Damage Identification for a Reduced Scale Spatial Steel Frame

This paper reports on an experiment carried out in the laboratory of Civil and Environmental Engineering of the University of Florence, where a reduced scale (approximately 2.5 m width × 2.5 m depth × 4.8 m height), three storey steel spatial frame was instrumented and progressively damaged by cutting one of its pillars just above the foundation; the deepness of the cut could be measured and was taken as the entity of the damage. A series of 12 accelerometers was installed on the frame and 3 more accelerometers were installed on the ground; acceleration measurements induced by ambient vibrations were recorded as the frame was progressively damaged. Acceleration signals were subsequently analyzed in both time and frequency domain (SSI, EFDD, neural networks): a frequency shift and a change in the corresponding modal shapes were observed as the damage increased. The damaged truss was also modelled by means of a FEM code and the frequency shift and the change in the modal shapes were reproduced. A genetic algorithm approach is currently being tested to try and indentify the location and entity of the damage, and is reported on