转炉终点锰预报模型的开发

Development of a Prediction Model for BOF End Manganese Content

探讨转炉炼钢终点锰预报模型的建模技术 模型分为两种 :代数学模型和人工神经网络模型 代数学模型采用多元线性回归方法建模 ,该模型简单、可视 ,但仿真结果并不理想 人工神经网络模型在选取适当输入参数的基础上 ,通过对转炉生产的历史数据进行训练 ,求得合理优化的网络权重 ,可对转炉终点锰含量进行离线预报 ,该模型的仿真结果很好 ,当预报误差精度|ΔMn|≤ 0 .0 2 5 %时 ,预报命中率超过 95 % 。

The modelling technology for prediction of BOF end manganese content is described. There are two kinds of models: algebra model and artificial neural network model. The algebra model is established with multiple linear regression equation.This model is simple and visible,but its emulation result is not satisfactory. On the basis of selecting some suitable input parameters, the artificial neural network model trains the BOF's previous practical data to obtain the logical optimum net weights,then predicts the BOF end manganese content off line. The artifical neural network model's emulation result is satisfactory. When the prediction error precision |△Mn|≤0.025%,the shoot ratio is above 95%.The artificial neural network model provides gists and technical supports for the development of BOF's online manganese content prediction and fast tapping model in the future.