Computer Simulation Of The Closed Blast Furnace Control

Imperial smelting furnace has many advantages, such as it can smelt zinc and lead together, it is applicable to smelt complicated and mass sinter containing zinc and lead. Investment and energy is less, but recovery rate of metal is high. However imperial smelting process is very complicated, it come down to interactions among gas, solid and liquid. The whole process occurs continuously, control effect dally over long time. Process control more depend on experience of workers at present. For the sake of improving ISP抯 control level and the prediction of production, realizing standardization work, Through analyzing substantial and thermodynamic balance, this thesis establishes mechanism model. For monishing gap between mechanism model and industrial process, the thesis searches object model on the basis of fitting industrial practice fuzzy neural networks. Learning from other strong points to offset one weakness, the two models combine to form adaptive predictive control model. By inputting component of sinter and sinter quantity, the whole predictive control model can output producing matter quantity, for example furnace gas quantity and component, can advise to input the quantity of bottom air(first air) and top air(secondary air), can calculate gasifying rate of coke etc, can draw trendline diagram including important influencing factor. The calculating result tallies with industrial practice, for example the air temperature increases every 100 0C,coke input quantity can decrease 35?7 kilogram fuzzy neural network combining simple applied high-learning-precision back propagation neural network and solving uncertainty fuzzy system is used in the thesis, the network can be used for the direction of industrial process and off-line control. It can be used for realizing and holding furnace status; for optimizing the output of control model and laying the foundation for on-line control, On- line adjustment is used. Some useful attempts are made for the design of fuzzy neural network.