Early damage detection in wind-excited suspension bridges

Continuous structural health assessment based on data recorded by monitoring sensors permanently installed in structures is an important challenge of the modern civil engineering. Among the various methodological approaches for structural health monitoring, methods of multivariate statistical analysis are currently receiving a deep attention because of their ability to early detect, in a fully automatic way, possible anomalies in the structural behaviour using either static or dynamic structural output response data. With this paper, the authors propose the use of a mathematical technique based on principal component analysis and novelty detection for early damage detection in wind-excited long-span bridges. The effectiveness of the proposed technique is investigated by application to pseudo-experimental vibration response data, obtained by means of a continuum model of damaged suspension bridge under wind loading and accounting for aeroelastic effects and measurement noise.