The intrinsic relations between energy dissipation，energy release，and strength and abrupt structural failure of rocks during failure process are discussed. It is emphasized that the failure of rock is a process in which energy dissipation and energy release occur simultaneously. The energy，part of work done by external forces，dissipates to produce damage and irreversible deformation inside rock and makes the material deteriorate and its strength drop down eventually. The strain energy released in rock volume plays the pivotal role of abrupt structural failure of rock. The concepts of energy dissipation，energy release，strength loss and abrupt structural failure of rock are defined. On the basis of these concepts，the criteria for strength and abrupt structural failure of rock are proposed. The critical stresses at the time of abrupt structural failure of rock under various stress states can be determined by these principles. The energy criteria have been applied to analyzing the failure condition of surrounding rock masses of a circular tunnel in practice.

Based on laboratory data of unsaturated flow in clay，the constitutive model of unsaturated flow coupled model for clay considering plastic strain hardening and Hoek-Brown model is applied to simulate the swelling and shrinking behavior of clay shaft during excavation，concrete lining stage and long time ventilation. With the help of numerical simulation，the flow process in clay from initial saturated to desaturated and then to resaturated is stated. Furthermore，the sensitivity analysis of Biot coefficient，capillary pressure and water relative permeability with saturation degree is discussed；and the results show that the hydraulic parameters have effects on unsaturated zone in surrounding rock mass，especially the water relative permeability，while the mechanical behavior of clay has low effect on unsaturated flow law in surrounding rock mass.

Some hydraulic engineerings have serious hidden troubles or lose their function. By absorbing advanced theories and methods from America etc.，and combining the features in China，the theory system on life diagnosis is put forward for the important hydraulic engineerings. This system includes three rules，namely safety rule，engineering benefit rule，ecology and environment rule. Based on the above rules，this paper proposes the theories，methods and technologies on life diagnosis for the important hydraulic engineerings.

In 3D finite element analysis of structure failure process，large scale numerical analysis has increased the demand for high-performance computing. The element-by-element(EBE) method for distributed memory processors(DMP) is formulated based on the Jacobi-preconditioned conjugate gradient(J-PCG) method. For data exchange，a scheme which only gathers and scatters necessary data is advised to make EBE method available for distributed-memory parallel computers. In this way，it will dramatically reduce data exchange and consequently improve efficiency of parallel computing. At the same time，the formation of global stiffness matrix can be avoided；greatly reducing the requirement for the storage，and the assignment of jobs can be done automatically. A 3D parallel finite element code is developed using MPICH and C/C++ language. Numerical tests on cantilever beam indicate that they are correct. Then it is applied to the finite element analysis of Xiluodu arch dam project and numerical analysis of fracture process of four-point shear test based on grid refining technology. It is the same in essence for EBE method and domain decomposition in task allocation. The results show that for the analysis of the 3D irregular and complicated structures like arch dams，the finite element EBE method is effective and reliable

Particle swarm optimization (PSO) algorithm is a stochastic global optimization technique and has become the hotspot of evolutionary computation because of its excellent performance and simplicity for implementation. In light of the fact that it is hard to determine the parameters of a constitutive model—cohesion weakening and frictional strengthening (CWFS) model，which performs excellently in modeling the extent and depth of brittle failure zone for hard rock under high in-situ stress condition，a new method is presented to identify parameters of CWFS model using PSO. At first，the stochastic values of parameters are initialized and the difference in failure zone between the value computed and the datum measured is regarded as fitness value to evaluate quality of the parameters. Then the parameters are updated continually using PSO until the optimal parameters are found. Thus parameters are identified adaptively during computation. The results of applications to two real tunnels，i.e.，Mine-by tunnel in Canada and Taipingyi tunnel in China，show that the method is feasible and efficient for identifying constitutive parameters and predicting the extent and depth of brittle failure of hard rock under high in-situ stress condition with high precision.

The collapse of the Cypress Viaduct is employed to test the following structural collapse analysis model：a mixed hinge model based on generalized functions is proposed to describe discontinuous displacements；the governing equations are built in the current unstressed configuration；a Lagrange multiplier approach is adopted for the analysis of collisions. After both the viaduct parameters and the ground acceleration recorded at Emeryville are input to the corresponding program，the changing coordinates of each member end during the collapse process are obtained. Visualized clearly by the Visualization of Dynamic Demo Show，the collapse process is consistent with the speculation of the Earthquake Engineering Research Center from UC Berkeley，which proves the validity and utility of the present models. Based on the results，some issues，such as the reason of the collapse，sharp changes of the frame accelerations when collisions happen，and the effects of earthquake lasting times on the weak parts of the structures，are discussed.

The stress equilibrium equation，the continuity equation for water and the elastoplastic matrix are established. By using the Galerkin method，all the governing equations are discretized in the space domain and time domain respectively；and a two-dimensional FEM code for coupled hydro-mechanical elastoplastic analysis of saturated-unsaturated porous medium is developed. By introducing one set of special computation equations related to the unsaturated state，a numerical computation is carried out for the assumed coupled hydro-mechanical problem in an unsaturated soil body. The displacements，the pore pressures，the effective principal stresses，the flow velocities and the plastic zones at different times in the soil body are investigated；and the validity of the code is tested and verified qualitatively.

The Three Gorges Reservoir area is located at the upper reaches of the Yangtze River，where the geological hazards occur with high frequency，especially the major landslides are fully developed at the bank slopes along the river. These landslides are characterized by great amount，extensive distribution and big magnitude，often causing serious loss and casualty in local areas. How to predict landslides is the key to prevent hazards and reduce economical loss and casualties. The rainfall is the dominant exterior factor triggering the landslides. Based on statistic analysis of the data of 112 landslides(volume≥104 m3)，the relationship between landslide and rainfall is derived. It is suggested that the temporal prediction criterion of landslides in the local area be established by the maximum 24-hour rainfall. In terms of the characteristics of landform，lithology，geological structures etc.，the Three Gorges Reservoir area is divided into two parts for landslides prediction. Using the maximum 24-hour rainfall in 72 hours as the prediction index，the occurrence probabilities of landslides are determined in different rainfall conditions for the two subareas. The distinct difference of landslides probabilities between the two subareas indicates that their different mechanism of rainfall induced landslides. Based on the coupling analysis of the geology and meteorology，it is possible to improve the precision of the landslide forecast.

After impoundment up to 139 m elevation in the Three Gorges Reservoir in June 2003，many geological hazards were triggered inevitably. Lots of deformation and failure phenomenon on the bank slope are dooming in new Fengjie，a county located on the Yangtze River. The complexity and diversity of deformation and failure modes of the bank slope bring many difficulties to effectively prevent instability of the bank slope and to release the threat to lives. Based on the research on rock and soil sort of the bank slope and its structural characteristics，the bank slope，which is 21.7 km long totally in 175 m elevation，is divided into 21 sections. The sections are researched respectively on the lithology，obliquity，direct relationship between slope and terrain，interrelation of different rocks and soils，and then，they are reduced to 13 structural types. Through detailed field investigation on the characteristics of the 13 structural types of the bank slopes，6 potential deformation and failure patterns are presented，such as bending-cracking-toppling，shearing-sliding，sphenoid falling-sliding and weathering-flake for rock bank slope，cambered sliding and part sliding-falling for gravel-clay bank slope. These 6 possible patterns will provide guidance for choosing proper protection measures so as to maintain stability of the bank slope.

Based on the comprehensive analysis of the creep experimental data of left bank slope shale samples at Longtan Hydropower Station and the weathered degree of rock masses，rock characteristics and joint distribution in local area，a creep model of shale rock masses of the left bank is set up. The uniform design method and FLAC3D are used to simulate the excavation and creeping process on the simplified geological model of the No.72 test-tunnel，which is located in the central portion of the left bank high slope. With the deformation monitoring data of the No.72 test-tunnel as the aim，the intelligent inversion is made for the creeping parameters by way of combined neural network and genetic algorithm. With these parameters，numerical simulation is conducted. Results indicate that the monitoring displacements not used in inversion are similar to the calculated displacements. This reveals the reasonability of the rheology model and parameters of left bank high slope shale rock masses of Longtan Hydropower Station as well as the important bridging function of intelligent displacement back analysis in the process of the transition from creeping experimental parameters of small-scale rock masses to those of the local large-scale rock masses.

薛强博士论文摘要