Intelligent Flow Pattern Recognition of Horizontal Gas-Liquid Two-Phase Flow Utilizing Distributed Optical Carrier-Based Microwave Interferometry

Accurate identification of gas-liquid two-phase flow patterns is crucial for the safety and efficiency of industrial production and energy transportation. This paper proposes a novel method for horizontal gas-liquid two-phase flow pattern recognition based on distributed optical carrier-based microwave interferometry (OCMI) technology and convolutional neural network (CNN) architecture. A theoretical model is established to characterize the force acting on distributed sensing fiber in gas-liquid two-phase flow for the first time. Subsequently, the strain and effective refractive index change in dual-coated sensing fiber under fluid-induced force are modeled. Furthermore, a response model is proposed for OCMI signal under gas-liquid two-phase flow, and related numerical simulation studies are conducted. Based on a specially designed deployment scheme of distributed sensing fiber in the pipeline, the OCMI signals at various sampling points under different flow patterns are experimentally acquired. A one-dimensional CNN (1D-CNN) is constructed to perform convolution along the scanning frequency dimension, and a frequency-spatial CNN (FSCNN) is developed to perform convolution separately along the frequency dimension and spatial dimension. The results demonstrate that the FSCNN enables more effective identification of gas-liquid two-phase flow patterns, achieving an accuracy higher than 91%. The proposed models and schemes provide valuable theoretical and practical references for the research and application of distributed fiber sensing technologies in multiphase complex flow systems.