Friction Stir Welding and Processing of Wrought and Cast

Aluminum Alloys: Heat Transfer Modeling and

Process Optimization

A thesis submitted to the faculty of

WORCESTER POLYTECHNIC INSTITUTE

in partial fulfillment for the

Degree of Master of Science

in

Materials Science and Engineering

by

___________________________

Yi Pan

May 2014

APPROVED:

_____________________________________

Professor Diana A. Lados, Advisor and Director, Integrative Materials Design Center

_____________________________________

Professor Richard D. Sisson, Jr., Materials Science and Engineering Program Head

_____________________________________

Professor John M. Sullivan, Jr., Committee Member

2

Abstract

Aluminum alloys are considered as potential candidates for replacing traditional ferrous

structural material in ground and air vehicles, with large savings in fuel consumption and lower

greenhouse gas emissions. Friction stir welding (FSW) is a relatively new (1991) joining

technique, where all the processing is conducted in solid-state. Materials are mechanically stirred

and mixed together, resulting in welds with promising behavior and properties, even without the

need of post-weld heat treatment. In the Integrative Materials Design Center (iMdc) at WPI,

FSW of Al and Mg alloys with different strengthening systems was systematically conducted to

understand and quantify the changes in microstructure and mechanical properties associated with

various processing conditions. In this project, efforts were concentrated on optimizing the

processing parameters in FSW and developing a comprehensive understanding of the changes in

tensile and hardness properties caused by microstructural changes.

To utilize Al alloys in ground vehicle and aircraft components, problems associated with welding

need to be investigated and solved, including weld quality and strength and the overheating

effect extending from then weld into the base materials.

In Chapter 1, the thermal history and heat transfer in FSW were investigated. A heat transfer

model for FSW was established, combining analytical and numerical analyses, and validated on

three wrought and cast Al alloys (6061, A356, and 319). Thermal history in FSW of different

alloys under various processing parameters was studied and the results were further used as

inputs for the modeling work. To visualize the results of the computational modeling work

simulations were performed with Matlab.

In Chapter 2, the study of FSW was focused on four wrought and cast Al alloys – 6061, A356,

319, and A390. These were selected for overall comparisons between materials with different Si

content and strengthening mechanisms. For each alloy, processing parameters domains that

provide good welds were identified and qualified using a new quality index. Microscopy studies

were conducted on welds to understand the changes in grain structure and secondary phase

morphology/distribution. Hardness and tensile tests were conducted to evaluate material’s

response after FSW. Pre-weld heat treatment effects were also investigated.

3

Acknowledgements

I would like to thank my advisor, Professor Diana Lados, for her guidance and support

throughout the whole duration of this project. Her insight and continuous help led to the

development of a logical and comprehensive work with valuable results for the scientific and

industrial communities. She served as my mentor and role model, giving me the opportunity to

work on an exciting project and great inspiration and motivation to overcome challenges

throughout my graduate studies and succeed.

I would like to thank Professor Richard Sisson and Professor John Sullivan for serving as thesis

committee members. Your knowledge and assistance have driven me to learn and excel in my

work.

I would also like to give my appreciation to the members of iMdc consortium and my colleagues

Anthony Spangenberger, Anastasios Gavras, Xiang Chen, Ye Cao, Yuwei Zhai, and Hayley

Sandgren. It is your enthusiasm and generosity that keeps me affirmative and helps me walk

through this journey. I am grateful for all the help that I have received from you. I also need to

thank Professor Torbjorn Berstrom and his colleagues in the manufacturing lab of the

mechanical engineering department. Without your help I would not have been able to design all

the necessary apparatus and conduct the FSW experiments in my project.

Thanks to the consortium members who focused and made great contributions to the project at

iMdc. Thanks to their efforts in preparation of experimental samples and any convenience of

using lab facilities in the whole path of the study.

Finally I would like to thank my parents for their kind and constant support to me which was so

necessary for me, while being away from home. In addition, I would like to thank my friends

that helped me to overcome various problems, from trivia to important. Without you, I would not

have been able to progress and concentrate on my studies. Thank you.

4

Table of Contents

Friction Stir Welding and Processing of Wrought and Cast Aluminum Alloys: Heat Transfer Modeling

and Process Optimization ............................................................................................................................. 1

Abstract ......................................................................................................................................................... 2

Acknowledgements ....................................................................................................................................... 3

Nomenclature in Chapter 1 ........................................................................................................................... 9

Chapter 1 ..................................................................................................................................................... 10

Heat Transfer Modeling and Thermal History Analysis in Friction Stir Welding of Wrought and Cast

Aluminum Alloys........................................................................................................................................ 10

1. Introduction ......................................................................................................................................... 10

2. Experimental Procedure ...................................................................................................................... 11

2.1. Materials and Processing ............................................................................................................. 11

2.2. Friction Stir Welding Setup ......................................................................................................... 12

2.3. Temperature Measurements ......................................................................................................... 13

3. Model Description and Results ........................................................................................................... 14

3.1. Model’s Assumptions .................................................................................................................. 14

3.2. One-Dimension Analytical Model ............................................................................................... 15

3.3. One-Dimensional Analysis and Solutions ................................................................................... 17

3.4. Determination of the Coefficients in the Heat Resource Term .................................................... 18

3.5. Two-Dimensional Temperature Distribution Simulation ............................................................ 20

4. Model Validation ................................................................................................................................ 22

4.1. Thermal Profiles at Different Positions on Testing Specimen ..................................................... 23

4.2. Validation of Model on Different Materials ................................................................................ 24

5. Discussion ........................................................................................................................................... 25

6. Conclusions ......................................................................................................................................... 27

References ............................................................................................................................................... 28

Chapter 2 ..................................................................................................................................................... 32

Friction Stir Processing of Aluminum Alloys: Weld Quality Evaluation and Effects of Processing

Parameters on Microstructure and Mechanical Properties .......................................................................... 32

5

1. Introduction ......................................................................................................................................... 32

2. Experimental Procedure ...................................................................................................................... 34

2.1. Materials and Microstructures ..................................................................................................... 34

2.2. Methodology: Friction Stir Processing and Testing ..................................................................... 35

3. Results and Discussion ....................................................................................................................... 37

3.1. Weld Quality Evaluation .............................................................................................................. 37

3.2. Optimization of Processing Parameter Domains ......................................................................... 40

3.3. Microstructures of the Nugget ..................................................................................................... 44

3.4. Micro-/Macro-Hardness Measurements....................................................................................... 49

3.5. Tensile Properties Within and Across the Nugget ....................................................................... 53

4. Conclusions ......................................................................................................................................... 56

References ............................................................................................................................................... 58

Future Work ................................................................................................................................................ 62

6

List of Figures

Figure 1. Friction stir welding and processing (FSW) modified from reference [ (B.Heinz & B.Skrotzki,

Characterization of a Friction-Stir-Welded Aluminum Alloy 6013, 2002) ]. ............................................. 13

Figure 2. (a) Fixture used for FSW and (b) FSW tool. ............................................................................... 13

Figure 3. Thermocouple positions for experiments. ................................................................................... 14

Figure 4. Heat transfer model setup and boundary conditions. ................................................................... 16

Figure 5. Temperature profiles of FSP at Position 1 tested in A356 .......................................................... 20

Figure 6. Mesh profile for the cross-section used in numerical simulation. ............................................... 20

Figure 7. Transient temperature distribution of cross-section area (x=x0) in A356 processing with 1000

RPM-2 mm/s at different times: (a) 20 sec and (b) 30 sec. ......................................................................... 21

Figure 8. Temperature profiles of A356-F processed with 1000 RPM-2 mm/s. ........................................ 22

Figure 9. Simulations of temperature at Position 2 of A356-F ................................................................... 23

Figure 10. Temperature profiles at different positions after FSW: 319-F (a) 800 RPM-2 mm/s and (b)

1000 RPM-2 mm/s; 6061-T651 (c) 1200 RPM-2 mm/s and (d) 1400 RPM-2 mm/s. ................................ 24

Figure 11. Simulation of thermal history with different processing parameters ......................................... 25

Figure 12. Temperature profiles at different positions with respect to weld nugget of A356 prepared under

1000 RPM-2 mm/s ...................................................................................................................................... 26

Figure 13. Microstructures of (a) 6061, (b) A356, (c) 319, and (d) A390 Al alloys. ................................. 35

Figure 14. (a) Fixture used for FSP, (b) FSP tool, (c) tensile specimen. .................................................... 36

Figure 15. Weld quality index model for (a) internal tunnel defect and (b) surface defect. ....................... 39

Figure 16. Cross-sections and quality indices for 6061-T651 alloys after FSP. ......................................... 40

Figure 17. Cross-sections and quality indices for as-cast A356 alloys after FSP. ...................................... 41

Figure 18. Cross-sections and quality indices for as-cast 319 alloys after FSP. ......................................... 41

Figure 19. Cross-sections and quality indices for as-cast 390 alloys after FSP. ......................................... 42

Figure 20. Grain structures within in the nugget ........................................................................................ 46

Figure 21. Band spacing measurements ...................................................................................................... 49

Figure 22. Banding spacing measurements for pre-weld heat treated samples .......................................... 49

7

Figure 23. Hardness profiles across FSPed zones ....................................................................................... 51

Figure 24. Hardness profiles across FSPed zones of pre-weld heat treated samples .................................. 52

Figure 25. Tensile properties within the welds (a) 6061-T651, (b) A356-F, (c) 319-F, (d) A390-F and

across the welds (e) 6061-T651, (f) A356-F and (g) 319-F. ....................................................................... 54

Figure 26. Fracture locations in tensile specimens ..................................................................................... 56

8

List of Tables

Table 1. Chemical compositions of all studied alloys in wt% .................................................................... 12

Table 2. Thermal and physical properties of studied Al alloys ................................................................... 18

Table 3. Chemical compositions of all studied alloys in wt% .................................................................... 34

Table 4. Microstructure characterization of base materials ........................................................................ 35

Table 5. T6 heat treating parameters for the studied alloys ........................................................................ 36

Table 6. Weld quality classifications .......................................................................................................... 40

Table 7. Tested and optimized domains of rotation and traverse speeds for all studied alloy without and with pre-weld heat treatment....................................................................................................................... 43

Table 8. Grain and Si particle size in the nugget after FSP without and with pre-weld heat treatment ..... 47

Table 9. Tensile properties of base and FSPed materials with different parameters .................................. 55

9

Nomenclature in Chapter 1

̇ Heat generation rate in unit volume

Unit volume

Density of the alloy

Heat capacity of the alloy

Thermal conductivity of the alloy

Thermal conductivity of fixture material

Surface heat transfer coefficient

Tool rotation rate

Tool traverse rate

Time

Distance of tool travel

Tool effective distance

Nugget temperature

Boundary temperature

Halfwidth of workpiece

Length of workpiece

Workpiece dimensions at “i” location

Surface area at specific boundary “j”

10

Chapter 1

Heat Transfer Modeling and Thermal History Analysis in Friction Stir

Welding of Wrought and Cast Aluminum Alloys

Abstract

Friction stir welding (FSW) is a novel technique used for materials joining and microstructural

refinement. Owing to the solid-state character of the process, FSW has significant advantages

over traditional, fusion welding, including reduced part distortion and overaging. In this study, a

novel heat transfer model was developed to predict weld temperature distributions and quantify

peak temperatures under various combinations of processing parameters for different wrought

and cast Al alloys. Specifically, a one-dimensional (1D) analytical analysis was first developed

to characterize and predict temperature changes within the weld nugget and then a two-

dimensional (2D) numerical simulation was performed with Matlab to evaluate the temperature

distribution in the weld cross-sections. The model was validated by measuring the actual

temperatures near the weld nugget with thermocouples, and good agreement was obtained for all

studied cases.

Keywords: Friction stir welding, processing parameters, heat transfer modeling, temperature

distribution

1. Introduction

Friction stir welding (FSW) is a solid-state welding technique first invented at The Welding

Institute (TWI, UK) in 1991 by Thomas’ group [ (W.M.Thomas, et al., Application No.

9125978.8, December 1991), (Dawes & Thomas, November/December 1995)]. The technique

allows for similar as well as dissimilar materials joining, with no solidification cracking,

oxidation, shrinkage and porosity, that are typically present in parts joined by fusion welding.

The great advantages of FSW have been widely recognized and later on, in 2005, the process

was slightly modified by Mishra [ (R.S.Mishra, Friction stir welding and processing, 2005)]

(Friction Stir Processing, FSP) to refine surface microstructure, reduce surface defects, and

improve mechanical properties. Nowadays, FSW and FSP are used in aluminum and other

materials processing, resulting in great energy savings in the transportation industry.

Even though FSW has been around for a few years, the magnitude and effect of the heat that is

generated during processing on the resulting microstructure and properties are still not well

understood. The heat during processing is primarily generated by friction and this is strongly

related to the processing parameters. It has already been accepted that the rotation speed plays

the most significant role on heat generation. Chen et al. [ (C.M.Chen & R.Kovacevic, 2003)]

established a heat resource model to quantify the effect of rotation on friction behavior in their

simulation work. In their model, the heat generation rate was associated with the tool radius,

angular velocity, and friction coefficient at the interface between the tool and workpiece.

11

A suitable amount of heat is required to soften the material while avoiding significant precipitate

growth. Heat generation explicitly controls the material flow patterns in welds. Studies on

material and plastic flow suggested that the distribution of secondary phases is alternated

completely under close working parameters [ (P.Dong, F.Lu, J.K.Hong, & Z.Cao, 2001)]. The

generated heat has also a direct influence on weld’s microstructures. Previous investigations

have focused on identifying the relationship between the peak temperature and microstructure

changes in aluminum alloys during processing [ (C.G.Rhodes, M.W.Mahoney, W.H.Bingel,

R.A.Spurling, & C.C.Bampton, Scripta Mater., 1997), (L.E.Murr, G.Liu, & J.C.McClure, 1998)].

Abnormal grain growth has been observed in post-weld heat treated alloys [ (S.Janaa,

R.S.Mishra, J.A.Baumann, & G.Grant, 2010)]. The effect is related to recrystallization at

elevated temperatures during post-weld heat treatment.

Heat generation during FSW is complicated due to the motion of the heat resource which is

associated with material flow and alters the thermal and physical properties of the workpieces. In

addition, heat conduction is affected by the type of FSW along with different geometries of the

working material. Defining the proper boundary conditions (for convection and conduction) is

one of the challenging tasks in modeling. Several models on thermal analysis [ (D.Deng,

Numerical simulation of welding temperature field, residual stress and deformation induced by

electro slag welding, 2012), (G.Buffa, Numerical procedure for residual stresses prediction in

friction stir welding, 2011), (Y.J.Chao, Heat Transfer in Friction Stir Welding - Experimental

and Numerical Studies, 2003)] are efficient in their methods of treating boundary elements with

an energy balance condition. With proper boundary element arrangement, the boundary

condition can also be solved in particular complicated heat flux and steady/non-steady combined

conditions. This also makes modeling more flexible and applicable to different shape of

workpiece.

Previous modeling efforts usually require complicated mathematical solutions and many

calculations, which make their actual application inconvenient. Thus, there is a need for

developing a simple and conveniently to use model to predict heat generation and thermal

profiles in FSP. In the present study, a new model, that takes into account tool rotation and

traverse effects on heat generation, was developed and validated on different wrought and cast

Al alloys: 6061, A356, and 319. Instead of solving three dimensional (3D) partial differential

equations, reasonable assumptions were made to simplify the problem and yield solutions for

different engineering applications. Microstructural changes in Al alloys were interpreted based

on the thermal history predicted by the model.

2. Experimental Procedure

2.1. Materials and Processing

The material used for friction stir processing and temperature measurements was A356-F

aluminum alloy made by Chrysler LLC. Two other materials, 319-F cast Al alloy made by

Chrysler LLC and 6061-T651 by Alcoa Inc. were processed and used for model validation.

12

The chemical compositions of each alloy are shown in Table 1. Studied alloys all belong to the

hypoeutectic family. The 6061-T651 has the lowest Si and highest Mg contents. The

strengthening mechanisms in 6061-T651 is the formation of Mg-Si precipitates. This alloy was

T651 heat treated (solutionized for 1 hour at 540°C, quenched in water at 25°C, and aged for 8

hours at 175°C).

A356 and 319 alloys have the same Si content. The major differences between them are the Cu

content and strengthening system. In 319, the higher Cu content makes the alloy harder and more

brittle than A356. Also, the two alloys, A356 and 319, are characterized by different

strengthening systems (Mg-Si versus Al-Cu). The chemical compositions and strengthening

precipitates have specific effects on mechanical properties, which subsequently affect the

mechanical stirring pattern and temperature distributions during processing.

Table 1. Chemical compositions of all studied alloys in wt%

6061 A356 319

Min Max Min Max Min Max Al 96 98.6 91.1 93.3 85.8 91.5 Si 0.40 0.8 6.5 7.5 5.5 6.5

Mg 0.8 1.2 0.25 0.45 - 0.10 Cu 0.15 0.40 - 0.20 3.0 4.0 Fe - 0.7 - 0.20 - 1.0 Mn - 0.15 - 0.10 - 0.50 Cr 0.04 0.35 - - - - Ni - - - - - 0.35 Zn - 0.25 - 0.10 - 1.0 Ti - 0.15 - 0.20 - 0.25

2.2. Friction Stir Welding Setup

The sketch of the welding process is shown in Figure 1, showing two parallel plates that were

clamped together. Rectangular plates were machined from each alloy with the dimensions: 203

mm × 127 mm × 18 mm (Length × Width × Height). The workpieces were mounted in a specially designed holder made out of Al, with dimensions: 254 mm × 203 mm × 25 mm

(Length × Width × Height). The fixture was always clamped rigidly on the working platform with screws along the tool passing direction in order to prevent lateral movement during FSW,

Figure 2(a).

FSW was conducted using a HAAS VM3 milling machine with three free axes (x, y, and z

directions). The spindle load (vertical) reached the peak value during the first penetration of the

tool and then decreased and remained constant during tool’s traverse motion. The cutting force in

the tool motion direction (horizontal) increased first and then was maintained constant during

processing.

13

Figure 1. Friction stir welding and processing (FSW) modified from reference [ (B.Heinz &

B.Skrotzki, Characterization of a Friction-Stir-Welded Aluminum Alloy 6013, 2002) ].

The tool was made out of steel suitable for FSW of Al alloys. The geometry of the tool and

holder are shown in Figure 2(b). The tool consists of a plain shoulder with a diameter of 18 mm

and a conical threaded pin with a diameter of 10 mm and length of 13 mm, respectively. The tool

tip has right-hand threads, and a counter-clockwise rotation direction was used for processing.

The pin penetration depth was 13 mm.

(a) (b)

Figure 2. (a) Fixture used for FSW and (b) FSW tool.

2.3. Temperature Measurements

Experiments were conducted to measure the temperatures at the nugget. Holes with a diameter of

3 mm and depth of 40 mm were drilled prior to welding from the sides of the workpieces,

perpendicular to the traverse direction, and 36 gauge K-type thermocouples were inserted into

the workpiece.

14

The layout of the thermocouples is shown in Figure 3; all thermocouples were beaded at the tip.

Three thermocouples (1, 2 and 4) were placed close to the edge of the pin path to record the

temperature at the nugget. Location 1 was on the advancing side, 25 mm from the starting point

in x-direction. Locations 2, 3 and 4 were on the same y-z plane, 50 mm from the starting in x-

direction. The chosen cross-sectional y-z plane is close to the mid-line of the workpiece, where

hardness and microstructure are also characterized.

A symmetric set of thermocouples was placed on the retreating side, as indicated by 1’, 2’, 3’

and 4’ respectively. Temperatures were measured and recorded with a four-channel thermometer

during the entire process from the moment the tool started to penetrate at x=0 till the time when

tool was removed from the workpiece. Time delay of the temperature was 1 second for all

measurements and did not influence the temperature distribution inside the material.

Figure 3. Thermocouple positions for experiments.

3. Model Description and Results

3.1. Model’s Assumptions

Five important assumptions were made for the proposed model:

1) Heat generation rate by tool rotation and translation is constant. The heat resource contributes

to temperature fluctuations only close to the studied plane (ahead and behind it). In the modeling

work, an effective range of heat resource is introduced and further explained.

2) Heat is primarily consumed by the temperature change of the workpiece. Chao [ (Y.J.Chao,

Heat Transfer in Friction Stir Welding - Experimental and Numerical Studies, 2003)] suggested

that the heat conducted by the tool merely represents 5% of the total amount of heat. In actual

experiments, the tool material (tool steel) has lower thermal conductivity than the workpiece (Al).

It is rational to neglect the heat flux into the tool to simplify the model. The energy of materials

15

mixing consumed by plastic deformation has also small effect on temperature changes compared

to friction generated heat. Heat radiation in this case is negligible, as the measured temperature is

far below 500°C.

3) The studied Al alloys have constant and isotropic thermal properties throughout the entire

weld. This assumption reduces calculation time and increases computational efficiency. Based

on the accuracy of prediction, the alteration of thermal and physical properties by material flow

and heat during processing may occur and will be discussed later.

4) The gap between the fixture and workpieces is negligible. The abrupt variation of temperature

at fixture (top and bottom surfaces) is therefore small enough to be ignored. The temperature

gradient at the interface between fixture and workpieces is also trivial due to the relatively small

thickness of the workpiece compared to fixture dimensions, which makes it a constant value

during computations.

5) Ambient temperature is assumed constant at all four side surfaces.

3.2. One-Dimension Analytical Model

The temperature distribution was calculated from the fundamental Fourier’s heat equation, Eq.

(1), with a unit volume (V0) heat resource rate term.

( ) ̇ (1)

With particular interest in the tool traverse direction (x-axis direction), the above equation can be

simplified into a one-dimensional (1D) equation as shown in Eq. (2).

(

) ̇ (2)

By selecting the center of the tool as origin of the new coordinate system (a moving coordinate

system), the thermal gradient term

will be zero. Therefore, the functional volume of heat

resource in the one-dimensional (1D) case was transformed into the unit length in x-direction.

The equation was further simplified and rearranged, Eq. (3).

̇ (3)

Temperature, T, indicates the temperature of the nugget where tool stirring takes place. For the

one-dimension case, the temperature throughout the whole nugget is assumed to be constant at

any point. Therefore, nugget temperature, T, is only a function of position, x and time, t.

In the actual experiments, thermocouples are placed at certain positions to record the temperature

change during tool motion. As a result the temperature profile measured at specific cross-section

planes are set to a constant position (xi). Thus, during the simulation the temperature function T

becomes only a function of time, t. The net heat resource in this case is determined by two parts:

- heat generation due to tool rotation and traverse motions

- heat dissipation due to convection and conduction to fixture and to ambient air.

16

To describe the behavior of net heat, Eq. (4) was developed and its physical meaning is discussed.

̇ (

) ∑ ( )

( ) ∑ (

)

(4)

In Eq. (4), Q represents the heat from friction as a result of tool rotation and traverse motions.

Fundamentally, Q can be further separated into two terms. Tool rotation is the main part of heat

resource. As Chao [ (Y.J.Chao, Heat Transfer in Friction Stir Welding - Experimental and

Numerical Studies, 2003)] stated in his work that the heat generation rate is proportional to the

radius of the tool and the angular speed, it is believed that the heat generation rate is proportional

to ω (ω-rotation speed in RPM) as Q1ω. Tool traverse motion also takes place, and the contribution to heat generation can be represented as Q2υ (υ-traverse speed in mm/s).

The total frictional heat is the summation of both terms; Q is equal to Q1ω and Q2υ. The coefficients Q1 and Q2 are combined functions of downward spindle forces, friction coefficients,

and other tool geometry factors, which were constant in this study.

Due to the moving coordinate system, the heat resource needs modification according to the

position change, as indicated by (

). The change rate of heat resource is related to the

traverse speed of the tool.

Figure 4. Heat transfer model setup and boundary conditions.

Two types of boundary conditions were applied, considering the actual fixture in the experiments.

The workpieces were exposed to ambient environment via the four side surfaces and were

attached to the fixture on top and bottom surfaces. For the four side surfaces (shown as shadow

surfaces in Figure 4) the convection mode dominates heat flux and the dissipating rate is affected

by the temperature difference between the surface and ambient temperatures. The surface heat

17

conduction coefficient, h, is assumed to be constant and Ai represents the surface area of each

side. The relationship between boundary temperature and nugget temperature is shown in Eq. (5).

( ) ( ) (5)

The solution is the first term in Fourier’s series. In Eq. (5), represents workpiece dimensions at different side surfaces (for front and back sides , and for left and right sides ). The value of is determined using Eq. (6), obtained from heat conduction within internal workpieces, with the solution only for range (0, π/2).

(6)

3.3. One-Dimensional Analysis and Solutions

The final solution is given in the differential form by substituting the heat resource into Eq. (3),

as shown in Eq. (7).

[ ( )] (7)

In Eq. (7), Term 1 and Term 2 change as the tool moves. The terms are defined below.

{

( )

(8)

{

∑ ( )

∑ ( )

(9)

The variables U and W are rearranged to reduce the complexity of the solution and their

definition is given in Eqs. (10,11).

(10)

∑ (

)

(11)

The xt (as a constant for a specific xi and traverse speed v) is defined as the effective range of

heat resource.

∑ ( )

[ ∑ ( )

( )] (12)

18

The tool heat resource only works in the range from 0 to xt. Beyond this range the heat effect

caused by tool is eliminated and heat dissipation starts to control the temperature distribution.

In this term ( ) the exponential coefficient n was estimated using the geometry of the workpiece. As the major heat transfer occurs in both length and width directions, the ratio of

length over width largely affects the tool motion behavior. Several values of n were simulated

and compared with real conditions. A value n=3 provided very good agreement between the

predicted and actual temperature profiles. The estimated n value was associated to the aspect

ratio of length over width, approximately (

)

.

The cooling coefficient C is used for practical convenience. As the workpiece was heated up

during processing, the ambient temperature would also increase around the workpiece. Therefore,

the heat dissipation rate would normally decrease, and as a result, an extra coefficient C was used

to account for this effect. Typically, the cooling coefficient is around 0.2, which was estimated

from the temperature difference between experiments and simulations.

3.4. Determination of the Coefficients in the Heat Resource Term

In this work, an A356-F aluminum cast alloy was used in the experiments. Two other Al alloys,

wrought 6061-T651 and cast 319-F, were used in the model validation experiments and

simulations. The thermal and physical properties of each alloy required for the analysis are listed

in Table 2 [ (J.G.Kaufman, Aluminum alloy database, 2004)].

Table 2. Thermal and physical properties of studied Al alloys

Density

Thermal

conductivity Heat capacity

Surface heat

conductivity

ρ (g/cm3) k (W/m∙˚K) Cp (J/Kg∙˚K) h (W/m2∙˚K)

6061-T651 2.7 167 900

100 A356-F 2.68 160 963

319-F 2.79 112 963

Q1 and Q2 term coefficients in heat resource were evaluated using the following procedure. The

measured temperatures were first plotted against time at Position 1, Figures 5(a-d). According to

Eq. (11), at the very beginning of processing, the thermal gradient at the boundary between

fixture and workpiece is relatively small. Therefore, at the time when the tool was passing by

Position 1 the heat dissipation could be neglected, since heat dissipation is small compared to the

amount of heat generated. The experiments were conducted with various rotational speeds and

results are shown in Figure 5. In each case, the temperature changing rate is extracted and plotted

against different rotational speeds. A linear fit was applied to the experimental data and the

results are shown in Figure 5(f). The coefficient of rotation heat resource, Q1, is determined from

the slope of the curve as shown in Figure 5(f).

19

[

] (13)

A similar procedure was followed for the evaluation of traverse heat resource, Q2. In this case,

experiments were conducted with various traverse speeds and results are shown in Figure 5(g). A

linear fit was applied to the experimental data of the temperature changing rate.

[

] (14)

In order to make comparisons between the amounts of heat resource caused by tool rotation and

traverse, respectively, dimensionless parameters of both Q1 and Q2 were introduced as shown

below.

[ ] [

] (15)

[ ] [

] (16)

In these expressions, 1000 RPM and 1 mm/s were chosen as standard values for tool rotation and

traverse speeds, respectively. Relative tool rotation and traverse speeds were obtained by

dividing the actual speeds by the baseline. If another pair of rotation and traverse speeds, for

instance 800 RPM and 2 mm/s, were chosen, the dimensionless parameters, Φ1 and Φ2, would be

15.2 and 3, respectively. In this case the proportion of heat generated by tool traverse increases.

Based on the dimensionless parameters, it is apparent that tool rotation is responsible for most of

the heat generated during processing. Changes in the rotational speed during the processing may

result in more fluctuations in heat generating rate compared to the fluctuations due to traverse

speed changes.

(a) (c)

20

(d) (e)

(f) (g)

Figure 5. Temperature profiles of FSP at Position 1 tested in A356:

(a) 600RPM-2mm/s, (b) 1000RPM-2mm/s, (c) 800RPM-1mm/s, (d) 800RPM-2mm/s,

(f) linear fit for rotation heat resource, and (g) linear fit for traverse heat resource.

3.5. Two-Dimensional Temperature Distribution Simulation

The temperature distribution simulation was conducted based on a Lagrangian incremental

formulation. The sectional-area at distance xi from the penetration point of workpiece was

meshed into 255 tetrahedral elements and shown in Figure 6. The positions of the nodes had x-

symmetry throughout the whole area. Each element has an average of 6 mm of link (coarse

mesh) at boundary. The elements within the nugget were refined, half of original element size (3

mm), to predict temperature changes with more accuracy.

Figure 6. Mesh profile for the cross-section used in numerical simulation.

Y direction (in)

Z d

irec

tio

n (

in)

21

The thermal history during FSW conducted with 1000 RPM and 2 mm/s advancing and traverse

speeds, respectively, was simulated and analyzed with Matlab. Temperature contours in the

cross-section area are shown in Figures 7(a,b) for times of 20 and 30 seconds. The highest

temperature locations are at the center of the nugget and top surfaces where the tool shoulder was

in contact with the workpiece. The temperature of the workpiece gradually increased as the tool

was approaching the plane of interest. After 30 seconds, Figure 7(b), the tool was exactly at the

plane of interest and the heating up rate was maximized. Highest temperature in the nugget was

~200°C. The temperature has a symmetric distribution in the horizontal direction (y-axis). In the

perpendicular direction (z-axis), the heat flux downwards is larger in the area close to the nugget.

This behavior was attributed to the higher heat dissipating rate at the contact interface between

the workpiece and fixture.

(a)

(b)

Figure 7. Transient temperature distribution of cross-section area (x=x0) in A356 processing

with 1000 RPM-2 mm/s at different times: (a) 20 sec and (b) 30 sec.

Similarly, comparisons were made between simulated and experimental profiles. The results are

shown in Figure 8. Positions 2 and 3 are arranged in horizontal direction (y-axis) and Positions 2

and 4 are arranged in vertical direction (z-axis). In the temperature rising stage both sets of data

show similar trends. Some deviations occurred at the peak temperature position, which is the

result of accumulation of heat at local position. The highest temperature at Position 2 measured

in experiments is higher than in modeling. As the tool was passing by, the area closer to the tool

may have accumulated more heat, and due to the delay of conduction the local temperature

would exceed the theoretical value. Furthermore, the heat flux opposite to the tool motion

Y direction (in)

Y direction (in)

22

direction contributed to this phenomenon. Peak temperature values at Positions 3 and 4 are lower

than the simulated ones. As Positions 3 and 4 are farther from the center than Position 2, heat

accumulation effects were less dominant. In the next stage of cooling, the simulated temperatures

decrease faster than the experimental data. This is due to the assumption that the ambient

temperature is constant, 25°C. In fact, the ambient temperature also rises during processing and

thus lowers the cooling rate. By comparing all the experimental data and simulations results, the

differences in temperature were considered reasonable. The simulations could also predict the

highest temperature at different locations during processing in good agreement with the actual

experimental conditions.

Figure 8. Temperature profiles of A356-F processed with 1000 RPM-2 mm/s.

4. Model Validation

Temperature is function of both position and time, and therefore the validation of the proposed

model consists of two parts. In the first part, temperature profiles were generated for the same

material/sample (A356-F) at Position 2 using various combinations of rotation and traverse

speeds. At Position 2, heat dissipation was not negligible so the comparisons between

experimental and simulated data are of practical value. Temperature profiles for different

processing parameters were plotted together and the results are discussed in Figure 9. In the

second part, different materials (6061-T651 and 319-F) and a wide range of processing

23

parameters at both Positions 1 and 2 were simulated and compared to the actual temperature

profiles of these materials, Figures 10,11. The heat generation rate coefficients Q1 and Q2 were

therefore verified and modified for a more general application. The results for both validation

studies are discussed in Sections 4.1 and 4.2.

4.1. Thermal Profiles at Different Positions on Testing Specimen

Figure 9 shows the results from the comparison between experimental and simulation data at

Position 2. The highest temperature was reached with 800 RPM and 1 mm/s speeds, Figure 9(a).

This is due to the longer duration of processing under low traverse speed. The peak temperatures

in 600 RMP, 800 RPM and 1000 RPM with traverse speed of 2 mm/s, Figures 9(b-d), gradually

increase from 174°C to 177°C and then to 203°C. There is good agreement between modeling

and experimental results in all tested cases. In all cases, the simulation profile also yields a good

prediction of the locations and values of the peak temperatures. In the case of 1000 RPM, the

cooling part of the curve has more deviation from the experimental results compared to the other

three cases. This deviation is due to the heat accumulation effect as explained earlier. In addition,

a higher thermal gradient at the boundary between the workpiece and environment enhanced the

heat dissipation rate.

(a) (b)

(c) (d)

Figure 9. Simulations of temperature at Position 2 of A356-F:

(a) 800 RPM-1 mm/s, (b) 600 RPM-2 mm/s, (c) 800 RPM-2 mm/s, and (d) 1000 RPM-2 mm/s.

24

4.2. Validation of Model on Different Materials

FSW temperature measurements were conducted on both wrought 6061-T651 and cast 319-F

alloys. Four cases were evaluated and results are compared in Figure 10. Both 6061 and 319

alloys have similar friction properties with A356. Also, as A356 and 319 alloys have similar

processing parameter domains that achieve good material flow and produce defect-free welds,

the processing parameters used in A356-F (800 RPM-2 mm/s & 1000 RPM-2 mm/s) were

selected again for 319-F. From Figures 10(a,b), the peak temperatures at Position 2 during the

tests are approximately 160°C and 200°C, respectively. Temperature distributions and peak

values in both cases were compared with the simulations in Figure 11, and the results showed

good agreement.

(a) (b)

(c) (d)

Figure 10. Temperature profiles at different positions after FSW: 319-F (a) 800 RPM-2 mm/s

and (b) 1000 RPM-2 mm/s; 6061-T651 (c) 1200 RPM-2 mm/s and (d) 1400 RPM-2 mm/s.

A higher rotation speed was used for 6061-T651 (1200 RPM-2 mm/s & 1400 RPM-2 mm/s),

which produce good quality welds for this material, and the experimental temperature profiles

are shown in Figures 10(c-d). Higher peak values of the temperature were obtained for this material using these two sets of parameters (220°C and 250°C, respectively). The theoretical

values of heat generation rates in each case are 20 J/s and 23 J/s in unit volume based on the heat

generation model, Eqs. (13,14). Thus, the highest simulated temperatures in the nugget are

228°C and 258°C. The calculations show again good agreement with the experimental data.

25

(a) (b) (c)

Figure 11. Simulation of thermal history with different processing parameters:

(a) Position 1 (25 mm) under traverse speed of 2 mm/s, (b) Position 2 (50 mm) under traverse

speed of 2 mm/s, and (c) Position 2 (50 mm) under traverse speed of 1 mm/s.

5. Discussion

The heat distributions on advancing and retreating sides show differences and the phenomenon is

due to higher stirring rate on the advancing side. This effect was ignored in the simulations, and

will be briefly addressed here. The practical stirring rate on the advancing side is (ωr+υ) mm/s and on retreating side the stirring rate is (ωr-υ) mm/s, where r is the radius of the tool. Typically, the chosen rotation speed is 1000 RPM and the radius of the tool tip is 5 mm. Thus, the velocity

for tool rotation is approximately 500 mm/s on both sides, but much larger than the traverse

speed.

Heat generation on different sides (advancing versus retreating) is shown in Figure 12. The

temperature profiles are quite different. As velocity differences on the two sides are small, the

material flow on each side during processing is responsible for the observed differences in the

temperature distributions. Based on material flow modeling work for an eutectic Al alloy (11.35%

Si, 1.49% Cu, and 1.46% Mg) [ (S.Tutunchilar, M.Haghpanahi, M.K.BesharatiGivi, P.Asadi, &

P.Bahemmat, 2012)], the material flow pattern changes from top (contact with tool shoulder) to

bottom (probe tip) and from advancing to retreating sides. The highest material velocity is

around 70 mm/s close to shoulder on top side, which results in relatively large changes to friction

behavior between tool and workpiece. On advancing side, high-velocity areas are smaller than on

the retreating side. The flow velocity on the retreating side is higher than on the advancing side

as a result of more material extracted by the tool, which indicates less friction occurring on the

retreating side.

26

(a) (b)

Figure 12. Temperature profiles at different positions with respect to weld nugget of A356

prepared under 1000 RPM-2 mm/s: (a) advancing side and (b) retreating side.

Changes in physical and thermal properties during processing also influence the modeling results.

The most important factors controlling the property changes are the strain rate and temperature

change. De [ (P.S.De, N.Kumar, J.Q.Su, & R.S.Mishra, 2011)] built a strain and stain rate model

as a function of the changes in tool rotation speeds. In their model, the deformation and

deformation rate are calculated in the recrystallized/nugget zone (radius and depth) as a function

of advance per revolution (APR) of the tool. Variations in strain and strain rate were observed on

both retreating and advancing sides. On advancing side, a high accumulated strain was obtained

which indicates that a high extrusion process was experienced under the motion of the tool.

Temperature effect can also make a difference during the processing procedure and change the

local plastic deformation behavior. High strain rate and low temperature make deformation more

difficult. El-Mag [ (E.El-Mag, 1994)] described a method for calculating the stress associated

with variable strain rates and temperatures:

̇ ̇ {

[ (

)

]

} (17)

In Eq. (17), the values of , , p and q are associated with material type and transient conditions. is a thermal activated component of stress. The equation can be transformed into Eq. (18) to evaluate the changes in yield stress at different strain rates:

(

)

(

̇

̇ )

(18)

Assuming a linear relationship (as p and q equals to 1), the equation is further simplified. The

change in the yield stress is determined by the change in strain rate under linear relationships to

( ̇

̇ ) at constant temperature. The sensitivity of the mechanical properties change is even

higher at elevated temperature. It could be concluded that in the FSW the highest stress change

would occur on the advancing side. This may explain why more welding defects are encountered

on the advancing side.

27

Combining the thermal profile predictions with microstructural changes, the proposed model can

be further extended to also determine grain evolution and precipitation growth during FSW,

which will be investigated in future studies.

6. Conclusions

A heat transfer model for FSW in wrought and cast Al alloys was developed using both

analytical and numerical methods. Mathematical formulations were derived to predict

temperature profiles within weld nugget during FSW based on Fourier’s heat transfer law.

Temperature profiles under different processing parameters, different materials, and different

pre-weld heat treatment conditions were experimentally measured. Verification and validation of

the model was performed and good agreement between predicted and simulated thermal profiles

was observed. The effects of tool rotation and translation were investigated and discussed. Most

of the energy input during the process comes from tool rotation. The heating up rate is affected

by other factors including the type of material, heat treatment, workpiece geometry, and material

flow pattern.

28

References A.C.Ugural, & S.K.Fenster. (2003). Advanced Strength and Applied Elasticity, 4th Edition.

A.Evans, C.S.Marchi, & A.Mortensen. (2003). Metal matrix composites in industry: an introduction and a

survey. Dordrecht, Netherlands: Kluwer Academic Publishers.

A.Larsen. (2012). Estimating the workpiece-backingplate heat transfer coefficient in friction stir wedling.

Engineering computations:International journal for computer-aided engineering and software,

pp. Vol. 29 No. 1,pp. 65-82.

B.F.Chanel, & D.A.Lados. (2011). Friction Stir Welding in 6061 Aluminum Alloys: Microstructure, Residual

Stress, and Fatigue Crack Growth Mechanisms. Metallurgical and Materials Transactions A..

B.Heinz, & B.Skrotzki. (2002). Characterization of a Friction-Stir-Welded Aluminum Alloy 6013.

Metallurgical and Materials Transactions B, pp. 33(3) pp. 489-498. .

C.Dawes, & W.Thomas. (November/December 1995). TWI Bulletin 6, p. 124.

C.G.Rhodes, M.W.Mahoney, W.H.Bingel, R.A.Spurling, & C.C.Bampton. (1997). Scripta Mater., p. 36

(1997) 69.

C.I.Chang, C.J.Lee, & J.C.Huang. (2004). Scr. Mater., pp. Vol 51,p 509–514.

C.M.Chen, & R.Kovacevic. (2003). Finite element modeling of friction stir welding—thermal and

thermomechanical analysis. International Journal of Machine Tools & Manufacture, pp. 43 (2003)

1319–1326.

C.M.Rejil, I.Dinaharan, S.J.Vijay, & N.Murugan. (2012). Microstructure and sliding wear behavior of

AA6360/(TiC + B4C) hybrid surface composite layer synthesized by friction stir processing on

aluminum substrate. Materials Science and Engineering A, 552 (2012) 336– 344.

D.Aruri, A.Kumar, & B.Kotiveerachary. (n.d.). Effect of post-process artificial aging treatment on tensile

properties of SiC particle reinforced AA6061-T6 surface metal matrix composite fabricated via

Friction stir process. International Journal of Mechanical Engineering, ISSN : 2277-7059 Volume

1 Issue 1.

D.B.Miracle. (2005). Metal matrix composites – From science to technological significance. Composites

Science and Technology, 65 (2005) 2526–2540.

D.Deng. (2012, 09). Numerical simulation of welding temperature field, residual stress and deformation

induced by electro slag welding. COMPUTATIONAL MATERIALS SCIENCE, pp. Volume 62, pp. 23 -

34.

Dawes, C., & Thomas, W. (November/December 1995). TWI Bulletin 6, p. 124.

29

E.El-Mag. (1994, 9). Colloque C8, supplement au Journal de Physique 111 . JOURNAL DE PHYSIQUE IV, pp.

Volume 4,C8 149-170.

G.Buffa. (2011). Numerical procedure for residual stresses prediction in friction stir welding. Finite

Elements in Analysis & Design, pp. Volume 47, Issue 4, pp. 470 - 476.

G.Faraji, & P.Asadi. (2011). Characterization of AZ91/alumina nanocomposite produced by FSP.

Materials Science and Engineering A, 528 (2011) 2431–2440.

H.Bisadi, & A.Abasi. (2011). Fabrication of Al7075/TiB2 Surface Composite Via Friction Stir Processing.

American Journal of Materials Science, 1(2): 67-70.

H.Uzun. (2007). Friction stir welding of SiC particulate reinforced AA2124 aluminium alloy matrix

composite. Materials and Design, 28 (2007) 1440–1446.

J. Arbegast William. (2008). A flow-partitioned deformation zone model for defect formation during

friction stir welding. Scr.Mater., 58:372–6.

J.B.Fogagnolo, E.M.Ruiz-Navas, M.H.Robert, & J.M.Torralba. (2002). 6061 Al reinforced with silicon

nitride particles processed by mechanical milling. Scripta Materialia, 47 (2002) 243–248.

J.G.Kaufman. (2004). Aluminum alloy database.

J.Guo, P.Gougeon, & X.G.Chen. (2012). Microstructure evolution and mechanical properties of dissimilar

friction stir welded joints between AA1100-B4C MMC and AA6063 alloy. Materials Science and

Engineering A, 553 (2012) 149– 156.

J.Guo, S.Amira, P.Gougeon, & X.G.Chen. (2011). Effect of the surface preparation techniques on the

EBSD analysis of a friction stir welded AA1100-B4C metal matrix composite. MATERIALS

CHARACTERIZATION, 62(2011)865–877.

J.Hemanth. (2009). Development and property evaluation of aluminum alloy reinforced with nano-ZrO2

metal matrix composites (NMMCs). Materials Science and Engineering A, 507 (2009) 110–113.

J.M.Root, D.P.Field, & T.W.Nelson. (2009). Crystallographic Texture in the Friction-Stir-Welded Metal

Matrix Composite Al6061 with 10 Vol Pct Al2O3. METALLURGICAL AND MATERIALS

TRANSACTIONS A, VOLUME 40A, SEPTEMBER 2009—2109.

K.N.Krishnan. (2002). Mater. Sci. Eng. A, p. 327 (2002) 246.

K.V.Jata, K.K.Sankaran, & J.J.Ruschau. (2000). Metall. Mater. Trans.A, p. 31 (2000) 2181.

L.E.Murr, G.Liu, & J.C.McClure. (1998). J. Mater. Sci., p. 33 (1998) 1243.

M.Amirizad, A.H.Kokabi, Gharacheh, M., R.Sarrafi, B.Shalchi, & M.Azizieh. (2006). Evaluation of

microstructure and mechanical properties in friction stir welded A356+15%SiCp cast composite.

Materials Letters, 60 (2006) 565–568.

30

M.Raaft, T.S.Mahmoud, H.M.Zakaria, & T.A.Khalifa. (2011). Microstructural, mechanical and wear

behavior of A390/graphite and A390/Al2O3 surface composites fabricated using FSP. Materials

Science and Engineering A, 528 (2011) 5741–5746.

M.Rittner. (2000). Metal matrix composites in the 21st century: markets and opportunities. Norwalk, CT:

BCC, Inc.

M.Salehi, M.Saadatmand, & Mohandesi, J. (2012). Optimization of process parameters for producing

AA6061/SiC nanocomposites by friction stir processing. Trans. Nonferrous Met. Soc. China.

N.Sun. (2009, May). Friction Stir Processing of Aluminum Alloys.

P.Dong, F.Lu, J.K.Hong, & Z.Cao. (2001). Coupled thermomechanical analysis of friction stir welding

process using simplified models,. Science and Technology of Welding and Joining, pp. 6 (5) (2001)

281–287.

P.S.De, N.Kumar, J.Q.Su, & R.S.Mishra. (2011). Fundamentals of Friction Stir Welding, Welding

Fundamentals and Processes,. ASM Handbook, Vol 6A,, pp. p 186-200.

R.Ramesh, & N.Murugan. (2012). Production and Characterization of Aluminium 7075 – T651 Alloy / B4C

Surface Composite by Friction Stir Processing. International Journal of Engineering and

Advanced Technology (IJEAT), ISSN: 2249 – 8958, Volume-2, Issue-1, October 2012.

R.S.Mishra. (2005, August 18). Friction stir welding and processing. Materials Science and Engineering.

R.S.Mishra, & Z.Y.Ma. (2005). Friction Stir Welding and Processing. Materials Science and Engineering: R:

Reports,, 50(1-2) pp. 1-78.

Rejil, C., I.Dinaharan, S.J.Vijay, & N.Murugan. (2012). Microstructure and sliding wear behavior of

AA6360/(TiC + B4C) hybrid surface composite layer synthesized by friction stir processing on

aluminum substrate. Materials Science and Engineering A, 552 (2012) 336– 344.

S.Benavides, Y.Li, L.E.Murr, D.Brown, & J.C.McClure. (1999). Scripta Mater, p. 41 (1999) 809.

S.H.Kazi, L.E.Murr, K.V.Jata, M.W.Mahoney, R.S.Mishra, S.L.Semiatin, et al. (2001). Friction Stir Welding

and Processing. TMS, (p. p. 139). Warrendale, PA, USA.

S.Janaa, R.S.Mishra, J.A.Baumann, & G.Grant. (2010). Effect of process parameters on abnormal grain

growth during friction stir processing of a cast Al alloy. Materials Science and Engineering A, pp.

528 (2010) 189–199.

S.Tutunchilar, M.Haghpanahi, Givi, M., P.Asadi, & P.Bahemmat. (2012). Simulation of material flow in

friction stir processing of a cast Al–Si alloy. Materials and Design, 40 (2012) 415–426.

T.W.Clyne, & P.J.Withers. (1994). An introduction to metal matrix composites. Paperback Pub.in 1994 ed.

31

W.H.Hunt, & D.B.Miracle. (2001). Automotive applications of metal matrix composites. In D.B.Miracle, &

S.L.Donaldson, ASM handbook. Materials Park: ASM International (pp. Composites, vol. 21. 2001.

p. 1029–32.).

W.H.Hunt, & D.R.Herling. (FEBRUARY 2004). METAL MATRIX COMPOSITES. ADVANCED MATERIALS &

PROCESSES.

W.M.Thomas. (2001). Aluminum 2001—Proceedings of the TMS 2001 Aluminum Automotive and

Joining Sessions. TMS 2001 Aluminum Automotive and Joining Sessions, (p. p. 213).

W.M.Thomas, E.D.Nicholas, J.C.Needham, M.G.Murch, P.Templesmith, C.J.Dawes, et al. (December

1991). Application No. 9125978.8.

Y.J.Chao. (2003, 2). Heat Transfer in Friction Stir Welding - Experimental and Numerical Studies. Journal

of Manufacturing Science and Engineering.

Y.J.Kwon, N.Saito, & I.Shigematsu. (2002). Friction Stir Process as a New Manufacturing Technique of

Ultrafine Grained Aluminum Alloy. Journal of Material Science Letters, 21pp. 1473-1476.

Y.Li, L.E.Murr, & J.C.McClure. (1999). Mater. Sci. Eng. A, p. 271 (1999) 213.

Y.S.Sato, H.Kokawa, M.Enmoto, & S.Jogan. (1999). Metall. Mater. Trans. A, p. 30 (1999) 2429.

32

Chapter 2

Friction Stir Processing of Aluminum Alloys: Weld Quality Evaluation and

Effects of Processing Parameters on Microstructure and

Mechanical Properties

Abstract

Friction stir processing (FSP) is a solid-state technique widely used for localized

microstructure/property modifications and repair in the aerospace sector. Understanding the

effects on microstructure and static and dynamic properties is critical in design for structural

integrity. In this study, four aluminum alloy systems (wrought 6061 and cast A356, 319, and

A390) were friction stir processed using various processing parameters in both as-fabricated and

pre-weld heat treated (T6) conditions. The effects of processing and heat treatment on the

resulting microstructures, hardness/micro-hardness, and tensile properties were systematically

investigated and mechanistically correlated to changes in grain size, characteristic phases, and

strengthening precipitates. Tensile tests were performed in room temperature air both along and

across the processing zones. Optimum processing parameter domains that provide both defect-

free welds (evaluated using a new quality index) and good mechanical properties were

determined for each alloy and associated to the thermal history of the process. The results of

these studies will be presented and discussed, together with recommendations for design and

materials/process optimization.

Key words: Friction Stir Processing, Al alloys, weld quality index, microstructure, static

mechanical properties, micro-/macro-hardness

1. Introduction

Friction Stir Welding (FSW) was invented at The Welding Institute (TWI) in 1991 UK by

Thomas et al. [ (W.M.Thomas, et al., Application No. 9125978.8, December 1991) (C.Dawes &

W.Thomas, November/December 1995)]. The technique can be used for both similar and

dissimilar materials joining without many of the drawbacks encountered in fusion welding. A

non-consumable tool with shoulder and pin is used for material stirring during processing. Later

on, in 2005, FSW technique was slightly modified by Mishra [ (R.S.Mishra, Friction stir welding

and processing, 2005)] (Friction Stir Processing, FSP) to refine surface microstructure, reduce

surface defects, and improve mechanical properties of the materials. Nowadays, FSP is used for

aluminum and other materials, resulting in great energy savings in the transportation industry,

and understanding its fundamentals and advantages on mechanical properties are therefore

important considerations.

33

Microstructural changes during mechanical stirring have significant effects on mechanical

properties and these are largely controlled by the processing parameters, especially rotation and

traverse speeds [(R.S.Mishra, Friction stir welding and processing, 2005) (W.M.Thomas,

Aluminum 2001—Proceedings of the TMS 2001 Aluminum Automotive and Joining Sessions,

2001)].

The mechanical stirring associated with high-temperature exposure within the weld zone results

in characteristic microstructures after FSP, and three distinct are observed: dynamically

recrystallized zone (DXZ) also known as nugget zone, thermo-mechanically affected zone

(TMAZ), and heat affected zone (HAZ). Both DXZ and TMAZ experience intense plastic

deformation during FSP, and therefore the grain structures and secondary phases are

mechanically refined. Dynamic recrystallization is driven and affected by the strain rate as well

as the total thermal energy during FSP. [(S.Benavides, Y.Li, L.E.Murr, D.Brown, & J.C.McClure,

Scripta Mater, 1999) (Y.Li, L.E.Murr, & J.C.McClure, Mater. Sci. Eng. A, 1999) (B.Heinz &

B.Skrotzki, 2002) (S.H.Kazi, et al., Friction Stir Welding and Processing, 2001)].

The shape of the nugget and TMAZ depend on tool geometry, processing temperatures, and

thermal conductivity of the material (R.S.Mishra, Friction stir welding and processing, 2005).

Within the nugget, “onion ring (or bands) material flow patterns are observed, which can be

correlated with the rotation and traverse speeds, and used to monitor the distribution of

secondary phases. Subgrain structure changes within the nugget and TMAZ have also been

reported in other studies. Dislocation and subgrains are usually eliminated during

recrystallization [(C.G.Rhodes, M.W.Mahoney, W.H.Bingel, R.A.Spurling, & C.C.Bampton,

1997)]. According to Jata et al., insufficient heat input and faster traverse welding/processing

conditions may increase dislocation density and alter the mechanical properties within the welds

[(K.V.Jata, K.K.Sankaran, & J.J.Ruschau, Metall. Mater. Trans.A, 2000)].

As opposed to TMAZ, HAZ has no plastic deformation, but is thermally affected by the process.

The properties within this zone can be influenced by the coarsening of strengthening precipitates

in heat-treatable alloys. A decrease in hardness of the material is frequently observed. The width

of the HAZ is determined based on the extent of the decrease in hardness. For Al alloys, a critical

temperature of 250˚C is the threshold for changing existing precipitate structures [(Y.S.Sato, H.Kokawa, M.Enmoto, & S.Jogan, Metall. Mater. Trans. A, 1999)]. Thus, tool rotation speed

can be adjusted to prevent overaging in the HAZ.

Changes in microstructures typically lead to changes in mechanical properties. Tensile properties

within the weld nugget (refined zone) usually show improvements in toughness [ (A.L.Biro &

D.A.Lados, 2012) (B.F.Chenelle & D.A.Lados, 2010)]. Also, for practical applications, tensile

properties in the traverse direction (across the weld) are important and need to be understood and

correlated to processing conditions and resulting microstructures. In these cases, low strength

may be the result of: 1) defects within mechanically stirred zones, 2) poor bonding between

different weld zones and 3) initial defects in base materials.

In this study, experimental investigations on wrought and cast Al alloys processed with different

combinations of rotation and traverse speeds were made, with and without pre-weld heat

treatment. A new quality index was developed to qualify the welds based on the morphology of

the defects and the stress concentration associated with them. Proper FSP processing domains

34

were determined for each alloy depending on the quality index classifications. Grain size, band

spacing, and secondary phases were characterized and associated with processing parameters.

Micro-/macro-hardness were measured across the welds and tensile properties both within and

across the nugget were interpreted based on processing conditions and resulting heat input.

2. Experimental Procedure

2.1. Materials and Microstructures

Four Al alloy (wrought 6061 & cast A356, 319, and A390) were used in this study. The alloys

were tested both without and with pre-weld heat treatment (T6). Chemical compositions of the

alloys are shown in Table 1. The selected materials have major chemical differences in three

elements: Si, Mg, and Cu. Based on Si content, alloys 6061, A356 and 319 belong to the

hypoeutectic, and A390 belongs to the hypereutectic family of the Al-Si system. Alloy 6061 has

the lowest Si content, while A356 and 319 have a similar Si content ~6%, and characteristic

eutectic Si phases. Alloy A390 was chosen for a high Si content comparison in hypoeutectic

family of the Al–Si system, having both eutectic and primary Si phases. In A390 alloy, the

primary Si phase adds additional complexity to the original microstructure, as well as to the FSP

microstructure in terms of phase morphology and distribution.

The 6061 alloy has the highest content of Mg among all the materials, and its strengthening

precipitate system is Mg-Si. Alloy A356 has moderate Mg content, and its strengthening system

in also Mg-Si. The Mg content in 319 is low and the dominant strengthening mechanism results

from the formation of metastable Al2Cu phase (θ’ and θ’’) due to the high Cu content. Alloy

A390 has similar Cu content and strengthening system with 319, and due to the high level of Si

the material exhibits high hardness and brittle behavior. In addition, the presence of Cu can

affect significantly the mechanical properties in cases with pre-weld heat treatment.

Table 3. Chemical compositions of all studied alloys in wt%

6061 A356 319 A390

Min Max Min Max Min Max Min Max Al 96 98.6 91.1 93.3 85.8 91.5 75.2 79.6 Si 0.40 0.8 6.5 7.5 5.5 6.5 16 18

Mg 0.8 1.2 0.25 0.45 - 0.10 0.45 0.65 Cu 0.15 0.40 - 0.20 3.0 4.0 4.0 5.0 Fe - 0.7 - 0.20 - 1.0 - 0.50 Mn - 0.15 - 0.10 - 0.50 - 0.10 Cr 0.04 0.35 - - - - - - Ni - - - - - 0.35 - - Zn - 0.25 - 0.10 - 1.0 - 0.10 Ti - 0.15 - 0.20 - 0.25 - 0.20

35

Metallographic specimens were ground and polished to 1 μm alumina powder and subsequently etched for microstructural characterization. The alloys were etched with Barker’s reagent at

(6061: 25V for 40 seconds & cast alloys: 20V for 30-40 seconds) to reveal the grain structure.

Optical and scanning electron microscopy was used for microstructural characterization of all

alloys.

The original microstructure of the studied alloys (before FSP) are shown in Figure 13. Low and

high magnification pictures were taken, and different characteristic phases were identified. The

size of secondary dendrite arm spacing (SDAS) and eutectic/primary Si particles were measured

as shown in Table 4.

(a) (b) (c) (d)

Figure 13. Microstructures of (a) 6061, (b) A356, (c) 319, and (d) A390 Al alloys.

Table 4. Microstructure characterization of base materials

Grain size (μm) SDAS (μm) Porosity (%)

Si Particle Size (μm)

Eutectic

Primary Average size

Aspect ratio

6061 400*80 N/A N/A N/A N/A

A356 694 60 1.05 43 0.4 N/A

319 801 63 1.2 55 0.4 N/A

A390 751 N/A 1.3 60 0.3 96

2.2. Methodology: Friction Stir Processing and Testing

All materials were machined into plates with the dimensions L × W × H = 203 mm × 127 mm × 18 mm. The wrought Al alloy 6061 was used in the T651 temper. The cast alloys were FSP in

both as-cast and with a pre-weld T6 heat treatment (A356-F/T6, 319-F/T6, and A390-F). The

Al(CrMnFe)Si

36

pre-weld heat treatment was used to improve the initial materials strength and understand

precipitate growth during processing. Different T6 heat treatments were used for each alloy as

listed in Table 5.

Table 5. T6 heat treating parameters for the studied alloys

Solutionizing T (°C) Solutionizing t (hr) Aging T (°C) Aging t (hr)

6061 540 1 175 8(Artificial) A356 540 1.5 153 12(Natural)/12(Artificial) 319 495 4 180 12(Natural)/4(Artificial)

(a) (b)

(c)

Figure 14. (a) Fixture used for FSP, (b) FSP tool, (c) tensile specimen.

A HAAS VM3 mill was used for the FSP experiments, and the experiments were conducted in

room temperature air. The fixture system used in the project was shown in Figure 14(a). Natural

cooling was allowed to avoid damage of the tool by cooling shrinkage and hardening of

materials around tool tip. The plates were FSPed using a tool made of tool steel, in Figure 14(b),

with a plain shoulder and a conical pin with right-hand thread. The shoulder has a diameter of 18

mm, and the pin has a length of 13 mm and a diameter of 10 mm. The total penetration depth

during FSP was 13 mm.

37

A single pass was used for all experiments. Advancing and retreating sides were marked based

on the counter-clockwise direction of tool rotation. The tool rotation speed ranged from 600 to

1600 RPM for wrought alloys and from 800 to 1200 RPM for all cast alloys. Tool traverse

speeds were from 1 to 3 mm/s for all hypoeutectic alloys and 0.17 to 0.64 mm/s for hypereutectic

alloy due to its brittle nature, caused by the high Si content.

Vickers hardness (HV) micro-hardness tests were conducted, using a 200 g load for 10 s, from

the base material to the nugget on both sides of the processed zone. Brinell macro-hardness tests,

using a 100 kg load via a spherical steel indenter of 1.6 mm in diameter were conducted, and the

measurements converted to the Vickers hardness scale (HV). The position of the hardness

measurements in the samples was 6 mm beneath the nugget surface.

Tensile tests were performed in accordance to the ASTM-E8 standard. Tensile specimens were

extracted from both longitudinal (within) and traverse (across) directions of the processing zones,

shown in Figure 14(c). The dimensions of the tensile bars were: 25.4 mm (gauge length), 5 mm

(gauge thickness), 6.5 mm (fillet radius), 13 mm (grip length), and 13 mm (grip width). For

longitudinal specimens, the whole gage area was within nugget. For traverse specimens, the

middle of the gage was located at the center of the nugget.

3. Results and Discussion

Specimens prepared with different processing parameters exhibit various flow patterns.

Traditionally welds/processed zones without pores or other defects are considered good welds.

For welds that have defects, the stress concentration around defects will be used to evaluate their

quality and understand the impact on mechanical properties.

3.1. Weld Quality Evaluation

With the aim to reduce the manufacturing cost of FSP products in industry while meeting the

requirements from the safety perspective, an effective quality evaluation method was developed

based on a simple model shown in Eq. (1).

Weld quality was quantitatively evaluated for a variety of combinations of rotation/traverse

speeds for all the studied alloys. A quality index was developed based on the aspect ratios of the

defects in the weld cross-sectional area, also considering the orientation of the defect with

respect to the load application. The quality index of the weld depends on both the size and shape

of the defects. The general equation of the quality index is given in Eq. (1).

( ) ̅ (1)

In Eq. (1), the defect area fraction ( ) and average morphology coefficient of defects ( ̅) must be defined. The defect area fraction, f, is calculated by the ratio of defects area and weld area as

shown in Eq. (2).

⁄ (2)

38

Morphology coefficient M is related to the stress concentration when a load is applied

[(A.C.Ugural & S.K.Fenster, Advanced Strength and Applied Elasticity, 4th Edition, 2003)]. To

develop the quality index model, defects are assumed to be elliptical for simple calculations of

stress concentration factors. When an elliptical defect is subjected to a simple tension load on

both sides, the stress at defect tip, perpendicular to defect orientation, is given by Eq. (3).

[

( )

] (3)

Stress at the defect tip can be then calculated based on the size of the defect obtained from

micrographs and Eq. (3) (defect measurements should be conducted with proper methods, as

indicated in Figure 2). The angle is important for a defect and is defined by elliptical semi-major axis, , elliptical semi-minor axis, , and half focus, . The relationships between these parameters are:

The corresponding stress concentration factor is calculated as the ratio of maximum and minimum stress on the boundary of the defect/pore in the tangential direction. It is obtained by

substituting boundary conditions into Eq. (4).

(

) (4)

For a universal load from all directions the circular shape has the lowest ( ); the stress concentration factor increases when the curvature radius decreases. Thus, an acceptable

morphology coefficient will be obtained in the case of a round defect. Morphology coefficient is

unacceptable when the defect is elongated in certain directions.

Irregular defect shapes are usually encountered in actual processing, and in such cases a defect

might contain tips in several directions. The morphology coefficient was determined by the

defect tip in the most elongated direction. Subsequently, the morphology coefficient of internal

tunnel defects was calculated by the square root of its aspect ratio measured from micrographs,

as shown in Figure 14(a).

√

(5)

In the Eq. (5), Length is measured in elongated direction and Width is measured perpendicular to

the elongated direction. Thus, the theoretical varies from 0 (needle-shape defect) to 1

(circular shape defect). In real cases, morphology coefficients under tension loading will be

lower than the measured/calculated values due to the irregular shape and multiple defect tips.

39

(a) (b)

Figure 15. Weld quality index model for (a) internal tunnel defect and (b) surface defect.

Surface defects are considered half of an internal defect. They have similar but slightly higher

stress concentration factors compared to internal defects, which increase when the curvature

radius at the tip decreases. Therefore, the surface morphology coefficient is also defined by

aspect ratio, using the relative distances from the surface, Eq. (6).

(

)

(6)

The average morphology coefficient was defined to include contributions from both internal

tunnel defects ( ) and surface defects ( ) . For a weld containing multiple defects, the average morphology coefficient ̅ is given by Eq. (7).

̅ ∑

∑

(7)

In this equation, n and m are the numbers of internal tunnel and surface defects, respectively.

The equation balances the effects of each defect, taking into account both the size and shape.

Larger defects with poor morphology coefficients contribute more to the decrease in mechanical

properties. Single defect area fractions are needed for each inspected defect. are single defects area fractions defined as given in Eq. (8).

(8)

The quality index was developed using image analysis on cross-sectional areas of the welds, but

it could also be applied in combination with other non-destructive detection approaches such as

X-ray diffraction. By acquiring the internal defect/porosity size and shape, the quality index can

be calculated with the measured aspect ratio of these defects.

FSP eliminates pre-existing porosity (especially in the cast alloys) from the nugget. Thus, the

defects considered in this analysis were all considered to be a result of inappropriate processing

parameters. For all the studied cast alloys, the existing pores only occupied about 1% of the

volume, and they were removed during processing. The wrought 6061 alloy did not have any

porosity. Therefore, the model can be used for the evaluations of all alloys studied in this project.

40

3.2. Optimization of Processing Parameter Domains

The quality index was used to characterize the quality of the FSPed materials. Welds are

characterized according to the quality index and classified into four groups: poor, fair, good, and

perfect as generically exemplified in Table 6. The welds with good and perfect quality are

considered qualified, based on safety factors.

Table 6. Weld quality classifications

Quality index 0.0-0.6 0.6-0.8 0.8-0.9 0.9-1.0

Weld quality

Poor Fair Good Perfect

Cross-sectional weld areas for all studied alloys processed with different combinations of

rotation and traverse speeds are shown in Figures 16-19.

800 RPM 1000 RPM 1200 RPM 1400 RPM 1600 RPM

1

mm/s

0.46 0.96 0.95 0.89

2

mm/s

0.64 0.93 0.94 0.93

3

mm/s

0.37 0.56 0.95 0.98 0.96

Figure 16. Cross-sections and quality indices for 6061-T651 alloys after FSP.

41

600 RPM 800 RPM 1000 RPM 1200 RPM

1 mm/s

0.47 0.53 0.70 0.88

2 mm/s

0.70 0.84 0.95 0.73

3 mm/s

0.61 0.91 0.62 0.54

Figure 17. Cross-sections and quality indices for as-cast A356 alloys after FSP.

600 RPM 800 RPM 1000 RPM 1200 RPM

1 mm/s

0.51 0.75 0.65 0.58

2 mm/s

0.73 0.96 0.93 0.65

3 mm/s

0.72 0.62 0.89 0.65

Figure 18. Cross-sections and quality indices for as-cast 319 alloys after FSP.

42

800 RPM 1000 RPM 1200 RPM

0.17 mm/s

0.89 0.94 0.41

0.25 mm/s

0.47 0.96 0.74

0.42 mm/s

0.56 0.67 0.58

0.64 mm/s

0.43 0.63 0.51

Figure 19. Cross-sections and quality indices for as-cast A390 alloys after FSP.

Domains of “appropriate processing parameters” were identified for each alloy, as shown in

Table 7, and correlated to material flow during processing. For 6061-T651, higher rotational

speeds are required in order to generate sufficient heat to soften and flow the material. Low

stirring rate (less than 1000 RPM) may introduce porosity close to the advancing side of welds.

The proper rotational speeds for this alloy range from 1200 RPM to 1600 RPM combined with

selected traverse speeds (1-3 mm/s). From the cross section of the welds, it was observed that

higher rotation speeds increase the width of the thermo-mechanically affected zone (TMAZ).

Better ductility also broadens the parameter domain of this material compared to the cast alloys.

In as-cast alloys, most good and perfect welds occurred for the rotation speed of 1000 RPM.

Narrower domains of proper parameters are due to the lower ductility of these alloys. It was also

observed that above certain limits of rotation speed, severe surface, lack-of-fill defects occurred.

Weld quality indices of both A356 and 319 with pre-weld heat treatment were also characterized

and compared with those in as-cast conditions in Table 7. The weld quality improves

significantly with the pre-weld heat treatment. For A356-T6, the tested domain of processing

parameters (rotation speeds from 600 RPM to 1200 RPM and traverse speeds from 1 mm/s to 3

mm/s) results in good and perfect welds. This indicates that the pre-weld heat treatment creates

more material flow during friction stirring of A356 and lowers the possibility of forming weld

defects. In 319-T6, weld quality also improves, but less than in A356-T6. Good and perfect

welds are limited to certain combinations of rotational and traverse speeds. Thus, it can be

43

concluded that pre-weld heat treatment improves the weld quality and enlarges the proper

processing domains for all alloys.

In alloys 319-F/T6 and A390-F, a triangular distribution of proper processing parameters was

observed, as indicated in Table 7. As from the heat transfer model we have understood the

relationship between processing parameters and heat generation rate and high temperature

duration in FSW/FSP, it is verified by experimental results that a certain amount of heat energy

is required for obtaining good and perfect welds via optimizing the combination of tool rotation

and traverse speeds. In both cases, the lower limits of heat input can be calculated v