An Intelligent Approach to Determine Component Volume Percentages in a Symmetrical Homogeneous Three-Phase Fluid in Scaled Pipe Conditions

Abstract

Over time, the accumulation of scale within the transmission pipeline results in a decrease in the internal diameter of the pipe, leading to a decline in efficiency and energy waste. The employment of a gamma ray attenuation system that is non-invasive has been found to be a highly precise diagnostic technique for identifying volumetric percentages across various states. The most appropriate setup for simulating a volume percentage detection system through Monte Carlo N particle (MCNP) simulations involves a system consisting of two NaI detectors and dual-energy gamma sources, namely 241Am and 133Ba radioisotopes. A three-phase flow consisting of oil, water, and gas exhibits symmetrical homogenous flow characteristics across varying volume percentages as it traverses through scaled pipes of varying thicknesses. It is worth mentioning that there is an axial symmetry of flow inside the pipe that creates a homogenous flow pattern. In this study, the experiment involved the emission of gamma rays from one end of a pipe, with photons being absorbed by two detectors located at the other end. The resulting data included three distinct features, namely the counts under the photopeaks of 241Am and 133Ba from the first detector as well as the total count from the second detector. Through the implementation of a two-output MLP neural network utilising the aforementioned inputs, it is possible to accurately forecast the volumetric percentages with an RMSE of under 1.22, regardless of the thickness of the scale. The minimal error value ensures the efficacy of the proposed technique and the practicality of its implementation in the domains of petroleum and petrochemicals.

Keywords

• three-phase
• symmetrical homogenous flow
• volumetric percentage
• MLP neural network
• scale thickness independent