Jeong Seong Lee

Phase boundary estimation in electrical impedance tomography using the Hooke and Jeeves pattern search method

In industrial processes, monitoring of heterogeneous phases is crucial to the safety and operation of the engineering structures. Particularly, the visualization of voids and air bubbles is advantageous. As a result many studies have appeared in the literature that offer varying degrees of functionality. Electrical impedance tomography (EIT) has already been proved to be a hallmark for process monitoring and offers not only the visualization of the resistivity profile for a given flow mixture but is also used for detection of phase boundaries. Iterative image reconstruction algorithms, such as the modified Newton–Raphson (mNR) method, are commonly used as inverse solvers. However, their utility is problematic in a sense that they require the initial solution in close proximity of the ground truth. Furthermore, they also rely on the gradient information of the objective function to be minimized. Therefore, in this paper, we address all these issues by employing a direct search algorithm, namely the Hooke and Jeeves pattern search method, to estimate the phase boundaries that directly minimizes the cost function and does not require the gradient information. It is assumed that the resistivity profile is known a priori and therefore the unknown information will be the size and location of the object. The boundary coefficients are parameterized using truncated Fourier series and are estimated using the relationship between the measured voltages and injected currents. Through extensive simulation and experimental result and by comparison with mNR, we show that the Hooke and Jeeves pattern search method offers a promising prospect for process monitoring.

An approximate formula for the capacitance–void fraction relationship for annular flows

A capacitance method is applied to measure the space-averaged void fraction in concentric annular flows. From the analytical solution of the governing electrical field equations, the expression for the capacitance in terms of a given void fraction is derived. Also, a closed form formula to predict the void fraction from the capacitance measurement is proposed. The relationship between the capacitance and the void fraction is successfully compared with static phantom experiments.

Particle swarm optimization technique for elliptic region boundary estimation in electrical impedance tomography

In this study we consider the recovery of smooth elliptic region boundary in electrical impedance tomography (EIT). The assumption made is that the resistivity profile is known a priori but some of the geometric information is missing. This missing information may for example be shape, size and location. This leads to a nonlinear ill‐posed inverse problem. In this study we propose a population based parallel evolutionary computation technique, particle swarm optimization (PSO). We formulate the forward problem as mapping of a set of Fourier coefficients representing boundary shapes to the resistivity profile in the domain. Then PSO is used to iteratively seek the boundary configuration minimizing a cost functional. The final goal of this study is to recover the region boundary with only one measurement data (i.e., one current pattern) contaminated with measurement noise along with a very high contrast ratio. The simulation results are also provided to assess the merit of PSO technique.

Electrical Resistance Imaging of Two‐Phase Flow Through Rod Bundles

This paper considers the dynamic electrical resistance tomography (ERT) for the visualization of phase distribution in heat transfer systems with rod bundles. Typical measurement techniques like conductivity probes, ultrasonic and radiological imaging suffer from difficulties in the measurements due to these internal structures. On the other hand ERT imaging does not get affected by the presence of internal structures. These internal structures can be used as internal electrodes to further improve the dynamic ERT performance. This paper discusses the current injection protocols for such systems. The main goal is to find the optimal current patterns to reconstruct the conductivity distribution. We provide extensive simulation results to evaluate the performance of different current patterns.