Summary: Published at the SENSOR 95 Kongreß. N¨urnberg, Germany, May 1995.
VisionGuided Flame Control 1
Hans Burkhardt, Lars Oest, Wenjing Tao
Technische Universit¨at HamburgHarburg, Technische Informatik I,
Harburger Schloßstr. 20, D21071 Hamburg
This paper is concerned with monitoring and control of combustion processes using image processing
techniques. For closedloop control we investigated two different approaches, namely optimization tech
niques as well as fuzzy logic and neural networks. Both control strategies are able to ``learn'' an optimal
burning state which is set up by a human expert. Advantages and disadvantages are discussed and expe
rimental results are presented.
Optimization of combustion is a very demanding task. A thorough analysis of such a complex fluid dynamical,
reactive process is the basis for a reduction of pollutant emission or for an enhancement of efficiency. Common
measurement techniques like local point measurements or exhaust analysis give only little information for
an interpretation of the whole burning process. Optical sensor arrays, used e.g. in video cameras, however,
provide an enormous amount of information, which can be used to analyze a process or for online control.