Bulletin of the American Physical Society
68th Annual Meeting of the APS Division of Fluid Dynamics
Volume 60, Number 21
Sunday–Tuesday, November 22–24, 2015; Boston, Massachusetts
Session M27: Experiments: Analysis, Image Processing and Algorithms |
Hide Abstracts |
Chair: Brenden Epps, Thayer School of Engineering, Dartmouth College Room: 308 |
Tuesday, November 24, 2015 8:00AM - 8:13AM |
M27.00001: On the Singular Value Decomposition of Measured Data Brenden Epps Singular value decomposition (SVD) is a well-known mathematical tool that can be used to decompose an ensemble of velocity field data into spatiotemporal modes that may reveal coherent flow structures. The proper orthogonal decomposition (POD) is a special case of the SVD, used when the data are uncorrelated in time (as in a turbulent flow). Although the SVD and POD have been widely used in fluid mechanics, Epps and Techet (2010, ExpFluids 48:355–367) were among the first to consider how experimental error affects the results of the SVD. This talk revisits that paper and provides mathematically-rigorous bounds on the errors in the computed singular values and spatio-temporal mode shapes. Given experimental data with unknown error, a procedure is presented to (i) determine the root mean square measurement error and (ii) determine “error bars” for the singular values and vectors. [Preview Abstract] |
Tuesday, November 24, 2015 8:13AM - 8:26AM |
M27.00002: Iterative Blind Deconvolution Algorithm for Deblurring PSP Image of Rotating Surfaces Anshuman Pandey, James Gregory Fast Pressure-Sensitive Paint (PSP) is used in this work to measure unsteady surface pressures on rotating bodies, with iterative image deblurring schemes being developed to correct for image blur at high rotation rates. A significant amount of rotational blur can occur in PSP images acquired in the lifetime mode when the time scale of luminescent decay is long relative to the rotational speed. Image deblurring schemes have been developed to address this problem, but are not currently able to handle strong pressure gradients. Since the local point spread function at each point on the rotor depends on the unknown pressure, restoring such an image is a spatially-varying blind deconvolution problem. An iterative scheme based on the lifetime decay characteristics of PSP has been developed for restoring this image. The scheme estimates the spatially-varying blur kernel without filtering the blurred image and then restores it using classical iterative regularization tools. The resulting scheme is evaluated using computationally-generated pressure fields with strong gradients, as well as experimental data with strong gradients in luminescent lifetime due to a nitrogen jet. Factors such as convergence, image noise, and regularization-iteration count are studied in this work. [Preview Abstract] |
Tuesday, November 24, 2015 8:26AM - 8:39AM |
M27.00003: Machine learning and synthetic aperture refocusing approach for more accurate masking of fish bodies in 3D PIV data Logan Ford, Abhishek Bajpayee, Alexandra Techet 3D particle image velocimetry (PIV) is becoming a popular technique to study biological flows. PIV images that contain fish or other animals around which flow is being studied, need to be appropriately masked in order to remove the animal body from the 3D reconstructed volumes prior to calculating particle displacement vectors. Presented here is a machine learning and synthetic aperture (SA) refocusing based approach for more accurate masking of fish from reconstructed intensity fields for 3D PIV purposes. Using prior knowledge about the 3D shape and appearance of the fish along with SA refocused images at arbitrarily oriented focal planes, the location and orientation of a fish in a reconstructed volume can be accurately determined. Once the location and orientation of a fish in a volume is determined, it can be masked out. [Preview Abstract] |
Tuesday, November 24, 2015 8:39AM - 8:52AM |
M27.00004: The Anatomy of Fourier-Based Correlation Image Velocimetry and Sources of Decorrelating Errors Matthew Giarra, Pavlos Vlachos Particle image velocimetry (PIV) algorithms have recently been applied to photographs captured using a variety of techniques including schlieren, synchrotron x-ray, and microscope imaging. While the characteristics of these types of images differ greatly from those of particle images, virtually no analysis has been done to determine how these differences affect the performance of Fourier-based cross correlation (CC) algorithms. Here, we analyze schlieren, x-ray, and traditional PIV images to show that the signal-to-noise ratios (SNR) of their CCs vary across spectral wavenumbers, and that the assignment of a single SNR to the CC is an oversimplification that obfuscates the underlying source of the decorrelating errors. We will show that the failure of traditional algorithms to distinguish correlated from uncorrelated wavenumbers introduces secondary CC peaks that increase measurement uncertainty by decreasing the correlation peak-height ratio, and can cause the measurement to fail by overtaking the true peak. Finally, we introduce a new algorithm that mitigates these issues and increases measurement accuracy by automatically discriminating correlated wavenumbers with no a priori information about the images' contents. [Preview Abstract] |
Tuesday, November 24, 2015 8:52AM - 9:05AM |
M27.00005: Empirical mode decomposition profilometry: small scale capabilities and comparison to Fourier Transform Profilometry Guillaume Lagubeau, Pablo Cobelli, Tomasz Bobinski, Agnes Maurel, Vincent Pagneux, Philippe Petitjeans Fringe projection profilometry is an instrument of choice for the instantaneous measurement of the full height map of a free-surface. It is useful to capture interfacial phenomena such as droplet impact and propagation of water waves. We present the Empirical Mode Decomposition Profilometry (EMDP) for the analysis of fringe projection profilometry images. It is based on an iterative filter, using empirical mode decomposition, that is free of spatial filtering and adapted for surfaces characterized by a broadband spectrum of deformation. Examples of such surfaces can be found in nonlinear wave interaction regimes such as wave turbulence in gravity-capillary water waves. We show both numerically and experimentally that using EMDP improves strongly the profilometry small scale capabilities compared to traditionally used Fourier Transform Profilometry. Moreover, the height reconstruction distortion is much lower: the reconstructed height field is now both spectrally and statistically accurate. [Preview Abstract] |
Tuesday, November 24, 2015 9:05AM - 9:18AM |
M27.00006: New Reconstruction Accuracy Metric for 3D PIV Abhishek Bajpayee, Alexandra Techet Reconstruction for 3D PIV typically relies on recombining images captured from different viewpoints via multiple cameras/apertures. Ideally, the quality of reconstruction dictates the accuracy of the derived velocity field. A reconstruction quality parameter $Q$ is commonly used as a measure of the accuracy of reconstruction algorithms. By definition, a high $Q$ value requires intensity peak levels and shapes in the reconstructed and reference volumes to be matched. We show that accurate velocity fields rely only on the peak locations in the volumes and not on intensity peak levels and shapes. In synthetic aperture (SA) PIV reconstructions, the intensity peak shapes and heights vary with the number of cameras and due to spatial/temporal particle intensity variation respectively. This lowers $Q$ but not the accuracy of the derived velocity field. We introduce a new velocity vector correlation factor $Q_v$ as a metric to assess the accuracy of 3D PIV techniques, which provides a better indication of algorithm accuracy. For SAPIV, the number of cameras required for a high $Q_v$ are lower than that for a high $Q$. We discuss $Q_v$ in the context of 3D PIV and also present a preliminary comparison of the performance of TomoPIV and SAPIV based on $Q_v$. [Preview Abstract] |
Tuesday, November 24, 2015 9:18AM - 9:31AM |
M27.00007: How To Efficiently Sample Data For Computation Of Statistics Barton Smith, Douglas Neal The mean of a sample is a random variable with a variance that is the variance of the measured variable divided by the number of samples, assuming the samples are independent. Ensuring independent samples requires that the sampling period is greater than twice the integral time scale $T_{u}$. This time scale is the integral over all time of the autocorrelation $\rho $ of the signal. Three signals are analyzed to investigate the convergence of the mean and other statistics. One is the velocity in a turbulent jet measured using a hot-wire. The other two are pressure signals generated by flow through a confined array of cylinders. To determine $T_{u}$, 11 sets of 100,000 data points were acquired at high rate. The value of $\rho $ for each record were averaged together. For the chaotic signal from the jet, $\rho $ was generally positive or slightly negative. The pressure signals contain a coherent component that caused the autocorrelation to become a damped oscillation. In this case, the scheme of integrating $\rho $ from 0 to the time where $\rho $ became negative gives poor results. Instead, an exponential was fit to the envelope of $\rho $ and integration was carried out to the point where this exponential function became small. Data were acquired in the jet and cylinder array at rates above and below those consistent with the 2$T_{u}$ criterion. Even at 8 times the recommended rate the mean converged at the predicted rate [although the error magnitude was larger than the theory] and that at sampling period of 2$T_{u}$ the error in the mean was well predicted by theory). Similar results were found for convergence of the variance. [Preview Abstract] |
Tuesday, November 24, 2015 9:31AM - 9:44AM |
M27.00008: Modal Energy Flow Analysis the Highly Modulated Wake of a Wall-mounted Square-based Pyramid Robert Martinuzzi, Zahra Hosseini, Bernd Noack We present the first modal energy flow analysis of a time-resolved 3D velocity field from experimental PIV data for a highly modulated wake of a square-based pyramid protruding a boundary layer. The underlying low-order representation is optimized for resolving the base flow variation as well as the first and second harmonics associated with vortex shedding, thus generalizing the triple decomposition of Reynolds {\&} Hussain (1972). The analysis comprises not only a detailed modal balance of turbulent kinetic energy as pioneered by Rempfer {\&} Fasel (1994) for POD models, but also the companion mean-flow energy balance. The experimental results strikingly demonstrate how constitutive elements of mean-field theory (Stuart, 1958) near laminar Hopf bifurcations are still strongly expressed in a turbulent wake characterized by highly modulated, quasi-periodic shedding. The results emphasize, for instance, the stabilizing role of the mean-field manifolds, as explored in the POD model of Aubry et al. (1988). The proposed low-order representation of the flow and modal energy analyses may provide a novel framework for characterizing highly anisotropic wakes and vortex interactions; yielding important insights and reference data for computational turbulence modeling, e.g. URANS. [Preview Abstract] |
Tuesday, November 24, 2015 9:44AM - 9:57AM |
M27.00009: Flow classification using machine learning on sparsely sampled experimental flow visualization data Zhe Bai, Steven L. Brunton, Bingni W. Brunton, J. Nathan Kutz, Eurika Kaiser, Andreas Spohn, Bernd R. Noack In this work, we consider a data-driven approach for characterizing the transitional separation bubble using video images and dimensionality reduction with supervised classification techniques to discriminate between an actuated and an unactuated flow. Flow visualizations are captured using the hydrogen bubble technique along a smooth ramp in a low-speed water tunnel, and instabilities are excited in the actuated case by oscillating a thin horizontal wire inside the boundary layer upstream of the separation. We apply clustering techniques, including the linear discriminant analysis (LDA) in a POD/PCA reduced subspace, to classify the baseline and controlled cases of the flow field from image data. With sparse subsampled pixel measurements, similar classification performance is obtained compared to that of the full-resolution images. Next, we demonstrate a sparse sensor optimization algorithm to locate a small set of pixels that optimally inform the classification task. With 5-10 specially selected sensors, the median cross-validated classification accuracy is $\geq 97\%$, as opposed to a random set of 5-10 pixels, which result in classification accuracy of 70-80\%. The methods developed here apply broadly to high-dimensional data from fluid dynamics experiments. [Preview Abstract] |
Tuesday, November 24, 2015 9:57AM - 10:10AM |
M27.00010: Image Processing Method of the Motion-Capturing PSP/TSP for the Measurement of a Free-Flight Object Masato Ishii, Hideki Goya, Takeshi Miyazaki, Hirotaka Sakaue The motion-capturing PSP/TSP system consists of a two-color PSP/TSP and a high-speed color camera. Red and green luminescent images are acquired simultaneously as signal and reference outputs by this system. Simply by rationing the red and the green images, we can obtain a pressure/temperature distribution on the surface of a target object. This system is applied to measure the surface pressure/temperature of a free-flight object. However, an acquired image includes motion blur, focus blur and random noise around the object. We discuss image processing methods and evaluations to optimize those uncertainties. Three types of the edge detect methods are used, which are the sobel, the laplassian and the canny. We will also show the evaluation results to discuss an optimized image processing for the motion-capturing PSP/TSP system. [Preview Abstract] |
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