Prediction System Of Silicon Content In Hot Metal Based On Integrated Neural Network

In iron making, detection of silicon content has significant impact. On the one hand, it is beneficial to measure the quality of pig iron. On the other hand, it can be used to predict the temperature of blast furnace. However, the silicon content in hot metal is not known until the steel samples are sent to laboratory to be tested. So, it is necessary to predict silicon content advanced and eliminate measurement delay.Single neural network are mostly used in the prediction model, which has complex network structure. So, the neural network is very hard to be trained effectively, and the prediction accuracy is low. To solve above problems, integrated prediction model and combination forecasting model are proposed here.Integrated prediction model decomposes the complex task into several simple sub tasks, which handled by sub networks. The processed results of each sub task are putted together into decision and fusion network to analysis and process, then the final result is obtained.Combination forecasting model includes grey forecasting model and integrated prediction model. Raw data of each parameter is no longer used by combination forecasting model as training and prediction samples any more. Instead of this, the prediction value of raw data obtained by the gray prediction model is used. The prediction value of raw data is more regular, whose randomness weakened.MATLAB simulation experiments results demonstrate the correctness and effectiveness of the two models. Combination forecasting model has not obtained such a good prediction effect as integrated prediction model. The reason lies in raw data's randomness is over weakened by combination forecasting model. Some information should be included by raw data is lost. However, integrated neural network is applied in integrated prediction model, which has simple network structure. So it is easy to be trained effectively. And the ability of fuzzy inference is given full use of by fuzzy neural network in integrated prediction model. As a conclusion, integrated prediction model can be used as the prediction model of silicon content with high accuracy and low relative error.In order to improve prediction models" practical value, PC management system of silicon content is developed. User interface is friendly and clear which programmed by C++language. Algorithm design is implemented by MATLAB which run in background as control module. It even can be used for those who are not familiar with the MATLAB. The PC management system has realized effective prediction and scientific management of silicon content in hot metal.