Begell House Inc.
Journal of Flow Visualization and Image Processing
JFV
1065-3090
27
4
2020
DEVELOPING PARTICLE IMAGE VELOCIMETRY SOFTWARE BASED ON A DEEP NEURAL NETWORK
359-376
10.1615/JFlowVisImageProc.2020033180
Wojciech
Majewski
Microvec Pte Ltd., Singapore
Runjie
Wei
Microvec Pte Ltd., Singapore
Vivek
Kumar
Department of Aerospace Engineering, Indian Institute of Technology, Kanpur
208016, India
particle image velocimetry
estimation of fluid motion
deep neural networks
As an experimental technique for fluid mechanics, particle image velocimetry (PIV) can extract global and quantitative velocity field from images. With the development of artificial intelligence, designing PIV method based on deep learning is quite promising and worth exploring. First, in this paper, the authors introduce the optical flow neural network based on one proposed in the computer vision community. Second, a data set including particle images and the ground truth fluid motion is generated to train the parameters of the networks. This leads to a deep neural network for PIV which can provide estimation of dense motion (down to maximum one vector for one pixel) with the high degree of efficiency. The featuring of particle image extracted by the neural network is also investigated in this paper. It is found that feature matching improves the accuracy of estimation. The proposed network model is firstly evaluated by a synthetic image sequence of turbulent flow. An experiment measuring the flow over an aerofoil is used to validate the practicability. The experimental results indicate that compared with the traditional cross correlation method, the proposed deep neural network has advantages in accuracy, spatial resolution, and efficiency.
MULTI-SCALE SAFETY PROTECTIVE GEAR DETECTION UNDER POWER CONSTRUCTION SCENE
377-395
10.1615/JFlowVisImageProc.2020033142
Bo
Yang
The Electric Power Research Institute of State Grid Beijing Electric Power Company
Huan
Xie
The Electric Power Research Institute of State Grid Beij ing Electric Power
Company
Hongbin
Li
State Grid Beij ing Electric Power Company
Nuohan
Li
Beijing Jindian United Power Supply Consulting Co., Ltd.
Anchang
Liu
State Grid Beij ing Electric Power Company
Zhigang
Ren
The Electric Power Research Institute of State Grid Beij ing Electric Power
Company
Kuan
Ye
The Electric Power Research Institute of State Grid Beij ing Electric Power
Company
Rong
Zhu
The Electric Power Research Institute of State Grid Beij ing Electric Power
Company
Xuezhi
Xiang
Harbin Engineering University
power construction
deep convolutional neural network
small object detection
multi-scale object detection
residual module
In order to solve the problem of small object detection under power construction scene, a multi-scale object detection (MSOD) method based on deep convolutional neural network is proposed to realize the detection of worker's safety protective gears such as helmet and vest. Inspired by Darknet53, we firstly improve the capability of feature extraction of the network through adding residual modules, which can achieve more abstract features of the objects and improve the detection accuracy of large and medium targets. Based on the redesigned feature extraction network, an improved multi-scale object detection network is designed through re-planning the structure of multi-branch network, which can improve the detection accuracy of small targets. The experimental results show that the proposed multi-scale object detection method can achieve the higher accuracy and can be used in engineering practice.
INTERACTION CHARACTERISTICS OF MULTIPLE SWEEPING JET ACTUATORS IN SERIES
397-425
10.1615/JFlowVisImageProc.2020031039
Abbishek
Gururaj
Department of Aerospace Engineering, Indian Institute of Technology Kanpur,
Kanpur, Utt ar Pradesh, 208016, India
Rohan Ajit
Kulkarni
Department of Aerospace Engineering, Indian Institute of Technology Kanpur,
Kanpur, Utt ar Pradesh, 208016, India
Kamal
Poddar
Department of Aerospace Engineering, Indian Institute of Technology Kanpur,
Kanpur, Utt ar Pradesh, 208016, India
Sanjay
Kumar
Department of Aerospace Engineering, Indian Institute of Technology Kanpur,
Kanpur, Utt ar Pradesh, 208016, India
fluidic oscillators
Coanda effect
multiple jets
particle image velocimetry
interaction height
An external flow field of a single oscillator is characterized and the interaction characteristics of an array of oscillators in quiescent environment are studied in this experimental investigation. Time-averaged and phase-averaged velocity fields of a single oscillator are studied to characterize the external flow field. A spacing between the oscillators and volume flow rate were varied to study their effects on the jet interaction characteristics. Sweeping jet oscillation similarity was ensured using frequency characteristics of the oscillators. Two-dimensional particle image velocimetry (PIV) was used to characterize the external flow field. The method of zero-crossing was used for phase-averaging and the characteristics of the jet exiting from the oscillator at various phases were studied. The interaction height of the oscillating jets is influenced by the spacing between the oscillators and is found to increase with the spacing. The volume flow rate does not significantly affect the interaction height. The change in the spacing alters the deficit in the velocity between the peaks, while the change in the volume flow rate spatially shifts the profiles in the streamwise direction as the flow momentum is directly dependent on the volume flow rate. The deficit in velocity between the peaks increases with the spacing; however, an increase in the flow rate at the same spacing decreases this deficit.
CHARACTERISTICS OF SOLIDIFICATION-DRIVEN DOUBLE-DIFFUSIVE LAYERS IN MIXTURES
427-451
10.1615/JFlowVisImageProc.2020032771
Virkeshwar
Kumar
Department of Mechanical Engineering, Indian Institute of Technology Bombay,
Mumbai, 400076, India
Atul
Srivastava
Department of Mechanical Engineering, Indian Institute of Technology Bombay, Powai, Mumbai
400076, India
Shyamprasad
Karagadde
Department of Mechanical Engineering, Indian Institute of Technology Bombay,
Mumbai-400076, India
buoyancy-driven convection
double-diffusive layer
solidification
inter-ferometry
particle image velocimetry
In aqueous mixtures, the double-diffusive convection has the ability to form stepped profiles of temperature and composition in the vertical plane, and it is termed double-diffusive layer (DDL). The DDL typically consists of an alternating convecting layer separated by a high-gradient interface. The total height of a double-diffusive layer is influenced by the Prandtl number, which increases with the composition of the salt. In the present study, the influence of initial composition of ammonium chloride (wt.% NH4Cl) on the thickness of the double-diffusive layers during bottom-cooled solidification is investigated. The formation of double-diffusive layers was observed using a Mach-Zehnder interferometer, which quantified the stepped composition profiles, and full-field composition gradient field in 2D. The convective flows in the double-diffusive layers were captured using particle image velocimetry (PIV). The results revealed the decreasing of the convective layer thickness with increasing initial composition, while the contrary was observed for the thickness of the high-solute gradient interface. The insights are applicable to developing new understanding of the composition distribution in mixing oceanic currents as well as the formation of layered structures in igneous rocks.
INDEX VOLUME 27
101-104
10.1615/JFlowVisImageProc.v27.i4.50