**1. Introduction**

As a widely used spatter-free welding technology, pulsed gas metal arc welding (P-GMAW) can achieve directional transition of spatter-free droplets with low heat input through current pulse [1]. However, current pulses with the constant frequency of P-GMAW could not effectively stir the weld pool with the heat-sensitive and high viscosity liquid metal such as stainless steel and aluminum alloys, which often results in the formation of structure defects such as coarse grains, pores and cracks [2]. To solve this problem, many arc-based welding techniques have been developed for advanced materials joining [3,4]. DP-GMAW was developed based on P-GMAW to assisting weld pool oscillation [5]. By periodically changing the output current, double-pulsed gas metal arc welding (DP-GMAW) leads to the periodic change of current pulse frequency. Through the whole process, not only the stable transfer mode of "one drop per pulse" can be obtained [6], but also the frequency oscillation and stirring effect of weld pool can be obviously improved [7]. Therefore, to some extent, DP-GMAW can refine the grains [2], reduce the cracking sensitivity [8,9] and porosity of welds [3,4] and improve the weld formation and joint performance.

Figure 1 shows the waveform of DP-GMAW, and the double pulse period consists of a peak thermal period and a base thermal period. The heat and mass transfer process of the heat pulse is determined by the waveform parameters of double-pulse current such as pulse frequency, current difference and duty ratio of two phases. To study the evolution rules of DP-GMAW weld formation and microstructure, various methods were carried out. Yao and Zhou et al. systematically investigated the influence of current waveform parameters of DP-GMAW on the weld waviness of austenitic stainless steel [10]. They discussed the regularity of DP-GMAW weld formation by using gray theory analysis [11] and further optimized the welding parameters of austenitic stainless steel [12]. Compared with P-GMAW, DP-GMAW has a wider adjustment range, broader root gap configuration and stronger solute agitation with the same heat input rate. Wang et al., suggested that increasing the frequency of thermal pulse of DP-GMAW could reduce dendrite size [13]. Wang studied the influence of the current amplitude of thermal pulse on the geometry, cooling rate, solidification parameters of the aluminum alloy weld pool and weld metal grain size from both experimental and numerical simulation aspects. It was proved that DP-GMAW could increase the cooling rate of the weld pool with the same heat input [14]. In the investigation on welding procedure of ferritic stainless steel and austenitic stainless steel, Shen [8] and Devakumaran [9] both confirmed that DP-GMAW could effectively inhibit the growth of HAZ (Heat Affected Zone) grains and promote the transformation of columnar grains to equiaxed grains in the weld zone. Anhua Liu et al. analyzed the dynamic process of weld pool shape of the aluminum alloy with the aid of high-speed camera [15]. The results showed that the size of weld pool changed synchronously with the frequency the thermal pulse, and the addition of thermal pulse obviously changed the behaviors of weld pool. As the frequency of thermal pulse increased, the grain size of weld metal decreased, and the eutectic Mg2Si precipitates in the weld zone were evenly distributed.

**Figure 1.** Schematic diagram of current waveform of double-pulsed gas metal arc welding (DP-GMAW).

It should be pointed out that the heat and mass transfer process during welding determines the dynamic behavior of the weld pool. Besides, the solidification behavior of weld pool is greatly controlled by the dynamic behavior of weld pool and welding heat input process. So far, studies on the effect of DP-GMAW on grain refinement of the weld mainly concentrated on the optimizing welding procedure. However, little research was conducted on the dynamic characteristics of weld pool in DP-GMAW, and the relationship between the dynamic behavior of weld pool and welding metal microstructure can hardly be established. Therefore, extensive research work needs to be carried out to analyze the influence of waveform parameters of DP-GMAW on the dynamic behavior of weld pool and its relationship with welding metal microstructure.

In this paper, with the help of a laboratory-made high-speed welding photography system, the influence of waveform parameters of DP-GMAW on weld pool oscillation behavior of austenitic stainless steel was studied. The purpose of this study is to explore the internal relationship between weld pool oscillation behavior and welding metal microstructure and explain the action mechanism of grain refinement.
