Abstract: Stages of run-up or run-down of rotating machinery are non-stationary. Then, the methods of time-frequency analysis, such as short-time Fourier transform,wavelet transform and so on, will be employed to analyze the corresponding test signals. To obtain the laws of the instantaneous frequency of multi-components signals during the run-up or run-down periods, the method based on peak search is widely used. However, the results through direct searching are often worse in precision, due to the interference of strong noise and adjacent components of signal. The hidden markov models are applied to solve the problem. Simulation indicated a better result could be obtained by the method based on HMM.

Key words: HMM, Instantaneous frequency, Peak search, Time-frequency analysis