Acceleration-based deflection angle and deflection estimation of beams in data-driven models

This study aims at proposing a method to estimate the deflection angle (rotation) and deflection of bridges using acceleration measurements without integrating acceleration in the time domain. It applies multiple measurements, filtering, and fitting techniques for a constrained local optimum rather than integrating acceleration in the time domain. First, the acceleration of the bridge is filtered to obtain the time history of the deflection angle at each measurement point. Then, the Karush–Kuhn–Tucker (KKT) algorithm as a constrained least squares (CLS) method is used to estimate the shape curve of the deflection angle and deflection of the bridge. The validity of the proposed acceleration-based deflection angle and deflection identification method is verified through an in-house moving vehicle experiment on a model bridge. The method develops a data-driven model based on a small amount of observational data, enabling accurate predictions of the angular rotation and deflection of beam structures.