Damage quantification in moment resisting frame using phase space reconstructed from independent component sources

Abstract

In the present study, a damage detection algorithm is proposed based on the amalgamation of independent component analysis (ICA) and phase space topology (PST) reconstruction. An experimental investigation is conducted by inducing damage in steel moment resisting frame (SMRF) and exciting it with ambient white noise base excitation. The experimental study is conducted for a healthy state and multiple progressive damage states. The damage in SMRF is manifested progressively by loosening the bolts of the beam-to-column and column-to-column bolted connections. The algorithm consists of two segments: In the first segment, the modal responses are obtained from the acquired response with the help of ICA. The optimal embedding parameters for modal responses are evaluated with the help of average mutual information function and singular system analysis. The second segment deals with the reconstruction of PST with the help of modal responses and embedding parameters. Lastly, the damage sensitive features (DSFs) are evaluated to identify and quantify the induced damage. The DSFs evaluated in the present study are, namely, (a) change in phase space topology (CPST), (b) Mahalanobis distance between phase space topology (MDPST), and (c) Itakura distance (ID). The CPST and MDPST measure the dissimilarity between the PSTs of healthy and damage states, whereas ID measures the dissimilarity between the set of auto-regressive coefficients of healthy and damage states of SMRF. The DSFs are able to detect and quantify the commencement and evolution of the damage in the SMRF subjected to ambient base excitation.