Emission Prediction Of A Combustion Process Based On Flame Radicals And Soft Computing Techniques

China is the world’s largest coal producer and consumer country, what determins that the main characteristics of atmospheric pollution is coal-burning pollution. Particulate pollutant and gaseous pollutants, including nitrogen oxides (NOx), sulfur dioxide (SO2),which are produced by the coal combustion, have polluted atmospheric environment seriously of human survivals. A flame of combustion contains abundant information. It not only is the direct reflection of the relationship of the combustion stability and the state of the economy,but also directly affects the emissions of combustion pollutants. Flame radicals appear in the flame front and they are the main symbols of heat release in hydrocarbon fuel combustion. Based on the digital imaging of flame radical, this paper presents the feature extraction in order to predict NOx emissions.It has important engineering significance for the optimization of combustion, reducing pollution and improving air quality.In this paper, the main content is as follows:Firstly, the application of digital image processing technology in industry and digital image acquisition method of four kinds of flame radicals CN*, C2*,OH*å'ŒCH*are introduced.The wavelet threshold value de-noising method is used to process the noise of image. Meanwhile, the fuzzy clustering methods(FCM) is used for image segmentation and areas of imaging of flame radicals and a luminance value is extracted as the image features.Secondly, based on the feature of image, combining the combustion flame temperature, BP neural network is adopted to establish predictive models of NOx emissions of combustion pollutants. Meanwhile, the neural network variable selection method is used to reduce BP network input variables and the prediction model is built to predict NOx emissions. The simulation results show that BP neural network has good effect on NOx prediction.After variable selection, neural network modeling has better prediction results.Thirdly, we have introduced the basic principle of extreme learning machine (ELM), mathematical models and model algorithm. The ELM is also used to eatablish prediction modeling for NOx emissions. The simulation results show that the prediction accuracy of ELM is better than BP neural network, and ELM has a faster convergence speed.Finally, the two methods of prediction of NOx emissions are summaried and the future work is presented.