Skip to main content
Fachgebiete
Automobil + Motoren
Bauwesen + Immobilien
Business IT + Informatik
Elektrotechnik + Elektronik
Energie + Nachhaltigkeit
Finance + Banking
Management + Führung
Marketing + Vertrieb
Maschinenbau + Werkstoffe
Versicherung + Risiko
DE
EN
Bücher
Zeitschriften
Weitere Formate
Webinare Technik
Webinare Wirtschaft
Kongresse
Veranstaltungskalender
Podcasts
Awards
Jetzt Einzelzugang starten
Zugang für Unternehmen beantragen
Erweiterte Suche
Anmelden
nach oben
Experiments in Fluids

Tipp

Weitere Artikel dieser Ausgabe durch Wischen aufrufen

Erschienen in: Experiments in Fluids 8/2019

01.08.2019 | Research Article

Singular value decomposition of noisy data: mode corruption

verfasst von: Brenden P. Epps, Eric M. Krivitzky

Erschienen in: Experiments in Fluids | Ausgabe 8/2019

Einloggen, um Zugang zu erhalten

Abstract

Although the singular value decomposition (SVD) and proper orthogonal decomposition have been widely used in fluid mechanics, Venturi (J Fluid Mech 559:215–254, 2006) and Epps and Techet (Exp Fluids 48:355–367, 2010) were among the first to consider how noise in the data affects the results of these decompositions. Herein, we extend those studies using perturbation theory to derive formulae for the 95% confidence intervals of the singular values and vectors, as well as formulae for the root mean square error (rmse) of each noisy SVD mode. Moreover, we show that the rmse is well approximated by \(\epsilon /\tilde{s}_k\) (where \(\epsilon\) is the rms noise and \(\tilde{s}_k\) is the singular value), which provides a useful estimate of the overall uncertainty in each mode.

Graphic abstract

Vorheriger Artikel A robust calibration method for seven-hole pressure probes
Nächster Artikel Extended flexible trailing-edge on the flow structures of an airfoil at high angle of attack

Sie möchten Zugang zu diesem Inhalt erhalten? Dann informieren Sie sich jetzt über unsere Produkte:

Springer Professional "Wirtschaft+Technik"

Online-Abonnement

Mit Springer Professional "Wirtschaft+Technik" erhalten Sie Zugriff auf:

  • über 102.000 Bücher
  • über 537 Zeitschriften

aus folgenden Fachgebieten:

  • Automobil + Motoren
  • Bauwesen + Immobilien
  • Business IT + Informatik
  • Elektrotechnik + Elektronik
  • Energie + Nachhaltigkeit
  • Finance + Banking
  • Management + Führung
  • Marketing + Vertrieb
  • Maschinenbau + Werkstoffe
  • Versicherung + Risiko

Jetzt 90 Tage mit der neuen Mini-Lizenz testen!

Jetzt informieren

Springer Professional "Technik"

Online-Abonnement

Mit Springer Professional "Technik" erhalten Sie Zugriff auf:

  • über 67.000 Bücher
  • über 390 Zeitschriften

aus folgenden Fachgebieten:

  • Automobil + Motoren
  • Bauwesen + Immobilien
  • Business IT + Informatik
  • Elektrotechnik + Elektronik
  • Energie + Nachhaltigkeit
  • Maschinenbau + Werkstoffe



 


Jetzt 90 Tage mit der neuen Mini-Lizenz testen!

Jetzt informieren
Anhänge
Nur mit Berechtigung zugänglich
Fußnoten
1
The SVD is related to the biorthogonal decomposition (Aubry 1991) and the method of empirical orthogonal functions (Loren 1956). The POD (Berkooz et al. 1993; Holmes et al. 1996, 1997) is related to the Karhunen–Loève transform (Karhunen 1946; Loève 1978), principal components analysis (Pearson 1901), the method of empirical eigenfunctions, and the method of snapshots (Sirovich 1987).
 
2
The number of data sites D is the number of individual pieces of data at each time step. For example, consider sampling two-dimensional velocity data on an \(I \times J\) grid of field points; then \(D = 2IJ\) is the total number of data sites.
 
3
Ideally, \(\mathbf{E}\) contains i.i.d. noise drawn from a Gaussian distribution, but herein we also consider \(\mathbf{E}\) containing spatially correlated noise, as occurs in PIV data.
 
4
In terms of the POD eigenvalues \({\tilde{\lambda }}_k = \tilde{s}_k^2\), the threshold criterion (3) requires \({\tilde{\lambda }}_k > \epsilon ^2 TD\).
 
5
Note that the reconstructed singular values \(\bar{s}_k\) could be used in place of the noisy ones \(\tilde{s}_k\), but we find this makes little difference in the predicted rmse.
 
6
Proof: since \(\mathbf{U}\) is orthogonal (\(\mathbf{U}\mathbf{U}^\intercal = \mathbf{I}\)), we can write (41) as \(\mathbf{H}= \mathbf{U}{\varvec{{\Lambda }}} \mathbf{U}^\intercal\). At the same time, \(\mathbf{H}= \mathbf{A}\mathbf{A}^\intercal = \mathbf{U}\mathbf{S}\mathbf{V}^\intercal \mathbf{V}\mathbf{S} \mathbf{U}^\intercal = \mathbf{U}\mathbf{S}^2 \mathbf{U}^\intercal\).
 
7
Kato uses the notation: \(\chi\), \(\mathbf{T}\), and \(\mathbf{T}(\chi )\) for \(\epsilon\), \(\mathbf{H}\), and \({\tilde{\mathbf{H}}}\), respectively. Kato and Venturi use \(\mathbf{S}\) for \(\mathbf{Q}\).
 
8
If \(\mathbf{H}\) has repeated eigenvalues, then Eq. (84) represents the weighted mean of such eigenvalues. In this case, the present theory then needs to be modified (via Kato’s reduction theory). However, these modifications complicate the analysis and prevent one from simplifying the results into forms as simple as, for example, equation (87).
 
9
Although matrix \(\mathbf{W}^{(1)}\) refers to mode k, we have omitted the subscript k to facilitate referring to its \(im\text {th}\) element as \(W^{(1)}_{im}\). The \(i\text {th}\) element of vector \({\tilde{\mathbf{u}}}_k\) is \(\tilde{U}_{ik}\).
 
10
Note that all odd “powers” of \(\mathbf{E}\) average to zero, so \(\langle W^{(1)}_{im}U_{mk} \rangle = \langle W^{(3)}_{im}U_{mk} \rangle = \big \langle ( W^{(1)}_{im}U_{mk} ) \, ( W^{(2)}_{in}U_{nk} ) \big \rangle = \dots = 0\).
 
11
Note that again all odd “powers” of \(\mathbf{E}\) average to zero, so \(\langle N^{(1)}_{im}V_{mk} \rangle = \langle N^{(3)}_{im}V_{mk} \rangle = \dots = 0\).
 
12
The author prefers to interpolate using a piecewise cubic Hermite interpolating polynomial, pchip, because it provides continuity of the function and its first derivative while not being susceptible to overshoots as in a cubic spline. In Matlab, \(g' = \texttt {pchip}(x,g, x')\) returns g(x) evaluated at \(x'\).
 
Literatur
Aubry N (1991) On the hidden beauty of the proper orthogonal decomposition. Theor Comput Fluid Dyn 2:339–352 CrossRef
Beltrami E (1873) Sulle funzioni bilineari. English translation by D. Boley is available as Techical Report 90-37, University of Minnesota Department of Computer Science, Minneapolis, MN, 1990
Benaych-Georges F, Nadakuditi RR (2011) The eigenvalues and eigenvectors of finite, low rank perturbations of large random matrices. Adv Math 227:494–521 MathSciNetCrossRef
Berkooz G, Holmes P, Lumley JL (1993) The proper orthogonal decomposition in the analysis of turbulent flows. Annu Rev Fluid Mech 25:539–575 MathSciNetCrossRef
Breuer K, Sirovich L (1991) The use of the Karhunen-Loève procedure for the calculation of linear eigenfunctions. J Comput Phys 96:277–296 MathSciNetCrossRef
Brindise MC, Vlachos PP (2017) Proper orthogonal decomposition truncation method for data denoising and order reduction. Exp Fluids 58(4):28 CrossRef
Cagney N, Balabani S (2013) On multiple manifestations of the second response branch in streamwise vortex- induced vibrations. Phys Fluids 25(7):075110 CrossRef
Charonko JJ, King CV, Smith BL, Vlachos PP (2010) Assessment of pressure field calculations from particle image velocimetry measurements. Meas Sci Technol 21(10):105401 CrossRef
Cohen K, Siegel S, McLaughlin T, Gillies E (2003) Feedback control of a cylinder wake low-dimensional model. AIAA J 41(7):1389–1391 CrossRef
Dawson STM, Hemati MS, Williams MO, Rowley CW (2016) Characterizing and correcting for the effect of sensor noise in the dynamic mode decomposition. Exp Fluids 57(3):42 CrossRef
Druault P, Bouhoubeiny E, Germain G (2012) POD investigation of the unsteady turbulent boundary layer developing over porous moving flexible fishing net structure. Exp Fluids 53:277–292 CrossRef
Epps BP, Krivitzky EM (2019) Singular value decomposition of noisy data: noise filtering. Exp Fluids (accepted)
Epps BP, Techet AH (2010) An error threshold criterion for singular value decomposition modes extracted from PIV data. Exp Fluids 48:355–367 CrossRef
Feng LH, Wang JJ, Pan C (2011) Proper orthogonal decomposition analysis of vortex dynamics of a circular cylinder under synthetic jet control. Phys Fluids 23(1):014106 CrossRef
Gandhi V, Bryant DB, Socolofsky SA, Stoesser T, Kim JH (2015) Concentration-based decomposition of the flow around a confined cylinder in a UV disinfection reactor. J Eng Mech 141(12):04015050 CrossRef
Holden D, Socha JJ, Cardwell ND, Vlachos PP (2014) Aerodynamics of the flying snake chrysopelea paradisi: how a bluff body cross-sectional shape contributes to gliding performance. J Exp Biol 217(3):382–394 CrossRef
Holmes P, Lumley JL, Berkooz G (1996) Turbulence, coherent structures, dynamic systems, and symmetry. Cambridge University Press, Cambridge CrossRef
Holmes PJ, Lumley JL, Berkooz G, Mattingly JC, Wittenberg RW (1997) Low-dimensional models of coherent structures in turbulence. Phys Rep 287:337–384 MathSciNetCrossRef
Jordan C (1874a) Mémoire sur les formes bilinéaires. J Math Pures Appl 19:35–54 MATH
Jordan C (1874b) Sur la réduction des formes bilinéaires. Comptes Rend Acad Sci 78:614–617 MATH
Kato T (1976) Perturbation theory for linear operators. Springer, Berlin MATH
Kourentis L, Konstantinidis E (2012) Uncovering large-scale coherent structures in natural and forced turbulent wakes by combining PIV, POD, and FTLE. Exp Fluids 52:749–763 CrossRef
Kriegseis J, Dehler T, Pawlik M, Tropea C (2009) Pattern-identification study of the flow in proximity of a plasma actuator. In: 47th AIAA aerospace sciences meeting, p 1001
Li RC (1998) Relative perturbation theory: (i) eigenvalue and singular value variations. SIAM J Matrix Anal Appl 19(4):956–982 MathSciNetCrossRef
Loève M (1978) Probability theory. Springer, Berlin MATH
Lorenz EN (1956) Empirical orthogonal functions and statistical weather prediction. Tech. rep., MIT
Marchenko VA, Pastur LA (1967) Distribution of eigenvalues for some sets of random matrices. Mat Sbornik 114(4):507–536
Marié S, Druault P, Lambaré H, Schrijer F (2013) Experimental analysis of the pressure-velocity correlations of external unsteady flow over rocket launchers. Aerosp Sci Technol 30:83–93 CrossRef
Mokhasi P, Rempfer D (2004) Optimized sensor placement for urban flow measurement. Phys Fluids 16(5):1758–1764 CrossRef
Nguyen TD, Wells JC, Mokhasi P, Rempfer D (2010) POD-based estimations of the flowfield from PIV wall gradient measurements in the backward-facing step flow. In: Proceedings of ASME 2010 3rd joint US-European fluids engineering summer meeting and 8th international conference on nanochannels, microchannels, and minichannels
Pearson K (1901) LIII on lines and planes of closest fit to systems of points in space. Lond Edinburgh Dublin Philos Mag J Sci 2(11):559–572 CrossRef
Rajaee M, Karlsson S, Sirovich L (1994) Low-dimensional description of free-shear-flow coherent structures and their dynamical behaviour. J Fluid Mech 258:1–29 CrossRef
Rowley CW, Mezic I, Bagheri S, Schlatter P, Henningson D (2009) Spectral analysis of nonlinear flows. J Fluid Mech 641:115–127 MathSciNetCrossRef
Schmidt E (1907) Zur theorie der linearen und nichtlinearen integralgleichungen. I teil. Entwicklung willkurlichen funktionen nach system vorgeschriebener. Math Annal 63:433–476 CrossRef
Sirovich L (1987) Turbulence and the dynamics of coherent structures. Part 1: coherent structures, Part 2: symmetries and transformations, Part 3: dynamics and scaling. Q Appl Math 45:561–590 CrossRef
Stewart GW (1978) A note on the perturbations of singular values. Tech. Rep. TR-720, University of Maryland
Stewart GW (1990) Perturbation theory for the singular value decomposition. Technical Report UMIACS-TR-90-124, CS-TR 2539, University of Maryland
Strang (2009) Introduction to linear algebra, 4th edn. Wellesley-Cambridge Press, Wellesley MATH
Sylvester JJ (1889a) A new proof that a general quadric may be reduced to its canonical form (that is, a linear function of squares) by means of a real orthogonal substitution. Messenger Math 19:1–5 MATH
Sylvester JJ (1889b) On the reduction of a bilinear quantic of the nth order to the form of a sum of n products by a double orthogonal substitution. Messenger Math 19:42–46
Utturkar Y, Zhang B, Shyy W (2005) Reduced-order description of fluid flow with moving boundaries by proper orthogonal decomposition. Inte J Heat Fluid Flow 26:276–288 CrossRef
Venturi D (2006) On proper orthogonal decomposition of randomly perturbed fields with applications to flow past a cylinder and natural convection over a horizontal plate. J Fluid Mech 559:215–254 MathSciNetCrossRef
Venturi D, Karniadakis GE (2004) Gappy data and reconstruction procedures for flow past a cylinder. J Fluid Mech 519:315–336 MathSciNetCrossRef
Weyl H (1912) Das asymptotische verteilungsgesetz der eigenwerte linearer partieller differentialgleichungen (mit einer anwendung auf die theorie der hohlraumstrahlung). Math Annal 71:441–479 MathSciNetCrossRef
Yildirim B, Chryssostomidis C, Karniadakis G (2009) Efficient sensor placement for ocean measurements using low-dimensional concepts. Ocean Model 27:160–173 CrossRef
Metadaten
Titel
Singular value decomposition of noisy data: mode corruption
verfasst von
Brenden P. Epps
Eric M. Krivitzky
Publikationsdatum
01.08.2019
Verlag
Springer Berlin Heidelberg
Erschienen in
Experiments in Fluids / Ausgabe 8/2019
Print ISSN: 0723-4864
Elektronische ISSN: 1432-1114
DOI
https://doi.org/10.1007/s00348-019-2761-y

Weitere Artikel der Ausgabe 8/2019

Experiments in Fluids 8/2019 Zur Ausgabe

Research Article

High resolution volumetric dual-camera light-field PIV

Research Article

Effect of bubble distribution on wall drag in turbulent channel flow

Research Article

Extended flexible trailing-edge on the flow structures of an airfoil at high angle of attack

Letter

Wavelet-based algorithm for correction of beam-steering artefacts in turbulent flow imaging at elevated pressures

Research Article

Analysis of a textural atomization process

Research Article

A scanning particle tracking velocimetry technique for high-Reynolds number turbulent flows

Premium Partner

MKVS GbR Nordson ViscoTec Hilger u. Kern GmbH BCD Chemie GmbH bdtronic GmbH Merz+Benteli TARTLER GmbH Robatech Hermann Otto GmbH Ruderer Klebetechnik GmbH Henkel AG Fere Xometry

    Marktübersichten

    Die im Laufe eines Jahres in der „adhäsion“ veröffentlichten Marktübersichten helfen Anwendern verschiedenster Branchen, sich einen gezielten Überblick über Lieferantenangebote zu verschaffen. 

    Zur Marktübersicht
    Bildnachweise