Structural health monitoring based on time series analysis

In this paper a monitoring technique based on the analysis of time history response of the structure is proposed. The method is based on the availability of: responses recorded in a number of locations of the undamaged structure and responses recorded in the same locations during or after a damaging event. For each configuration of the building an interpolation error of the responses in terms of absolute acceleration can be defined. The variation of the interpolation error between two different configurations of the structure is assumed as a parameter able to measure damage to the building. The proposed method has been applied using responses recorded on the Factor Building at UCLA for which a numerical model has been built on the base of an existing model and calibrated using results of an identification procedure. Several different damage scenarios have been simulated with the numerical model and responses have been used to check the ability of the proposed method to detect damage. Results show that for different damage patterns the damage detection function herein proposed is able to not only locate the storey where damage occurs but also to give an estimate of the amount of damage. The proposed method does not require any finite element modeling of the structure or knowledge of its mechanical characteristics hence it can be implemented in a "on line" damage detection warning system that, after a damaging event, can provide, nearly in real time, reliable information about the location of damage