ABSTRACT
If a flame accompanies a combustion process, it reflects the chemical and physical phenomena that accompany the combustion process being examined. Flame properties, such as the geometry (size, position), radiation properties (the emission spectrum, irradiation distribution), and flicker frequency depend on the conditions in the location where the combustion process is proceeding. These conditions include the combustion chamber geometry, the burner type, fuel flow, air to fuel ratio, fuel granularity, etc (Fristrom, 1995). Analyzing the changes in the shapes of the flames that accompany variable burner input parameters, makes it possible to detect the different states of the combustion process, both in the laboratory (Lu et al. 2005) and under industrial conditions (Marques & Jorge 2000, Beak et al. 2001). Tens of parameters, resulting in a large feature vector, can be used to define shape. The problem is which shape features are most sensitive to variations in the burner input parameters and which can be neglected in the assessment of the combustion of biomass and pulverized coal. Size reductions of the feature vector, as well
as finding the most adequate shape features, can be obtained using principal component analysis (PCA). In this paper, only some of the geometric features of the flame image were taken into consideration.
