Year: 2017
Type: conference-paper
Abstract: Experimental and numerical simulations provide nowadays a great amount of detailed spatio-temporal data, which needs to be properly examined to achieve a better description of the turbulence dynamics.New investigative tools are hence continuously required to handle and properly interpret such big-data.In this context, complex network theory -by combining graph theory and statistical physics -recently turned out to be a powerful framework to analyze complex systems, such as turbulent flows [4], [5].In this work, a DNS of a fully-developed turbulent channel flow [1],[3] is investigated through the natural visibility graph (NVG) method [2].A subset of the simulation grid domain is firstly selected, acquiring all the available temporal data for the velocity field, (u, v, w), and for the kinetic energy, K.The time-series of each selected grid-point is then mapped into a network by means of the NVG method.In particular, two data values constitute a pair of linked nodes of the network if the straight line connecting the two data points lies above the other in-between data.The degree centrality, k, quantifying the visibility of nodes, is the first metric studied.The transitivity, T r, and the newly introduced mean link-length, d 1n , are then evaluated as indicators of the inter-visibility and mean temporal distance among nodes, respectively.The metrics are averaged along the directions of homogeneity of the flow (i.e., x and z), thus they only depend on the wall-normal coordinate, y + (see Fig. 1a).The visibility-based networks inherit the temporal structure of the corresponding time-series, as we observe the trend of the metrics is closely related to the flow properties along y + .In this way, different temporal features of the time-series are mapped in the networks and the metric trends (Fig. 1a) allow one to shed light on how the temporal structure of the series changes moving along y + (see Fig. 1b).Although intrinsically simple to be implemented, the visibility graph-based approach then offers a promising support to the classical methods for accurate time-series analyses of inhomogeneous turbulent flows.
Source: PORTO@Iris
Language: English
Subfield: Computational Mechanics
Field: Engineering
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W2752066942