This work provides an in-depth investigation into the use of a knowledge-guided tensor decomposition method to compactly represent coherent structures from complex and turbulent flowfields. By utilizing a guided decomposition, the method is able to produce components that are useful for the identification and discovery of spatiotemporal structures in the flow where an unguided tensor decomposition cannot. Through three investigative cases—flow over a stationary cylinder, flow over a NACA 4412 airfoil, and turbulent channel flow—we demonstrate the ability of knowledge-guided tensor decomposition to produce components that represent coherent structures in the flow and gain much insight into the characteristics of the knowledge-guided tensor decomposition method presented here. First, knowledge-guided tensor decomposition provides tensor components better suited for the identification of coherent structures in the flow compared to those provided by unguided tensor decomposition. The components created by knowledge-guided tensor decomposition have spatial modes that are modulated by temporal modes and are organized by physical characteristics, such as turbulent kinetic energy or frequency content, which is not the case for unguided tensor decomposition. Furthermore, it is shown that providing guidance information reduces the variability of the results. In general, knowledge-guided tensor decomposition shows less variability to the initial guess compared to unguided tensor decomposition. These benefits are accompanied by improved or equal accuracy in the prediction of aerodynamic coefficients as well.


    Zugriff

    Zugriff prüfen

    Verfügbarkeit in meiner Bibliothek prüfen

    Bestellung bei Subito €


    Exportieren, teilen und zitieren



    Titel :

    Knowledge-Guided Tensor Decomposition for Approximation of and Discovery in Complex and Turbulent Flows


    Beteiligte:
    Coley, Christopher (Autor:in) / Metzler, Joy (Autor:in)

    Erschienen in:

    Erscheinungsdatum :

    01.04.2025




    Medientyp :

    Aufsatz (Zeitschrift)


    Format :

    Elektronische Ressource


    Sprache :

    Englisch





    Statistical Analysis of the Knowledge-Guided Tensor Decomposition

    Metzler, Joy J. / Coley, Christopher | AIAA | 2023




    Large eddy simulation of complex turbulent flows

    Weber, C. / Ducros, F. / Corjon, A. | AIAA | 1998