Cluster-based reduced-order descriptions of two phase flows

Abstract

Dynamical system representations of two different multiphase flow regimes are extracted using a clustering algorithm. Experiments were conducted in a flow loop using an X-ray tomography system to obtain cross-sectional phase fraction fields of two distinct two-phase regimes: Dispersed flow and slug flow. Low-order representations are used for classification of the dynamics and construction of the cluster-based model through Markov chain theory. The number of partitions required to describe the flow provides insight into the flow nature and complexity, such as; the dispersed flow regime requires a larger number of clusters to present the full dynamical cycle than the slug flow regime. The flow features of each regime are obtained with the optimal number of clusters and the transition among these clusters highlights the hidden dynamics of the probabilistic model. Compressive measurements are achieved through the optimized point measurements to compute the transition dynamics and perform the classification task.