An Experimental Study on the Stochastic Model Updating of a Structure with Irreducible Parameter Variability and Fixed But Unknown Hyperparameters

The paper provides an experimental evaluation of the sensitivity and Bayesian multilevel approaches to model updating of a system with irreducible (aleatory) parameter variability and fixed but unknown (epistemic) hyperparameters. The research is carried out on a system with three degrees of freedom having coupling springs that facilitate the deliberate introduction of aleatory variability by means of a positioning parameter. In this way, measured natural frequencies from multiple tests are used to recover the tested positional-parameter distributions. Finally, a different arrangement of the system is used to validate the updated model. The contribution of this research is in the experimental comparison of two philosophically different model updating approaches, which is afforded by means of an experimental rig specifically designed so that two natural frequencies are statistically independent. This enables control to be achieved over experimental uncertainty in a straightforward way, thereby illustrating the strengths and possible deficiencies of the two approaches