Nonparametric damage detection in a multistory building through the use of neural networks

摘要

A practical structural damage detection methodology is pesented for the health monitoring of structure-unknown systems. The authors have developed a neural network-based approach for the detection of changes in the characteristics of structure-unknown system. The application of the methodology to a multi story building relies upon the availability of measurements at all stories. In this paper, the methodology is generalized by considering more practical situations, where measurements at only a limited number of stories are available. The new approach utilizes the modal information of the underlying structure to estimate response at unmeasured atories. The performance and limitations of the proposed detection method was explored through numerical simulations with a multi degree-of-freedom model.