Structural damage diagnosis under varying environmental conditions-part II: local PCA for non-linear cases

It is well known that changes in vibration features of structures due to damage may be masked by the effects of environmental variations. This influence has to be eliminated in the structural health-monitoring process, especially when a long-term in situ monitoring is expected. In the companion paper [1] a linear method based on principal component analysis (PCA) has been proposed and has shown encouraging results for linear or even weakly non-linear cases. The present paper concerns a further extension of the proposed method to handle non-linear cases, which may be encountered in some complex structures. The method involves a two-step procedure, namely a clustering of the data space into several regions and then the application of PCA in each local region. The application of local PCA allows performing a piecewise linearisation of the non-linear problem. A close look at the choice of the distortion function used in data clustering leads to two new clustering strategies. Whereas the first strategy is specifically suitable for the application treated in this paper, the second one is more general. The local PCA-based damage detection method is applied for the structural health monitoring of a real bridge using vibration data measured in situ over a one-year period.