INSTANTANEOUS MODAL IDENTIFICATION UNDER EARTHQUAKE-INDUCED VIBRATION FOR STRUCTURAL ASSESSMENT

Tracking the health condition of existing structures has become a growing necessity in order to optimise preventive conservation strategies and reduce the maintenance costs associated with retrofitting interventions. In this context, the dynamic characterisation of the structural system through the execution of ambient vibration tests and the application of Structural Health Monitoring (SHM) protocols is fundamental to keep under control the structural performance under operational conditions and timely identify deviations from the expected behaviour. Conventional vibration analysis methods work very well for this purpose, but they are based on the principle of mode superposition, which does not apply in the presence of strong nonlinearities and non-stationary input excitations, such as earthquake actions. This poses remarkable limitations to the possibility to exploit operational modal analysis methods for seismic SHM purposes. Strong ground motions can lead to major damages and collapses, but repeated earth-shakings of small or medium intensity can yield cumulative damages – often unseen to the naked eye – which must be detected in real time in order to safely manage rescue operations and support civil protection activities in the aftermath of natural disasters. This paper arises to address the afore-mentioned drawback by proposing a novel non-parametric method for time-dependent modal analysis in the presence of nonlinear earthquake-induced damage phenomena. Making use of output-only vibration measurements in non-stationary conditions, the proposed algorithm can track the fundamental dynamic features of a structure as time-dependent functions for the complete observation period, enabling to catch instantaneous variations in the modal estimates during transient phases. The efficiency of the method is tested through the case study of a historic masonry tower, whose geometrical and mechanical features have been defined on the basis of a statistical evaluation derived from the literature.