An Improved Point Estimate Method for Probabilistic Risk Assessment

Rosenblueth's probabilistic point estimate method is a computationally straightforward technique for uncertainty analysis of engineering problems. It is capable of estimating the statistical moment of any order of a model output involving several stochastic variables that are correlated or uncorrelated and symmetric or asymmetric. However, in multivariate problems with more than two stochastic variables involved, the Rosenblueth's method is not able to provide a unique solution, rather an approximate solution to indeterminate problems. This is attributed to the fact that the number of unknowns to be solved is larger than the number of governing equations that are provided. An improved point estimate method is proposed to circumvent this drawback and increase the computational efficiency as well. It is demonstrated that the proposed point estimate method preserves the capabilities of the Rosenbleth's PE method but resolves the associated non-uniqueness problem.