P. L. Moore
Paper presented at JOM 11, 25-28 May 2003
Synopsis: The ability to measure accurately the temperature of the molten pool during welding allows the cooling rate of the weld to be known. In turn, the cooling rate determines the resulting weld microstructure and hence mechanical properties. Although there are potentially many methods for recording cooling curves during welding, thermocouples are a straight-forward method since they have been used extensively in arc welding. However, adapting these methods to laser welding and laser/arc hybrid welding poses particular problems. Hence, a novel method of harpooning a thermocouple directly into the molten pool has been developed for laser and hybrid welds. The results of this technique are evaluated and compared to other temperature recording methods. Thermocouple harpooning is found to be an effective way of gathering accurate cooling data from laser and hybrid welds.
Reliable cooling data from welds are important for a number of reasons. It is the cooling rate that directly determines the microstructure and hence properties of the weld.  By knowing the cooling profile of a weld accurately, its effect on the weld properties can be isolated. Cooling data also are used in a number of mathematical models, which can lead to understanding and prediction of weld microstructures and/or properties.
2. Temperature Measurement Techniques
Harpooning a thermocouple into the molten weld pool has long been used to obtain cooling data from arc welds.  The thermocouple position is fixed relative to the workpiece and is plunged into the molten pool as the arc passes. This allows the cooling of the weld metal itself to be recorded and so can provide very reliable data during solidification and subsequent cooling. The strength of a laser welded joint is determined by the weld, as well as the heat affected zone (HAZ) properties. For this reason thermocouple harpooning was investigated for laser welds. The harpooning method, however, poses greater problems when laser welding. It is not possible for an operator to be next to the laser during welding on health and safety grounds; the risks from laser radiation and burns are too great.  Hence, a mechanical harpooning method that can be triggered safely from outside the welding cell must be used. Also, the molten pool of a laser weld tends to be much smaller than that of an arc weld, and the need to prevent the thermocouple touching the plasma around the laser beam (or the arc during arc welding) restricts the area of the molten pool that can be harpooned. In hybrid welds the molten pool is slightly larger, but space around the weld is restricted by the presence of two process heads. Despite these difficulties, thermocouple harpooning was used successfully during laser welding and hybrid welding.
Infrared (IR) temperature recording devices use measurements of the emissivity of a surface to determine the surface temperature. IR devices are especially suited to applications where conventional sensors cannot be used, for example when dealing with moving objects, temperatures too high for standard thermocouples or other contact sensors (over 2300°C), or if there is another risk such as high voltage or contamination.  IR cameras show the temperature distribution over a field of view, this makes them attractive for general monitoring of equipment e.g. to indicate hot spots or heat leakage. Temperature readings using IR devices are easier for materials with low reflectivity (materials such as aluminium that have high reflectivity can pose difficulties), and the transition between molten and solid metals gives a discontinuity in the emissivity readings.  This is a problem when recording cooling curves for the whole melting and fusion weld cycle. For an IR device to measure a temperature accurately, the atmosphere between the device and sample surface must allow the IR waves topass through it easily. When there is dust or fume between the device and sample, interference will occur and the device will not be able to measure the temperature of the surface.  Although IR temperature measurement devices are costly and need complicated calibration, once set up, large amounts of data can be recorded.
3. Experimental approach
Three temperature measurement methods were used during the bead-on-plate laser welding of carbon manganese steels, and the results were evaluated. These methods were:
- mechanically peened thermocouples in the HAZ;
- thermocouples harpooned directly into the molten pool;
- remote viewing of the weld using an IR camera.
The thermocouples from both harpooning and peening methods were connected to a data logger and PC in order to record the cooling curves. The data logger was set to record at least 5 readings per second from each thermocouple. All three methods were used to monitor the cooling of 4 kW CO 2 laser welds on C-Mn steel plates of 5 mm thicknesses. When thermocouples are attached to a plate surface the following factors must also be taken into consideration.
- The temperature will be recorded at the first place along the thermocouple that the two wires touch. If a junction is made further up the thermocouple for any reason, the cooling curve recorded will not be that from the molten pool or HAZ.
- The response time of the thermocouple will depend on the diameter of the wire. A thinner wire will give a shorter response time; generally thermocouples made from wire 0.8 mm diameter or thinner will perform satisfactorily.  Thermocouples used in this work were made from wire under 0.5 mm diameter, and the time lag to reach the peak temperature was typically under 0.3 s.
Data obtained from these three methods were compared for their ease of interpretation and applicability for making microstructural predictions. The methods were also evaluated for their relative ease of use. Mechanical peening was chosen as the thermocouple surface attachment method for a robust joint to the thick section material being welded ( Fig.1). The narrow HAZ in laser welds made it difficult to position the thermocouples accurately. However thermocouples peened at distances of 2-3 mm from the weld centreline often gave peak temperatures above 800°C,which defines the HAZ. Photographing the plate, together with a scale rule and taking measurements from the magnified image was a simple method to calculate the position of surface thermocouples relative to the weld fusion boundary (asin Fig.1).
Fig.1. Mechanically peened thermocouples in weld heat affected zone (HAZ)
Harpooning a thermocouple into a laser weld pool is a new application of this technique. The set-up is shown schematically in Fig.2. The thermocouple was held in a plunger, fixed to the laser welding head. This figure also illustrates the aiming position of the thermocouple into the rear of the molten pool. Fig.3 shows a thermocouple harpooning into the back of a laser weld molten pool; the welding direction is from the top left to bottom right of the picture. The thermocouple tip embedded in this weld is shown more clearly in Fig.4. As the plunger was set up to be in constant alignment with respect to the laser and molten pool, it could be fired once the laser keyhole had had time to stabilise. Initial harpooning trials used a spring-loaded thermocouple on a plunger, and its release was triggered by pulling a cord attached to the plunger. This method, although crude, gave a quick and easy way of evaluating the feasibility of the experimental method. Thermocouple harpooning was used with a 4 kW CO 2 laser, welding 16 mm thick structural steel plate, and also used during high power Nd:YAG laser welding of the same material. The 9 kW Nd:YAG laser beam was obtained from three Nd:YAG lasers (two 3 and a 4 kW)combined to give a single beam. High power Nd:YAG lasers generally have a larger molten pool than high power CO 2 lasers, which gives a larger target area for harpooning. For safety, the only way of monitoring the harpooning of Nd:YAG laser welds was via video feedback; the CO 2 laser welds were observed both through the laser cell window and with a video feedback monitor.
Fig.2. Schematic diagram showing the thermocouple harpooning set-up
Fig.3. Thermocouple (held in plunger) embedding in the rear of the weld pool
Fig.4. Thermocouple tip securely embedded in the weld shown in Fig.3
An Infrared camera from AGEMA called Thermovision ® 900 was used to monitor a 4 kW CO 2 laser weld in structural steel.  In order to limit the amount of data handled, the images of the workpiece during welding and subsequent cooling were captured at about 1 per second. The camera came with its own system controller unit for data processing. The IRcamera could be set up fairly easily, but the camera itself must be cooled with liquid nitrogen.
The hybrid welding set up is shown in Fig.5. The laser and metal-active-gas (MAG) welding heads were held at a fixed angle to each other in a bracket, with zero process separation and the MAG leading. The harpooning plunger was attached to the rear of the samebracket. Cooling curves were taken from Nd:YAG laser/MAG hybrid bead-on-plate welds, as well as separately from both constituent Nd:YAG laser and MAG welds, in order to determine the effect of combining the two processes. The part ofthe laser weld to aim at was along the weld centreline at a distance of around 3mm from the laser beam impingement point, in the rear of the molten pool. For hybrid welds, the distance from the impingement point for aiming was nearer6mm. The angle of attack was between 30° and 60°, determined by limitations of clamping to the welding head, and the requirement of a steep enough angle for successful harpooning. If the thermocouple was aimed too close tothe laser beam, it would be destroyed by the plasma and give erratic data. If too far away, the thermocouple would not fuse in the molten pool, and so no weld cooling curve would be recorded.
Fig.5. Diagram of the hybrid laser/arc welding set up
4. Data comparison
Cooling curves obtained concurrently from both mechanical peening in the HAZ (recording a peak temperature above 800°C), and from molten pool harpooning of a thermocouple, plotted from a 4 kW CO 2 laser weld in 5mm thick steel plate, are shown in Fig.7. The harpooned curve recorded a peak temperature around the steel's melting point, and gave a smooth cooling profile. For steels, the cooling time between 800°C and 500°C can be linked to the resulting weld microstructure. Recording a peak temperature at the steel's melting point allows confidence that the whole thermal effect on the weld metal austenite grain formation and transformation has been captured. The HAZ curve peaked at around900°C but showed a similar cooling trend. It can be seen, though, that one of the HAZ peened thermocouples (labelled *) was close enough to the fusion zone edge to record a very similar curve to the harpooned one. However, both HAZ thermocouples were peened at the same distance from the centreline (about 3mm). The difference was due to changes in the molten pool width along the weld, so that one was actually touching molten metal and the other was not. This illustrates the variability of HAZ methods, and the difficulty in predicting the required positions of the thermocouples prior to welding. It is possible to have high confidence in the data from the harpooned thermocouple since it is clear that the weld bead cooling profile itself has been recorded.
The peak temperature of the harpooned curve was just below the actual melting point of steel, even though the thermocouple had embedded into molten metal. This was due to the slight temperature lag in the recorded data from the thermocouple tip's thermal mass. It can be seen from Fig.7 that the agreement between the HAZ peened thermocouple data, and that from harpooning when both are recording the temperature of molten metal was good. All three curves show a very similar cooling behaviour below about800°C, although this would not always necessarily be the case. The agreement between the curves leads to the conclusion that the quality of the data gathered from both harpooning and peened methods is very similar. However, the error in measuring the position of a peened thermocouple means that the peened thermocouple might not be recording a cooling curve from the desired location. The amount of confidence in the thermocouple cooling data is one of the main differences between the peening and harpooning methods. However, both the thermocouple methods allowed a measure of the cooling time between 800°C and 500°C to be determined.
Fig.7. Cooling curves obtained from a 4 kW CO 2 laser weld in 5 mm thick steel sheet using both harpooning and peening thermocouple methods
Cooling curves obtained from Nd:YAG laser welds were of a similar quality to the CO 2 laser welds. Fig.8 shows harpooned cooling curves obtained from the same 16 mm thick steel plate for an 8.9 kW Nd:YAG laser weld, and a 4 kW CO 2 laser weld, both bead-on-plate. When welding thicker material with the same power as the 5 mm sheet, the cooling time between 800°C and 500°C decreased (from 3.2 s to 1.8 s) since the thicker plate conducts heat in three dimensions instead of just two. Using a higher laser power (8.9 kW instead of 4 kW) on the thick material increased this cooling time to 2.8 s as the effect of heat conduction in the plate through-thickness direction was reduced.
Fig.8. Harpooned cooling curves from an 8.9 kW Nd:YAG and a 4 kW CO 2 laser weld in 16 mm thick steel plate
When using the IR camera it was found that the interference of the laser keyhole plasma with the IR radiation during welding led to indistinct images being captured ( Fig.9a). The dark region in the middle of the image was a null reading due to this plasma interference. However, the images captured immediately after the laser beam had been switched off showed a clear visual cooling profile along the weld line ( Fig.9b). This IR camera could show the temperature profile along a line scan on the image (e.g. along the weld bead), but this data could not be superposed onto the cooling with time curves recorded from the thermocouples. The IR camera was not useful for microstructural predictions since a cooling time between 800°C and 500°C could not be calculated.
Fig.9. Infra red camera images: a) during laser welding b) just after the laser was turned off, showing the cooling profiles. Temperatures are in °C
The laser, MAG and hybrid welds are illustrated in Fig.10, showing the macro profiles of the separate arc weld component and laser weld component, compared to the resulting hybrid weld when combined. The MAG weld did not penetrate into the plate, and was quite peaky due to thefiller wire. The Nd:YAG laser weld penetrated deeply, but had a narrow profile and flat surface since there was no filler addition. The hybrid weld showed a similar penetration to the laser weld, but also had the peaked surface profilefrom the filler wire addition. Fig.11 shows a diagram relating the heat flow in the welds to their macro profiles; the strengths of the arrows indicate the amount of heat flow per second, and the direction of heat flow.
Fig.10. Macro profiles of a) MAG-only, b) laser-only; and c) the resulting hybrid weld profile. Scale is in mm
Fig.11. Diagram of heat flow amount and direction during cooling of MAG weld (A), laser weld (B) and hybrid weld (C)
The cooling data recorded from the MAG-only, laser-only and hybrid welds are shown in Fig.12. The MAG-only weld had a heat input of 215 kJ/m, while the laser-only weld's heat input was 281 kJ/m. The hybrid weld was made by directly adding these two welding sources and had a heat input of 496 kJ/m. It took 1.3seconds for the MAG weld to cool from 800°C to 500°C, and this resulted in a fine microstructure of predominantly acicular ferrite, with some grain boundary ferrite. The composition of the filler wire helped to promote thistype of microstructure since the high oxygen content in the metal-cored filler wire forms inclusions in the molten pool for nucleation of intragranular phases. The laser-only weld cooled from 800°C to 500°C in 1.3 s, the sameas the MAG weld. However, the laser weld had a microstructure of predominantly ferrite with an aligned second phase, and grains oriented perpendicular to the weld depth. This occurred due to the deep and narrow profile of the laserweld, and the lack of composition control from the filler wire, that was possible for the arc weld. Comparing the cross-sectional area of the laser weld to the arc weld, its fusion zone is about twice the area of that of the arc weld,so twice the metal has solidified in about the same time. This caused the aligned ferrite type microstructure to form in this weld. When hybrid welding bead-on-plate the cooling rate increased to 2.9 s, almost in direct proportion tothe increase in heat input when the arc and laser were combined. The microstructure contained predominantly intragranularly nucleated phases, including acicular ferrite, due to the presence of inclusions arising from the use of the MAGfiller wire. However, further down the weld the microstructure contained more ferrite with an aligned second phase, like the laser weld. It is a combination of the different heat cycles that each weld experiences ( Fig.11), and the chemical composition control available from filler additions, that results in the different microstructures between the hybrid weld and its constituent processes.
Fig.12. Cooling curves from hybrid weld, and its constituent MAG and laser welds
5. Method comparison
The problem with the mechanical peening method is the relatively large error in measuring the thermocouple's position relative to the edge of the weld fusion zone. A tolerance of ±0.5 mm is not unusual since this might be thediameter of the hole, and can be greater if the bulky thermocouple insulation must fit inside the hole. The relative distance of a thermocouple from the fusion boundary might alter, even if the distance from the weld centreline isknown, due to slight variations in the weld width. Even small differences in distance from a laser weld will drastically alter the cooling profiles obtained, and is particularly problematic when obtaining data from the narrow HAZ inlaser welds compared to arc welds. For harpooning, adjustment of the clamps gave adequate control to position the thermocouple as required. There was some difficulty in aiming the thermocouple into the correct part of the molten pool.Initially, about one in six harpooning attempts was successful, resulting in the thermocouple being fused into the molten pool. This improved to about 1 in 3 when video feedback was used on the CO 2 laser, to allow better viewing of the welding and to identify possible reasons for the failed attempts. When harpooning Nd:YAG laser welds, video feedback was always used, and gave similar success rates (1 in3). Hybrid bead-on-plate welds gave similar success rates to the Nd:YAG laser welds. However, when harpooning the bead-in-groove hybrid welds it took almost ten attempts to obtain a successful cooling curve. This was due to the moltenpool being much narrower in the vee preparation, giving a much smaller target area, as well as being a millimetre or two down into the groove restricting access. The IR camera can be used to give useful qualitative information but doesnot give a simple method of obtaining accurate cooling data that can be usefully processed. It also suffered interference from the laser during welding, giving a noisy image.
6. Summary & conclusions
When a thermocouple becomes embedded within the weld after harpooning, it is possible to have high confidence in the data obtained. The cooling curve is that from the actual molten pool itself once the thermocouple has become stuckfast in the weld after harpooning. Although it is not possible to guarantee that a good cooling curve will be obtained from every harpooning attempt in laser and hybrid welding, the quality of the data is of a high standard andaccurately measures the cooling of the weld metal. This makes the approach very worthwhile, although best suited to the acquisition of a smaller number of critical cooling curves, rather than for large scale data acquisition from ahigh number of welds. Cooling curves recorded from welds using different laser power on different material thickness allow an understanding of how the cooling time will be affected by these process parameters. When the cooling curveswere taken from both a laser/arc hybrid weld, and its constituent laser and arc welds, it was possible to determine the effect of the process combination on the weld cooling time between 800°C and 500°C, and the resultingmicrostructure.
The overall conclusions from this work are:
- Both CO 2 and Nd:YAG laser welds, and laser/MAG hybrid welds can be successfully harpooned with a thermocouple, and good quality cooling curves obtained.
- Confidence in the data obtained, and the relatively low cost of the equipment, makes thermocouple harpooning of laser welds an attractive method for data acquisition.
- Infrared temperature measurement methods suffer with interference of the laser plasma preventing readings from the molten pool to be made during welding.
- Thermocouple harpooning of laser welds is recommended for small-scale data acquisition, rather than for large numbers of data sets.
- Rapid cooling in autogenous laser welds, that results in an unfavourable microstructure of ferrite with an aligned second phase, can be improved by using a laser/arc hybrid process. This results in an improved acicular ferrite weld microstructure, as a result of the slower cooling and filler addition composition control.
- The cooling time between 800°C and 500°C for a weld can be increased (preventing the formation of hard meta-stable phases), by using thinner material when welding with the same laser power, or by increasing the laser power for welds in thick plate, since this reduces the heat conduction in the through-thickness direction.
I'd like to thank my PhD supervisors, Dr Dave Howse at TWI and Dr Rob Wallach at the University of Cambridge. This work forms part of a larger collaborative project at TWI. Thanks are given to BP for providing the materials, and toEPSRC and the DTI for funding this work.
- Moore, P. L., Howse, D. S. and Wallach, E. R. 2002. Microstructure and properties of autogenous high power Nd:YAG laser welds in C-Mn steels. 6 th International Conference on Trends in Welding Research, Pine Mountain, Georgia, USA.
- Terasaki, T., Hall, G. T. and Pargeter, R. J. 1991. Cooling time and predictive equation for estimating hydrogen diffusion in CTS test welds. Transactions of the Japan Welding Society. 22(1):52-56.
- Steen, W. M. 1998. Laser Materials Processing. 2 nd Edition. p.323 Springer Verlag.
- Gruner, K. 2001. Principles of non-contact temperature measurement. Raytek ® www.raytek.com
- Henrikson, P. and Ericsson, M. 2002. Non-contact temperature measurements using an infrared camera in aerospace welding applications. 6 th International Conference on Trends in Welding Research, Pine Mountain, Georgia, USA.
- Eller, F. C. 1983. Thermocouples-essential devices for proper control of local weld heat treatment. Welding Journal 62(12):38-40.
- AGEMA Infrared Systems. 1992. Thermovision ® 900 Series Users Manual-part 2.