80 t转炉终点预报TSVR模型精度

 转炉炼钢是一个复杂的高温物理化学反应过程。在冶炼过程中不能连续检测钢的成分。所以,准确地预报终点的碳质量分数和温度对于提高终点命中率是非常有意义的。基于广西某钢厂80 t转炉炼钢实际生产数据,建立了终点碳质量分数和终点温度的孪生支持向量回归机(TSVR)预测模型,对100个炉次的实际生产数据进行了模型的训练,另外30个炉次的数据用于验证模型的精度。结果表明,预测误差Δω([C])≤ 0.01%的命中率为93.3%,Δt≤15 ℃的命中率为96.7%,双命中率为90%。与BP神经网络模型相比,TSVR模型的终点碳质量分数和终点温度命中率均比BP神经网络模型高。

End-point prediction TSVR model accuracy of 80 t BOF steelmaking

Basic oxygen furnace (BOF) steelmaking is a complex process with high-temperature physicochemical reactions. The composition of steel cannot be detected continuously during smelting, accurately predicting the carbon mass percent and temperature of the end-point is very meaningful to improve the end-point hit rate. The actual production samples of 80 t BOF were collected from one steel plant in Guangxi province. A twin support vector regression (TSVR) prediction model was established to realize the prediction of end-point temperature and carbon mass percent. The training process was carried out by using actual production samples of 100 heats, and the other 30 heats were adopted to verify the accuracy of the prediction model. The results show that the hit rate of the model with Δw([C])≤0.01% is 93.3％ and the hit rate of the model with Δt≤15 ℃ is 96.7％. In addition, the double hit rate of the model is 90%. By comparing with the BP neural network model, the end-point carbon content and the end-point temperature hit rate of this model are higher than those of the BP neural network model.