Gearbox fault diagnosis under fluctuating speed conditions using vibration analysis

Abstract

Condition monitoring and early fault diagnosis of gearbox using vibration analysis has gained signi cant importance in order to avoid catastrophe. In the real-life operating conditions, the speed of the gearbox uctuates to compensate applied load, resulting a behavioral change in gearbox vibration signal. Basically, gearbox vibration signal consists of shaft rotation frequency, tooth meshing frequency their harmonics, fault transients and noise. The non-stationary behavior of gearbox vibration signal due to uctuating speed leads to the ine ectiveness of some of the prominent signal processing techniques like discrete Fourier transform (DFT) and time synchronous averaging (TSA).One of the approaches to minimize the e ect of uctuating speed is by resampling the acquired vibration signal for every revolution. These resampled time instants have been used for modifying TSA which has resulted in the increase of signal to noise ratio (SNR). Keeping in the view of early gearbox fault diagnosis, two condition indicators (CIs) are proposed in this thesis. Both limited and high speed uctuation with di erent gear faults and di erent loads have been used to evaluate the proposed CIs.The signal entropy based on probability density function (PDF) fails to detect fault at uctuating speed. In this thesis, root mean square (RMS) based PDF has been proposed. Entropy features when evaluated by proposed RMS based PDF were able to detect defect for uctuating speed conditions. In an another attempt, dynamic time warping (DTW) of intrinsic mode functions (IMFs) decomposed by empirical mode decomposition (EMD) and frequency domain averaging (FDA) of warped IMFs has improved the performance of DFT for fault diagnosis under uctuating speed conditions.Under uctuating speed conditions, cyclic changes have been observed in nth order moment of a vibration signal. Such non-stationary gearbox vibration signal are characterized as cyclostationary signals. Conventional CIs like RMS and kurtosis fail to sense faults, both for time domain and angular domain signals. Therefore,CIs based on cyclostationarity (ICS2Y and DCS ) have been explored for gear fault diagnosis. It has been observed that ICS2Y and DCS have detected cracks present in the gear tooth under uctuating speed conditions.To acquire gearbox vibration signals, accelerometer is placed conventionally on the bearing housing. This leads to distortion/ corruption of signal due to signal transfer paths and interference from other sources in the acquired vibration signal. Therefore,to minimize the distortions due to transfer paths, accelerometer has been placed internally on a gear collar. Placing accelerometer internally has resulted in the increased strength of vibration signals. The e ect of transfer paths on the gearbox vibration signal has also been studied and consequently its e ect on the performance of fault diagnosis is presented. To summarize, this thesis proposes signi cant modi cations in the signal processing techniques, two novel CIs, and a novel method to place accelerometer for gear fault diagnosis. The results hence obtained, will add value to the knowledge towards signal processing approaches and their e ectiveness on gearbox fault diagnosis under uctuating speed conditions.