【摘要】 建立在多元回归基础上的传统热分析法存在预测模型函数形式有过强的人为限定,以及对生产条件变化缺乏自适应的问题,为了解决上述问题,本文把BP神经网络算法用于热分析中,构造预测铁液C、Si含量BP神经网络。利用38组实验数据对网络进行训练,获得预测网络模型,然后采用该模型对8组验证样本进行预测,预测C、Si含量的绝对误差分别为0.12%和0.16%。结果表明:该方法能够避免建立模型时人为限定,提高预测精度;对于不稳定的生产条件,具有较强的学习能力和适应能力。更多还原

【Abstract】 In order to eliminate the factitious limit on function form and improve adaptability to unstable production conditions in the traditional thermal analysis based on multiple regression,a BP neural network algorithm for thermal analysis has been built and used in predicting carbon,silicon content in molten iron.38 groups of the experimental samples were used to train the network and 8 groups are used to verify the network.The absolute errors of predicted C,Si content were 0.12% and 0.16% respectively.The results show that this method can avoid the factitious limit on modeling to improve prediction accuracy,as well as has strong learning ability and adaptability for unstable production conditions.更多还原