Forces to be reckoned with - an examination of what acts where, and by how much, during the friction stir welding process

TWI Bulletin, November/December 2000

After graduating from Cambridge University with a Degree in Metallurgy, Richard Johnson joined UKAEA. Later whilst working at the Institute of Physics Fulmer Research Institute he was awarded a PhD in Metallurgy and Materials Science by Surrey University. He then rejoined UKAEA for 24 years before joining TWI in 1998 as a Senior Project Leader in the Electron Beam, Friction and Forge Processes Department.

Nathan Horrex is a Senior Research Technician, who has worked in the Friction and Forge Processes Group for three years, mainly involved with equipment development and research and development of friction based processes. He joined TWI from college and holds a Higher National Diploma in Engineering Mechatronics.

Friction stir welding (FSW) was developed at TWI and resulted in a patent application in December 1991. ^[1] The process operates by a rotating tool being plunged into softer material and then traversed through it. The material is plasticised by the friction heat generated by the tool as it rotates and, without any melting, the material is swept around the tool from the front to the back in order to effect a weld ( Fig.1). As Richard Johnson and Nathan Horrex report TWI has been performing an increasing amount of work for Industrial Members as well as developing a further understanding of the technology through its Core Research Programme.

As part of the latter, a need has been identified to quantify the FSW forces and torques developed by the process in order to transfer the technology to industrial equipment, including robots and machines operating under closed loop control. Force measurement has been performed previously by the use of strain gauges and a unidirectional force table, ^[2] but only measuring the downward and traversing forces. There is instrumentation now available to monitor force and torque data more directly on the FSW tool. This report summarises the findings of a short programme of work to evaluate such instrumentation.

Objectives

• To evaluate a commercially available dynamometer for the measurement of the horizontal and vertical forces and the torques generated during FSW.
• To compare the forces and torques generated during the FSW of different aluminium alloys.

Experimental approach

FSW machine and materials

6082-T6 aluminium alloy plates of 6.3mm thickness were selected as the initial material for the assessment, and these plates were friction stir welded in a standard welding jig. These were followed by 5083-O alloy plates of the same thickness. Some of the trials were conducted on a stir in plate basis (similar to the bead on plate welds in arc welding) rather than welding two plates of material together, in order to accommodate the small force table in use for some trials, Table 1.

Table 1: Compositions, hardnesses and room temperature mechanical properties of materials used (TWI analysis S/98/372)

The FSW tool was made of H13 tool steel, of a design incorporating a profiled pin that had been used at TWI on previous projects.

Force measurement equipment

Kistler Instruments has developed a range of force measurement devices based on the signals generated by piezo-electric crystals in compression. These range from small tables with four load cells, that can measure forces in the three orthogonal axes and can derive the moments along these axes, to rotary dynamometers that can measure forces in three axes and also measure the torque generated by the process. These units have found reasonably wide usage in the machine tool industry, but there has been limited experience to date of their application to FSW, although NASA and Boeing have recently reported the results of using this type of instrument. ^[3,4] The advantages of using a dynamometer include the capability to measure the torque developed during the process as well as the forces in three orthogonal directions, and also being able to monitor very long length welds which would otherwise require very large force tables.

The dynamometer used was the 9123 model, which has a load limit of 20kN in the vertical and ±5kN in the horizontal axes, and up to 200Nm torque. There was an operational temperature restriction of 60°C, because of the electronics mounted in the dynamometer. FSW operates in aluminium alloys by generating temperatures of up to 450-500°C in the workpiece and around the tool tip, and commensurately higher values in copper and steels. For this initial short programme it was decided not to incorporate a thermal barrier. In view of the restrictions imposed by the size and limitations of the force measurement systems, it would be important to operate the unit only for short duration welding runs at relatively slow traverse speeds, with the welding conditions not necessarily reaching equilibrium. This meant that any time- or temperature-dependent drift in load measurement was unlikely to be recorded during these short weld runs. In some of the weld runs, a Kistler 9257B force table was also used, with vertical and horizontal force capabilities of 30kN and ±20kN respectively, and the force measurements of the two units were compared. The small size of the table restricted the weld length to only 100mm. To assist in limiting the temperature rise seen by the units, a gas cooling jet was aimed at the tool during and immediately after completion of the weld run. The test arrangement is seen in Fig.2.

Experimental procedure

The plates were bolted into a holding jig that was placed on the machine bed, or on the force table which in turn was bolted to the machine bed. The FSW head was static whilst the plates were traversed underneath it. All of the weld trials were performed with the FSW tool tilted backwards at an angle of 1.5°, so that the rear of the weld is kept under compression, except for some trials at 0 and 3° in order to assess the effect of tilt angle on the forces. Most of the work was conducted on 6082-T6 plate, but some trials were conducted on 5083-O plate in order to assess the welding forces on an alloy that requires more energy to produce an equal length of weld in the same thickness of material. The welding was effected by plunging the FSW tool into the material to a constant depth. Figure 3 shows the friction stir welds in 6082-T6, and also shows a surface flaw being produced when the downforce is too low, and sound welds produced when the downforce is increased, whilst Fig.4 shows some friction stir welds in 5083-O.

Table 2: Friction stir welding force and torque measurements using Kistler instrumentation (9123 dynamometer and 9257B force table)

Results

Initial dynamometer evaluation trials on 6082-T6

These results are summarised in Table 2, welds 1-8. The rotational and traverse speeds used were initially low, at 710 rev/min and 40 mm/min, to ensure that the forces on the dynamometer were not too high. As the forces measured were well within the capability of the instrument, the process parameters were progressively increased up to 1400 rev/min and 224 mm/min. As the traverse speed increased, both the horizontal and vertical forces being developed were seen to increase from about 0.6kN and 6kN respectively to about 1.3kN and 10kN respectively. The torque also increased from 10 to 16Nm.

In Figs.5 and 6 the force traces are annotated to be virtually self-explanatory. There is a small movement in the X and Y traces (horizontal forces), and in the Z trace (vertical force or download) and in the torque trace when the FSW tool pin touches the surface of the plate and is plunged into the material. A larger movement in the vertical force trace is observed when the tool shoulder, being a much greater area, touches and is plunged slightly into the material. As soon as the welding traverse begins there is a large sinusoidal motion recorded on the X and Y force traces for the dynamometer, and a step change in the equivalent force traces on the force table print out, as well as smaller changes in the Z force and torque traces. When the weld traverse is stopped the X and Y force traces decay to a low level, and as the FSW tool is lifted out of the material the torque trace and all of the force traces decay to zero.

It was apparent that the forces measured in the horizontal axes by the dynamometer outputs were very similar. It was also apparent that relatively high vertical forces were achieved when first the FSW tool pin and more especially when the shoulder touched the workpiece, sometimes reaching 70% of the unit's limit of 20kN. Care was taken subsequently not to exceed this limit, by very slowly lowering the tool into the workpiece. In general the plots showed, even with a dwell time of 5-10 seconds before starting the weld, that equilibrium conditions were not always being approached during the short weld runs.

Trials with the dynamometer and force table on 6082-T6

Trials on 6082-T6 plate to assess the effect of tilt angle

Most trials were performed with the FSW tool tilted 1.5° away from the direction of welding, and a comparison was made with tilt angles of 0, 1.5 and 3°. The results are summarised for welds 22-24 in Table 2, which were also stir in plate welds. All three welds were similar in appearance with regard to weld surface roughness, although the trial at 0° tilt had a surface flaw. The force and torque plots were all in close agreement, with the marginally lowest horizontal forces being developed with the head tilted at 1.5°. But the downforces and torques were effectively the same for all three tilt angles, as can be seen in Figs 7 and 8 for welds 22 and 23 and in Table 2.

Comparative trials on 5083-O

The FSW trials on 6.3mm 5083-O plate are summarised for welds 26-34 in Table 2 and a typical force plot given in Fig.9. The welding speeds adopted for 5083-O were lower, based on previous experience, and started at 500 rev/min and 40 mm/min, increasing to 710 rev/min and 224 mm/min respectively. The horizontal forces measured were up to 2.5kN and the downforce was up to 15kN, which were 50-100% higher than those measured in the 6082-T6 studies, whilst the torques were also higher at up to 40Nm. Above 80 mm/min traversing speed there was no evidence of surface flaws, and in this material there were no significant differences in the force plots between FSW stir in plate runs and welding two plates together.

Discussion

Dynamometer and force table trials

The 9123 dynamometer measures the forces at a rate of 4kHz, and therefore produces a plot oscillating between ±maximum in every revolution of the FSW tool. Additionally the FFP monitoring equipment operating at 10-1000Hz could not consistently measure the force at 90° to the traverse direction when the force in the direction of welding was at its maximum. A faster sampling rate in the monitoring equipment would have made this possible, but a different dynamometer would be beneficial in simplifying the initial data output by monitoring only the peak load for each revolution of the tool.

The force table plots almost exactly mirror the contours linking successive peaks plotted by the dynamometer, without the sinusoidal oscillation produced by the continuous readout of the latter. In this sense the force table plots showed more clearly how the rotation of the FSW tool generated a sideways force almost as large as that in the direction of welding, at the most some 5-10% lower than that generated in the weld direction. Previous workers have also reported significant sideways forces, sometimes approaching the same force as that generated in the welding direction. ^[3,4] It is an important design requisite for multi-directional FSW machines, and especially when specifying the use of robots, to be able to withstand these forces adequately.

6082-T6 trials

The work performed on 6082 was largely in agreement with previous work at TWI, ^[2] with the horizontal and vertical forces and the torque all increasing with traverse speed. The horizontal forces also increased with rotational speed at a constant traverse speed, and the torque marginally increased. In general the downforce applied and the angle of tilt at 1.5° adopted were somewhat lower than those used previously at TWI, ^[2] and it was apparent from some welds with surface flaws that in this short programme of work the aluminium alloy was not always being fully consolidated, but increasing the downforce corrected the effect. The rate of lowering the FSW tool into the workpiece is controlled by the operator, and affects the peak downforce developed, but the plunge depth of the tool into the material will affect the equilibrium downforce whilst welding, and this and the horizontal forces and torques generated will be governed by how the material responds to the FSW process. Because of the temperature limitations of the dynamometer, the rotation and traversing speeds used in this work were towards the lower end of the spectrum of operating conditions normally used with this tool, so some differences in parameter measurements were not totally unexpected when comparing the results with previous work. ^[2] There may also be minor variations in the composition of the material across the plate and between plates of the same composition and heat treatment. In this work with short weld runs the equilibrium processing conditions had generally not been achieved.

There was very little difference between the force plots for the trials conducted with the FSW tool at 0, 1.5 and 3° tilt angles. However, the trial conducted at 0° tilt produced a surface flaw, which was probably a consequence of the heel of the tool not being plunged into the material adequately.

5083-O trials

The limited work conducted on 5083-O showed the same trends as seen in 6082-T6, with the forces and torque increasing with traverse speed. The rotational speed was varied only between two values, and at the higher speed the forces had increased slightly but the torque had dropped more significantly. It is possible that the greater energy being provided at the higher rotational speed is able to develop a softer plasticised zone around the tool pin, with a resultant drop in the torque, but then an increase in traverse speed generates higher forces because of the greater rate at which cold material is being introduced into the plasticising region. In general the forces were up to 100% higher than those measured in the 6082-T6 plates of the same thickness, and the torque was 100-200% higher, which suggests that 5083-O requires more energy to effect a friction stir weld equivalent to one in 6082-T6 material. Additional trials would establish whether there is a critical rotational speed window where good quality welds at low torques are achievable for an energy-efficient process, as it has been estimated that the torque developed during FSW operations can represent half of the energy consumed. ^[3]

Conclusions

• The dynamometer measurements were comparable with those of the force table, except the former produced pronounced oscillating plots of data because of its continuous force monitoring during each revolution of the FSW tool.
• The sideways horizontal force generated in FSW of 6082-T6 and of 5083-O was measured to be at least 90-95% of that generated along the direction of welding.
• The forces and torques generated in the FSW of 5083-O plates were higher than those measured when processing 6082-T6 in the same plate thickness, confirming the indications in previous TWI work.

Recommendations

• Further work needed on the quantification of the forces and torques developed during FSW, in particular extending the weld lengths to ensure equilibrium processing conditions are achieved, and replicating the optimised processingparameters developed for different materials.
• FSW machines and robots should be designed to be capable of accommodating the sideways forces which are equivalent to those generated by the process in the direction of welding.

References