【摘要】 针对传统预测模型建立中的弱点,分析了转炉炼钢的输入输出参数,建立了三输出转炉碳温分类预测模型;根据不同数据范围所对应的不同规律,对数据源进行了交叉分段处理;为提高模型训练速度,采用LM算法进行训练,同时针对该算法容易陷入局部极小和过学习的缺陷,采用了多种方法进行处理,实验仿真证明,将训练数据源进行分段建立分类预测模型,能显著提高网络的泛化能力,结合三输出模型结构,有效地提高了转炉终点碳含量和温度的预测百分比。更多还原

【Abstract】 Aiming at the weakness in traditional prediction model establishment,a prediction model with three outputs was developed by analyzing the input and output parameters.According to the laws corresponding with the different data,the train data were splitted to two cross sections.In order to enhance the training speed of the model,LM algorithm was used.For the algorithm vulnerable to a local minimum and over learning,a variety of methods were introduced.The simulation shows that the network capacity was enhanced significantly by cross splitting.The hitting percentage was improved effectively with the three outputs model.更多还原