Identification of wind turbine natural frequencies using narrow-band decomposition methods

摘要

Wind turbine vibration analysis remains an important task, since changes in the non-stationary oscillation modes are indicators of the machine condition. However, the non-stationary vibration signals present modulation phenomena, noise contamination and mode mixing. To cope with this issue, several techniques aim to decompose the signal into simpler functions. In this paper, a two-stage methodology is proposed to decompose the signal, identifying the main frequency bands, fundamental modes and its harmonics. Firstly, the signal is divided into frequency sub-bands using ensemble empirical mode decomposition, which decomposes the signal into a set of intrinsic mode functions. Nonetheless, these frequency bands are overlapped, entailing a difficult separation of the oscillation modes. Therefore, taking into account that each frequency sub-band comprises several oscillation modes, a second stage is rendered using a novel order tracking approach, for which the reference shaft speed is not required. The proposed methodology allows the different oscillation modes compounding the vibration signal to be identified and the relationship between those components associated with modulation phenomena and its natural frequencies to be obtained. The outcomes show that the natural frequency identification task is successful when both stages are used together, rather than separately.