A boundary-element approach for the complete-electrode model of EIT illustrated using simulated and real data

Abstract

Electrical impedance tomography (EIT) allows non-invasive monitoring and control of dynamic industrial processes. Although reconstruction of tomographic images is useful for process visualization, for industrial process control, estimation of key process parameters is frequently more appropriate than visualization. Often such parameters comprise geometric information. The proposed direct problem solution method naturally encodes geometric information. The complete-electrode model for the direct problem of electrical tomography is solved with a boundary-element approach assuming a piecewise constant conductivity distribution with unknown conductivity and unknown interface location. The boundary-element discretization of the complete-electrode model is presented. In practical applications it is essential to provide an assessment of reliability of the estimates. To achieve this aim, confidence intervals are constructed in a manner which allows sufficient flexibility for the summary of non-Gaussian distributions. Hypothesis testing is proposed to detect changes in electrical impedance on the electrode surface which can be used to identify malfunctioning or poorly contacting electrodes. The proposed data analysis is applied to both computationally simulated and real experimental data.