Under different operating conditions, the welding current and arc voltage output waveforms are required because the welding process requires different metal transfer modes. At the same time, the welding process affected by many factors to produce random, occasional information is also reflected in the welding current or voltage waveform data composition, so that the data information structure derived from various influencing factors in GMAW process is superimposed by a number of different natures of probability density functions. In the past, the expression of arc welding parameters was generally premised on the law of large numbers, and the statistical value of the data was calculated according to the general assumption of normal or approximate normal. However, this statistical hypothesis and its calculation method, the data information processing under GMAW conditions is only partially feasible, for the arc welding information based on wave control or pulse current mode is no longer valid in most operating conditions, not only will lead to parameter mean, heat input and other physical errors, but also make the arc welding process occasional outlier data and its hidden key information may not be known. On the analysis of the current and arc voltage data of GMAW processes, this paper explains the reasons for the different data distribution properties of the welding process, and puts forward the way to reconstruct the expression of welding parameters and improves the scientific and effectiveness of data statistics analysis in the current development of welding to automation and informationization, so as to prevent and avoid misreading and misleading the information understanding and application of arc welding process.