Optimizing Friction Stir Welding via Statistical Design of Tool Geometry and Process Parameters C. Blignault, D.G. Hattingh, and M.N. James (Submitted February 22, 2010; in revised form May 12, 2011) This article considers optimization procedures for friction stir welding (FSW) in 5083-H321 aluminum alloy, via control of weld process parameters and tool design modifications. It demonstrates the potential utility of the ‘‘force footprint’’ (FF) diagram in providing a real-time graphical user interface (GUI) for process optimization of FSW. Multiple force, torque, and temperature responses were recorded during FS welding using 24 different tool pin geometries, and these data were statistically analyzed to determine the relative influence of a number of combinations of important process and tool geometry parameters on tensile strength. Desirability profile charts are presented, which show the influence of seven key combi- nations of weld process variables on tensile strength. The model developed in this study allows the weld tensile strength to be predicted for other combinations of tool geometry and process parameters to fall within an average error of 13%. General guidelines for tool profile selection and the likelihood of influ- encing weld tensile strength are also provided. Keywords destructive testing, quality control testing, welding 1. Introduction Previous investigations have shown that the measurement of tool forces, temperature and torque during friction stir welding (FSW) is a valuable aid in better understanding and character- ization of process dynamics (Ref 1-7). Such measurements facilitate optimized process control and efficient tool design, in turn leading to more efficient and faster welding which is characterized by reduced tool forces, torque and temperature and high mechanical property values. Once process data can be monitored and recorded reliably, it is possible to use statistical techniques to explore tool and joint optimization. The use of process forces as a statistical process control tool during FSW has also been previously highlighted by Arbegast (Ref 4). The present authors developed a rotating transducer that measures lateral forces during welding and that is attached directly to the spindle of a milling machine that has been converted to a FSW platform. The system records real- time measurements of process variables close to the tool and its main components are a sensing element, tool holder, telemetry system and data-logger as illustrated in Fig. 1. The design, development and calibration procedure of the system has previously been reported in the literature (Ref 1, 5, 7). The system allows real-time assessment of the influences of modifications to tool geometry and to changes in process parameters such as tool feed and speed, either continuously or at various points along the weld seam. Such information could greatly reduce the current high level of empiricism involved in choosing FSW parameters. Welding forces on the tool can be presented in the form of a lobed polar plot as a function of tool angle during rotation, whose area is related to energy input during welding. Figure 2 shows a typical polar plot of the bending force on the tool acting opposite to the direction of travel of the tool. This force is measured during each rotation of the tool, referenced to a particular position (in the present case the 270º position as shown in Fig. 2). It provides a 2D representation of the maximum and minimum bending force vectors experienced by the tool during each revolution. A similar polar plot can be obtained for the transverse bending force and hence a resultant force plot can also be produced. This ‘‘force footprint’’ (FF) has been proposed (Ref 5, 6) to present a graphical indication of aspects of the macroscopic plastic flow processes of entrain- ment, mixing and forging during tool rotation. It is believed that these effects are manifested in the area of the plots, and the magnitude and angular rotation of the lobe apogees. The FF C. Blignault—Deceased. C. Blignault, D.G. Hattingh, and M.N. James, Faculty of Engineering, the Built Environment & Information Technology, Nelson Mandela Metropolitan University, Port Elizabeth 6001, South Africa; and M.N. James, School of Marine Science & Engineering, University of Plymouth, Drake Circus, Plymouth, Devon PL4 8AA, UK. Contact e-mails: Danie.hattingh@nmmu.ac.za and mjames@ plymouth.ac.uk. Abbreviations FF force footprint FS friction stir FSW friction stir welding GRM general regression model GUI graphical user interface NRF National Research Foundation RSM response surface model UTS ultimate tensile strength NMMU Nelson Mandela Metropolitan University JMEPEG (2012) 21:927–935 ÓASM International DOI: 10.1007/s11665-011-9984-2 1059-9495/$19.00 Journal of Materials Engineering and Performance Volume 21(6) June 2012—927