Moment quadrature method for uncertainty quantification of fatigue damage prognosis

Abstract

The study proposes a new method for uncertainty quantification of fatigue damage prognosis using very few, typically 4, model evaluations, while achieving an accuracy comparable to the Monte Carlo method normally requires 106 model evaluations. Higher order statistics such as skewness and kurtosis can reliably be obtained using the method with 9 to 16 model evaluations. The proposed method provides a viable solution to problems where the computational demand is prohibitively high for conventional methods and/or real time computation is needed. Validations are made using several fatigue examples with realistic testing data, and results are compared with Monte Carlo simulations.