Enhanced flow regime identification with SVM using electrical capacitance tomography (ECT) in fluidized beds

Y. Ru, S. Mylvaganam

Telemark University College (TUC) Faculty of Technology, Department of Electrical Engineering, Information Technology and Cybernetics, N-3901 Porsgrunn, Norway

Abstract

Fluidization is widely used in commercial operations such as mixing, heating, facilitating reactions, etc. In fluidized beds, the focus is normally on determining the flow velocity of air, size and density of particles so as to achieve the needed degree of fluidization. Formation of bubbles in the process of fluidization is a common phenomenon. When bubbling occurs in fluidized beds, some of the parameters of interest are bubble size/velocity, bubble coalescence etc. Study of bubbles using non-intrusive twin plane electrical capacitance tomographic (ECT) system has been done before. Using process tomography in fluidization processes can help to understand and optimise the processes by identifying crucial parameters and observing their behaviours. In this paper, an ECT image based analysis is presented. Glass particulates are used in fluidization studies at different flow rates of air at ambient temperature. Two different image texture analysis methods and SVM classification are applied on the ECT images. The analysis helps to identify the flow regimes between bubble, slug and turbulence. Both image texture based analysis and SVM classification algorithms clearly lead to clusters in the plots indicating that a non- intrusive sensor system based on ECT and data fusion using SVM can help to identify flow regimes in fluidized beds handling materials with permittivities within the sensitivity of the ECT-system. The method has potential of being useful with other tomographic modalities.

Keywords: ECT; image texture; SVM flow regime identification