GNSS-free structural displacement estimation through heterogeneous sensor fusion

Abstract

Displacement plays a vital role in the monitoring and control of structures. Though the Global Navigation Satellite System (GNSS) has been commonly adopted to measure structural displacement, it has limited accuracy and sampling rate, and cannot work under GNSS-denied environments. This dissertation aims at GNSS-free structural displacement estimation through heterogeneous sensor fusion, and three types of techniques are proposed for different structures: (1) For structures moving out-of-plane relative to a nearby fixed target, a displacement estimation technique is proposed by fusing FMCW millimeter-wave radar and accelerometer. The primary contributions lie in the automated selection of the best target and estimation of a conversion factor for the radar, and improvement of radar-based displacement estimation by developing an acceleration-aided adaptive phase unwrapping algorithm. (2) For structures moving in-plane relative to a nearby fixed target, two displacement estimation techniques are proposed by fusing an accelerometer and vision camera. The contributions of the first technique lie in the automated scale factor estimation, development of an adaptive multi-rate Kalman filter to fuse asynchronous acceleration and vision measurements, and improvement of feature matching by developing a region of interest updating algorithm and two automatic mismatch rejection algorithms. The contributions of Technique II lie in the development of a hybrid computer vision algorithm and automated initial calibration for scale factor estimation and active pixel selection. (3) Without requiring a no nearby fixed target, two displacement estimation techniques that fuse accelerometer and strain gauges are proposed for beam-like bridge structures and submerged floating tunnels, respectively. The main contributions of the two techniques lie in the derivation of strain-displacement transformation without requiring true mode shapes of a structure and the automated estimation of unknown parameters associated with the proposed strain–displacement transformation. The displacement estimation performance of the proposed techniques was numerically and experimentally validated. In addition, the effectiveness of all algorithms included in these techniques was investigated, and the application range and limitations of each technique were discussed.