Identification of high void fraction flows using conductivity measurements and machine learning techniques

Abstract

This study explores the application of a droplet-capable conductivity probe (DCCP-4) combined with a Kohonen self-organizing map (SOM) for identifying flow regimes in high void fraction flows. The DCCP-4, designed to accurately measure the electrical conductivity of multiphase flows, is adept at capturing the intricate details of droplet behavior within these systems. The research demonstrates the effectiveness of the combined DCCP-4 and SOM method to identify different types of annular flows. The integration of the DCCP-4 probe's precise measurements with the SOM's robust clustering capabilities results in an advanced methodology for flow regime identification. The unsupervised neural network is used to categorize high void fraction flow data into three

flow regimes. High-speed camera footage is also employed to visually corroborate the findings. Statistical distributions related to droplet, bubble, and ligament measurements are also presented to further highlight the differences between the flow regimes. This approach not only enhances the accuracy of flow characterization in multiphase systems but also provides valuable insights into the underlying physical phenomena driving these flows. The findings have significant implications for optimizing industrial processes where high void fraction flows are prevalent to safety, such as in chemical reactors, oil and gas pipelines, and nuclear reactors, by improving monitoring and control strategies.