Experimental evaluation of state estimation with fluid dynamical models in EIT imaging of 3D pipe flow

Antti Lipponen1, Aku Seppänen1 and Jari P. Kaipio1,2

1Department of Physics and Mathematics, University of Eastern Finland, P.O.B. 1627, FI­70211 Kuopio, Finland, Antti.Lipponen@uef.fi

2Department of Mathematics, University of Auckland, Auckland 1142, New Zealand, jari@math.auckland.ac.nz

ABSTRACT

In this paper, we experimentally evaluate the feasibility of the state estimation approach with fluid dynamical models to imaging of three­dimensional flows in process tomography. In the state estimation approach, the reconstructions are computed based on the state space representation which consists of two models: the observation and the evolution model. In the experiment, the conductivity distribution in 3D pipe flow was monitored with electrical impedance tomography (EIT). The evolution of the target was modelled with the convection­diffusion equation and the observations with the complete electrode model. We also employed the approximation error approach to compensate the errors caused by coarse discretization, truncation of computational domain, and unknown contact impedances. All reconstructions were based on absolute imaging? that is, no reference measurements were utilized in the reconstructions. The results confirm that reliable tomographic reconstructions of nonstationary targets can be obtained with state estimation. In addition, the study demonstrates the importance of accounting for the model inaccuracies and uncertainties in absolute imaging: enhancing the state­space representation with the approximation error models yields a significant improvement in the reconstructions.

Keywords state estimation, electrical impedance tomography, absolute imaging, approximation error approach