Updated long-term Bayesian Monitoring Strategy for time-dependent deflections of I-35W Saint Anthony Falls Bridge

The goal of structural health monitoring is to assist in the maintenance and management of structural systems by providing a process to identify problematic behavior. Current long term structural health monitoring systems provide information on whether the structure is continuing to function as expected and identify potential anomalous behavior. In this case, long-term anomalies for bridges are defined as a change in behavior that lasts longer than a month, examples being time-dependent deterioration, unexpected post-tensioning loss, or translation of the superstructure on the bearings. To identify anomalous behavior, an accurate prediction of the structural response needs to be made to determine whether the measured data falls outside of expected bounds. For the I-35W Saint Anthony Falls Bridge in Minneapolis, Minnesota, monitoring efforts were focused on detecting anomalies in the deflections at the expansion joints measured by the linear potentiometers. While short-term monitoring approaches use a rigorous Bayesian framework to account for uncertainties, previous long-term structural health monitoring of the Bridge was not robust due to uncertainty in time-dependent displacement predictions over long durations. A Bayesian statistical framework was adopted to account for the uncertainty in the time-dependent predictions. This is particularly important as time-dependent behavior is dominated by temperature effects and noise. This work presents a method to compute a more reliable bounding interval for accurate long-term anomaly prediction through a Bayesian statistical framework. The resulting bounds and anomaly detections are compared with the previous approach on existing data and applied to artificially generated perturbations.