摘要
针对常规Elman网络泛化能力差的缺点,以及工业生产中对高精度板形预测模型的需要,用人工蜂群算法(ABC)代替误差反传算法训练Elman网络,建立了一个基于Elman网络的板形预测模型.神经网络的隐层节点数通过经验公式和仿真试验来确定.通过仿真验证,用人工蜂群算法训练的Elman网络在同等条件下比常规Elman网络具有更强的泛化能力,其板形预测精度更高.
Because the ability of generalization of routine Elman network is poor, and the need of industry for flatness prediction model with high precision is eager, a flatness prediction model based on Elman network that is trained by artificial bee colony （ABC） algorithm instead of error back propagation algorithm is presented. The number of the hidden node is determined by experiential formula and simulation experiments. The results of simulation show that the model trained by ABC has a more powerful ability of generalization than routine Elman network under the same conditions and the accuracy of flatness prediction is higher.
出处
《沈阳大学学报（自然科学版）》
CAS
2012年第3期38-42,共5页
Journal of Shenyang University：Natural Science
基金
国家自然科学基金资助项目(50675186)
关键词
人工蜂群算法
ELMAN网络
板形预测模型
泛化能力
artificial bee colony algorithm
Elman network
flatness prediction model
the ability ofgeneralization
