

In-Operation Identification of a Wind Turbine Structure via Non-Stationary Parametric Models
Abstract
The in-operation identification of a wind turbine structure via non-stationary parametric Time-dependent AutoRegressive (TAR) models is presented. The identification is based on a single vibration acceleration response signal obtained at the tower of a NegNicon NM52/900 wind turbine under normal operating conditions. Following preliminary non-parametric analysis, Smoothness Priors TAR (SP-TAR) and Functional Series TAR (FS-TAR) models are identified and analyzed. The results of the study confirm the cyclo-stationary and/or non-stationary nature of the dynamics, and indicate three main types of dynamical behavior: That due to the wind characteristics, those due to the low and high speed shafts of the transmission system, and strongly cyclo-stationary dynamics associated with blade rotation.