Aimed at the problem of vibration fatigue failure of Stype metal welded bellows,the vibration signals under relevant working conditions were collected,and the shorttime Fourier transform was used to edit the signals to identify the damage region,and the signals were interpolated and spliced at the end of the region equidistant to obtain the edited signal for fatigue analysis of parts.For the original signal and edited signal,the number of damage cycles was counted by the rain flow counting method,and the average energy of the signal was calculated.The edited signal was loaded into the bellows,which is further verified by the simulation calculation,and the damage distribution nephogram of the bellows under the influence of the edited signal and the original signal was obtained.The energy cycle of the vibration signal in different stress ranges under different rotational speeds was analyzed.The results show that the edited signal obtained by signal reduction effectively retains the damage segment and reduces the length of the signal,which indicates that the signal reduction method based on shorttime Fourier transform can effectively shorten the time of fatigue experiment and accelerate the research and development cycle.At low speed,the radial stress of bellows is larger and the number of radial large stress cycles is higher,and the bellows are more prone to vibration fatigue.