Response surface-based structural damage identification using dynamic responses

Abstract

Response surface methodology (RSM) is efficient in updating a model as compared to a finite element model (FEM) based on updating techniques. This feature of RSM can be used to replace the time-consuming FEM-based updating step with a response surface-based surrogate model. In addition, RSM also solves the problem of slow convergence and time consumption. The present study deals with the applicability of RSM for damage identification in a six-storey shear building. At first, the theoretical overview of RSM has been provided, followed by a detailed description of the proposed algorithm for damage identification. For this purpose, different types of design of experiments (DoEs) have been used for generating the response surface (RS) model. A simulation study has been carried out on a six-storey numerical building model, along with an experimental investigation on a three dimensional six-storey miniature model of a shear building in a laboratory environment. For executing the methodology, dynamic responses of the structures are analyzed to get the translational modal parameters (frequencies and mode shapes) using frequency domain decomposition (FDD). The RS equations have been generated using these modal parameters through various DoEs (namely, central composite design, Box-Behnken design, and D-optimal), and they have been found to be equally efficient in identifying the damage. A comparative study over RSM based damage detection using different modal parameters have been presented. It has been observed that the RS model generated using the first three translational modal frequency and the corresponding first two mode shapes are quite efficient in detecting as well as localizing the damage in the shear building with proper quantification.