Flame detecting and diagnosis of combustion in modern coal-fired boilers are very important to the safe and economical operating of power generation unit. So the effective real-time detection of combustion flame stability and in-time judgment are necessary. One of the important characteristics of combustion flame is a combination of flame jet fluctuation and flickering of flame radiation, which is a time random variable and reflects the combustion conditions. In this paper, after the data are acquired through the tests in our university’s laboratory, the power spectrum analysis using algorithm of fast Fourier transformation (FFT) and a self-organized neural network are applied into a diagnostic system for combustion conditions. At first, a time series of radiation intensity values of the flame, which fluctuate at a mean intensity value with a certain frequency are obtained through the photoelectric sensor. And then the time signal is converted to the power spectrum signal through the processing of FFT. Under the stable and unstable combustion conditions, the spectral intensity of the low frequency component of the converted signal has distinct magnitude. According to this method, software for the power spectrum analysis and the self-organized neural networks has been developed.

This content is only available via PDF.
You do not currently have access to this content.