实验流体力学

contents

2022, 36(3)

Abstract(198) PDF(53)

Abstract:

2022, 36(3): 0-1.

Abstract(235) HTML (166) PDF(31)

Abstract:

Modern design of experiment and its development in aerodynamics

HAI Chunlong, HE Lei, MEI Liquan, QIAN Weiqi

2022, 36(3): 1-10. doi: 10.11729/syltlx20220005

Abstract(2187) HTML (392) PDF(127)

Abstract:
The scientific experimental design method can significantly improve the quality and efficiency of scientific research and industrial production. This paper introduces and summarizes the research progress of modern test design methods. Firstly, this paper summarizes the differences between OFAT（One Fact at A Time） method and MDOE（Modern Design Of Experiments） method in the wind tunnel test in three aspects: the test purpose, organization strategy and test results, and analyzes the advantages of MDOE method. Secondly, it summarizes the status quo of the MDOE method in three aspects: the experimental design, model establish-ment and result analysis. Then we demonstrate the experimental design method with standard functions and aerodynamic examples. Finally, some key scientific problems and future research directions are discussed.

Research on intelligent optimal design method of wind tunnel test scheme

JIANG Jinjun, ZHOU Xuanchi, CHEN Lianzhong, CUI Ning, JIANG Yan

2022, 36(3): 11-19. doi: 10.11729/syltlx20210155

Abstract(1755) HTML (243) PDF(97)

Abstract:
Advanced aircraft research requires rapid acquisition of correct-key data on the aerodynamic shape. The development of our country's aerospace equipment has gradually entered a stage of independent innovation. The exploration, innovation, and optimization of the aerodyna-mic design for a new generation of aircraft need to be supported by matching wind tunnel test capabilities. The development trend of high-performance and high-precision advanced aircraft has put forward higher requirements on the “quantity” and “quality” of wind tunnel test data, which means the design of traditional wind tunnel test schemes and data analysis modes have become increasingly unsatisfactory. It is necessary to make breakthroughs in wind tunnel test design and test data analysis technology. Based on aircraft ground test analysis, this paper uses the support vector machine（SVM） to model and analyze the test data, develop wind tunnel test plan optimization, design methods and test data intelligent analysis methods, and explore the new internal correlation between the aerodynamic data and aircraft geometric parameters. Building a wind tunnel test auxiliary design and analysis system to improve the efficiency of wind tunnel test and the accuracy of test data provides technical support for the aerodynamic design of high-performance aircraft.

Using modern design of experiments method for hypersonic wind tunnel test

YOU Wenjia, WANG Huijie, HAN Renkun, CHEN Gang

2022, 36(3): 20-32. doi: 10.11729/syltlx20210179

Abstract(1573) HTML (375) PDF(83)

Abstract:
Modern Design of Experiments （MDOE） is an important technical approach to improve wind tunnel test efficiency. Although the modern design of experiment method based on Latin Hypercube Sampling has high theoretical efficiency, the practical efficiency of the random sampling points designed by it decreases significantly when the automatic attitude control system of the wind tunnel model is coordinated. In this paper, a modern design of experiment method based on the Stratified Latin Hypercube design is proposed to meet the requirements of the multi-variable wind tunnel test design in the automatic attitude control system. It is applied to the two-variable test and three-variable test design of the 6 Mach number wind tunnel model. The results which were compared with the One Factor at A Time （OFAT） method show that MDOE method only needs about 20% of the sample size of OFAT method in the two-variable test and only needs about 30% of the sample size of OFAT method in the three-variable test. Compared with the classical Latin Hypercube method, the Stratified Latin Hypercube method developed in this paper combined with the existing wind tunnel test equipment can effectively reduce the change of test runs, improve the test efficiency and shorten the test period.

Investigation on artificial intelligence for the prediction of aeroacoustic performances and controlling parameters optimization of aircraft

WU Junqiang, YANG Dangguo, ZHANG Lin, GONG Tianchi, ZHOU Fangqi, WANG Yan, LI Yang

2022, 36(3): 33-43. doi: 10.11729/syltlx20210073

Abstract(1712) HTML (264) PDF(110)

Abstract:
Accurate aerodynamic characteristic data under different conditions is the prerequisite and fundamental guarantee for the fast design of a flight vehicle, the improvement of a control system, the evaluation of performances and performance appraisal. The cross synthesis between the machine learning technology （ML） based on deep neural network （DNN） and fluid mechanics is developing fast and has achieved remarkable progresses in the modification of turbulence models, modeling of systems, prediction of the aerodynamic and aeroacoustic characteristics, optimization of control parameters and reconstruction of the flow field. To effectively apply the powerful representative capability of DNN, according to the demand of intelligent optimization and design of weapon bays, this paper first established a database of aerodynamic loads for flows past cavities and then built deep forward neural network model for the prediction of aerodynamic loads. To enhance the robustness of the model, random search and Bayesian optimization are introduced during the training of the model. Numerical results show that the trained DNN model is able to predict the aerodynamic loads and aeroacoustic characteristics accurately and efficiently, which provides a useful tool for the prediction and control of the aeroacoustic characteristics of the cavity.

Predictive analysis of flow control in high-speed complex flow field based on machine learning

YU Baiyang, LYU Hongqiang, ZHOU Yan, LUO Zhenbing, LIU Xuejun

2022, 36(3): 44-54. doi: 10.11729/syltlx20210168

Abstract(1573) HTML (254) PDF(81)

Abstract:
The flow control actuator is the core of the active flow control technology. The design level and performance of actuator directly determine the application direction and effect of active flow control. In order to obtain the action law of the flow control actuator, a large number of experiments are needed to study the influence of excitation parameters on control effect parameters, and the experimental cost is large. In this paper, the experimental data of jet shock control in reverse plasma synthesis are used, and the Gaussian process regression model in machine learning is used to obtain the mapping law from the actuator parameters （head cone diameter, cavity volume, discharge capacitance and outlet diameter） to the control effect parameters （maximum out of body distance）. We compare the prediction effects of Gaussian process regression under various kernel functions, and analyze the influence of actuator parameters on control effect parameters by using the characteristic importance analysis method. The results show that for this small sample problem, Gaussian process regression with the quadratic polynomial kernel function Poly2 obtains the highest accuracy; in characteristic importance analysis, the head cone diameter has the greatest influence on the maximum separation distance, followed by discharge capacitance and cavity volume. The influence of these two parameters is similar, and the influence of the outlet diameter is the least. The work of this paper can provide some guidance for the setting of various parameters of the actuator in the flow control experiment of the high-speed complex flow field.

Deep reinforcement learning for the control of airfoil flow separation

YAO Zhangyi, SHI Zhiwei, DONG Yizhang

2022, 36(3): 55-64. doi: 10.11729/syltlx20210085

Abstract(1696) HTML (348) PDF(82)

Abstract:
A jet closed-loop control system based on Deep Reinforcement Learning （DRL） was built, and an experimental study was carried out on the separation flow control at high angles of attack on the NACA0012 airfoil. The airfoil chord length is 200 mm and the wind speed was 10 m/s. The Reynolds number was 1.36×10⁵ based on the chord length. The jet actuator was arranged on the upper surface of the airfoil and the solenoid valve was used for stepless control. The pressure coefficient of the airfoil surface and the action output of the agent itself were taken as the observation of the agent. The pressure coefficient of the trailing edge of the airfoil was used as the reward function to train the agent. Our results showed that the trained agent successfully suppresses the flow separation at high angles of attack and the cost-effectiveness ratio is reduced by 50% compared with steady blowing. At the same time, the agent could also stabilize the pressure coefficient of the trailing edge near the target value. The state input and the change of the reward function also have different effects on the final training effect.

Unsteady aerodynamic modeling research and virtual flight test verification

CHEN Xiang, ZHAN Jingxia, CHEN Ke, WEI Zhongcheng, CAO Yuan

2022, 36(3): 65-72. doi: 10.11729/syltlx20210143

Abstract(1454) HTML (212) PDF(76)

Abstract:
Unsteady aerodynamic modeling involves aerodynamics, flight mechanics, flight control and other fields, which is the key of aircraft high angle of attack database. The traditional aerodynamic model is composed of static aerodynamic, rotating balance and dynamic derivative data, which cannot describe unsteady aerodynamics exactly. Recurrent Neural Network （RNN） structure is a kind of neural network structure for processing and predicting sequence data, which is widely used in the field of artificial intelligence. RNN has the same time–dependent characteristic as unsteady aerodynamics. The application of RNN on the unsteady aerodynamic modeling has been researched. The forced motion test and virtual flight test have been used for unsteady aerodynamic model’s verification. In the forced motion wind tunnel test, comparing the aerodynamic forces of cobra maneuver, the result demonstrates that the RNN model is more accurate than the traditional model. In the virtual flight test, comparing the movement parameters curve of the wind tunnel test and simulation, the results also demonstrate that the RNN model is closer to the wind tunnel test than the traditional dynamic derivative model.

Predict the mass distribution of hypervelocity impact debris using deep learning

ZHOU Hao, LI Yi, LAN Shengwei, LIU Hai

2022, 36(3): 73-78. doi: 10.11729/syltlx20210125

Abstract(1636) HTML (129) PDF(72)

Abstract:
Efficiently predicting the evolution of hypervelocity impact（HVI） debris is crucial in the design of spacecraft protective structures. To this end two deep learning models, namely the de-convolutional architecture and the multi-layer perceptron architecture, are compared. Training data come from smoothed particle hydrodynamics （SPH） simulations. Deep learning models include four controllable labels, namely the projectile velocity （range 3–8 km/s） , the radius （range 2–8 mm）, target plate thickness （range 1–4 mm） and time instant （range 1–12 μs）. It is found that the two architectures are both accurate if the labels are confined within the range of training data. Out of the training data labels range, the de-convolutional architecture extrapolates poorly, though it can capture the granular property of the HVI debris. At the same time, the multi-layer perceptron architecture homogenizes the local mass distribution in the HVI debris and achieves much better extrapolation capability. By learning the mass distribution between time 1 μs and 12 μs, the multi-layer perceptron architecture can predict the mass distribution at time 24 μs with reasonable accuracy. Prediction time of the de-convolutional architecture is several milliseconds and prediction time of the multi-layer perceptron architecture is several seconds.

Comparison of machine learning data fusion methods applied to aerodynamic modeling of rocket first stage with grid fins

XU Chenzhou, DU Tao, HAN Zhonghua, ZAN Bowen, MOU Yu, ZHANG Jinze

2022, 36(3): 79-92. doi: 10.11729/syltlx20210154

Abstract(1575) HTML (309) PDF(95)

Abstract:
Machine learning data fusion method has attracted significant attention recently in aerodynamic database construction since it makes a trade-off between high prediction accuracy and low fitting cost by fusing samples of different fidelities. But the research on methods for complex engineering project is not sufficient. In this paper, several commonly used variable-fidelity models （VFMs） of data fusion are applied to the control law design in the rocket first stage landing area control project with grid fins. Based on wind tunnel tests of partial test states, combined with CFD simulation results, VFMs successfully predict the whole aerodynamic characteristics of grid fins. Here, our objective is to compare the performances of these four VFM methods （AS-MFS, Co-Kriging, HK, MFNN） and the results show that: Gaussian exponential function is more suitable for aerodynamic modeling problems; Co-Kriging has the best performance in the interpolation of aerodynamic data; HK model has high prediction accuracy for interpolation but has poor performance for extrapolation; MFNN model can obtain smoother and more reasonable results in the extrapolation region.

Fine reconstruction method of airfoil surface pressure based on multi-source data fusion

ZHAO Xuan, PENG Xuhao, DENG Zichen, ZHANG Weiwei

2022, 36(3): 93-101. doi: 10.11729/syltlx20210166

Abstract(1835) HTML (342) PDF(112)

Abstract:
Laying out pressure taps on the surface of the wind tunnel test model is an important means to obtain the surface pressure distribution. However, due to the limited space location and experimental cost, it is usually difficult to arrange enough pressure taps on the complex model surface to obtain complete surface pressure distribution information. Hence, the accuracy of the lift and moment calculated by the direct integration may fail to meet expectations. In this paper, to obtain the high-precision pressure distribution through less pressure test data, a method of combining the sparse wind tunnel test data and the numerical simulation data is proposed. Firstly, the proper orthogonal decomposition （POD） technique is used to extract the low-dimensional feature of the pressure distribution of the numerical simulation data, which is called the POD basis. Then, by applying the compressed sensing algorithm, the coordinates of the basis function are obtained with the sparse wind tunnel pressure measurement data, and finally transformed into the physical space to reconstruct the pressure distribution. The accuracy of the method is verified by the steady fixed airfoil variable state or with variable geometry in conjunction with variable flow state examples, and the reconstructed results can accurately match the experimental results. The developed reconstruction method largely solves the problem of fine reconstruction of distributed load under the condition of limited space and sparse observation.

Reconstruction of turbulent fields based on super-resolution reconstruction method

JIANG Hao, WANG Bofu, CHONG Kai Leong, LU Zhiming

2022, 36(3): 102-109. doi: 10.11729/syltlx20210185

Abstract(1883) HTML (256) PDF(177)

Abstract:
It is an important issue to obtain detailed flow fields from limited flow fields data. The convolutional-neural-networks-based super-resolution reconstruction methods developed in recent years are effective methods to obtain detailed flow fields. The efficient sub-pixel convolutional neural network（ESPCN） method is used to reconstruct Rayleigh–Bénard（RB） convection numerical simulation data and turbulent boundary layer（TBL） experimental measured data, and obtain high resolution flow fields data. The reconstructed high resolution flow fields data obtained using ESPCN is then compared to the results from the traditional super-resolution reconstruction method, the bicubic interpolation method. The results indicate that the flow fields reconstructed by the ESPCN method and the bicubic method agree well with the original high-resolution flow fields data when the down-sampling ratio is small. But, when the down-sampling ratio is large, the accuracy of the flow fields reconstructed by the ESPCN method is significantly better than that constructed by the bicubic method. In addition, the ESPCN method has a better performance than the bicubic method in areas with large gradients.

A high-resolution velocity field predicted method in near wall boundary layer by CNN

WANG Shaofei, PAN Chong, QI Zhongyang

2022, 36(3): 110-117. doi: 10.11729/syltlx20210142

Abstract(1435) HTML (185) PDF(68)

Abstract:
A high-resolution velocity field prediction method in the boundary layer based on CNN-PTV is designed and verified in this paper. This method includes two stages, training process and prediction process. In the training process, a CNN model is trained by the particle image pairs. By optimizing the parameters, the exact ensemble particle movements are predicted by the CNN model. In the prediction process, a synthetic image including just a single particle is imported to the CNN model to estimate the flow information at this particular pixel space. Thus, a high-resolution velocity field is predicted. Comparing with the single-pixel ensemble correlation method, the CNN-PTV method has a higher precision. And the results of CNN-PTV method is insensitive to the frame numbers and particle density.

Spatio-temporal reconstruction method of flow field based on deep neural network

HAN Renkun, LIU Ziyang, QIAN Weiqi, WANG Wenzheng, CHEN Gang

2022, 36(3): 118-126. doi: 10.11729/syltlx20210124

Abstract(1759) HTML (529) PDF(190)

Abstract:
The flow field PIV measurement method cost a lot, but the measurement results have low spatial and temporal resolution. The spatio-temporal reconstruction method of flow field based on experimental and numerical simulation data is studied. In order to realize the high-resolution spatio-temporal reconstruction of the experimentally measured low-resolution data, a flow field spatio-temporal reconstruction method based on deep neural network is proposed. A hybrid deep neural network based on convolutional neural network and long-short-term memory neural network is constructed. This hybrid deep neural network is trained to learn the spatio-temporal evolution features of the flow field. After the training is completed, it can be used to reconstruct the experimental data into spatio-temporal high-resolution results. The test results show that when the spatial high-resolution reconstruction is performed alone, the mean square error between the reconstructed flow field and the ground truth flow field is about 0.0065, and the number of data points is 51 times more than that of the input field. When the flow field is reconstructed to high resolution in time and space at the same time, the mean square error be maintained at about 0.065, and the density in the time dimension is 5 times more than that of the input field. It is proved that this method can greatly improve the efficiency of the experiment and save the cost of the experiment.

2022 Vol. 36, No. 3