Composite rotor blades of wind turbines are subjected to high dynamic load. This load can cause damages which can accumulate over time to critical structural damage. A system detecting defects early and reliably helps to react fast and to avoid critical damage. Such a method will enable the wind turbine operator to increase operational safety and minimize the economical burdens caused by downtime, maintenance as well as repairs and replacement. One promising damage detection approach is acoustic emission event detection. In this regard acoustic emission events are stress waves emitted by a damage process. While other acoustic emission approaches use ultrasonic surface accelerations as input signals, we propose to use the airborne sound in audible frequencies and a sophisticated signal processing which can handle environmental noise. We were able to perform a full scale rotor blade fatigue test until blade failure where one 44cm long continuous crack occurred. The test was continued and the crack was further increased. The evaluation of the airborne sound recordings shows that the continuous crack as well as parts of the crack propagation emitted cracking sounds. We modified our detection algorithm based on audio features in the time-frequencypower space by a power impulse feature. With this modification the algorithm detects the continuous crack as well as parts of the crack propagation without false alarms.
doi: 10.12783/SHM2015/340