Research And Development Of Prediction Model And Control-Guidance Expert System Of Sinter Chemical Composition

Sinter is the main raw material of blast furnace. The stability of sinter chemical composition has directly influences on sinter quality and the stable operation of the furnace operation. So it is nesserary to control the chemical composition to be stable.The stability of chemical composition is mainly affected by raw material composition. It needs a long time to get the chemical composition of sinter from raw material charging. It is highly correlated among sinter chemical compositions. So it is complex to control the stability of the chemical composition. Currently, the approach which combines forecasting and expert system is mainly used to control sinter chemical composition.GM(1,1), BP neural network model and grey neural network are developed for sinter chemical composition forecasting. Small sample is required in gray prediction model. The accuracy of neural network model is high, however, it requires large amount of data for modeling. Grey neural network (GNN) combines advantages of both. Therefore, the forecasting method of using gray model under small sample condition and GNN with medium sample and neural network model with large sample is proposed based on sintering characteristics. The hit ratio of gray model is over 85%, and the hit ratio of GNN is over 92%.According to the main raw materials, combined with sintering theory and practical production conditions, sinter chemical composition control expert system is built. The expert system includes knowledge base which is based on database technology and forward reasoning machine. The raw material structure is divided into two categories and eight sub-categories, knowledge bases of sub categories were constructed, so that versatile of chemical composition control expert system is greatly improved.Software of sinter chemical composition prediction and controling-gudinance expert system is developed by mixed programming approach in VC++6.0 and Matlab, combining with database technology. The system includes Prediction models, expert systems and system application. The software is of Practicality, versatility, user-friendly, and easily operated. The user can choose the appropriate production model, load the corresponding knowledge base, and connect the data acquisition interface, according to the situation of the plants. Forecasting and controlling-guidance of sinter chemical composition are realized. The system can apply for high forecast accuracy, and great directive function has been achieved.