Monitoring of Civil Engineering Structures Based on H∞ Estimation

Abstract

Subspace methods, a field of numerical mathematics, are applied to system identification of mechanical structures. The efficacy of these methods, applicable to both deterministic and stochastic model estimation, has been proven on numerous occasions in practical and research applications. The system identification process yields multidimensional state space systems, which serve as the mathematical foundation for the models. In the context of stochastic system identification for mechanical systems, particularly those exposed to ambient noise sources such as wind and traffic, output-only methods are employed. Since the system inputs are unknown, the model can’t be completely determined. By applying ?∞ optimization methods, models with a theoretical normalized input are obtained. These state space systems can be interpreted as a digital twin at a designated time point. In this contribution the method “Autonomous Model Order Selection” (AMOS) is linked with the structural health monitoring (SHM) method “State Projection Estimation Error” (SP2E), which leads to an automatic and continuous SHM approach. The output of the SHM method is a damage localization indicator, which is combinable to a digital twin. This approach is verified with laboratory measurements and a large-scale experiment on a real bridge.