Assessment of the Combustion Behavior of a Pilot-Scale Gas Turbine Burner Using Image Processing

Abstract

Experimental investigations were performed on a non-premixed liquid fuel-lean burner. The present work aims to the development of a methodology for the recognition of flame instability regimes in industrial and aeronautical burners. Instability, in fact, is an unpleasant aspect of combustive system that negatively impacts on combustion efficiency. The online monitoring of the occurrence of instability conditions, permits to adjust combustion parameters (as fuel or air mass flow, temperature, pressure, etc.) and to stabilize again the flame. High speed visualization systems are promising methods for on-line combustion monitoring. In this study two high speed visualization systems in the visible range and in the infrared spectral region were applied to characterize combustion efficiency and flame stability. Different processing techniques were used to extract representative data from flame images. Wavelet Decomposition and Spectral analysis of pixel intensities of flame images were used for feature extraction. Finally a statistical analysis was performed to identify the most unstable regions of the flame by the pixel intensity variance.