Diagnosis of misalignment faults by tacholess order tracking analysis and RBF networks

Misalignment faults in rotating machinery may have very harmful effects if not early identified. This study deals with the diagnosis of such faults when the machinery operates under varying speed conditions. Aiming at the problems arising from the ever increasing rotating speed and its consequential influences on vibration signals, this paper proposes an instantaneous frequency estimation based order tracking analysis scheme. It takes the advantages of adaptive instantaneous frequency estimation, correlation analysis and re-sampling technique in angular domain. As a pretreatment before feature extraction, the proposed scheme can remove the effects of varying speed by transform non-stationery signals in time domain to quasi-stationary signals in angular domain without the presence of the tachometer. Afterwards, Wavelet packet decomposition is performed on the acquired signal to obtain its energy distribution in different frequency spans. Finally, RBF networks are employed for faults classification based on the wavelet packet energy ratios of the re-sampled vibration signals. Experiments in different levels of misalignment faults are conducted to verify the validity of the proposed approach.