A time-frequency method based on novel oddsymmetric window functions for mechanical equipment fault diagnosis

Abstract

Under time-varying operating conditions, signals generated by mechanical equipment usually exhibit strong non-stationarity, frequency-modulation and amplitude-modulation (FM-AM), and contamination by background noise. These characteristics pose three main challenges for time-frequency analysis (TFA): the instantaneous frequency trajectories potentially vary rapidly, weak fault-related components can be obscured by broadband noise, and conventional TFA methods suffer from energy leakage and blurred time-frequency ridges. These issues make it difficult to obtain a concentrated and interpretable time-frequency (TF) representation for machine condition monitoring. To address these challenges, a novel TFA method, referred to as the synchronous general linear Chirplet transform based on odd symmetric window function (SGOT), is proposed in this paper. Different from the traditional methods that are based on even-symmetric window functions, the proposed SGOT leverages the TF energy leakage associated with an odd-symmetric window functions to effectively characterize the TF properties of signal. Moreover, an odd symmetric synchronous Chirplet operator (OSSCO) is proposed for better fitting the signal TF ridges to achieve high TF aggregation performance for accurately extracting the TF energy of the signal. Test results show the SGOT method not only performs well in analysing strong FM-AM signals but also exhibits outstanding anti-noise performance compared with existing approaches. Furthermore, the proposed method has been applied to different machine condition monitoring case studies, such as rolling bearing vibration signal, tapper bearing inner race fault vibration signal and robot joint current signals. These case study results demonstrated that the SGOT method achieves excellent performances in the fault diagnosis of rolling bearings and tapered bearing as well as in capturing change behaviours of robotic joints, which further confirms the significant advantages of SGOT method in extracting and representing the TF features of nonstationary signal.