Scholarly article on topic 'Multi Objective Optimization of Friction Stir Welding Parameters for Joining of Two Dissimilar Thin Aluminum Sheets'

Multi Objective Optimization of Friction Stir Welding Parameters for Joining of Two Dissimilar Thin Aluminum Sheets Academic research paper on "Materials engineering"

CC BY-NC-ND
0
0
Share paper
Academic journal
Procedia Materials Science
OECD Field of science
Keywords
{"Friction stir welding" / "Tailored blanks" / "Multi objective optimization" / "Taguchi-grey method."}

Abstract of research paper on Materials engineering, author of scientific article — R.K. Kesharwani, S.K. Panda, S.K. Pal

Abstract The article presents multi objective optimization of parameters affecting weld quality in tailored friction stir butt welding of 2.0mm thin dissimilar sheets of AA5052-H32 and AA5754-H22 using Taguchi grey based approach. The L9 orthogonal array has been used to design the experiments, and the experiments have been conducted in a laboratory stage vertical milling machine by varying tool rotational speed, worktable translational speed, tool shoulder diameter and tool pin geometry. After welding, the weld strength and percentage elongations have been evaluated using uniaxial tensile test. Based on the experimental data, empirical relations among the parameters correspond to each output feature has been developed using simple regression method. Optimum levels of parameters have been identified using grey relation grade.

Academic research paper on topic "Multi Objective Optimization of Friction Stir Welding Parameters for Joining of Two Dissimilar Thin Aluminum Sheets"

Available online at www.sciencedirect.com

ScienceDirect PrOCSd ¡0

Materials Science

Procedia Materials Science 6 (2014) 178 - 187

www.elsevier.com/locate/procedia

3rd International Conference on Materials Processing and Characterisation (ICMPC 2014)

Multi Objective Optimization of Friction Stir Welding Parameters for Joining of Two Dissimilar Thin Aluminum Sheets

R. K. Kesharwania*, S. K. Pandab, S. K. Palc

a,b,cIndian Institute of Technology Kharagpur, Kharagpur-721302, India

Abstract

The article presents multi objective optimization ofparameters affecting weld quality in tailored friction stir butt welding of 2.0 mm thin dissimilar sheets of AA5052-H32 and AA5754-H22 using Taguchi grey based approach. The L9 orthogonal array has been used to design the experiments, and the experiments have been conducted in a laboratory stage vertical milling machine by varying tool rotational speed, worktable translational speed, tool shoulder diameter and tool pin geometry. After welding, the weld strength and percentage elongations have been evaluated using uniaxial tensile test. Based on the experimental data, empirical relations among the parameters correspond to each output feature has been developed using simple regression method. Optimum levels of parameters have been identified using grey relation grade. © 2014ElsevierLtd.Thisisanopenaccessarticleunder the CC BY-NC-ND license (http://creativecommons.Org/licenses/by-nc-nd/3.0/).

Selection and peer review under responsibility of the Gokaraju Rangaraju Institute of Engineering and Technology (GRIET) Keywords: Friction stir welding; Tailored blanks; Multi objective optimization; Taguchi-grey method.

CrossMark

1. Introduction

Since last few decades, the increase in demand of automobiles and shortage of materials and fuels forced engineers and researchers to find new light weight materials and advanced manufacturing processes to reduce the vehicle weight. Presently, tailor welded blanks (TWBs) are widely being used in body in white (BIW) structure to reduce the weight and manufacturing cost without compromising on strength, rigidity and crashworthiness. A TWB is fabricated using two or more sheet metals of different thickness, shape, material and/or coating by welding with each other in a single plane before forming [Kinsey and Wu (2011)]. Aluminum alloys with magnesium as the major alloying elements constitute a group of non heat treatable alloys (5XXX series) with medium strength, high ductility and excellent corrosion resistance. This wrought Al-Mg alloys are used as structural materials in automotive, aircraft, and marine applications. While welding this alloy using conventional welding techniques; vaporization of Mg takes place. This cause compositional changes in the weld region [Rooks (2001); Anon (1994)].

* Corresponding author. Tel.: +91-8900162924; fax: +91-3222-282996. E-mail address: ramkumar31kesharwani@gmail.com

2211-8128 © 2014 Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.Org/licenses/by-nc-nd/3.0/).

Selection and peer review under responsibility of the Gokaraju Rangaraju Institute of Engineering and Technology (GRIET) doi: 10.1016/j.mspro.2014.07.022

Hence, friction stir welding (FSW) may be the feasible solution to overcome this problem and may retain the Mg content on weldaments as it is a low heat input solid state welding process.

FSW was invented at The Welding Institute (TWI), UK in 1991 [Thomas et al. (1991)]. In contrast to conventional friction welding where two objects which are to be joined are rubbed with each other with high contact pressure, in FSW a third body is rubbed against two firmly clamped objects to be joined in the form of a non-consumable rotating tool having probe at the tip. This tool is plunged into the joining region of two objects and subsequently translated along the joining line till the end and ultimately plunged out from the region [Mishra & Ma (2005); Nandan et al. (2008)]. At one side where tool rotation direction remains same as the welding direction is called advancing side (A.S.) and the other side is called retreating side (R.S.). Further in FSW the microstructures are broken up into four distinct zones - (a) base metal (BM), (b) heat affected zone (HAZ), (c) thermo mechanically affected zone (TMAZ) and (d) nugget zone (NZ). However, depending upon the peak temperature reached during welding sometimes HAZ may remain absent [Threadgill (1997)] .

It is important to note that application of aluminum alloys in fabrication of tailored welded blanks reduces vehicle weight. However, while fabricating these blanks using FSW technique mainly tool rotational speed, tool translational speed (along the joining line), plunge force and tool geometry play very vital role in producing good quality joints which can withstand the deformation without premature failure during post weld forming [Arora et al. (2009); Arbegast & Hartley (1998); Cavaliere et al. (2006); Kim et al. (2006); Hirata et al. (2007); Cavaliere et al. (2008); Ren et al. (2007); Arora et al. (2011); Kumar et al. (2012)]. Hence, in the present work optimizations of these parameters have been done followed by transverse tensile test.

2. Experimental Procedure

AA5052-H32 and AA5754-H22 sheets of 2.0 mm thickness have been used for fabrication of dissimilar metal tailored friction stir butt welded blanks, because of their wide applications in automotive car body panel and marine industries etc. Sheets of dimensions 200 mm X 100 mm were cut from bulk sheet volume using band saw such that maximum dimension coincides with the rolling direction, and the cut edges were filed to make them smooth. In order to remove oxide layers the faying surfaces were rubbed using silicon carbides emery paper (grade P400), and subsequently acetone was applied for cleaning of the rubbed surfaces.

A column and knee type vertical milling machine (BFW, VF 3.5) having 18 numbers of rotational speeds of spindle (ranges 35.5-1800 rpm), 21 numbers of feeds of worktable (ranges 16-1600 mm/min for longitudinal and transverse, and 4-400 mm/min for vertical) and 11 kW of maximum power capacity; was used for the FSW. A high carbon steel plate was used for backing plate and stainless steel (SS)-316 was used for FSW tool. Elemental compositions of the working sheet metals and tool material are shown in the Table 1 and 2, respectively.

Table 1. Elemental compositions (in wt %) of aluminum alloys used in the present experiment.

Materials Mg Si Mn Cr Fe Cu Zn Ti Al

AA5052-H32 AA5754-H22 2.63 2.54 0.118 0.11 0.061 0.083 0.212 0.224 0.27 0.234 0.051 0.03 0.025 0.037 0.0041 0.0095 Bal. Bal.

Table 2. Elemental composition (in wt %) of SS-316 FSW tool used in the present experiment.

Elements C Cr Mn Mo Ni P S Si Fe

Wt. % 0.06 16.27 1.87 2.22 11.60 0.027 0.018 0.66 Bal.

After a number of trial experiments three different tool rotational speeds (1120, 1400, 1800 rpm), three different worktable translational speeds (50, 125, 200 mm/min), three different tool shoulder diameter (12, 15, 20

mm), and three different tool pin profiles (circular, triangular and square) were selected for carrying out further FSWs. All FSW tools were designed with a pin height of 1.8 mm keeping the shoulder surfaces flat for all the welding experiments. However, the diameter of the circular tool pins was selected to be one third of the shoulder diameter as per the suggestion of Arora et al. (2011). The square and triangular pins were designed such that the static volume was equal to that of the circular pin for a particular shoulder diameter (as shown in Fig. 1 a). All the welding was done keeping the FSW tool shoulder perpendicular to the sheet faying surface (to be welded) without tilting the spindle. That means the tool axis was in line with the faying surfaces. A constant tool shoulder plunge depth (equal to 0.1 mm) was applied during the FSW experiments. In initial trial experiments, it was observed that the harder sheet metal (AA5052-H32) should be kept at the retreating side for achieving visual defects free welds like hot cracking etc. Figure 1b shows a defective weld, which was achieved during FSW when AA5052-H32 was kept at A.S. Hence, the AA5052-H32 sheet was kept at the R.S. during all the dissimilar FSWs. All the samples were weld along the rolling direction of the sheet metals. Taguchi L9 orthogonal array was applied to reduce the number of experiments. So it was four factors, three levels design. Factors and their levels are shown in Table 3. Tool rotational speed, worktable translational speed, tool shoulder diameter, and tool pin profiles have been nomenclatured as factor A, B, C and D, respectively. The pin profile (factor D) has been levelled according to the numerical values obtained from ratio of tool pin dynamic volume to the static volume. The detailed experimental design with factors level and their combinations are shown in Table 4. For each set of combination of parameters two welding replicates have been conducted. From each welded sheets two tensile test specimens were cut. Hence, for each set of combination of parameters there were four results. Out of which best of three results were taken for the analysis. Each set of FSW experiment has been nomenclatured as 'FSW_' followed by combination of factors levels for that experiment (as shown in seventh column of Table 4.).

Table 3. Factors and their levels

A B C D

Factors Tool rotation speed (rpm) Work table feed (mm/min) Tool shoulder dia. (mm) Pin cross sectional shape Tool pin geometry Dynamic vol./Static vol.

Level 1 1120 50 12 Circular 1

Level 2 1400 125 15 Square 1.6

Level 3 1800 200 20 Triangular 2.4

Table 4. Designed experiments with Taguchi L9 orthogonal array

Input array

Sr. Tool Feed Tool Tool pin geometry FSW designation

No. spindle rotation (rpm) (mm/min) shoulder diameter (mm) Cross sectional shape (Dyn.vol./ Stat. vol.)

1 1120 50 12 Circular 1 FSW 1111

2 1120 125 15 Square 1.6 FSW_ "1222

3 1120 200 20 Triangular 2.4 FSW_ 1333

4 1400 50 15 Triangular 2.4 FSW_ 2123

5 1400 125 20 Circular 1 FSW_ "2231

6 1400 200 12 Square 1.6 FSW_ 2312

7 1800 50 20 Square 1.6 FSW_ 3132

8 1800 125 12 Triangular 2.4 FSW_ 3213

9 1800 200 15 Circular 1 FSW_ 3321

Figure 1. (a) FSW tool with pin geometry used in the present study; (b) FSW when AA5052-H32 was kept at A.S. 2.1. Tensile testing

Subsize tensile test specimen as per the ASTM E8/E8M (2012) guidelines (shown in Fig. 2) were cut using 'wire EDM' from each welded samples and parent metals. Tensile test specimens were cut in all three directions (0 45° and 90° to rolling direction) from parent metals. The tensile specimens from the welded samples were cut transverse to the welding direction only. All tensile tests were carried out using 'Tinius Olsen (H50KS)' universal testing machine at 2 mm/min constant cross head speed.

tCCrrm

SOrrm 3Srrm 3Qrtm

V / i

* I

Figure 2. ASTM E-08 subsize tensile test specimen.

2.2. Weld macrostructure

Generally FSW appears defect free from the top and bottom surfaces, but defects like kissing bond, micro holes, porosity, worm holes and tunnel defects occur inside the weld. So, to examine the weld it is required to see the weld macrostructure. For this, samples were sectioned perpendicular to weld direction i.e. in plane of the weld cross sections and subsequently hot mounted using nonconducting epoxy powders followed by wet grinding and polishing. Keller's reagent for 30 seconds by dipping action was applied for etching. These etched samples were seen for macrostructures on stereo zoom microscopy (Leica, S8APO) using bright field imaging technique.

3. Results and discussion

3.1. Ultimate strength and % elongations

000 0.02 0.04 0.06 O.OB 0.10 0.12 0.14 0.16 0 18 Engg. Strain

Figure 3. Typical engineering stress strain curves for both BM and TFSWBs.

Figure 4. Locations of fracture in transverse tensile test specimens (a) FSW_1111, (b) FSW_3132

Figure 3 shows engineering stress strain curves for both base metals (i.e. AA5052-H32 and AA5754-H22) and three typical cases of FSWs (i.e. best, moderate and worst welds). AA5052-H32 showed ultimate tensile strength (UTS) of 199.93 MPa and 16 % elongation where as AA5754-H22 showed UTS of 179.61 MPa, and 12.8 % elongation. FSW_1111 showed lowest UTS and elongation i.e. 134.29 MPa UTS and 2.8 % elongation. FSW_2123 showed 162.14MPa UTS and 5.5 % elongation, so it has been considered as moderate weld. FSW_3132 showed 182.53 MPa UTS and 12 % elongation, so it has been considered as best weld. FSW_1111 got fractured very near to

the weld center line (Fig. 4a). The reason was presence of kissing bond (Fig. 5a) and lack of joining at the bottom most layer of the sheet. So the joint got separated easily at a very low load during transverse tensile test, and resulted in lowest UTS and % elongation. The reason behind lack of joining at the bottom layer in FSW_1111 was insufficient metal flow due to circular pin and smallest shoulder diameter with lowest rpm of tool rotation. In FSW_2123 and FSW_3132 fracture took place at TMAZ (Fig. 4b). Open literature in this regards suggests that during FSW, at TMAZ, orientation of grains takes place whereas in NZ fine equiaxed grains are achieved (Fig. 5b) and in HAZ only grain coarsening takes place. The grain orientation at TMAZ becomes almost 90° to the orientation of grains in BM due to large plastic deformation for FSW using square and triangular pin profiles. So stress concentration takes place at TMAZ during transverse tensile test due to oriented grains. This stress concentration results in localized strain. Hence necking takes place at this region and samples get fractured.

Figure 5. (a) Optical macrograph of FSW_1111 (showed lowest UTS); (b) Optical macrograph of FSW_3132 (showed highest UTS)

Based on experimental observations following relations among parameters have been developed using simple regression method for calculating UTS; as:

UTS = 141 + (7.65 X A*) — (0.82 Xg) + (2.30 X C) + (1.62 X D) (1)

Similarly for % elongation:

% Elongation = 1.16 + (2.82 x A) + (0.277 xB)+ (0.937 xC) + (0.641 x D) (2)

3.2. Multi-objective optimization of parameters

A grey based approach has been used to handle both weld strength and ductility values for optimization of the parameters. According to first step of the approach, all experimental data have been normalized in the range from zero to one [Julong (1989)]. The experimental data and their normalized values can be seen from Table 5 and 6 respectively. Data sequence for UTS and % elongations, which are larger the better performance characteristic; have been normalized as follows:

(k) = (fc)-min^f (fc)

J maxi? (k)-minx? (k) '

where k = 1 to n, i = 1 to 9, n is performance characteristic and i is trial number.

Table 5. Input (coded) and Output values after experiments

Input array Output array

Sr. No. A B C D Result-1 Weld strength (UTS, Mpa) Result-2 Result-3 Result-1 Elongation (%) Result-2 Result-3

1 1 1 1 1 134.292 152.668 159.043 4.5125 5.36956 6.1488

2 1 2 2 2 177.613 153.721 165.369 13.15625 11.15625 10.75

3 1 3 3 3 153.066 154.066 157.505 12.4375 9.0625 12

4 2 1 2 3 159.439 158.913 166.3969 5.65 5.816 7.125

5 2 2 3 1 153.423 157.081 162.343 7.65625 8.875 11.0625

6 2 3 1 2 172.15 146.124 162.874 12.78125 6.84375 8.59375

7 3 1 3 2 177.263 177.528 180.523 13.03125 13.59375 14.125

8 3 2 1 3 170.992 162.4469 170.777 12.59375 11.375 10.79365

9 3 3 2 1 168.163 162.381 175.029 13.84375 11.3125 12.02813

Table 6. Normalized values of response.

UTS % E

Sr. No. Result-1 Result -2 Result -3 Result -1 Result -2 Result -3

1 0 0.208381 0.066817 0 0 0

2 1 0.241912 0.341646 0.926323 0.703618 0.576866

3 0.433369 0.252898 0 0.849297 0.449034 0.733582

4 0.580481 0.407241 0.386302 0.121902 0.054284 0.122389

5 0.44161 0.348905 0.210183 0.336906 0.426235 0.616045

6 0.873894 0 0.233252 0.886135 0.17925 0.306531

7 0.991921 1 1 0.912927 1 1

8 0.847164 0.519771 0.576592 0.866042 0.730217 0.582339

9 0.781861 0.517673 0.761317 1 0.722617 0.737109

After calculating normalized values, the grey relation coefficients (fc) (shown in Table 7) were calculated as:

fc (ft) =

&oi (k)+ CAma,

where Aoi(k) = (fc) — xl[ (fc)|| is the difference of the absolute values x„ (fc) and xl[ (fc) , ( e [0 — 1] is distinguishing coefficient; 0.5 is widely accepted. Amin = minVjeiminVjek\\xÔ (fc) — xf (fc)y is the smallest value of Aoi, Amax = maxVjeimaxVjek\\xÔ (fc) — xl[ (fc)|| is largest value of Aoi.

After calculating grey relation coefficients, the grey relation grades (shown in Table 8) were obtained as:

Yi= ^ZJUfcCfc)

where 'ft' is the grey relation grade and 'n' is the number of performance characteristics. The higher value of grey relation grade is near to the optimum process parameters.

Table 7. Grey relation coefficients of the response

UTS % Elongation

Sr. No. Result-1 Result -2 Result -3 Result -1 Result -2 Result -3

1 0.333333 0.387111 0.348874 0.333333 0.333333 0.333333

2 1 0.397428 0.431647 0.87157 0.62784 0.541633

3 0.468766 0.400929 0.333333 0.768399 0.475753 0.652386

4 0.543762 0.457557 0.448955 0.362819 0.345849 0.362947

5 0.472416 0.434369 0.387652 0.429888 0.465651 0.56564

6 0.798587 0.333333 0.394712 0.814512 0.378573 0.418947

7 0.984098 1 1 0.851683 1 1

8 0.765889 0.510085 0.541473 0.788695 0.649533 0.544863

9 0.696244 0.508995 0.676881 1 0.643184 0.655401

Table 8. Grey relation grade and ranking

Sr. No. Grey relation grade Ranking

1 0.344886 9

2 0.64502 3

3 0.516594 6

4 0.420315 8

5 0.459269 7

6 0.52311 5

7 0.97263 1

8 0.633423 4

9 0.696784 2

3.2.1. Optimum level of factors

The average grey relation grade for each level of the factor was computed. Higher grey relation grade implies better quality characteristics. Based on the higher grey relation grade optimum level of each controllable factor was determined. The average grey relation grade and the optimum levels of factors were listed in Table 9. The optimum levels of factor based on grey relation grade was found A3B1C3D2.

_Table 9. Average grey relation grade at each level_

Factor Level-1 Level-2 Level-3 Max. - min.

A 0.502167 0.467565 0.767613* 0.300047593

B 0.579277* 0.579237 0.57883 0.000447508

C 0.500473 0.587373 0.649498* 0.149024622

D 0.500313 0.713587* 0.523444 0.213273563

*Indicates optimum level of factors

Table 10. Analys is of variance for means

Source of variation SS DF MS F P

A 0.161687248 2 0.080843624 439367.521 58.26948

B* 0.000000368 2 0.000000184 0.000133

C 0.033619398 2 0.016809699 91357.0594 12.11589

D 0.082174869 2 0.041087435 223301.275 29.6145

Error 0 0

Total 0.277481883 8

Error* 0.000000368 2 0.000000184

*Indicates pooling

The higher F value in the ANOVA table (i.e. Table 10) indicates that the factor is highly significant in affecting the process response. In this investigation, tool rotational speed was found most significant factor and played a major role followed by tool pin geometry and tool shoulder diameter; whereas effect of worktable feed was found insignificant factor.

3.2.2. Predicted optimum condition

Based on experiments, the optimum level setting was found A3B1C3D2. So the predicted grey relation grade can be calculated as:

f=rm+ Zi=i(fi - Ym) (6)

Where ym is the total mean grey relation grade, ft is the mean grey relation grade at the optimum level, and o is the number of the main design parameters that affect the quality characteristics.

So, predicted grey relation grade = 0.874251356411638

Here it is not required to run the confirmation experiment, because the optimized factor level experiment already exists within the designed experiment. The existing grey relation grade at the optimum condition is 0.97263; whereas the predicted grey relation grade is 0.87425. Hence; the difference is only 0.1 (approx.). This variation occurs due to neglecting the nonlinear effects in four factor three level Taguchi L9 orthogonal array.

4. Conclusions

The analysis presents effect of tool rotational speed, worktable translational speed, tool shoulder diameter and tool pin geometry on weld quality. Based on the analysis following conclusions can be made.

• According to applied grey based approach, 1800 rpm of tool rotational speed, 50 mm/min worktable translational speed, 20 mm of tool shoulder diameter and square pin geometry are the optimum parameters for fabrication of AA5052-H32 and AA5754-H22 dissimilar 2.0 mm thin tailored friction stir butt welded blanks.

• FSW using 1800 rpm of tool rotational speed, 50 mm/min of worktable translational speed, 20 mm of tool shoulder diameter and square tool pin geometry gives maximum weld strength (UTS = 175 MPa, approx) and maximum % elongation (13.854, approx).

• Location of fracture in uniaxial tensile test of the welded sample using optimized parameters is at TMAZ, which confirms relatively higher strength of the weld NZ.

In future, these thin aluminum TFSWBs will be deformed through biaxial stress state in laboratory scale deep

drawing and stretch forming set up to further validate the failure location for the interest of industrial applications.

Acknowledgements

The support of 'Industrial development and testing laboratory Kolkata' is acknowledged by the authors for

providing the facility of optical emission spectroscopy to evaluate elemental composition of BM and FSW tool

material.

References

ASTM E8/E8M, 2012. Standard Test Methods for Tension Testing of Metallic Materials 1. , (C), pp.1-27.

Anon, 1994. Marine Applications of Aluminum Alloys. Key to metals database. Available at: http://www.keytometals.com/Article99.htm [Accessed June 12, 2012].

Arbegast, W.J. & Hartley, P.J., 1998. Proceedings of the Fifth International Conference of Trends in Welding Research. In Pine Mountain, GA, p. 541.

Arora, A. et al., 2009. Strains and strain rates during friction stir welding. ScriptaMaterialia, 61(9), pp.863-866.

Arora, A., De, A. & DebRoy, T., 2011. Toward optimum friction stir welding tool shoulder diameter. Scripta Materialia, 64(1), pp.9-12.

Cavaliere, P. et al., 2006. Effect of welding parameters on mechanical and microstructural properties of AA6056 joints produced by Friction Stir Welding. Journal of Materials Processing Technology, 180(1-3), pp.263-270.

Cavaliere, P., Squillace, A. & Panella, F., 2008. Effect of welding parameters on mechanical and microstructural properties of {AA6082} joints produced by friction stir welding. Journal of Materials Processing Technology, 200(1-3), pp.364-372.

Hirata, T. et al., 2007. Influence of friction stir welding parameters on grain size and formability in 5083 aluminum alloy. Materials Science and Engineering: A, 456(1-2), pp.344-349.

Julong, D., 1989. Introduction to Grey System Theory. The Journal of Grey System, 1(1), pp.1-24.

Kim, Y.G. et al., 2006. Effect of welding parameters on microstructure in the stir zone of FSW joints of aluminum die casting alloy. Materials Letters, 60(29-30), pp.3830-3837.

Kinsey, B.L. & Wu, X., 2011. Tailor welded blanks for advanced manufacturing First Edit. B. L. Kinsey & X. Wu, eds., UK: Woodhead Publishing Limited.

Kumar, R., Singh, K. & Pandey, S., 2012. Process forces and heat input as function of process parameters in AA5083 friction stir welds. Transactions of Nonferrous Metals Society of China, 22(2), pp.288-298.

Mishra, R.S. & Ma, Z.Y., 2005. Friction stir welding and processing. Materials Science and Engineering: R: Reports, 50(1-2), pp.1-78.

Nandan, R., Debroy, T. & Bhadeshia, H., 2008. Recent advances in friction-stir welding - Process, weldment structure and properties. Progress in Materials Science, 53(6), pp.980-1023.

Ren, S.R., Ma, Z.Y. & Chen, L.Q., 2007. Effect of welding parameters on tensile properties and fracture behavior of friction stir welded Al-Mg-Si alloy. Scripta Materialia, 56(1), pp.69-72.

Rooks, B., 2001. Tailor-welded blanks bring multiple benefits to car design. Assembly Automation, pp.323 - 329.

Thomas, W.M. et al., 1991. Friction stir butt welding.

Threadgill, P.L., 1997. Macro and microstructural features of friction stir welds in various materials. TWI Bulletin, pp.30-33.