Filtern
Erscheinungsjahr
- 2023 (1) (entfernen)
Dokumenttyp
- Vortrag (1) (entfernen)
Sprache
- Englisch (1)
Referierte Publikation
- nein (1)
Schlagworte
- Wind turbine blade (1) (entfernen)
Organisationseinheit der BAM
Eingeladener Vortrag
- nein (1)
The European Green Deal and the global fight against climate change call for more and larger wind turbines in Europe and around the world. To meet the increasing demand for maintenance and inspection, partly autonomous methods of remote inspection are increasingly being developed in addition to industrial climbers performing the inspection.
Rotor blades are exposed to extreme weather conditions throughout their lifetime of 20 years, and the leading edge erodes over time. These erosion damages change the aerodynamic features of blades and can cause structural damages. The estimated annual energy production (AEP) losses caused by erosion damages are between 0.5% and 2% per year. The classification of the severity of a rain erosion damage and the quantification of the resulting AEP losses for cost efficient repair and maintenance efforts and improved power production of wind turbines are subject of scientific research.
For the inspection of wind turbine rotor blades, passive thermography is an option that has been used to detect both internal damage [3, 4] as well as erosion on the surface [5, 6]. The advantage is that, given suitable boundary conditions, not only the rain erosion damage itself but also temperature differences caused by the resulting turbulences can be observed on the surface of the blade. Turbulences reduce the efficiency of the rotor blades and result in AEP losses. Optimised thermography inspections can contribute to detect and to evaluate rain erosion damages. The thermal inspection lasts 10 minutes per turbine and is performed while the turbine is in full operation, avoiding downtime and lost opportunities for the turbine owner which are usually caused by conventional blade inspections. The timely inspection procedure is complemented by an automatic data evaluation and results in a considerable number of inspected wind turbines in a certain period of time. A fully convolutional network (FCN) is implemented for the automated evaluation of thermal images.
In the presented study, more than 1000 thermographic images of blades were annotated, augmented and used to train and test the FCN. The aim is the precise marking of thermal signatures caused by erosion damage at the leading edge. The area size of the detected temperature difference caused by turbulences was used to identify and categorise damages. Certain strategies were adopted to group small individual indications as one large damage, in order to develop simplification rules based on realistic thermal imaging resolution.