Detection of damage using temporal variation in natural frequencies and principal component analysis

Abstract

Detection of structural damage is an important issue. Some techniques require knowledge of the undamaged state or in-depth knowledge of the construction materials used. The application of principal component analysis to the temporal variation in natural frequencies is a promising technique. With an undamaged structure a single principal component will suffice to explain the change in modal frequencies with weather variables, whereas more components are needed in the case of damaged structures. Like all other detection techniques, it has advantages and drawbacks. The main advantages are the ability to work without knowledge of the previous state of the structure, the relatively short monitoring time required (about 10 days), low cost, and low number of instrumentations. The main drawback is the need for more studies to correlate degree of damage with the number of principal components. This paper presents the results obtained for six structures: Mallorca Cathedral, four university buildings in Barcelona, and a reduced model structure in a laboratory. We also propose a semiautomatic algorithm to track modal frequencies over time and add a discussion of operational modal analysis in real buildings.