Acoustic emission-based damage mechanism identification and warning for 3D-printed patch-reinforced scarf joints in segmented composite wind turbine blades

Abstract

This study addresses the challenge of assessing damage in segmented wind turbine blade adhesive joints, where external reinforcement improves performance but introduces new critical failure interfaces. By testing a laboratory-scale segmented composite blade with a scarf adhesive joint in two configurations—an unreinforced baseline and one reinforced with externally bonded, 3D-printed CF/PEEK patches—under bending-dominated monotonic loading to fracture with continuous acoustic emission (AE) monitoring, we employed unsupervised learning to separate AE event populations and identify both reinforcement-related interfacial activity and inherent joint degradation mechanisms. Post-failure analysis confirmed that reinforcement increased the ultimate load and deformation tolerance, with the dominant failure mode being debonding at the patch-skin interface rather than patch rupture. To enable proactive warning, a composite AE state quantity (cr) combining the b-value and normalised cumulative energy was constructed for severity grading, and an intelligent CNN-LSTM model was trained to deliver a four-level damage warning; validation showed the model achieved high overall performance (macro-averaged F1-score: 0.96) with exceptional recall (0.99) for the severe damage stage. The trained model was further tested on the unreinforced blade dataset as an external test set, showing reasonable transferability with strong sensitivity to severe damage. These results demonstrate that AE-based mechanism identification can be linked to reliable warning of interface-dominant damage in reinforced segmented bonded joints.