Epistemic uncertainty in reliability-based design optimization

Reliability-based design optimization (RBDO) needs to take into account of both aleatory and epistemic uncertainties. It is critical to explore these uncertainty sources and evaluate their impacts on RBDO. This paper provides a comprehensive study of uncertainties, in which the uncertainty sources are listed, categorized and their impacts are discussed. Epistemic uncertainty is of our interest, which is due to lack of knowledge and can be reduced. We specifically discuss the epistemic uncertainties due to unknown constraint function and unknown random variable distribution. The strategies to address epistemic uncertainty are summarized. An I-beam case study is employed to illustrate the impact of epistemic uncertainty on RBDO, in which a Kriging model is used to approximate the unknown true constraint function and the root-mean-square error (RMSE) parameter estimate is used to replace the unknown distribution parameters.