Multiphase flow reconstruction in oil pipelines by capacitance tomography using simulated annealing

ABSTRACT A highly optimized simulated annealing (SA) algorithm is applied to reconstruct permittivity images of real two-phase gas-oil flows through a cylindrical vessel using electrical capacitance tomography (ECT). ECT yields low-accuracy images but is robust, inexpensive and much faster than many other tomography processes. This non-intrusive method essentially measures non-conductive system distributions and is applied in oil industry processes such as mixing or stirring vessels, fluidized bed reactors, separator tanks and pipelines carrying multiphase flows. A forward problem is solved at each step of an iterative algorithm to solve the inverse problem using simulated annealing (SA). Comparisons with linear methods like The Projected Landweber technique are discussed. In this paper we introduce a finite volume discretization with local mesh refinements in a cylindrical configuration close to the electrodes in order to improve resolution in the calculation of capacitances, and to avoid problems with resolution at the centre of cylindrical container when finite differences are used. This discretization has the advantage of a conservative formulation as used in finite element methods and features the flexibility of mesh refinement close to the electrodes. Thus, improvement of local accuracy is achieved without increasing prohibitively the number of mesh points. Performance of the forward problem resolution is compared with finite element based methods and experimental data. We show that the non linear version of SA provides better reconstructions of three-phase flows than the Landweber method.

Saved in:
Bibliographic Details
Main Authors: Martin,R., Ortiz-Alemán,C., Rodríguez-Castellanos,A.
Format: Digital revista
Language:English
Published: Universidad Nacional Autónoma de México, Instituto de Geofísica 2005
Online Access:http://www.scielo.org.mx/scielo.php?script=sci_arttext&pid=S0016-71692005000300241
Tags: Add Tag
No Tags, Be the first to tag this record!