Methods for Identifying and Modeling Strain Anomalies in Prestressed Concrete Structures

Abstract

Aging infrastructure poses significant concerns in terms of both safety and sustainability. Utilizing techniques to accurately monitor structures, or potentially prolong their lifespan, can be extremely beneficial. Structural health monitoring can provide strain and temperature data from active structures in our urban environment, which can provide greater insight than traditional physical inspection methods.

Streicker Bridge on Princeton University's campus is a 15-year-old structure with embedded sensors that have been collecting data since its creation. This strain and temperature data has allowed members of the SHM lab group to create a data-driven predictive model of strain. Comparing the predictive model to the measured data, many anomalies in strain data have been detected throughout both the main span and the southeast leg. The aim of this thesis is to provide a method of predicting the evolution of long-term anomalous strain behavior in prestressed concrete structures. In this thesis, we propose two possible diagnoses of different strain anomalies in Streicker Bridge. The first involves anomalies in the southeast leg, which were thought to be due to the settlement of one of the bridge's arch supports. After computational modeling using a finite element model, it was found to be unlikely that the anomaly was due to settlement. The second anomaly diagnosis involves a crack formation followed by long-term creep recovery in the main span of the bridge, which has not been thoroughly investigated in previous literature. Following this diagnosis, four new data-driven predictive models were created to forecast the future of the anomalous strain values at two locations in the main span of the bridge. These models resulted in the successful prediction of strain approximately 300 days in the future at locations undergoing anomalous behavior within an accuracy of 16 \(\mu\)m/m .

Future work concerning the prediction of anomalous strain behavior includes expanding the models to predict values at other sensor locations, using more advanced modeling methods to mitigate error, and predicting for longer periods.