Chaos Analysis And Feature Extraction In Sintering System Of Coal-fired Furnaces And Kilns

Coal-fired kilns and furnaces are key production equipments and main energy consumption equipments in basic industrial fields such as metallurgy,chemical industry,electric power,cement and active lime.As the stability of coal-fired process(sintering process)determines the quality of products,the consumption of pulverized coal and the emission level of pollutants,it is the key to realize “energy-saving and emission-reduction”.However,the large number of physical and chemical reactions,material and energy exchange in the sintering process limit the cognition of the intrinsic kinetic mechanism of coal-fired kilns and furnaces,and the practical industrial application effect.In this paper,the chaotic identification and application of the sintering system of alumina rotary kiln are studied from the perspective of nonlinear systems,further to pave the way for stable control of coal-fired kilns and furnaces.The main work and innovations are as follows:(1)Researches of the chaos in many fields are mainly based on mathematical equations or small prototypes,which are not suitable in coal-fired kilns and furnaces.In this paper,based on the actual thermal data from the sintering system of alumina rotary kiln,the chaotic identification of the sintering system in the coal-fired kiln with unknown mathematical equations is studied.Among them,according to the influence of data source,data length and data sampling period on the results,the sintering temperature data and kiln head temperature data with different data length and sampling periods are selected firstly.Secondly,several candidate dynamic trajectories of each set of data are constructed by using the phase space reconstruction method.Thirdly,based on the nature of chaotic characteristic quantities,the suitable dynamic trajectory is determined from the candidate dynamic trajectories of each set of data.Finally,the sintering system of alumina rotary kiln is determined to be a fifth-order system with chaotic characteristics.(2)Noise in coal-fired kilns and furnaces interferes with signal detection and control strategies,moreover the types and characteristics of noise are unknown.In this paper,chaotic features and multifractal features of noise data extracted from five kinds of thermal data are analyzed.Experimental results show that the noise in the sintering system of alumina rotary kiln is not the white Gaussian noise or the monofractal color noise,but has chaotic and multifractal characteristics.And one reason why the noise is difficult to predict effectively is discovered.(3)In view of the disadvantages of high computational complexity,without considering dynamic features of the system and lack of interpretability of deep learning models applied to coal-fired kilns,an integrated prediction framework with high prediction accuracy,low computational complexity and strong interpretability is proposed to realize accurate on-line prediction of kiln head temperature chaotic time series.Where,in order to response time delay behavior of the sintering system,the phase space reconstruction method is used to achieve the high-dimensional dynamic trajectory of the sintering system of the rotary kiln.And Volterra filter is used to approximate the dynamic relationship.As the time-varying characteristic of the parameters of the sintering system will weaken the performance of the fixed model,the sliding window technique is used to improve the Volterra filer through updating the kernels of the Volterra filter before each prediction to realize dynamic prediction.Furthermore,considering the practical industrial data contains unreliable values,each predicted value is corrected by the offset compensation technique as the final output.(4)The poor production environment inside coal-fired kilns can conceal much useful information from flame images(like,the color,texture,gradient),which hinders the performance of the method based on flame images or videos for identifying the sintering condition.In this paper,a chaotic feature extraction method based on the average gray value sequence of the flame video is proposed to accurately identify the changing trend of the flame temperature to further improve the automation level of coal-fired kilns and furnaces.Moreover,the effectiveness of this method is not only explained based on chaos theory and entropy theory but also verified by standard data and flame video data of alumina rotary kiln.