Abstract:
This paper presents a vibration~based damage detection method that utilises frequency response functions (FRFs) to identify added mass on a two-storey framed structure. Added mass is used to simulate frequency changes due to structural damage. Artificial neural networks (ANNs) are employed to map changes in FRFs to locations of the added mass. In order to obtain suitable inputs for neural network training, principal component analysis (PCA) techniques are adopted to reduce the size of the FRF data and to filter noise. A hierarchy of neural network ensembles is used to take advantage of individual measurement characteristics from different sensors. The method is tested on laboratory and numerical models of a two-storey framed structure. From the two kinds of structures, FRF data are determined and compressed utilising PCA techniques. The PCAreduced FRFs are then used as input patterns for training and testing of ANN ensembles predicting different locations of added mass.
