Estudo multivariado de sinais de vibração em uma turbina eólica em escala reduzida

Abstract

The main issues addressed in accurately modeling random vibration in wind turbines are due to the complexity of the structure, applied loads, and the operating environment. In this study, collected signals of structural vibrations of a reducedscale wind turbine were analyzed, firstly with a balanced set of three blades, and then with another set of three unbalanced ones for the , that is the TSR (Tip Speed Ratio). Aiming to understand the dynamics of change in vibration signals and to extend concepts and methods recently developed in statistical physics, extensive use of the DFA (Detrended Fluctuation Analysis); FFT (Fast Fourier Transform) and Hurst method (R/S); that in recent years has been cementing itself as a very soughtout technique for the determination of fractal scale properties and detection of longrange correlations in non-stationary noise time series; and the PCA (Principal Component Analysis) which is one of the simplest multi-variable methods. It was observed that the FFT method showed better results than the other methods, followed by DFA and, finally, the R / S. The λ = 6 blades showed the best results for rotation of 1100 RPM at the PCA, with cumulative variance of 82.39% and 82.34% in the FFT and DFA methods in the first two components. The λ = 7 blades showed better results for rotations of 600 RPM, 900 RPM and 1200 RPM at the PCA with variance 81.14%, 84.06% and 83.21%, respectively, in FFT method and variance 81.94% at 1200 RPM rotation in DFA method. It was found that in all methods with an unbalanced rotor blade showed greater average amplitude oscillation, followed by two unbalanced rotor blades and, finally, with balanced rotor.