Identification for a Model Frame Structure Using Vibration Measurements from FBG Displacement Sensors

In this study, a neural networks-based structural parameters identification methodology without any mode shapes and frequency extraction is proposed and validated using vibration-induced displacement measurement from a Fiber Bragg Grating (FBG) sensor. The theoretical base of the proposed method is explained based on the discrete solution of structural dynamic response. To facilitate the parametric identification process, a reference structure is assumed firstly and its dynamic response is determined by numerical simulation. A displacement-based neural network (DNN) is constructed and trained using the simulated response to forecast the vibration displacement of the reference structure. With the support of the trained DNN and associated structures, a parametric evaluation neural network (PENN) is constructed to describe the mapping between an evaluation index and structural parameters. A two-story model frame structure on a shaking table is employed as an illustrative structure to validate the proposed approach using vibration displacement response measurement from a FBG sensor, which measure the relative displacement response of the middle point of the column of the first floor. Parametric identification results by the proposed methodology are compared with them from traditional frequencies extraction and show that the proposed methodology can identify the inter-storey stiffness of the whole frame structure within acceptable accuracy even only the displacement measurement at one point of the model structure is known. The proposed algorithm may be an applicable method in practice for and damage detection and structural model updating.