LAUSR

Abstract Over the past few decades, the Constitutive Relation Error (CRE) theory has proven to be a valuable numerical technique for model identification in engineering systems. Under certain conditions, this technique allows predicting the global state of connections starting from an observed structural vibratory response. These conditions often assume deterministic knowledge about input parameters and mechanical modelling. As with most inverse problems, the inherent uncertainties in the system can drastically affect the parameters of interest, posing an additional difficulty in the identification process. The present study proposes a simple and direct way to include and evaluate the effect of uncertainties in the CRE-based process of identifying parameters of semi-rigid structural connections. The numerical investigations are carried out in two parts. The first, from an exclusively deterministic perspective, shows the main characteristics of the method such as the possibility of using partial data and the inclusion of flexible support conditions. The second part explores the limits of the suggested formulation in a non-deterministic context considering both measurement noise and material uncertainty. The results show that the presented CRE formulation can identify the parameters properly in various situations. However, high noise levels in the reference measures may deteriorate identification process.