Research On Air Preheater Hot Spots Detection System Based On Image Features

This thesis is mainly concerned with mechanism of recombustion of air preheater and air preheater hot spots detection technique based on image features .Air preheater is the important heat exchanger in power plant units.Recombustion accident can be caused by inadequacy combustion of fuel or badly heat-dispersed condition aroused by low air or gas velocity after boiler outage. At present, thermocouple or infrared sensors are used as temperature measurement units in air preheater hot spots detection system. Whether or not recombustion caused was judged by the result of temperature measure value compared to the alarm threshold value set in advanced. However, when fire is brought out in air preheater, temperature degree covers a great range . So, it is hard to set appropriate value to detect all the hot spots at any time and much more experience and working conditions is depended on, therefore, sometimes warnings are lost or it makes a mistake to alarm.Aiming at the deficiency in existing air preheater hot spots system ,a new method of image is presented to detect air preheater recombustion in this thesis. By means of a experiment setting made ahead to simulate the air preheater working condition, full and accurate image information of air preheater at different stage is obtained by low-illuminati-on CCD(charge couple device). And then,by using FFT(Fast Fourier Transformation) frequency domain analysis, wavelet packet transformation and Fisher linear discriminant analysis separately,images are analyzed and processed and image features can be received accordin-gly.Afterwards, Euler distance acts as classifying criterion and experimental results show that the three methods can classify fire alarm effectively. Finally, performance of the three classification identifier is compared by means of ROC(receiver operation characteristic), error rate and calculation complexity.The result shows that FFT frequency domain analysis is a better one.