Bulletin of the American Physical Society
71st Annual Meeting of the APS Division of Fluid Dynamics
Volume 63, Number 13
Sunday–Tuesday, November 18–20, 2018; Atlanta, Georgia
Session M30: Experimental Techniques: Extending Capabilities of PIV Based Measurements |
Hide Abstracts |
Chair: Steve Beresh, Sandia National Lab Room: Georgia World Congress Center B402 |
Tuesday, November 20, 2018 8:00AM - 8:13AM |
M30.00001: Abstract Withdrawn
|
Tuesday, November 20, 2018 8:13AM - 8:26AM |
M30.00002: A measurement of particle translation and rotation using magnetic particle tracking Xingtian Tao, Huixuan Wu Optical based diagnostic methods are ubiquitous in flow measurements. For example, the particle image velocimetry (PIV) is able to provide 2D or 3D information of an entire flow field with high resolution. However, the optical methods have two limits: they can hardly be used in an opaque environment, and most of them are not able to measure the spin rate of individual particles, which plays a key role in granular flow. To address these issues, a magnetic particle tracking (MPT) technique is developed. The principle is to reconstruct the position and orientation of a magnetic dipole based on its magnetic field distribution. A magnetic particle in motion has six degrees of freedom. Hence, the measurements at six or more positions in the field are sufficient for the reconstruction (assuming each position provides one field strength signal). The MPT does not need illumination, facilitating the measurement in the opaque environment, and more importantly, it is capable to measure the particle spin rate. In addition, the MPT has no harmful radiation and the measurement system is robust and easy to operate. Therefore, the MPT will have a wide application in the study of particulate flows. |
Tuesday, November 20, 2018 8:26AM - 8:39AM |
M30.00003: Development and Validation of Helium-Filled Soap Bubble System for Time-Resolved Velocimetry Paul Allen Swiney, Harrison Taylor, Lokesh Silwal, Vrishank Raghav Traditional tracer particles used for Particle Image Velocimetry (PIV) imposes a limitation during time-resolved PIV measurements, where the laser pulse energy is nominally low. One solution to overcome this limitation is the use of neutrally buoyant helium-filled soap bubbles (HFSBs) as tracer particles, which have a much higher light scattering intensity than traditional particles. To this end, a HFSB system was designed and fabricated at Auburn University to enable time-resolved velocity field measurements. An orifice type nozzle was designed similar to previous investigations and was fabricated using a metal 3D printer with sufficient spatial resolution. The HFSB system was characterized by measuring the bubble diameter and production rates as a function of the nozzle design parameters. The HFSB system will be used to measure velocity field data for canonical flows, as a validation of the tracing fidelity of HFSBs, when compared to traditional particles. The system will be optimized to ensure appropriate time response characteristics of the tracer particles for such canonical flows with the eventual goal of applying it to flows commonly associated with rotorcraft. |
Tuesday, November 20, 2018 8:39AM - 8:52AM |
M30.00004: Robust Principal Component Analysis of Corrupted Flow Fields Isabel Scherl, Kazuki Maeda, Brian Polagye, Steven L Brunton When particle image velocimetry (PIV) is used to quantitatively measure fluid flows, noise and spurious velocity vectors can reduce measurement quality and degrade subsequent analysis. Standard post-processing methods involve subjective outlier identification and interpolation, which can discard useful data and produce non-physical flow fields. Here, we present a new method to clean and de-noise PIV measurements using robust principal component analysis (RPCA). RPCA can uncover and isolate the low-rank coherent structure of data, objectively identifying and removing noise and sparse outliers. We demonstrate the RPCA algorithm on simulations and experiments where the flow physics are well-known. We analyze the turbulence spectra of flow measurements from high-fidelity numerical simulations before and after RPCA to show that dominant coherent structures are not degraded. Additionally, the dimension of the low-rank subspace and percentage of non-zero elements in the sparse subspace are analyzed to determine how aggressively to filter. |
Tuesday, November 20, 2018 8:52AM - 9:05AM |
M30.00005: Data assimilation for particle image velocimetry based on lattice Boltzmann method Dong Kim, Kyung Chun Kim Experimental and numerical approaches in PIV measurements have limitations in uncertainty that arise in experimental setups and mathematical models. In this regard, data assimilation for PIV that can complement experimental data and numerical model have received much attention in the fluid mechanics community. In the post-processing of PIV data, the most widely used approaches are linear interpolation, POD and Kalman filtering which are reduced order of modeling using general signal processing techniques. However, these techniques do not consider the flow physics. In this study, we propose a new data assimilation method for PIV using LBM. We solve the flow governing equations directly and non-iteratively using the PIV data as initial or boundary conditions to resolve finer temporal and spatial velocity vectors. In this study, GPGPU operation was performed using openCL for parallel data analysis. With a comparison of the proposed method to existing experimental and DNS data base, the performance of LBM based data assimilation method is verified more effective to recover missing data of PIV results. |
Tuesday, November 20, 2018 9:05AM - 9:18AM |
M30.00006: Pressure reconstruction using velocity measurement-error based generalized least-squares Jiacheng Zhang, Melissa Brindise, Carlo Scalo, Pavlos Vlachos A pressure reconstruction method using the velocity measurement error to inform generalized least-squares (GLS) is introduced. An overdetermined linear system of equations is formulated to relate pressure with pressure gradients calculated from velocity field. Pressure field is obtained by solving the system using GLS with a weight matrix based on the accuracy of the pressure gradients. The weight matrix is generated based on pressure gradient errors by propagating the velocity errors calculated from spurious velocity divergence for incompressible flow. The pressure reconstruction method is demonstrated and analyzed using velocity fields from artificial images. The method is applied to flow in the geometry of a cerebral aneurysm. 3-dimensional velocity field is obtained using PTV with STB method. The pressure reconstruction method is validated by comparing the calculated pressure field with direct pressure measurements and CFD simulation. Application to in-vivo flow measurement by MRI is also included. |
Tuesday, November 20, 2018 9:18AM - 9:31AM |
M30.00007: Robust low noise sensitivity of PIV-based pressure measurement by omnidirectional integration Jose R Moreto, Xiaofeng Liu
We derived an analytical expression for the error propagation from the PIV-based pressure gradient to the integrated pressure by the omnidirectional integration method. The pressure calculation for the boundary points is an iterative process and the error decreases for each iteration. The inner domain pressure is a one-step calculation which uses the boundary point solution. For the inner domain, the error is an average of the boundary error and the integration truncation error. The analysis shows that the omnidirectional integration provides an effective mechanism to reduce the sensitivity to random noise. We verified those results using a direct numerical simulation (DNS) database of isotropic turbulence flow, with a homogeneously distributed random noise added to the entire field of DNS pressure gradient. The random noise has a magnitude varying randomly within the range of ±40% of the maximum DNS pressure gradient. A total of 1000 statistically independent noise distributions achieved by using different random number seeds. Three different methods are compared. The average error is 0.15±0.07 for the Poisson approach, 0.028±0.003 for the Circular Virtual Boundary method and 0.027±0.003 for the Rotating Parallel Ray method, indicating the validity of the expressions derived. |
Tuesday, November 20, 2018 9:31AM - 9:44AM |
M30.00008: Efficient reconstruction of flow field with pressure from particle tracks: VIC# Young Jin Jeon, Markus Müller, Dirk Michaelis, Bernhard Wieneke A data assimilation-based method, which supplements additional constraints and coarse-grid approximation to Vortex-in-Cell-plus (VIC+), is proposed. An instantaneous pressure field is optimized in addition to and simultaneously with vorticity, velocity and acceleration fields. Not only a disparity between measurement and reconstruction at each particle position but also grid-based residuals from the additional constraints are minimized by an iterative optimization procedure. The additional constraints are derived from the Navier-Stokes equation, continuity equations, and vector calculus identities. Larger flow structures are mainly reconstructed by the coarse-grid approximation which consists of multiple schemes with various grid sizes, and thus computation time is saved. The coarse-grid approximation also contributes to stable convergence because of refined initial guesses for finer-grid schemes and therefore leads to smaller residuals at the finest-grid scheme. Experimental 4D PTV data obtained by Shake-the-Box (STB) is dealt with by the present method, Vortex-in-Cell-sharp (VIC#), and its reconstructed fluid structure, pressure field, and computational performances are analyzed. |
Tuesday, November 20, 2018 9:44AM - 9:57AM |
M30.00009: Velocity Estimation in the Shear Layer of a Mj=0.6 Jet Using Deep Neural Networks Andrew S. Tenney, Mark N Glauser, Zachary P Berger In this presentation, we recast the Linear Stochastic Estimation method to take advantage of the predictive power of Deep Neural Networks (DNNs), and offer a quantitative comparison to traditional Linear Stochastic Estimation (LSE). High frame-rate Particle Image Velocimetry (PIV) (10kHz) was used to measure the flow-field of a Mach 0.6 axisymmetric jet. Time-series of the stream-wise and cross-stream components of velocity were recorded at 11 locations, spanning 0.1 to 0.75 diameters in the radial direction, simulating the output of crosswires at those locations. A reduced order model of the turbulent velocity fluctuations in the shear layer is then produced using Proper Orthogonal Decomposition (POD). The resulting model is used to train both a DNN and an LSE model to predict the velocity within the shear layer, given a subset of the raw velocity measurements. We show that, on average, the DNN is able to predict the velocity fluctuations more accurately than LSE, because of the non-linear relationship between the conditional and unconditional events. |
Tuesday, November 20, 2018 9:57AM - 10:10AM |
M30.00010: Using PIV Data to Determine Turbulent Viscosity for Data-Driven Turbulence Modeling Steven Jay Beresh, Nathan E Miller, Barton L Smith Turbulent viscosities have been calculated from stereoscopic particle image velocimetry (PIV) data for a supersonic jet exhausting into a transonic crossflow. Image interrogation must be optimized to produce useful turbulent viscosity fields. High-accuracy image reconstruction should be used for the final iteration, whereas efficient algorithms produce spatial artifacts in derivative fields. Mean strain rates should be calculated from large windows (128 pixel) with 75% overlap. Turbulent stresses are optimally computed using multiple (more than two) iterations of image interrogation and 75% overlap, both of which increase the signal bandwidth. However, the improvement is modest and may not justify the considerable increase in computational expense. The turbulent viscosity may be expressed in tensor notation to include all three axes of velocity data. In this formulation, a least-squares fit to the multiple equations comprising the tensor generated a scalar turbulent viscosity that eliminated many of the artifacts and noise present in the single-component formulation. The resulting experimental turbulent viscosity fields will be used to develop data-driven turbulence models that can improve the fidelity of predictive computations. |
Follow Us |
Engage
Become an APS Member |
My APS
Renew Membership |
Information for |
About APSThe American Physical Society (APS) is a non-profit membership organization working to advance the knowledge of physics. |
© 2024 American Physical Society
| All rights reserved | Terms of Use
| Contact Us
Headquarters
1 Physics Ellipse, College Park, MD 20740-3844
(301) 209-3200
Editorial Office
100 Motor Pkwy, Suite 110, Hauppauge, NY 11788
(631) 591-4000
Office of Public Affairs
529 14th St NW, Suite 1050, Washington, D.C. 20045-2001
(202) 662-8700