Real time structural modal identification using recursive canonical correlation analysis and application towards online structural damage detection

A novel reference free approach for identifying structural modal parameters using recursive canonical correlation analysis (RCCA) is proposed in this paper. Using first order eigen perturbation (FOEP) approach, the method provides eigenspace updates at each instant of time from which the linear normal modes (LNMs) of a vibrating system can be determined. Modal assurance criterion (MAC), examined in a recursive framework, is utilized to compare the theoretical modes with the estimated LNMs in order to verify the accuracy of the obtained modes. The eigenspace updates provide basis vectors that can be subsequently utilized to determine potential damage induced to the system. The set of transformed responses obtained at each time stamp is fit using time varying auto regressive (TVAR) models in order to aid as damage sensitive features (DSFs) for identifying temporal and spatial patterns of damage. The proposed RCCA algorithm based on the FOEP approach can accommodate multi-directional dependencies of a system. Numerical simulations carried out on linear systems demonstrate the efficacy of the proposed method in recursively separating the structural modes. Comparative case studies with the batch CCA based method illustrate the superlative performance of the RCCA scheme towards recursively identifying the modes of a system, an aspect that is significantly missing in the context of online structural health monitoring (SHM) literature. The extension of the method in detecting spatio-temporal patterns of damage for both linear and nonlinear systems demonstrates the effectiveness of the proposed scheme as a real time damage detection technique. Application of the method towards experimental setups devised under controlled laboratory environment in identifying structural modes and determining the change of state of the structure greatly demonstrate the robustness of the method as an ideal candidate for baseline free, real time SHM. (C) 2019 Elsevier Ltd. All rights reserved.