摘要
基于卸荷岩体力学分析方法建立了考虑岩体强、弱卸荷区地下洞室二维ADINA有限元模型,建立了BP神经网络模型,采用搜索算法确定了训练误差最小网络训练参数,将训练好的网络保存并用于仿真,利用工程监测的位移资料反演了岩体主要参数,参考卸荷岩体力学参数折减方法确定了卸荷区岩体参数,并将反演的参数代入有限元模型计算,计算位移值与监测位移值的对比结果表明,二者吻合较好,反演的参数可靠性较大。
Based on the unloading rock mass mechanics analysis, two dimensional ADINA finite element method of underground chamber is built with considering strong and weak excavation disturbed zone. BP neural network model is established. Its parameters are determined by the search algorithm which leads to the minimal training error. The trained model is saved for simulation. Project displacement monitoring data is used to inverse the main rock mass parameters by the BP neural network. The parameters of rock mass in the excavation disturbed zone are determined by the analysis method of unloading rock mass mechanics. Inversion parameters are put into the finite element model. Compared with the computed displacement and monitored value, it finds that the value of computed displacement is close to the monitored value. Inversion parameters have high reliability.
出处
《水电能源科学》
北大核心
2010年第3期98-100,114,共4页
Water Resources and Power
关键词
地下洞室
参数反演
BP神经网络
岩体卸荷
位移
underground chamber
inversion parameter
BP neural network
unloading rock mass
displacement
