Extraction of Fault Features of Machinery Based on Fourier Decomposition Method

Separations of close-frequency signal components are important for accurate diagnosis of machinery faults. Conventional methods for processing vibration signals of defective machinery can be divided into adaptive methods and parametric methods. Empirical mode decomposition (EMD), an adaptive method, suffers from a bottleneck of performance in adaptively separating close-frequency signal components. Parametric methods, such as ensemble EMD (EEMD) and variational mode decomposition (VMD), face difficulties in parameter setting. As a result, these conventional methods demonstrate limited capabilities to extract fault features from machinery vibration signals. To overcome these deficiencies, this paper proposes a novel method for extraction of fault features of machinery based on Fourier decomposition method (FDM). Firstly, this paper demonstrates adaptive narrow-band filtering properties of FDM both in low frequency and in high frequency by examining Gaussian white noise. Afterwards, this paper numerally proves that FDM can transcend the bottleneck of performance of EMD in adaptively separating close-frequency signal components. Furthermore, the proposed method is compared with the method based on EMD, EEMD or VMD by investigating a turbine gearbox vibration signal. The results show that the proposed method outperforms the others in feature extraction of machinery vibration signals.