|
The main purpose of this paper is to improve the CWS slurry performance and establish a high-precision CWS coal blending slurry performance prediction model , the application of the regression model and neural network model predictive models of the preparation and development of the CWS basic research . First of all , through the experimental study of the the coal physicochemical characteristics parameters of coal slurry ability to study the relationship between the coals of Mad, Aad and Oad several factors with coal slurry ability . The experimental results can be found in coal slurry ability is determined by a variety of factors , but the more complicated the relationship between several factors , in a separate influencing factors to analyze a coal slurry ability unscientific 's . Due to the different nature of single coal , the different coal mixed into a slurry process interaction , mutual restraint , so coal blending into a slurry can not be a simple superposition of the components of coal characteristics , but showing a very complex nonlinear characteristics of neural network technology is the emerging market to solve the the blending nonlinear problem of effective method . Through the analysis of the nature of coal slurry performance , choose ten factors regression analysis forecast , compared with the neural network model . 10 factor , the nine factor , five - factor , four-factor and three- factor linear and nonlinear analysis , forecast results , the best five - factor linear regression model , predict the results of the error of 1.69% . A total of 10 factors , nine factors , the five factors , the four-factor and three-factor combination of coal slurry performance factors of neural network predictive analysis . Best model of the number of factors by comparing each input parameter and error , found that the best results of the five - factor neural network prediction model , the prediction error to the level of 0.49% , significantly lower than the five - factor linear regression model each input factor the number of neural network prediction model results than the corresponding results of the regression equation is better .
|