A dynamics-based tool for the analysis of experimental two-phase flows

This paper aims at presenting a novel approach for the analysis of experimental void fraction time series detected from two phase flows and to flow pattern identification. The main scope is to address the complexity of the observed dynamics on the basis of the representation in phase space of the attractors of the experimental time series, allowing an appropriate description of the complex structure of the nonlinear behaviours of the process and, eventually, a systematic research of hints of a possible chaotic source of the system dynamics. rst step of the proposed approach is the reconstruction of an n-dimensional representation state space on the basis of Takens’ theorem; the complex but regular attractors obtained in this way are noisy, mainly as a consequence of the high order dynamics associated to the secondary flow of small dispersed bubbles. ore, as a second step, Principal Component Analysis (PCA), also called Singular Value Decomposition (SVD), has been applied to the n-dimensional state space in order to determine the singular values of the state space and to project the attractor onto a new space spanned by the principal vectors. In this way it is possible to separate the dominant features of the system dynamics from noise-like dynamics, and to obtain unfolded phase portraits of the various flow patterns. As a final step, in order to achieve a deeper understanding, the attractors in the principal component phase portrait has been analysed by means of Poincaré maps, which have led to the observation of low order system dynamics.