Friction stir welding

TWI Bulletin, January - February 2003

How big is the process torque? - a simple prediction

Simon Smith joined TWI in 1989 with a BSc in Mechanical Engineering and a PhD in computer modelling of creep crack growth. He is currently the TWI Technology Manager for Numerical Modelling. Over the last two years Simon has been heading the TWI research on modelling of FSW.

Rick Leggatt is Section Manager, Finite Element Analysis in TWI's Structural Integrity Department. He joined TWI in 1978 and his qualifications are MA, PhD (distortion in welded steel plates), Eur Ing, C Eng, MIMech E, SenMWeldI. Rick is responsible for TWI's finite element analysis activities and is involved in consultancy work on residual stress and distortion. His 24 years of specialist experience comes from close involvement in research projects and consultancy work.

Ruth Sanderson joined TWI as a project leader in the Finite Element Analysis section in July 1998. Previously she studied for a BSc (Hons) in mathematics at Warwick University. Her main focus has been the development of numerical methods for simulations of guided waves. She has published four papers on the topic and has recently won the ABAQUS young engineers' prize for one of the papers. Ruth is also currently involved in general stress analysis, weld modelling and fracture mechanics modelling.

Conventional rotational friction welding joins parts by spinning one part and forcing it into the other. The relative motion at the interface breaks up any obstacles to joining and generates heat. As Simon Smith, Rick Leggatt and Ruth Sanderson report high temperature permits material from both sides to form a good interface, but how do you predict the torque at the welding tool?

An early publication on conventional friction welding by Vill' in 1954 presented data from numerous tests conducted on low carbon steel. His results are presented in Fig.1 and 2. They show that an optimum weld speed exists. Welds take longer to produce at speeds above the optimum speed.

Vill' states that this appeared to be counter-intuitive to some workers. The optimum weld speed appears to be related to the power liberated as heat (see Fig 2) which drops continuously from 500revs/min to a minimum at the preferred welding speed. TWI (Nicholas and Ellis 1968) presented similar results in another report.

Studies have now been undertaken to investigate the process window for Friction Stir Welding (FSW) using the same method of making welds under a range of welding conditions, principally by varying the linear and rotation speeds of the weld.

Dawes et al (684/1999) presented the process window for 6mm thick 5083-O showing good welds can be made up to 600revs/min for a travel speed of 100 to 200mm/min. Johnson (716/2000) performed a more detailed study. He measured the torque, drag and transverse loads on welds in 6082-T6 and three other aluminium alloys. There seemed to be a trend in the results presented by Johnson that better welds were made at higher applied torque.

These data suggest trends that are fundamental to friction welding. It is important now to determine whether the behaviour of friction welds can be predicted using a simple synthesis of the process. This report provides a description of a simple model of FSW. The model is based on the heat flow from a moving heat source and the temperature dependence of the material strength under the FSW tool.

Elements of the method

Vill' showed that an optimum welding time exists for friction welding of low carbon steel. He thought this meant that there were two opposing mechanisms during friction welding. One mechanism dominates at low speed and the other dominates at high speed. Conditions where multiple mechanisms effect a result have been rationalised by Ashby and co-workers by generating what he called mechanism maps. For example, Fig.3 presents the deformation map for aluminium alloy 1100 on axes of shear strain rate versus temperature.

Ashby uses the basic physical equations of each process to find the dominant process as function of the axes of the map. A method of predicting the conditions that occur in FSWs is presented below which provides a possible basis for generating a mechanism map for the process. The basic physical equations also prove to be useful alone as a predictive tool.

FSW mixes material across the weld line. The process needs the material to be at the correct temperature. The material is too stiff when its temperature is low and voids form because the weld line is not fully closed. The material can also be too weak when it is very hot. At high temperatures the material can be thrown from the weld or a large flash can develop. The first mechanism of the model is therefore the temperature dependence of the material strength.

FSW mainly shears the material under the tool, so an important quantity is the shear yield strength of the material. For 6082-T6 the room temperature uniaxial tensile yield strength is around 240MPa and the melting temperature is about 520°C. For simplicity, the model will assume a shear yield strength that varies from 120MPa at room temperature to zero at 500°C.

To make use of the material strength assumption the model must have a prediction of the welding temperatures. Many models of FSW heat flow have been proposed. At TWI Mike Russell speaking at INALCO '98 has used the Rosenthal equations to predict the temperature accurately outside the tool shoulder.

The Rosenthal equation for heat flow from a moving heat source on a large, thick plate is therefore used in the model. A simple approach is adopted. It is assumed that the temperature predicted at the outer edge of the shoulder determines the strength of the material. The stirring of the material under the tool will tend to even out the temperature, so it is considered that the temperature at the edge of the tool is reasonably representative of the region under the tool. The temperature prediction could be more accurately evaluated, but this value is assumed to be sufficient to introduce the general approach.

The final element of the model is to couple the heat flow solution to the material strength. Expressing the applied torque as a function of the temperature dependent yield strength does this. The proposed approach assumes that the shear strength under the tool shoulder is equal to the material strength at the temperature predicted by Rosenthal for the edge of the tool shoulder. The shear strength is the shear resistance preventing tool rotation, so the torque is predicted by integrating the shear resistance across the tool shoulder.

A fully coupled model has therefore been developed. The heat input is a function of the material strength and the material strength is a function of the heat input.

Results

The full set of equations was processed in Mathcad. The result was equations for torque and power as a function of the material properties, the tool shoulder radius and the welding speeds (travel and rotation). The thermal properties assumed for 6082-T6 are 1.8x10 ⁷m ²/s for thermal diffusivity and 180W/m°C for thermal conductivity. The predicted power is plotted against the tool rotation speed in Fig.4. The figure shows that a feedback between heat generated and material resistance has been achieved. At speeds over 1000revs/min the power consumed becomes constant. The reason is seen in Fig.5. The torque developed during welding falls steadily from 500revs/min.

Figure 5 also shows results presented by Johnson which reveal a remarkable agreement between the model and the test data. Most of the tests were conducted at the same speed (1000revs/min). The variation between the results at this speed arises from the applied down force (values between 5.8 and 17kN were used). This variation can be accommodated in future developments of the model through a multiaxial yield criterion. This is likely to reduce the shear resistance of the material and reproduce the measured behaviour.

Discussion

A remarkably accurate model of FSW has been developed from a simple synthesis of the process. The torque needed for welding 6082-T6 was accurately predicted at 700, 1000 and 1400revs/min. A similar process could be used to predict the other forces (drag and transverse loads). The proposed method relies on loads arising from plastic deformation. Some detailed models of material movement during FSW have assumed that material resistance arises from viscosity (Bendzsak et al, 2000). It is clear that a viscous material law is likely to generate a different prediction of torque versus speed from those presented here. Research on the dominant deformation mechanisms is therefore needed. The research could be presented in the form of mechanism maps as used by Ashby to delineate deformation mechanisms in materials under pure loading conditions. A simple model, like the one presented in this article could be used to determine the temperature and strain rate near the FSW as a function of welding travel and rotation speeds. Existing results could then be used to present a map with travel speed on one axis and rotation speed on the other. The FSW deformation mechanisms would be presented as regions on the map. The FSW map could then be extended to delineate problematic areas such as when the weld is too cold or too hot.

The viscous-like processes may dominate the near tool behaviour, yet the material surrounding the viscous zone could be deforming plastically. This would be like a dust particle moving by convection inside a moving car. The particle has some motion arising from convection, but the dominant mechanism for the movement of the particle is the motion of the car. The equation method presented here could investigate the possibility that more than one material resistance mechanism occurs under an FSW tool. Plasticity occurring outside a viscous zone could still be correctly predicted by the synthesis presented in this article. The proposed FSW map would be able to reveal this possibility.

Conclusions

A simple model of FSW has produced an equation to predict the torque needed for FSW. The model has shown remarkable accuracy when compared to measurements made on 6082-T6 welded at 710, 1000 and 1400revs/min. Based on the simple method adopted here a FSW deformation mechanism map is proposed. The map would be of great use to FSW welding engineers.

References

For more information and detail of related papers please contact the authors at TWI.