Classification Algorithms for Damage Identification in Framed Structures by Means of Natural Frequencies

The identification and localization of structural damage plays an important role in the health monitoring of civil structures. However, when the number of potentially damaged structures is high it can be difficult to distinguish which ones have actually suffered damage and where further investigations need to be directed. It can be therefore useful to have procedures and algorithms that allow indicating automatically, as a preliminary step of deeper investigations, the possible occurrence of damage and its approximate position. In this context, the paper explores the variations of the natural frequencies for plane and 3D framed structures due to localized damage within macro-regions into which the structure can be divided. Based on numerically generated data, it is shown how the position of the damage within different macro-regions produces different and recognizable scenarios with respect to the variations of the natural frequencies. This allows to make use of classic data classification procedures, through which it is possible to automatically obtain indications on the macro-region of the structure subject to damage, thus limiting the extent of the investigations to be carried out on site.