Sinter Chemical Composition Based On Bp Neural Network And Gray Neural Network Prediction Model

In the modern steel enterprises,the sintering process of blast furnace material is one of the best important production process.The chemical composition of sintering production such as alkalinity, FeO, MgO, CaO, SiO₂, drum index has a direct effect on production and economic benefits of whole steel enterprise.Therefore almost every steel factory is equipped with many instruments and automatic control systems in its sintering plant for its producton process control.But the complexity of sintering production process makes difficult to be described by a set of mathematic models. Since this process often has large time delay and dynamic time-varilabilityit.is hard to perform control tasks of total sintering process by using conventional control models.This problem has been consided as one of the key difficulties in the production process of steel enterprises.The main research gain of the thesis lies in:Firstly, a prediction model of the chemical composition in sintering process based on BP neural network is proposed to judge the trend of the chemical composition. The application result shows that the prediction with this method can achieve higher robust, better utility and expensive value.Secondly, a grey neural network model were proposed on the basis of the models.The fluctuation of data sequence is weakened by the grey theory and the neural network is capable of processing non-linear adaptable information, and the GNN is a combination of those advantages. The results reveal, the alkalinity of sinter can be accurately predicted through this model by reference to small sample and information. It was concluded that the GNN model is effective with the advantages of high precision, less samples required and simple calculation.Thirdly, the control strategies of the composition of sintering product are mainly introduced.It adoped the strategies of zone optimization and took the alkalinity as the center,which decreased the non-accurate control due to error of data measurement. Also avoided the frequent adjustment of process control .It realized the stable run of sintering process and avoided the mistake-operation.The last, A back-propagation neural network simulation tool is developed.A mothed of hybrid programming with VC and Matlab is introduced.which is based on COM .By means of Matlab COM Builder in Matlab 6.5,Matlab's M —function files can be converted into a COM,then applied in VC.The integrated software can not only take full advantages of VC and Matlab,but also run independently without Matlab.The process of realization and a sample programe both indicate that method of is sample and convenient.