Application And Research Of Improved Wavelet Neural Network In The Acoustic Pyrometry System

In large coal-fired power plant,to operate safely and efficiently,to lower pollution,and to realize the essence of the combustion,the distribution and visualization of the furnace temperature field is important in practical value and scientific meaning.However,the processing of combustion in power plant furnace is instantaneous,random and in hard environment,and the scale of furnace is large. At present,an effective monitoring measurement on furnace temperature field has not been established.Further theory and experiment study is developed of the real-time monitoring technique on furnace temperature field based on acoustic theory.The measuring principle of acoustic temperature measurement is indirect measuring the temperature through measure the sound travel-time.So accurate sound travel-time is the premise and assurance of acoustic temperature measurement.The time delay estimation(TDE) of weak sound signal under the condition of strong noise is studied.The disadvantages of some traditional methods of TDE are shown.A new TDE method based on improved fourth order cumulants is given.The simulation tests of TDE are done based on the ML cross-correlation method and Wavelet Neural Networks method respectively under different signal noise ratio(SNR).The results show that the influence of correlated Gaussian noise can be restrained from the non-Gaussian signal based on the improved fourth order cumulants even under poor SNR.Reconstruction algorithm is image key to achieve high-precision reconstruction of temperature field at the acoustic method furnace flame temperature field study of image reconstruction.In this paper,through analysis and research the traditional temperature field reconstruction algorithms,the problems and shortcomings are pointed out.Combining the studied knowledge, proposed the use of real-time correction of wavelet neural network algorithm to achieve the reconstruction of temperature field.The simulation by wavelet neural network to improve the temperature field reconstruction algorithm has a better reconstruction effects.