System Identification of Civil Engineering Structures through Wireless Structural Monitoring and Subspace System Identification Methods.

Abstract

Recent dramatic catastrophic failures of civil engineering infrastructure systems such as the I-35 Bridge collapse (Minneapolis, MN, 2007) and PG&E gas pipeline explosion (San Bruno, CA, 2010) has called attention to the need to better manage these complex systems to ensure their safe usage by society. Structural health monitoring has emerged over the past decade as an active interdisciplinary research field dealing with the development and implementation of sensing technologies and data processing methods aimed to perform condition assessment and damage detection of engineered structural systems. While many technological advance have been made over that period, some technological hurdles still remain. For example, high costs and laborious installations of monitoring systems hinder their wide-spread adoption. Furthermore, there exists a lack of generalized data processing algorithms. The thesis addresses these fundamental bottlenecks. At the core of the dissertation is the advancement of wireless sensors for cost-effective structural monitoring. A wireless sensor node designed explicitly for monitoring civil infrastructures is introduced and deployed on operational bridge structures. Numerous intrinsic advantages of wireless sensors are illustrated including decentralized computing architecture, reconfigurable installation, and use for mobile sensing. To process the large tracts of structural response data created by wireless monitoring systems, subspace system identification techniques recently developed in the field of control theory are explored. While subspace system identification enjoys exclusive superiority of black-box estimation, a physical interpretation of the estimated black-box model remains unresolved. The thesis proposes a theoretical methodology for physical interpretation of black-box models while linking physical system parameters with grey-box models. Having established subspace identification as a powerful data processing tool, embedment of these methods in the wireless sensors is explored for autonomous in-network execution. A novel approach to unified dynamic load monitoring of a moving heavy truck and bridge response is proposed for experimental observation of vehicle-bridge interaction. By leveraging the mobility of wireless sensors installed in the truck, a permanently deployed wireless bridge monitoring system collects time-synchronized data on the bridge loading and response. The thesis also discusses a data processing algorithm for the identification of bridge systems excited by a moving vehicle to analyze vehicle-bridge interaction.