The initial in situ stress field is the fundamental factor causing the deformation and failure of underground engineering and is an important basis for the feasibility analysis, design, and construction of underground engineering. However, it is difficult to obtain the whole in situ stress field of large-scale underground engineering in difficult and dangerous areas by field measurement. In view of the fact that the measured in situ stress components (σ_xx, σ_yy, σ_zz, τ_xy, τ_xz, τ_yz) of Sichuan-Tibet Railway in China are linear with the buried depth, a method is proposed to solve the in situ stress by applying corresponding loads to all unit bodies in the calculation area based on BP neural network and FLAC^3D. Through this method, the in situ stress of the tunnel is inverted. The results show that both the maximum principal stress and minimum principal stress increase with the increase of buried depth, and when the tunnel passes through faults or anticlines, the main stress will suddenly drop. Furthermore, compared with the results of the multiple linear regression method, it is found that the proposed method has higher accuracy; especially for the simulation of the maximum horizontal principal stress and vertical stress, the average relative error is reduced by 26.44% and 77.27%, respectively. The research in this paper can provide a new idea for the initial in situ stress inversion of engineering.

中文翻译：

---

基于BP神经网络的初始地应力场反演方法及单元体的荷载施加。

初始地应力场是引起地下工程变形和破坏的根本因素，是地下工程可行性分析，设计和施工的重要基础。然而，通过现场测量很难获得大型地下工程在困难和危险区域的整个现场应力场。鉴于此原位应力分量测定（以下事实σ _XX，σ _yy的，σ _ZZ，τ _XY，τ _XZ，τ _YZ）川藏铁路线与埋深成线性关系，提出了一种基于BP神经网络和FLAC ^3D对计算区域内所有单元体施加相应载荷来解决地应力的方法。通过这种方法，隧道的原地应力被反转。结果表明，最大主应力和最小主应力均随埋深的增加而增大，当隧道通过断层或背斜时，主应力会突然下降。此外，与多元线性回归方法的结果相比，发现该方法具有较高的准确性。特别是对于最大水平主应力和垂直应力的仿真，平均相对误差分别降低了26.44％和77.27％。本文的研究可以为工程的初始地应力反演提供一个新的思路。