Abstract
Modal parameters such as natural frequencies, modal shapes, and the damping ratio are useful to understand structural dynamics of mechanical systems. Modal parameters need to be estimated under operational conditions for use in structural health monitoring. Therefore, operational modal analysis (OMA) without input signals has been proposed to easily extract modal parameters under operational conditions. Recently, OMA for under-determined systems with more active modes than measurement outputs has been investigated to reduce the number of sensors. This study proposes the OMA framework for under-determined systems based on Bayesian CP (CANDECOMP/PARAFAC) decomposition of second-order statistical data. The proposed method enables us to extract the modal parameters from under-determined systems without tuning the number of active modes, because the rank of the tensor data corresponding to the number of active modes is automatically determined via Bayesian inference. The effectiveness of this method is demonstrated using artificial vibration data of a mass-spring system under operational and under-determined conditions.
