Data fusion for electrical spectro-tomography

Abstract

Electrical tomography has demonstrated the fast estimation of spatial distributions in a range of industrial processes, but has limitations in applications which feature multiple components. There is a need to discriminate between such components, or to identify specific materials, while retaining fast performance to permit the tracking of process dynamics. To deliver this need a method is outlined that utilizes a compressed, wideband excitation signal to produce response data. These are then pre-processed into segmented datasets for frequencies of interest, which in turn yield frequency-banded tomograms. These data must be fused to gain process component discrimination or identification and hence deliver the benefits. The paper describes a number of potential data fusion methods: exploiting a look-up table of known process components; and identification via a process model. Simulation based trials are described and their results are presented in overview and evaluated as a key step in the development of spectro-tomography as a new process sensing and measurement method.