Time series analysis of vibration signal in the accelerating process of gear box

In the accelerating or decelerating process, the vibration signal is non-stationary signal. The primary characteristic is violent vibration, and the fault information can stand out. Usually the special analysis techniques of non-stationary signal such as STFT, Gabor transforming, Wigner-Ville distribution and WT are applied to obtain the varying law of signal relative to frequency and time and increase the accuracy of analysis result, but in the vibration analysis of gear box, the angle-domain analysis method is preponderant. Sampling frequency changes along with the rotate speed, and the non-stationary signal relative to frequency is changed into the stationary signal relative to the angel of rotor, and then the power-spectral analysis method is applied. The classical power spectrum estimation methods such as period map method, autocorrelation methods are applied widely in engineering application, but the limitations of these methods exit such as low variance performance and resolution. This paper adopts the method of parameter model power spectrum estimation to analysis the signal, and the resolution of frequency is increased.