Abstract: Detection of cycle stages of the current waveform in DP-GMAW process is the precondition for quantitative feature extraction and power supply evaluation. However,the wide variety of DP-GMAW modulations results in the large difficulties for detection,and related achievements have not been reported. This paper put forward a self-adapting detection method based on the average local current. First,the local time range was estimated,which was followed by the calculation of the average local current sequence. Then,according to the sequence, the double-threshold value method was applied to marking the pulse trains. Finally,each pulse train was subdivided into cycle stages. Tests show that with the above method,five typical types of waveforms during the DP-GMAW process could be accurately divided into phases of pulse train,and the markers of cycle stages within a pulse train were almost all correct,which gave full support to the subsequent feature extraction and other calculations and studies.