Statistical bridge damage detection using girder distribution factors

A hypothesis testing framework is introduced for bridge damage detection, which enables a rigorous, decision-oriented approach for detection of bridge damage when it exists. A bridge damage detection hypothesis test is developed using girder distribution factors (GDF) under operational, output-only strain monitoring. GDFs are calculated from measured strain data collected during traffic events at the Powder Mill Bridge in Barre, Massachusetts. A sample of GDFs is drawn to establish a baseline over the course of one week, representing the probabilistic behavior of a healthy bridge under normal operating conditions. A new sample can be compared with the baseline at the end of each day, providing a timely and effective operational damage detection method. A calibrated finite element model is used to simulate damaged bridge GDF samples under four damage scenarios. The damaged bridge GDF samples are compared with the healthy baseline sample using the rank-sum test, and the results are employed to develop a damage index capable of alerting bridge owners of potential damage. A simple bootstrap resampling scheme is used to evaluate the probability of issuing a false alarm (Type I error), as well as the likelihood of not issuing an alert when the bridge is damaged (Type II error). A three-dimensional statistical bridge signature is developed to aid damage localization and assessment. Nonparametric prediction intervals corresponding to a baseline signature are generated using the bootstrap method, creating an envelope of possible baseline bridge signatures. When a bridge signature falls outside the baseline bridge signature envelope, damage is detected. Damage was successfully identified for all four artificial damage cases considered. The overall damage detection method is designed to alert bridge owners when damage is detected and to provide a probabilistic tool to aid damage assessment and localization while controlling for both Type I and Type II errors. (C) 2015 Elsevier Ltd. All rights reserved.