Image reconstruction by electrical impedance tomography and void-fraction estimation by artificial neural network for behavior analysis of gas-liquid flow

Abstract

Gas-liquid two-phase flow are frequently encountered phenomena in various engineering field, and monitoring of flow condition in pipeline is required for accident prevention and flow control in the application field. The objectives of this research are the estimation of void-fraction and comparison with the performance by the two methods. One is estimation by image reconstruction using the electrical impedance tomography method. The other is estimation of void-fraction by comparing pre-training data using a neural network. Training and image reconstruction of the artificial neural network were performed based on current data obtained through multiple electrodes phantom for void-fraction and the results were compared.