摘要
用神经网络模型代替传统的数学模型,达到提高轧制参数预报精度的目的。在分析了轧制原理的基础上设计了神经网络冷连轧参数预报模型,并针对前向网络反向传播算法(BP)收敛速度缓慢和易陷入局部极小点的缺点,将有全局寻优特性的模拟退火算法(SA)与之结合得到具有较快收敛速度和较高逼近精度的神经网络轧制参数预报模型,提高了网络的快速性和精确性。最后以轧制力预报为例,证明了该方法收敛速度快,稳定性好,可信度高,具有较好的应用前景。
The neural network (NN) models can be used to enhance the prediction of rolling parameters instead of traditional mathematical models. ANN model for tandem cold rolling parameter prediction was designed on the base of rolling theories. Aiming to avoid the disadvantages of the Back-Propagation(BP) Algorithm, of which the speed of constringency is lower and the local minimum is easy to be got, a global optimization algorithm called Simulated Annealing Algorithm is used to help the neural network to possess at high speed and precisely. The high speed of constringency, stability and reliability of the method was confirmed by rolling force prediction.
出处
《钢铁》
CAS
CSCD
北大核心
2008年第7期55-58,90,共5页
Iron and Steel
关键词
冷连轧
轧制力预报
BP神经网络
模拟退火算法
tandem cold rolling
rolling force prediction
BP neural network
simulated annealing algorithm