论文部分内容阅读
针对传统逻辑漏钢预测系统稳定性差、收敛速度慢、收敛精度低等缺点,建立具有自组织、自学习等功能的误差反向传播BP神经网络预测模型.采用变步长并加入动量项、防振荡项等方法,使网络训练过程能够跳出局部极小,加快了收敛速度.系统改变以往只将温度数据作为输入参数的传统,将拉速、中间包钢水温度作为考虑因素,扩大了漏钢因素的考虑范围.实验结果表明,采用BP神经网络对某炼钢厂实际数据进行漏钢预测,预报结果准确,具有较好的在线应用前景.
Aiming at the shortcomings of the traditional logic breakout forecasting system, such as poor stability, slow convergence rate and low convergence precision, a BP neural network prediction model with self-organization and self-learning function is established.By variable step size and adding momentum term, Oscillation and other methods to make the network training process can jump out of local minimum, speed up the convergence rate.System change in the past only the temperature data as the input parameters of the traditional, the pull speed, tundish temperature as a factor to expand the breakout Factors to consider.Experimental results show that using BP neural network to predict the actual steel breakout of a steelmaking plant, the forecast result is accurate and has a good online application prospect.