Cluster-Based Reduced-Order Modeling of Dispersed and Slug Flows

Two-phase dispersed and slug flows in a pipe are investigated using a cluster-based reduced-order analysis technique. The data are obtained via 2-D construction of X-ray computed tomography measurements. Moving-average is used to obtain the fluctuating phase fraction of the periodic slug flow field. Flow fields are analyzed and reduced-order descriptions are achieved. A cluster-based reduced-order model algorithm is implemented on the phase fraction field and the proper orthogonal decomposition time coefficients. The algorithm partitions the basis into clusters, minimizing distances between the data points inside a cluster and maximizing the distances between the clusters. Cluster-based reduced-order model links a cluster analysis and a Markov chain model. The dynamical model is presented based on the transition process between the extracted clusters. Results display coherent features near the center of the pipe, the liquid-liquid interface, for the dispersed flow case while the slug flow shows coherent structures that correspond to the periodic formation of the slug.