Abstract: A new time series decomposition method is introduced. It is deduced from S-transform; the algorithm is used to separate one dimension time series into two dimension local time series. By researching on the different separating component, the local characteristic and implication distinction of the time series can be gained. Though numerical simulations experiment, it is proved that the algorithm is feasible on extracting characteristic information. A statistical characteristics parameter is defined based on the time series decomposition method. It is used for the data analysis of different fault gear; the result indicates the method has an excellent effect on machinery fault classification.

Key words: Fault diagnosis, Gear, Time domain decomposition, Time series