جستجوی مقالات مرتبط با کلیدواژه « design point » در نشریات گروه « عمران »

Monte Carlo simulation (MCS) is a useful tool for computation of probability of failure in reliability analysis. However, the large number of samples, often required for acceptable accuracy, makes it time-consuming. Importance sampling is a method on the basis of MCS which has been proposed to reduce the computational time of MCS. In this paper, a new adaptive importance sampling-based algorithm applying the concepts of first-order reliability method (FORM) and using (1) a new simple technique to select an appropriate initial point as the location of design point, (2) a new criterion to update this design point in each iteration and (3) a new sampling density function, is proposed to reduce the number of deterministic analyses. Besides, although this algorithm works with the position of design point, it does not need any extra knowledge and updates this position based on previous generated results. Through illustrative examples, commonly used in the literature to test the performance of new algorithms, it will be shown that the proposed method needs fewer number of limit state function (LSF) evaluations.

This paper focuses on the design point and the failure probability of problems with continuous random variables. The charged system search (CSS) algorithm is utilized as the optimization tool to achieve minimum reliability index under limit state function. In order to acquire the optimal solution, random variables such as elastic modulus, loads, and geometric parameters are selected as decision variables of the problem which are optimized by means of the CSS algorithm. This algorithm is inspired by the Coulomb and Gauss’s laws of electrostatics from physics. In order to evaluate the accuracy and efficiency of this algorithm, several numerical examples are studied and the results are compared to those of the existing methods. The proposed method is capable of finding a design point over the failure surface and calculates the reliability index with a reasonable accurately. As the proposed framework enforces low computational time and holds a satisfactory convergence rate, it is a competent methodology to calculate different types of reliability problems.