Design of experiment study on hardness variations in friction stir welding of AM

S.Richmire,K.Hll,M.Hghshens,∗

a Department of Mechanical Engineering,University of North Dakota,Grand Forks,ND 58202,USA

b Department of Chemical Engineering,University of North Dakota,Grand Forks,ND 58202,USA

Abstract Identificatio of process parameters,their effects and contributions to the outcomes of the system using experimental approach could be a daunting,time consuming,and costly course.Using proper statistical methods,i.e.,Taguchi method,could significantl reduce the number of required experiments and statistical significanc of the parameter can be identified Friction stir welding is one of those welding techniques with many parameters which have different effects on the quality of the welds.In friction stir welding the tool rotational speed(RPM)and transverse speed(mm/min)influenc the strength(i.e.,hardness distribution)of the stirred zone.In this study,these two factors are investigated to determine the effect they will have on the hardness in the stirred zone of the friction stir welds and how the two factors are related to one another for as-cast magnesium alloy AM60 with nominal chemical composition of Mg-(5.5–6.5)Al-(0.24–0.6)Mn-0.22Zn–0.1Si.Experimental data was taken at three different tool rotational speeds and three different transverse speeds.The data obtained was then analyzed using a 32 factorial design to fin the contribution of these parameters.It was determined that both tool rotational speed and transverse speed possess significan effects on the stir zone hardness.Also,the interactions between the two factors were statistically assessed.

Keywords:AM60 Mg alloy;Design of experiment(DOE);Statistical method;Friction stir welding.

1.Introduction

1.1.Friction stir welding parameters and DOE

Friction Stir Welding(FSW)is considered a solid state joining process because the heat generated by the process does not reach the melting temperature of the materials being joined.Developed in 1991[1],FSW allows for the joining of materials that,by traditional welding technologies,are considered to be non-weldable(i.e.,dissimilar material,Mg alloys,some aluminum alloys,etc.)[2].In the FSW process,a non-consumable cylindrical tool with a shoulder and pin is used as a stirrer.In this study,the FSW tool is fied to a displacement controlled milling machine chuck and is rotated clockwise along the longitudinal axis.It is worth mentioning that the FSW is not always performed on a milling machine.The work pieces to be welded are butted up(or overlapped)to one another and clamped in order to stay stationary and withstand the large applied forces.The rotating tool pin is then inserted into the work piece at the faying surface,and the shoulder is forced into contact with the work piece surface.The tool is then moved transversely along the faying surface.As the tool rotates and moves along the joint,the material is softened and stirred together forming a weld without reaching the melting point of the base materials[1,3].At the end of the joint the tool is retracted out of the work piece.Four steps of the FSW have been shown in Fig.1.

Fig.1.Four steps of the FSW including initial set up,tool plunging,traverse,and tool withdrawal.

Complex physical processes like FSW usually involve large number of inter-related variables that directly or indirectly affect the weld quality and the performance.These variables,in the FSW process,are classifie as tool related,joint design related,and machine(welding)related parameters as well as a few others,which do not fall into any of these categories.Tool related parameters include the tool geometry and material type of the tool.Joint design includes the joint type(butt,lap,or fillet)Machine parameters include tool rotation speed(ω,rpm),transverse speed(V,mm/min),axial force(insertion depth),tool tilt(angle of the spindle),and the offset from the center of the joint.Other factors that could influenc the weld quality include the initial temperature of the material(pre-heat or cooling),as well as the cooling rate.It is literally impossible to study the effect of these parameters all together or one by one considering inter-relation among them since an unusually large number of tests must be performed while the overall contributions of the interrelating parameters may not be necessarily detectable.According to engineering logic,one should be able to extract as much information as possible form a process(system)like FSW based on fewest number of essential experiments.

Around 1920 a British scientist,Ronald A.Fisher,originated the concept of design of experiment(DOE)by proposing a systemic approach to extract as much information as possible from experimental tests[4,5].His approach was welcomed by experimentalists around the world as,via employing an appropriate DOE approach,redundant tests and observations are eliminated and therefore time,investment and human power are saved largely.Beside this,the main parameters of the system,their relations,and their contributions toward outcome of the system can be systematically identifie and recorded in a timely manner.In the next sections,we employ a DOE approach to assess the effect of two important FSW parameter,ωandV,on the local hardness measurements of an FSWed die-cast AM60 Mg alloy.

1.2.The alloy used in this study

The Mg alloy used in this study is AM60,a high-pressure die cast alloy which contains 6 wt%aluminum(Al)as the main alloying element and 0.5 wt%Mn.Al is soluble in Mg from 2.1 wt%at room temperature up to 12.6 wt%at the eutectic temperature of 437°C(see Fig.2)[6].Al helps to improve the strength and the castability of the alloy.The microstructure of AM60 consists of course grains of alpha magnesium(α-Mg),and an intermetallic beta phase(β-phase)distributed along the grain boundaries(see Fig.3)[7].β-phase consists of Mg17Al12and forms due to the fast cooling conditions,which are typical of the casting process.Non-equilibrium solidificatio occurs and the intermetallicβphase,which possesses a higher Al concentration than theα-Mg phase,forms and is distributed along the grain boundaries which improves strength of the AM60 alloy[7].Fig.3 shows the microstructure of the as received AM60.The base material(BM)characteristically possesses large grains made up ofα-Mg andβ-phase.This alloy is commonly used for automotive die castings for safety components such as instrument panel structures and seat frames.This alloy offers excellent ductility,energy absorbing properties,strength and castability[8,9].

There are a handful of reported literatures on FSW of AM60 implying that the weld quality of AM60 is largely affected by tool rotational speed and tool transverse speed.According to the published reports,there is a narrow range of parameters that produces good quality welds[10–12].Outside this narrow span,inner voids and/or lack of bonding are observed in the FSWed parts.According to the literature,higher tool rotational rates and lower transverse speeds show better quality welds[1,11].

Fig.2.Mg–Al phase diagram[5].

Fig.3.The Microstructure of as-cast AM60 Mg alloy,(a)OM image,(b)SEM micrograph.Twins are clearly observed in the SEM micrograph.

The FSW process,typically,creates three distinct microstructural zones when welding metallic materials,these zones are the stirred zone(SZ)which is in the center of the weld,the thermomechanically effected zone(TMAZ)which is on either side of the SZ,and the heat affected zone(HAZ)which is located to either side of the TMAZ(see Fig.4).Among these three zones,the SZ is highly subjected to intense plastic deformation applied through a rotating tool which creates high frictional heat.This results in a dynamically recrystallized fin grain structure in the SZ.The TMAZ is subjected to plastic deformation but does not experience recrystallization to the same extent as the SZ because of insufficien deformation strain.That is,the TMAZ might experience a partial recrystallization which might also experiences a coarsening of precipitates at the grain boundaries,due to the high temperatures experienced.This zone is characteristic of elongated grains in an upward flwing pattern around the SZ.The HAZ goes through a thermal cycle but is not subjected to any plastic deformation.The optical images of the SZ,TMAZ,and SZ/TMAZ interface,showing both equiaxed and elongated grains,in the FSWed AM60 are depicted in Fig.5.

1.3.Objectives statement

This study aims at assessing the effects of tool rotational speed and tool transverse speed on the hardness trend within the stirred zone of the FSWed AM60 Mg alloy.It will determine statistically,using an appropriate DOE approach,if the hardness in the stirred zone(SZ)will significantl vary when using different tool rotational speeds and transverse speeds.Throughout the study the following questions will be investigated:

1.What effects do tool rotational speed and transverse speed have on the hardness of the friction stir welded joints and are they significant

Fig.4.Various zones produced due to the FSW run,(a)base metal,(b)heat-affected zone,(c)thermo-mechanically affected zone,and(d)weld nugget zone(stirred zone)[16].

Table 1 Data collection scheme including different tested rotational and travel speeds and the corresponding hardness values.

2.Are the interactions between the tool rotational speed and transverse speed significant

3.What are the contributions of these parameters(RPM and mm/min)on hardness quantities?

2.Sample preparation procedures

The samples were prepared by welding two similar sheets of AM60 together using the FSW principles in butt config uration utilizing a displacement controlled milling machine.Fig.6 shows an example of a successful FSW run showing advancing and retreating side.The FSW trails were performed at several tool rotational and transverse speeds as provided in Table 1.Upon the completion of FSW tests,a small crosssection of each weld condition was cut from the weld and mounted using an epoxy resin mixture.The samples were then mechanically polished until mirror-like surfaces were achieved.Five hardness values were taken from the SZ of each sample using a Vickers hardness tester at 300 g and 15 s and the average number was recorded.For microstructural assessment,a separate set of samples were polished and etched using an etchant made up of 4.2 g picric acid,10 mL water,10 mL acetic acid,and 70 mL ethanol and images were recorded using an Advanced Metallurgical Microscope model EQ-MM-500T-USB.

The two factors that were studied in the experiment were tool rotational speed and transverse speed.These factors were chosen because they are the two parameters that,according to the literature,possess the largest effect on the quality of the friction stir welds and are fairly convenient to measure and keep consistent throughout the entire experimentation.To minimize the effect of other contributing factors,the tool type,geometry,tool tilt,and joint design were all kept constant;same type of material(similar manufacturing history)was used and no post processing(i.e.,heat treatment)was performed on the FSWed sheets.It should also be noted that the axial depth and offset from the center of the joint are factors that could not be precisely measured but were kept as close as possible to the same for every friction stir weld.

3.Designing experiments

In the parametric design of the experiments the factorial method is used to study the effect of tool rotational and travel speeds on the hardness of the joint.To analyze the effect of the two parameters,a 32factorial design was chosen.The 32factorial design has three level of each of the two factors.The study used fied factors,three tool rotational speeds(635,1200,and 1530 RPM)and three transverse speeds(48,220,and 422 mm/min),resulting in 9 experimental units.Each of the 9 experimental units contains 5 replicates.With three levels of each factor,there are now points in the middle that allow for a curved firesponse function.The design used is shown below in Table 2 and Fig.7.This design allowed for an economical investigation into the cause and effect relationships of the significanc of the two variables at hand.

4.Result and analysis of data

Table 3 shows sum and mean of the collected hardness data at different tool rotational and tool travel speed.The main effects(αiandβj)are calculated by Eqs.(1)and(2):

Fig.5.Microstructure of the FSWed AM60 alloys at different zones.

Fig.6.Example of successful FSW completed.

Table 2 The 32 Factorial arrangement;0,1,and 2 stand for low,medium and high quantities,respectively.

The main effects are calculated in Table 4 and displayed graphically from the Minitab output in Fig.8.

The interaction effects are calculated using Eq.(3).

The interaction effects are calculated in Table 5 and displayed graphically from the Minitab output in Fig.9.

Fig.7.A 32 design dcheme(0,1,and 2 stand for low,medium,and high quantities,respectively).

Analysis of variance(ANOVA)[13–15],a collection of statistical models and their associated procedures(such as"variation"among and between groups)used to analyze the differences among group means,were done by using Eqs.(4)–20 in Appendix A,consideringn(number of replicates)=5,a(number of levels tool rotational speed)=3,andb(number of levels of the transverse speed)=3.ANOVA calculations were also made in Minitab.The results from the output calculations are displayed in Table 6 and the output from Minitab is displayed in Appendix B.

Upon calculating the results,the model adequacy needed to be verified Fig.10 exhibits four different plots used to verifythe adequacy of the model.These plots include the normality assumption,residuals,histogram,and the residuals versus observation order,that verify the adequacy of the model.None of these plots exhibit anything out of the ordinary.The normality plot shows no severe indication of non-normality,nor evidence of any possible outliers.The residuals plot reveals nothing abnormal or of unusual interest.Histogram plots show the central tendency,spread,and variability.This histogram also shows nothing from the ordinary as with the residuals versus observation order.

Table 3 Sum and mean of collected data.

Fig.8.Minitab output showing the main effects plot for the hardness values.

Fig.9.Minitab output showing the interaction effects plots for the hardness values.

The purpose of this study was to determine if changing the tool rotational speed and the transverse speed influence the hardness of the friction stir welded joints.Looking at the ANOVA results(Table 6)theFrotationalspeed=158.37 andFTransverse speed=92.83,for bothFcritical=3.26 with aP-value of almost 0.Therefore,the main effect of both the tool rotational speed and the transverse speed are significantl different and do influenc the hardness.Also,looking at the interactionsFinteractions=75.29 with theFcritical=2.63 and again aPvalue of almost 0,it can be concluded that there is a significan interaction between the tool rotational speed and the transverse speed.From the main effects plots(Fig.10),it is shown that the lower tool rotational speed and transverse speed yield higher hardness value.Looking at the interaction plots(Fig.9),it is visually shown there are significan interactions between the tool rotational and transverse speeds by the lack of parallelism of the lines.

Fig.10.Plots of model adequacy checking.

Table 4 Main effects.

Table 5 Interaction effects between parameters.

Table 6 ANOVA statistical estimates.

5.Summary

Friction stir welding was used to join two similar sheets of AM60(cast magnesium alloy)in a butt joint.It was desired to determine the effect of tool rotational speed and transverse speed on the hardness in the stirred zone of the friction stir welds.This was accomplished by firs creating welds of each variation of transverse speed and tool rotational speed,followed by hardness tests in the stirred zone of the welds.The experimental data collected was then analyzed in a 32factorial design.The main effects and interaction effects were calculated.An ANOVA analysis was done on the data,along with a normal probability plot to show that the data came from a normally distributed population.

The main effects were calculated by hand and in Minitab.The main effects plot from Minitab show that as tool rotational speed and the transverse speed is decreased,the hardness value increases.The interaction effects were also calculated by hand and in Minitab.The interaction plots from Minitab show that the interactions are significant At lower tool rotational speeds(RPM)and higher transverse speeds(mm/min),the hardness values are low.At lower tool transverse speeds(mm/min)and higher tool rotational speeds(RPM),the hardness value is again low.When both factors are high,the hardness is at its lowest point.Also,in agreement with the main effects,when both factors are at their lowest setting,the hardness is at its highest value.

Through the ANOVA analysis it was determined that the main effects were significantl different and both have an effect on the hardness.It was also found that there is a significant interaction between the tool rotational speed and transverse speed.Both of these conclusions are in agreement with the main effects and interaction effects calculations and plots.

Time restraints made it impossible to continue past the current point.It would have been ideal to optimize tool rotational and transverse speed in this experiment.Given more time,it is recommended to determine the line of steepest ascent as well as to further explore the maximum area and determine the optimum values.It is also recommended to repeat the DOE experiment to determine the maximum fatigue life.

Appendix A

A1.Sum of squares equations

A2.Degreed of freedom equations

A3.Mean square equations

A4.F-value equations

Appendix B

B1.Two-way ANOVA:hardness versus RPM,mm/min

B2.One-way ANOVA:hardness versus RPM

B3.One-way ANOVA:hardness versus mm/min