Development and Evaluation of a Robot Guided Friction Stir ...

In today’s automotive industry, reducing vehicle weight to increase energy efficiency and also improve vehicle safety is of central importance [1, 2]. The use of mixed material design in car body engineering is constantly increasing, as it enables optimal weight, rigidity, and cost by combining different materials [3,4,5]. This poses a considerable challenge for the joining technology, as conventional methods often reach their limits. Furthermore, the relatively new approach of using mega or giga-castings in automotive manufacturing introduces additional challenges: the reduced number of parts and joints, variability in component and material properties, and geometric inconsistencies all demand for different or new joining technologies [6,7,8]. Especially for lightweight materials such as high-strength aluminum or magnesium alloys, friction stir welding (FSW) is a widely used state-of-the-art technology with superior strength properties of the resulting welds [3, 9]. Irregularities in the weld seam, such as hot cracks and pores, associated with fusion welding of high-strength aluminum alloys of the 6XXX [10] and 7XXX [11] series, do not occur with friction stir welding due to the lack of a weld pool [12, 13].

Check now

Furthermore, FSW is considered an ecofriendly and energy efficient technology, as it does not require consumables such as wires and shielding gases and does not emit harmful fumes or radiation to operators and workers in the surrounding area [14]. When comparing FSW with other welding processes such as gas-metal arc welding, studies showed a decrease in energy consumption by 42 % when welding aluminum and up to 50 % when welding aluminum to steel [15, 16].

Despite its advantages, the wide use of friction stir welding is often limited by the high process forces and relatively slow feed speeds compared to fusion welding [17], which increases process times. The high process forces can also lead to a deflection of the machine frames, which in turn can lead to process flaws [18,19,20]. This problem occurs in particular with highly flexible welding equipment, such as welding guns for RSW[21], friction stir spot welding (FSSW) [18] and robot-assisted FSW [20]. In these applications, the requirements for the lowest possible weight and high accessibility conflict with the demand for low deflection and high rigidity of the system.

As a result, there have been a variety of patents and developments in this field in the past that have aimed to realize friction stir welding as a gun application. Due to the comparatively low strength of the spot weld seams, different variants for enlarging the weld seam were considered, from which the Swing and the Stitch FSW have evolved [22,23,24]. The solutions developed range from simple patent ideas to prototype applications and series applications in the automotive industry. One of the more prominent examples of FSW is the welding gun used with the Honda Accord, which has a spherical anvil and was used to weld the engine subframe from a mixed aluminum-steel joint [16, 25, 26].

At the Materials Testing Institute (MPA) of the University of Stuttgart, a friction stir welding gun was developed to manufacture short stitch welds with an integrated linear kinematic [27]. The aim of the development is to enable flexible production of short friction stir welds as an alternative to conventional spot joints such as RSW and rivets in car body manufacturing.

As reported in the literature, FSW spot welding process modifications such as refill friction stir spot welding (RFSSW) or Swing-FSW show a higher static and fatigue strength compared to conventional RSW, as shown by Okamoto [22], Badarinarayan [24] and Gale et al. [28], or riveting, as shown by Lakshmi et al. [29] with materials and thicknesses commonly used in the automotive environment. Furthermore, the strength properties of the spot welds can be increased by expanding the weld area [24, 30], which is the motivation of the stitch-welding approach presented here. In Fig. 1, the kinematic principle of the friction stir welding gun is shown, as well as possible stitch weld configurations.

A first version of the friction stir welding gun developed showed a significant deflection at high plunge forces. Due to the insufficient rigidity of the anvil and the frame and the resulting change in angle in relation to the rotation axis of the tool, tunnel defects emerged in the outer part of the weld zone, as shown in Fig. 2.

Based on these initial experiments, a version 2.0 welding gun was developed, which involved scaling the screw connections and changing the layout of the components. In addition, the entire design space of this optimized version was reduced and the projection was increased to meet industrial requirements.

In order to understand the origins of the deflections and to investigate the design weaknesses of the first version of the welding gun, this study uses a digital twin (DT) approach for a design study on stiffness optimization. For this purpose, the first demonstrator (1.5) and the advanced welding gun version (2.0) are digitized and tested with the real force–deflection curve of their physical twin in order to identify and optimize areas of high deflection. Digital image correlation is used to calibrate and validate the DT with their physical twins.

The two variants fundamentally differ in their design-variant 1.5 represents a feasibility demonstrator, while version 2.0 already fulfills basic requirements such as a reduced installation space and enhanced accessibility. One of the main requirements for both welding guns was their applicability with conventional industrial robots (e.g. KUKA KR210), which led to a weight limit of 200 kg. The structure and components of the two versions of the welding gun are shown in Fig. 3. The main differences between version 1.5 and 2.0 of the Steppwelder are the arrangement of the linear guides of the main spindle, the orientation of the feed actuator and an increase in the cantilever of the anvil from 235 mm in version 1.5 to 320 mm in version 2.0 for improved accessibility. In version 2.0, the linear guides of the spindle have been attached to the rear of the spindle at a 45\(^{\circ }\) angle in order to utilize the increased load capacity of the guide carriages around the longitudinal axis. The wall thicknesses of the welded frame components have also been increased in version 2.0. The main frame of the welding gun 2.0 is bolted to the side of the rear square profile with 32 M12 bolts. In version 1.5, this connection is realized at the front using 10 x M10 screws, which led to increased resilience and thus also angular deviations. However, the total weight of both welding guns is the same with ca. 180 kg. The increase in weight due to the increase in wall thickness of the frame parts is compensated for by the use of a more lightweight integrated spindle motor. In general, initial optimizations can already be expected to result in improved rigidity.

The C-frame consists of a square steel tube, in which the drive of the anvil is installed, and a welding frame for the main frame, which carries the spindle. As the welding gun frame has to absorb and dissipate all process forces, the main focus here is on the highest possible rigidity; optimizations in regard to weight, design space and maximum rigidity are the most promising here.

Due to the complexity of machinery and components, optimizing design and stiffness is often an elaborate and costly task. With the help of DTs, such tasks can be simplified without the need for component manufacturing and testing on a physical machine. The use of data for decision making in the engineering design process is also the objective of this study. With a DT, future designs of the welding gun and a predictive analysis of the deflection areas and overall stiffness can be made using real-time data like plunge force, feed speed and position angle.

The high process forces during friction stir welding are a challenge both for the machine structure and for the spindles. Conventional spindles in production technology are usually designed for the predominant radial forces in milling and turning and focus on high concentricity (e.g. for drilling applications). This results in restrictions with regard to the selection of parameters and components when friction stir welding is to be carried out with higher axial forces on systems that are not designed for the process forces (axial and radial) occurring with friction stir welding. Limited information is available in the literature on spindle and bearing concepts specifically for friction stir welding.

For example, in the work of Luo et al., two quad-tandem arrangements with a large spacing between the ball bearings are used in order to be able to withstand the highest possible axial forces, which, however, increases the spindle’s design dimensions [36]. Grätzel et al. developed a two-piece demonstration spindle for separate control of the tool shoulder and pin during friction stir welding, which consists of two spindles arranged in series [37]. On the one hand, this variant makes it possible to reduce machine vibration by rotating in opposite directions, and on the other hand, it makes the application much more flexible, as welding can also be carried out with a stationary shoulder. However, the spindle has a design space that is not suitable for application in a welding gun.

Industrially available spindles are characterized by high achievable axial forces at low speeds (through the use of torque motors) or medium achievable axial forces at higher speeds. A characteristic common to all available concepts is the relatively large design space, which indicates a scaling of conventional spindle or ball bearings.

A compact spindle design therefore requires a dedicated bearing concept. The spindle of the welding gun presented in this work is shown in Fig. 4. The main bearing consists of a pair of tapered roller bearings in an X arrangement, while a cylindrical roller bearing is used as the floating bearing. The rotary drive is a 28 kW frameless spindle motor with a maximum speed of 10,000 rpm. With this concept, axial forces of above 20 kN can be realized at a maximum torque of 48.1 Nm with an outer diameter of 147 mm and a length of 526 mm. With the ER25-collet integrated in the shaft, welding tools can be used with a diameter of up to 18 mm.

The Schaeffler company software tool bearinX was used to calculate the bearing configuration. Various possible bearing concepts were investigated and calculated, taking into account the maximum design space while retaining the shaft geometry. The assumed load collective used for bearing calculation is shown in Tab. 1 and includes conservative axial and radial forces, as well as moments that typically occur during friction stir welding, as well as examples of FSW applications on sheet metal combinations relevant to the automotive sector are listed in correspondence with the forces. The force values were based on the first friction stitch welding tests and refer to tools with shoulder diameters between 10–14 mm for 6XXX aluminum alloys with sheet thicknesses between 1–3 mm. Future application-related tests are likely to be carried out with shoulder diameters of less than 8 mm, therefore the force collective is to be considered conservative. The simulation of the frame design in the following section was not carried out with the load collective from Tab. 1, but a more conservative value of 14 kN in Z-direction. As shown in Fig. 5, tapered roller bearings have the significantly longest service life with an assumed lifetime lubrication for the given design space and requirements and are therefore used in all design variants of the welding gun. If higher load collectives are expected to occur, for example when using larger sheet thicknesses or tools, the tapered roller bearings might still be the best choice. As an alternative, the diameter of the bearing could be increased if the design space is not limited.

4.1 Numerical Modeling

For this study, two versions of the friction stir welding gun were built as a complete DT in the ABAQUS simulation software and compared with each other. In order to ensure reasonable simulation durations and a converging solution, simplified models with individual components of the respective welding gun were calculated beforehand. Five pre-simulations were considered for each version, including the mere analysis of the anvil and adding more components with each simulation. As part of these pre-simulations, a starting increment step of 0,001 s was found to be sufficient to obtain a converging solution for the respective versions, and the maximum increment step was set to 0,5 s to achieve reasonable simulation durations. The runtime for one simulation of the complete model was ca. 6 h containing 110 increments. The meshing of the respective assemblies was carried out with C3D8R elements, using an adaptive mesh with mesh sizes between 0,1 mm and 2,8 mm for the frame parts and mesh sizes between 6 mm and 12 mm for parts which were of minor interest such as the spindle or robot flange. The screw connections of the components were modeled with single-bolt connections and a preload force according to the standards for the respective screw size VDI  and DIN EN ISO  [42, 43]. The simulation consisted of two steps: (1) preloading of the screws and (2) applying the closing force according to the measured force over time ramp. The first step also included a small plunge force of 100 N to ensure contact between the screwed parts, as the first models showed convergence errors due to the definition of force-induced contact. The second step was realized through a linear actuation of the plunge motor, therefore closing the welding gun with a plunge force of 14 kN.

After validation using comparative optical measurements, the DT of version 2.0 was further tested in simulations with higher forces applied in the Z-direction to explore its potential applications and limitations. During the plunge phase of FSW, the Z-direction force can temporarily exceed the target value significantly. Table 2 summarizes the axial force values of the literature for material and thickness combinations relevant to the automotive industry. These values show that forces can reach up to 15 kN, although most applications require considerably lower forces.

The feed motor installed in this setup has a nominal maximum force of 14 kN but is capable of briefly generating up to 20 kN. To account for this, an extended simulation was performed with a plunge force of 22 kN and a welding force of 20 kN. The maximum deflection at the virtual measurement points was calculated accordingly to the previous simulation with a welding force of 14 kN.

You will get efficient and thoughtful service from World Wide Welding.

The force signal at the beginning of the force–deflection curve was smoothed before being used as an input into the linear-elastic numerical simulation. The frame parts of the welding gun were modeled using the material S355JR with a yield strength of 355 MPa, while the material 42CrMo4 with a yield strength of 550 MPa was selected for the force transmission components. Weld seam geometries in the frame parts were neglected to simplify the modeling process and a detailed modeling of the linear rails and linear carriages was omitted in order to enable the overall model to be meshed and converge. Strength grade 8.8 with a yield strength of 640 MPa was specified for all screws. The spindle assemblies have been simplified to a single solid part with the material properties of S355JR. In order to be able to compare the two versions with each other, measuring points with fixed distances were defined on the welding gun frame. Care was taken to ensure that the points had a similar distance and position despite geometric differences in the gun frames.

4.2 Physical Testing

The physical models of the welding guns were measured under load using optical strain measurement to evaluate the quality of the simulation results. The load was applied using the plunge motor and thus closing the gun until the force sensor showed a load of 14 kN. The reference points were arranged on the parts of the frame in a manner equivalent to the simulation model, as shown in Fig. 6. The optical strain measurement was performed with a GOM ARAMIS system, which was previously calibrated to an absolute measurement accuracy of 0.01 mm. The plunge force was measured using a load cell consisting of strain gauges, which is located in the force flow of the weld gun frame between the spindle and the plunge motor. The strain gauges in use had a gain factor of 2 mV/V and were measured with a 4-channel analog input EtherCAT terminal in a full-bridge arrangement. Calibration of the full bridge was carried out using a testing device, especially for closing forces on welding guns. The progression of force over time was recorded during the tests and served as an input variable for the deflection simulation of the DT. The optical measuring system was positioned so that a front view of the welding gun was possible. Figure 7 provides an overview of the experimental test setup of the optical strain measurements.

5.1 Comparison of the Deflection Behavior of Steppwelder 1.5 and 2.0

Figure 8 shows the stress distribution (von-Mises equivalent stress) in both welding guns at the maximum joining force of 14 kN. The stresses of the screw preloading step can be seen around the screw heads and add to the stress induced by the loading force in those areas.

Due to the increased wall thicknesses in version 2.0 - both in the main gun frame, the square profile and in the anvil - the overall stresses are significantly lower here. It is also noticeable that the stress level in the welded construction of the top frame 2.0 is negligible, whereas in version 1.5 these assume values in the range of 90–100 MPa in the area of the screw joints. The anvils of both versions show a similar stress distribution; in version 2.0, the longer projection and the increased wall thickness of the component compensate for each other.

However, a qualitative comparison of deflection behavior reveals significant differences between versions 1.5 and 2.0, as can be seen in Fig. 9. Although the overall deflection is reduced by a factor of approximately 3 as a result of the optimization measures, a change in the virtual pivot points can also be seen. In version 1.5, the upper frame of the welding gun bends around the center of the mounting surface on the square profile (P1). The anvil also deforms around the center of its mounting points on the profile. This indicates that the compliance of the designed geometry, especially in the linear rails and screws, is too high and leads to undesired deflections. These effects cannot be observed in version 2.0, where the virtual pivot point of the upper frame of the gun is approximately in the center line in the back of the gun (P2). Only displacements at the tip can be recognized in the anvil 2.0; no significant deflection can be observed in the direction of the square profile. It can therefore be concluded that the anvil behaves more like a cantilever and that the displacements no longer occur in the linear rails. This cantilever anvil can be further stiffened by topology optimization.

5.2 Calibration and Validation of the Digital Twin

The simulation results are compared with the optical strain measurements in order to validate and, if necessary, calibrate the DT of the two welding guns. For this purpose, the general deflection is compared in Fig. 10. In addition, parts of the deflection that result from the upper frame and parts that originate from the anvil are evaluated. For both simulations, an underestimation of the experimentally determined deflection is evident, for version 1.5 by 0. mm (1.778 mm versus 1. mm) and for version 2.0 by 0. mm (0. mm versus 0.633 mm). The division into upper frame proportion and anvil does not exactly match the measurements in both simulations. Version 1.5 overestimates the proportion of the top frame component and underestimates the proportion of deflection in the anvil, while these ratios are the exact opposite in version 2.0. One reason for this is the simplified design of the linear rails in the DT, which does not include or allow any information on possible clearance. In addition, the welded construction of the upper frame was simplified in both variants without taking into account the geometry of the weld seams and mounting holes.

Looking at the overall deflection of the two versions, it can be concluded that the DTs are well suited for preliminary assessments and optimizations with regard to stiffness behavior.

The maximum deflection over time and force over time plots at the respective measurement points are shown in Fig. 11. In this graph, the extended simulation results are also shown, following the same trend of overestimating the stiffness of the upper frame and underestimating the stiffness of the anvil. However, it can be expected that the overall deflection values fit those of the physical twin, indicating that version 2.0 is capable of handling the highest possible forces that this welding gun design can apply. Furthermore, the total deflection of 0,74 mm at 20 kN force is still below version 1.5.

The aforementioned observations can be confirmed once again by looking at the individual measuring points. Figure 12 shows the values of the deflection of, respectively, 8 measuring points on the upper frame and the anvil of each version. The significant reduction of the deflection both in the upper frame from 1.223 mm (version 1.5) to 0. mm (version 2.0) as well as in the anvil from approx. 0.7 mm (version 1.5) to approx. 0.3 mm (version 2.0) indicates that the first intuitive optimization of the welding gun has been successful. When comparing the individual points from simulation to experiment, it is also clear that the deviations in the simulation are greatest at those points where the maximum deformations are present. In order to determine which parts of the deflection originate from the linear rails and screws and which from the mere elastic deformation of the steel component, the anvil of version 1.5 was simulated separately. The results show that the elastic deformation accounts for approximately one-third of the deflection (0,238 mm at measuring point 9), which also indicates that there is unwanted clearance on the linear rails. Further investigations of the material properties of the materials used are necessary for the exact calibration of the DT.

In the simulations, investigations were also carried out on the relationship between the measurement inaccuracy when the measuring points were applied to the welding guns. For this purpose, various deviations up to 1 mm measurement inaccuracy were assumed and the resulting deviation of the deflection was analyzed using a path evaluation along the nodes. However, the resulting deviations in the positions of the measurement marks were in the range of a thousandth of a millimeter and therefore have no impact on the deviation of the simulation results. The influence of friction between the spindle tool and the anvil and possible deviations in the Young’s modulus of the welded frame components were also investigated. Varying these values has a very significant influence on the simulation results. However, without knowing the exact conditions and correlations, it is not expedient to adjust these values. For a highly accurate simulation, the DT would first have to be set up in even greater detail (consideration of weld seams and detailed design of the linear guides), and the friction values of the material pairings would have to be determined. Furthermore, the clearance of the bearings in the spindle and a possible non-linear behavior of the linear rails and carriages have to be investigated.

This paper introduced the design of a novel welding gun for friction stir stitch welding in two different design stages and investigated the deflection behavior of these two versions. A bearing concept especially fit for the load collective occurring with friction stir welding was presented, calculated and compared to conventional spindle bearing concepts. For design optimization, a DT of both versions of the welding gun was created as a simulation model of the entire assembly, which was subjected to the real force-time profile of the accompanying optical deflection measurements.

The results show a good correlation of the overall deflection between simulation and experiment, although there are individual deviations in the distribution of the deflections within the components of the welding gun. The overall deflection in the Z-direction could be reduced by a factor of 3 from the initial version 1.5 to the optimized version 2.0. The main decrease in deflection was observed in the upper frame (factor 5), indicating a design weakness in the structure and its mounting points to the rest of the welding gun. The stiffness of the anvil has been improved by a factor of 2,7 in the experimental measurements by using larger linear rails and screws and increasing the thickness of the part.

It can be concluded that the use of DT is generally suitable for design and stiffness analyses and their optimization, but the models require more detailed implementation when considering using the results for real design optimization.

The deflection behavior of the optimized version 2.0 was simulated in an extended simulation with the maximum force which the plunge motor can apply. With the total deflection being 0,74 mm, this design shows its ability to weld a wide variety of material combinations used in the automotive industry.

Whether the current deflection of version 2.0 still results in tunnel defects will be investigated in subsequent welding tests with accompanying X-ray and strength analyses of the resulting weld seams. Furthermore, in the future, the twisting of the frame through forces in the feed direction will be investigated. These investigations are carried out on the robot-guided version 2.0 of the welding gun, which can produce fully automated weld seams in any configuration and is shown in Fig. 13. Additionally, the potential of topology optimization will be investigated, especially optimizing the anvil and top frame parts using the existing simulation files in ABAQUS together with the integrated TOSCA solver.

This website uses services for operation, statistical analysis, display and sharing of content from social networks and interest-based advertising. These services make the website usable, improve your browsing experience, enable interactions with networks and show relevant advertising messages. Activate the desired services in the respective categories to accept them. More information

By clicking on "Submit" or "Select all and submit", you agree that your data will be processed by the corresponding services.

The activated services do not affect your browser configuration. Depending on the browser configuration, cookies from these services are deleted when the browser is closed or must be deleted manually if they are not to be valid beyond the specified time period.

By activating the respective services, you consent to us providing you with information and offers on social media, search engines, or display networks, potentially across devices. You can withdraw this consent at any time, free of charge and without formal requirements. Only the transmission costs according to basic rates apply (e.g., , fax, , postage).

For more robotic friction stir weldinginformation, please contact us. We will provide professional answers.