Application of pattern recognition for damage classification in scarf repairs

Bonded repairs in composite laminates have been proposed as an alternative approach to mechanically fastened repairs since they show reduced local stresses and improved structural performance for aerospace applications. More specifically, scarf repairs offer great advantages compared to external patch repairs since they provide higher stiffness by matching ply to ply the original structure and by reducing stress discontinuities in the repaired region. The current work presents a damage detection study of a scarf repair under tensile loading with guided ultrasonic waves. The study focuses on the selection of appropriate signal features and on their subsequent investigation through outlier analysis. The limits of the applied outlier analysis are interpreted through Principal Component Analysis and optimised through the concept of Principal Curves. The obtained results are compared with images recorded by 3-Dimensional Digital Image Correlation and conclusions about the efficiency of the damage detection strategy are deduced.