基于优化BP神经网络算法的隧道岩土体力学参数分析方法研究  

Study on Tunnel Geotechnical Physical Parameters Analysis Method based on Optimized BP Neural Network Algorithm

作　　者：龚盛 杨柱 张国鹏 罗曦 陈兴周[1] GONG Sheng;YANG Zhu;ZHANG Guopeng;LUO Xi;CHEN Xingzhou(Department of Architecture and Civil Engineering,Xi'an University of Science and Technology,Xi'an 710054;PowerChina Northwest Engineering Corporation Limited,Xi'an 710065,China)

机构地区：[1]西安科技大学建筑与土木工程学院,西安710054 [2]中国电建集团西北勘测设计研究院有限公司,西安710065

出　　处：《西北水电》2022年第5期133-137,共5页Northwest Hydropower

基　　金：国家自然科学基金(U1965107、51979218);陕西省自然科学基金项目(2018JM5118).

摘　　要：在隧道稳定工程稳定分析中,岩土体力学参数的准确性对工程设计有着重要影响,以获取隧道岩土体力学参数指导支护设计,保障隧道施工及运营期的安全性。采用有限差分数值计算模型构建位移与岩土参数的训练样本集,结合CO-RDPSO算法优化BP神经网络的初始权值和阈值进行训练,将实际监测位移带入训练好的神经网络反演岩土体力学参数,利用反演参数带入有限差分模型计算得到隧道位移,并将其与监测位移进行对比分析,并对该方法进行实例验证,结果表明,基于有限差分方法反演得到的参数能较好地反映隧道开挖后土体的位移沉降,变形规律与监测数据基本一致,CO-RDPSO算法优化BP神经网络具有一定的可行性。In the study of tunnel stability,the accuracy of the mechanical parameters of rock and soil has an important influence on the engineering design.In order to obtain the mechanical parameters of the rock and soil in the tunnel to guide the support design and ensure the safety during construction and operation of the tunnel.The finite difference numerical model is used to construct the training sample set of displacement and geotechnical parameters.Combined with the CO-RDPSO algorithm,the initial weights and thresholds of the BP neural network are optimized for training,and the actual monitoring displacement is brought into the trained neural network for geotechnical inversion.The inversion parameters are substituted into the finite difference model to calculate the displacement of the tunnel,and compare it with the monitoring displacement.The method is further verified by a numerical example.The results show that the parameters obtained by inversion based on the finite difference method reflect well the displacement and settlement of the soil mass after tunnel excavation,and the deformation law is basically consistent with the monitoring data.Therefore,the BP Neural network optimized by CO-RDPSO algorithm is feasible.

关 键 词：隧道工程 反分析方法 BP神经网络 粒子群算法 交叉算子 

分 类 号：TU452[建筑科学—岩土工程]