International Journal of Machine Tools & Manufacture 57 (2012) 46–54

Contents lists available at SciVerse ScienceDirect

International Journal of Machine Tools & Manufacture

0890-69

doi:10.1

n Corr

Tel.: þ1

E-m

journal homepage: www.elsevier.com/locate/ijmactool

Effect of process parameters on the rate of abrasive assistedbrush deburring of microgrooves

George Mathai n, Shreyes Melkote

The George W. Woodruff School of Mechanical Engineering, Georgia Institute of Technology, Atlanta, GA, United States

a r t i c l e i n f o

Article history:

Received 12 October 2011

Received in revised form

13 February 2012

Accepted 20 February 2012Available online 27 February 2012

Keywords:

Micromilling

Deburring

Loose abrasive brushing

55/$ - see front matter & 2012 Elsevier Ltd. A

016/j.ijmachtools.2012.02.007

espondence to: 813 Ferst Drive, MaRC 380, Atla

404 683 3262; fax: þ1 404 894 4133.

ail address: georgekm@gatech.edu (G. Matha

a b s t r a c t

Burrs in micromilled parts need to be removed for proper functioning of the parts. Abrasive assisted

brushing presents a fast and effective method for deburring these parts. The deburring rate in abrasive

brushing depends on the workpiece material, abrasive grit size, type and rotational speed of the brush.

In this paper the effects of these variables on the rate of deburring micromilled grooves in copper and

tool steel workpieces are studied experimentally. It is found that the burr removal rate is proportional

to the initial burr height and deburring speed. An empirical model derived from the experimental data

is used to predict the deburring time. Groove depth change after deburring is less than 8 mm and

surface roughness improves by about 100 nm in all cases studied.

& 2012 Elsevier Ltd. All rights reserved.

1. Introduction

Presence of burrs in micromilled parts prevents their properworking. In micro fluidic channels they can increase the resis-tance to flow [1]. In biomedical applications, burrs increase thechance of autoimmune rejection of the inserted component [2].

Burr formation in machining has been minimized by eithermodifying the machining process or via a subsequent finishingoperation. The machining process can be modified by proper toolpath planning, changing the tool nose geometry and exit angle,and proper selection of feeds and speeds so that burrs do not formor they are removed during machining [3–7]. Burrs can also bereduced using alternate machining processes such as EDM. Whilethese techniques can reduce the burr size, the workpiece stillrequires some deburring after machining. Commonly employeddeburring operations at the macroscale include brushing, abrasivepolishing and abrasive peening [8]. Laser deburring has also beenapplied [9]. Magnetic brush deburring of microburrs has also beenstudied and shown to have promise for deburring holes [10].At the microscale, ultrasonic wet peening has been used to deburrtool steel. However, it does not fully eliminate the burrs [11].Deburring using a magnetorheological slurry [12] has been usedto deburr grooves in stainless steel and brass. This method,however, requires additional equipment and a system to managethe magnetorheological fluid.

ll rights reserved.

nta, GA 30332, United States.

i).

Brush deburring of micromilled features presents a simple andinexpensive method for deburring of micromilled parts. Theprocess involves reciprocating a rotating nylon brush over thepart to be deburred in the presence of an abrasive slurry (Fig. 1).It can also be coupled with mechanical polishing to achieve bothdeburring and the required surface finish. Another advantage ofthis method is that it can be performed on the same machine usedto produce the micro part. In spite of these advantages, abrasiveassisted brush deburring of micromilled features has not beenstudied.

In this paper, we seek to experimentally characterize thedeburring rate in abrasive assisted brush deburring as a functionof the process parameters. The process parameters considered inthis study are the abrasive grit size, abrasive type and spindlespeed. A previous study of the process focused on the time takenfor complete deburring of micromilled grooves [13]. Since thetime for complete deburring varies with process parameters, theireffect on surface finish and change in groove depth is confoundedwith deburring time. This difficulty is overcome in the currentstudy by keeping the deburring time constant. Moreover, thepresent study takes into account the effect of initial burr heighton the deburring rate and examines two different workpiecematerials.

2. Experiment design

A full factorial experiment was conducted to quantify theeffects of abrasive grit type, size and spindle speed. The brush wasrotated at a spindle speed that is limited at the lower end bymachine capability and the desired burr removal rate. It is limited

burrs

abrasivegrits

bristles

groove

direction ofrotation

x

y

Fig. 1. Schematic of abrasive assisted brush deburring process. Brush reciprocation in

z direction.

Table 1Characteristics of abrasive grits used for deburring.

SiC Diamond

3 mm 1.25 mm 3 mm 1 mm

Zeta potential (mV) �29.8 �26 �31.1 �35.5

Mean size (mm) 2.93 1.08 2.16 1.22

Polydispersity index 0.628 0.244 0.698 0.305

10µm

entry

exit5mm

stylusburr

rootgroove floor

Fig. 2. (a) Top view of artifact to be deburred and (b) burr measurement.

G. Mathai, S. Melkote / International Journal of Machine Tools & Manufacture 57 (2012) 46–54 47

at the higher end by flaring of the bristles, which also reducesburr removal rate and damages the brush. When abrasives areused, high rotational speeds result in the particles being sweptaway from the area where burrs need to be removed. The lowestspindle speed possible in the setup used is 5000 rpm, while thehighest speed recommended by the brush manufacturer is15,000 rpm. Hence, the three speeds chosen for the experimentwere 5000 rpm, 10,000 rpm and 15,000 rpm.

Abrasive slurry was applied to the groove to increase thedeburring rate. Commonly used abrasives are diamond, siliconcarbide and alumina. Alumina is usually used for very smallamounts of material removal in polishing operations [14]. Dia-mond slurries have long life and usually yield a much bettersurface finish. However, they are expensive. In this study, siliconcarbide (Black SiC, Washington Mills) and diamond (AdvancedAbrasives Corp.) abrasive particles were used to assist the brush-ing process. Two nominal grit sizes (1.25 mm and 3 mm) wereconsidered for the process. The slurry consisted of one part byweight of abrasive powder in 10 parts by weight of de-ionizedwater. This ensured that the material removal is by purelymechanical means and not by chemical action. The de-ionizedwater also helps to cool the brush and extend brush life.

Burr properties and removal rates vary with the hardness ofthe work material. Hence, burr removal in copper alloy 110 (99.9%Cu, 95 HV) and A2 tool steel (215 HV) are studied in this paper.All experiments were replicated three times.

The abrasives used for deburring were characterized in adynamic light scattering system (Zetasizer Nano, Malvern Instru-ments) yielding the measurements listed in Table 1. The meansizes are close to the specified grit sizes. Polydispersity indexmeasures the dispersion of particle sizes. Values above 0.7 indi-cate a broad distribution of particle sizes. Note that largerparticles have a much higher dispersion than smaller particles.The zeta potential measures the likelihood of the grits agglomer-ating. Potentials whose absolute values are above 30 mV areconsidered stable and hence less likely to agglomerate. Diamondshows much better stability than SiC. The SiC grits exhibit a mildtendency to agglomerate.

The normal force applied by the brush was limited at the lowerend by very low burr removal rates leading to long process times.It was also limited at the higher end by excessive buckling ofthe nylon bristles of the brush and damage to the workpiece.

The buckling of the bristles causes the sides of the bristles insteadof their ends to come into contact with the workpiece. A priorstudy [13] indicated that deburring time is more sensitive tochanges in speed than changes in the normal load. Hence, in thisstudy, the normal load was fixed at 1 N.

3. Experimental procedure

3.1. Sample preparation

Micro-grooves were milled using a micro milling machinedesigned and built in-house and are shown schematically inFig. 2(a). The artifact to be deburred consists of a 5 mm longgroove that is 50 mm deep and is cut in a single pass by a 500 mmdiameter, two flute carbide end mill (PMT, TR-2-0200-S) spun at60,000 rpm with a feed rate of 500 mm/min. The entry and exitportions of the groove and associated burrs are affected by toolentry and exit dynamics. Hence, entry and exit sections areprovided in the grooves to allow the 5 mm section being studiedto remain unaffected by entry and exit transients. Since groovedepth can be affected by workpiece positioning errors and surfaceirregularities, the grooves were produced on a pre-milled refer-ence surface.

The burr removal rate was evaluated by partially deburring thegrooves for a fixed time. The difference in burr height before andafter deburring was then calculated. Since the deburring time isfixed, the difference in burr height can be used to calculate theburr reduction rate.

Abrasive assisted brush deburring is a fairly quick process.Hence, if the burrs produced in micro milling are small, they willbe removed in a very short time and the difference in burr heightreduction due to change in the process parameters cannot beeasily measured. In order to accentuate the effects of theseparameters, it was necessary to produce fairly large burrs. Burrheight generally increases with tool wear and feed [4,15]. Hence,the tool used for machining the reference surface was also usedfor machining the grooves at a high feed, thus wearing-in the tool.This resulted in much larger burrs than would be seen if a newtool was used at feeds recommended by the tool manufacturer.

3.2. Burr measurement

Nominally, the calculation of burr removal rate requires thevolume of the burr to be known. Typical approaches for burrmeasurement include optical, tactile and destructive methods[16]. Optical methods include the volume under the rollover burrscommonly seen in micromilling [3], leading to an error in thecalculated volume. Tactile measurement devices give line tracesof the surface and hence, cannot be used to calculate the burrvolume but can quantify burr height. Destructive methods such assectioning [16], makes subsequent deburring impossible. Hence,there is no reliable method to calculate burr volume. Material

G. Mathai, S. Melkote / International Journal of Machine Tools & Manufacture 57 (2012) 46–5448

removal rate in macroscale deburring has been calculated bymeasuring the weight of the sample before and after deburring[17]. The abrasive assisted brush deburring process also involvesa polishing process. Hence, measuring the sample weight yields ameasure of the total material removed by polishing anddeburring.

A metric often used to characterize burr size [6,15] and burrreduction rate [10,18] is the burr height, which can be measuredaccurately by a profilometer or a laser scanner either in-processor off-line. This feedback can then be used in the subsequentdeburring operation to predict deburring time. Hence, in thisstudy, deburring rate has been characterized by the change inburr height during the operation.

The initial burr height in turn depends on tool wear. Conse-quently, there will be some variation in burr height from the firstgroove to the last groove machined using the same tool. Hence,subsequent grooves in the order of machining are deburredat different spindle speeds. Thus, any variation in burr heightdue to tool wear is distributed more or less equally over allspeeds. Moreover, the model used to evaluate the effect of speedseparates the effects of burr height from the effects of spindlespeed.

Burr height is measured before and after deburring usinga surface profilometer (Taylor-Hobson, Form Talysurf) with a397 mm diameter spherical sapphire stylus. The large size of thestylus tip makes it easier to pick out the high points of the burr ina small region. The stylus was aligned with the length of thegroove and positioned on top of the burr using a microscopeunder 50X magnification.

The burr height at any point along the groove length varies inthe direction of the groove width. Hence, in order to find thehighest point of the burr at any point along the length of thegroove, three traces offset 10 mm from each other in the directionof the groove width were taken. The burr height at any point wastaken as the highest value for that point among the three scans(see Fig. 2b). Burr heights were measured on both the upmillingand downmilling sides of the groove. Similarly, groove depth wasmeasured before and after deburring with the same profilometerbut with a 2 mm effective diameter diamond stylus tip.

machiningmarks

downmillingside

upmillingside

top surface

200 µm

100 µm

100 µm

Before

groove

machiningmarks

Fig. 3. Burrs in milling of A2 tool steel before and after deburring with 3 mm SiC grits at

3.3. Deburring

The micromilled grooves were deburred using a nylon endbrush (Dremel, Bristle brush 405) mounted in a Dremel spindle(Dremel 4000). The brush had a diameter of 3.5 mm and 50 mmdiameter bristles. It was positioned above the artifact to bedeburred and spun at the required speed and reciprocated alongthe groove length at 300 mm/min. A few drops of abrasive slurrywere applied on the artifact. The normal force exerted by thebrush was kept constant at 1 N using a counterbalance weightsystem. The spindle is free to move vertically in a lubricatedbushing. The spindle weight is balanced by the counterbalanceand the normal force between the workpiece and the brush.Hence, the normal force between the brush and the surface can becontrolled by changing the counterbalance weight. The normalforce was monitored using a piezoelectric force platform (KistlerMiniDyn 9256C2).

Deburring experiments were carried out for a constant dura-tion of 65 s in all experiments reported here. This duration wasselected such that deburring was stopped before complete burrremoval, thus allowing for the estimation of burr height reduc-tion. The burr reduction rate was calculated by dividing the burrheight reduction by the deburring time. The effects of spindlespeed, grit size and abrasive type on burr height reduction ratewere then evaluated.

4. Experimental results

4.1. Burr types

Figs. 3 and 4 show the typical burrs observed in micromillingof tool steel and copper respectively. The burr size seen in copperis much larger than that seen in tool steel because copper is moreductile [8]. The burrs in the grooves can be categorized as tear andPoisson burrs. Tear burrs are caused when material tears apartinstead of being sheared by the tool [8]. Poisson burrs form wherethe material is plastically ploughed up near the edge instead ofbeing sheared to form a chip. Typically, in both materials the burr

After

abrasive grits

side wall

marks

groove

groovemachining

machiningmarks

top surface

polishedsurface

100 µm

100 µm

100 µm

15,000 rpm (a) both sides of groove, (b) down milling side and (c) up milling side.

500 µm

300 µm

300 µm

Before After

root

machining marks

downmilling side

upmilling side

groove floor

top surface

top surface

groove floor

polished surface

abrasionmarks

Fig. 4. Burrs in milling of copper before and after deburring with 3 mm SiC grits at 15,000 rpm (a) both sides, (b) down milling side and (c) up milling side.

direction of motionof bristle

freetrajectory ofbristle tip

unshielded regionshieldedregion

Region IRegion IIRegion III

upmillededge

shieldedregion

Region II Region III

Region I direction of

motion of bristles

incline

burrs

0123456su

rface

pro

file

(µm

)

7

scan length (mm)1 2 3

top surface

Region Iunshielded region

200 µm

shielded region

Fig. 5. Effect of shielding of surface by burrs. (a) Image of surface, (b) trace of surface of a completely deburred groove and (c) schematic of motion of brush.

G. Mathai, S. Melkote / International Journal of Machine Tools & Manufacture 57 (2012) 46–54 49

is a combination of tear and Poisson burrs. The tear burr isattached to the groove edge at the burr root. At the root, the burris formed by material being ploughed by the tool edge. As a result,the root is much thicker than the rest of the burr and also thetoughest to remove. The root region gets exposed during debur-ring and can be seen clearly on partially deburred edges. Sincetool steel is harder and less ductile than copper, the tool steel

burrs have a much more well defined shape than the copperburrs. It can also be seen that burrs on the downmilling side of thegroove are larger than burrs on the upmilling side. This is becausethe torn burr gets cut on the upmilling side and is swept towardsthe downmilling side. On the upmilling side, the burrs are mostlyof the Poisson type, although a few small tear burrs are alsopresent.

varying initialburr height

almost constantburr root heightafter deburring

Fig. 6. Sample trace of groove edge before and after deburring of A2 tool steel at

15 krpm with 3 mm SiC grits.

G. Mathai, S. Melkote / International Journal of Machine Tools & Manufacture 57 (2012) 46–5450

4.2. Visual inspection of deburred grooves

Fig. 3 shows images of a groove in tool steel deburred with3 mm SiC grits at 15 krpm for 65 s. It can be seen that burrs on thedown milling side are larger than on the up milling side. The burrsare feathery with a mostly thin cross section except at the root.It can be seen that machining marks on the top surface of thegroove have been removed by the deburring process. However,machining marks on the groove base are still present afterdeburring. Note that, prior to imaging, any residual abrasives onthe deburred surfaces were removed using isopropyl alcohol anda light brush (Crayola 3515).

Fig. 4 shows a groove in copper before and after deburringwith 3 mm SiC grits at 15 krpm for 65 s. A much larger residualburr is seen in this case because the initial burrs were fairly large.The downmilling side burrs have a smoother surface afterdeburring. This suggests that while the slender burrs are removedby breaking or tearing, thicker burrs and burr roots are graduallyremoved by abrasive action of the grits. The upmilling side burrsare removed quickly due to their slender shape to give a welldefined edge. Machining marks have been removed from the topand bottom surfaces of the groove. However, unlike tool steel,scratch marks on the surface due to action of the abrasive grits aremuch more pronounced. This is due to the fact that copper issofter than tool steel.

Fig. 5(a) shows an unbrushed area next to the burrs (Region I).This can be explained by noting that in the presence of a largeburr, the bristle follows the trajectory shown in Fig. 5(c). When itcontacts the burr it bends and travels along the top of the burr.When the tip of the bristle reaches the end of the burr it isreleased and follows a curved path until it contacts the workpieceagain at the start of Region II. Hence, there is a small region nearthe burr that is shielded from action of the abrasives embedded inthe bristles. A consequence of this is that the unshielded regiongets abraded more than the shielded region. This causes adepression in the top surface from the groove edge towards theouter edge of the brushed area. This incline can be seen in a fullydeburred groove shown in Fig. 5(b). It is also seen that Region II isabraded more than Region III edge although both surfaces areexposed to abrasives for the same time. The bristle in Fig. 5(b)behaves like a bent cantilever spring that is released from theburr. It is hypothesized that the excessive abrasion in theunshielded region is due to the impact of the bristles againstthe surface after being released from the burr. For the sectionshown in Fig. 5(b), bristle impact causes a depression about 5 mmin depth.

4.3. Burr reduction rate

A trace of a downmilled tool steel edge deburred with 3 mmSiC grits and 15 krpm spindle speed is shown in Fig. 6. It can beseen that there is significant reduction in burr height in a veryshort time. While the initial burr height varies significantly, theheight of the residual burr is fairly constant. Hence, for the samedeburring time, the height reduction is greater for tall burrs thanfor short burrs. Hence, it is inferred from the plot that burr heightreduction is proportional to initial burr height.

This behavior can be explained by examining the shape of theburrs shown in Fig. 3. The burrs in A2 tool steel tend to be slenderand tear-off or break upon contact with the bristles and/orabrasive leaving only the burr root behind. If the burr root heightis nearly the same for both large and small burrs, then when a tallburr is broken by the combined action of the bristles andabrasives it results in a larger burr height reduction than whena small burr breaks. Hence, tall burrs show a higher burrreduction rate than short burrs. However, it is expected that this

trend will change when the burr root is removed. In this case, thedeburring rate will be proportional to the area of the root incontact with the brush.

These findings are consistent with the work of other research-ers [12] who suggest that there are two types of burr removalmechanisms: extensive yielding (Type 1) and abrasive wear (TypeII). Extensive yielding occurs for thin sheet-like burrs that bendunder the action of the deburring force and undergo failure at theroot when the stress exceeds the failure stress of the burrmaterial. For an average pressure P0 acting on a burr of height x,thickness tb and inclined to the reference surface at an angle y, thestress s generated at the root during Type 1 mechanism is givenby Eq. (1). It can be seen that the stress is proportional to the burrheight x when all other factors are constant. For Type II mechan-ism, it is suggested in [12] that the burr root material is plasticallydeformed without any material removal by stress concentrationbeyond yield. The abrasion marks on the surface of the materialseen in Figs. 3 and 4 however seem to suggest that material isremoved by chip formation. In this case, the proportionalitybetween burr removal rate and initial burr height will not hold.Furthermore, it has been observed in the present study thatremoval of sheet-like burrs occurs due to mode I and III fatiguefracture at the burr root

s¼ P0tbx2

c2siny ð1Þ

4.4. Statistical model for burr reduction rate

A quantitative evaluation of deburring speed can be obtainedby fitting a linear regression model to the burr reduction rate y inmm/min. The factors in the statistical model given by Eq. (2) arethe initial burr height x in mm, spindle speed N in krpm, abrasivegrit size a in mm and abrasive grit type b [19]. The effect of grittype is captured by setting b to 0 if SiC grits are used and to 1 ifdiamond grits are used. Since the burr height reduction rate atevery point along the groove edge depends on the initial burrheight at that location and the initial burr height cannot bechosen apriori, the initial burr height is a covariate for burr heightreduction rate

y¼ a0þa1xþa2Nþa3aþa4bþa5Naþa6Nbþa7abþE ð2Þ

A comparison between the regression fit to the experimentaldata for deburring of A2 tool steel with 3 mm SiC abrasives at15 krpm is shown in Fig. 7(a). An almost linear dependenceof burr reduction rate on the initial burr height can be observed.

G. Mathai, S. Melkote / International Journal of Machine Tools & Manufacture 57 (2012) 46–54 51

A cursory look at Fig. 7 may suggest that the model does not fitthe data accurately. However, note that the linear model is fittedto the data for the entire experiment design while the 2D plot inFig. 7 allows for only one subset of the data. The fitted equationhas an R2 value of 0.76. This value is low probably due to theuncontrolled nature of the free abrasives based process.A possible remedy to this problem is to use abrasive impregnatedbrushes for better process control.

The regression analysis in Table 2 indentifies initial burr height x,brush rotational speed N and grit type b to be statistically significantat a 10% level. Similar results are presented for copper in Table 3. Itis interesting to note that the coefficient of the initial burr height a1

is about 0.5 for both copper and tool steel. This is explained by thetypical shape of the burrs, which tend to be thick at the root andthin away from the root. Therefore, they are more prone to bendand/or break right above the root. The value of the coefficient a1

seems to suggest that the burr either breaks or bends at abouthalf its initial height. The analysis also indicates that diamondabrasives have a higher burr reduction rate than SiC. Changing from

Table 2Coefficients for linear regression model for burr reduction rate for tool steel.

Intercept a0 Initial burr height

a1 (mm)

Speed a2

(krpm)

Size a3 (mm)

Coefficients �7.68 0.49 1.02 �1.54

p-value 0.28 0 0.1 0.59

Table 3Linear regression model coefficients for burr reduction rate in copper.

Intercept a0 Initial burr height

a1 (mm)

Speed a2

(krpm)

Size a3 (mm)

Coefficients �41.89 0.53 2.43 4.28

p-value 0 0 0.03 0.42

burr

redu

ctio

n ra

te (µ

m/m

in)

initial burr height (µm)

burr

redu

ctio

n ra

te (µ

m/m

in)

initial burr height (µm)

Fig. 7. Burr height reduction rate as a function of initial burr height for deburring

at 15 krpm with 3 mm SiC grits (a) A2 tool steel and (b) copper.

SiC to diamond grits causes an increase in deburring rate of about13 mm/min in tool steel and of about 19 mm/min in copper. This canbe explained by noting that diamond grits are harder than SiC gritsand maintain their sharpness longer than SiC.

The deburring rate also increases with speed as indicated bythe positive coefficient a2. A higher brush rotational speed impliesa higher number of passes of the brush over the burr per unit timefor a constant feed along the length of the groove. This implies ahigher number of passes of the abrasives on the burr and hence ahigher burr reduction rate. This increase in material removalsufficiently overcomes any reduction in material removal due toscattering of the grits by the centrifugal forces. In polishing 90%material removal is assumed to come from two body interactionand only 10% from three body interaction [20]. Hence, materialremoval is assumed to occur due to abrasives embedded in thebristles (Fig. 8). The centrifugal forces are not strong enough todisengage the grits embedded in the bristles as observed fromimages of the bristles after deburring. Furthermore, the rotationalkinetic energy of the grits should be high enough to overcome theresistance offered by the densely packed bristles. Since speed andtype are significant, for tool steel the speed-type interaction termis also indicated as being slightly significant.

Models used for polishing may be applied to analyze abrasiveassisted brush deburring since both are free abrasive processesthat use a soft backing media. It has been suggested in [21] that inpolishing involving material removal (i.e. chip formation insteadof bulk plastic flow), the Mulhearn model [20] used for polishingwith emery paper may apply, which suggests that the materialremoval rate is proportional to the depth of penetration of theabrasive. A similar microcutting model [22] based on Archard’swear equation is shown in Eq. (3), which shows that the materialremoval rate dV=dt is proportional to the relative velocity vr

between the grit and the workpiece (L is the load on the abrasives,H is the hardness of the workpiece, and kw is a constant). This is inagreement with the experimental results presented in the present

Type a4

(SiC¼0, Dia¼1)

Speed� size a5 Speed� type a6 Size� type a7

13.05 0.3 �0.79 0.03

0.05 0.23 0.11 0.99

Type a4

(SiC¼0, Dia¼1)

Speed� size a5 Speed� type a6 Size� type a7

19.34 0.17 1.04 1.81

0.12 0.7 0.26 0.64

SiC grits

nylonbristle

10 µm

Fig. 8. SEM image of a nylon bristle with 3 mm SiC abrasive grits embedded in it.

G. Mathai, S. Melkote / International Journal of Machine Tools & Manufacture 57 (2012) 46–5452

study. It is also suggested in [22] that the depth of penetration isproportional to hardness of the grit. Based on Mulhearn’s model,increased penetration would also mean higher material removalrate, which supports the findings in the present study that theharder diamond grits yield a higher burr reduction rate than SiC

dV

dt¼

kwLvr

Hð3Þ

It is surprising to note that the effect of grit size is not verysignificant. While this warrants further investigation, one probableexplanation is that for a constant grit concentration by mass,smaller grits would have a larger number of abrasives in contactwith the workpiece. However, larger grits also remove morematerial per pass for similar indentation depth. It has been shownthat abrasion rate in polishing is independent of grit size up to acritical grit size below which abrasion rate is proportional to gritsize [20]. For the experiments in [20] the critical value was 50 mm.However, the load in this case was much higher than in the presentstudy at (about 5 N with SiC abrasives paper on a cold-drawn steelworkpiece). Another possibility for the insignificance of grit sizecould be the narrow range of sizes used in the experiment. Notethat a wider range of abrasive grits sizes cannot be used in thisstudy due to excessive damage caused to the micro-scale featureby larger grit sizes under the deburring conditions used. It has beenobserved that 6 mm diamond grits can cause over 50% reduction inthe groove depth, which is clearly undesirable.

A comparison between the regression model coefficients fortool steel and copper shows that the coefficients for copper aremuch higher than tool steel for all the factors. The effect of speed ismuch more pronounced in copper as indicated by the much lower

Fig. 9. Verification of model: full deburring with 3 mm SiC grits at 15 krpm for predic

before and after deburring for copper, (c) tool steel groove cross section profile after d

p-value for copper. The higher sensitivity of copper to the debur-ring parameters is attributed to the lower hardness of copper. Aregression fit for only the significant factors is performed on thedata and the resulting equations are shown in Eq. (4) for tool steeland Eq. (5) for copper. The speed-type interaction factor has beenretained for copper for the sake of consistency. Surface roughnessin all cases improves by at least 100 nm and the mean change indepth is usually about 2 mm with a maximum of 4 mm for 3 mmdiamond abrasives used on a copper workpiece at 15 krpm

y¼�10:45þ0:49xþ1:62Nþ13:04b�0:79NbþE ð4Þ

y¼�33:72þ0:54xþ2:78Nþ22:86bþ1:04NbþE ð5Þ

5. Verification

The deburring rate predicted by the model in Eq. (2) is verifiedexperimentally for tool steel and copper. The conditions chosen forthis experiment are 3 mm SiC grits with a spindle speed of 15 krpm.The deburring rate for these conditions is given by Eq. (4) for toolsteel and Eq. (5) for copper. A trace of the burr height for a groove istaken and the time t in minutes required for complete deburring ofthe edge is obtained by dividing the burr height x by the deburringrate as shown in Eqs. (6) and (7). The grooves are deburred for thehighest value of time obtained from these equations. For tool steel,this time comes to be 107 s while for copper it is 81 s. Note thateven though the burr reduction rate y is always proportional to theinitial burr height x, deburring time t may have a positive ornegative exponential relationship with initial burr height dependingon the values of the coefficients in the model. However, in bothcases, the quantity of interest is the maximum time required to

ted time. (a) Burr height before and after deburring for tool steel, (b) burr height

eburring and (d) copper groove cross section profile after deburring.

G. Mathai, S. Melkote / International Journal of Machine Tools & Manufacture 57 (2012) 46–54 53

deburr a particular edge i.e. the maximum value of t from Eqs. (6) or(7) given the values of x along the edge

t¼x

�10:45þ0:5xþ24:3ð6Þ

t¼x

�33:72þ0:54xþ41:7ð7Þ

The burr height before and after deburring is shown in Fig. 9 for thedownmilling side of the groove, which is generally characterized bylarger burrs. It can be seen that most of the groove has beencompletely deburred. There is a significant residual burr at thebeginning and end of the groove where the tool rounds the corner ofthe entry and exit section (see Fig. 2a). The model drasticallyunderestimates the time taken to deburr this region. This can beexplained by the fact that the portion of the burr at the corners isheavily strain hardened and hence the root of the burr for thisregion is much thicker than the straight portion of the groove.In calculating the coefficients of the model, about 80% of the datacomes from the straight portion of the groove instead of the corner.Hence, the model predicts the time for deburring the straightportion of the groove much better than the corners. Unlike toolsteel, the copper specimen shows an almost uniform and completeremoval of the burrs. Some small residual burrs of the order of 3 mmare visible at the ends of the groove for reasons similar to thosediscussed for tool steel.

Note that the straight portion of the groove lies between 1 and4 mm of the trace along the groove edge. This portion is almostcompletely deburred. Table 4 summarizes the characteristics of theburrs in this region before and after deburring. It can be seen thatthere are a few residual burrs in this region with a maximum height

Table 4Burr characteristics for region between 1 and 4 mm of trace length.

Tool steel Copper

Before After Before After

Mean (mm) 173 1.3 143 1.24

Standard dev. (mm) 23.6 1.1 31.1 0.6

Max (mm) 234.5 3.5 239.6 3.04

Befor

Afterno edge damage

top surface

machining marks

top surface

top surface

top surface

to

200 µm 2200 µm

200 µm200 µm

Fig. 10. Burrs in milling of before and after full deburring time predicted by model for 3 mm

tool steel, up milling side, (c) copper, down milling side, (d) copper, up milling side. Aft

(g) copper, down milling side and (h) copper, up milling side.

of about 3.5 mm for tool steel. Thus, the model for tool steel slightlyunderestimates the burr reduction rate in this region. This is due tothe fact that the burr root is much more difficult to remove than thefeathery burr. It is interesting to note that even though the burr rootwas not explicitly considered by the model, most of the root hasbeen removed (the standard deviation of the edge height is about1 mm). The edge condition after deburring for copper in the straightportion (between 1 and 4 mm) of the groove length has undulationsof the order of a few microns which is about half the size of theabrasive grit used for deburring and may be because of abrasivemarks. Both copper and tool steel show a slight dip in the profiletowards the center due to less strained material and smaller burrstowards the center. SEM images scan of the groove are shown inFig. 10. The burrs in copper have been completely removed fromboth edges. Irregularities along the edge are mostly due to irregula-rities along the wall of the groove. The mean change in groove depthis 1.3 mm for tool steel and 3.4 for copper.

As mentioned earlier, the brush is much larger than the grooveand the burrs on the downmilling side are larger than on the upmilling side. Hence, the upmilling side is deburred for a longerperiod than required. The edge radius was measured by takingfour profilometer traces across the groove width (Fig. 9c and d)and the results are summarized in Table 5.

The residual burr size was measured from the zero datum,which is the level of the unbrushed region. The surface to theright of the residual burr drops below this datum because ofmaterial removal from the top surface as discussed earlier (Fig. 5).It can be seen from Fig. 10 that the radius is about 6 mm and isfairly constant. The maximum initial burr height on this edge is

emachining marks

no machining marks

top surface

top surface

p surface

200 µm200 µm

200 µm00 µm

SiC grits at 15,000 rpm. Before deburring: (a) A2 Tool steel, down milling side, (b) A2

er deburring: (e) A2 tool steel, down milling side, (f) A2 tool steel, up milling side,

Table 5Residual burr and edge radius for region between 1 and 4 mm of trace length

along the straight portion of the groove.

Material Tool steel Copper

Groove side Downmilling Upmilling Downmilling Upmilling

Radius/residual burr Residual burr Radius Radius Radius

Mean (mm) 2.2 6.1 6.1 5.8

Standard dev. (mm) 1.3 0.5 1.6 3.6

G. Mathai, S. Melkote / International Journal of Machine Tools & Manufacture 57 (2012) 46–5454

about 90 mm, which gives a predicted deburring time of 60 s. Thegroove was deburred for 132 s. Hence, in the remaining 72 s, theedge radius obtained is about 6 mm. Unlike the tool steel speci-men, both edges in the copper specimen show edge radii. Table 5indicates that the edge radius on the upmilling side is larger thanon the downmilling side because it has smaller burrs.

6. Conclusions

This paper focused on experimentally evaluating the effects ofspindle speed, grit size and type on the rate of burr heightreduction in abrasive assisted brush deburring of micromilledgrooves in tool steel and copper.

A linear regression model was fitted to the deburring rate toquantify the effect of initial burr height and process parameters. Itcan also be used to predict the time required for completelydeburring a micromilled groove based on the maximum initialburr height present. It is concluded that the deburring rate isproportional to the initial burr height with an almost constantproportionality for the conditions chosen in this study. An inferencedrawn from this is that the burrs either bend or break at about halfthe initial burr height. Deburring rate increases with spindle speedand is higher for diamond than SiC. No clear trend is evident for theinfluence of grit size. The empirically derived model was experi-mentally verified and shown to slightly underestimate the deburringtime for tool steel and slightly overestimate the time for copper.

Groove depth change was small for all conditions examined inthis study and was maximum for the largest diamond grit at thehighest speed. The edge of the groove on the upmilling sideexhibits a small radius due to the action of the abrasives. Largeburrs shield some of the area next to them from the action ofbristles resulting in a slope in the surface and a depression in theunshielded region. The study showed that brush deburring caneffectively remove large burrs in a few minutes with improve-ments in surface finish.

The regression model can be further improved by consideringType I and wear mechanisms separately and developing aregression equation for each. The process can also be improvedby concentrating the action of abrasives to the burrs and usingbrushes with impregnated abrasives to achieve better processcontrol. These improvements and further mechanistic modelingwill be the subject of future work.

References

[1] M. Ali, et al., Prediction of burr formation in fabricating MEMS componentsby micro end milling, Advanced Materials Research 74 (2009) 247–250.

[2] G. Thomas, H. Crol, New Options for Micromachining Medical Devices,Medical Device and Diagnostic Industry Magazine, 2000. Available: /http://www.mddionline.comS.

[3] G. Chern, et al., Study on burr formation in micro-machining using micro-tools fabricated by micro-EDM, Precision Engineering 31 (2007) 122–129.

[4] K. Lee, D. Dornfeld, Micro-burr formation and minimization through processcontrol, Precision Engineering 29 (2005) 246–252.

[5] O. Olvera, G. Barrow, Influence of exit angle and tool nose geometry on burrformation in face milling operations, Proceedings of the Institution ofMechanical Engineers, Part B: Journal of Engineering Manufacture 212(1998) 59–72.

[6] D.A. Dornfeld, S. Min, A review of burr formation in machining, in: Presentedat the Proceedings of the CIRP International Conference on Burrs, Kaiserslau-tern, Germany, 2009.

[7] K. Lee, D. Dornfeld, An experimental study on burr formation in micro millingaluminum and copper, Transactions of NAMRI/SME 30 (2002) 255–262.

[8] L.K. Gillespie, Deburring and Edge Finishing Handbook, SME Press, 1999.[9] S.H. Lee, D.A. Dornfeld, Precision laser deburring, Journal of Manufacturing

Science and Engineering 123 (2001) 601.[10] R. Madarkar, V.K. Jain, Investigation into magnetic abrasive micro deburring

(MAMDe), in: Proceedings of the Fifth International Conference on Precision,India, 2007, pp. 307–312.

[11] A. Kienzler, et al., Burr minimization and removal by micro milling strategiesor micropeening processes, in: Proceedings of the CIRP International Con-ference on Burrs, 2009, pp. 237–243.

[12] K.-I. Jang, et al., Deburring microparts using a magnetorheological fluid,International Journal of Machine Tools and Manufacture 53 (2012) 170–175.

[13] G.K. Mathai, S.N. Melkote, Deburring of microgrooves by abrasive brushing,in: Proceedings of the International Conference on Micromanufacturing,Madison, WI, 2010, p. 1979 (1–7).

[14] G.F. Vander Voort, Metallography: Principles and Practice, ASM International,Materials Park, Ohio, 1999.

[15] L.K. Gillespie, et al., Burrs Produced by End Milling, Bendix Corp., Kansas,1976.

[16] V. Franke, et al., Burr measurement: a round robin test comparing differentmethods, Burrs—Analysis, Control and Removal 5 (2010) 167–178.

[17] H. Fujita, et al., Magnetic-field-assisted finishing with axial vibration—

deburring on internal holes and internal finishing of pipes with rectangularcross-sections, Key Engineering Materials 329 (2007) 261–266.

[18] V. Cariapa, et al., Aspects of process model for automatic control of edge-deburring with filamentary brush, Journal of Manufacturing Science andEngineering 114 (1992) 294–300.

[19] C.F. Wu, M.S. Hamada, Experiments: Planning, Analysis and Optimization,John Wiley and Sons, 2009.

[20] T.O. Mulhearn, L.E. Samuels, The abrasion of metals: a model of the process,Wear 5 (1962) 478–498.

[21] R.L. Aghan, L.E. Samuels, Mechanisms of abrasive polishing, Wear 16 (1970)293–301.

[22] J.-Y. Lai, et al., Evolution of copper-oxide damascene structures in chemicalmechanical polishing, Journal of The Electrochemical Society 149 (2002)G31–G40.