Two-Phase Flow Pattern Identification by Embedding Double Attention Mechanisms into a Convolutional Neural Network

There are inevitable multiphase flow problems in the process of subsea oil-gas acquisition and transportation, which two-phase involving gas liquid is given much attention. The performance pipelines equipment systems greatly affected by various patterns. As a result, correctly efficiently identifying pattern pipeline critical for oil industry. In this study, two attention modules, convolutional block module (CBAM) efficient channel (ECA), introduced into neural network (ResNet50) to develop gas–liquid identification model, named CBAM-ECA-ResNet50. To verify accuracy efficiency proposed collection images vertical selected as dataset, data augmentation employed on training set enhance generalization capability comprehensive model. Then, comparison models similar model obtained adjusting order number modules positions inserting other different modules. Afterward, ResNet50 all applied classify identify images. CBAM-ECA-ResNet50 observed be highest (99.62%). addition, robustness complexity satisfactory.