Tacholess Instantaneous Rotational Frequency Estimation via NSTFT and Joint NLMS–EMD Adaptive Filtering

Gearbox is a key transmission system in the wind turbine; especially, the main shaft’s operating state has an important influence on the wind turbine’s intelligent maintenance. In recent years, many vibration‐based methods for detecting a gearbox’s main shaft fault have been proposed. Considering that the gearbox often operates under wind speed fluctuation conditions, how to acquire a precise instantaneous rotational speed is an inescapable problem. As in many working conditions, we are unable to install an encoder in the wind turbine gearbox; the instantaneous rotating frequency (IRF) estimation based on vibration signal analysis is necessary. However, in the existing IRF estimation method, the antinoise performance is poor, and the key parameters need to be selected manually which will lead to incorrect diagnoses. To address the problem, we propose a gearbox’s main shaft IRF extraction method that combines nonlinear short‐time Fourier transform (NSTFT) joint adaptive filtering. Simulation and experiment prove it can effectively deal with noise interference and multicomponent coupling phenomenon. Quantitative analysis of the error between the estimated IRF and the actual IRF was conducted. The results indicate that the mean squared error (MSE) is 0.63, and the mean absolute percentage error (MAPE) is 3.42%. For the simulated signal without joint filtering, the estimation of the IRF yielded the MSE of 3.88 and the MAPE of 9.28%. When the signal was processed solely with normalized least mean square (NLMS) filtering, the corresponding MSE and MAPE were 1.61% and 6.21%, respectively. The method integrating the nonlinear short‐time Fourier transform with adaptive chirp mode decomposition achieved the MSE of 1.13 and the MAPE of 4.7% in the IRF estimation. A comparative analysis demonstrates that the algorithm proposed in this paper achieves optimal performance.