A model-based approach for statistical assessment of detection and localization performance of guided wave–based imaging techniques

This article aims at providing a framework for assessing the detection and localization performance of guided wave–based structural health monitoring imaging systems. The assessment exploits a damage identification metric providing a diagnostic of the structure from an image of the scatterers generated by the system, allowing detection, localization, and size estimation of the damage. Statistical probability of detection and probability of localization curves are produced based on values of the damage identification metric for several damage sizes and positions. Instead of relying on arduous measurements on a significant number of structures instrumented in the same way, a model-based approach is considered in this article for estimating probability of detection and probability of localization curves numerically. This approach is first illustrated in a simplistic model, which allows characterizing the robustness of the structural health monitoring system for various levels of noise in test signals. An experimental test case using a more realistic case with an artificial damage is then considered for validating the approach. A good agreement between experimental and numerical values of the damage identification metric and derived probability of detection and probability of localization curves is observed.