Subscribe to our newsletter to receive the latest news and events from TWI:

Subscribe >
Skip to content

Development of a low cost Friction Stir Welding Monitoring System

Kathryn Beamish, Andy Ezeilo and Simon Smith

TWI Ltd, Granta Park Great Abington, Cambridge, CB21 6AL, UK

Peter Lewis
Applied Measurements Ltd UK

Peter Cheetham
Sigmapi Systems Ltd UK

Paper presented at 6th International Symposium, Friction Stir Welding, Saint-Sauveur, Canada, 10-13 Oct.2006.


Industrial take up of friction stir welding has been limited to a few industry sectors that have sufficient capital to invest in the high facility and technology costs. The need for purpose built FSW machines, can make it difficult for product manufacturers to justify implementation of the technology. An alternative approach is to adapt milling machines. However, standard milling machines lack the process monitoring capabilities required to ensure high quality friction stir welded joints. LOSTIR, a project part funded by the European Commission, has developed a low cost FSW monitoring system that can be retro fitted to milling machines to facilitate their application to FSW.

The LOSTIR device consists of 4 components; a general purpose FSW tool; a tool holding system; a weld monitoring system; and a display software system. It is targeted at those wishing to join 2-8mm thickness aluminium alloys. The device specification is based on existing process information and current FSW machine specifications. To gain maximum benefit from such a system, there is a need to link process parameters, rotational speed, downforce, torque etc. and the quality performance of the welded joint. As such the system development has been supported by a programme of materials property determination and the appropriate use of advanced modelling techniques.

This paper describes the development of the LOSTIR device and a review of its key features and capabilities.

1. Introduction

Friction Stir Welding (FSW) has significant advantages over other joining techniques including good mechanical properties, low distortion, and an ability to weld some materials that cannot be welded by other methods. The development and take up of FSW over the last decade has been significant. Most current uses involve the joining of Al alloys, for applications including: airframes and aircraft components, ship decking and structures, rail carriages, automotive components, bridge components, pressure vessels, and space launch systems. In addition to Al components, development of FSW has recently been reported for the joining of magnesium, copper, steels, and titanium alloys (1-3).

However, small businesses wishing to introduce FSW are faced with large set-up costs as equipment costs for new purpose built FSW machines are significant and as such the application of FSW is currently limited to industry sectors where investment in the technology has been supported. It is well known that standard milling machines can perform FSW operations. However there is a lot of uncertainty regarding the quality of the joints produced as there is little correlation between the milling machine parameters (torque, forces, feeds etc) and the quality of the joint (defects, heat affected zone, residual stresses, distortions etc.). For example, Fig. 1 shows the possible effect of a low down-force on the FSW tool on the work-piece resulting in a flaw beneath the weld surface. This highlights the need to know these forces during the joining process.

Fig. 1 - Visually acceptable weld containing a large central void caused by a lack of down-force. This form of welding defect can be avoided by measuring the down-force during the process, and ensuring that it remains within certain pre-defined limit
Fig. 1 - Visually acceptable weld containing a large central void caused by a lack of down-force. This form of welding defect can be avoided by measuring the down-force during the process, and ensuring that it remains within certain pre-defined limits.

Basic milling machines are not able to measure and control these parameters, and, as such, the FSW joint quality produced by these systems cannot be assured. Where a monitoring unit is retrofitted to milling machines they tend to be very expensive as they are rarely designed for friction stir welding but adapted from cutting machines. At the other end of the spectrum, purpose built FSW machines are available and are currently being used in production. These machines overcome the uncertainties by incorporating a range of force measuring and process control systems.

Users of the low cost system will be able to evaluate FSW with low start up costs and minimise their financial risks. This will make the evaluation and implementation of FSW for existing and new products more viable. In addition because friction stir welding can join dissimilar materials, in a way which conventional processes cannot achieve, the LOSTIR device will also be able to influence design and development of new products.

2. Requirements of milling Machines for FSW

The aim of the LOSTIR project was to develop a FSW device that would fit a broad range of milling machines, taking into account the architectural, structural and operational constraints of these milling machines. An outline specification was drafted based on current knowledge of aware of the process conditions necessary in order to achieve a satisfactory FSW in 2-8mm thickness aluminium alloys. These are stated in table 1.

In addition a study was conducted of up to 100 milling machines in terms of their ability to meet the criteria in table 1. Selecting a production milling machine for a use other than that for which it was designed will always involve compromises. Milling machines are normally targeted at specific applications, which govern machine architecture and the maximum workpiece size. For example, the standard universal head can only position the tool either horizontally or vertically. A multiple position milling head or a tilting worktable are required to achieve tool tilt unless using a full CNC system. Machine characteristics including forces, torque, spindle speed and traverse speed were examined to determine suitability for friction stir welding. While modern machining centres may not have the required rotation and traverse speed ranges or may not have the required torque characteristics at high rotation speed these factors can be largely overcome by fitting suitable motors, bearing etc. Machine load (Fz) was identified as the key parameter in machine selection. For most machines the maximum admissible load is in the range 10-20kN. The stiffness of the machine directly affects joint quality, the higher the cantilever distance of the welding head is, the lower the stiffness. Gantry type machines offer maximum stiffness but are not practical propositions for most firms. Older traditional milling machines, available on the second-hand market, offer a more practical alternative.

Table 1 Machine Specifications for Friction Stir Welding

ParameterSpecified Range
Spindle speed range 0 - 3000 rpm
Z axis traverse speed 0 - 1500 mm/min
X axis traverse speed 0 - 3000 mm/min
Z axis travel 500 mm
Z axis max workpiece size 750 mm
X axis travel 2000 mm
Y axis travel 2000 mm
Spindle tilt angle 0 - 5º
Z axis load 0 - 30 KN
X axis load 0 - 20 KN
Spindle torque 0 - 80 N-m

3. Components and features of the LOSTIR device

The Lostir device is made up of 4 components; a force monitoring system, a FSW tool holder, a control and logging system and FSW tools appropriate for low downforce FSW. The LOSTIR device is shown in Fig. 2, and the specification given in table 2.

Fig. 2 The LOSTIR Device
Fig. 2 The LOSTIR Device

Table 2 System Specification

Material to be joined (for the purposes of system validation) 2-8mm thickness aluminium alloys
Weld parameters to be monitored and recorded Down force Fz
Lateral force Fx
Torque Mz
2 thermocouple ports
Rotation speed 0-3000rpm
Traverse rate 3000mm/min max
Down Force Fz 50kN
Lateral Force Fx 25kN
Torque Mz 10-100Nm
Operating temperature Max 60C at base of weld monitoring system
Max 400C at base of tool holding system
Temperature measurement 1 port provided for thermal cut out 1 port provided for in-process temperature measurement of tool/shoulder
Machine interface ISO 50 taper
Size Measuring device: 150mm dia. X 100mm Tool holder:
Measuring accuracy ±5%
Data transfer Scan at 10Hz
Record at 1Hz
Auto-log desirable
RS485 data stream
Tool material H13 tool steel

3.1 General Purpose FSW Tool

Friction stir welding tools have been developed capable of producing good quality friction stir welds on commercial milling machines for the materials investigated using the measurement system to provide process monitoring. The tools have been designed to operate within the target range of torques, forces and processing rates that can be achieved by standard milling machines. The force exerted during the welding process is influenced by the dimensions and profile of the tool shoulder. The FSW tool has a profiled shoulder to permit welding at 0 tilt, negating the need for a specialised head or tilting table. However, the use of a plain concave shoulder profile and appropriate tool tilt angle can reduce downforce by 30%. Probe features, which control material movement through the section of the weld, and hence weld quality, do not strongly influence the forces and torques developed. However, the addition of appropriate tip features can reduce downforce as a good quality weld can be achieved with the tip of the tool probe a greater distance from the backing bar.

3.2 Tool Holding System

A tool holding system has been designed to securely hold the FSW tool during the welding process. In addition the Tool Holder has been designed to dissipate as much heat as possible in order to protect the sensitive electronics and sensor equipment mounted in the weld monitoring housing from the sever temperatures experienced by the FSW tool. Finite Element Analysis was used to support the design requirement. Cooling fins and a ceramic heat shield were introduced to ensure effective shielding of the electronics. Subsequent tests on the device under extreme loading conditions confirmed that the final design was satisfactory.

3.3 Weld Monitoring System

A weld monitoring system has been developed to accurately measure the vertical and horizontal forces and torque on the tool. The sensor is machined from one piece of high grade stainless steel, heat treated for maximum strength and stability. The sensor design allows for various taper sizes to be attached to accommodate the requirements of the user. The data gathered can be directly linked to the acceptance or otherwise of the weld. In addition, the device has the capability to monitor two user defined temperatures via thermocouples, one to be attached to the FSW tool, the second will act as a safety cut out to protect the integral telemetry circuit, monitoring the temperature at the interface between the tool holder and the weld monitoring system. All of the power requirements for the sensor are transmitted using wireless digital telemetry, this also supports two way communication for calibration and data collection. The sensor has an annular rotating antenna, with the information being transmitted to a static calliper style coupling module. In addition there is a separate housed electronics module that processes the information into a suitable format for interfacing with a PC.

3.4 Display Software

The information gathered by the Lostir device is displayed to the operator in a clear and straightforward manner using a laptop PC running Labview, a sample display screen format is shown in Fig. 3. The instrument panel displays real time numerical values of forces, torque, the temperature adjacent to the system electronics and (if desired) the tool temperature. The system also has the capability to add real-time event markers to allow correlation between process conditions/stages and the recorded data. The main display screen has buttons to start and stop recording of data. Alternatively an automatic trigger facility exists for initiating the recording of data. The display also shows the current captured data values for the weld in progress indicating whether they are within the acceptable range for satisfactory welding. The display also has a multi-graph facility where the user can select which sensor values are displayed.

Fig. 3 Example display screen showing the screen format
Fig. 3 Example display screen showing the screen format

4. Numerical modelling for process parameter selection

The task of establishing suitable FSW process parameters for new and existing applications is usually a costly and lengthy exercise involving an extensive test programme. The ability to simulate the FSW process using analytical and numerical modelling techniques for selecting process parameters introduces significant opportunities for investigating and optimising FSW process parameters. In order to support the development of suitable models a programme of work was conducted using two alloys, each alloy representing a different sector of the FSW market: AA2024 (Aerospace), and AA6082 (Extrusion and sheet fabrication). Appropriate materials property data for modelling purposes were obtained. In addition a range of FSW test welds were made some using acceptable welding parameters and some outside of the acceptable range to identify the conditions for defect formation, again to support the validation of the models.

A combination of mathematical and computer-based, numerical modelling was used during the Lostir project to predict the forces, torques and temperatures associated with the range of welding procedures represented in table 2.

Fig. 4 Representative compression stress-strain curves for AA6082-T6
Fig. 4 Representative compression stress-strain curves for AA6082-T6

The scope of work was designed to focus the development of suitable results and modelling techniques for the practical analysis of FSW. Materials property data were generated using a combination of hot compression tests, elevated temperature tensile tests and split Hopkinson’s bar tests. Representative compression stress-strain curves for AA6082-T6 are shown in Fig. 4. However, the modelling methods work reasonably well using easily obtained properties such as room temperature values for thermal conductivity and material yield strength, together with the melting point of the alloy. This means that models developed could be utilised by Lostir users in the future without the need for the measurement of a wide range of materials data. A couple of sample predictions are shown in Figs. 5 and 6. Figure 5 shows the predictions of welding power made by mathematical expressions for heat flow and torque based upon simple mechanics and steady state heat flow. The predictions are compared with measurements made during the Lostir project, and a good agreement is observed. The results of Fig. 6 were generated using the Computational Fluid Dynamics (CFD) software FLUENT (4). Data such as that presented in Fig. 4 can be used in the CFD model however the results in Fig. 6 use a simple material property set as input data to the model. The CFD model predicted the welding torque with accuracy similar to the mathematical model, but it also predicted the material flow patterns around the tool. The model results will be combined with weld quality parameters from the tests to provide a predictive system for weld quality.

Fig. 5 Comparison of welding torque weld data (symbols) and predictions from mathematical expressions for AA2024-T3 (grid) for a range of welds
Fig. 5 Comparison of welding torque weld data (symbols) and predictions from mathematical expressions for AA2024-T3 (grid) for a range of welds
Fig. 6 Prediction of temperature contours from the CFD, FLUENT model for a AA6082-T6 weld made at 1737rpm and 2450mm/min
Fig. 6 Prediction of temperature contours from the CFD, FLUENT model for a AA6082-T6 weld made at 1737rpm and 2450mm/min

5. Weld monitoring as a quality indicator

Instrumented welds were made to simulate five key quality control problems: tool wear, lack of penetration, joint line gap, and void formation. All welds were made using an MX Triflute™ tool manufactured from AISI-H13 tool steel. Baseline welding conditions for the two alloys were: AA6082 rotation speed 100rpm, traverse speed 1000mm/min, tool tilt 2.5°; AA2024, rotation speed 350rpm, traverse speed 210mm/min, tool tilt 2.5°. A series of welds was made to assess the error level that resulted in a visible defect when a metallographic section was examined by optical microscopy (table 3).

Table 3 Impact of quality issues

Quality IssueAlloyCritical error levelDefect
Tool wear 6mm AA6082-T6 Reduction in pin length of 0.5mm Root flaw
Lack of penetration 6mm AA6082-T6 Reduction in plunge depth of 0.5mm Root flaw
Joint line gap 6mm AA6082-T6 1.8mm Submerged void
Void formation (incorrect tool tilt) 6mm AA6082-T6 2° (-0.5°) Submerged void
Tool wear 4mm AA2024-T4 Reduction in pin length of 0.25mm Root flaw
Lack of penetration 4mm AA2024-T4 Reduction in plunge depth of 0.15mm Root flaw
Joint line gap 4mm AA2024-T4 1.8mm Submerged void
Void formation (incorrect rotation speed) 4mm AA2024-T4 375rpm Surface breaking void

The effect of these process variations on the weld signature (forces and torque) and mechanical properties of the welds has been assessed. The impact of incorrect welding procedure is clearly visible when downforce data is compared. Similar, but less marked, differences were visible in the measured torque and traversing force data. Figure 7 compares weld downforce data for AA2024 aluminium alloy. A worn tool generates less force reaction from the backing bar. When a reduced plunge depth is used the distance between pin and backing bar increase resulting in the same reduced downforce as a worn tool. An additional reduction in downforce is observed, as the tool shoulder is not in full

Fig. 7 Influence of incorrect welding parameters on weld downforce for 4mm AA2024-T4
Fig. 7 Influence of incorrect welding parameters on weld downforce for 4mm AA2024-T4

contact with workpiece. The presence of a joint line gap of 1.8mm resulted in the largest reduction in downforce. The joint line gap was simulated by removing pockets 0.9mm x 440mm long starting 100mm from the edge of the plate, with support upstands of 10mm length between the pockets. The spikes in the data correspond to welding through these supports. The selection of an increased tool rotation speed also showed a reduction in measured downforce. More work is required to determine if this is a true response as values differed only by approximately 3kN.

6. Conclusions

A device has been designed and manufactured which provides accurate monitoring of forces, torque and tool temperature during friction stir welding. Weld monitoring can be used as a weld quality indicator. Quality issues such as tool wear, lack of tool penetration and an excessive joint line gap can all be identified using the weld downforce signature. In addition it is now possible to use modelling of the FSW process to assist in feasibility and joint quality investigations.

7. Acknowledgements

The work described in this paper originated from a collaborative project part funded by the European Commission. The authors would therefore like to thank the European Commission for providing this funding as well as the consortium partners as follows; Petr Belsky and Vaclav Kafka of the Aeronautical Research and Test Institute of the Czech Republic for conducting all material properties tests; Olga Mishina and Seppo Tuovinen of Sapa Technology Sweden and Andrew Wescott of BAE Systems UK for supplying raw materials and industrial applications advice; Adam Pietras of Instytut Spawalnictwa Poland for conducting all structural integrity tests, Asun Rivero of Fatronik System Spain for designing and manufacturing the FSW Tool Holder, Paul Wilson of Suffolk Precision Ltd UK for manufacturing the FSW tools used in the project and finally Mike Russell of TWI for his input and direction in the technical programme. The authors would also like to thank Mike Slack of FLUENT for his technical support.

8. References

  1. Kallee S W, Nicholas E D and Thomas W M: ‘Friction stir welding – invention, innovations and applications’. Tube International, Vol. 21, No. 117, July-August 2001.
  2. Dawes C J and Thomas W M: ‘Development of improved tool designs for friction stir welding of aluminium’. Paper presented at 1st Symposium on Friction Stir Welding, Thousand Oaks, CA, USA, 14-16 June 1999.
  3. Nicholas E D and Kallee S W: ‘Friction stir welding – a decade on’. IIW Asian Pacific Welding Congress. Proceedings, NZIW 2000 Annual Conference and WTIA 48th Annual Conference, Melbourne 29 October-2 November 2000.

For more information please email: