C. H. J. Gerritsen, N. C. Sekhar, K. M. Egland*
TWI Ltd., Granta Park, Great Abington, Cambridge CB1 6AL, UK.
*Caterpillar Inc, Caterpillar Technical Center, Peoria IL 60656, USA.
Presented at Joining of Materials, JOM-10, Helsingør, Denmark, 11-14 May 2001
One of the main advantages of Nd:YAG laser sources is that the wavelength of the laser light (1.064µm) allows the beam to be guided via low-loss optical fibres. These flexible fibres can then be connected to robot-mounted beam focusing systems, providing three-dimensional processing capabilities. With higher power Nd:YAG lasers commercially available, multi-kilowatt laser beams can thus be used for 3D welding of thick-section steel.
This paper describes the laser welding of a 2200x700x300mm steel demonstration component, containing 2D and 3D joint configurations, using a robot-manipulated laser welding head. A single 4kW Nd:YAG laser source has been used with weld penetration ranging from 6.4 to 7.5mm. Additionally, a unique beam combining facility coupling the output from three 4kW Nd:YAG lasers into a single beam has been used for stake welds through the 8mm thick steel cover plate, resulting in a weld penetration of about 10mm. The welding procedures for the actual component are briefly described in the paper and the challenges for using 3D robotic laser welding in an industrial environment are discussed.
In industrial applications, products tend to be three-dimensional, containing different joint configurations (e.g. butt, fillet, stake) of different geometries (linear, planar or three-dimensional) in different orientations (e.g. downhand, horizontal, vertical). In this respect, one of the key features of Nd:YAG laser sources is the wavelength of the laser light (1.064µm), which allows the beam to be guided via flexible optical fibres, with relatively small energy loss. This is a great advantage when compared to CO 2 lasers, where the beam has to be guided via (moving) mirrors. Consequently, 3D robotic processing with CO 2 lasers is somewhat cumbersome, whereas with Nd:YAG lasers, the flexible optical fibre should make robotic processing relatively straightforward, enabling easier 3D processing.
The robotic welding capabilities of Nd:YAG lasers are already used in the automotive industry for, for example, the welding of thin sheet in body-in-white applications such as roof to side panel joining. However, the power levels of commercially available Nd:YAG lasers have been limited to about 4kW, making the robotic laser welding of thicker sections (>6mm for steel) difficult to apply. Nevertheless, the process is attractive for the yellow goods and shipbuilding industries, for example, because of the high joint completion rates and low thermal distortion on offer.
These considerations led TWI to develop a unique Nd:YAG facility, as part of a Group Sponsored Project, with a range of articulated-arm robots, three 4kW Nd:YAG lasers, and a beam combiner unit (BCU). The latter allows the output of three lasers to be optically combined into a single fibre, resulting in up to 9kW of laser power at the workpiece. A large demonstrator component was designed to investigate robotic 2D and 3D welding of thicker sections using these Nd:YAG laser sources. The component contained joints of various configurations, dimensions and orientations. The laser welding of this component and the challenges facing industrial robotic laser welding (particularly of butt joints) are described in this paper.
2. Demonstration component
A schematic of the demonstration component is shown in Fig.1
. The component comprised the following parts:-
- Two rectangular C-Mn steel plates of dimensions 2200x700x8mm thick. These were used as base and cover plate (a in Fig.1).
- Two C-Mn steel tubes (b) of height 284mm, internal diameter 305mm, each with a wall thickness of 6.35mm (1/4"). Each tube had been EDM wire cut into two pieces in a saddle joint configuration, and was to be welded back together around the 360° circumference of the tube. The cutting procedure provided a seam that varied from the normal position to up to 60° to the tube surface. Thus the required penetration over the circumference of the tube varied between 6.4 and 7.3mm.
- Two laser cut, curved, rectangular C-Mn steel plates (c) of 445x284x12mm thick.
Fig.1. Schematic of 2200x700x300mm steel demonstration component:
a) cover plates
c) curved rectangular plates
The surface mill scale was left on, whilst the oxide on the laser cut edges was removed. The joint areas were cleaned with acetone prior to welding. The welds to be made were of several types: butt welds in the PC (2G) position (saddle joints on tubes), T butt welds effectively in the PB (2F) position (welding of tubes and curved plates to base plate) and stake welds in the PA (1G) position (welding through cover plate into tubes and curved plates). All welds were single-sided, except for the first of the curved plates, which was welded from both sides.
Three beta-site CW Multiwave-Auto TM Nd:YAG lasers manufactured by GSI-Lumonics Ltd were used. For the T butt and butt joints, the output from a single laser source was used, providing 3kW at the workpiece. The beam was delivered to the welding torch using a step index optical fibre of 0.6mm in diameter. The optics used provided X1 imaging, resulting in a focused spot size of approximately 0.6mm in diameter. A coaxial nozzle with an orifice of 18mm in diameter was used for delivery of the shielding gas (helium).
For the stake welds through the cover plate, all three lasers were used simultaneously, their beams optically combined using a beam combiner unit (BCU). The combined beam was delivered to the welding torch using a 1mm diameter step index optical fibre. The laser power at the workpiece was approximately 8.2kW. In this instance, X1 focussing optics gave a focused spot size of 1.0mm diameter. Argon shielding was delivered through a coaxial nozzle with a 25mm diameter orifice.
3.2. Ancillary equipment
The welding torch in both cases was mounted on a 6-axis Kawasaki Heavy Industries JS-30 robot on a floor-mounted traverse. The robot was used to move the welding head along the seams, while the workpiece itself was kept stationary.
Clamping and fixturing were kept simple, mainly using plates, bars, threaded rods and welded nuts to hold down the components during welding. The tubes were internally tacked to the base plate using MMAW. The curved plates were laser tack welded to the base plate and one of the saddle welds was laser tacked as well. Due to their slightly elliptic shape after cutting, the saddle joints had a mismatch of approximately 0.75mm at the peaks and troughs. A screw-jack on the inside of the tubes was used to minimise the mismatch.
4. Procedure development
Prior to the actual execution, welding of the component was modelled using UltraArc from Delmia. Within this software, the cell, robot, demonstration component and fibre were modelled, and the kinematics simulated. This allowed issues such as accessibility, reach and collisions between robot and component to be evaluated, and the welding sequence and the best position for the component relative to the robot to be determined. This software was not, however, used for programming the robot. (Some of the difficulties that can influence the success of off-line programming are highlighted in section 6.1.) Instead, all welding runs were programmed manually, initially using a mechanical pointer mounted on the welding head, followed by the laser's diode pointer.
4.2. Weld procedure development
One spare of each of the different parts in the component was used for weld procedure development. The following parameters were varied: travel speed, beam-to-joint alignment and angle relative to the seam. Two typical weld cross sections produced during procedure development are shown in Fig.2 & Fig.3. The welding parameters chosen from this work to weld the actual component are listed in Table 1.
Fig.2. Cross section of double-sided T butt joint at non-optimised conditions, therefore showing only partial penetration. Scale in millimetres
Fig.3. Cross section of saddle weld, showing mismatch. Laser beam entry from the left. Scale in millimetres
Table 1: Parameters used for the different welds in the demonstration component
|Weld||Weld type||Weld position||Speed [m/min]||Focal position||Beam angle|
||At surface, on seam.
||0° to joint faces; i.e. at constantly changing angle to horizontal.
||At surface, ~0.5mm above seam.
||18° to horizontal.
||At surface, ~0.5mm above seam.
||18° to horizontal.
||90° to horizontal.
The completed component can be seen in Fig.4. Externally, all welds appeared sound, although in some places the top bead surface was not smooth, particularly where mismatch or gaps had been present. The first saddle joint, which was not tacked, was welded in two 190° runs. This resulted in an opening up of the second half of the joint due to local thermal expansion. Nevertheless, the welds showed acceptable penetration and fusion. As a first approximation, the distortion of the final component (measured across the cover plate) was found to be a maximum of 3mm between the extremities (2200mm). The distortion is, in fact, more likely to result from variations in height of the internal plates and tubes (difference up to 1.5mm) than from actual thermal distortion. No cross sections were taken from the component.
Fig.4. Picture of finished 2200x700x300mm demonstration component
6.1. Issues to be addressed in an industrial scenario
The welding of this component has highlighted many challenges that will be encountered when attempting robotic laser welding, particularly in the case of butt joints. In general, most of these issues arise from beam-to-joint alignment. Because the energy spot in laser welding is very small (generally <1mm in diameter), the beam-to-joint alignment needs to be precise, otherwise imperfections such as lack of penetration and lack of fusion may occur (as the joint may easily be missed). This especially applies to butt joints, where the seam has a square edge preparation, resulting in a very narrow 'target'. With stake welds, it is generally less of a problem.
Some of the factors affecting the beam-to-joint alignment are:-
- Joint configuration. Some joint configurations require better beam-to-joint alignment than others do. The alignment is most critical for square edge butt joints, and probably most tolerant in the case of stake welds. This should be taken into account in the design phase of a component.
- Part-to-part variation. Parts should be manufactured and machined to tight tolerances. This is particularly true in the case of autogenous laser butt welding, as only very small gaps can be tolerated. Consideration should be given to residual stress levels and patterns present in materials due to previous processing (e.g. rolling, tube welding, stamping) as these may give rise to part-to-part variation (cf. earlier mentioned mismatch in the saddle joints).
- Part placement and fixturing. The position and orientation of the parts relative to the robot should be accurate and repeatable. If the alignment is inconsistent, the programmed path will not follow the joint line.
- Dynamic accuracy. The dynamic accuracy of the robot should ensure proper alignment is kept when the robot is moving, even at high travel speeds. This is particularly important with laser welding, as generally better accuracy is required, and often at higher travel speeds than for many other processes.
- Off-line programming. If off-line programming is to be successful, the model should match the real component, working area and equipment in totality. This is particularly important when considering part placement, and time should be planned for calibration of the robot.
- Metal movement during welding. Local heating and weld thermal cycles can result in the movement of components during welding. Pre-weld tacking can resolve this to a certain degree, as may suitable clamping of the components.
Peripheral processing equipment: Basic laser welding has a very useful feature in that it is axi-symmetric. This term is used to describe processes that are symmetrical around the tool centre point (e.g. the focal point of the laser beam) and therefore direction independent. As a result of this, the robot used to manipulate the processing head can choose from some different configurations to follow a certain seam. Unfortunately this is not the case when peripheral equipment is used (e.g. wire feed nozzles, plasma control jets). The process then ceases to be axi-symmetric, and the tool centre point is no longer a point but a vector, which means the process is direction specific. This imposes significant constraints on the robot, especially if the joint geometry varies in all three dimensions (e.g. saddle joints). The same applies even without peripheral equipment, but to a lesser degree, when the laser beam is not perpendicular to the component surface. To overcome these constraints, it is often necessary to include additional degrees of freedom for the robot to maintain its flexibility and dexterity.
Accessibility: To not impair the accessibility, clamping should be kept as close to the component as possible, as anything that sticks out increases the likelihood of a collision with the robot arm, or entanglement of the fibre. Similarly, the processing head should be kept as small as possible, which is particularly important for welding in corner areas. Modelling should be used to determine the best robot orientation and welding sequence. Although the fibre optic cable is flexible to some extent, it cannot be twisted or bent over too small a radius. It is generally preferable to pre-wind the fibre in such a manner that it unwinds during the welding sequence.
6.2. Remedies and recommendations
Sensor technology: The problem of joint alignment is not exclusive to laser welding, but persists with all automated welding processes. Different methods of following the seam have already been developed over the years (for example tactile, eddy current and vision-based sensors) and these are also applicable to laser welding.
Currently, the vision-based systems are best suited for laser welding, because of their high signal update rate and resolution. Sensors are available which can track seams with a width of less than 0.1mm. There are different varieties of sensors, for example projecting a single stripe, multiple stripes or a circular projection. From the single stripe to the circle, the information that can be gained from monitoring its reflection increases. However, the use of vision-based sensors for seam tracking of non-linear joints is bound by some limits. Only curves above a certain radius and limited angular deviations can be followed, as the seam must remain inside the field of view of the sensor. The deviation that can be tracked therefore mainly depends on whether the welding head is rotated to follow the seam (and thereby keep it in view), as well as the field of view of the seam tracker and the 'look-ahead' distance. The field of view is defined as the length of the laser stripe, and the 'look-ahead' distance as the distance between the laser stripe of the seam tracker and the keyhole ( Fig.5). The larger the look-ahead distance, and the smaller the field of view, the smaller the allowable deviations become. This is particularly of importance for laser welding as the required resolution generally results in smaller fields of view.
Fig.5. Schematic of features affecting seam tracking (in top view)
A more difficult problem is the use of seams that are not perpendicular to the workpiece surface, as was the case in this demonstrator for the saddle welds. Although a seam tracker might be able to follow the seam, it will provide no information on the angle of the seam with the workpiece surface in the case of a square, closed butt preparation. Also, although the laser beam may be angled to the surface, the seam tracker ideally would remain perpendicular to it.
Hybrid laser-arc welding: Two of the major drawbacks of the laser welding process are the required accuracy in positioning of the laser with respect to the joint line, and the requirement for very good part fit-up. The advantages obtained with a hybrid laser-arc welding system, in terms of increased weld width, could be used as an alternative method. When using a laser-plasma hybrid, the two processes could even be made to be coaxial, retaining axi-symmetry.
Spot size: Comparably, the use of a larger spot, either through use of a different focusing optic, or an out-of-focus spot, can be used to relax the criteria of beam-to-joint alignment and, to a small extent, joint fit-up. However, the lower power density that results will give a slower travel speed and/or shallower penetration.
Self-locating jigging: A method of reducing the effort required for accurate part placement is to use joint designs that are amenable to self-jigging. This can reduce the assembly and improve accuracy and repeatability.
Simultaneous welding: To minimise distortion and decrease part movement during welding, one approach that has been proven is the use of two laser sources simultaneously, for example two beams 180° apart around the circumference of a pipe when welding a girth weld.
Synchronised movement: To resolve accessibility issues when welding smaller components, synchronised movement of the welding head and the workpiece may result in better access of joints that are difficult to reach.
Diode pumped Nd:YAG lasers: Because of the better beam quality of diode pumped over lamp pumped Nd:YAG lasers, the first can be used with a smaller diameter fibre. Consequently, they can use longer focal lengths than lamp pumped Nd:YAG lasers for the same focused spot size. Conversely, use of a shorter focal length with these lasers can result in a narrower welding head. This may be useful for welding of certain joint configurations in areas that are difficult to access.
The three-dimensional welding capability of Nd:YAG lasers has been demonstrated by welding a 2200x700x300mm steel demonstration component containing two- and three-dimensional joints. The work flagged up some of the challenges that can be encountered when attempting robotic 3D laser welding in an industrial mass-production environment, particularly in the case of butt joints. These challenges, which are mainly related to the beam-to-joint alignment required for laser welding, are discussed and possible solutions are highlighted.
The help of Andy Hassey, who performed the welding described in this paper, is gratefully acknowledged. Nigel Smith, Paul Hilton and Frank Nolan are thanked for their fruitful discussions. Furthermore, the authors would like to thank the sponsors (the UK Department of Trade and Industry and a group of industrial sponsors) for permission to publish this paper.