Summary: Published at the fourth meeting of the European Concerted Action on Process Tomography (ECAPT). Bergen, Norway,
Application of a New Iterative Minimization Method to VisionGuided
Flame Control 1
L. Oest, H. Burkhardt
Technische Informatik I, Technische Universit¨at HamburgHarburg, D21071 Hamburg, Germany
ABSTRACT: This paper presents a new iterative minimization method for camerabased
flame control applying digital image analysis. The method is an extension to the Gauss
Newton iteration. The controller is able to ``learn'' process states which are set up by a human
expert. The controller properties are discussed in terms of convergence of the optimization
method. Experimental results are presented to emphasize the controller properties.
KEYWORDS: Closedloop Control, Image Processing, Minimization, Combustion.
Combustion is a turbulent reactive fluid dynamic process. The interaction of chemical reactions, heat
transfer and fluid dynamics makes it impossible to model these multiphase flows in the full spatial
domain. Local point measurements give only little information about the process state. On the other
hand, information about spatial distributions, e.g. temperature, concentrations and velocity fields, directly
indicates the origin of undesired burning conditions.