Obtaining reliable diagnosis results based on modal parameter changes is a challenge due to their low sensitivity to local damages and uncertainties related to their estimation, especially under ambient excitation. In non-destructive testing, the reliability is quantified through probability of detection (POD) curves, which are often limited to damage detection and cannot be applied to structural health monitoring applications where no data from the damaged state is available. To fill this gap, a method is developed in this paper that allows one to create probability of localization curves (POL curves) based on measurements from undamaged structures. The approach is based on statistical damage localization tests and requires a finite element model. For proof of concept, the method is applied to a simple numerical structure, demonstrating that it is a powerful tool to analyze the performance of SHM systems before damage occurs. The findings demonstrate that the POL increases with an increasing number of observed modes of vibration, an increasing measurement duration, an appropriate sensor layout, and low measurement noise levels.
