A novel approach for measuring the void fraction in stratified air-water systems utilizing an 8-blade capacitance-based sensor, sinogram, and a deep neural network

Measuring void fraction (VF) in a pipeline is crucial for ensuring operational efficiency, safety, and environmental responsibility in various engineering applications.There are several methods commonly used to measure VF in multiphase flow systems.Capacitance sensors are a dependable and practical option for measuring VF, providing benefits such as versatility, sensitivity, cost-effectiveness, and ease of use.In this study, simulations were performed to produce different VF levels of an air-water stratified two-phase flow, ranging across 31 distinct VF values from completely full to entirely empty.Moreover, an 8-blade concave capacitive sensor was designed and utilized for VF measurements.In order to use the power of the finite element method (FEM), COMSOL multiphysics was employed to produce the desired void fractions and measure the capacitance value of each pair of electrodes.The capacitance values of these electrode pairs were measured, resulting in the creation of sinograms corresponding to different VF.These sinograms were utilized as inputs for a deep neural network (DNN) developed in Python, specifically a Multilayer Perceptron model, to estimate VFs.Furthermore, to enhance user understanding, sinograms were employed to reconstruct fluid images using the back-projection method.The results demonstrated an accuracy of 0.002, a significant improvement over previous methodologies in VF measurement.