Research On Predictive Model Of Iron Ore Sintering Capabilities

With the rapid development of modern iron and steel industry, the iron ore resource is changed largely. How to use the variational domestic and overseas iron ore resource reasonably, sintering production has become a hotspot research in order to meet the need of the iron and steel production.By analyzing and summarizing the object of research,a new prediction method,which include neural network system and database system, is proposed. and the whole framework of the prediction-model of the iron ore sintering capabilities is ascertained. The database system of the iron ore sintering capabilities is designed including the memory technology of the database system and accessing technology of the ADO. The keys of neural network modeling such as getting the effectual training samples, in-out of the neural network, the framework of the neural network, the learning arithmetic of the neural network, the evaluating method of precision model are researched. In allusion to the intrinsic shortcoming of BP arithmetic,a quick BP the neural network learning arithmetic with amending error,appending momentum and adaptive variable step size linear reinforcement is presented and designed by programme. A general software system to predict the iron ore sintering capabilities including the database subsystem,the modeling subsystem and the predicting subsystem is realized and developed.A neural network prediction-model of the iron ore sintering capabilities with the 12-34-4 structure is built by this software system. The prediction result of random unknown samples is satisfactory by the model: the prediction score of tumbling intensity is 86.87%, the prediction score of sintering speed is 83.57%, the prediction score of sintering yield is 91.47%, the prediction score of sintering product is 85.57%.