Abstract: Based on consumption of electric power, oxygen and carbon powder of Consteel arc furnace, the offset of measuring of metal weight system for Consteel arc furnace due to consumption of refractory of furnace is compensated by es tablished BP artificial neutral network model and Matlab simulation software calculation. Results showed that the absolute value of metal sample data reduced error was ≤0.000 8 t by Levevberg-Marquadt optimum calculation training with BP ar tificial neutral network model.

Key words: Artificial Neutral Network, Consteel Arc Furnace, Metal Weight, Simulation Calculation

摘要： 根据Consteel电弧炉的电能、氧气和碳粉消耗，通过建立的BP人工神经网络模型和Madab仿真软件 计算补偿因炉衬耐火材料消耗而引起的Consteel电弧炉称重系统测算钢水重量的偏差。结果表明，经过Levevberg- Maxquardt优化算法训练后,BP人工神经网络模型对钢水样本数据还原误差绝对值≤0.000 8 t.

关键词: 人工神经网络 , Consteel电弧炉 , 钢水重量 , 仿真计算