ABSTRACT
In stochastic model updating, hybrid uncertainties are typically characterized by the distributional p-box. It assigns a certain probability distribution to model parameters and assumes its hyper-parameters as interval values. Thus, regardless of the updating method employed, the distribution family needs to be known a priori to parameterize the distribution. Meanwhile, a novel class of the random variable, called staircase random variable, can discretely approximate a wide range of distributions by solving moment-matching optimization problem. The first author and his co-workers have recently developed a distribution-free stochastic updating framework, in which model parameters are considered as staircase random variables and their hyper-parameters are inferred in a Bayesian fashion. This framework can explore an optimal distribution from a broad range of potential distributions according to the available data. This study aims to further demonstrate the capability of this framework through a simple numerical example with a parameter following various types of distributions.
