Diagnostics of Wind Turbine by Detectivity

Hjorth’s parameters were initially proposed in literature, to characterize the morphology of an ECG signal in the field of medicine. They consist of three scalar values that summarize in a compact form time and frequency properties of an input signal. They are the Activity defined as the variance of the time-domain signal, the Mobility related to the variance in the frequency content and the Complexity which is related to the shape of the signal itself. In a previous paper, the authors introduced a new parameter called Detectivity that fuses together the Hjorth ones. Detectivity can be seen as the total gain of the actual Hjorth parameters with respect to a reference condition (i.e. healthy machine). In this paper, Detectivity is used to analyse a wind turbine dataset, publicly available from the LuleåUniversity website, in order to monitor the occurrence of damage in one of the six tested turbines. The challenge is to make the Detectivity robust in the face of a significant change in the instantaneous rotation speed of the turbine. A down-sampling pre-processing allows to filter data corresponding to a limited speed range. The cumulant of the Detectivity value evolves in a different trend for the faulted turbine.