Predictive Model Of Coke Qualities Based On Nerve Network

It is a complicated process of coal blending for coke making. The mixed coal quality ,which is determined by the quality and proportioning ratio of single coal, has directinfluence on the coke quality index. So, the relation among the variables is very complex.The nonlinear relation makes the traditional prediction, which changing rule and characteris enacted at first, can't fill with the actuality demand. Artificial nerve network can providenew idea and method to solve these problems for its function of nonlinear and self-study.In this paper, on the basis of the summing-up and analysis of the coal blending theoryand method and the researching actuality of coke quality prediction in the world, the singlecoal quality analysis and carbonization experiment is done, and also, a optimizing projectof coal blending is designed and the mixed coal quality analysis and carbonizationexperiment is done. On the other hand, the experiment of analyzing the influencing factoron the coke quality is put up aiming at the problems of coke quality prediction in ourcountry. The experiment result indicate that in all the factors, the feed coal quality hasmore impact on the coke quality, which can lead to variety and fluctuation; but productiontechnology has less influence on it.Therefore, on the basis of this analysis, the factors that can influence coke quality areinduced filtrated and optimized. At last, the predicting models of mixed coal quality andcoke quality are set up by the way of nerve network, and the relation between input andoutput of the models are confirmed. The experiment data educated these two models, andat the same time, the experiment is done. The result of analysis indicate that the nervenetwork, which is on the basis of statistic, can describe the complicated relation betweencoal blending and coke making. The result of nerve network prediction is consistent withexperiment. The higher precision of this prediction proves that nerve network prediction isfeasible. That is to say, this research provides a new method for improving the coke qualityprediction veracity and scientific explore.