Recognition And Diagnosis Of Boiler Flame Based On Wavelet Neural Network

The wavelet theory has been developed and perfected for ten years. Because of its particular merit it is widely used in many fields, such as signal processing, image processing, communications and physical geography, etc. it is similar to the function approximantion ability of Fourier series in the Fourier analysis. The function approximantion ability of the wavelet series in the wavelet analysis is one of the most important characters in the wavelet theory. Neural Network (NN) offers people configuration and algorithms, which are learning and memories and simultaneous disposal. For the recent years, people have been attracted by the universal approximants of NN, which are able to approach all functions. NN has some unique merits for pattern recognition and function approximants. But wavelet analysis has the capability of stepwise fine description and distilling characters for non-smooth random signal. The two tools form wavelet Neural Network, namely WNN. WNN was brought forward by Zhang Qinghua, etc. as the concept of former Neural Network approaching all functions in 1992[17]. A few years recently, the concept of combining wavelet with Neural Network has been being advanced and improved[18-23].WNN is used in the function study by Zhang Jun etc., who investigated function approximants of the WNN and compared it with other methods[20].The precision of wavelet approximants is analyzed from the theory by B.Delyon, etc.[23]. The former(Back-propagation) Neural Network is analyzed and colligated by the means of discrete imitational wavelet transformation[18]. Lately, WNN and its characters are investigated by Jiao Licheng etc[21]. According to inherent analysis of wavelet transformation and NN, we make use of the linear combination of different stimulant functions to construct corresponding wavelet function and deduce the conclusion that the wavelet function(stimulant function) of NN and the stimulant function of common NN are essentially the same. Consequently WNN is introduced by the academic analysis and comparison of wavelet transformation, NN with the function approximants of NN, we think WNN has obvious predominance in the aspect of function approximants. Besides, we use the three imminent manners to validate a typical function. It is pivotal that improving automatization level of electric power system consists in the intelligent degree of burning state supervision in the boiler of fire electricity factory. It is essential for the boiler of electricity factory to set up and maintain burning flame steady and symmetrical in the hearth. If coal powder was continuously supplied to the buring container, when all or partial coal powder which is burning is going out, the phenomena of deflagrating would take place and the production would be influenced badly because of the accumulation of coal powder in the hearth. Therefore, effective inspection, recognition and diagnosis of flame must be conducted. It is quite difficult in recognizing the burning flame state with existing simple method. But WNN is excellent in the aspect of self-study and non-linear division and can be used to recognize flame image when the characters aren't determinate and the circumstance is abominable. A method on fast recognition and exact diagnosis of the boiler flame is proposed using image processing and Wavelet Neural Network. The image of boiler flame is captured by the system of image processing. The character parameter of image for boiler flame used for burn diagnosis is attracted. Adopting the orthogonal least square (OLS) to select the proper wavelet construct Wavelet Neural Network. The character parameter to attract burn image is the input vector of Wavelet Neural Network. The method of image processing and diagnosis based on Wavelet Neural Network has the exact and fast excellence, forecasting the state of burn blame via training and testing. Similarly, the technology of image processing and Wavelet Neural Network is proposed to realize the monitoring of the NOx for fast recognition of the boiler flame. The method of image processing based on Wavelet Neural Network for the monitoring of NOx has the exact and fast excellence, distilling the state of burn blame via training and testing. Within the scope of visible spectrum, the color of high-temperature object changes with the temperature. A method of temperature measurement...