Advanced welding analysis methods applied to heavy section welding with a 15 kW fibre laser

A. F. H. Kaplan and G. Wiklund Luleå University of Technology, Department of Applied Physics and Mechanical Engineering, Luleå, SE 971 87, Sweden

E-mail: alexander.kaplan@ltu.se; Web-site: www.ltu.se/tfm/produktion

Abstract For butt joint welding of 8 mm and of 16 mm thick stainless steel a 15 kW fibre laser was applied, achieving full penetration for a welding speed of 7 m/min and 3 m/min, respectively. Optics with a focal length ranging from 500 mm to 150 mm focus the laser beam to a spot diameter of 0.65 to 0.2 mm with a focal depth of 18 to 4 mm and a cw-power density of 4 to 48 MW/cm2. Although narrow sound welds can be achieved, heavy spatter ejection along with underfill can take place at the top and root side, particularly for high power density and for low line energy. Various advanced analysis methods were developed. Spatter was observed by high speed imaging of the weld pool surface and of the keyhole. Quantitative evaluation of the drop size and flight trajectories enabled categorisation into four spatter modes, revealing favoured conditions for spatter suppression. Modelling subsequent to the drop evaluation provided additional analysis. An illustrative theoretical description and the formulation of a standard documentation methodology were developed. This aims to generalise and transfer knowledge as a guideline for spatter suppression in other laser welding situations.

Introduction Laser beam welding is often characterized by deep narrow welds due to the highly focused laser beam with high power density. Such beam enables to drill a vapour capillary, the keyhole, in the material where the energy can be distributed over depth - in contrast to any other welding methods (except the electron beam). The achievable depth depends on the power and focusability of the laser beam.

For decades almost all high power lasers installed were of the CO2-laser type (60 % of the annual unit sales) or of the Nd:YAG-laser type (40%) [1]. Advantages of the CO2-laser are, beside its lower price, achievable power up to 20-50 kW and high beam quality, nowadays even for high power. The lamp-pumped Nd:YAG-laser is commercially available merely up to 4 kW (although prototypes up to 12 kW exist), but its short wavelength λ (1.06 µm compared to 10.6 µm) bears several advantages. The beam can be guided by an optical fibre, hardly any interaction (plasma shielding) with the metal vapour or shielding gas takes

place, the absorption is different and the focusing potential is accordingly higher (factor 10). However, in practice the Nd:YAG beam quality is poor. Thus each choice of laser requires a compromise.

Although the high power diode laser, emerging in the 90’s, still focuses unsatisfactorily for deep welding, it has been useful for pumping other laser types. For the DPSSL (diode pumped solid state laser), the disc laser (both Nd:YAG) and the Yb:fibre laser it enables much higher beam quality, thus better focusing in the 1 µm-range. The compromise between power and beam quality is less restrictive, approaching the properties of electron beams.

From 2002 the fibre laser has linearly conquered market shares (from the Nd:YAG laser) up to actually more than 20% (of total). Several hundred fibre lasers in the kW-range are installed worldwide. Basically fibre lasers up to 30-50 kW can be purchased, with good beam quality, up to 3 kW even close to the diffraction limit. The fibre laser has the potential for much deeper welds and much higher welding speed, as providing both, high power and high beam quality. However, the extreme power densities and geometrical process zone properties can cause quality problems [2]-[6].

The present study focuses on welding of heavy section stainless steel with a 15 kW fibre laser, see also [6]. Only BAM Berlin (20 kW), BIAS Bremen (17 kW) [3] and the laser manufacturer IPG itself have (30 kW) so far used higher laser welding power in the 1 µm wavelength range (i.e. except the CO2-laser). Improved and new methods will be applied for advanced research analysis of the process.

Experimental set-up Stainless steel AISI 304 (CrNi 1810) of 8 mm and 16 mm thickness is welded - mainly bead-on-plate to simulate butt welds with zero gap for tube manufacturing [6]. A 15 kW Yb:fibre laser (IPG Laser, YLR-15000, cw/continuous wave, diode-pumped) is used. The characteristic of the focused beam is described below. The processing head was tilted 7° ahead to reduce the risk of backscattering damage of the fibre. Ar is applied as the top and root shielding gas. The parameters varied are the beam power P, the welding speed v, the focal plane position z0 and the focal length f of the optics. Corresponding process indicators are the line energy E = P/v and the power density at the surface

I(z=0) = P/w2π for a beam radius w=f(z-z0). To achieve manifold information from a single weld pass, during the welds the focal plane position z0 was usually varied from –10 mm to +10 mm, thus one weld pass represents all focal plane positions and skips the thermal focus shift problem. The weld pool was observed by high speed imaging with a Redlake HS-X3 camera at 2000 or 4000 frames per second, filtered for the illumination wavelength (diode laser, 808 nm). For the images shown it has to be considered that the camera was inclined 51° from the side. The (horizontal) resolution is 15.2 µm per pixel.

Methodological approach

The methodological approach is as follows: First the focused beam, i.e. the tool is characterized. A

15 kW fibre laser beam is unusual. Its focusing characteristic is essential for the process. The welds are carried out in a systematic manner for 8 and 16 mm thickness and are then analysed and compared. The variation of the focal plane position z0 along the 100 mm weld reduces the number of experiments significantly. Visual analysis of the top and root surface appearance of the weld is carried out as well as microscopic analysis of the etched weld cross sections. From visual analysis parameter trends are identified, in particular by graphically studying parameter combinations like power density or line energy to accomplish generally valid criteria.

The main quality defects studied were lack of penetration and underfill (from spatter). High speed imaging of the weld pool and keyhole is a very powerful method, so far studied for fibre laser welding up to 10 kW [5]. Although even the qualitative information from high speed imaging is highly interesting, in the present study we try to quantitatively evaluate the images (which was rarely done before) in order to obtain additional and rational information. In earlier studies we experienced that mathematical modelling starting from high speed imaging facts is a very powerful, complementary tool as it keeps the level of uncertainty low (in contrast to pure modelling) and provides additional information not accessible from experiments. E.g. the keyhole collapse was derived from the measured (X-ray) switch-off shape during pulsed laser welding or photodiode monitoring of the pool emissions has been analysed from measured weld surface motions [7]. We regard generalisation of discovered knowledge as essential, but widely missing as not being obvious. As a first approach we introduced the Bifurcation Flow Chart, BFC, where the essential findings of how to suppress defects can be inserted in a standardised qualitative flow chart, capable to be combined with findings from very different cases. The BFC (described in detail in [7], [8]) is applied to the present study by identifying general trends, criteria, categories and then documenting them by the BFC.

Laser beam characterisation

The 15 kW Yb:fibre laser beam (cw, plug efficiency 30%, λ = 1070 nm) is delivered by a 30 m long optical fibre with a core diameter of 200 µm. The Beam Parameter Product BPP = 10.4 mm·mrad corresponds to a beam quality K = 0.033 (K = 1.0 would be the ideal, diffraction limited beam). The collimator length is 150 mm. Five different focal lengths f can be chosen, related to the collimator length thus providing five different magnifications M (of the fibre diameter) and five combinations beam diameter vs. Rayleigh length (zR, focal depth), as shown in Fig. 1(a).

(b)

Figure 1: Focused 15 kW fibre laser beam (BPP = 10.4): (a) Rayleigh length as a function of the spot diameter for

the five different optics (focal length f, Magnification ratio M) available and compared to an ideal diffraction-limited beam (K = 1.0), (b) focused beam radius w, (c) average power density I along the beam propagation direction z.

(a)

As the workpiece thickness studied was d = 8 mm and 16 mm, the longest focal lengths, 300 mm and 500 mm, and thus Rayleigh lengths zR, ±4 mm and ±10 mm, were chosen so far. The beam radius w(z) and the average power density I(z) = P/w2π along the beam propagation direction z, with z0 = 0 defined as the focal plane, are plotted in Fig. 1(b),(c).

The focused beam as the tool is basically characterised by P, I, w, zR (and z0). The compromise of properties is excellent for the here studied laser. In fact the properties are better than needed for the here applied thicknesses. The power and focusability of the laser beam is clearly superior to lamp-pumped Nd:YAG-lasers and similar to CO2-lasers (but providing the above mentioned advantages; in the meanwhile even superior fibre lasers are available).

Problems experienced for the new generation fibre (and disc) lasers are (i) sensitivity to pollutions of the optics, resulting in beam scattering and absorption, degrading the beam quality and damaging the optics, (ii) backreflections from the workpiece damaging the fibre end, and (iii) thermal shift of the focal plane [9]. The latter was identified here to be less than 2 mm and was overcome by linear variation of a larger focus range, from z0 = −10...+10 mm.

Welding results

For 8 mm plate thickness, P, v and z0 were varied and f=300 mm (more details can be found in [6]) was compared to f=500 mm. More than 100 parameter cases were studied, some of them by varying z0 linearly or v stepwise. For 16 mm thickness only a few welds were conducted so far, varying v and f. Typical weld seam cross sections are shown in Fig. 2. Fig. 2(a)-(c) shows how increasing speed narrows the weld but leads to underfill. Fig. 2(d)-(f) shows for 15 kW the sensitivity to the focal plane position, either causing lack of penetration or underfill at the top or root. For 15 kW and 7 m/min the process window for sound welds was very narrow, while for 7.5 kW and 3 m/min (thus similar line energy) the process was very robust.

Welding of 16 mm thick plates can sag through, see Fig. 2(g), when the line energy is too high or lack penetration, Fig. 2(h), for improper focusing. Welds of good quality, Fig. 2(i), were achieved at 2.75 m/min with the 300 mm optics. Thus even deeper welds seem possible.

Process window analysis

In order to try to identify more general criteria and guidelines, different parameter combinations were studied. Figure 3 shows the top and root surface appearance of two welds where the focal plane z0 was linearly varied from -10 mm to +10 mm from left to right (more details in [6]). The quality along the weld (i.e. corresponding to different focal planes) was classified, showing the much more robust processing range (with respect to z0) for 7.5 kW.

(a) (b) (c) (d) (e) (f)

(g) (h) (i)

Figure 2: Cross-section of typical welds; Thickness 8 mm: (a) P = 7.5 kW / v = 1 m/min / z0 = -1.5 mm, (b) 7.5/3/0, (c)

10/5/-1, (d) 15/7/-7, (e) 15/7/-4, (f) 15/7/-0.5 (all: f = 300 mm); Thickness 16 mm: (g) v = 0.9 m/min, f = 500 mm,

(h) 1.75/500, (i) 2.75/300 mm.

Beside the average power density of the beam at the surface

( )⎥⎥⎦

⎤

⎢⎢⎣

⎡

⎟⎟

⎠

⎞

⎜⎜

⎝

⎛⎟⎟⎠

⎞⎜⎜⎝

⎛+== π

2

020 10

RzzwPzI , (1)

the line energy

vPE =' (2)

and variations of the line energy were studied, in particular

( )vdzwPE0

''' = . (3)

Figure 3: Top and root surface appearance (z0 varied) for (a) 15 kW/7m/min and (b) 7.5 kW/3 m/min, identified quality ranges (+/-; Full/Partial Penetration), range of

modes, selected imaging interval ( ) for the modes studied Various graphs of combinations of two properties are shown in Fig. 4, trying to identify clear threshold criteria between full and partial penetration. While from the direct parameters power vs. speed, Fig. 4(a), a minimum power of 5 kW was obtained (for 8 mm), speed and focal plane position require more sophisticated analysis. Although E’’’ > 30 J/m3 was a promising threshold criteria for penetration for a series of seven welds [6], this criterion remained only applicable for sufficiently high power density I > 20 MW/cm2 for the whole set of experiments, Fig. 4(d). Further criteria (broken lines) are plotted in Fig. 4, to some extent being boundaries of reliable domains.

High speed imaging of the melt pool/keyhole

From high speed imaging a qualitative analysis of the melt pool and keyhole behaviour is possible. In particular, four different modes of spatter ejection were distinguished, see Fig. 5 and correspondingly Fig. 3. For two of them a spatter ejection sequence is shown in Fig. 6. From the high speed

images geometrical properties can be quantitatively evaluated. Figure 7 shows clear trends of these properties, depending on the spatter mode. For 7.5 kW a very robust process is achieved (Mode IV) where only small drops are occasionally ejected, while for 15 kW different modes with frequent ejection of heavy drops at different flight angle take place, as also quantified in Table 1. From the imaging evaluation (see also [6]), mathematical modelling enables further analysis, e.g. the derivation of the kinetic energy of the drops that turns out to be lower than the energy for creating new surfaces (surface tension/curvature), Fig. 7(c).

Figure 4: Graphical process window analysis for identifying full penetration limits (broken lines): (a) laser

power vs. welding speed; (b) speed, (c) line energy E’ and (d) energy property E’’’, all vs. surface power density

I0(z=0); (all for d = 8 mm, except three circles: d = 16 mm).

(a)

(b)

Figure 5: HS image of the four types of spatter identified: (a) Mode I (15 kW, z0<0), (b) Mode II (15 kW, z0 = 0), (c) Mode III (15 kW,z0>0), (d) Mode IV (7.5 kW,z0=0).

Mode II Mode IV

Figure 6: High speed images of a typical spatter sequence (1 ms-steps): (a) Mode II (15 kW), (b) Mode IV (7.5 kW).

TABLE 1: AVERAGE DROP PROPERTIES FOR

THE FOUR SPATTER MODES

Mode Frequenc

y [Hz]

Volume [mm3/ms]

Speed* [m/s]

Angle* [°]

Energy [10-8J/s]

I 424 105 1.63 48 4.7

II 737 388 1.45 61 41.5

III 1000 427 1.62 71 41.0

IV 196 5 3.05 4 1.7* average per drop, otherwise accumulated per second

0

2

4

6

0.0 0.1 0.2 0.3 0.4 0.5 0.6Drop volume V [mm3]

Dro

p sp

eed

U [m

/s]

Mode I, 15 kW, z0 < 0Mode II, 15 kW, z0 = 0Mode III, 15 kW, z0 > 0Mode IV, 7.5 kW, z0 = 0

1.E-11

1.E-10

1.E-09

1.E-08

0.0 0.1 0.2 0.3 0.4 0.5 0.6Drop volume V [mm3]

Ener

gy [J

]

Mode IMode IIMode IIIMode IVSurf. tens.

Figure 7: Evaluation of ejected drops as a function of the measured drop volume for the four modes: (a) drop speed, (b) travelling angle, (c) kinetic and surface tension energy

of the drop. High speed imaging for welding 16 mm thick plates is shown in Fig. 8. Occasionally (sometimes periodically) a substantial melt volume builds up at the top. Due to its height of 0.5-1 mm the keyhole formation starts higher up.

Figure 8: High speed images (250 µs-steps) of welding 16 mm thick steel; a melt zone emerges around the keyhole.

(>1)

(>1)

(a)

(b)

(c)

Theoretical description

When magnifying the high speed images from Fig. 5, it can be seen that the (dark) melt domains ahead of and beside the keyhole are very thin for 15 kW/7 m/min, Fig. 9(a), but significantly wider for 7.5 kW/3 m/min, Fig. 9(b). We state the hypothesis that a thicker melt film facilitates to relax the momentum continuously generated by evaporation through the laser beam at the keyhole wall, thus larger geometrical melt environments facilitate the reduction of spatter. The theory is correspondingly illustrated in Fig. 10. Moreover, a Bifurcation Flow Chart (BFC) was developed, see Fig. 11, to document the observed phenomena - and the criteria and conditions of the theory. These approaches aim to facilitate knowledge transfer and to stimulate scientific discussions.

(a) (b)

Figure 9: High speed images of the melt channel (dark) ahead and beside the keyhole: (a) Mode I, (b) Mode IV.

Figure 10: Illustrated theoretical description (geometry) of boiling and momentum flow as the origin for accelerating

the melt to a column before necking and drop ejection.

Figure 11: Bifurcation Flow Chart (BFC) qualitatively describing the essential origins of spatter / underfill.

Conclusions

The 15 kW/200 µm fibre laser provides high power and power density and a deep focus, suitable for deep welds

8 mm thick stainless steel was welded with 7 m/min, 16 mm with 2.75 m/min, but the process window is narrow

The Rayleigh length and the beam diameter are critical Important criteria are power density and line energy;

from according graphs general limits can be derived For power densities of the order of 10 MW/cm2 spatter

and thus underfill is likely to occur at the top and root From high speed imaging different spatter modes can be

identified; their quantitative properties can be derived Less spatter occurred for wider melt beside the keyhole With a BFC, standardised quantitative documentation,

generalisation and combination of findings is facilitated The narrow process windows demand further studies Advanced methods enable more powerful analysis

Acknowledgements

We acknowledge support from VINNOVA/Jernkontoret, Fibertube project, no. 34012, from Outokumpu Stainless ARC as the main industrial partner, from the Knut and Alice Wallenberg Foundation, project no. KAW 2007-0119, from VINNOVA, The Swedish Innovation Agency, project no. 2006-00668, and from the Kempe foundation.

References

[1] D. A. Belforte, “Economic review and forecast: Keeping the economy,” Industrial Laser Solutions, pp. 4-11, January 2009.

[2] I. Miyamoto, S. Park, and T. Ooie, “Ultrafine keyhole welding processes using single-mode fiber lasers,” in Proc. LMP, Part A, 2003, pp. 203-212.

[3] F. Vollertsen, and C. Thomy, “Welding with fiber lasers from 200 to 17000 W,” in Proc. ICALEO, Oct. 2005, Jacksonville, FL, Orlando: LIA, 2005, pp. 254-263.

[4] X. Jin, P. Berger, and Th. Graf, “Multiple reflections and Fresnel absorption in an actual 3D keyhole during deep penetration laser welding,” J. Phys. D: Appl. Phys., vol. 39, pp 4703-4712, 2006.

[5] Y. Kawahito, M. Mizutani, and S. Katayama, “Elucidation of high-power fiber laser welding phenomena of stainless steel and effect of factors on weld geometry,” J. Phys. D: Appl. Phys., vol. 40, pp. 5854-5859, 2007.

[6] A. F. H. Kaplan, E. M. Westin, G. Wiklund, and P. Norman, “ Imaging in cooperation with modelling of selected defect mechanisms during fibre laser welding of stainless steel,” Proc. ICALEO, Oct. 2008, Temecula, CA, Orlando: LIA, 2008.

[7] P. Norman, H. Engström, and A. F. H. Kaplan, “Theoretical analysis of photodiode monitoring of laser welding defects by imaging combined with modelling,” J. Phys. D: Appl. Phys., vol. 41, p. 195502 (e-9p), 2008.

[8] A. F. H. Kaplan, P. Norman, and G. Wiklund, “The bifurcation flow chart as a method for a generalised theory, applied to spatter defects during laser welding,” in Proc. 12th NOLAMP, August 2009, Copenhague, Denmark: FORCE Technology, 2009, in press.

[9] F. Abt, and F. Dausinger, “Focusing of high power single mode laser beams,” in Proc. LIM 4, Munich, D: WLT, June 2007, pp. 321-324.