[en] This paper presents a comparative study on several approaches of structural damage diagnosis based on vibration measurements. Stochastic subspace identification method is used to identify modal parameters and to generate a Kalman prediction model, which are taken as damage-sensitive features for structural damage detection. A statistical process control technique based on principal component analysis (PCA) is also presented. An improvement and enhancement of PCA are proposed. It is assumed that without damage, structural responses should remain approximately in a hyperplane defined by the principal directions of data. Damage localization is explored with these methods. As only the measured output signals are needed, the methods are convenient for an on-line monitoring. The efficiency and limitation of the proposed methods
are illustrated by a practical application.
