Damage identification on a numerical two-storey framed structure using ambient vibration response analysis and artificial neural networks

This paper presents a damage identification method based on ambient floor vibration measurements in multi-storey buildings. The proposed method uses ambient response vibration data to fannulate a damage index based on Frequency Response Functions (FRFs), which is used as input parameter to artificial neural networks (ANNs), to identify locations and severities of damage in a two-storey framed structure. By adopting principal component analysis (PCA) techniques, the size of the derived damage index is reduced in order to obtain suitable patterns for ANN training. A hierarchy of neural network ensembles is designed to take advantage of individual characteristics of measurements from different floor locations. The proposed method is tested on finite element models of a complex two-storey framed structure inflicted with notch-type damage of different locations and severities (in total six damage cases). The results of the study show that the proposed algorithm is capable of accurately and reliably identifying damage in complex multi-storey structures based on response-only ambient floor vibration measurements