BTP Prediction Model Based on ANN and Regression Analysis

In this paper, data mining technology is adopted to find correlations from massive production data to predict burning through point (BTP) of sintering process. A hybrid BTP prediction model is presented which is based on artificial neural network and multi-linear regression error compensation algorithm. In this model, the final prediction result is calculated based on both the prediction value from ANN model and the compensation value from linear regression model. Compared with the pure ANN model, the maximal prediction error is reduced from 3.85 meter to 1.3 meter and the prediction errors are kept in the range of 1 meter with the probability of 98.89%. Finally, the prediction values and correlated variables are sent back to L1 system through BTP control model to realize a closed-loop control system.