沈阳地铁开挖引发地面沉降的BP神经网络反分析

以沈阳地铁一号线重启区间隧道为研究背景，采用基于均匀试验设计法的有限元数值分析法确定隧道围岩土体物理力学参数与地面沉降之间的关系，并以此作为神经网络的输入样本，通过BP神经网络对样本的训练、学习，建立隧道围岩土体力学参数与地面沉降之间的映射关系，然后利用这种映射关系，根据地铁开挖引发的地面沉降实测值反演岩土体的物理力学参数，最后根据参数反演结果，建立有限元应力应变模型预测地面沉降，并与实测值相比较，以检验BP神经网络地面沉降位移反分析方法的有效性和合理性。

BACk ANALySIS BASED ON BP NEURAL NETWORk OF gROUND SETTLEMENT By ShENyANg METRO EXCAVATION

In this paper, taking Zhonggong Street Station-qigong street station of Shenyang metro line 1 as the research background, and using numerical analysis method based on the uniform experimental design method, the relationship between mechanical parameters of surrounding rock of tunnel and the ground settlement is determined, which is regarded as the neural network input samples. Based on BP neural network training and learning, the mapping relationship between soil mechanics of tunnel surrounding rock parameters and the ground settlement is established. Using this mapping relationship, the parameter inversion of rock and soil is carried on according to ground settlement value caused by ground subway excavation. According to the results of parameters inversion, a finite element model for stress and strain is established for predicting the surface subsidence, and compared with the measured values to test the Validity and rationality of the displacement back-analysis by BP neural network.