On the basis o f the very good t ime- frequency domain featur es of wavelet, this paper studies, the continuous wavelet coefficients of turbine mechanical vibration signal. Considering feature ex tracting, this paper puts forward new stat istics and proves that this kind of one-o rder moment is very effective with 8 kinds of fault analysis results given. Preparator y research show s that t his quant itat ive analysis met hod can be used for extr acting features of vibr atio n siganl.