In terms of the frequent occurrence and much trouble in governance of the disaster caused by water and mud inrush in deep and long tunnels, the risk assessment model based on normal grey cloud clustering method was proposed. Taking the Jigongling Tunnel of Fanba Expressway as an example, firstly the evaluation target was divided into 8 clustering indices and 4 grey categories according to the grey clustering method. In order to avoid the defects that the traditional whitenization weight functions could not give a good description of system's randomness and ambiguity, the cloud model was introduced to improve it. Then the whitenization weight values were discretized by using the one-dimensional forward cloud generator to simulate the uncertainties in engineering, and the normal grey cloud whitenization weight functions were established. Afterwards, combined with the engineering data of Jigongling Tunnel collected on site, the clustering weight of each clustering index was analyzed under specific engineering and the clustering coefficient of the target was determined. Lastly the risk of water and mud inrush in Jigongling Tunnel was evaluated using the model. The results, which showed that the risk of water and mud inrush in target D₁, D₂ and D₃ was respectively medium, extremely high and high, were compared with the excavation data. The two coincided with each other well which indicated that the model had a certain engineering value and could provide reference for related engineering.

A probabilistic approach is proposed for face stability analysis of subway tunnels, which begins with establishment of multiple failure mechanisms, including active failure and passive failure. In terms of the complex conditions during subway construction, the buried depth, surcharge load imposed on the ground, and pore water pressure are introduced to account for the extrinsic factors that probably affect the face stability. The probabilistic analysis is performed by consolidating the multiple failure mechanisms into response surface method (RSM). The cohesion and internal friction angle of soils are considered as random variables, and the remaining parameters are considered as nonrandom variables due to their low variability or determinacy for specific projects. The influence of nonrandom variables and random variables on failure probability is investigated in detail. Finally, a reliability-based design chart is provided to evaluate the target supporting pressure against tunnel face according to different coefficients of variation of random variables. In short, by taking into account of all the possible face failures in one performance function, the proposed approach has the virtue of catching the global behaviors of face failure when multiple failure mechanisms occur at tunnel face.