Verification and validation of dielectric mapping technique for non‐destructive evaluation of polymer matrix composites

Abstract The rapid increase in use of polymer matrix composites in different industries underscores the need for reliable non‐destructive evaluation techniques to characterize small‐scale damage and prevent structural failure. A novel dielectric technique exploits moisture‐polymer interactions to identify and track damage, leveraging differences in dielectric properties between free and bound water. This technique has demonstrated the ability to detect low levels of damage, but the localization accuracy has not yet been evaluated. This work utilizes unsupervised machine learning to assess the technique's ability to identify the damage boundary following a low‐velocity impact event. Bismaleimide/quartz and E‐glass/epoxy laminates were impacted via drop tower to induce varying levels of damage, and subsequently inspected via dielectric technique at several moisture levels by weight. Resulting data was processed via k‐means clustering and the identified damage boundary was compared to a boundary obtained from backlit images and scanning electron microscopy. Accuracy was quantified using developed metrics for damage centroid and boundary identification. The technique averaged 93.9% accuracy in determining the damage center and 77.5% accuracy in identifying the damage boundary. Results indicated the technique's effectiveness across varying moisture levels, particularly in damage centroid identification. Localization accuracy was shown to be insensitive to moisture content, improving the technique's practical capabilities. Further analysis revealed potential for delineation of delaminations. Highlights Low‐velocity impact of two material architectures. Damage boundary determined and validated via scanning electron microscopy. Detected damage site via dielectric technique compared to damage boundary. High technique accuracy revealed; >90% centroid localization accuracy. Potential for delamination delineation observed.