Recursive identification of frequency-amplitude model for damage detection in structures with non-linear behaviour

Most damage identification methods used in the field of structural health monitoring are based on comparing the modal parameters estimated at different time intervals. Since natural frequencies can be evaluated even by using limited instrumentation and does not require any signal synchronization procedure, frequency-based identification methods are of great interest for applications in real-time monitoring by means of low-cost sensing systems. In the last decades, several damage-sensitive features have been proposed in order to conduct damage identification minimizing the occurrence of errors under varying environmental conditions. However, the amplitude of input excitation can also lead to short-term frequency variations, due to a non-linear behaviour of civil structures. This paper proposes a procedure aimed at damage detection in non-linear structures based on frequencies and amplitudes of modal responses, estimated assuming a locally-linear structural behaviour. It is shown that, by analysing a set of instantaneous parameters, it is possible limit the problems related to an incorrect identification of the structural state of health when the effects of non-linearities on modal parameters prevail over those of an actual state of damage. The method proposed is applied to a set of data recorded on a reinforced concrete structure with strongly non-linear behaviour.