Estimation of coefficient of variation for structural analysis: The correlation interval approach

The paper is focused on the efficient estimation of the coefficient of variation for functions of correlated and uncorrelated random variables. Specifically, the paper deals with time-consuming functions solved by the non-linear finite element method. In this case, the semi-probabilistic methods must reduce the number of simulations as much as possible under several simplifying assumptions while preserving the accuracy of the obtained results. The selected commonly used methods are reviewed with the intent of investigating their theoretical background, assumptions and limitations. It is shown, that Taylor series expansion can be modified for fully correlated random variables, which leads to a significant reduction in the number of simulations independent of the dimension of the stochastic model (the number of input random variables). The concept of the interval estimation of the coefficient of variation using Taylor series expansion is proposed and applied to numerical examples of increasing complexity. It is shown that the obtained results correspond to the theoretical conclusions of the proposed method.

Automated summary

The paper focuses on efficient estimation of the coefficient of variation for functions of correlated and uncorrelated random variables. It shows that modifying Taylor series expansion for fully correlated random variables can significantly reduce the number of simulations. The concept of interval estimation is proposed and applied to numerical examples, with results corresponding to the theoretical conclusions.