Adaptive Control Hybrid Welding with a High Brightness Laser

Paper presented at International conference on power beam processing technologies (ICPBPT2010) Oct. 25-29, 2010 Beijing, China.

Abstract

Hybrid laser-MAG welding with real-time adaptive control has been demonstrated. Hybrid welding conditions have been developed to produce ISO 13919-1 class B (stringent) quality butt welds between 8mm thickness steel plates, using a 5kW 6mm.mrad Yb fibre laser combined with conventional arc welding equipment. A laser vision seam tracking system has enabled a 7-axis robot, manipulating the hybrid welding head, to track butt joints in real time during welding. The laser vision system has also provided information on the variations in joint fit-up, in particular the width of any gaps, or height of any mismatches present. This information has then been used to automatically adjust the robot position or speed, or arc welding parameters, depending on the fit-up detected. In this way, the tolerance of the hybrid process has been augmented, producing stringent quality welds over a wider range of joint fit-up cases than when using fixed conditions alone.

Introduction

Hybrid welding can offer significant advantages over other welding processes, including high welding speeds, consistent weld quality and reduced distortion. Nevertheless, guaranteeing productivity and quality targets can be a challenge. Requirements on part positioning, edge preparation and fit-up are more stringent than those for arc welding. These can augment preparation and welding fixtures costs.

Joint tracking systems increase the robustness of hybrid welding to joint placement, and are already used in some industrial applications. Vision-based tracking sensors can also relay details of joint fit-up in real time to the welding equipment, enabling adaptive control of the process. Parameters can be adjusted with changes in fit up, maintaining and guaranteeing weld quality, and relaxing demands on preparation and fixturing.

Adaptively controlled hybrid welding has already been described, using moderate to high power CO₂ lasers ^[1-5] or lower power Nd:YAG lasers ^[6,7], as opposed to the latest high power fibre-delivered lasers now available. The current work has developed and demonstrated adaptive control of robotic hybrid welding of butt joints in structural steels using a high brightness (6mm.mrad) 5kW Yb fibre laser. The results can be applied to high speed, high quality, hybrid welding using these new lasers, for example in shipbuilding, steel fabrication or pipe welding.

Experimental

An IPG YLS-5000 Yb fibre laser was used in welding trials, with a calculated focused spot diameter of 0.28mm. The beam focusing optics were mounted on a Kawasaki FS-060L robot. The optics were protected from fume and spatter using an air-knife and an appropriate cover slide. All trials were performed in the flat (PA, 1G) position, with the laser beam at 90° to the workpiece surface. A measured laser power of 4.8kW at the workpiece was used. When hybrid welding, an ESAB PSF 410MW metal active gas (MAG) welding torch was also used, with an AristoFeed 30 wire feeding unit, U8 controller and AristoMIG 450 arc power source. All welds were made using a laser-leading configuration. The separation between the laser and the arc was set to 2mm. A contact tip to workpiece distance of 15mm was used, with a work angle of 60° (pushing) for the MAG torch. A synergic pulsed metal transfer mode was also used, with a 1.2mm diameter A18 grade solid wire consumable. The weld top bead was shielded by 20l/min of argon through the torch. A small slot was ground in to the shroud of the torch, to allow the laser beam to pass without being clipped by the shroud. The weld under bead was shielded by 2l/min of argon underneath the joint line. All trials were carried out on S355 grade structural steel plates, 8mm in thickness, which had been laser cut to size before welding, acetone degreased, disc ground back to bright metal, and then re-degreased.

Experiments consisted of:

• Gas shielded autogenous melt run and butt welding trials, at 0.9-2.2m/min, using laser focus positions between -4mm (4mm below plate surface) and +4mm. These trials determined a suitable joint preparation and welding speed for the hybrid trials that followed.
• Determining hybrid conditions producing ISO 13919-1:1997 Class B welds (the most stringent weld quality class in this standard) over flush, close fitting joints.
• Gauging the tolerance of these hybrid conditions to joint mismatch or joint gap.
• Developing other conditions, producing class B welds over joints with greater amounts of gap or mismatch.
• Using all the conditions as control points in adaptively controlled trials. A Servo-Robot Digi-I/S laser vision sensor with V300 control unit, communicating with a Kawasaki D+ robot controller was used. Welds were made over joints with either varying amounts of joint gap or joint mismatch, with welding conditions being changed in real time.

A tracking accuracy of <0.2~0.3mm was required, given the narrowness of the weld roots made in close fitting joints. This was achieved by calibrating the tracking sensor position with respect to the robot tool centre point, sighting both in turn on a small diameter hole drilled in a calibration block.

Changes in welding (robot) speed and robot position were effected by relaying joint fit-up information from the V300 unit to the D+ controller, then referencing a look-up table written within the D+ controller software. For example, this table may call for a reduction in robot speed and the introduction of an offset of the robot tool centre point off of the joint line, when a joint mismatch over a given height was detected. Conversely, changes in wire feed rate were effected using an ADAP software module in the V300 unit. A program written within this module, with support from Servo-Robot, generated an analogue output voltage from the V300 unit. The voltage value was made to scale with joint gap width. These values had been calibrated to correspond to an appropriate range of wire feed rates. The voltage signal was sent, via an ESAB RA23 remote adaptor, to the arc welding power source. This was just one way of changing the wire feed rate in real time as a function of the joint gap detected. A suitable time delay between the generation of the signal and its transmission to the arc welding power source was used, to allow the robot tool centre point to catch up with the point at which the laser vision sensor had collected the information on joint gap.

Results

Without plume dispersion, stable full penetration was not achieved in 8mm thickness plate in the autogenous melt run trials. Penetration was unstable, producing a 'stitching' weld root. This indicated periodic, uncontrolled, collapses of the laser welding keyhole. Autogenous trials were also carried out using a broad root face V butt joint preparation, with a 6mm root face and 60° included angle. Stable, full penetration was achieved, at 1.3-2.2m/min. The most visually acceptable weld roots profiles were achieved using the laser focus positioned 2-4mm above the top of the root face.

Hybrid trials followed on close fitting, flush plates, using the same V joint preparation and laser focus position. A steady arc and a visually acceptable weld top bead appearance was achieved, with slightly convex profile and minimal undercut, using a wire feed rate of 7m/min and a -3V arc voltage trim. Radiography indicated internal porosity contents to Class B. Fig.1 shows a cross-section of a weld made with these conditions. This weld had a cap undercut of 0.16mm and an excess penetration of 0.5mm, ie to class B.

These conditions were also used to weld joints with up to a 2mm high mismatch. Figs.2a and 2b show sections from such welds with 0.6 and 1.0mm mismatches. In particular, these figures show that a Class B root and cap profile were achieved to a mismatch of 0.6mm (Fig.2a), with the root toe blend angle becoming re-entrant on the higher plate when the mismatch increased to 1.0mm (Fig.2b).

At still higher mismatch values, the weld root was both re-entrant and also undercut. In terms of internal quality, the radiographs of these welds indicated that class B quality was achieved to mismatch values of at least 0.6mm, with chain or clustered porosity and lack of penetration defects present at higher values of mismatch. Thus, the conditions suitable for welding flush, close fitting joints, could be used to produce Class B welds for a mismatch up to, at most, 0.6mm.

These same conditions were also used to weld joints with gaps. In terms of the weld profile, Figs.3a and 3b show cross-sections at nominal gap widths of 0.3 and 0.6mm width, respectively. In this weld a small amount (0.3mm) of mismatch has been introduced inadvertently. Nevertheless, as Fig.3a shows, the depth of underfill is 0.26mm at a nominal gap width of 0.3mm (ie to Class B), becoming 0.56mm deep at a nominal gap width of 0.6mm (outside of class B requirements). The root profile was acceptable up to a gap width of at least 1.2mm. Only one 0.7mm diameter sidewall pore was detected in the radiograph of this weld, ie the internal quality was also to Class B. The results suggested that the base conditions could also be used to weld joints with gaps up to, at most, 0.3mm wide.

A range of other welding conditions was examined, to improve the weld root profile achieved over a joint mismatch >0.6mm in height. One or more of the following changes was introduced:

• A reduction in welding speed.
• Deliberately offsetting the laser beam off of the joint line (by offsetting the welding head).
• Using different laser focus positions.

An acceptable root toe blend angle and internal weld quality was achieved up to a value of 0.5mm mismatch, when welding at 1.2m/min. This was similar to the result achieved at 1.6m/min. Sections through a weld made at 1.6m/min, but with the laser beam offset 0.5mm on to the lower plate, showed that an acceptable root blend was achieved to a mismatch value of at least 0.6mm, ie again comparable with the results achieved with the laser aligned on the joint. Nevertheless, sections taken of a weld made at 1.2m/min with the laser offset 0.5mm, all contained lack of sidewall fusion defects and sidewall porosity. This suggested that a lack of internal defects could not be guaranteed when offsetting the beam. Offsetting by either 0.5mm or 1mm on to the higher plate also introduced internal defects, although a 1mm offset led to a slight improvement in the weld root toe blend angle for mismatch values up to ~0.7-1.0mm. The sensitivity of internal defects to offset has been noted previously. ^[4]

 

Focusing the laser 2 or 4mm above the top of the root face, or 2mm below it, resulted in a very slight improvement, in terms of weld cap and root profile and internal quality. Acceptable weld root profiles were achieved to a mismatch value of ~0.8mm, as shown in Fig.4. Focusing the laser directly on top of the root face did not lead to any improvements.

Two welds were also made using the laser focused either +4mm above the root face or on top of the root face, with the welding speed reduced to 1.2m/min and the laser aligned 1mm from the joint line on to the higher plate. However, these conditions resulted in inconsistent penetration in both cases.

Taken all together, these results suggested that an adaptive change in laser focus position with joint mismatch, would be the best strategy to be tested adaptively, for a modest (<25%) improvement in mismatch tolerance, with any deliberate offsets being kept to <0.5mm.

A second range of welding conditions was examined to increase the tolerance to joint gap. One or more of the following changes was introduced:

• Increasing the wire feed rate and/or arc voltage trim.
• Reducing the welding speed.
• Focusing the laser 4mm above the top of the root face.

At 1.6m/min, joint gap bridging was improved by increasing the wire feed rate to 11m/min, but class B weld cap profiles were not achieved at gap widths >0.5mm. Using a wire feed rate of 13m/min, or 11m/min but with 0V trim, resulted in unstable top bead profiles. Better results were achieved reducing the welding speed to 1.2m/min. The depth of underfill over a nominal gap width of 0.5mm reduced to only 0.2mm. Welding at <1.2m/min generally resulted in unacceptable amounts of top bead underfill and excessive penetration. Further improvements were made when welding at 1.2m/min with an increase in wire feed rate. For example, Fig.5 shows the weld profile over a nominal gap width of 0.5mm, when welding at 1.2m/min with a wire feed rate of 9m/min, with a cap undercut to class B, (c.f. Fig.3b).

Using a laser focus position 4mm above the top of the root face did not improve the weld profile, despite changes in welding speed or arc conditions.

In general, with two notable exceptions, the internal quality of the welds made over joint gaps in the range 0~1mm was to class B. Linear indications were detected at gap widths beyond which weld underfill or undercut would render the weld quality outside of Class B. In addition, porosity content, and mean pore size, appeared to increase with increasing arc energy. In particular, when using a wire feed rate of 13m/min, and/or when using a 0V trim, the internal quality was not to Class B.

Taken together, these results on gap dependence suggested that adaptive changes in both welding speed and wire feed rate would be the best strategy to be tested adaptively, for a possible ~200% improvement in gap tolerance.

Adaptive trials were first performed over joints with mismatch, changing welding speed, laser focus position and/or laser beam offset during welding. Unless deliberately offsetting the beam, it was assumed the beam always impinged on the joint line, as seam tracking was also used in these trials. Most of the welds made had an acceptable internal quality and weld cap profile, albeit with a re-entrant weld root profile at higher values of mismatch. Offsets >0.3mm resulted in unacceptable quality welds. The most successful result was achieved by adaptively reducing welding speed, to cope with joint mismatch. Figs.6a and 6b show the weld profiles achieved from the same weld, over joint mismatches of 1.0 and 1.2mm, by adaptive reduction of welding speed from 1.6 to 1.2mm. As Figs.6a and 6b show, an acceptable weld root profile has been achieved with this approach, to a mismatch of at least 1.0mm.

Adaptive trials were also performed over joints with a tapering joint gap, changing welding speed and/or wire feed rate during welding. The most successful result was achieved by adaptively reducing the welding speed and increasing the wire feed. Figs.7a and 7b show the weld profiles achieved at joint gaps of 0.6 and 0.9mm, from the same weld. As Figs.7a and 7b show, an acceptable weld cap profile is achieved to a gap width of at least 0.6mm, with this approach, twice that achieved without adaptive control. An ISO class C weld was achieved over a 0.9mm wide gap, but at gaps >1mm, weld cracks and lack of sidewall fusion defects were detected.

Finally, adaptive trials were carried out over joints with a combination of both mismatch and gap. Once again, the welding speed and wire feed rate were changed adaptively when a certain gap width threshold was exceeded, as this occurred prior to the mismatch threshold being exceeded.

As Figs 8a and 8b show, a class B weld could be made over a combined mismatch/gap of 0.65/0.5mm, with this approach, and a class C weld over 1.0/0.85mm, respectively.

Discussion

In the current work, the gap tolerance of the hybrid process with a high brightness Yb fibre laser without using adaptive control is 0.3mm, to produce class B welds over broad root face V butt joints in 8mm steel plate. This compares with reports in 8mm plate using similar output power CO₂ or Nd:YAG lasers of 0.2 ^[5] or 0.8mm^[7], respectively. This suggests that the low tolerance to gap of Yb fibre lasers when hybrid welding is comparable with CO₂ lasers, perhaps arising from the similarity in their focusability.

Gap tolerance has been increased in other work to 1.5-1.6mm for CO₂ laser-based hybrid welding^[4,7], by increasing the wire feed rate by ~55-60%, or to 1.2mm for Nd:YAG laser-based hybrid welding^[5], by reducing the welding speed by ~15%. In the current work, the maximum gap tolerance for class B welds has been doubled by using adaptive control. However, this maximum is still only 0.6mm. This more modest result can be attributed to the V butt joint configuration used, necessitating the use of high wire feed rates to achieve a weld profile with acceptably low levels of underfill. With an alternative choice of joint configuration, weldable with higher output power levels, a greater maximum gap tolerance for class B welds would be anticipated.

Tolerance to joint mismatch appears to have been less reported. Full fusion through a butt joint between 11.2mm thickness plates with a 1.4mm mismatch has been achieved using a 15kW fibre laser,^[8] but the weld profile quality class achieved was not reported. In the current work, class B welds have been achieved across mismatch values of up to at least 1mm, using adaptive control.

The individual tolerance limits to either gap or mismatch are reduced when both are present. The limits are 0.5 and 0.65mm for class B welds, respectively, compared with 0.6 and 1mm, if only gap, or mismatch, is present. Nevertheless, class C welds can be achieved over combinations of 0.85, and 1mm, respectively, with adaptive control.

Conclusions

The conclusions of the current work are:

1. The joint fit-up tolerances when hybrid welding using a high brightness, multi-mode, multi-kilowatt Yb fibre laser appear similar to those achieved using a CO₂ laser under the same conditions.
2. Hybrid welding using such a 5kW fibre laser can produce ISO 13919-1 class B butt welds in 8mm thickness steel when either joint gaps of 0.3mm width, or mismatches of 0.6mm height, are present.
3. With adaptive control of welding parameters, the tolerance limits for class B welds can be extended to 0.6 or 1mm, respectively.
4. Hybrid welding can also be adaptively controlled to cope with combinations of gap/mismatch, to 0.5/0.65mm (class B) or 0.85/1mm (class C).

Acknowledgement

The research leading to these results has received funding from the European Community's Seventh Framework Programme (FP7/2007-2013) under grant agreement no. 222289.

References

1. C. Bro and O. Madsen: Proc. 8th NOLAMP Conference, 13-15/8/01, Copenhagen, Publ. DK-2800 Lyngby, Denmark.
2. N.J. Orozco, P.A. Blomquist, R.B.Rudy and S.R.Webber: Paper 304, Proc. ICALEO 2003, 13-16/10/03, Jacksonville, Publ. Orlando, FL 32826, USA.
3. N.J. Orozco, P.A. Blomquist, R.B.Rudy and S.R.Webber: Paper 304, Proc. ICALEO 2004, 4-7/10/04, San Francisco, Publ. Orlando, FL 32826, USA.
4. H.S. Kim, Y.S. Lee, Y.S. Park, J.K. Kim and J.H. Shin: pp. 165-169 in Proc. LIM 2003, 24-26/6/03, Munich, Publ. D-70331 Stuttgart, Germany.
5. G. Shi and P. Hilton: Welding in the World, 2005, 49, pp. 75-87.
6. J.P. Boillot, J. Noruk and F. Arsenault: Fabricator, 2004, 34, 7, pp. 24-26.
7. S.G. Shi, P.A. Hilton, S.J. Mulligan and G. Verhaeghe: Welding and Cutting, 2005, 4, 6, pp. 345-350.
8. C. Thomy, T. Seefeld, F. Vollertsen, E. Vietz: Welding J., 2006, 85, 7, pp. 30-33.