Nonstationary bubble shape determination in Electrical Impedance Tomography combining Gauss–Newton Optimization with particle filter

Abstract Electrical Impedance Tomography (EIT) is a technique utilized in multiphase flow for identifying flow patterns and void-fraction calculation. An image made of the conductivity of the domain is formed by combining an inverse problem and measurements. Recursive statistical inversion, such as the particle filters, is scarce when dealing with EIT. For that matter, it is proposed a recursive Gauss–Newton Optimization Sequential Importance Resampling (GNOSIR) filter aimed to reconstruct vertical slug flow. It combines the Optimization with an Image Processing algorithm to improve the prior information when the posterior density is far from likelihood. The Level Set Method describes the contour of the inclusion using an elliptical Level Set Function to avoid the curse of dimensionality. Results demonstrate that the GNOSIR filter improves the estimation when there is a sudden and fast evolution of the inclusion’s contour between states.