暗挖大跨地铁车站增量位移反分析研究

Back Analysis of the Incremental Displacement of a Long-Span Bored Metro Station

文章采用均匀设计法对围岩密度、弹性模量、泊松比、粘聚力、内摩擦角5个因素的16个水平进行了数值仿真试验设计,并借助有限差分法对暗挖大跨地铁车站施工进行了三维计算,获得了不同试验设计条件下不同工序的地表沉降位移值。以计算结果中临时支撑拆除引起的11个观察点的增量位移作为BP神经网络的输入集,以相应工况的岩土体参数作为BP神经网络的输出集,完成了BP神经网络的训练和学习工作;然后将现场实测到的临时支撑拆除导致的位移作为BP神经网络的输入参数,反演出了相应的岩土体参数;最后将该套反演参数代入FLAC3D进行三维仿真计算,得到了岩土体的位移结果,并采用现场实测的7个测点的地表沉降增量对该套参数的合理性进行验证。结果表明,以临时支撑拆除这一工序引起的地表沉降增量为考察指标,采用均匀设计法和BP神经网络可以快速有效地获得暗挖大跨地铁车站围岩的等效物理力学参数,指导暗挖大跨地铁车站的设计与施工,保证工程施工和周边环境安全。

In this paper, a uniform design method is adopted to design a numerical simulation test with 16 levels of 5 factors（surrounding rock density, elasticity modulus, Poisson ratio, cohesive force and internal friction angle） and a 3D calculation is conducted for the construction of a long-span bored metro station by the finite difference method to obtain the values of surface subsidence for different experimental conditions and different procedures. The incremental displacements of 11 observation points induced by the demolition of the temporary support and corresponding rock-soil parameters are used as the input set, and the output set of a BP neural network helps to complete the training and learning. Additionally, the measured displacements induced by the demolition of the temporary support are taken as input parameters and the corresponding rock-soil parameters were inversed. The displacements of the rock-soil mass were obtained by putting all the inversion parameters into FLAC3D for a 3D- simulated calculation, and the rationality of those parameters are verified by measured increments of surface subsidence at 7 measuring points. The results show that the equivalent physical and mechanical parameters of the surrounding rock of a long-span bored metro station can be effectively obtained by a uniform design method and BP neural network in light of the surface subsidence increments induced by the temporary support demolition.