Evaluation of accuracy and efficiency of some simulation and sampling methods in structural reliability analysis

Numerous simulation and sampling methods can be used to estimate reliability index or failure probability. Some point sampling methods require only a fraction of the computational effort of direct simulation methods. For many of these methods, however, it is not clear what trade-offs in terms of accuracy, precision, and computational effort can be expected, nor for which types of functions they are most suited. This study uses nine procedures to estimate failure probability and reliability index of approximately 200 limit state functions with characteristics common in structural reliability problems. The effects of function linearity, type of random variable distribution, variance, number of random variables, and target reliability index are investigated. It was found that some methods have the potential to save tremendous computational effort for certain types of limit state functions. Recommendations are made regarding the suitability of particular methods to evaluate particular types of problems.