Stochastic Load Identification And Reliability Analysis Of Underground Structures

Underground space structures such as metro, utility tunnel, etc., are important components of a city's lifeline system. So ensuring the safety and functional performance of large underground structures has already been an important subject of researches on disaster prevention and mitigation for underground engineering of cities. In this context, the safety monitoring and pre-warning of underground structure should be put more attention on.In this thesis, safety monitoring, load identification and reliability analysis of underground structures are systematically studied in turn. Firstly, for the deficiency of traditional monitoring technologies, new types of sensors, e.g. fiber Bragg grating sensor, diaphragm pressure sensor, wireless network sensor, etc., are investigated in detail. Since all of the above mentioned technologies can realize large-scale and distributed monitoring, so a new path of in-situ monitoring can be created with their help.Secondly, spline function approximation of actual complex distribution of loads on underground structures is introduced and the loads are identified with this approach as basis. Then the application of Genetic Algorithms (GA) to optimal inverse analysis is discussed. Compared with classical optimal algorithms, the colony searching characteristics of GA helps to find the global optimal solution more easily. Moreover, the stochastic load identification approach based on probability density evolution method (PDEM) is proposed. The recently developed PDEM can reveal the substantial probability information among sampling points in a stochastic system, so it can not only truly reflect the objective system, but also significantly lower the difficulty and work of problem solving. Utilizing this stochastic load identification approach, the statistical characteristic value and the probability density function of the identified loads can be obtained at the same time.Thirdly, the PDEM based reliability analysis of underground structures is developed. Using this method to calculate the probability density surface of structure response is more accurate and more efficient than using traditional stochastic analysis methods. Furthermore, the system reliability analysis of structures based on equivalent extreme value event is introduced and applied to designs of underground structures under complex failure criteria. The results clearly prove the validity of the proposed method. Also a human-computer interaction interface of the load identification and stochastic reliability analysis codes is fulfilled for possible future users.Finally, problems need further studies are discussed.