Uncertainties quantification of fatigue load mixture model using hierarchical Bayesian models

Abstract

There are two primary challenges in current fatigue load probability modeling. Firstly, it is difficult to measure the estimation errors in the mixture model for fatigue loads, particularly in the tails with low-probability and high-stress levels. Secondly, the component number of the mixture model cannot be observed. To address these challenges, this research introduces the hierarchical Bayesian mixture model and the Dirichlet process prior. A relative error measure is proposed to reveal the errors of the density tails. The results of an illustrative example show significant relative errors in the tails and a discrete distribution of the mixture number.