中文摘要：

贯穿白鹤滩水电站地下洞室群的层间错动带是发育于各岩流层顶部凝灰岩层中的缓倾角、贯穿性的错动构造，受到不同地质作用，在错动带内部形成夹碎裂岩和泥岩等不同构造。由于其形成过程和内部构造相对复杂，因此其力学行为的描述也较为困难，为了直观合理地揭示其力学特性，借助两体理论，将该地质构造考虑为一种上下接触、中间夹软弱带的“夹心饼干”结构。基于层问错动带现场试验规律，建立错动带非线性弹性一硬化本构模型，基于FLAC3D软件中的INTERFACE单元，利用FISH语言实现对接触单元的自动搜索和判断，从而对切向刚度以及摩擦角计算参数进行动态调节；并且依据现场试验成果，利用神经网络对Ks和φ进行回归反演。并且基于该接触模型计算分析了白鹤滩水电站地下洞室群的围岩稳定性，并且利用动态安全系数对错动带的稳定性进行量化评价。

英文摘要：

An interlamination disturbed be station, characterized with low-angle dip an It runs through the underground caverns of Baihetan hydropower d perfoliate formation developed from various tops of tuff. In fact, it is a disturbed shear zone of blended cataclasite and mudstone, etc. under complicated geological processes. As such formation process and internal structure are relatively complicated, the mechanical behaviors are generally difficult to describe. To reveal the structure behaviors more intuitively and reasonably, this paper, using the two-body theory, considers a sandwich biscuit element with its mutual-contacting upper and lower layers and a weak belt clamped in the interlayer, and develops a nonlinear elastic-harden contact constitutive model for this structure based on the field test data. Then, we study this contact element, search and judge it automatically with the INTERFACE element and FISH language of FLAC3D, so that the parameters for calculation of its shear stiffness and friction angle can be regulated dynamically. The coefficients Ks and ~0 are analyzed with a neural network technique for regression of the field test data. Comparison with the field data shows that this model better reflects the mechanical properties of the contact interface. The model was applied to the Baihetan hydropower station for stability analysis of the adjacent rock of its underground caverns, and also for quantitative evaluation of the stability of the disturbed belt based on the dynamical safety parameters.