[en] We propose a complex network-based method to distinguish complex patterns arising from experimental horizontal oil–water two-phase flow. We first use the adaptive optimal kernel time–frequency representation (AOK TFR) to characterize flow pattern behaviors from the energy and frequency point of view. Then, we infer two-phase flow complex networks from experimental measurements and detect the community structures associated with flow patterns. The results suggest that the community detection in two-phase flow complex network allows objectively discriminating complex horizontal oil–water flow patterns, especially for the segregated and dispersed flow patterns, a task that existing method based on AOK TFR fails to work. - Highlights: • We combine time–frequency analysis and complex network to identify flow patterns. • We explore the transitional flow behaviors in terms of betweenness centrality. • Our analysis provides a novel way for recognizing complex flow patterns. • Broader applicability of our method is demonstrated and articulated