Investigations on the formation of pores during laser beam welding of hairpin windings using a high-speed x-ray imaging system

The efforts to make the energy supply sustainable have become even clearer with the 2022 climate protection program, which was adopted in June 2021. The transport sector plays a significant role in this, accounting for around 1/5 of the greenhouse gases emitted.¹ The switch to climate-neutral energy carriers such as electric drives is essential for achieving these targets.² It is expected that about 150 · 10⁶ electric vehicles will be registered worldwide by 2030.³ The reliable and economical production of batteries and electric drives is crucial for the manufacture of electric vehicles.

To achieve a high efficiency and performance of electric drives, a rectangular wire, so-called hairpins, has become established. Depending on the design of the electrical drive, more than 150 pairs of hairpins are needed for one stator.⁴ A process with a high efficiency and, consequently, a short cycle time, such as laser beam welding, is needed to join the high number of hairpin ends.⁵ 

For a good electrical connection that can withstand high currents and mechanical stress, the largest possible connection area is required. This connection area should have a low porosity and, if possible, create no spatters during the joining process, which can cause damage to other components.⁴ Due to the high thermal conductivity of copper, a high energy density is required as the weld zone cools down quickly. In addition, a short process time and, consequently, a high welding speed are necessary for high efficiency. Because of small welding geometries required in hairpin welding due to the size of the workpiece, high frequent changes of the direction and crossings of the transition area at the same position affect the behavior of the keyhole. This differs the process enormously from line welding. Due to these high dynamics in the laser welding of copper hairpins, pore formation is a huge issue for joining processes.

The formation of pores has been described mainly for line welds so far. Rapp and Schiebold^6,7 have already divided the formation of pores into metallurgical and process-related causes. It is noted that the formation of metallurgical pores, in particular, depends on the material characteristics. Alter et al.⁸ have demonstrated this effect on laser-welded copper line seams and Heider⁹ has demonstrated this effect on hairpins for different oxygen contents in the copper material but without protective gases. Horrigan¹⁰ showed that the solubility of oxygen in copper increases with increasing temperature, which is why pores are formed in the solidification process due to the reduced oxygen solubility.¹¹ In contrast, process pores are generated, in particular, by the parameters of laser beam welding. Matsunawa et al., Kaplan et al., and Lin et al.^12–14 have already described the formation of pores during the laser beam welding of aluminum and zinc and Heider¹⁵ for copper. The keyhole grows at the bottom and is constricted by the flow of the melt, and finally, a pore is formed. Lin and Fetzer^14,16 showed that the welding speed has a particular influence on this effect. Alter et al.⁸ demonstrated on line welds for copper that the use of inert gas (argon and nitrogen) can reduce metallurgical pores, especially for electrolytic tough pitch copper (Cu-ETP). Gläßel¹⁷ did not found any influence of protective gases on porosity in the hairpin welding of Cu-ETP. It should be noted that a comparison of the welding results was only made by measuring the resistance of the welded hairpins.

For hairpins, Bocksrocker et al.¹⁸ showed that using beam shaping is a promising approach to avoid the formation of pores. However, only one copper wire was used for this investigation. It has already been demonstrated on hairpin-pairs in Ref. 9 that static beam shaping (using a ring/core shape) reduces the pore volume. In this case, only a core laser power of 40% was used. The behavior of the keyhole during welding and the description of the pore formation process in the special application of hairpin welding are still not investigated. In addition, the effects of different beam intensities on the keyhole should be analyzed for hairpins, as only one intensity distribution was used for copper welds so far. For visualization of keyhole during the process, high-frequency x-ray imaging has proven to be useful in different cases.^19–22 Therefore, this work takes the influence of different beam intensity profiles on the formation of pores with hairpin pairs into account. Specifically, the behavior of the keyhole in the transition area from one pin to the other is evaluated. In addition, the welding speed and the use of a protective gas are included in the experiments as relevant variables for influencing the pore formation during welding.

In order to investigate the pore formation in welding two copper hairpins, experiments using an x-ray imaging system to radiograph the processing zone are carried out. For these investigations, hairpins made out of oxygen-free copper (Cu-OF) are chosen in order to eliminate the influence of metallurgical pores in the best possible way. As protective gases are mainly expected to reduce metallurgical pores,⁸ Cu-ETP is used for experiments with protective gases to see a clear effect on pore formation.

A disk laser (TruDisk 8001 by TRUMPF, Ditzingen, Germany) with a wavelength of 1030 nm is used for the experiments. The beam is transported to the processing optics by means of a two-in-one beam²³ delivery fiber with a core diameter of 100 μm and a ring diameter of 400 μm (BrightLine by TRUMPF, Ditzingen, Germany). A scanner optics (PFO33 by TRUMPF, Ditzingen, Germany) is used to position the beam on the surface of the hairpins. Using an aspect-ratio of 1:1.7 results in a beam diameter in the beam waist of the core beam $d f , c$ of 170 μm and in a beam waist of the ring beam $d f , r$ of 680 μm. The focus position of the ring and the core beam is set on the pin surface. An x-ray imaging system is utilized to determine the shape of the vapor keyhole and to observe the formation of pores during the welding process by radiographing the processing zone in the transverse x-direction. The high speed camera records 5000 frames per second in the following experiments and is described in more detail in Ref. 19. Figure 1 shows the schematic layout [(a) an x-ray tube FXE−225.48 by Comet Yxlon, Hamburg, Germany, with a minimum spot size of 6 μm and an acceleration voltage of 150 kV with a current of 566 mA; (e) a scintillator by Hamamatsu, Shizuoka, Japan, with 728 × 728 pixels with a resulting resolution of 58 pixels mm⁻¹; and (f) a high speed camera SA5 by Photron, Tokyo, Japan]

The hairpins are clamped gapless and positioned underneath the processing optics. For analyzing the use of the protective gas in a comparison to welding without gas, a nozzle is added. The initial process is running without an additional protective gas. Figure 2 pictures the experimental setup.

A major challenge is the high density of the copper with a pin width of around 3 mm, which is why the automatic postprocessing of images is required. A typical processed image of the x-ray recording is shown in Fig. 3. In order to minimize the background noise, the gray-values of a time-averaged image [Fig. 3(a)] of the pins are subtracted from the gray-values of the captured images [Fig. 3(b)]. Due to subtraction [Fig. 3(c)], bright areas are created when the pins are melted, in which there is no more material after melting. The melted shape of the pins can be recognized by a black background, which is indicated by a white dashed line in Fig. 3(c). The contour of the pins of the time-averaged image before welding remains visible in the image even after melting because of the described subtraction.

By exploiting the information about the attenuation of x rays contained in the gray values of the acquired image, it is possible to infer pores and the keyhole. The change in gray values due to a pore is plotted in Fig. 4. The values between the two red dashed lines describe the noise of the gray value. In the transition area of the pins, a small gap can be identified as black (gray value: 0). This gap is also visible in Fig. 3(c). During the welding process, the keyhole can also be identified in the same way as the pore.

The geometry of the hairpins and the chosen laser beam path are shown in Fig. 5. The following strategy is used for all experiments. Three different contours are utilized to weld the pair of pins. The first two ellipses are used to melt each pin. The second ellipse is repeated less often, since this pin has already been heated up when the first was melted. A power ramp is used to pierce the solid pins. The third geometry is a circle positioned center over the two pin ends. This contour combines two small melting baths into one large one.²⁴ Two different experiments with and without beam shaping are carried out to determine the influence of static beam shaping on the formation of pores. For the initial experiment, only the core of the beam delivery fiber is used (Sec. III A). The experiment is, henceforth, referred to as “single spot weld.” The settings of single spot welds are listed in Table I.

For experiments with static beam shaping (Sec. III B), the laser power and the number of scans are adjusted with the intention to keep the melted volume approximately constant. The used total laser power P_max for exemplary three different intensity distributions in beam shaping is illustrated in Fig. 6. Intensity distribution steps between plotted distributions are linearly approximated. If the welding speed is changed in the experiments, the line energy $P m a x v$ is kept constant for each intensity distribution. As the laser provides a maximum power of 8000 W, intensity distributions with a low core power P_C cannot be fully mapped with high welding speeds.

is specified.²⁵ 

In the initial single spot weld experiments, the keyhole geometry and the formation of pores are primarily observed for the third contour (N3) in x-ray recordings. Figure 8 illustrates the course of a welding process in a series of images. The time indicated specifies the elapsed time, which is relative to the start of the welding process. The first image shows how the first ellipse (N1—total duration 34 ms) melts the right pin. In the second image, the second geometry (N2—total duration 29 ms) has already created a joined melt pool. The other images visualize the growth of the keyhole during the third, circular geometry (N3—total duration 49 ms). During the circular motion, the keyhole and, thus, the weld penetration depth grow steadily as indicated in Fig. 8 (bottom right). The keyhole penetrates deeper into the material in a spiral pattern (pictured in the last image), up to a maximum welding depth of approximately 2.3 mm. The scale shown applies to all images in this series. The total duration of the welding process is 112 ms.

By analyzing the x-ray videos, it is possible to retrace the pore formation in the hairpin welding process in detail. In the recordings, the process of formation of pores, in which pores detach from the lower end of the keyhole, can be observed. It is noticeable that this process mainly takes place in the plane between two pin ends. Figure 9 shows the formation of such a pore in a series of images. In the first image, the keyhole is located in the left pin and is fully formed. The second image already visualizes how the keyhole is strongly bent and elongated at the lower end. The opening of the keyhole is already above the right pin, with the lower end extending beyond the cleavage plane and into the left pin. In the following two images, the keyhole widens significantly before finally collapsing, leaving behind the pore visible in the last image. Here, the process leading up to the detachment of the pore takes only about 7 ms. The remaining pore is measured with a width of 0.90 mm from the x-ray image. Assuming an ideal sphere for pore geometry, a pore volume of 0.38 mm³ can be estimated. The CT measurement leads to a pore volume of 0.42 mm³ and pictures a spherical shape. The pore analysis of the CT-measurement of this exact pin is illustrated in Fig. 7(b).

To determine the influence of static beam shaping on the formation of pores, the second experiment is carried out. First, the influence of intensity distribution between the core and ring beam is observed. In Fig. 10, the resulting keyhole geometry using three different intensity distributions in the core and ring beam can be seen. It points out how the keyhole geometry changes its shape with a change in the intensity distribution. All three images picture the keyhole during the fourth repetition of geometry N3 of the welding process at the same position. Figure 10(a) shows the shape of the keyhole for the single spot weld. At the lower end, the keyhole is wider and easily visible. The keyhole opening is very narrow. While increasing the power in the ring beam to 30%—Fig. 10(b)—it can be recognized that the keyhole opening at the surface of the pin widens. The keyhole changes from a saxophone-shape to a cylindrical shape. This is expected to make the keyhole more stable.¹⁵ While increasing the power in the ring even further to 60%—Fig. 10(c)—the keyhole becomes even wider.

In order to investigate the influence of the intensity distribution on the porosity of the welding result, the core power is reduced from 100%, in steps of 10%, to a minimum of 20%. Three samples are welded at each step. The welding speed remains constant (Table I). The power is adjusted according to Fig. 6. Figure 11 shows the produced melt volume of each intensity distribution level. Since the melt volume V_Cu still varies slightly by using a linear approximation of the total laser power for different intensity distributions, small deviations in the melted volume must consequently be taken into account. To eliminate this effect in the pore analysis and to be able to compare all parameters used, the relative pore volume V_pore(%) is utilized. The measured data indicate a high repeatability in the melt volume V_Cu in each intensity distribution level.

Figure 12 plots the result of the relative pore volume. An average value is also calculated for each intensity distribution. The mean values are charted in a trend line (red). Only three samples per intensity distribution are evaluated in the test series, but a clear trend is indicated. In addition, as the interval of set intensity distributions is narrow at 10%, the trend of the results is considered to be representative. Possible outliers that can occur with a larger number of samples are not taken into account. In particular, for a power in the core beam P_C ≥ 90% ⋅ P_max, a high mean value is obtained with simultaneously high scattering of welded samples. Specifically, core powers in the range of 40% P_max ≤ P_C ≥ 70% ⋅ P_max show a mean relative pore volume of V_pore(%) ≤ 0.2%. The lowest scatter is achieved with P_C = 40%⋅P_max.

The results of the investigation with variations in the welding speed are described. For this experiment, the intensity distributions P_C = (40%, 70%)⋅P_max as a result of the previous experiment and the single spot weld P_C = 100%⋅P_max are used. Figure 13 illustrates the result. It is evident that the relative pore volume indicates a similar progression of all three investigated intensity distributions. For welding speeds v  ≤  400 mm s⁻¹ and v  ≥  700 mm s⁻¹, the scattering of three repetitions per intensity profile is higher than that for welding speeds in the range of 500 mm s⁻¹ ≤ v ≤ 600 mm s⁻¹. In addition, with 500 mm s^−-1  ≤  v  ≤  600 mm s⁻¹, the relative pore volume is the lowest [V_pore(%) < 0.5%]. Furthermore, the intensity profiles influence the relative pore volume. The mean value of three repetitions for each intensity profile (Fig. 13, dashed lines) shows that for an intensity distribution P_C = 40%⋅P_max, the relative pore volume is the lowest over all considered welding speeds.

From the CT measurements, it can be observed that the pores are located particularly in the root of the welding seam. As described above (Fig. 9), this is also the location where the pores form when the keyhole passes the transition area. This formation of process pores was observed especially at the end of N3, which is at the end of the welding process. It is assumed that the saxophone-shape (especially for P_C = 100% ⋅ P_max) of the keyhole inflates the keyhole at the bottom due to multiple absorption of the laser beam. The transition area is an irregularity, and thus, the inflated keyhole strips the pore at the gap. Since this happens mostly at the end of the welding process, the pore has no time to move up and outgas before solidification begins. For slower welding speeds, this process occurs even easier because of higher temperature gradients around the keyhole. These process pores have a comparatively high pore volume. Figure 14 shows the location of pores for welding speeds 200 mm s⁻¹ and 500 mm s⁻¹ for P_C = 100% ⋅ P_max and 70% ⋅ P_max, respectively.

In summary, the use of static beam shaping can reduce the pore volume by opening the keyhole at the surface of the samples. Because the vapor can flow out easier, the pressure in the keyhole remains stable and the risk of pore formation when the keyhole is crossing the pin ends is reduced. In particular, a lower scatter of porosity compared to single spot welding can be expected when using static beam distribution. However, the welding speed for all beam distributions should not be chosen under 400 mm s⁻¹. Consequently, a high laser power is required to realize the combination of high welding speeds and high percentages of the ring power.

In the third investigation, the influence of a protective gas on porosity in comparison to no protective gas will be considered. By shielding the melt pool from the ambient atmosphere, a lower absorption of ambient gases (especially oxygen) is expected.^8,10 For this, a nozzle is utilized in the experimental setup (Fig. 2). The inert gases argon 5.0 and helium 5.0 are used for this purpose (nozzle diameter: 15 mm, distance nozzle to hairpin: 35 mm, flow rate: 20 l h⁻¹). Ten samples each are welded and plotted as a boxplot in Fig. 15. On average, the relative pore volume is reduced by >50% for the usage of both protective gases. Helium exhibits a significantly lower scatter in the results.

However, the experiment shows that the influence of protective gas [V_pore(%) range 0 … 1.8%] on the pore volume has a comparably smaller role than the optimization of the parameter set [V_pore(%) range 0 … 4.0%] with static beam shaping and welding speed. The reason for this is that process pores have high volumes and are mainly influenced by the shown process parameters. For reducing metallurgical pores when using Cu-ETP, a protective gas as shown in this experiment is effective. Unfortunately, these metallurgical pores are barely visible in the presented setup of x-ray imaging. Higher resolutions and higher frequencies are required to trace these pores.

The laser beam welding of hairpin windings is influenced by numerous factors. High connection areas and low pore volumes are the goals for welding results. The purpose of the present experiments is to investigate the pore formation during hairpin welding and to analyze the effects of static beam shaping in combination with welding speed. In addition, the use of protective gas was considered. In summary, the following results can be noted:

• In the welding process presented, the keyhole increases in welding depth in the form of a helix. Specifically, when moving over the two pin ends, process pores can detach from the keyhole bottom.

• The use of static beam shaping reduces pore volumes, because of the widening of the keyhole opening. The optimum intensity distribution for copper hairpins is between 40% and 70% power in the core. A reduction of approximately 60% of the relative pore volume was realized compared to single spot welding with 100% power in the core.

• The optimum welding speed for the presented welding geometry was above 400 mm s⁻¹. It must be taken into account that high percentages of power in the ring in combination with high welding speeds require a high total power of the laser.

• Protective gas can help to reduce the metallurgical pore volume. However, this is cost intensive in production. In the experiments presented, a reduction in the relative pore volume of approximately 50% could be achieved for Cu-ETP.

For future studies, higher resolutions at higher image frequencies are desirable. This can be achieved by synchrotron radiation that provides higher energy levels. With that also, the solidus line and smaller pores are detectable and traceable.

The authors have no conflicts to disclose.

Markus Omlor: Conceptualization (lead); Data curation (equal); Investigation (lead); Methodology (equal); Project administration (equal); Visualization (equal); Writing – original draft (lead); Writing – review & editing (equal). Eveline Nicole Reinheimer: Conceptualization (supporting); Methodology (equal); Writing – original draft (supporting); Writing – review & editing (equal). Tom Butzmann: Conceptualization (supporting); Data curation (equal); Investigation (supporting); Visualization (equal); Writing – review & editing (supporting). Klaus Dilger: Supervision (equal); Writing – review & editing (supporting).