چكيده به لاتين
Existence of uncertainty in real world has caused researchers to pay more attention to this subject. Different examples of uncertainty in structural engineering can be observed in external loads exerted on the structure, properties of construction material and details of the geometry of the structure. Thus, in order to reach a realistic analysis of the behavior of a structure, uncertainty must be somehow taken into account. Structural reliability theory provides a tool for including the effects of uncertainty in the procedure of structural analysis. The main duty of reliability methods is to compute probability of failure or reliability index such that the level of reliability to a structure can be evaluated.
The purpose of this thesis is to extend the reliability methods in order to reach efficient methods in computation of probability of failure and reliability index. In the proposed methods the main emphasis is concentrated on improving the capability to encounter nonlinear problems and on increasing efficiency. Since, in the field of reliability, the so-called approximate methods can express balanced performance in accuracy and computational cost, they are considered the most practical methods in this field. Therefore, by relying on approximate methods as the basis of the current research, the proposed methods try to develop and extend them to more capable and efficient methods for different problems including linear and nonlinear ones. On this way two reliability methods have been proposed. The first one, which falls absolutely in the class of approximate methods, is a generalization for one of the most important basic methods of this class. By the proposed generalization, the capababilities of the basic method have been promoted such that the possibility for controlling the convergence speed has been created. Besides, in the problems the basic method cannot even converge, the proposed method has created the possibility for transmission from divergence to convergence. The second proposed method benefits from the combination of the concepts of several classes of reliability. This hybrid method can be used when the first proposed method encounters limitations due to insufficient information about limit state function. In fact for such situations, in which the aforementioned limitations usually lead researchers to the time-consuming simulation methods, the present research proposes the hybrid method which is much more practical than simulation methods. The analysis results of the commonly used examples of the reliability literature implies the appropriate performance of the proposed methods, each compared to its own corresponding methods.
