Global–Local Structure Analysis-Based Status Monitoring for Oil–Gas–Water Flow

Water-based dispersed (WD) oil-gas–water three-phase flow is a complex nonlinear, dynamic, and multiscale process, which is widely encountered in oil industries. Focusing on the global and local structures embedded in flow status, a graph-based global-local statistical pattern analysis (GGLSPA) framework is proposed for status monitoring. Statistics pattern analysis (SPA) is used for multisensor feature extraction. On this basis, the global variance information of three typical flow statuses is extracted through principal component analysis (PCAs), outlining the external shape of flow status ensembles. The local neighborhood information is simultaneously extracted through local preserving projections (LPP) to retain the detailed differences of distinct flow statuses. Moreover, a novel local monitoring indicator named weighted k-nearest neighbors-based local ratio density (WKNN-Based LRD) is constructed to realize global-local joint monitoring together with the global monitoring statistic Hotelling’s T-squared (<inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">${T}^{{2}}$ </tex-math></inline-formula>). Dynamic experiments show that the proposed framework effectively addresses the problem of confusion between the WD bubbly flow status, the local status of liquid-plug in WD plug flow, and the local status of liquid-slug in WD slug flow for typical flow statuses. In addition, for transitional flow statuses with strong dynamics and continuity, a temporal fluctuation indicator S is introduced to assist the global-local spatial monitoring result, further achieving effective characterization of transitional flow process.