Of Sinter Chemical Composition Of Artificial Neural Network Forecasting System

The steady control of the sinter chemical composition has puzzled sintering fields for many years. The mainly reason is that there are many factors which influence sintering process and the long time delay. So the key to realization of controlling sinter chemical composition steadily lies in the prediction.The predictive models of sinter chemical composition based on artificial neural network are established by comprehensive utilization of sintering theory and artificial neural network technology.On the base of the analysis and comparison of the predictive methods of sinter chemical composition, combined with the characteristic of the sintering process, the predictive models of the chemical composition are proposed to establish by the technology of artificial neural network. By studying the artificial neural network technology, the three-level former network is used, and the BP algorithm is improved. The structure self-organized algorithm is adopted, and the defect in the algorithm is improved.Several models of artificial neural network are established for every kind of the sinter chemical composition. The model structures are founded by initialization training. The appropriate number of adaptive learning samples is researched, which solves the real-time problem of the practical application of the artificial neural network model.This application software is developed by the VC++ which is a kind of visual software developing tool with powerful functions. The software is tested by the sintering production data, and the rate of region hit is over 90%, which shows the models can predict sinter chemical composition precisely. It lays the foundations of the next online control of the sintering chemical composition.