IOP PUBLISHING MEASUREMENT SCIENCE AND TECHNOLOGY
Meas. Sci. Technol. 21 (2010) 015502 (6pp) doi:10.1088/0957-0233/21/1/015502
Electrical capacitance tomography with a
non-circular sensor using the dbar method
Zhang Cao
1
,LijunXu
1
, Wenru Fan
2
and Huaxiang Wang
2
1
School of Instrumentation and Opto-Electronic Engineering, Beihang University, Beijing 100191,
People’s Republic of China
2
School of Electrical Engineering and Automation, Tianjin University, Tianjin 300072, People’s
Republic of China
E-mail: zh cao@hotmail.com, lijunxu@buaa.edu.cn, fanwenru@tju.edu.cn and hxwang@tju.edu.cn
Received 17 August 2009, in final form 5 October 2009
Published 27 November 2009
Online at stacks.iop.org/MST/21/015502
Abstract
In this paper, a new treatment of the dbar method for electrical capacitance tomography with a
non-circular sensor is presented. It is a direct algorithm of image reconstruction and the gray
value at any pixel of the reconstructed image is obtained directly and independently. The
major calculations of image reconstruction in the dbar method were implemented in a circular
sensor. As a result, complicated calculations of the scattering transform required by a
non-circular sensor were avoided. A unique conformal transformation was used to map a
circular sensor onto a non-circular sensor, e.g. a square sensor. In the experiments, a square
sensor was constructed and the reconstructed results validated the feasibility and effectiveness
of the dbar method in electrical capacitance tomography with a non-circular sensor.
Keywords: dbar method, conformal transformation, non-circular sensor, electrical
capacitance tomography
(Some figures in this article are in colour only in the electronic version)
1. Introduction
Electrical capacitance tomography (ECT) is a technique that
uses a set of capacitances measured by a multi-electrode sensor
array surrounding an industrial vessel or a pipe to reconstruct
the distribution of permittivity inside. It has been proved to
be a very suitable technique for imaging processes involving
multiphase flows (Olmos et al 2008, Wang et al 2008a). The
main difficulty for ECT sensors is the ‘soft field’ nature of
the problem. The core of the ECT technique is to estimate
the material property at any position in the region of interest.
In the literature, most of the image reconstruction algorithms
are iteratively implemented based on the sensitivity theorem
(Geselowitz 1971). Some existing non-iterative methods are
usually simplifications of the iterative ones, e.g. pre-iteration
method (Wang et al 2004), linear back projection (LBP) (Yang
and Peng 2003), etc. In recent years, direct algorithms have
been introduced to electrical impedance tomography (EIT) by
using the nonlinear Fourier transforms (Isaacson et al 2004)
or the value range identification (Schmitt 2009), termed as
the dbar method and the factorization method, respectively.
The factorization method is only suitable to reconstruct the
disturbed distributions with a connected background, e.g.
multi-rods. If the region to be reconstructed is of a ring shape,
for example, the factorization method fails to work (Schmitt
2009). The dbar method has no requirement on the shape
of the region of interest. However, the sensor configurations
used in EIT are different from those applied in ECT. The dbar
method implemented in EIT requires a sinusoidal wave-like
current pattern, in which all the values of the current injected
at each electrode generate a discrete approximation of the sine
function (Isaacson et al 2004). The ECT sensor has an earthed
screen; each electrode is usually either excited or grounded for
hardware implementation, so the dbar method implemented in
EIT cannot be applied directly to ECT. Moreover, the sensors
and associated reconstruction techniques in ECT are often of
and for circular shapes. There exists a demand for non-circular
sensors, e.g. the cross-sections of pipes in the power industry
or chemical reactors are square or rectangular (Wang et al
2008b,Liuet al 2008).
No result of ECT using the dbar method has been reported
yet. The dbar method requires the calculation of the scattering
0957-0233/10/015502+06$30.00 1 © 2010 IOP Publishing Ltd Printed in the UK