论文部分内容阅读
针对电弧炉炼钢终点的重要性,建立了基于MATLAB的BP神经网络对电弧炉炼钢终点碳含量及温度预报的模型,运用传统试凑法和单向调节法相结合的方法,可快速地找到最佳隐含层节点数,节省网络训练时间,具有实用性。对比预报线性回归相关系数可知,在0.01显著水平下终点碳含量和终点温度的相关系数均通过显著性检验;终点碳含量和温度的预报值和实测值的整体相关系数分别为0.899和0.820,均高于0.75,表明预测值与实测值的相关性十分显著,该网络模型具有极好的预报性;终点碳含量和温度在误差范围为±0.02%和±10℃内的预报命中率达到91.2%和94%,表明运用神经网络对电弧炉炼钢终点进行预报是有效的、可行的。
Aimed at the importance of EAF steelmaking endpoint, a model based on MATLAB for BP neural network to predict the carbon content and temperature at the end of EAF steelmaking was established. By combining the traditional trial and error method and one-way adjustment method, The best hidden layer nodes, save network training time, with practicality. The correlation coefficient of linear regression of forecasting forecast shows that the correlation coefficient between the end point carbon content and the end point temperature has passed the significance test at 0.01 significant level. The overall correlation coefficients between the predicted and measured end point carbon content and temperature are 0.899 and 0.820, respectively Higher than 0.75, indicating that the correlation between the predicted value and the measured value is very significant, and the network model has excellent predictability; the prediction yield of the end point carbon content and temperature within the error range of ± 0.02% and ± 10 ℃ reaches 91.2% And 94%, respectively, indicating that it is effective and feasible to predict the EAF steelmaking endpoint by using neural network.