A G-function-based probabilistic design methodology for system reliability

摘要

A new method is developed in this paper to design structures with system reliability constraints. In this method, the g-functions instead of the reliability functions are approximated around the current design and their corresponding MPPs (most probable points) with respect to both the design and random variables. Design variables that are parameters of a random variable are separated from the random variable through a simple transformation. The probability of failure of a structural system is computed using an efficient importance sampling technique that approximates each limit states using a quadratic polynomial with adaptive curvature and position. This system reliability analysis method provides both the accuracy and efficiency needed for the purpose of probabilistic design. A cantilever beam subjected to lateral and vertical random loading is designed using the conventional and the proposed method against excessive lateral and vertical deflection and yielding. The proposed method is five times more efficient than the conventional approach. At the optimal design not all the failure modes are equally important. Therefore, it is important to consider the multiple failure modes as a system not individually during design.