also necessary to set a filter function and a sub-grid stress model to simulate the influence of small-scale vortices; the mathematical simulation model input parameter three-dimensional model is input to complete the establishment of the mathematical simulation model.

[0064] 303. Use a mathematical simulation model to perform mathematical simulation on the internal flow field of the atmospheric turbulence simulation device to obtain simulation data;

[0065] In this embodiment, after the mathematical simulation model is constructed, the simulation calculation is started. During the calculation process, the simulation model will solve the Navier-Stokes equations to simulate the complex pressure distribution and temperature distribution inside the device; because the refractive index of the visible light wavelength is a function of temperature and atmospheric pressure, the refractive index distribution field of the air at different positions in the simulation device can be calculated through the temperature distribution and pressure distribution, and then the atmospheric refraction structure constant and coherence length data inside the device can be calculated through the refractive index distribution field; the atmospheric refraction structure constant and coherence length data are output as simulation data to the mathematical simulation model.

[0066] See also Figure 4 , a fourth embodiment of the flow field detection method in the embodiment of the present invention includes:

[0067] 401. Perform data verification on the simulation data and the original experimental data to obtain a simulation error;

[0068] In this embodiment, in order to evaluate the accuracy of the simulation data, it is necessary to verify the simulation data. First, a series of key experimental measurement points are selected, and these points should be able to fully reflect the main characteristics of the internal flow field of the atmospheric turbulence simulation device. Then, the experimental original data and the simulation data at these measurement points are compared. In order to quantify the error between the simulation data and the experimental data, the error calculation method can select the root mean square error (RMSE), mean absolute error (MAE) and relative error, etc. By calculating these error indicators, the simulation error is finally obtained.

[0069] 402. Adjust the mathematical simulation model according to the simulation error to obtain an optimized mathematical simulation model;

[0070] In this embodiment, according to the distribution of simulation errors, the input parameters of the mathematical simulation model such as boundary condition parameters and grid resolution are adjusted to improve the mathematical simulation model. After these adjustments, we re-run the mathematical simulation model and compare and verify it with the original experimental data again until the simulation error is reduced to an acceptable level, thereby obtaining an optimized mathematical simulation model.

[0071] 403. Perform mathematical simulation on the internal flow field of the atmospheric turbulence simulation device based on the optimized mathematical simulation model to obtain optimized simulation data;

[0072] In this embodiment, the optimized mathematical simulation model is configured to perform mathematical simulation on the internal flow field of the atmospheric turbulence simulation device again to obtain simulation output data; the optimized mathematical simulation model input parameters are recorded; the simulation output data and the mathematical simulation model input parameters are merged to obtain optimized simulation data.

[0073] 404. Merging the optimized simulation data and the original experimental data to obtain training data;

[0074] In this embodiment, the optimization simulation data and the experimental original data are combined into a unified data set, and this data set will be used as the input data of the PINN model, that is, the training data.

[0075] See also Figure 5 , a fifth embodiment of the flow field detection method in the embodiment of the present invention includes:

[0076] 501. Determine the type of deep learning model from a preset model library according to the original experimental data to obtain a basic model;

[0077] In this embodiment, according to the characteristics of the experimental raw data, a suitable deep learning model type is determined from a preset model library; the deep learning model types include deep neural networks (DNN), convolutional neural networks (CNN), recursive neural networks (RNN), and long short-term memory networks (LSTM); deep neural networks are usually selected so that they can process complex physical function relationships; and the basic model is obtained according to the deep learning model type.

[0078] 502. Obtaining deep learning configuration information from the original experimental data, and setting network parameters of the basic model according to the deep learning configuration information to obtain a configuration model;

[0079] In this embodiment, deep learning configuration information is extracted from the experimental raw data, and the configuration information includes the number of network layers, the number of neurons in each layer, the activation function type, the optimizer type and its parameters; the network parameters of the basic model are set according to the configuration information to obtain the configuration model.

[0080] 503. Obtain boundary condition constraints and physical equation constraints from experimental raw data;

[0081] In this embodiment, the boundary condition constraints include the temperature of the top plate and the bottom plate, the laser parameters, and the size of the atmospheric turbulence simulation device. The boundary condition constraints are used to limit the boundary conditions of the physical field; the physical equation constraints are based on the physical equations of fluid dynamics, such as the Navier-Stokes equations. The physical equation constraints are used to impose constraints on deep learning that lack prior information on fluid dynamics.

[0082] 504. Constraining the configuration model according to boundary condition constraints and physical equation constraints to obtain a deep learning model;

[0083] According to the boundary condition constraints, the boundary condition loss term can be obtained. The boundary condition loss term is used to measure the performance of the deep learning model in the boundary area and improve the deep learning model's ability to recognize complex and fuzzy boundaries. According to the physical equation constraints, the physical information error term can be obtained. The physical information error term is used to measure whether the prediction results of the deep learning model satisfy the physical equation. The data fitting loss term and the physical information error term are combined into a first loss function. The configuration model training is guided according to the first loss function, so that the configuration model can deeply learn the physical field information that conforms to the data and satisfies the physical equation, and then obtain the deep learning model.

[0084] See also Figure 6 , a sixth embodiment of the flow field detection method in the embodiment of the present invention includes:

[0085] A training set and a validation set are extracted from the training data. In this embodiment, the training data is first divided into a training set and a validation set, and the divided training set and validation set have similar data distribution and characteristics. A common division ratio is that 80% of the training data is used to establish a training set, and the remaining 20% ​​is used to establish a validation set. The training set is used to train a deep learning model, and the validation set is used to evaluate the performance of the deep learning model in practical applications.

[0086] 602. Train the deep learning model according to the training set to obtain multiple candidate models;

[0087] In this embodiment, the deep learning model is trained using a training set; the second loss function and the third loss function are first extracted from the training set; the second loss function is used to measure the difference between the prediction results of the deep learning model and the optimized simulation data; the third loss function is used to measure the difference between the prediction results of the deep learning model and the original experimental data; the second loss function and the third loss function are used to guide the training of the deep learning model; hyperparameters such as the learning rate and batch size need to be adjusted during the training of the deep learning model; by changing the model architecture, hyperparameters or training methods, multiple candidate models are ultimately generated.

[0088] 603. Evaluate multiple candidate models through the validation set to obtain evaluation results, and obtain an optimized deep learning model according to the evaluation results;

[0089] In this embodiment, a validation set is used to evaluate each candidate model, and the performance indicators of the candidate model on the validation set are calculated, including but not limited to accuracy, recall rate, etc.; the performance indicators of different candidate models are compared, the advantages and disadvantages of the models are analyzed, and evaluation results are obtained; based on the evaluation results, the candidate model with the best performance is selected as the baseline model, the baseline model is further trained, and the network architecture, hyperparameters and training methods of the baseline model are adjusted to finally obtain an optimized deep learning model.

[0090] See also Figure 7 The seventh embodiment of the flow field detection method in the embodiment of the present invention includes:

[0091] 701. Obtain inversion input data from an atmospheric turbulence simulation device;

[0092] In this example, the input parameters of the atmospheric turbulence simulation device are preset first, and then the atmospheric turbulence simulation device is started, the output data of the atmospheric turbulence simulation device is obtained from the sensor in the atmospheric turbulence simulation device, and the input parameters of the atmospheric turbulence simulation device and the output data of the atmospheric turbulence simulation device are integrated to obtain the inversion input data.

[0093] 702. Input the inversion input data into the optimized deep learning model, and use the optimized deep learning model to perform inversion calculation on the internal flow field of the atmospheric turbulence simulation device to obtain inversion data;

[0094] In this example, the inversion input data is input into the optimized deep learning model, which will use the data to perform inversion calculations, and then reconstruct the flow field inside the atmospheric turbulence simulation device, obtain the characteristics of the flow field, and finally output the inversion data; the inversion data may include key information such as velocity distribution, pressure distribution, and temperature distribution of the flow field, and the inversion data can reflect the situation of the flow field inside the atmospheric turbulence simulation device; based on the inversion data, the experimenters can control the atmospheric turbulence simulation device in real time; at the same time, considering the limitation of the number of sensors inside the atmospheric simulation device, the data that can be measured in the experiment has limitations and may not reflect the actual situation of the flow field inside the atmospheric simulation device; and through the inversion calculation of the flow field inside the atmospheric simulation device by the deep learning model, more comprehensive inversion data can be output, which can make up for the shortcomings of the experimental data and provide a more accurate and reliable basis for the experiment.

[0095] The flow field detection method in the embodiment of the present invention is described above. The flow field detection device in the embodiment of the present invention is described below. Figure 8 , an embodiment of the flow field detection device in the embodiment of the present invention includes:

[0096] An acquisition module 801 is used to acquire raw experimental data from an atmospheric turbulence simulation device;

[0097] The simulation module 802 is used to establish a mathematical simulation model according to the original experimental data and obtain simulation data from the mathematical simulation model;

[0098] The verification module 803 is used to verify the simulation data and the original experimental data to obtain training data;

[0099] Establishing module 804, for establishing a deep learning model;

[0100] A training module 805 is used to train the deep learning model according to the training data to obtain an optimized deep learning model;

[0101] The inversion module 806 is used to perform inversion calculation on the internal flow field of the atmospheric turbulence simulation device by optimizing the deep learning model to obtain inversion data.

[0102] In this embodiment, the acquisition module 801 obtains the experimental original data from the atmospheric turbulence simulation device, and then the simulation module 802 is used to establish a mathematical simulation model based on the experimental original data, and the internal flow field of the atmospheric turbulence simulation device is mathematically simulated to output the simulation data; the verification module 803 is used to compare the simulation data and the experimental original data to verify the accuracy of the simulation data. The verification module outputs a simulation error according to the difference between the simulation data and the experimental original data, and the mathematical simulation model is adjusted according to the simulation error. The input and output of the adjusted mathematical simulation model are integrated to obtain training data; the establishment module 804 is used to establish a deep learning model, and the training module 805 trains the deep learning model with the training data to obtain an optimized deep learning model. Finally, the inversion module 806 inputs the inversion data into the optimized deep learning model, and the internal flow field of the atmospheric turbulence simulation device is inverted and calculated by the optimized deep learning model to obtain inversion data. The trained deep learning model can better simulate the internal flow field of the atmospheric turbulence simulation device, and thus obtain more realistic flow field data. The deep learning model has a faster computing speed than the mathematical simulation model. The inversion data obtained based on the deep learning model can match the internal flow field characteristics of the atmospheric turbulence simulation device in real time. Experimenters can adjust the atmospheric turbulence simulation device in real time based on the inversion data to obtain better experimental results.

[0103] See also Fig. 9 Another embodiment of the flow field detection device in the embodiment of the present invention includes:

[0104] An acquisition module 801 is used to acquire raw experimental data from an atmospheric turbulence simulation device;

[0105] The simulation module 802 is used to establish a mathematical simulation model according to the original experimental data and obtain simulation data from the mathematical simulation model;

[0106] The verification module 803 is used to verify the simulation data and the original experimental data to obtain training data;

[0107] Establishing module 804, for establishing a deep learning model;

[0108] A training module 805 is used to train the deep learning model according to the training data to obtain an optimized deep learning model;

[0109] An inversion module 806 is used to perform inversion calculation on the internal flow field of the atmospheric turbulence simulation device by optimizing the deep learning model to obtain inversion data;

[0110] In this embodiment, the acquisition module 801 includes: a first acquisition unit 8011, which is used to acquire basic data of each sensor from the atmospheric turbulence simulation device, and perform data cleaning and noise reduction on the basic data to obtain experimental observation data; a second acquisition unit 8012, which is used to acquire experimental input data from the atmospheric turbulence simulation device; a first merging unit 8013, which is used to merge the experimental observation data and the experimental input data to obtain experimental original data.

[0111] In this embodiment, the simulation module 802 includes: an analysis unit 8021, which is used to perform data analysis on the experimental original data to obtain simulation model type information; a first modeling unit 8022, which is used to perform modeling based on the simulation model type information and the experimental original data to obtain a mathematical simulation model; a first simulation unit 8023, which is used to use the mathematical simulation model to perform mathematical simulation on the internal flow field of the atmospheric turbulence simulation device to obtain simulation data.

[0112] In this embodiment, the verification module 803 includes: a verification unit 8031, which is used to perform data verification on simulation data and experimental original data to obtain a simulation error; an optimization unit 8032, which is used to adjust the mathematical simulation model according to the simulation error to obtain an optimized mathematical simulation model; a second simulation unit 8033, which performs mathematical simulation on the internal flow field of the atmospheric turbulence simulation device based on the optimized mathematical simulation model to obtain optimized simulation data; a second merging unit 8034, which is used to merge the optimized simulation data and the experimental original data to obtain training data.

[0113] In this embodiment, the establishment module 804 includes: a second modeling unit 8041, which is used to determine the type of deep learning model from a preset model library according to the experimental original data to obtain a basic model; a configuration unit 8042, which is used to obtain deep learning configuration information from the experimental original data, and set the network parameters of the basic model according to the deep learning configuration information to obtain a configuration model; a third acquisition unit 8043, which is used to obtain boundary condition constraints and physical equation constraints from the experimental original data; a constraint unit 8044, which is used to impose constraints on the configuration model according to the boundary condition constraints and the physical equation constraints to obtain a deep learning model.

[0114] In this embodiment, the training module 805 includes: an extraction unit 8051, which is used to extract a training set and a verification set from the training data; a training unit 8052, which is used to train the deep learning model according to the training set to obtain multiple candidate models; an evaluation unit 8053, which is used to evaluate the multiple candidate models through the verification set to obtain evaluation results, and obtain an optimized deep learning model according to the evaluation results.

[0115] In this embodiment, the inversion module 806 includes: a fourth acquisition unit 8061, used to obtain inversion input data from the atmospheric turbulence simulation device; an inversion unit 8062, used to input the inversion input data into the optimized deep learning model, and use the optimized deep learning model to perform inversion calculations on the internal flow field of the atmospheric turbulence simulation device to obtain inversion data.

[0116] above Figure 8 and Fig. 9 The flow field detection device in the embodiment of the present invention is described in detail from the perspective of modular functional entities. The flow field detection device in the embodiment of the present invention is described in detail from the perspective of hardware processing.

[0117] Fig.10 is a schematic diagram of the structure of a flow field detection device provided by an embodiment of the present invention. The flow field detection device 900 may have relatively large differences due to different configurations or performances, and may include one or more processors (central processing units, CPU) 810 (for example, one or more processors) and a memory 920, and one or more storage media 930 (for example, one or more mass storage devices) storing application programs 933 or data 932. Among them, the memory 920 and the storage medium 930 can be short-term storage or permanent storage. The program stored in the storage medium 930 may include one or more modules (not shown in the figure), and each module may include a series of instruction operations in the flow field detection device 900. Furthermore, the processor 910 can be configured to communicate with the storage medium 930, and execute a series of instruction operations in the storage medium 930 on the flow field detection device 900 to implement the steps of the flow field detection method provided in the above-mentioned method embodiments.

[0118] The flow field detection device 900 may also include one or more power supplies 940, one or more wired or wireless network interfaces 950, one or more input and output interfaces 960, and / or one or more operating systems 931, such as Windows Server, Mac OS X, Unix, Linux, FreeBSD, etc. It will be appreciated by those skilled in the art that Fig.10 The structure of the flow field detection device shown does not constitute a limitation on the flow field detection device, and may include more or less components than shown in the figure, or combine certain components, or arrange the components differently.

[0119] The present invention also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. Instructions are stored in the computer-readable storage medium. When the instructions are executed on a computer, the computer executes the steps of the flow field detection method.

[0120] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the above-described system, device, or unit can refer to the corresponding process in the aforementioned method embodiment and will not be repeated here.

[0121] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or the whole or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk and other media that can store program code.

[0122] Finally, it should be noted that the above description is only a preferred example of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or replace some of the technical features therein by equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A flow field detection method, characterized in that: The following steps are involved: Acquire experimental raw data from atmospheric turbulence simulation devices; Establish a mathematical simulation model based on the original experimental data, and obtain simulation data from the mathematical simulation model; Perform data verification on simulation data and experimental original data to obtain training data; Building deep learning models; The deep learning model is trained according to the training data to obtain an optimized deep learning model; By optimizing the deep learning model, the internal flow field of the atmospheric turbulence simulation device is inverted and the inversion data is obtained; The method of establishing a mathematical simulation model according to the original experimental data and obtaining simulation data from the mathematical simulation model includes: performing data analysis on the original experimental data to obtain simulation model type information; performing modeling based on the simulation model type information and the original experimental data to obtain a mathematical simulation model; and performing mathematical simulation on the internal flow field of the atmospheric turbulence simulation device using the mathematical simulation model to obtain simulation data. The data verification of the simulation data and the original experimental data to obtain the training data includes: performing data verification on the simulation data and the original experimental data to obtain a simulation error; adjusting the mathematical simulation model according to the simulation error to obtain an optimized mathematical simulation model; performing mathematical simulation on the internal flow field of the atmospheric turbulence simulation device based on the optimized mathematical simulation model to obtain optimized simulation data; and merging the optimized simulation data and the original experimental data to obtain training data; The fast inversion capability of the optimized deep learning model significantly shortens the calculation cycle of the mathematical simulation model and improves the simulation efficiency, so that experimenters can perform real-time control of the atmospheric turbulence simulation device based on the inversion data, respond to experimental needs in real time, and improve the accuracy and efficiency of the experiment.

2. The flow field detection method according to claim 1, characterized in that: The step of obtaining raw experimental data from the atmospheric turbulence simulation device includes: The atmospheric turbulence simulation device includes a plurality of sensors for detecting the interior of the device; Obtain the basic data of each sensor from the atmospheric turbulence simulation device, clean and reduce the noise of the basic data to obtain experimental observation data; Obtain experimental input data from an atmospheric turbulence simulation device; The experimental observation data and the experimental input data are merged to obtain the experimental original data.

3. The flow field detection method according to claim 1, characterized in that: The establishing of the deep learning model comprises: Determine the type of deep learning model from the preset model library based on the original experimental data to obtain the basic model; Obtaining deep learning configuration information from the original experimental data, and setting the network parameters of the basic model according to the deep learning configuration information to obtain a configuration model; Obtain boundary condition constraints and physical equation constraints from experimental raw data; Constraints are imposed on the configuration model according to boundary condition constraints and physical equation constraints to obtain a deep learning model.

4. The flow field detection method according to claim 1, characterized in that: The deep learning model is trained according to the training data to obtain an optimized deep learning model, including: Extract training and validation sets from the training data; The deep learning model is trained according to the training set to obtain multiple candidate models; Multiple candidate models are evaluated through the validation set to obtain evaluation results, and an optimized deep learning model is obtained based on the evaluation results.

5. The flow field detection method according to claim 1, characterized in that: The inversion calculation of the internal flow field of the atmospheric turbulence simulation device is performed by optimizing the deep learning model to obtain inversion data, including: Obtain inversion input data from an atmospheric turbulence simulation device; The inversion input data is input into the optimized deep learning model, and the optimized deep learning model is used to invert the internal flow field of the atmospheric turbulence simulation device to obtain the inversion data.

6. A flow field detection device, characterized in that: include: An acquisition module is used to obtain raw experimental data from an atmospheric turbulence simulation device; A simulation module is used to establish a mathematical simulation model based on the original experimental data and obtain simulation data from the mathematical simulation model; The method of establishing a mathematical simulation model according to the original experimental data and obtaining simulation data from the mathematical simulation model includes: performing data analysis on the original experimental data to obtain simulation model type information; performing modeling based on the simulation model type information and the original experimental data to obtain a mathematical simulation model; and performing mathematical simulation on the internal flow field of the atmospheric turbulence simulation device using the mathematical simulation model to obtain simulation data. The verification module is used to perform data verification on the simulation data and the original experimental data to obtain training data; the data verification on the simulation data and the original experimental data to obtain training data includes: performing data verification on the simulation data and the original experimental data to obtain simulation errors; adjusting the mathematical simulation model according to the simulation errors to obtain an optimized mathematical simulation model; performing mathematical simulation on the internal flow field of the atmospheric turbulence simulation device based on the optimized mathematical simulation model to obtain optimized simulation data; merging the optimized simulation data and the original experimental data to obtain training data; Build modules for building deep learning models; A training module, used to train the deep learning model according to the training data to obtain an optimized deep learning model; The inversion module is used to perform inversion calculations on the internal flow field of the atmospheric turbulence simulation device by optimizing the deep learning model to obtain inversion data.

7. A flow field detection device, characterized in that: The flow field detection device comprises: a memory and at least one processor, wherein the memory stores instructions; at least one of the processors calls the instructions in the memory so that the flow field detection device executes each step of the flow field detection method as described in any one of claims 1-5.

8. A computer-readable storage medium having instructions stored thereon, characterized in that: When the instructions are executed by the processor, the various steps of the flow field detection method according to any one of claims 1 to 5 are implemented.

Images