Because of the unobserved sudden fatigue failure of the large-scale and heavyburden supporting shafts, the normal product line is affected seriously. Based on analyzing the relation ship between this type of supporting shafts construction characteristics and vibration signal characteristic, the model of the vibration signal is established by the method of time series.Residual (${\sigma }_a^2$) and normal residue square sum (NRSS) are adopted to identify the fault state, and the degree of supporting shafts' fatigue crack is diagnosed effectively.The result of experiment shows that the method of time series using ${\sigma }_a^2$ and NRSS as the index to diagnose the fatigue crack of large-scale and heavyburden supporting shafts is more sensitive and effective than the normal method using time-frequency domain of amplitude, and has very strong practicability.