Flow Features Recognition of Horizontal Two-Phase Flow Instability Based on Machine Learning

When the passive heat removal system discharges the core residual heat, the steam in the pipeline condenses due to the direct contact with the low-temperature water, resulting in the instability of the gas-liquid interface and even the condensation-induced water hammer (CIWH). To investigate the capacity for heat and mass transfer in direct contact condensation, it is necessary to identify the two-phase flow pattern in the pipeline. Though the visual experimental device, the two-phase flow image in the pipeline is collected. The mean average of the gray value of each line of the local image is extracted as the flow feature, and the BP neural network method is used to identify the local images of different features. The characteristics of different flow patterns are analyzed by using the flow pattern feature map of the continuous image in time. Results indicate that the BP neural network has the ability to accurately identify three types feature, exhibiting an accuracy rate of 95%. The feature map demonstrates distinguishable characteristics of various flow patterns.