The analysis of random systems with combined parametric and non-parametric uncertainty models

In recent years there have been advances in the use of non-parametric models of uncertainty to predict the response statistics of random structures. With this approach, it is assumed that various components of the system are sufficiently random for universal statistical distributions to be applied directly to the natural frequencies and mode shapes, thus obviating the need for a detailed description of the underlying random physical parameters. A hybrid finite element (FE)/statistical energy analysis (SEA) method has previously been developed based on this approach, in which components are considered to be either deterministic or highly random. In the present paper this method is extended by applying parametric uncertainty models to components that are not highly random. Both probabilistic and possibilistic models are considered, and the method is illustrated by application to an example built-up plate system.