DETERMINATION OF VOLUMETRIC FRACTION IN OIL-WATER TWO-PHASE FLOW THROUGH WIRE-MESH SENSOR AND ARTIFICIAL NEURAL NETWORKS

A technique to measure volumetric fraction of liquids in oil-water pipe flow based on electrical permittivity obtained by a homemade wire-mesh sensor aided by an artificial neural network is presented in this work. The artificial neural networks use the superficial velocities of fluids as inputs and the volumetric fraction based on models of electrical permittivity as outputs. In order to validate the results, they were compared with data obtained with the technique of quick closing valves. Furthermore, two-phase-flow tomographic images were compared and validated with images recorded by a high-speed video camera. The experiments were carried out in a test line formed by a glass pipe of 25.4 mm i.d. and 12 meters long, where different flow patterns were observed. The working fluid were oil of viscosity 220 mPa.s and density of 828 kg/m3 and tap water. The results showed the great potential of using a feedforward backpropagation neural network to improve the capabilities of the wire-mesh sensor to obtain important information about the characteristics of liquid-liquid two-phase flow, such as the volumetric fraction and tomographic images.