Application of wavelet transform in detection of burned fuels in power boiler

In power boiler control, it is very important do distinguish signals corresponding to different air/fuel ratios and different flames (pulverized coal, mazout). Since the signals obtained from flame monitoring systems are unstable, wavelet transform has been utilized in filtering these signals. Problems arise with choosing the most adequate wavelet to analyzed signal. In this paper, we investigate signals which reflects sudden cut off of mazout in industrial burner (pulverized coal is still present). Because amplitude measures are unreliable, we have been concentrated on 'frequency' measures. A method of choosing the proper wavelet and decomposition level have been proposed.