Development Of Flame Detector Based On Chaos Theory And Support Vector Machine

The burner is one of the most important equipments of thermal power generation. When the burner flame is out, and if fuels are still supplied, these will result in the accumulation of the fuels, which would cause the boiler explosion. Therefore, it is important to improve the detection of the burners' flames for ensuring the safe and reliable running of the boiler.At present, the photoelectric flame detector is widely used in the fame detecting of the burner of the boiler in the power station. Its principle is to distinguish the combustion conditions of the fuels according to the frequency and the strength of the radiate of the flames, and its merits are small physical volume, cheap price and convenient installment. Meanwhile, most flame detectors can only detect the combustion status of a kind of fuel, but can't distinguish the categories of the fuels. In the actual usage, coal burner and oil burner need respectively to be provided with respective flames detector.Based on the chaos theory, this paper analyzes the flames radiates, which contain the stable combustion condition of pure coal, the stable combustion condition of the pure oil, the stable combustion condition of the mixture of the coal and oil and the unstable combustion of the flame, and which are sampled from power station on the scene. The result indicates that radiates of the flames combustion is chaotic time series; we can withdraw the chaotic characteristic eigenvectors (correlation dimension and largest Lyapunov exponent), and these two parameters can be taken as the characteristic parameters of conditions distinguishing. From the two-dimension figure that has correlation dimension and largest Lyapunov exponent we can see that the chaotic eigenvectors for different fuels in different burning statuses appear in different vary areas, so not only the combustion conditions of the fuel can be distinguished, but the combustion categories of fuels can also be distinguished.In order to carry on burnable diagnosis more efficiently, this paper takes the correlation dimension and largest Lyapunov exponent as the inputs of the Support Vector Machine and the classification model based on Support Vector Machine is built up after training and studying. The imitating and verification are carried on the actual testing data, and the result indicates that the model based on SVM is effective to distinguish the four combustion conditions containing the stable combustion condition of pure coal, the stable combustion condition of pure oil, the stable combustion condition of the mixture of coal and oil and the unstable combustion condition of flame, and its accuracy is 100%.According to the flames combustion diagnose model mentioned above, this text designs the flame detector, whose core is the microprocessors of PIC18F2525 and function conclude the sampling of the flames signals, processing, diagnosing, identify and protection etc. Not only the flame detector developed by this paper can identify the combustion conditions of fuel, but also can identify the categories of the fuels, which provides solutions and supports for developing the common detector of the coal burner and oil burner.