Predicting Premature Failures in Small Wind Turbines With Recurrence Plots

Abstract

Abstract This paper presents the application of the recurrence plot as an alternative for preprocessing the raw data. The recurrence plots can extract the nonlinear and transient response and are sensitive to slight variations in the signal frequency, amplitude, and waveform. Thus, it is an alternative technique for improving the sensitivity; consequently, the prognostic algorithms can predict with better resolution. The data were obtained from an experimental 12 m wind turbine. The transmission was instrumented with three accelerometers and three gyroscopes; the generator's current and voltage were monitored. The difficulty in producing the phase plane using acceleration data is its integration to obtain the kinetic and potential signal energies. This limitation is overcome by integrating the data using the empirical mode decomposition and the shift principle. The results show good sensitivity for predicting variations in the operating conditions and are the basis for other prognostic analyses.