Image Reconstruction in Process Tomography Using Hierarchical Self-organising Maps

D Benchebra, P V S Ponnapalli, R Deloughry

Department of Engineering and Technology, Manchester Metropolitan University, John Dalton Building, Chester Street, Manchester M1 5GD, U.K., p.v.s.ponnapalli@mmu.ac.uk

ABSTRACT

Electrical Capacitance Tomography (ECT) has become a useful tool in the measurement and monitoring of industrial systems. A typical ECT system would consist of sensing elements, measurement and signal conditioning electronics and a software module for the construction of images. Image reconstruction using the measured capacitance values is an important part of ECT systems. Common techniques used in this process include Linear Back Propagation (LBP), Singular Value Decomposition methods, variations of Tikhonov methods and Landweber algorithms. However, only the LBP has been found to be useful for real-time implementation. The images obtained using LBP suffer from soft-field effects and lower resolution due to the approximations involved in the LBP algorithm. This paper presents initial results in applying Artificial Neural Networks to enhance the quality of images obtained using LBP. In this context, the use of Self-Organising Maps (SOMs) is proposed as a filtering mechanism to improve the quality of the reconstructed image. Initial results indicate that SOMs can help in reducing the uncertainty in the predictions of area and average permittivity of material in a pipe using ECT measurements.

Keywords ECT, Neural Networks, Self-Organising Map, Process Tomography