Modelling of Multiphase Processes Using Tomographic Data for Optimisation and Control

H. Tabe, S. J. R. Simons, J. Savery, R. M. West#, R. A. Williams#

Department of Chemical Engineering, University College London, Torrington Pl., London WC1E 7JE, UK. Email: stefan.simons@ucl.ac.uk.

# Virtual Centre for Industrial Process Tomography, Camborne School of Mines, University of

Exeter, Redruth, TR15 3SE, UK.

Abstract - Tomographic sensors are ideally suited to the on-line control of multiphase processes. Little work to date has been undertaken to determine what type and style of information is required from an image to provide effective process control. In this paper, a possible modelling strategy is presented; namely, a combination of Principal Component Analysis (PCA) and Neural Networks (NN) is used to convert multivariate data from tomographic images into useful information suitable for the control and optimisation of chemical processes.

Keywords: Multivariate data analysis, neural networks, principal component analysis, process tomography.