摘要
传统神经网络算法不可避免会出现局部极值问题,可能导致训练失败。因此,作者在分析了隧道拱顶下沉规律及其主要影响因素的基础上,采用基于遗传算法的神经网络建立了隧道拱顶下沉时序的预测模型。模型在綦万高速公路麒麟寺隧道施工中成功应用,结果表明采用基于遗传算法的神经网络能够有效避免局部极值问题,收敛速度较快,能对隧道拱顶下沉时序进行较为准确的预测。
In the traditional neural network algorithm the partial extremal problem can＇t be avoided, which may cause the failure of the training. After analyzing the vault crown settlement tendency of the tunnel and the factors affected it, the Genetic algorithm neural network model for predicting the vault crown settlement has been created. The model has been applied successfully to predict the vault crown settlement tendency of the Qilinsi Tunnel in the Qiwan Expressway, which proves that in the Genetic algorithm the partial extremal problem can be avoided and it is feasible to predict the settlement tendency of tunnel with the Genetic algorithm neural network model and this method is convenient and correct.
出处
《地下空间与工程学报》
CSCD
2006年第4期547-550,共4页
Chinese Journal of Underground Space and Engineering
基金
国家自然科学基金重点资助项目(No.50334060)
国家自然科学基金青年基金资助项目(No.50104013)
关键词
遗传算法
神经网络
隧道
拱顶下沉
genetic algorithm
neural network
tunnel
vault crown settlement
