Blast Furnace Thermal State Based On Artificial Neural Network Prediction Model

Blast Furnace Thermal Condition Expert System is used widely in homeland and overseas Blast Furnace, the knowledge base is the core part of the expert system and is key to the capability of the expert system. the expert system belongs to computer software system and simulates experts to solve problems by means of application the knowledge and experiences of experts in some fields.BFES is a real-time expert system based on sensors. In Blast Furnace Expert System, knowledge is described by lots of rules of if...then...and rules -matching method is adopted for process inference. However, in traditional expert system, knowledge is acquired and retrieved by means of person's summary and transplant, it has the characters of taking-time, low efficiency, bad capacity of self-study and self-organization, limited knowledge, few layers, bad real-time. Being directed against these shortcomings of the traditional expert system, the neural network, one field of the artificial intelligence, is developed rapidly. the neural network is a non-linearity dynamical system, it has the features of distributed storage and parallel disposal, the structure and function of singe nerve cell is simple and limited . but the neural network system composed of lots of never cells has a strong function, it has better self-adaptation, better self-organization, and better self-study than traditional expert system.With the puddling technology developing in BF, the combination of NN with Expert system has become the most important developing trend. The combination can exert its own advantage fully. The simple method is to use NN to deal with these blurred and half-baked data or to use NN to enhance the judgement.The craftwork and operation of the forth BF in Xiang tan iron & steel has been analyzed and the soul knowledge of BFES has been studied and researched. by means of analyzing the factors effecting the BF thermal condition, the combination of BPNN with expert rules is discussed and the prediction and control model of Hot metal Si-content is put forward in the paper. on the basis of control theory and user manualof Si-content and experiences of puddling experts, The Expert knowledge base of Hot metal Si-content has been established. It has been trained and corrected and it is described by by lots of rales of if...then...and rules.What else, prediction and control model, the application programme of the neural network prediction model has been programmed in Matlab and the database has been developed in Oracle.