The Hilbert-Huang transform (HHT) has been used in many fields for time-frequency analysis of the non-linear and non-stationary signal. But when implementing the HHT to analyze the measurement signals from actual offshore wind turbines, due to the large rigidity and weak wave energy, the results will suffer a serious mode-mixing problem and contain a lot of noises which will lead to a bad result. In this paper, a new time-frequency analysis method based on the Hilbert transform is introduced, the Intrinsic mode functions (IMF) obtained by empirical mode decomposition (EMD) is redefined by introducing a combination of state-space model and energy gridding. To demonstrate the proposed method, one numerical example of three slow varying frequency components was used. Specifically, a field test on an offshore wind turbine located in the Putuo of Zhejiang Province of China is introduced, and measured data are used to verify the effectiveness of the proposed method. The result demonstrates that the proposed method has a better performance when implemented on the offshore wind turbines.
Due to the stability and availability of the wind sources in the marine environment, the offshore wind energy has become one of the fastest growing renewable energy in the world. But the cost of investment is much higher than that of onshore wind energy (Ren et al., 2018). And because of the impact of the harsh marine operation environment, offshore wind turbines are vulnerable to the process of exceeding wearing and accelerated degradation in for their key components in the offshore wind turbines is accelerated, making the structures vulnerable to be destroyed (Li and Peng, 2016). For example, there were nearly 1000 safety accidents of offshore wind turbines in 2014 (Seyr and Muskulus, 2016). Therefore, a real-time monitoring of the offshore wind turbine becomes more and more important.
Traditional signal processing methods, such as fast Fourier transform, require the signals should satisfy the hypotheses of stationary and linear. However, when analyzing non-stationary or non-linear signal, it is not enough to only obtain the information in time domain or in frequency domain. It is also necessary to understand the variation of spectrum with time. The traditional method is no longer suitable and effective signal processing techniques are needed to analyze the signals (Sejdic et al., 2009). Therefore, by establishing a joint function of time and frequency, the frequency energy intensity of the signal at different times can be described, this is also the source of the idea of time-frequency analysis. In recent years, with the increasing attention to nonlinear vibration of structures (Xing, 2000), the frequencies of the components of nonstationary signals and their time-varying characteristics can be effectively revealed by implementing the time-frequency analysis, and it also becomes a novel and practical technical for analyzing structural health monitoring in vibration systems (Alisaraei et al., 2016). Through the time-frequency analysis, the characteristics of the signal can be displayed, and by selecting a stable time period and taking targeted measures, the correctness of the modal analysis results is ensured (Liu et al., 2016), thus avoiding the occurrence of major accidents.
