UC Berkeley

The vibration-based structural health monitoring paradigm is investigated from the perspective of the mathematically defined dynamical system. Practical implementation is described as conducted on BRACE2 (Bridge Rapid Assessment Center for Extreme Events), a dynamic web application platform commissioned by the California Department of Transportation that monitors bridges and can be seamlessly extended to buildings. The development of the BRACE2 platform has served as the primary driver of the presented research.

System identification furnishes vibration-based structural health monitoring with a quantitative problem statement. Vocabulary, notation, and convention employed to describe a physical system's state-space representation are developed from control theory and Newtonian mechanics. This scientific language is then used to describe the investigated structural systems. From a state-space representation, two classes of metrics are derived: (I) modal properties and (II) response reconstruction. Subsequently, algorithms are presented for obtaining a system realization from measured vibrations. Following the theoretical discussion is the demonstration of a software library implementing the proposed methods.

An inventory of 22 instrumented bridges and eight instrumented buildings is collected for structural health monitoring within the BRACE2 framework. The modal properties obtained from system identification methods are investigated for use as health metrics. It is shown that the computed metrics can be used to distinguish between different structural systems in the inventory, and that for structures with available large acceleration records, there is a clear relationship between anomalous, potentially damaging events and a change in the computed metrics. Then, the reconstruction of response histories from system realizations is investigated for real-world structures. The study shows that reasonable response reconstructions for unseen events can be obtained from the system realizations. Finally, a detailed example is presented for one bridge with a large dataset of recorded earthquake events and whose geometric, material, and construction properties are studied in depth and modeled. The model of this case-study bridge constitutes a full digital twin in the BRACE2 framework.