1. Introduction

Laser welding has received great attention as promising joining technology with high quality, high precision, high aspect ratio (penetration depth to weld width), high speed, good flexibility and low distortion, in addition to the recognition of easy and wide applications due to congeniality with a robot, reduced man-power, full automation, systematization, and production lines and so on. Thus, applications of laser welding are increasing [1]. In laser welding there are two main processes, namely heat conduction welding and deep penetration welding. In deep penetration laser welding, a thin capillary, called keyhole, is generated in the welded material at high intensity values of the laser beam (>10⁶ W/cm²). The formation of the keyhole enhances absorption of the laser energy [2,3]. Moreover, multiple reflection absorption phenomena on the keyhole wall are highly geometry dependent and vary with laser wavelength [4,5]. Deep penetration laser welding is a very complex process and only partially understood [5]. Recently, high-power and high-brightness Yb fiber and disk lasers have been developed for deep penetration laser welding with maximum output power up to 100 kW [6]. The new solid state laser with short wavelength has great potential for welding thick section steel of industry [1,7]. However, welding defects, namely porosity, spatter, undercut and root sagging are prone to be generated in high-power laser welding of thick plate [8,9]. The fundamental causes of the welding defects lie in the dynamic behaviors of the keyhole and the melt pool in deep penetration laser welding [10,11]. It is usually considered that the keyhole instability is associated with this irregular flow of molten material around the keyhole, nevertheless, the relevant mechanisms of the keyhole instability has not been proposed up to now [11]. Therefore, understanding of the keyhole dynamics in deep penetration laser welding is necessary if one wants to improve the weld seam quality, especially, during high-power laser welding of thick plate.

Direct observation of the keyhole is not feasible when welding nontransparent materials such as metals. However, efforts have been made to experimentally observe the keyhole shape. The keyhole produced during welding of transparent soda-lime glass has been observed [12], but the captured image was not clear or stable due to the small temperature difference between the melting and vaporization points. Keyholes produced during deep penetration laser welding of metal (steel) have been observed using an online X-ray facility [13]. However, the contrast of the captured radiographs was too low to distinguish the keyhole shapes. Following this, several experimental investigations [14,15] were carried out to determine whether the dynamic keyhole shapes produced during deep penetration laser welding of various metals could be observed using an X-ray transmission imaging system with a high-speed video camera. Unfortunately, most of the keyholes that were obtained were not clear enough to use for a quantitative study. In addition, welding experiments employing water and ice [16] have been performed to observe the capillary behavior and the flow field in the weld pool, even though there are significant differences in the material properties of water and metals. Recently, a special transparent material, borosilicate glass (GG17), has been exploited to observe the keyhole in deep penetration laser welding with a CO₂ laser source [17–19], and a “sandwich” specimen [18–20] using this material allows for direct observation. However, in deep penetration welding, the internal keyhole cannot be observed directly since the GG17 glass melts and evaporates via direct absorption of the CO₂ laser. The resulting molten layer and hot plume of glass will attenuate the emission spectrum of the plasma and prevent observation of the liquid keyhole wall. Moreover, this specimen is somewhat different from actual laser welding conditions of a compact metal, since one or more thin metal films are fixed between two pieces of GG17 glass and form a loose multilayer construction. As a result, the welding pool is prone to collapse, which affects the stability of the welding process and could produce misleading results [20].

In this paper, we introduce a modified sandwich specimen to directly observe the keyhole during deep penetration laser welding with a 10 kW fiber laser. The modified sandwich specimen consists of one sheet of stainless steel and one piece of GG17 glass. The GG17 glass does not absorb light with a wavelength of about 1 μm (e.g., light from fiber, Nd:YAG, and disc lasers) but melts via heat conduction from the welding pool. In this way, the keyhole can be captured in real-time by a complementary metal-oxide-semiconductor (CMOS) high-speed video camera (5000-20 000 frames per second (fps)) without being hidden by the molten layer or a hot plume of glass. We are thus able to directly observe a clear image of the keyhole wall during laser welding of a metal for the first time. The hydrodynamic phenomena surrounding the keyhole wall are also captured and analyzed. The formation of related welding defects, namely bubbles, undercuts and spatters, is analyzed based on the recorded images. Furthermore, the emission spectrum of the laser-induced metallic vapor plasma plume inside the keyhole is accurately detected. Based on the measured spectral lines, the characteristics of the keyhole plasma plume such as the electron temperature, electron density, ionization degree and pressure can be accurately calculated.

2. Experimental setup

The experimental setup is shown in Fig. 1. Experiments are performed with a continuous wave (CW) fiber laser (IPG YLS-10000) with a maximum power of 10 kW. The beam parameter product (BPP) was 7.5 mm·mrad for a processing fiber 200 µm in diameter. The laser beam emitted from the end of the optical fiber was collimated by a lens with a 150 mm focal length and then focused onto the specimen surface using a lens with a 300 mm focal length. The final focal spot size was 0.4 mm in diameter. Motion direction of welding head via a machine tool is defined as the welding direction.

 

Fig. 1 Schematic of the experimental setup: (a) Setup for observing the longitudinal keyhole with a high-speed camera system. (b) Setup for capturing the keyhole plasma plume with a spectrometer system.

Download Full Size | PDF

The spectrometer system consists of a SpectraPro-2356 spectrograph, an area-array 1340 × 400 pixel charge-coupled device (CCD) detector (Princeton Instruments, PIXIS-400), and special optical fiber bundles. The spectrograph has a slit width of 3 mm and a grating of 1200 g·mm−1, the blaze wavelength of which is 300 nm, and the wavelength coverage is 62 nm. This gives the wavelength resolution of the optical emission spectroscopy (OEM) system as approximately 0.05 nm (62/1340). A standard CMOS high-speed video camera (Photron Fastcam SA4) with a highest frame rate of 500 000 fps was used to observe the keyhole, and a diode laser (λ = 808 nm) with a maximum power of 30 W was used to illuminate the welding zone. To block scattered light from the welding process, which would otherwise saturate the image, a narrow band-pass interference filter was placed in front of the camera lens.

The welding specimens were austenitic stainless steel (Type 304) 5 mm in thickness and a GG17 glass. The modified sandwich specimen is made from one sheet of stainless steel and one sheet of GG17 glass both of which have a size of either 40 × 40 × 5 mm³ or 40 × 12 × 5 mm³, which are subject to partial penetration welding and full penetration welding, respectively. The GG17 glass has a low thermal expansion coefficient and excellent heat resistance, which prevents it from shattering during the laser welding [19].

3. Experimental results

Figure 2 shows images of the longitudinal keyhole in the modified sandwich specimen taken during the 10 kW laser welding. The side keyhole wall can be seen clearly, and the keyhole is surrounded by a thin molten layer. The front keyhole wall is inclined slightly in the welding direction. The significance of these figures is that we can directly observe the laser-induced metallic vapor plasma plume located inside the keyhole. The keyhole wall is not smooth but rather rough with typical gauffer structures owing to the surface tension (capillary collapse) [21], as shown in Fig. 2. Moreover, we can also see micro-droplet being torn off the liquid wall of the keyhole. This may be attributed to the small-scale capillary waves induced on the keyhole wall due to the capillary-evaporative instability of the liquid surface [22]. The ejected micro-droplet flies toward the rear keyhole wall in a direction perpendicular to the front keyhole wall, as illustrated in Fig. 2 (indicated by A). The micro-droplet can be pushed downward by the expanded metallic vapor generated at the moving shelf [23] on the front keyhole wall, which is moving at high speed (see the points indicated by A in Figs. 2(c)-2(d)). Such a micro-droplet is then redirected toward the rear keyhole wall by the lower intense metallic vapor generated at the shelf (see point A in Figs. 2(e)-2(g) and the red arrow in Fig. 2(e)). The size of the micro-droplet that arrives at the rear keyhole wall is much smaller than that ejected from the front keyhole wall, as shown in Figs. 2(a) and 2(g). These micro-droplets lose some volume through evaporation when passing through laser radiation field inside the keyhole, thus causing an increase in the metallic vapor pressure and may even metallic vapor bursts [24]. All these observations are consistent with previous experimental observations [25] and theoretical analysis [23,24,26].

 

Fig. 2 Direct observation of the longitudinal keyhole during 10 kW fiber laser welding of a modified sandwich specimen at a welding velocity of 1.2 m·min⁻¹ and a + 5 mm defocus (Media 1).

Download Full Size | PDF

In the set of images in Figs. 2(a)-2(g) and Media 1, we can identify gauffers that move down the keyhole owing to the recoil pressure of the metallic vapor. In addition, there is a bright zone on the front keyhole wall that is accompanied by a substantial amount of metallic vapor directed toward the rear keyhole wall, as shown in Figs. 2(c)-2(g). This behavior may be caused by the moving liquid shelf, which the laser beam irradiates directly, resulting in intense evaporation [23,27]. The top surface of this liquid shelf is subject to vapor ablation recoil pressure, and thus the melt underneath the shelf is quickly driven down the front keyhole wall [25,26]. The velocity of this violent downward flow is on the order of tens of meters per second [26,28]. On the rear keyhole wall, we can see that a bulge forms, which then moves up toward the surface. The same phenomenon observed using an X-ray method in [25] was defined as a deep depression in the keyhole wall that moved downward and was explained by intense evaporation that occurred at the hump or shelf on the front keyhole wall; the dynamic pressure of the metallic vapor jet created an indentation in the rear keyhole wall. Here, we hypothesize that the chain of bulges and constrictions on the rear keyhole wall is a result of the pressure unbalance with a deep keyhole. Specifically, the pressures in the plasma vapor flow inside a deep keyhole and in the molten pool play an important role for keeping the central region of the keyhole open [29]. It is not surprising that the interaction between the fluid pressure and the surface tension causes the rear keyhole wall to bulge and constrict during laser welding of a thick plate; in fact, the indentation induced by the intense metallic vapor jet generated on the shelf enhances the local deformation of the rear keyhole wall.

Figure 3 and Media 2 present images showing the formation of a bubble near the tip of the keyhole. The brightness in Fig. 3 has been increased to highlight the bubble. Dynamic fluctuations of the keyhole tip can be seen, and occasionally, indentations appear in the rear keyhole wall from the impact of an intense metallic vapor (Figs. 3(a) and 3(b)). At the same time, the violent downward flow of liquid along the front keyhole wall gives rise to some strong vortex flows through the interaction with the solid-liquid boundary of the bottom weld pool [26]. As a consequence, the hydrodynamic pressure due to the vortex flow of molten metal behind the keyhole attempts to close the keyhole, and a bubble is formed at the tip of the keyhole (Figs. 3(c)-3(f)). Interestingly, the bubble is compressed by the extended keyhole tail (Figs. 3(e) and 3(f)), and a pore forms in the weld.

 

Fig. 3 Direct observation of the longitudinal keyhole and plasma plume with formation of bubble during 10 kW fiber laser welding of a modified “sandwich” specimen at 1.5 m min⁻¹ welding velocity and 0 mm defocus (Media 2).

Download Full Size | PDF

During full penetration laser welding of the modified sandwich specimen, the tip of the keyhole opens and closes, as shown in Fig. 4 and Media 3. A large amount of metallic vapor is ejected from the opened keyhole, which is accompanied by the splashing of substantial melts from the bottom of the keyhole (Figs. 4(a)-4(c)). Consequently, the top surface of the molten pool moves downward to form a deep undercut on the top surface of the resulting weld. We note that we can observe an oblique melt column (Fig. 4(c)) that moves downward without producing metallic vapor to hang from the bottom melt pool, as shown in Figs. 4(d) and 4(e). This behavior may be attributed to the fast downward flow along the front keyhole wall induced by the local evaporation on the top surface of the shelf [28]. This long melt column eventually separates from the molten pool and splatters under the action of gravity.

 

Fig. 4 Direct observation of the keyhole outlet and melt ejected from the bottom during full penetration laser welding of a modified sandwich specimen with a 10 kW fiber laser at a welding velocity of 1.2 m·min⁻¹ and a −10 mm defocus (Media 3).

Download Full Size | PDF

The optical emission of the laser induced plasma plume inside the keyhole can be captured directly from a side view. Figure 5 shows the typical spectral lines of the keyhole plasma plume during 10 kW fiber laser welding of stainless steel. Most of the lines are atomic emission spectral lines, and the ionic spectral line is too small to be distinguished.

 

Fig. 5 Optical emission spectrum of the captured keyhole plasma plume based on a modified sandwich specimen during 10 kW fiber laser welding of stainless steel at a welding velocity of 1.2 m·min⁻¹ and a + 5 mm defocus.

Download Full Size | PDF

From this emission spectrum, the characteristics of the keyhole plasma plume such as the electron temperature, ionization degree, electron density, and pressure can be calculated [30,31]. Figure 6 shows a Boltzmann plot obtained from the Fe I lines in Fig. 5. This plot gives the electron temperature of the keyhole plasma plume to be ~5720 K. Subsequently, the degree of ionization of the keyhole plasma plume can be calculated using the Saha equation. At 5720 K, the degree of ionization is 0.75, and therefore, we conclude that the laser-induced plasma plume inside the keyhole is a weakly ionized plume.

 

Fig. 6 Boltzmann plot obtained from the Fe I lines of the keyhole plasma plume.

Download Full Size | PDF

4. Conclusion

By using a modified sandwich specimen containing GG17 glass, the keyhole characteristics during deep penetration laser welding using a 10 kW fiber laser could be directly and clearly observed. The keyhole wall and liquid motion along the keyhole wall were captured by a high-speed camera. The moving liquid shelf on the front keyhole wall and the accompanying fast downward flow and vapor jet were observed simultaneously. Micro-droplets torn off the keyhole wall due to a capillary evaporative instability and the resultant bursts of vapor were also visualized.

The direct observation results showed that the keyhole wall was not smooth but rough with typical gauffer structures. The local evaporation at the shelf and evaporation of the micro-droplets in the laser irradiation field caused metallic vapor bursts inside the keyhole. The pressures in the plasma vapor flow inside a deep keyhole and the molten pool interact with the keyhole wall surface tension and can cause the keyhole wall to oscillate; bulges and constriction of the rear keyhole wall were seen. The vortex flow of the molten metal behind the keyhole, induced by the fast downward flow along the front keyhole wall, exerted a hydrodynamic pressure on the fluctuating rear keyhole wall, closing the keyhole and forming a bubble. During full penetration welding, a large amount of metallic vapor was ejected from the opened keyhole, which was accompanied by the splashing of substantial melts and the formation of an undercut defect. Moreover, the optical emission spectrum of the laser-induced plasma plume inside the keyhole was acquired accurately by a spectrometer. The characteristic spectral lines of the plasma plume allowed the electron temperature, electron density, and pressure of the keyhole plasma plume to be calculated. The results presented here will assist in gaining a more detailed understanding of the keyhole effect and the mechanism responsible for welding-defect formation during deep penetration laser welding, as well as providing a possible technique for observing laser micromachining. More generally, this novel method also can be used to reveal the processes which appear in the depth of the material due to the high aspect ratio of the structures under investigation.