摘要
本文针对LF精炼炉冶炼过程中物理化学反应过程及传热过程的复杂性,采用混合模型对钢水温度进行软测量,将传统的机理模型与智能方法相结合,并采用改进AdaBoost.RT集成BP网络作为智能模型部分校正机理模型中难以准确获得的参数,再用机理模型进行预测。这种混合模型既克服了传统机理模型难以准确实现的不足也避免了"黑箱"模型过分依赖数据的缺陷。同时改进的AdaBoost.RT集成BP网络算法可以提高传统单神经网络的预测精度和稳定性。实验结果表明,此混合模型具有较好的预测结果,终点温度预测误差不大于±5℃的炉次大于85%。
This paper provides a new hybrid modeling method that combines traditional mechanism model with artificial intelligence method to predict the temperature of ladle furnace. In this new method, an improved AdaBoost. RT using BP network as a weak learning machine is presented to modify the parameters of the mechanism model. Then the temperature prediction result is obtained by the mechanism model. The traditional mechanism method can not predict the temperature exactly and the ＂black box＂ method depends on the data excessively. This new hybrid modeling method could overcome above disadvantages. At the same time, the new method overcomes the shortcoming of BP algorithm. Experiment results show that the new hybrid modeling of AdaBoost. RT could predict the molten steel temperature in LF more exactly, the number of furnace heats with the predictive error of molten steel end temperature in LF is not larger than ± 5 ℃ is greater than 85%.
出处
《仪器仪表学报》
EI
CAS
CSCD
北大核心
2008年第3期662-667,共6页
Chinese Journal of Scientific Instrument
