Bridge Structural Health Monitoring using Statistical Damage Detection and Advanced Load Rating Methods

Two structural health monitoring methods are applied to the Powder Mill Bridge (PMB), a fully instrumented bridge located in Barre, MA. The first topic describes a new nonparametric statistical method for evaluating bridge damage using girder distribution factors (GDFs) under operational strain monitoring. A calibrated finite element model is used to simulate damaged bridge GDF samples for four damage scenarios. The Wilcoxon rank-sum test is employed to compare a healthy baseline sample with a potentially damaged bridge sample. A damage index is established based on the rank-sum test statistic and is designed to alert bridge owners of potential damage. To further evaluate potential bridge damage, a 3D statistical bridge signature is developed for damage localization and assessment. Prediction intervals corresponding to a baseline signature are developed using the nonparametric bootstrap method, creating an envelope of possible baseline bridge signatures. This method enables a rigorous statistical comparison of a baseline bridge signature envelope with a potentially damaged bridge signature. Damage was successfully identified for all four damage cases. The overall damage detection method is designed to alert bridge owners when damage is detected and to provide a probabilistic tool for damage assessment and localization.;The second topic presents a comparison of three methods used to load rate the PMB. The first method employs the conventional load rating technique commonly used by bridge engineers. The second method uses strain measurements from a diagnostic load test to account for in-situ bridge behavior. The third method uses a finite element model of the bridge, which accounts for 3D structural system behavior. Advantages and disadvantages of these three methods are discussed. The advanced load rating methods are shown to produce higher ratings in comparison to the conventional approach.