The in situ stress distribution is one of the driving factors for the design and construction of underground engineering. Numerical analysis methods based on artificial neural networks are the most common and effective methods for in situ stress inversion. However, conventional algorithms often have some drawbacks, such as slow convergence, overfitting, and the local minimum problem, which will directly affect the inversion results. An intelligent inverse method optimizing the back-propagation (BP) neural network with the particle swarm optimization algorithm (PSO) is applied to the back analysis of in situ stress. The PSO algorithm is used to optimize the initial parameters of the BP neural network, improving the stability and accuracy of the inversion results. The numerical simulation is utilized to calculate the stress field and generate training samples. In the application of the Shuangjiangkou Hydropower Station underground powerhouse, the average relative error decreases by about 3.45% by using the proposed method compared with the BP method. Subsequently, the in situ stress distribution shows the significant tectonic movement of the surrounding rock, with the first principal stress value of 20 to 26 MPa. The fault and the lamprophyre significantly influence the in situ stress, with 15–30% localized stress reduction in the rock mass within 10 m. The research results demonstrate the reliability and improvement of the proposed method and provide a reference for similar underground engineering.

Surrounding rock deformation and consequent support failure are the most prominent issues in red-bed rock tunnel engineering and are mainly caused by the effects of unloading, rheology, and swelling. This study investigated the mechanical responses of two kinds of red-bed mudstone and sandstone under unloading conditions via laboratory observation. Volume dilation was observed on the rocks during unloading, and the dilatancy stress was linear with the initial confining pressure. However, the ratios of dilatancy stress to peak stress of the two rocks kept at a range from 0.8 to 0.9, regardless of confining pressures. Both the elastic strain energy and the dissipated energy evolved synchronously with the stress–strain curve and exhibited conspicuous confining pressure dependence. Special attention was paid to the evolution behavior of the dilatancy angle. The dilatancy angle changed linearly during unloading. When the confining pressure was 10 MPa, the dilatancy angle of mudstone decreased from 26.8° to 12.5° whereas the dilatancy angle of sandstone increased from 34.6° to 51.1°; when the confining pressure rose to 25 MPa, the dilatancy angle of mudstone and sandstone decreased from 45.8° to 17.4° and increased from 21.7° to 39.5°, respectively. To further understand the evolution of the dilatancy angle, we discussed the links between the variable dilatancy angle and the processes of rock deformation and energy dissipation.

Landslides are usually caused by rainstorms and geological processes such as earthquakes and may have a massive impact on human production and life. The hazard chain of landslide–river blockage–outburst flood is the most common hazard chain caused by landslides. A database based on existing landslide cases was established in this study to investigate the assessment formulas of the risk of river blockage, dam stability, and peak flood discharge after a dam has broken. A risk assessment model of the landslide–river blockage–breaching hazard chain was established, including the downstream vulnerability. The case of the Baige landslide verified the applicability of the model. This model can be used in a landslide-prone area to predict whether a relatively massive river blockage will form after the landslide occurs, whether the landslide dam formed by the river blockage will breach in a short time, and the impact of the outburst flood on the downstream area.

The void compression stage causes porous cement mortar to present special mechanical properties. In order to study the compaction behavior and the damage evolution of the porous material, cement mortar specimens with an average porosity of 26.8% were created and cyclic uniaxial compression tests were carried out. The irreversible strain accumulated in the tests was obtained by cyclic loading and unloading. As the secant modulus of the porous cement mortar increases with stress in the pre-peak deformation stage, its damage variable is defined according to the accumulated irreversible strain instead of modulus degradation. The strain-based damage indicator fitted with the damage evolution law is characterized by linear accumulation at the beginning and has an acceleration rate of about 0.3 in the pre-peak deformation stage, and the damage value converges to 1 at failure. Based on the Weibull distribution, a constitutive damage model of porous cement mortar is improved by considering both the damage evolution during the plastic deformation stage and the mechanical behavior in the compaction stage. The theoretical envelope curves obtained by the constitutive model are in good agreement with the experimental envelope curves of cyclic uniaxial compression in the compaction and pre-peak stages, and the average absolute error is about 0.54 MPa in the entire pre-peak stage, so the proposed damage constitutive model can characterize the damage-induced mechanical properties of porous cement mortar in the compaction and pre-peak stages.

The surrounding geological conditions and supporting structures of underground engineering are often updated during construction, and these updates require repeated numerical modeling. To improve the numerical modeling efficiency of underground engineering, a modularized and parametric modeling cloud server is developed by using Python codes. The basic framework of the cloud server is as follows: input the modeling parameters into the web platform, implement Rhino software and FLAC^3D software to model and run simulations in the cloud server, and return the simulation results to the web platform. The modeling program can automatically generate instructions that can run the modeling process in Rhino based on the input modeling parameters. The main modules of the modeling program include modeling the 3D geological structures, the underground engineering structures, and the supporting structures as well as meshing the geometric models. In particular, various cross-sections of underground caverns are crafted as parametric modules in the modeling program. The modularized and parametric modeling program is used for a finite element simulation of the underground powerhouse of the Shuangjiangkou Hydropower Station. This complicated model is rapidly generated for the simulation, and the simulation results are reasonable. Thus, this modularized and parametric modeling program is applicable for three-dimensional finite element simulations and analyses.