Abstract
The objective of the method described in this work is to provide an improved reconstruction of an original flow field from experimental velocity data obtained with particle image velocimetry (PIV) technique, by incorporating the local accuracy of the PIV data. The postprocessing method we propose is Kriging regression using a local error estimate (Kriging LE). In Kriging LE, each velocity vector must be accompanied by an estimated measurement uncertainty. The performance of Kriging LE is first tested on synthetically generated PIV images of a two-dimensional flow of four counter-rotating vortices with various seeding and illumination conditions. Kriging LE is found to increase the accuracy of interpolation to a finer grid dramatically at severe reflection and low seeding conditions. We subsequently apply Kriging LE for spatial regression of stereo-PIV data to reconstruct the three-dimensional wake of a flapping-wing micro air vehicle. By qualitatively comparing the large-scale vortical structures, we show that Kriging LE performs better than cubic spline interpolation. By quantitatively comparing the interpolated vorticity to unused measurement data at intermediate planes, we show that Kriging LE outperforms conventional Kriging as well as cubic spline interpolation.
Similar content being viewed by others
Notes
The error matrix is not a ‘nugget effect’, as it does not directly change the correlation matrix and does not represent a cascade of unresolved smaller flowscales [36].
To avoid negative or infinitely large \(\epsilon\), we implement: \(\epsilon_{{\textsc{snr}},i}=\frac{c}{\max(\rm{SNR}_i-SNR_{\rm min},0.01)}\).
A MLE optimization of the Kriging correlation ranges typically results in θ x = θ y = 0.26, and in the following, we use a fixed range of θ x = θ y = 0.25.
One might conclude that the Low-Fi UM slightly underestimates the measurement uncertainty for high PMR. This conclusion could be attributed to the extremely simple Low-Fi UM; a slightly more sophisticated Low-Fi UM might incorporate the uncertainty associated with the subpixel fit as a lower uncertainty bound–which would roughly amount to \(\epsilon_{\rm subpix} \approx 0.1 {\hbox{m/s}}\) in the present case.
Unlike in mpiv , PPR is defined in Davis.
References
Adrian R (2005) Twenty years of particle image velocimetry. Exp Fluids 39:159–169
Scarano F, Riethmuller ML (2000) Advances in iterative multigrid PIV image processing. Exp Fluids 29:S051–S060
Westerweel J, Scarano F (2005) Universal outlier detection for PIV data. Exp Fluids 39:1096–1100
Theunissen R, Scarano F, Riethmuller ML (2007) An adaptive sampling and windowing interrogation method in PIV. Meas Sci Technol 18(1):275
Elsinga G, Scarano F, Wieneke B, Oudheusden B (2006) Tomographic particle image velocimetry. Exp Fluids 41:933–947
Scarano F (2013) Tomographic PIV: principles and practice. Meas Sci Technol 24(1):012001
Percin M, Eisma H, van Oudheusden B, Remes B, Rujsink R, de Wagter C (2012) Flow visualization in the wake of flapping-wing MAV ‘DelFly II’ in forward flight, In AIAA fluid dynamics and co-located conferences and exhibit New Orleans
Raffel M, Willert CE, Wereley ST, Kompenhans J (2007) Particle Image Velocimetry: a practical guide (2nd ed). Springer, Berlin
Adrian R, Westerweel J (2010) Particle image velocimetry. Cambridge University Press, Cambridge Aerospace Series, Cambridge
Matheron G (1963) Principles of Geostatistics. Econ Geol 58:1246–1266
Gandin L (1965) Objective analysis of meteorological fields: Gidrometeorologicheskoe Izdatel’stvo (GIMIZ), Leningrad. Translated by Israel Program for Scientific Translations, Jerusalem
Cressie N (1990) The origins of Kriging. Math Geol 22(3):239–252
Cressie N (1993) Statistics for spatial data. Wiley: New York
Venturi D, Karniadakis GE (2004) Gappy data and reconstruction procedures for flow past a cylinder. J Fluid Mech 519:315–336
Gunes H, Sirisup S, Karniadakis GE (2006) Gappy data: to krig or not to krig?. J Comput Phys 212:358–382
Gunes H, Rist U (2007) Spatial resolution enhancement/smoothing of stereoparticle-imagevelocimetry data using proper-orthogonal decomposition based and kriging interpolation methods. Phys Fluids 19(6):064101-1–064101-19
Gunes H, Rist U (2008) On the use of kriging for enhanced data reconstruction in a separated transitional flat-plate boundary layer. Phys Fluids 20(10):104109-1–104109-19
Raben SG, Charonko JJ, Vlachos PP (2012) Adaptive gappy proper orthogonal decomposition for particle image velocimetry data reconstruction. Meas Sci Technol 23(2):025303
Lophaven S, Nielsen H, Søndergaard J (2002) DACE - A Matlab Kriging Toolbox. Report IMM-REP-2002-12, informatics and mathematical modelling, Technical University of Denmark, p 34, tech. rep
Stamatopoulos C, Mathioulakis D, Papaharilaou Y, Katsamouris A (2011) Experimental unsteady flow study in a patient-specific abdominal aortic aneurysm model. Exp Fluids 50:1695–1709
Spence C, Buchmann N, Jermy M (2012) Unsteady flow in the nasal cavity with high flow therapy measured by stereoscopic PIV. Exp Fluids 52:569–579
Tecplot (2012) Tecplot 360 User’s Manual. http://www.tecplot.com
Karri S, Charonko J, Vlachos PP (2009) Robust wall gradient estimation using radial basis functions and proper orthogonal decomposition (pod) for particle image velocimetry (piv) measured fields. Mea Sci Technol 20(4):045401
Casa LDC, Krueger PS (2013) Radial basis function interpolation of unstructured, three-dimensional, volumetric particle tracking velocimetry data. Meas Sci Technol 24(6):065304
JC for Guides in Metrology (JCGM) (1995) Guide to the expression of uncertainty in measurement (GUM)
Timmins B, Wilson B, Smith B, Vlachos P (2012) A method for automatic estimation of instantaneous local uncertainty in particle image velocimetry measurements. Exp Fluids 53:1133–1147
Charonko JJ, Vlachos PP (2013) Estimation of uncertainty bounds for individual particle image velocimetry measurements from cross-correlation peak ratio. Meas Sci Technol 24(6):065301
Wikle CK, Berliner LM (2007) A Bayesian tutorial for data assimilation. Phys D Nonlinear Phenom 230(1–2):1–16
Mardia KV, Marshall RJ (1984) Maximum Likelihood Estimation of Models for Residual Covariance in Spatial Regression. Biometrika 71(1):135–146
Kitanidis PK, Lane RW (1985) Maximum likelihood parameter estimation of hydrologic spatial processes by the Gauss-Newton method. J Hydrol 79(1–2):53–71
Kitanidis PK (1986) Parameter uncertainty in estimation of spatial functions: Bayesian analysis. Water Resour Res 22(4):499–507
Mardia KV (1989) Maximum likelihood estimation for spatial models. Proceedings from the symposium on spatial statistics: past, present, and future, pp 203–253
de Baar J, Dwight R, Bijl H (2013) Speeding up kriging through fast estimation of the hyperparameters in the frequency-domain. Comput Geosci 54:99–106
Chiles J, Delfiner P (2012) Geostatistics, Modeling Spatial Uncertainty, 2nd Edn. Wiley, New York
de Baar J, Dwight R, Bijl H Improvements to Gradient-Enhanced Kriging using a Bayesian perspective, Int J Uncertain Quantif, In production
Matheron G (1971) Les Cahiers du Centre de Morphologie Mathematique: The theory of regionalized variables and its application. Ecole Nationale Superieure des Mines de Paris
Metropolis N, Ulam S (1949) The Monte Carlo method. J Am Stat Assoc, pp 335–341
Mori N (2009) mpiv. http://www.oceanwave.jp/softwares/mpiv
Keane RD, Adrian RJ (1992) Theory of cross-correlation analysis of piv images. Appl Sci Res 49(3):191–215
Witten I, Frank E, Hall M (2011) Data mining, practical machine learning tools and techniques, 3rd Edn. Morgan Kaufmann, Burlington
Percin M, Eisma HE, de Baar JHS, van Oudheusden BW, Remes B, Ruijsink R, de Wagter C (2012) Wake reconstruction of flapping-wing MAV DelFly II in forward flight. Int Micro Air Vehicle Conf Flight Compet, July 2012, Braunschweig Germany
Acknowledgments
We are very grateful for the financial support by Technologiestichting STW, project number 10113 and 11023.
Author information
Authors and Affiliations
Corresponding author
Electronic supplementary material
Below is the link to the electronic supplementary material.
Appendix: Detailed mpiv settings
Appendix: Detailed mpiv settings
For the cross-correlation of the synthetic images in Sect. 3, we use the following general settings: nx_window = 32; ny_window = 32; overlap_x = 0; overlap_y = 0; iu_max = 0; iv_max = 0; dt = 0.01; piv_type = ’cor’; i_recur = 2; i_plot = 0; By default mpiv assigns NaN to vectors with a SNR or PPR below 3.00, however, we want to control the SNR min top-level. Therefore, in the files piv_cor.m and piv_crs.m , we change these settings: r_SNR = 0.00; r_PPR = 0.00.
Rights and permissions
About this article
Cite this article
de Baar, J.H.S., Percin, M., Dwight, R.P. et al. Kriging regression of PIV data using a local error estimate. Exp Fluids 55, 1650 (2014). https://doi.org/10.1007/s00348-013-1650-z
Received:
Revised:
Accepted:
Published:
DOI: https://doi.org/10.1007/s00348-013-1650-z