Spectral Method and Machine Learning approach to Wind Turbine damage detection

Resumo

Wind energy is one of the cleanest energy source currently used in the world, an energy source that interferes least with the environment. It is important to locate Wind Farms (WF) in such a way as not to limit the living space and not to reduce the comfort of people in the area. Due to the intensity of the wind and minimal human impact, offshore farms seem to be the required solution. Another aspect important from the point of operational reliability, is ensuring continuous working conditions due to the design and material solutions. Wind Turbine (WT) structures are exposed to the dynamic action of wind and waves from the sea, as well as to the corrosive environment, causing accelerated damage to WT. The action ensuring safe use of structural elements of WT is monitoring the technical condition of the structure based on the analysis of frequency response functions (FRF). At the design, as well the operational stage, it is important to predict the failure of the element. The paper presents the simulation results of the monitoring and prediction of damage of IEA 15-Megawatt offshore wind turbine using the analysis of changes in resonant frequencies in the FRF and the Machine Learning technique.