Model-Assisted Probability of Detection Using QL-Scan GUPI for Wall-Thinning Defects

Abstract This study presents a simulation-driven framework for quantitatively evaluating the probability of detection (PoD) of wall-thinning defects using a laser-based Q-switched Laser scan Guided Ultrasonic Propagation Imaging (QL-scan GUPI) system. QL-scan GUPI has demonstrated strong potential for guided wave visualization, but no model-assisted PoD (MAPoD) framework currently exists to assess its detection reliability. To address this gap, a full-model MAPoD methodology was developed following MIL-HDBK-1823A standards. High-fidelity finite element simulations using ANSYS Explicit Dynamics were conducted under varying defect sizes, depths, and sensor distances. Corresponding experiments were performed on aluminum plates with artificial square-shaped defects. Guided wave responses were analyzed using the Ultrasonic Energy Mapping (UEM) algorithm, and logistic regression models were employed for both hit/miss and signal response analysis. Results demonstrated strong agreement between experimental and simulation-based PoD curves, with consistent trends across varying defect conditions. The multivariate MAPoD surface revealed defect depth as the most significant factor, while sensor distance showed minimal influence. Simulation outputs provided conservative estimates compared to human inspectors, especially for small or shallow defects. This study establishes the first validated MAPoD framework for the QL-scan GUPI platform. The proposed methodology enables scalable, inspector-independent evaluation of guided wave imaging systems and provides a foundation for their integration into reliability-centered maintenance strategies in safety–critical industries.