ABSTRACT Structural Health Monitoring (SHM) systems based on Guided Ultrasonic Waves are pivotal for the transition from damage tolerance to condition‐based design in aerospace Composite Fiber Reinforced Polymers (CFRP) components. However, environmental and operational variability poses a critical challenge for baseline‐free methods. This work investigates the statistical variability of Lamb waves to validate a population‐based diagnostic approach. The study involved an experimental campaign on 35 CFRP plates subjected to Low‐Velocity Impacts (15 J) and monitored via a piezoelectric sensor network. Parallelly, a Finite Element (FE) model was developed to augment the database. The results highlight that the high intrinsic dispersion within the pristine population overlaps significantly with the damaged population distribution, preventing simple threshold‐based detection. This finding emphasizes the limitations of purely experimental statistical baselines and suggests the implementation of hybrid numerical‐experimental databases to train robust Machine Learning algorithms.
Pianese et al. (2026) studied this question.