Adv. Theor. Appl. Mech. , Vol. 6, 2013, no. 1, 13 - 26 HIKARI Ltd, www Adv. Theor. Appl. Mech., Vol. 6, 2013, no. 1, 13 - 26 HIKARI Ltd, www.m-hikari.com http://dx.doi.org/10.12988/atam.2013.222 Optimization of Friction Stir Welding Parameters for Joining Aluminum Alloys Using RSM Jawdat A. Al-Jarrah Mechanical Engineering Department, Faculty of Engineering Northern Border University, Saudi Arabia jawdatj@yahoo.com Sallameh Swalha Mechanical Engineering Department, Faculty of Engineering Northern Border University, Saudi Arabia On leave from: Mechanical Engineering Department Al-Balqa Applied University, Jordan sawalhasalameh84@yahoo.com Talal Abu Mansour Mechanical Engineering Department, Faculty of Engineering, Northern Border University, Saudi Arabia tabumansour@gmail.com Masoud Ibrahim Material and Chemical Engineering Department Faculty of Engineering, Northern Border University, Saudi Arabia On leave from: Industrial Engineering Department Fayom University, Egypt Ibrahim_64@yahoo.com Maen Al-Rashdan Mechanical Engineering Department, Al-Huson University College Al-Balqa Applied University, Jordan mmrashdan@yahoo.com

14 Jawdat A. Al-Jarrah et al Deya A. Al- Qahsi Industrial engineering Department, Faculty of Engineering Northern Border University, Saudi Arabia deya2941982@yahoo.com Copyright © 2013 Jawdat A. Al-Jarrah et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. Abstract The friction stir welding process (FSW) is a variant of the linear friction stir welding process in which the material is being welded without bulk melting. The FSW parameters such as tool rotational speed, welding speed, welding tool shoulder diameter, and welded plate thickness play a major role in determining the strength of the joints. A central composite rotatable design with four factors and five levels was chosen to minimize the number of experimental conditions. Empirical relationships were established to predict the yield tensile strength and the hardness of friction stir welded aluminum alloys by incorporating independently controllable FSW process parameters. Response surface methodology (RSM) was applied to optimize the FSW parameters to attain maximum yield strength of the welded joints. Keywords: Friction stir welding, Aluminum alloys, Response surface metho- dology, Optimization Nomenclature: FSW: Friction stir welding, N: Tool rotational speed, rpm V: welding speed, mm/sec D: Welding tool shoulder diameter, mm T: Plate thickness, mm YS: Yield strength, MPa Hv: Vickers microhardness RSM: response surface methodology

1. Introduction Optimization of friction stir welding parameters 15 Aluminum is the most prominent candidate to meet the challenges for future automotive regarding high strength/weight ratio, corrosion resistance, emissions, safety, and sustainability [Jayaraman,2009]. A high thermal and electrical conductivity cause problems in fusion and resistance welding of aluminum alloys[Guricic,2011]. Friction stir welding (FSW) is a solid state welding process and it considered the most significant development in metal joining techniques in the last decades, it was invented by The Welding Institute (TWI) of UK in 1991 as a solid-state joining technique, and it was initially applied to aluminum alloys[Kanwer,2011]. However, the extended application of this welding process in industry still requires accurate knowledge of the joining mechanism, and the metallurgical and mechanical transformations it induces in the base materials[Rodregues,2010]. Actually the effectiveness of the obtained joint is strongly dependent on several operating parameters[Jamshidi,2011]. First of all, the geometric parameters of the tool, such as the height and the shape of the pin and the shoulder surface of the head, have a great influence on both the metal flow and the heat generation due to friction forces[Frantinin,2010]. Secondly, both the rotating speed and the feed rate have to be selected in order to improve “nugget integrity” that results in a proper microstructure and eventually in good strength, fatigue resistance and ductility of the joint[Ericsson,2003]. In FSW process heat generated by friction between the surface of the plates and the contact surface of a special tool, composed of two main parts: shoulder and pin. Shoulder is responsible for the generation of heat and for containing the plasticized material in the weld zone, while pin mixes the material of the components to be welded, thus creating a joint[Elangovan,2007]. This allows for producing defect-free welds characterized by good mechanical and corrosion properties. The advantages of FSW are due to the fact that the process is carried out with the material to be welded in the solid state. Avoiding melting prevents the production of defects, due, for instance, to the presence of oxygen in the melting bath, and limits the negative effects of material metallurgical transformations and changes strictly connected with changes of phase. Finally, the reduced thermal flux, with respect to traditional fusion welding operations, results in a reduction in residual stress state in the joints and, consequently, in distortions in the final products [Fu,2012]. It is well known that the input welding parameters play a major role in determining the weld quality. As the process facts have not been disclosed so far, the selection of input parameters to join aluminum alloy is very difficult. Response Surface Methodology (RSM) is a collection of mathematical and statistical techniques useful for the modeling and analysis of problems in which a

2. Experimental Work = X1 X2 = X3 = X4 16 Jawdat A. Al-Jarrah et al response of interest is influenced by several variables and the objective is to optimize this response[Montgomery,2000]. The RSM is important in designing, formulating, developing, and analyzing new scientific studies and products. It is also efficient in the improvement of existing studies and products. In this investigation RSM will be used to reduce the number of experiments, in addition to build a numerical relation between the quality of welding and the welding parameters. 2. Experimental Work In this investigation, aluminum alloy sheets with thicknesses of 4, 5, 6, 7 and 8 mm were used. The sheet was cut to required size by power hacksaw cutting and followed by grinding to remove the burr. The main alloying elements in this alloy are 4.2 wt% Mg and 0.42 wt% Si. Butt joint configuration was used to fabricate the friction stir welds. The joint was initially obtained by securing the plates in position using mechanical clamps. A non-consumable tool made of high-carbon steel was used to fabricate the joints. Welding tools with flat cylindrical shoulder diameters of 18, 21, 24, 27 and 30 mm were used. A conventional milling machine was used as friction stir welding machine. where the rotational speed controlled to be 400, 700, 1000, 13000 and 1600 rpm. The welding speed also controlled to be 0.5, 1.0, 1.5, 2.0 and 2.5 mm/sec. A large number of trial experiments should be conducted to determine the working range of the above factors by varying one of the process parameters and keeping the rest of them at constant value. For the purpose of minimizing the experimental work, a simple and adequate experimental design named Response Surface Method (RSM), with the Box and Hunter is used [Box,2005]. In this study each parameter has five levels as mentioned above and shown in Table 1. According to a central composite-second-order rotatable design with 4 independent variables thirty one experiments were conducted with the combination of values that shown in Table 2. This consists of 16 corner points at ±1 level, 8 axial points at ±2 level and a center point at zero level repeated 7 times to estimate the pure error. The values of the levels of each FSW parameters used in this work were coded to simplify the experimental arrangements. The range of each parameter was also coded in five levels ( -2, -1, 0, 1, 2) using the following transformation equations: Rotational speed Welding speed Shoulder diameter Plate thickness N-1000 = 1 2 3 4 X1 300 V-1.5 X2 = 0.5 3D-24 X3 = T-61 X4

~~~~~ ~ ~~~~~ ~ ~~~~~ ~ ~~~~~~~ ~ ~~~~~~~ ~ ~~~~~~~ ~ ~~~~~~~ ~ Optimization of friction stir welding parameters 17 3 Developing an Empirical Relationship The out response surface micro-hardness and yield strength are a function of the FSW parameters such as a rotational speed, (N), welding speed (V), shoulder diameter, (D), and the plate thickness (T). The output response surface, y, expressed as: , , , 5 The second order polynomial (regression) equation used to represent the response surface, y, is given by ~~~~ ~~~ ~ ∑~~ ~~ ~∑~~~~~~ ~∑~~~~~~ 6 Where xi is the input parameter it can be one of N, V, D or T. and for factors the selected polynomial could be expressed as: ~ ~ ~ ~ ~~~~~~~~ ~~~~~ ~ ~~~~~~~ ~~~~~ ~ ~~~~~ ~ ~~~~~ ~ ~~~~~~~ ~ ~~~~~~~ ~ ~~~~~~~ ~ ~~~~~~~ ~ ~ ~~~~~~~ ~~~~~~~ ~ ~ 7 Where is the average of responses, and, ~~ , , , ..., are regression coefficients[Montgomery,2000] that depend on respective linear, interaction and squared term of factors. The values of coefficients were calculated using Matlab software. The multiple regressions can be expressed in a matrix form as ! "# $ 8 Where -& .&& .&' ... .&0 !& 5& 4 $& !' 5' $' ! ~ % * , " ~ , & .'& .'' ... .'0 3 , 5 ~ 6 8, $ ~ % * ⋮ , ⋮ ⋮ ⋮ ⋱ ⋮ 3⋮ ⋮ 9 Ynb ett xia xnq q Where y is an (nx1) vector of observations, X is an (nx(q+1)) matrix of level independent variable, b is an (qx1) vector of regression coefficient, and e is an (nx1) vector of random error[Bradley,2007]. If X is a (q x q) matrix, then the linear system y = Xb + e has a unique least squares solution given by bˆ = (X'X) -1 X' y . The estimated regression equation is : ; ~ "# 10 ! it can also be represented as B ~:~ ~ ~~ ~ < b>?xA? C 11 Using the observed values of the response as shown in Table 2, mathematical models which relate friction stir welding response (micro-hardness and yield strength) to friction stir welding parameters have been proposed. The less-significant coefficients were determined from further analysis using the student’s t-test, Table 3. Also, to check the adequacy of each model, the analyses of variance were carried out by using the F-ratio test, Table 4. The responses

18 Jawdat A. Al-Jarrah et al surface for the micro-hardness,(Hv) and yield strength,(YS) as functions of the four parameters used in this work are as follow: Hv = 53.472 - 1.975N + 0.967V - 0.758D + 0.283T - 0.307N2 - 0.382V2- 0.895D2 - 0.193T2 + 0.288NV + 0.012ND + 0.012NT - 0.025VD - 0.05VT - 0.075DT 12 YS = 92.571 - 3.583N + 0.667V - 3.083D - 2.75T - 0.58N2 - 1.045V2- 0.955D2 - 1.795T2 - 1.00NV + 1.375ND - 0.125NT - 0.125VD - 0.625VT - 0.000DT 13 Using the above models that were obtained by the RSM method, the relationships between the friction stir welding responses (micro-hardness and yield strength) and the significant variables are shown in Figs. 1 and 2. It is worth mentioning that each curve represents the effects of two input parameters while other two parameters were kept constant at level 0. In the following, the friction stir welding results will be discussed in terms of each of the friction stir welding parameters. In this study the welding parameters can be divided in two groups according to their contribution on heat generation during friction stir welding. The first group is the heat generated factors which are rotation speed and the shoulder diameter. Increasing the values of rotation speed or shoulder diameter lead to more heat induced in the welding region. The other group is the welding speed and the plate thickness, which reducing the rate of heat induced to the welding area during FSW process. So, the discussion of 3-D plot surface response will be built in mixing of these parameters to get a clear idea about the contribution of each factor to get free defect weld. Fig. 1 shows the 3D plot of surface response of the micro-hardness with welding parameters. The significant coefficients were found from T-test are used in micro-hardness regression model given in equation 12. Also the adequacy of model is tested using ANOVA analysis. It was found that the regression F-ratio of 1019.7 is significant. However, the F-ratio of the lack of fit is 1.44 which is insignificant. It has been found that the maximum micro-hardness happened at low rotation speed with high welding speed or with thick plates as shown in Figs. 1(a) and (c). Also, the maximum micro-hardness can be resulted from combination of low shoulder diameter with high welding speed or thick plates as shown in Figs. 1(e) and (f). Fig. 1(b) shows that the apex happened at a welding speed of 0.5 mm/sec and at shoulder diameter of 24 mm, while the rotation speed and plate thickness are kept at center values i.e 1000 rpm and 6 mm respectively. Fig. 2 shows 3D plots of the response surface for yield strength with different friction stir welding parameters. In yield regression model given in equation 4.10, the significant coefficients were used. The significant coefficients are resulted from T-test given in Table 4.2. Also the adequacy of model is tested using ANOVA analysis. It was found that the regression F-ratio of 16.93 is

Table 1: Welding Factors and their levels Optimization of friction stir welding parameters 19 significant, the F-ratio of the lack of fit is 1.86 which is insignificant. Also, in this model it was found that the interaction F-ratio of 2.19 insignificant. From Fig. 2 it was found that a general result for a plate thickness of 6 mm, the best combination to have maximum yield strength is 1000 rpm rotational speed with 1.5 mm/sec welding speed and a shoulder diameter of 24 mm. This result agreed with welding appearance quality and experimental results to have free defect joints. However, the quality of welded joints depends on controlling the rotation speed with welding speed to fill up the cavity behind the pin when moving forward. Fig. 2(a) shows that at high rotation speed, less effect of welding speed on yield strength, but it may affecting the appearance of the top joint due to excess of heat at low welding speed. Acknowledgment: Authors would like to express their sincere gratitude and thanks to deanship of research in Northern Border University, Kingdom of Saudi Arabia, for their supporting this project. Table 1: Welding Factors and their levels No. Factors Notations Units levels -2 lowest -1 low middle 1 high 2 highest 1 Rotational speed N rpm 400 700 1000 1300 1600 Welding speed V mm/sec 0.5 1.0 1.5 2.0 2.5 3 Shoulder diameter D mm 18 21 24 27 30 4 Plate thickness T 5 6 7 8

Table 2: Experimental design matrix and results 20 Jawdat A. Al-Jarrah et al Table 2: Experimental design matrix and results Exp. No. Coded factors Original Values Output response N V D I Hv YS (MPa) 1 -1 700 21 5 53.6 103 2 1300 49.2 96 3 55.2 109 4 51.7 95 27 52.4 6 47.8 90 7 53.8 98 8 50.3 94 9 54.6 97 10 49.7 91 11 55.8 12 88 13 52.9 14 48.4 86 15 54 16 50.8 85 17 -2 400 1.5 24 55.4 18 1600 50 83 19 1000 0.5 80 20 2.5 93 51.8 89 22 30 82 23 54.8 25 53.4 92 26 53.5 28 29 31

Optimization of friction stir welding parameters 21 Table3: Student's T-test* for the Two responses Coef. Value of coefficient Computed t- value Significant coefficient Hv YS bo 53.471 92.57 1324.2 122.01 b1 -1.975 -3.58 -90.6 -8.75 b2 0.967 0.667 44.3 1.627 0.667* b3 -0.758 -3.08 -34.8 -7.525 b4 0.283 -2.75 12.9 -6.712 b11 -0.307 -0.58 -15.4 -1.546 -0.58* b22 -0.382 -1.045 -19.1 2.783 b33 -0.895 -0.955 -44.8 -2.545 b44 -0.193 -1.795 9.6 4.781 0.193 b12 0.288 -1.00 10.8 -1.993 -1.00* b13 0.012 1.375 0.47 2.74 0.012* b14 -0.125 -0.249 -0.125* b23 -0.025 -0.94 -.025* b24 -0.050 -0.625 -1.87 -1.245 -.050* -0.625* b34 -0.075 -0.000 -2.8 -0.000* * The standard critical value of the t-test; t0.05, 26=2.06

Table 4: F-testa for the four responses studied in the present work 22 Jawdat A. Al-Jarrah et al Table 4: F-testa for the four responses studied in the present work Hv Source Sum of squares Degree of freedom Mean square F-ratio Regression 162.756 14 11.625 1019.7 Linear 131.77 4 32.9425 2889.7 Square 29.519 7.3796 647.3 Interaction 1.467 6 0.2446 21.456 Residual error 0.183 16 0.0114 Lack-of-fit 0.128 10 0.0128 1.422 Pure error 0.054 0.009 Total 162.939 30(=N-1) YS Sum of squares Degree of freedom Mean square 955.08 68.22 16.93 728.50 182.125 45.20 173.58 43.396 10.77 53.00 8.833 2.19 64.46 4.029 48.75 4.875 1.86 15.71 2.619. 1019.55 a) The standard valued of F- ratio for the significance level α =0.05 and degrees of freedom 3 and 5 is F0.05(3,5)=5.4, at degree of freedom 6 and 5 is F0.05(6,5)=5.0 and at degree of freedom 5 and 5 it is F0.05(5,5)=5.1.

Optimization of friction stir welding parameters 23 60 60 Micro-hardness, Hv Micro-hardness, Hv 55 55 50 50 45 45 40 40 2 2 1 2 1 2 1 1 -1 -1 -1 Welding speed Rotational speed Shoulder diameter -2 -2 -2 -2 -1 Rotational speed (a) (b) 58 56 Micro-hardness, Hv 56 Micro-hardness, Hv 54 54 52 52 50 50 48 2 48 2 1 2 1 1 2 1 -1 -1 -1 Plate thickness Rotational speed Plate thickness -2 -2 -2 -2 -1 Welding speed 2 (c) (d) 55 56 Micro-hardness, Hv Micro-hardness, Hv 54 50 52 50 45 48 2 2 1 2 1 1 1 -1 -1 Shoulder diameter -2 -2 Welding speed Shoulder diameter -2 -2 Plate thickness Shoulder diameter -2 Welding speed (e) (f) Plate thickness Fig.1: 3D plots of the response surface for micro-hardness with different friction stir welding parameters -1 -1

24 Jawdat A. Al-Jarrah et al 100 Yeild strength, MPa 95 110 90 Yeild strength, MPa 100 85 90 80 80 75 2 70 1 2 2 1 1 2 -1 1 -1 -1 -1 Welding speed -2Rotational speed Shoulder diameter -2 -2 -2 Rotational speed (a) (b) 100 95 Yeild strength, MPa 90 Yeild strength, MPa 90 80 85 80 70 75 60 2 70 1 2 2 1 1 2 1 -1 -1 -1 -1 Plate thickness Rotational speed Plate thickness -2 -2 -2 -2 Welding speed (c) (d) 95 Yeild strength, MPa 90 100 Yeild strength, MPa 85 90 80 80 75 70 70 2 60 1 2 2 1 1 2 -1 1 -1 -1 -1 Plate thickness -2Welding speed Plate thickness -2 -2 -2 Shoulder diameter (e) (f) Fig.2: 3D plots of the response surface for yield strength with different friction stir welding parameters

References Optimization of friction stir welding parameters 25 D. M. Rodregues, C. Leitao, R. louro, H. Gouveia, and A. Loureiro, High speed friction stir welding of aluminum alloys, Science and tech. of welding and joining, 15(2010), 676-681. G.E.P. Box, W. Hunter and S.J. Hunter, Statistics for Experimenters: Design, Innovation ,and Discovery , 2nd Edition, Wiley (2005) H. Jamshidi, , S. Serajzadeh, A. kokabi, Theoretical and experimental investigation into friction stir welding of AA 5086, Int. J. Adv. Manuf. Techn., 52(2011), 531-544. K. Elangovan, V. Balasubramanian, Influences of tool pin profile and tool shoulder diameter on the formation of friction stir processing zone in AA6061 aluminum alloy, Center for materials joining research, 2007. L. Fratinin, G. Buffa, and R. Shivipuri, Mechanical and metallurigical effect of in process but joints, Acta materialai, 58(2010), 2056-2067. M. Ericsson and R. Sandstrom, Influence of welding speed on the fatigue of friction stir welding and comparison with MIG and TIG, International Journal of Fatigue, 25(2003), 1379-1387. M. Guricic, G. Arakera, B. Pandurangan, A. hariharan, C. F. Yen, and B. A. chesseman, Development of a robust and cost-effective friction stir welding process for use in advanced military vehicles, J. of material engineering and performance, 20(2011),11-23. M. Jayaraman, R. Sivasubramanian, V. Balasubramanian, and S. Babu, Optimization of friction stir welding process parameters to weld cast aluminum alloy A413- an experimental approach, Intr. J of cast metal research, 22(2009), 367-373 M.L. Santella, T. Engstrom, D. Storjohann, T.Y. Pan, Effects of friction stir processing on mechanical properties of the cast aluminum alloys A319 and A356, Scripta Materialia 53(2005), 201–206. Montgomery, C. Douglas, Design and analysis of experiments: response surface method and design, New Jersey: John Wiley and Sons, Inc, 2000. N. Bradley, The response surface methodology, Master Thesis, Indiana University South Bend, mathematical Department, 2007.

26 Jawdat A. Al-Jarrah et al R.D. Fu, R.C. Sun, F.C. Zhang, and H.J. Liu, Improvement of formation quality for friction stir welded joints, Welding journal, 1(2012), 169-173. R.S. Mishra, Z.Y Ma, Friction stir welding and processing, Materials Science and Engineering, 50(2005), 1–78. S. Kanwer, Arora, S. pandey, M. Schaper, and R. Kumar, Effect of process parameters on friction stir welding of aluminum alloy 2219-T87, Int. J. adv. Manu. Tech., 50(2010), 941-952. Received: February 12, 2013