Abstract: Aiming at the problem that the trailing edge crack of the ultra-long wind turbine blade is difficult to identify in real time during its service, this paper takes a full-size 51.5 m glass fiber/epoxy blade as the object, and arranges 20 resonant acoustic emission sensors along the potential crack propagation path in the ground test of lead-breaking calibration-graded crack expansion-variable amplitude fatigue, and implements 5 MHz parallel acquisition and dynamic threshold triggering. The results show that the acoustic emission amplitude decays rapidly within0.2-0.5 m propagating distance, and local grinding can improve the signal strength; the change of the trigger threshold will cause obvious statistical drift of the acoustic emission characteristics, and the actual monitoring should adopt the background noise +3 dB baseline. The accumulated energy decays exponentially along the normal direction of the crack, forming a “peak-valley” spatial pattern, revealing the coupling mechanism of crack tip stress enhancement and wave attenuation; the crack propagates steadily within 40-60 h, and the high-frequency hotspot is confined to the crack tip. After entering the instability stage at 60-70 h, the high-frequency area expands rapidly along the span and chord directions. The three parameters of energy, center of mass frequency and amplitude are reduced in dimension by t-SNE to form a continuous spectrum band, which can clearly separate the three stages of crack initiation, steady state and instability. The statistical characteristics of acoustic emission show certain regularities in the three stages, providing a physically explainable criterion for the online early warning of cracks on the trailing edge of the blade.

Key words: acoustic emission (AE), wind turbine blade, trailing edge crack, attenuation test, threshold effect, damage evolution