A comparison of methods for in situ discrimination of imaged phase boundaries using electrical capacitance tomography

Abstract

The detection of hard boundaries using tomographic techniques is challenging due to the measurement resolution inherent in the hardware and smoothing effects created during image reconstruction. This paper is concerned with the development of data processing approaches which enable the use of electrical capacitance tomography (ECT) in real-time applications to visualise interfaces in liquid/liquid and solid/liquid systems based upon phase permittivity differences in media with a high di-electric continuum. The methodologies developed were applied to a series of phantoms to investigate their validity as a tool for imaging phase boundaries in two and three phase systems.In an ECT based tomogram, the interface between phases is exhibited as a transition region; by applying a threshold technique based upon known areas of each respective phase within the system, the transient region can be resolved into a sharp interface. The image error of a tomogram, defined as the deviation of all pixels from their theoretical value, has been calculated using a pixel-by-pixel approach; however this requires exact a priori knowledge and is unsuitable for in-line application; the areal method used in this paper requires global phase distribution information thereby allowing for real-time application. A range of threshold values were applied to tomograms of phantoms of varying geometry and the corresponding image error for each threshold value calculated using both the areal and pixel-by-pixel approaches given above. The threshold value yielding lowest image error from this range is further used in the binary images giving improved tomograms with approximately 40% increase in image accuracy when compared with a default threshold value. Close to the sensor wall, the image becomes distorted due to reconstruction errors arising from decreased density in the electrical field lines, resulting in a circular phantom appearing elongated by approximately 10% when positioned near the wall.