Signal optimization for recognition of gas–liquid two-phase flow regimes in a long pipeline-riser system

Abstract

Accurate and fast recognition of multiphase flow regimes is an urgent requirement for the flow assurance of oil/gas pipelines. Experiments on gas–liquid flow regimes are conducted on a 1657 m horizontal pipeline system with a 16.7 m S-shaped riser. The flow regimes are classified as severe slugging, oscillating flow and stable flow, based on quantitative criteria of riser pressure difference. Data from the accumulation stage of severe slugging is selected for training the model, and a new optimization scheme is proposed for the recognition process from feature extraction and selection to model construction and testing. The influence of the measurement distance and location on the recognition rate is revealed, and a new set of signal evaluation and optimization criteria is proposed, covering the recognition rate (higher than 90%), reliability and practicality. An accurate recognition rate over 90% is yielded by using above-water signals with a sample duration of 6.2 s.