Abstract: In this paper, a series of welding tests are carried out by adjusting the pulse waveform parameters of the welding power supply. Changes of the welding current signal are compared and analyzed after modifying pulse current peak coefficient, pulse current peak time coefficient, pulse current base value coefficient, pulse current rising coefficient and pulse current falling coefficient. Ensemble empirical mode decomposition (EEMD) method is adopted to decompose the welding current signal collected in real time during the welding process. The welding current signal and a series of intrinsic mode function (IMF) were analyzed in time-frequency domain. Then, the stability of welding process was evaluated according to the analysis results and weld surface morphology. The test results show that the EEMD method can be used to resolve the characteristic IMF closely related to the short-circuit transition process from the welding current signal. Compared with the unstable welding process, the characteristic IMF spectrum distribution is significantly different. The narrower the frequency distribution range of the characteristic IMF, the more stable the welding process, the smaller the welding splash, and the better the weld surface.