基于新变量选择方法的FNN预报转炉终点

考虑到机理模型能较准确给出转炉吹炼过程的定性规律，而难以给出可靠的定量关系，首先利用冶金机理模型和PLS方法分析影响转炉终点碳的因素，然后建立基于模糊推理神经网络的转炉终点碳预报模型。结果表明，此法能有效提高对转炉终点碳预报的命中率和网络的训练速度。在w(C)绝对误差±0.02%控制精度下命中率达94.12%，相对误差±10%控制精度下命中率达56.86%。

Fuzzy-Reasoning Neural Network System Based on a New Variable Selecting Method to Predicate End-Point of BOF

Considering that the metallurgical mechanism model can provide reliable quality variables during BOF blowing, but not present their quantitative relationship. The influence factors on end-point carbon of steel were analyzed by mechanism model and partial least squares method. Then based on fuzzy reasoning neural network system, a prediction model of end point carbon for converter was established. The results show that this method can effectively enhance the hit rate of end point carbon prediction and the training speed of network. The hit rate reaches 94.12% in absolute error range of ±0.02% and 56.86% in relative error range of ±10%.