Journal Of Materials Processing Technology 202 (2008)

Due to thevery different surface integrity induced by hard turning or grinding,the fatigue damage mechanisms could be significantly different fromother mechanical components because the f

J Kundr ´aka, B Karpuschewskib, K Gyania, V Banac, ∗

aDepartment of Production Technology, University of Miskolc, Miskolc, Hungary

bInstitute for Production Technology and Quality Assurance, Otto-von-Guericke University, Magdeburg, Germany

cAdvanced Technology Centre (ATC), Philips DAP bv., Drachten, The Netherlands

in some cases problems may arise in keeping the prescribed geometrical accuracy tigations were performed in a working environment in order to determine the attainablesize, form and positional accuracy obtained with hard turning Error sources of machiningerrors that occurred in hard turning and in grinding were taken into account, giving typicaldifferences between the two processes In the parts produced in series, size deviations weremeasured as well as out-of-roundness, cylindricity error and parallelism error of the bore’sgeneratrices The workpieces used for the investigation are disc-type parts with bores, i.e.,gears that are built into transmissions Our first measuring series evaluates the achievableaccuracy with hard turning while the second includes the comparison of grinding with hardturning The most important error sources are identified We present measures for keepingprescribed tolerances and propose methods for eliminating the means error source

Inves-© 2007 Elsevier B.V All rights reserved

The primary task of hard turning – as a finishing operation –

is to ensure the quality and reliability of the parts The quality

criteria of hard turning as a cutting operation can be found

in the technical drawings The most important quantities are

geometrical accuracy, surface topography, and the integrity of

the subsurface layer

Geometrical accuracy includes size errors, form errors and

positional errors Surface topography covers the drawing,

the roughness and the bearing curve of the surface Surface

integrity describes changes in the physical properties of the

material that result from machining (Barry and Byrne, 2002)

For analysis of the geometrical accuracy, four typical

char-acteristics of hard turning – as opposed to grinding – must be

outlined

∗Corresponding author.

E-mail address:viktoria.bana@hotmail.com(V Bana)

These are: (1) the significantly higher cutting force, (2) theomission of coolant, (3) the single point form generation, and(4) the minimum value of the depth of cut

The cutting force occurring in hard turning is higher thanthat in conventional turning or grinding The passive forceoccurring in hard turning – the component perpendicular tothe cutting speed – is a multiple of the main cutting force,while in traditional turning it is only a fraction of this value.The extraordinarily high passive force, which contributes tothe material removal, significantly loads the elements of themachining system, causing elastic deformation and deterio-rating the machining accuracy The disadvantageous effect ofthe high passive force must be compensated for by an increase

in machine tool rigidity

Hard turning can be done in dry conditions at relativelyhigh speed The relative high friction coefficient and the pas-

0924-0136/$ – see front matter © 2007 Elsevier B.V All rights reserved

doi:10.1016/j.jmatprotec.2007.09.056

sive force cause a significant friction force which transforms

into heat The other source of the generated heat is the high

cutting speed This cannot be reduced, because PcBN is only

effective at high temperatures The high temperature

gener-ated during material removal causes thermal expansion of the

workpiece, which also deteriorates the machining accuracy

The disadvantageous effect of the hot swarf falling on the

ele-ments of the machine tool must be compensated for with the

increase of “heat rigidity” of the machine (Schmidt and Dyck,

2000)

The surface generating element of hard turning is the

single-point tool tip, which shapes the surface of the

work-piece and is accompanied by significant force and heat

effects Under such conditions the single-point tool tip reacts

sensitively to any irregularities It abreacts the allowance

distinctions, the hardness differences and the other

hetero-geneities of the material with the creation of machining errors

On the other hand, grinding, where the surface generating

element is the linear generatrix of the wheel and the forces

are also lower, hardly detects any inhomogeneities in its path;

it simply eliminates them and creates a more accurate

sur-face

The fourth specification influencing the accuracy is the

depth of cut In hard turning this cannot be reduced arbitrarily,

although this is possible in grinding Moreover, grinding can

also be performed with zero depth of cut, called spark-out, to

eliminate any deflections of the system by continuously

reduc-ing the forces Because of the necessity of a minimum depth of

cut, hard turning is followed by higher forces than in grinding,

even in the finest smoothing operations

Due to the four typical characteristics of hard turning

men-tioned above, the tool tip is exposed to a substantially more

intensive physical–mechanical load than a single grain of the

grinding wheel The density of energy transmitted into the

workpiece on the tool tip is much higher Therefore, the

sta-tionary state is more difficult to maintain and the accidental

error sources arising in the material removal process are more

difficult to handle than in grinding Most unfavourable effects

can be eliminated by an increase in the robustness of the

machining system The main element of the machining

sys-tem is the machine tool, which must possess extraordinarily

high static and dynamic stiffness and must also withstand

the heating effect of the hot swarf falling upon it A

sec-ond element is the clamping device for the workpiece It has

a large influence on the geometrical accuracy in disc-type

components, which sometimes possess thin walls Extremely

rigid clamping devices and deliberate clamping forces

sys-tems must be applied The third element is the tool-clamping

device together with the tool, which offers few possibilities to

increase the rigidity The fourth element of the system is the

workpiece, which is given and bears the errors issuing from

the motions of system elements (T ¨onshoff et al., 1997)

Unintended error sources occurring in hard turning can be

divided into two classes: error sources depending on the load

and error sources independent of the load

Error sources depending on the load:

1 Cutting force, which creates elastic deformations;

2 Cutting heat, which causes distortions;

3 Tool wear, which increases the force and the heat;

4 Insufficient static rigidity of the machining system;(a) The machine tool with insufficient stiffness becomesdeformed from the force

(b) The workpiece’s clamping device of inadequate rigidity

is deformed

(c) The weak tool holder and tool are bent

(d) A workpiece with unsatisfactory stiffness is deformed

by the clamping force

5 Because of the inadequate dynamic stiffness of the ing system, low-frequency oscillations may cause formerror

machin-Error sources independent the load:

1 Uneven allowances cause force fluctuation and form errorand release stresses

2 Inhomogeneities of the workpiece material cause fromerror by the force fluctuation

3 Manner of the surface generation: in turning a point, ingrinding a line generates the surface

4 Construction of the clamping device and the deformingeffect of the clamping force may be a significant source ofform errors Instead of concentrated force it is more suitable

to clamp with distributed force

5 Number of clampings: hard turning has an advantagebecause with the single point tool and one clamping var-ious surfaces can be machined In grinding usually severalclampings are necessary, which is a source of significantform and positional errors

2 Experimental conditions

2.1 Type and main sizes of the components and accuracy prescriptions

This investigation was performed in a working environment

on gears with bores machined in batch The gears are builtinto the transmissions of motor vehicles, and are disc-typecomponents of different sizes The material of the gears iscase-hardened steel: 16MnCr5(AISI 5115) with hardness 62± 2HRC Four out of 20 measured gears are presented with themeasuring setup inFig 1in order to show the sizes and typesinvestigated

The accuracy prescription for the gears applies to the size,form and positional accuracy In many cases the size accuracy

of the bores is IT6, occasionally IT5 or IT7 For form errors,the out-of-roundness of the bore and the flatness error of thefaces are prescribed Furthermore, the parallelism of the bore’sgeneratrices and the axial run-out of the faces has to be keptwithin certain limits The specifications of form and positionalaccuracy are set by the internal standards of the factory onthe basis of functional conditions These are much stricterthan the prescriptions in the general standards For instance,

to the Hungarian standard MSZ ISO 2768-2:1991 permits of-roundness equal to half of the diameter, which is 2–3␮mlarger than the value prescribed by the factory

out-Table 1summarizes the prescriptions of accuracy for hardturned and ground surfaces for the 10 gears chosen as exam-

Fig 1 – Main types and sizes of gears A, B, C: Measuring planes of out-of-roundness.

Table 1 – Prescription of accuracy for hard turned/ground surfaces

Fig 2 – Interpretation of measuring results (a) Out-of-roundness, (b) cylindricity, (c) parallelism of the bore’s generatrix, (d) measuring setup for determination of flatness and axial run-out, (e) flatness and (f) axial run-out.

ples The root diameter/bore diameter (droot/d1) and bore

length/bore diameter (L/d1) ratios are also shown, since these

play a very significant role in the formation of the measuring

results

Although the prescriptions of accuracy according to the

internal standards of the factory do not include cylindricity,

cylindricity error was also measured and the alteration of its

value is presented In positive cases the cylindricity error

mir-rors the parallelism of the generatrices As the investigation

of parallelism is done in two-dimensional planes, the result

does not provide any reliable information about the

cylindric-ity error

The finish hard turning was performed on an advanced

machine tool suitable for the requirements of hard

turn-ing and was done in one clampturn-ing, in a hydraulic three-jaw

chuck The special jaws centralize on the pitch circle of

Fig 3 – Out-of-roundness, cylindricity and parallelism in hard turning.

Fig 4 – Generation of out-of-roundness in hard turning.

the gears and clamp with concentrated force The clamping

force is not known; only the pressure of the system, which

is 12 bar The finish grinding was done on several grinding

machines, with several clampings; a three-jaw chuck was

used

2.2 Machining conditions

The experimental conditions for machining of gear bores are:

(a)Hard turning

Machine tool: PITTLER PVSL-2 lathe

Cutting tool: PcBN CNGA 120408 BNC80

n=−6◦,˛n= 6◦,r= 95◦,

εr= 80◦, r␧= 0.8 mm(CAPTO C5-PCLNL-17090-12)Technological data: vc= 120–228 m/min,

f = 0.08–0.1 mm/rev,

ap= 0.1 mm(b)Grinding

Machine tool: SI-4/A internal grinder (VEB

BerlinerWerkzeugmachinenfabrik)Grinding wheels: CBN wheel 50× 32 × 20 9A 80

K7 V22 Bay stateCBN wheel 60× 36 × 20 9A 80K7 V22 Bay state

Fig 5 – Generation of cylindricity error in hard turning.

Fig 6 – Size accuracy of hard turned bores machined in sequence.

Technological data:

Wheel speed: vc= 25 m/s

Workpiece speed: vw= 18 m/min

Depth of cut: ae= 0.020 mm/double stroke,

roughing

ae= 0.020 mm/double stroke,finishing

Sparking out: 8–6 double strokeTraverse speed: vf,L= 2200 mm/min, roughing

(traverse speed)

vf,L= 2000 mm/min, finishingFeed rate: f = 13.7 mm/workrev, roughing

f = 12.0 mm/workrev, finishing

Coolant: half synthetic, Q = 50 l min−1

Fig 7 – Diagrams of hard turned and ground profiles.

measuring methods can be seen inFig 2 The determination

of the real geometrical forms was done by a 4 mm diameter

stylus head Out-of-roundness is either defined by the longest

distance between the fitting circle and the given points of the

real profile, or the instrument determines a so-called

refer-ence circle and out-of-roundness is determined as a value of

the largest positive local roundness deviation added to the

absolute value of the largest negative local roundness

devi-ation (Fig 2a) The last method is more frequently used and

it was applied in this study The reference circle is defined

and written out by the measuring computer on the basis of

the least square sum of the deviations Cylindricity error is

given as value of the largest positive local cylindricity

devi-ation added to the absolute value of the largest negative

local cylindricity deviation (Fig 2b) The computer generates

the real cylinder using the circumferential section method:

the instrument determines the out-of-roundness diagrams

in three or four planes perpendicular to the axis and from

them it applies a cylinder superficies built up from straight

lines This is the real cylinder form The parallelism error

of the generatrices is the absolute difference in local

diam-eter at the top and the local diamdiam-eter at the bottom of the

cylindrical future of the two associated lines fitted through

the two generatrix profiles obtained from an intersection of

a plane through the axis of the least squares reference

cylin-der and the cylindrical feature within the full extent of the

feature (Fig 2c) The flatness error of the face and its run-out

is determined during one scanning by analysing the

written-out profile in two different manners (Fig 2d) As the face can

be a plane either with or without run-out, both errors may

appear together In the measurement of flatness firstly a

ref-erence plane must be determined in the profile outstretched

into a plane The trace line of the reference plane (MMZRP)

is also defined by the least square sum method, by means of

regression analysis (Fig 2e) After this two reference planes

(OMZRP, IMZRP), parallel to the reference plane are fitted on

the real surface The distance between these two planes

deter-mines the flatness error (FLTt) In the measurement of the

axial run-out of the face the real surface is also touched

with two planes that are perpendicular to the datum axis

rather than parallel with the reference plane (Fig 2f) The

distance between these two planes defines the axial run-out

(R).

The diameter of the bores was measured by the Derby

Etalon 454 coordinate measuring machine The applied

sty-lus was the Renishaw Brown Shape TP-ES This measuring

machine measures the diameter of one circle with three

touches We repeated it on four circles and their average value

formed the real value for the diameter

up to 500, which are set regularly on the circumference pared with Gauss-type filters of 1–50 u/r or 1–15 u/r, this filtercan give a more-detailed idea of the form of the out-of-roundness It is allowed to compare the gained diagrams only

Com-if they were measured with the same filter

4 Out-of-roundness, cylindricity error and parallelism error in hard turning

The geometrical accuracy of hard turning was investigated inseries production One series each of three gears was mea-sured; the series contained 285, 60, and 200 pieces The gears

differ in droot/d1ratio Out-of-roundness, cylindricity and allelism measurements were performed at regular intervals:after each 25th piece for the 285-piece series, each 5th for the60-piece series, and each 10th for the 200-piece series Averagevalues of these measurements are given inFig 3 The measur-ing results truly reflect the geometrical errors issuing from thedeformation of the workpiece The “rigidity” of the workpiece,

par-the weakest element of par-the system, is defined by par-the droot/d1 ratio With the decrease of the droot/d1ratio, out-of-roundnessand cylindricity error increase The form errors increase notonly in absolute meaning but also its scattering significantlyrises However, the parallelism error of the generatrices hardlychanges, nor does its scatter; each measured value remainsmuch lower than the prescribed 0.006 mm

Out-of-roundness and cylindricity error are presented inFigs 4 and 5, where every measuring result is shown For the

gear with droot/d1= 2.83 ratio, which can be regarded as quiterigid, the out-of-roundness of 0.006 mm prescribed for qual-ity IT6 can be kept, but this cannot be guaranteed for the

Fig 8 – Summary of measured results.

gear of droot/d1= 1.34 ratio The reason for this is obviously

the clamping chuck, because the clamping method and force

are not suitable It is mechanical clamping with high

pres-sure instead of magnetic clamping On the newly developed

lathes, in addition besides 12 bar, two lower pressures are also

set with 8 and 4 bar As can be seen inFig 5, the generation

of the cylindricity errors also proves that the wall thickness

(droot/d1) must be taken into account in the selection of the

clamping force Although there are no prescriptions for dricity in the drawings, their magnitude must be controlled

cylin-Fig 9 – Diagrams of hard-turned and ground profiles.

sented inFig 6.The magnitude of scattering does not exceed

the whole tolerance range for the bores with accuracy IT6

This means that on the lathe the accuracy IT5 can be ensured

easily Moreover, if the possibility of size correction is used

maximally, it is possible to meet more restrictive tolerances

as well, with proper rigidity and good maintenance of the

machine tool Furthermore, nowadays machine tool factories

produce lathes not only for disc-type parts but also for

exter-nal, cylindrical, conical and shaped surfaces Their guaranteed

machining accuracy is IT5 with out-of-roundness of 1␮m and

a cylindricity deviation of 2␮m

6 Formation of accuracy in comparison

with grinding

Many factories have already changed to hard turning and,

therefore, in some cases grinding is only applied as a

nec-essary solution However, we were able to measure gears

produced with both hard turning and grinding.Fig 7

intro-duces the main data and measuring diagrams of the relatively

large The out-of-roundness diagrams apply to the plane

noted as C because, according to our experience, the

high-est out-of-roundness always appears here (Kundrak and Bana,

2003), due to the deforming effect of the three-jaw chuck

The characteristic out-of-roundness is a three-lobe form for

both processes Cylindricity error, like out-of-roundness, is

higher for hard turning The reason for this is the higher

clamping force and form generation with a single-point tool,

whereas in grinding the bore is conical At the point where

the grinding tool is first applied the diameter is larger, since

on this side the wheel is more overtravelled than on the other

side

The parallelism is better in hard turning but on the entering

side the diameter is smaller due to insufficient conduction of

the intensive heat The parallelism error of grinding is higher;

the expansion on the entering side can be explained by the

running out of the end stroke of the wheel to a larger extent

The final results of measurements, with the addition of

flatness and the axial run-out of face, are presented inFig 8

Although the out-of-roundness is higher in hard turning, it

still satisfies the quality prescribed for IT5 The parallelism

of the generatrices is also suitable The cylindricity tolerance

is not prescribed, therefore, its generation is less important

The flatness is adequate for hard turning, but for grinding it is

higher than the permitted limit The axial run-out is adequate

in both hard turning and grinding but much higher in grinding,

where it is about five times of the hard turned values These

results show that the ground gear does not fulfil the

require-hard turning but in grinding the bore diameter is higher on theentering side The reason for the conicity is that wheel over-travel occurs to a higher degree here than on the chuck-handside Here the parallelism of the generatrices for hard turning

is also better than in grinding

In hard turning, at entering a lower accumulation of heatcan be observed, which increases at outgoing In addition, onthe hard turned piece a softer mark was found on the scannedside that was abreacted by the single point tool by more mate-rial removal On the bore of the ground gear the result ofunequal overtravel of the wheel also appears

The results of all performed measurements are rized inFig 9 The out-of-roundness measured in planes A, B,

summa-and C is notable It is always the highest in plane C because

here the deforming effect of the clamping force is the highest.Moreover, for this gear neither hard turning nor grinding meetout-of-roundness limits, especially in the case of hard turn-ing The prescription for cylindricity is not given, therefore itsshaping is neutral The parallelism of the bore’s generatrices

is definitely acceptable

ComparingFigs 8 and 10, it is prominent that the flatnesserror and the axial run-out of the ground part face are 4–5times higher than those of the hard turned components Thisarises because of the grinding process applied The face has to

be machined on the internal grinding machine However, facegrinding can be regarded as a necessary procedure on thesemachine tools The armed apparatus that clamps the grindingspindle can be brought into work position outside As a con-sequence of this we must work with a light apparatus withlow rigidity, which leads to a higher degree of deformationfrom the grinding force Moreover, the working of the appa-

Fig 10 – Summary of measured results.

ratus that performs the dressing of the grinding wheel is not

always precise, because sometimes the dressing of the small

cup wheel is done by hand Because of the effort made for

rel-atively small sizes, the grinding quill is also weak Therefore

it is no wonder that the face plane is not precise

As for the axial run-out, the surface grinding of the face

is performed on a separate machine in a different clamping.While the clamping was done on the pitch circle in internalgrinding, for face grinding the clamping is on the oppositeplane surface The alteration of the clamping surface causes

Fig 11 – Diagrams of hard-turned and ground profiles.

Fig 12 – Summary of measured results.

the significant flatness error and axial run-out In hard

turn-ing such problems do not arise because the whole machinturn-ing

is done in one clamping, and therefore, the flatness error and

run-out are significantly lower The out-of-roundness of gear

E belonging to the bore quality IT5, whose magnitude exceeds

the prescribed values, can be explained by the inadequacy of

the clamping device For such a small wall thickness clamping

with concentrated force is not acceptable A clamping device

of some other construction is necessary; for this purpose

mag-netic and hydraulic clamping chucks are the most suitable

options

Finally we investigated one more hard turned and ground

pair, which agrees with the previous in everything but the bore

diameter’s tolerance, which is not IT5 but IT6 There are no

prescriptions for the flatness and the axial run-out of the face

The main data of the gear and the diagrams gained during the

measurement can be seen inFig 11

The summary review of the measuring errors can be seen

inFig 12 In the bore with quality IT6 the prescribed

out-of-roundness cannot be met with hard turning but can be with

grinding This is because the higher concentrated force

neces-sary for hard turning deforms the workpiece with small wall

thickness to such a degree that after release it possesses

out-of-roundness higher than that permitted In grinding, because

of the lower clamping force, lower form error will occur The

parallelism of the generatrices can be ensured for both

pro-highly concentrated clamping force from the clamping anism with of 12 bar pressure (uncontrollable) This problemcould probably be avoided if, instead of concentrated clampingforce, a magnetic or vacuum chuck working with distributedforce were applied In this case, the typical three-lobe form

mech-of the bores would disappear The other prescriptions for theform and positional accuracy can be ensured by hard turn-ing at IT5 level In grinding, the flatness and axial run-out ofthe faces are critical, and this is about 4–5 times higher than

in hard turning With the given equipment and technology itcan be hardly reduced However, conicity on the entering sidecan be avoided with the overtravel of the wheel occurring inlower degree

Kundrak, J., Bana, V., 2003 Geometrical accuracy of machining ofhardened bore holes In: Fourth Workshop on EuropeanScientific and Industrial Collaboration, May 28–30, University

The impact of surface integrity by hard turning vs grinding on fatigue damage mechanisms in rolling contact

Y.B Guo ⁎ , A.W Warren

Department of Mechanical Engineering, the University of Alabama, Tuscaloosa, AL 35487, United States

© 2008 Elsevier B.V All rights reserved

1 Background on rolling contact fatigue damage

1.1 Introduction

Since most mechanical components produced by hard turning and

grinding are widely used in rolling contact applications, the fundamental

knowledge of fatigue damage mechanism is necessary for understanding

manufacturing process effects Most research to investigate rolling contact

fatigue (RCF) damage were performed by creating artificial surface defects

such as cracks or depressions using Vickers or Knoop indentations The

purpose of doing this is to accelerate crack propagation and fatigue,

however, creating such artificial defects is not realistic and the results do

not necessarily compare with real-life scenarios The artificial defects will

change the surface integrity and therefore alter the nature or mechanism

of fatigue damage Thus, the fatigue damage in the presence of artificial

defects in literature may not reflect the true nature of real-life fatigue

process Therefore, fatigue tests performed without the presence of

artificial defects (excluding machining profiles resulting from turning and

grinding or minor scratches due to polishing) are necessary and more

representative of real-life applications

The majority of fatigue mechanisms investigated in other research

is based on the accelerated tests of various materials such as metals and

ceramics[1–3] Currently, very little research has been performed tostudy fatigue mechanisms of hard machined components Due to thevery different surface integrity induced by hard turning or grinding,the fatigue damage mechanisms could be significantly different fromother mechanical components because the fatigue life is very differentfor same surfacefinish[4] Though a few studies have examined thegeneral surface integrity factors such as roughness, residual stress, etc

on RCF, the basic limitations is that the conclusions are too general toexplain the specific difference of fatigue performance caused by hardturning and grinding[4,5]

The objective of the present RCF tests is to study the fatigue damagemechanisms for crack propagation in real-life tests of samples prepared

by hard turning and grinding Four types of test samples were prepared:

as turned (AT), turned and polished (TP), as ground (AG), and ground andpolished (GP) Each RCF test was ran until significant fatigue damage (i.e.surface spalling) had occurred at which time it was stopped The testsamples were then analyzed in terms of surface and subsurface fatiguedamage considering the effect of process induced surface integrity Afatigue damage mechanism has been proposed based on the experi-mental and simulation results

1.2 Fatigue crack appearance

Three possible mechanisms for surface fatigue crack growth havebeen proposed by Bower[6]: shear (Mode II or III),fluid pressurization

⁎ Corresponding author Tel.: +1 205 348 2615; fax: +1 205 348 6419.

E-mail address: yguo@eng.ua.edu (Y.B Guo).

0257-8972/$ – see front matter © 2008 Elsevier B.V All rights reserved.

Contents lists available atScienceDirect

Surface & Coatings Technology

j o u r n a l h o m e p a g e : w w w e l s ev i e r c o m / l o c a t e / s u r fc o a t

based “wedge effect”[3] (Mode I) The shear model assumes that

cracks are initiated by the Hertzian contact stress field occurring

around the circumference of the contact area For this model, thefluid

does not have an effect on crack growth The wedge effect model is

based on crack growth by fluid pressurization generated by the

entrainment offluid (i.e lubricant) into the high pressure contact

flowed into the pre-existing crack or surface defect and caused it to

open (Mode I) orfluid entrapment in which closure of the mouth

sealedfluid into the crack and contact motion generates fluid

pres-sure at the crack tip (Mode II) Mode II propagation can be severely

hindered in practice due to friction or interlocking of the crack faces

and the presence of afluid assists crack growth by reducing friction,

separating the cracks during compressive loading, or by creating

internal pressure This has been shown by Way[7]who found that RCF

pits only occurred when a lubricant was present

In rolling contact fatigue (RCF)[1–22], the initial cracks are nearly

always inclined to the surface at a shallow angle (≈20°) to the surface

For surface originated cracks, most research shows that the orientation

of the crack is highly dependent on the direction of the tractive force It

is usually observed that the crack grows in a direction opposite the

direction of the tractive force For subsurface originating cracks, it has

been suggested that the direction of crack growth is dependent on the

direction of rolling[8] Explanations for the direction of crack

pro-pagation have been proposed that consider residual stresses[8,9]

resulting from the machining processes Residual stresses could open

the crack in the observed orientation while others[9,10]speculate that

the cracks evolve by ductile fracture and that accumulating plastic

deformations control the crack directions The prediction of a preferred

crack orientation is made more difficult because the topography of the

contacting surfaces is constantly evolving The changes in surface

topography lead to dramatic differences in the contact pressure and

therefore the contact stresses driving crack propagation Even small

changes in surface topography lead to a substantial difference in

pres-sure and the presence of surface spalls or pitting can cause the prespres-sure

to greatly exceed the Hertzian pressure[8]

1.3 Fatigue mechanisms on crack initiation and propagation

1.3.1 Surface roughness effect

Surface roughness is critical to fatigue endurance of components

subjected to RCF The contact stress between two bodies in contact is

significantly affected by the surface topography of the contact zones

High surface roughness can cause local contact stresses to be

dra-matically higher than predictions based on the Hertz theory orfinite

element analysis and results in more rapid failure of components

Surface roughness profiles can also simulate pre-existing surface

flaws such as cracks The peak and valley profile of turned and ground

surfaces exhibit such a phenomenon and act as notches that

ex-perience locally high stress concentrations and are the preferred

locations for crack initiation During RCF, lubricant can be trapped in

the valleys of one contact surface which is then pressurized by theother contact surface leading to hydrostatic pressure and crack growth.The effect of surface roughness on fatigue damage is not fully under-stood, however, previous research[11]has shown that increasing thesurface roughness above a critical level leads to greater influence ofmicro-geometrical features on fatigue strength

1.3.2 Microstructure effectThe capability of a material to resist fatigue damage during rollingcontact is directly related to its microstructure Brittle materials such asceramics and cast irons usually fail from normal loads, while ductilematerials such as steels fail due to shear stresses The reason formaterials being either ductile or brittle is directly controlled by themicrostructure and atomic arrangement of the crystal lattice Previousresearch[12]has shown that the presence of a“white layer” induced

by phase transformation on hardened steels produced by abusivehard turning is detrimental to RCF life for high load applications Awhite layer is formed when cutting temperatures above the austeniz-ing temperatures are achieved and the martensite is then rapidlyquenched The resulting temperature effects leave an untemperedmartensite (UTM) region that is referred to as the white layer Duringrolling contact, the maximum normal stresses are located at thesurface, while maximum shear stresses occur at a certain depth in thesubsurface since friction coefficient is much less than 0.3 for thewell lubricated contacting surfaces The machined surface of the testsamples could be considered as brittle since the hardness of the heattreated AISI 52100 steel is 62 HRC In addition, strain hardening whichoccurs during hard turning and grinding increase the hardness in thenear surface by as much as 20% This is important because it meansthe location of crack initiation is dependent on the loading and micro-structure of the component For steels, the origin of cracks is usually atsurface asperities, subsurface inclusions, or cementite–ferrite bound-aries[13] It is important to note that the initial damage does notalways occur at the location of maximum shear stress, but can occur atany location where the shear stress is greater than the maximum shearstrength of the material

1.3.3 Residual stress effectResidual stresses can have a significant effect on the development

of fatigue damage [14–20] Residual stresses are caused by latticedistortions resulting from the machining process These lattice dis-tortions serve to increase the strength of the material by hinderingdislocation motion during loading Compressive residual stress will

Fig 1 RCF test setup.

Table 1 RCF fatigue test conditions Load (N) Peak Hertzian stress (MPa) Radius of contact circle (µm) Frequency (Hz)

impede crack growth by closing the crack tip, while tensile residual

stress aids crack growth by opening the crack mouth A mechanical

model[23]has been developed to quantify the effects of the residual

stress and hardness gradient by laser transformation hardening on

crack driving force During rolling contact applications, the

mag-nitude and distribution of residual stresses is constantly evolving

Voskamp[17]has shown that residual stresses develop in high carbon

steels (1% C) at sufficiently high loads and these stresses may influence

the direction of crack propagation Guo et al.[18,19]has shown that

the distinct residual stress profiles by hard turning and grinding only

affect near-surface fatigue damage rather than locations deeper in the

subsurface The residual stresses affect neither the magnitudes nor the

locations of peak stresses and strains below the surface at high load

applications The slope and depth of a compressive residual stress

profile are key factors for rolling contact fatigue damage Equivalent

plastic strain could be a parameter to characterize the relative fatigue

damage

2 Experimental procedures

Four types of test surfaces of as turned (AT), turned and polished

(TP), as ground (AG), and ground and polished (GP) were prepared

The AT and AG surfaces have equivalent surface roughness Ra 0.16 µm

and the TP and GP surfaces have Ra 0.06 µm[24] The experimental

RCF setup shown inFig 1was used to monitor RCF tests of ground

surfaces This test rig is capable of rotating up to 4000 rpm, thus

allowing the test to run in a reasonable time frame The load is applied

through the rotating shaft which also houses the slave washer The

test utilized eight chrome steel balls with the diameter of 5.56 mm

and hardness of 63 HRc A retainer was used to hold the balls and was

constructed of nylon to minimize AE noises during the test Slave

washers of the same material as the test samples were created byturning An 8 mm radius groove was machined on each slave washer

to maintain the ball's position during operation and provide a lowercontact pressure between the ball and slave washer The lower contactpressure in the slave washer would help ensure that fatigue failurewouldfirst occur in the test sample, outside of any other variables Thetesting conditions are shown inTable 1

In-situ monitoring of fatigue damage was accomplished using acomplete AE acquisition package that is versatile in the application offatigue damage analysis An AE sensor was attached to the test samplewith a holdingfixture, while vacuum grease served as the couplingmedia between the sensor and test sample The AE sensor has a

125 kHz resonant frequency and connects to an 18 bit PCI-2 dataacquisition board that was incorporated into a PC Before the datareached the PC it was passed through a preamplifier that was set at a

40 dB gain and a threshold of 45 dB was used The RMS/ASL timeconstant for all of the tests was 500 ms with a sampling rate of 10 Hz.The sampled AE parameters in this study include: absolute energy,RMS, amplitude, counts, and average frequency

The test sample is secured in a resting plate which is positionedwith four rods that maintain the rigs position and rigidity The restingplate is free to move up and down as it rests on a load cell, allowingaccurate measure of the applied load An acoustic emission sensor ismounted directly to the test sample Prior to startup, a high-temperaturemulti-purpose lithium complex grease was evenly distributed acrossthe test surface, slave washer track, and balls Additional lubricationwas added at regular intervals throughout the duration of the test.Rotational speed of the spindle and washer was monitored using anoptical tachometer

Parallelism and centricity between the slave washer and testsurface is critical to ensure that the pressure distribution is uniform

Fig 2 Wear track and pitting on the ground and polished sample.

across the test surface For this reason, a dial indicator gage was used

to measure the vertical and horizontal runout of the washer and test

surface Using this method it was possible to adjust the horizontal and

vertical displacements so that parallelism and centricity would be

achieved

3 Experimental results and observations

The fatigued samples were analyzed in terms of surface and

subsurface damage Surface damage reveals itself as pitting and/or

spalling and is formed from an accumulation of cracks that eventuallylead to fracture and material loss This type of damage can be observed

on the test surface by optical microscopy immediately after the RCFtests are completed Subsurface damage is only visible on the cross-section of the test samples and requires cutting and polishing in order

to view the crack characteristics such as depth and orientation.The significant challenge for this work is to determine the me-chanisms for crack growth during rolling contact by analyzing the postfatigue damage It is impractical to conduct an experiment of thisnature and document the progression of fatigue damage for every test

Fig 4 a Wear track profile of the turned and polished sample b Wear track profile of the as turned sample c Wear track profile of the ground and polished sample d Wear track profile of the as ground sample.

sample In order to view the subsurface cracks, the test specimen must

be destructively cross-section which renders the sample useless for

further loading by rolling contact The technique used in this study

involved subjecting the test specimens to rolling contact until spalling

occurred on the surface At this time the specimens were removed and

prepared for investigation by cross-sectioning The fatigue damage

(surface and subsurface cracks) were then observed and characterized

in terms of location, length, and orientation Comments and

specula-tions were then made regarding fatigue progression based on the

observed fatigue damage Conclusive or definite statements regarding

the nature of fatigue propagation could not be made due to the lack of

intermediate stages of fatigue damage

3.1 Surface spalling

Fig 2shows a section of the wear track created on the surface of a

GP sample For brevity, surface damage of each sample is not shown,

but the damage shown is representative of all test samples Surface

pits and wear debris are scattered throughout the wear track and

become more severe near the spalled region The width of the wear

track varies from 500 to 750 µm, the larger widths occurring in areas

with the most severe pitting The wear debris is created from material

loss of the ball and test sample asperities (mostly during the run-inperiod), and may be responsible for a large portion of the surface pits

by creating dents on the surface as the balls travel across them.Fig 3

shows an optical image of spalling that was formed on the surface of a

GP test sample Clearly seen is the substantial loss of material due tothe accumulation of damage that resulted in the observed spalling Onaverage, the widths of the surface spalls are between 700 and 850 µmdepending on how long the fatigue damage was allowed to accu-mulate Typically, tests are allowed to run for some time after fatiguedamage has been indicated to assure that there is sufficient damage toinvestigate Because the tests run for so long (2–5 weeks depending onthe specific test samples), the post fatigue duration times are notidentical, but are within 1–2 h For this reason, the width and depth ofthe surface spalls can vary to some extent between identical testingconditions Radial cracks are clearly seen near the center of the spallsand are shaped in the form of semicircles that are convex in thedirection of rolling and have an approximate radius of 135–180 µm.This value is similar to the value of the contact circle radius calculated

Fig 5 a Surface spall profile of the turned and polished sample b Surface spall profile of the as ground sample.

by Hertz theory as shown in[24] The depth of the spalls ranges from

40 to 60 µm, and is also dependent on the post-failure run time

An optical image of ball fatigue has been shown in a previous study

[24] Premature fatigue failure of the balls is due to the kinematics and

loading of the RCF test Although the balls are the same material and

hardness as the test surface and washer, the load experienced by them

is quite different While the sample surface and washer experience

cyclic loading at a frequency of≈173 Hz, the rollers are under constant

load and the interaction of surface topographies of the contactingsurfaces result in cyclic loading frequencies that are much higher thanthat of the test surface For this reason, the balls are predisposed to failmore rapidly than the other components of the test assembly.Line traces were made using a Taylor Hobson Talysurf 2000 surfaceprofiling machine to determine the width, depth, and profile of thewear tracks and spalled regions on the test samples.Fig 4showsseveral examples of the wear track profiles Typically, the wear trackshave a depth of 2 to 4 µm and range in diameter from 500 to 750 µm.There is some pile-up of material along the edges of the track that runparallel to the rolling direction due to plastic deformation (similar to

an indentation) accumulated during the test Surface spalls are createdwhen significant material loss has occurred.Fig 5shows examples ofspall profiles These images show the much larger depth (≈60 µm) ofthe spalled region when compared to the wear track

3.2 Subsurface cracksSubsurface cracks were investigated by cross-sectioning the testsamples tangential to the wear track as shown inFig 6in a locationwhere significant fatigue had occurred, mounting in epoxy, andpolishing to a mirrorfinish using alumina polishing compound Thesamples were also slightly etched using a 2% nital solution to view thecracks and microstructure in the SEM The subsurface damage shown

is representative of each testing condition and several observationscan be made

Figs 7–10show the representative subsurface fatigue damage ofthe GP, AG, TP, and AT samples, respectively Upon observing thecollective data, it is identified that there are two distinct crack types

Fig 8 Subsurface fatigue damage of the as ground sample.

Fig 9 Subsurface fatigue damage of the turned and polished sample.

Fig 11 Finite element simulated of shear stress S23 contour.

Thefirst is a significant crack that runs parallel to the surface at a

depth of 5.3 to 13.3 µm for nearly the entire length of the specimen

(Fig 7) This crack is referred to in this paper as the“main” crack The

other types of cracks are inclined cracks that extend from the surface

to a depth as much as 45 µm These cracks could extend from the

surface due to high stress concentration induced by sharp surface

asperities for ground surfaces and may be independent of the main

crack Smaller branching cracks are also observed connecting the

surface to the main crack and it appears as though sufficient branching

cracks in any region causes eventual fracture and loss of material The

occurrence of a main crack and surface cracks together is not observed

in all test specimens, but at least one or both appear in all fatigued

samples The mechanisms which form the distinct crack types are not

fully understood, and the purpose of this paper is to explain some

possible causes for the presence of these cracks Due to the location of

the main crack (5.3 to 13.3 µm below the surface), it is possible that it

is formed due to the magnitude of shear stress experienced by the

sample at this depth which could cause fracture or debonding of

subsurface inclusions or second phase particles located at this depth

The variation in main crack depth may also be influenced by the

presence (or formation) of residual stresses that would influence the

magnitude/distribution of shear stress experienced by the test

samples Also, surface topography will greatly influence the

distribu-tion of local stresses, and may affect the depth at which the shear

stress is large enough to initiate failure

Afinite element simulation using Abaqus [25] was conducted to

estimate the magnitude and distribution of the stresses resulting from

rolling contact A 3-dimensional axisymmetric mesh was generated to

simulate the experimental conditions A rigid roller (d=5.56 mm) was

used to apply the vertical load of 305 N and was given linear and angular

velocities of 3.14 m/s and 1129 rad/s, respectively, which correspond to the

experimental conditions Shear strength of AISI 52100 steel is

approxi-mately 924 MPa[5]and results of thefinite element analysis inFig 11

show that the depth at which this value occurs is≈15 µm as shown in

Fig.12 The reason that the observed depth is less than 15 µm is most likely

due to loss of material on the top surface that occurs during the RCF test or

because of stress variations caused by surface asperities It is important to

note that the FE results assume an idealflat surface (which is impractical),and future work will incorporate surface roughness As the main crackincreases in length, the region above the main crack may behave as adelaminated surface layer This will significantly affect the stressdistribution below the main crack because the structural material is nolonger continuous The delaminated region then becomes subjected tomuch higher stress and this may be the cause for formation of thebranching cracks When a sufficient number of branching cracks areformed connecting the main crack with the surface, small portions ofsurface material eventually are removed by the repeated loading Themechanism that causes the formation of surface cracks is quite different.These cracks appear to form at the surface and travel downward at ashallow angle to the surface that ranges from 24° to 36° and in a directionthat is usually the same as the direction of rolling The mechanism forgrowth is most likely due to the wedge effect As the ball approaches, itforces lubricant into the crack which then propagates due to high pressuregenerated as the ball travels across the crack face

From observation of the fatigue damage, it is apparent that theformation of the main crack and surface cracks are parallel processes.However as seen inFigs 7 and 8, the surface crack must be presentfirst in order to propagate deeper than the depth of the main crack.This is because the main crack causes the material to be discontinuousand surface cracks that form after the main crack forms will arrestafter intersecting it However, new cracks below the main crack may

Fig 13 Orientation of cracks represented by crack analysis tables.

Table 2

Crack orientation analysis for case A

Sample type θ (deg) d 1 (µm) Average θ (deg) Average d 1 (µm)

Turned and polished 43 28 36.71 20.50

42.5 23

44 24 18.5 8

type

θ (deg)

d 1

(µm)

d 2

(µm) Average θ (deg)

Average d 1

(µm)

Average d 2

(µm) Turned and

polished

15 15 7.5 35.07 12.61 6.07 36.5 10.5 7

37 7 2.5

35 5.5 1 30.5 12 4

18 23 Ground and polished 39 12.31 32.50 11.28

26 10.25

be initiated by repeated point loading caused by increasedflexibility

of the delaminated near-surface layer The mechanism for the main

crack is believed to be shear stress However, the mechanism for

surface cracks is a combination of shear stress (from the compressive

loading of the balls) and tensile stress induced by the presence of

pressurized lubricant that is trapped within the crack as the ball

contact patch travels across the crack mouth As shown in thefinite

element simulation in Fig 11, the shear stress contour on the top

surface is similar to that of the surface ring cracks observed inFig 3

The formation of surface cracks may be accelerated by tensile stress

concentrations at surface asperities or surface topography profiles

that act as notches It is believed that the presence of surface cracks

has little effect on the formation of the main crack

3.3 Depth and orientation of surface initiated cracks

The surface cracks for each test sample were analyzed according

to maximum depth and orientation with the top surface Four

possi-Fatigue is an accumulation of damage sustained from cyclic stress

In rolling contact applications, the mechanism of fatigue damage of acomponent is dependent on several factors including frequency andamplitude of loading, microstructure, presence of initial defects,mechanical properties, and residual stress For ductile materials such

as steel, fatigue fracture is caused by shear stress The maximum shearstress of a component subjected to rolling contact is located in thesubsurface, the depth of which can be estimated by the Hertz theory

or precisely by afinite element analysis

RCF is a cyclic loading process that is very complex and composed

of many stages For the current tests, the general progression of fatiguedamage can be outlined as follows:

(1) Initially, the crack free surface is subjected to rolling contact underspecified loading conditions Immediately after start-up, thesurface experiences elastic and plastic deformation as the ballstravel around the track The duration of the run-in period is specific

to each test and varies from 4–18 million cycles depending onseveral factors including surface roughness, hardness, and load

(2) Contact between the surface and balls eventually removes most

of the surface asperities of the test sample and eventually a

wear track is formed that becomes the preferred path of the

balls as they travel across the test surface The wear particles

that are removed from the surface roughness eventually are

pushed outside of the contact zone, or are displaced onto the

wear track If a ball travels across a wear particle, it can lead to a

surface depression and eventual ball or surface failure

(3) The maximum shear stress experienced by the test sample is

located in the subsurface as seen inFigs 11 and 12, the depth of

which is dependent on the load, contact area, and material

properties For materials such as AISI 52100 steel with subsurface

inclusions, voids, and interfaces with second phase particles,

fracture and/or delamination can occur at any depth for which

the shear stress is greater than the material shear strength The

maximum shear stress in the subsurface may be magnified by the

presence of second phase particles and/or inclusions This may

promote the main crack formation in the subsurface as shown in

Figs 7–10, especially the as-turned surface inFig 10 Because the

depth of the maximum shear stress is constant (for aflat

speci-men) below the surface of the wear track, a main fracture is often

observed at this depth and can run parallel to the surface

throughout the entire length of the wear track

(4) The mechanism for inclined or branching cracks nucleation

may be tensile stress located in the vicinity of surface asperities

while propagation occurs due to trapped lubricant that opens

the crack mouth as a result of hydrostatic pressure caused by

the ball as it travels across the crack mouth The shear stress S23

drives a surface crack propagating downward with an inclined

angle along the rolling direction

(5) The interaction between an inclined crack and the main crack

usually breaks the inclined crack The branching cracks from

the main crack may either propagate to or from the top surface

When a sufficient number of branching cracks reach the

surface, eventual material loss and rapid spalling occurs

The formation sequence of the main crack and inclined crack may

be different under the influence of distinct surface integrity by turning

vs grinding The examination on the SEM images of cracks suggests

two different fatigue damage mechanisms,Fig 14, for the turned and

ground surfaces The basic difference is that an initial or main crack

form in the subsurface for turned samples as shown inFig 10, while an

initial branching crack could start from ground surfaces and joins with

the subsurface main crack

5 Conclusions

Fatigue damage of four types of test samples in real-life RCF tests

was analyzed A fatigue damage mechanism was proposed based on

the experimental and simulation results The major results can be

summarized as follows:

• Wear tracks created by plastic deformation and material loss from the

rolling process were visible along the circumference of the test sample

surface Spalling occurs on the sample surface in regions of the wear

track where accumulated fatigue damage results in significant material

loss

• Subsurface damage was characterized by the formation of two distinctcrack formations: the Mode I main cracks that extend parallel to thesurface and surface cracks that are inclined at a shallow angle to thesurface The locations of main cracks are approximately equivalent tothe depth predicted by a FEA analysis for which the experienced shearstress was greater than the material shear strength The main crack isformed as a result of fracture or delamination of subsurface inclusions

or second phase particle interfaces located at this depth Possibleexplanations for the depth variation in the main crack location wereattributed to differences in residual stress, surface topography, andstatistical variation between the samples created by hard turning orgrinding

• Surface cracks extend from the surface at shallow angles ≈35° Thedirection of surface crack propagation is somewhat random, but themajority propagates in the direction of rolling The mechanism forsurface crack nucleation is tensile stress located at the vicinity ofsurface asperities while propagation occurs due to trapped lubricantthat opens the crack mouth as a result of hydrostatic pressure caused

by the ball as it travels across the crack mouth

• The formation sequence of fatigue cracks is different for the turnedand ground surfaces The basic difference is that an initial main crackform in the subsurface for turned samples, while an initial branchingcrack could start from ground surfaces and joins with the subsurfacemain crack

Acknowledgement

This research is based upon the work supported by the NationalScience Foundation under Grant No DMI-0447452

References

[1] R Dommarco, P Bastias, C Rubin, G Hahn, Wear 260 (2006) 1317.

[2] Y Wang, M Hadfield, Wear 225–229 (1999) 1284.

[3] K Kida, M Saito, K Kitamura, Fatigue Fract Engng Mater Struct 28 (2005) 1087 [4] F Hashimoto, A.W Warren, Y.B Guo, Ann CIRP 55 (2006) 81.

[5] D Schwach, Y.B Guo, Int J Fatigue 28 (2006) 1828.

[6] A Bower, Trans Am Soc Mech Engrs., J Trib 110 (1988) 704.

[7] S Way, J Appl Mech 57 (1935) A49.

[8] A Oliver, Proc IMechE 219 (2005) 313.

[9] J Ringsberg, M Loo-Morey, B Josefson, A Kapoor, J Benyon, Int J Fatigue 22 (2000) 205.

[10] F Franklin, I Widiyarta, A Kapoor, Wear 251 (2001) 949.

[11] E Siebel, M Gaier, Engrs Digest 18 (1957) 109.

[12] Y.B Guo, D Schwach, Int J of Fatigue 27 (2005) 1051.

[13] P Fernandez, Eng Failure Analysis 4 (1997) 155.

[14] Y Matsumoto, F Hashimoto, G Lahoti, Ann CIRP 48 (1999) 59.

[15] R Scott, R Kepple, M Miller, in: J.B Bidwell (Ed.), Rolling Contact Phenomena, Elsevier, Amsterdam, 1962, p 301.

[16] D Townsend, E Zaretsky, SAE Technical Paper Series 881291, 1988.

[17] A Voskamp, Trans Am.Soc Mech Engrs., J Tribology 107 (1985) 356.

[18] Y.B Guo, M.E Barkey, Int J Mech Sci 46 (2004) 371.

[19] Y.B Guo, M.E Barkey, Int J Fatigue 26 (2004) 605.

[20] A Oliver, H Spikes, A Bower, K Johnson, Wear 107 (1986) 151.

[21] N Govindarajan, R Gnanamoorthy, Mater Sci and Eng.: A 445–446 (2007) 259 [22] N Govindarajan, R Gnanamoorthy, Wear 262 (2007) 70.

[23] B.Q Yang, K Zhang, G.N Chen, G.X Luo, J.H Xiao, Surf Coat Technol 201 (2006) 2208.

[24] A.W Warren, Y.B Guo, Fatigue Fract Eng Mater Struc 30 (2007) 1.

[25] Abaqus, Inc., ABAQUS User's Manual, Ver 6.4, Pawtucket, RI, , 2003.

3-of a turned surface which was in sharp contrast with the random and isotropic nature 3-of aground surface In general, a gentle turned surface has higher values of amplitude param-eters (arithmetic mean, root mean square, maximum height of summits, maximum depth

of valleys, and 10-point height) than an abusively turned surface, whereas the opposite wastrue for the ground counterparts Only the gentle ground surface has a negative skewnesswhich means that the topography distribution is more biased towards the valley side Thelarger kurtosis value of the abusively ground surface implies a more peaked surface topogra-phy The gentle ground and abusively turned surfaces have a much larger bearing area ratioand therefore better bearing capacity than the gentle turned and abusively ground ones.The abusively ground surface has higher fluid retainability than other surfaces in terms ofmean void volume However, surface performance, such as wear and fatigue is dependentboth on surface topography as well as mechanical property and relies on the dominance ofthe individual aspect

© 2007 Elsevier B.V All rights reserved

Hard turning and grinding are competitive finishing processes

which produce distinct surface topography features due to

the inherent difference in the material removal process The

single point cutting tool with the defined geometry in

turn-ing will naturally produce a more anisotropic surface while

the multiple small abrasives of random geometry in grinding

will produce a more isotropic surface (Malkin, 1989; Marinescu

et al., 2004) The anisotropic turned surface is characterized

by a symmetric and periodic variation of peaks and

val-leys, whereas the isotropic ground surface shows a more

∗Corresponding author Tel.: +1 205 348 2615; fax: +1 205 348 6419.

E-mail address:yguo@eng.ua.edu(Y.B Guo)

unsymmetrical and random distribution of peaks and leys The study of surface topography is very essential as itdirectly impacts the component functionality, such as friction,wear, fatigue, and seal behavior The surface is the bound-ary between the contact components and its surroundings.Any interaction between the contact components takes place

val-at the surfaces Hence, the characterizval-ation of surface raphy is a critical factor when component functionality isconcerned

topog-The geometrical quality is one criterion for the surfaceintegrity (Guo and Warren, 2004) The state-of-the-art hardturning can achieve surface values which are, under certain

0924-0136/$ – see front matter © 2007 Elsevier B.V All rights reserved

doi:10.1016/j.jmatprotec.2007.05.054

conditions, equal to or even better surface roughness than

that by grinding (Pawlus, 1997; Kundrak and Bana, 2003;

Warren and Guo, 2006) The formation of the 3D turned

sur-face topography is the combined action of tool geometry, feed,

depth of cut, material microstructure, and machining system

dynamics In contrast, the geometries of abrasive particles

and grinding wheel may have more influence on a ground

surface topography Coarser particles tend to produce rougher

surfaces while fine abrasive particle produces a smoother

surface A ground surface topography is a function of process

parameters like in turning However, the relation between

the ground surface topography and process parameters is not

as deterministic as that of a turned one leading to a higher

isotropy

The bulk surface profile data in literature is limited to 2D

line tracing with a stylus A 2D profile may not be efficient

in characterizing the surface due to: (a) the reality of contact

surfaces is 3D in nature; (b) the parameter rash or correlation

problem associated with defining 2D parameters; and (c) the

unique effect of 3D surface topography on component

func-tionality Hence, a 3D surface topography analysis is highly

needed

Specific to hard turning and grinding, a comprehensive

3D characterization of surface topography generated by hard

turning versus grinding has not yet been done There is lack of

knowledge of the cause effects of the machining processes on

the 3D surface The scope of this work is limited to

geomet-rical features of the machined surfaces since the mechanical

and physical properties of turned and ground surfaces have

been well studied (Guo and Sahni, 2004; Guo and Janowski,

2004)

The objective of this paper is to comprehensively

char-acterize and compare 3D surface topography features of the

four types of representative hard turned and ground surfaces

at “extreme” machining conditions (gentle and abusive) This

characterization will shed light on the fundamental

relation-ship between surface topography parameters and functional

aspects

2.1 2D surface parameter comparison of turned and

ground surfaces

It has shown that hard turning and grinding produce different

surface structures and layers (Guo and Sahni, 2004)

Sur-face structures are consisted of micro-geometrical asperities

and valleys and macro-geometrical features, such as cracks,

undercuts, etc The surface structures are preferred locations

leading to a stress concentration and therefore crack

initia-tion Siebel and Gaier (Siebel and Gaier, 1956) have shown that

surface roughness would not influence on the fatigue strength

when the average surface roughness Rzis less than 1␮m

How-ever, this characterization of the surface influence by variable

Rz is in adequate, because other surface parameters (Siebel

and Gaier, 1956; Denkena et al., 2002; Borbe, 2001) generated by

hard turning and grinding are very different surface structure

Even though the notches of the turned and ground surfaces

differ in depth and in occurrence, fatigue tests (Denkena et

al., 2002; Renner et al., 2001) with a constant amplitude of thetest piece showed that the specimens which were hard turnedusing unworn tools exhibit a higher fatigue strength than theground specimens Only if a massive tool wear (VB = 200␮m)occurs does the fatigue strength of the hard turned test piecedrop below the fatigue strength of the ground ones

The single cutting edge in turning showed a nearly regularpeak and valley distribution in comparison to grinding withmultiple cutting edges The peak distance is much wider thanthat in grinding Several studies (Brinksmeier and Giwerzew,2003; Elbestawi et al., 2003; Seker et al., 2003) have shown thateven if the dimensions of the width of the roughness profileover the complete measured surface is the same, the distanceand the height between one single peak and valley is com-pletely different in hard turning and grinding

Both outer and inner hard turning and grinding ments (Pawlus, 1997; Kundrak and Bana, 2003; Warren andGuo, 2006; Klocke et al., 2005) have shown that turning at cer-tain machining conditions can achieve equivalent or better

experi-surface roughness Raand RMS than grinding The comparison

of other different surface parameters, such as bearing lengthcurve ratio (Kundrak and Bana, 2003) shows the dissimilaritybetween the turned and ground surface features Even if onevalue is the same, both surface topographies cannot be con-sidered as “nearly the same” Thus, one surface parameter isnot sufficient for characterizing a machined topography.The skewness values obtained for turned and ground sur-faces by Kundrak and Bana (Kundrak and Bana, 2003) wereboth negative Hence, both surfaces showed fluid retentionproperties which are good for wear resistance However, thehard turned surface had higher negative skewness than theground surface which might indicate a better fluid retentionproperty (Valasek, 1996) and, therefore, better wear resistancefor the hard turned surface From this point of review, hardturning would be a recommended finishing process It should

be realized that the above data was obtained from ments with a new tool An increasing tool or grinding wheelwear will significantly influence the surface topography and adifferent set of surface variables are expected However, veryfew data are reported for the cases with worn tools or grindingwheels

experi-2.2 Limitations of 2D surface profile analysis

There are several limitations to a 2D surface profile analysis:(a) As the real nature of contact in surfaces is 3D in nature, a2D profile can only show the surface roughness at a par-ticular plane Moreover, the profile may only pass over theshoulder of a summit and may not represent true peaksand valleys If the surface is strictly uniform and the laypattern is perpendicular to the plane of the profile, thenthe 2D profile may do justice to the surface However, such

a surface is rare in real world applications A 2D profile not represent an anisotropic nature Hence, a 2D profile isinadequate to show the complete and real nature of thesurface It is from this inadequacy of the 2D parametersthat the need for 3D parameters arises

can-(b) The parameter rash (Whitehouse, 1982) problem ciated with defining 2D parameters Parameter rash is

asso-Coolant type Dry turning Dry turning

manifested in two aspects First, parameters defined in

some national and international standards are found to

vary from country to country Second, the dominance of

manufacturers of measurement instrument on the

devel-opment of 2D parameters resulted in some problems of

defining parameters For example, the commonly used Ra

has no direct functional significance and is less significant

in statistics than the root-mean-squares RMS Third, some

parameters are correlated A well-known example of this

is the correlation of the arithmetic mean Raand the Rq

(c) Most importantly, 3D surface topography has a better

cor-relation with component functionality The effects of 2D

surface parameters on functionality have been studied

on wear (Suh and Nagao, 1976; Jahanmir and Suh, 1977),

scuffing (Kelley and Lemanski, 1967), pitting (Berthe et

al., 1977), and run in behavior and fatigue (Zhou and

Hashimoto, 1994) To evaluate the influence of surface

topography on component functionality, a set of 3D

sur-face parameters has to be studied to relate them with

functional aspects General 3D surface parameters have

been defined to explain the fluid retention property of the

surface which could explain potential wear and friction

behavior of the surface (Dong et al., 1994a) The basic

dif-ference between the 3D surface parameters between the

turned and ground surfaces has been poorly understood

Furthermore, if some surface parameters are same for

sur-faces machined by different processes then explaining the

functional aspects becomes even more difficult

Although surface topography has been studied in great

detail in terms of 2D parameters, a comprehensive

compar-ison on the nature of 3D surface topography generated byhard turning and grinding at different machining conditions isstill missing Additionally, the relationship between 3D surfacetopography and functional aspects is poorly understood

Work material of AISI 52100 steel discs of 76.2 mm diameterand 19.05 mm thickness were heat treated at the austeniz-ing temperature 815◦C 2 h, quenched in an oil bath at 65◦Cfor 15 min, and tempered at 176◦C for 2 h producing a finalhardness of 61–62 HRC The test samples were machined byboth face turning and grinding at machining condition inTables 1 and 2

Since the objective is to study and compare surface raphy obtained by turning and grinding a practical approach

topog-is, to generate surfaces at “extreme” machining conditionsincluding “best” and “worst” achievable machining condi-tions Since the machining conditions are very difficult todefine since it depends on the performance of machine toolsand tooling Consequently, the turning and grinding condi-tions were divided into two categories, i.e gentle and abusivemachining conditions in production scenarios For turningexperiments, the “best” achievable machining condition asshown inTable 1is to use a fresh round CBN cutting tool insert

at the gentle machining range to ensure that there would be nophase transformations on the machined surface The “worst”

or abusive turning condition was carried out using a worn ting tool with a flank wear of 0.5 mm and an increased cuttingspeed so that a white layer, a surface burn via phase trans-

cut-Table 2 – Grinding conditions

Ground fresh surface (GF) Ground surface with white layer (GWL)

Grinding wheel Al2O3(dia 254 mm) Al2O3(dia 254 mm)

Fig 1 – Fresh surface by hard turning (HTF).

formation, appears on the ground surface The face turning

experiments were conducted using a CNC lathe which

main-tains a constant cutting velocity to generate uniform surface

integrity

Gentle grinding operation was performed using an Al2O3

wheel which was dressed with a diamond wheel prior to the

grinding tests Ample coolant was used to minimize heat

gen-eration or thermal damage on the ground surfaces Abusive

grinding operation was performed using a dry Al2O3 wheel

without any coolant with an increased depth of cut and higher

table speed as shown inTable 2

The 3D surface topography of the machined surfaces was

measured using a Taylor Hobson Talysurf CLI 2000 3D surface

profiling system Due to the small geometrical features of the

precision machined surfaces, the measurement was carried

out using the inductive gauge with resolution of 10 nm and

measurement range of 2.5 mm The stylus was used to scan

across a set area of the workpiece The collected data is

pro-cessed using the signal processing software package to render

the surface topography maps and 3D surface parameters

4 Results and discussion

4.1 3D surface topography

The 3D topographies of the as machined surfaces are shown

inFigs 1–4 The turned surface topography inFigs 1 and 2

shows well-defined peaks and valleys, while the ground ones

in Figs 3 and 4 show a much more random surface This

is mainly because of the distinct difference in cutting edge

geometry between hard turning and grinding Turning uses

a single cutting edge with defined geometry to generate the

Fig 2 – Hard turned surface with white layer (HTWL).

Fig 3 – Fresh surface by grinding (GF).

machined surface and the turning parameters of feed andcutting edge geometry define the symmetrical topography

of the machined surface In contrast, grinding is carried outwith a grinding wheel with randomly distributed abrasives ofirregular geometry The size and shape of the abrasives andtheir spacing are the decisive parameters for the resultingsurface topography Hence, the parameters which affect thesurface topography in turning and grinding are quite different

in nature The microview of the turned surface is anisotropicand the ground one is more isotropic

The functional aspects of a machined surface are not onlydependent on the roughness but also on many other impor-tant aspects like the distribution of peaks, the sharpness ofthe surface, the bearing area ratio, and other spatial parame-ters The 3D topographies of the turned and ground surfacesclearly show the difference in spacing between the peaks, thesharpness of the peaks, and the randomness of the profile.These aspects of the surface can only be best described using

a set of 3D parameters

It can also be observed that the abusive turned surfacesshow much sharper and random peaks, and surface cracksthan those of the gently turned ones This aspect of the topog-raphy can be measured by the surface bearing area ratio.Figs 1 and 2show the 3D surface topography of the sam-ples produced by gentle and abusive hard turning The majordifferences in process parameter between the two cases arethe higher cutting speed and larger flank wear used for theabusive turning condition The higher cutting speed and toolflank wear cause higher temperatures and more chatter whencompared to the gentle turning condition The result is theincreased waviness and surface roughness along the cuttingdirection for the abusive turned sample The 2D surface pro-

Fig 4 – Ground surface with white layer (GWL).

Fig 5 – Surface roughness along feed direction of the HTF surface.

Fig 6 – Surface roughness along cutting direction of the HTF surface.

files of the hard turned surfaces along the cutting direction are

shown inFigs 6 and 8 It must be noted that all the 2D profiles

have represent pure roughness values, i.e the waviness

com-ponents have been filtered out The profile for the HTF surface

shows a lower roughness value than the profile for the HTWL

surface which is consistent with the 3D topography image

However, as the feed is constant for both cases the surface

topography is nearly identical with respect to distribution of

peaks and valleys

Figs 5 and 6 show the 2D surface profiles of the HTF

and HTWL surfaces along the feed and the cutting directions

respectively Ptis the maximum peak-to-valley height of the

2D surface profile The 2D profile along the feed direction

shows a very repeatable form of peaks and valleys whereas the

profile along the cutting direction shows a random nature The

profile has been taken along the valley portion of the

topog-raphy The random nature of the profile may be because of

random processes like machine vibration, surface tions of the material, etc We also notice that the maximumpeak-to-valley height along the feed direction is much greaterthan along the cutting direction This is because the highestand lowest points of the topography, i.e the ridges and valleysformed by turning are captured by trace along the feed direc-tion only Similar features are demonstrated byFigs 7 and 8which represent he HTWL surface

imperfec-Figs 5–8show that the 2D profile of the HTWL surface has

larger Pt values than those of the HTWL surface along bothfeed and cutting directions This may be because the HTWLsurface has been machined with a worn tool causing deepervalleys and more vibrations

Figs 3 and 4show the 3D surface topography of the groundsurfaces The GF surface appears to be more random andisotropic than the GWL surfaces Also the surface roughness

of the GF is lower than the GWL surface which may be inferred

Fig 7 – Surface roughness along feed direction of the HTWL surface.