Model-Uncertainty Impact and Damage-Detection Accuracy in Plate Girder

Abstract

The impact of model uncertainty on the damage-prediction accuracy of a nondestructive damage-detection technique when applied to a model plate girder is assessed. First, a damage-detection algorithm to locate and size damage from few mode shapes of structures is outlined. Next, locations and magnitudes of damage in the model plate girder are predicted by using the damage-detection algorithm. Finally, damage-prediction accuracy for the model plate girder is estimated as a function of model uncertainties that are most likely or exist in the damage-detection procedure of the model plate girder. By applying the approach to the model plate girder, it is observed that damage can be confidently located with a relatively small localization error and a relatively small false-negative (i.e., missing detection of true damage locations) error but a relatively large false-positive (i.e., prediction of locations that are not damaged) error. It is also observed that severity of damage can be estimated with a relatively large error.