You are on page 1of 8

Materials Science and Engineering A 426 (2006) 173180

Finite element modelling of shot peening process:


Prediction of the compressive residual stresses,
the plastic deformations and the surface integrity
M. Frija a , T. Hassine b , R. Fathallah c, , C. Bouraoui b , A. Dogui b ,
Laboratoire de Genie Mecanique
a Institut Superieur des Sciences Appliquees et de Technologie de Sousse, Cite Taffala, 4003 Sousse, Tunisia
b Ecole Nationale dIngenieurs de MonastirAvenue Ibn Jazzar 5019 Monastir, Tunisia
c Ecole Nationale dIngenieurs de Sousse, Cite Taffala, 4003 Sousse, Tunisia

Received 18 November 2005; accepted 30 March 2006

Abstract
This paper presents a numerical simulation of the shot peening process using finite element method. The majority of the controlling parameters
of the process have been taken into account. The shot peening loading has been characterised by using energy equivalence between the dynamic
impact and a static indentation of a peening shot in the treated surface. The behaviour of the subjected material is supposed to be elastic plastic with
damage. An integrated law of the damage proposed by Lematre and Chaboche has been used. The proposed model leads to obtain the residual
stress, the plastic deformation profiles and the surface damage. An application on a shot peened Ni-based super alloy Waspaloy has been carried
out. The comparison of the residual stresses, obtained by X-ray diffraction method and by finite element calculation, shows a good correlation.
The in-depth profile of the plastic deformations and the superficial damage values are in good agreement with the experimental observations.
2006 Elsevier B.V. All rights reserved.

Keywords: Shot peening; Residual stresses; Plastic deformations; Superficial damage; Surface imperfections; Finite element method

1. Introduction cial defects such as overlaps, scales, micro cracks and surface
roughness imperfections [69], which can induce significant
Controlled shot peening is the most widely used method decrease of the high cycle fatigue strength of the treated part
of pre-stressing mechanical components such as, automobile [10,11]. In order to predict the high cycle fatigue behavior of
leaf springs, aero-engine blades, gears, axle bearings, etc. [1]. the shot peened part (e.g. by using fatigue criteria), it is nec-
It consists of projecting small shots made of steel, ceramic or essary to predict completely the initial state of the shot peened
glass, at relatively high velocities (20100 m/s) at the surfaces surface conditions [12].
of the metallic parts. It leads, principally, to induce two favor- To predict the shot peening induced plastic deformations and
able effects: (i) compressive residual stresses in the near surface the residual stresses, a theoretical shot peening model, based
region and (ii) consolidation of the superficial treated layers, on the simplified method of elastic plastic calculations of Zarka
so as to ensure a better resistance to fatigue and stress corro- and Casier [13] has been proposed firstly by Guechichi et al.
sion [25]. However, in many cases, and especially where the [14] and improved by M.T. Khabou et al. and R. Fathallah et
shot peening parameters are not well chosen (large shot, over al. [15,16], respectively. It has the advantages to take into acco-
peening, high velocities, etc.), there is a risk of deteriorating or unt the majority of the controlling process parameters. However,
altering the integrity of the treated surface by inducing superfi- it is difficult to identify the coefficients used to model the
elastic-plastic behavior law of the treated material, particularly,
This work is a part of the Phd thesis of Mr. M. Frija supervised by T. Hassine
for the cases of non-linear work hardening materials.
and R. Fathallah.
Different authors [1719] have developed simple theoret-
Corresponding author. Tel.: +216 98 22 88 69; fax: +216 73 332 658. ical analysis of the shot peening process. Those analytical
E-mail address: raouf.fathallah@issatso.rnu.tn (R. Fathallah). approaches lead principally, to determine the residual stresses

0921-5093/$ see front matter 2006 Elsevier B.V. All rights reserved.
doi:10.1016/j.msea.2006.03.097
174 M. Frija et al. / Materials Science and Engineering A 426 (2006) 173180

and the plastic deformations induced in the first affected layers of elastic-plastic contact. It is defined by the normal projection
of the shot peened part. of the stream velocity to the normal projection of the return
Several models are conducted to simulate the shot peen- velocity ratio.
ing process by using the finite element method. Some of them
Vr sin
are axi-symmetric and consist to simulate an impact of one er = (1)
shot [20,21]. Others were developed by using three-dimensional Vi sin
finite element models in order to be more realistic [2225]. The coefficient of restitution, er , depends on the conditions of
Generally, the shots were supposed to be perfectly rigid. The the contact between the shot and the treated material. It depends
subjected material behavior law was supposed, in some cases essentially on the shot/material hardness ratio.
elastic plastic, and in other cases rate-sensitive material [26,27].
The majority of the cited shot peening finite element simula- 2.2. Calculation of energy stored by plastic deformation in
tions allow predicting the introduced plastic deformations and the peened material
residual stresses profiles in the near surface region.
The principal originality of this work is the prediction of During the impact, the movement of the shot can be a slid-
the superficial shot peening damage, characterized by the dam- ing, an in depth penetration, a spin, or a combination of these
age variable D, proposed by Lematre and Chaboche [28]. The motions. In order to simplify the present study, the following
damage has been supposed isotropic (does not depend on the assumptions have been taken into account: (i) the movement of
direction). The damage has been determined by using an inte- the shot particles is supposed to be an in depth penetration and
grated law, depending on the accumulated plastic deformation. the surface sliding was supposed to be negligible and (ii) the
This model permits, also, to predict and to characterize the shot dynamic impact of the peening shot is supposed to be equiva-
peening surface geometrical irregularities. lent to a static indentation. The energy restored in the plasticized
first outer layers by the static indentation is equal to that restored
2. Mechanics of shot peening process by the dynamic impact.
For the case of normal impact, the difference between the
2.1. Impact of two solids initial and the restored kinetic energies is given by the following
expression [16]:
During an impact between two solids, the two following 1
energy transformation steps can be distinguished: (i) the phase W(=90 ) = mVi2 (1 e2r ) (2)
2
of deformation where the stream kinetic energy is transformed
In our case, we assume that the shot is perfectly rigid, and it
into potential energy of deformation, and (ii) the phase of shot
cannot store plastic deformation. The difference between the
restitution. If the impact is completely elastic, the potential
two kinetic energies W(=90 ) is composed of the energy
energy is totally transformed into returned kinetic energy and if
needed to plastify the shot peened surface layers Wp and the
the contact is elastic-plastic, the potential energy is only partially
dissipated energy during the impact (by vibrations, superficial
transformed. If there is internal degrees of freedom (tempera-
heating, etc.) Wd :
ture, vibrations, deformations, etc.), a part of the stream kinetic
energy can be transformed into internal energy or an internal W(=90 ) = Wp + Wd (3)
energy can be transformed into kinetic energy.
Let us consider a shot with mass, m, impacting a semi-finite The energy dissipated during impact is not exactly determined.
body at a velocity, Vi , under an angle of impingement, , and W. Johnson [29] introduces an efficiency coefficient K, defined
returning at a velocity, Vr , under an angle,  (Fig. 1). The as the energy transmitted to the treated material to the supplied
impact is characterized by a coefficient, er , called the coefficient total initial energy ratio. It is estimated to be 0.80.9 for
of restitution or impact coefficient equals to 1 for an elastic mechanical impacts. Then, the energy transmitted by plastic
contact, 0 for a plastic contact and between 0 and 1 for the case deformation to the treated material can be expressed as follows:
1
Wp = K W = mVi2 (1 e2r )K (4)
2
3. Finite elements simulation of the shot peening process

3.1. Assumptions

The principal assumptions employed in the implementation


of the finite elements model are:

(i) The shot particles are considered rigid spheres of uniform


radius.
(ii) The diameter of the shot taken into account is the nomi-
Fig. 1. Energy transmitted during contact shot/material. nal diameter conventionally defined by the SAE J444 [30]
M. Frija et al. / Materials Science and Engineering A 426 (2006) 173180 175

Table 1
Table of the size distribution of the cast steel shots granulometry (S230)
Number of standard SAE J 444 Nominal dimension of the grains Rate cumulated on sieve no. ASTM Opening sieves ISO 565 (mm)

All pass with no. 18 1.00


10% max on no. 20 0.84
S230 0.6 mm
85% min on no. 30 0.59
97% min on no. 35 0.5

standard. It corresponds to the diameter of the sieve having (vi) The shot is supposed impinging the target surface at
a rate of 85% cumulated shots (Table 1). For example, in the normal incidence = 90 .
case of the S230 shot, 75% of the diameters are randomly
distributed between 0.6 and 0.84 mm. It is worth noticing, 3.2. Finite elements model
that the stream kinetic energy is qualified experimentally
by the Almen arc height [31]. The calculation using 0.6 or The simulation has been carried out by using a three-
0.84 mm diameter will involve a slight change of the veloc- dimensional finite element elastic plastic model. The analysis
ity. However, according to the Eq. (2), the stream kinetic was performed by the commercial finite element code ABAQUS
energy is the same in the both cases. [34]. In this work, a static analysis has been adopted using an
(iii) The area density of impacts is considered uniform. The energy equivalences, in order to simplify the application of the
residual stress profile is obtained by taking the residual coupled elastic-plastic damage analysis. Also, it is worth notic-
stresses values from the in-depth nodes located on the ing that the dynamic analysis needs an important computing time
central axis of the damped zone. We suppose that it is and leads, generally, to obtain instable results.
representative of a shot-peening with 100% coverage. The In order to take into account the effect of simultaneous
calculation should be repeated for the case of high harden- impacts and the effect of the elongation of the first affected lay-
ing materials, which are sensitive to peening coverage [32]. ers [18], the geometry of the part has been reduced on one cell
(iv) In our approach, shot peening is considered as a cold with 1 mm width, 1 mm length and 5 mm height dimensions. To
hardening surface treatment. We assume that the mechan- have semi-infinite solid condition suitable to the Hertz solution,
ical response of the processed material conforms to a geometrical sensitivity study has been carried out to optimise
elastic-plastic behaviour coupled with damage and the the 5 mm thickness of the target part.
treated material is not strain rate sensitive. Particular boundary conditions using four symmetry planes
(v) The velocity of the shot is assumed to be constant during applied on the vertical faces of the target has been considered
the impact. Its value has been determined by using (Fig. 2). The cell was meshed by means of eight-node brick
available curves [32] giving the Almen arc height [33] with solid elements. In order to improve the accuracy of FE solu-
shot velocity for a specific shot. Those curves are obtained tions, the contact region was refined with small elements of
by numerical simulations predicting the Almen arc height 0.01 mm 0.01 mm 0.01 mm.
obtained by the shot peening. The conventionally nom- A sensitivity study has been carried out to optimize the dimen-
inal shots diameter defined by the SAE J444 standard sions of the element in refined zone, by comparison of the
[30] is used in the calculation. They are experimentally analytical solutions of the elastic Hertz contact problem [29]
calibrated [32]. to those obtained by numerical resolution. The shot is modeled

Fig. 2. Finite element model.


176 M. Frija et al. / Materials Science and Engineering A 426 (2006) 173180

n , the value of the damage D(M, n ) is defined as:


S0
D(M, n ) = (5)
S
where S is the area of the intersection of the plane with the
representative volume element, and S0 the effective area of the
intersections of all micro-cracks or micro-cavities, which lie
within S.

3.3.2. The integrated model


The treated surface material behaviour law is considered to
Fig. 3. Definition of Lematre and Chaboche damage variable. be elastic plastic with isotropic hardening. The viscosity of the
treated material under the impact is supposed to be negligible.
We assume that the shot size is not important by comparison to
as a rigid sphere and defined by a reference point corresponding the treated part, so the heating should be principally located in
to its center. the projected shot and in the extremely first outer layer of the
affected surface. In order to predict the shot peening superficial
damage, a Lematre and Chaboche [28] three-dimensional duc-
3.3. Material behaviour law
tile plastic model of damage has been used in its integrated form.
It is expressed as follows:
3.3.1. Damage modelling      
The shape and the size of the shot peening defects can be
Dc 2 H 2
considered as discontinuities of the material at the first affected D= p (1 + )+3(12) D (6)
R D 3 eq
layer [12]. In this approach, the damage variable D, proposed
by Lematre and Chaboche [28], is used to take into account this where is the Poissons ratio, H the hydrostatic stress of the
effect. For a representative volume element of a damaged body applied stress tensor and eq the Von Mises equivalent stress.
at a point M (Fig. 3), oriented by a plane defined by its normal The three variables Dc , R and D are considered constants,

Fig. 4. Flow chart of the calculation.


M. Frija et al. / Materials Science and Engineering A 426 (2006) 173180 177

Fig. 5. The used stressplastic strain curve for the finite elements calculations.

where D is the initial critical deformation for damage and R Fig. 6. The calculated and the X-ray residual stresses profiles in depth of the
the deformation at rupture for witch the damage D is equal to its shot peened Waspaloy.
critical value Dc .
p is the cumulated plastic strain defined by the following
expression: 3.4. Shot peening loading
  1/2
2 P P In this work, we use a shot peening equivalent static inden-
p= : dt (7)
3 tation (Section 2), which is obtained by imposing depth pene-
tration of the rigid shot into the treated surface. The imposed
where P is the rate of plastic strain. penetration depth is calibrated incrementally by comparison of
In our study, the shot peening simulation is reported to a the stored energies determined by the finite elements calculation
serial of indentation. The damage parameter is evaluated from in the case of static indentation and that calculated by Eq. (2)
the degree of cumulus of plastic strain. in the case of dynamic impact. The friction between the shot
The elastic plastic calculation was made incrementally by and the treated surface has been characterized by the Coulomb
using the Von Mises criterion. After each increment, the damage friction model:
variable D is determined, and the Youngs modulus E in the
elastic law is updated as follows: Ff = Fn (9)

Et+t = Et (1 Dt ) (8) where Ff is the friction force, Fn the normal force and the
friction coefficient.
where Et+t is the youngs modulus at the instant t + t and Et
The coefficient of restitution er , the value of the contact effi-
and Dt are respectively the Youngs modulus and the damage
ciency K and the friction coefficient , which are difficult to
value at the instant t.
identify experimentally, are determined by calibration of the
The principal steps of the calculations are presented in flow
finite elements calculated and the X-ray diffraction analysed
chart of Fig. 4. This model is implemented in a subroutine
residual stress profiles.
UMAT in the finite element code ABAQUS [34].

Table 2
Damage parameters
Damage parameters Values

R 0.6
D 0.02
Dc 0.8

Table 3
Shot peening conditions
Almen Shot type Coverage Angle of Friction
intensity A % impingement ( ) coefficient

8A S230 100 90 0.01 and 0.2


Fig. 7. Von Mises equivalent plastic strain profiles vs. depth.
178 M. Frija et al. / Materials Science and Engineering A 426 (2006) 173180

4. Application is 1275 MPa, the total elongation at rupture is 25%, and the
hardness is 50 HRC [32].
The treated part is made of an aeronautical-based The plastic material behaviour law was determined by using
Nickel super-alloy material, Waspaloy (based on Ni0.04C the Holloman law [35]:
19.55Cr 1.37Al 2.95Ti 13.51Co4.25Mo0.0063B0.06Zr
= A np (10)
(wt.%)). It is well used to resist to high temperature, so its strain
rate sensitivity is supposed to be negligible in this work. The where A and n are two material constants, identified by using
mechanical properties of the material are: the initial Youngs the initial and ultimate stress strain values. Fig. 5 shows the
modulus E = 210,000 MPa, the Poissons ratio v = 0.3, the stress plastic strain curve with coefficients A = 1367.11 MPa and
yield stress at 0.2% is 1000 MPa, the ultimate tensile strength n = 0.05. The supposed damage parameters used in this appli-

Fig. 8. The obtained superficial finite elements damage iso-values under a shot peening indentation: (a) = 0.01 and (b) = 0.2.
M. Frija et al. / Materials Science and Engineering A 426 (2006) 173180 179

cation are presented in Table 2. The shot peening conditions


are presented in Table 3. To determine the shot velocity, we use
available curves, giving the Almen arc height versus velocity
for the S230 shot [32]. This abacus is determined by numerical
calculations, where the 0.6 mm conventionally normal diam-
eter is used. The Waspaloy is considered to be not sensitive
to the coverage [32]. So in this modelling, we proceed by none
repeated simultaneous calculations.

5. Results

Fig. 6 shows the residual stress profiles obtained by calcu-


lation and X-ray diffraction technique. A good correlation is
observed between the calculated finite elements and the anal-
ysed residual stresses.
The calibrated model corresponds to the values of K(1 Fig. 11. Comparison of the residual stress profiles obtained with and without
e2r ) = 0.2 and = 0.01. The value of er is estimated equal to damage model.
0.86 when K = 0.8 and equal to 0.88 when K = 0.9. These values agreement with the full width at half maximum (FWHM) of the
are physically coherent for the Ni base super-alloy Waspaloy X-ray diffraction peak profile [32].
case, which is a very hard material. Fig. 8 presents the iso-values of the calculated superficial
The calculated in-depth plastic deformations profile induced damage D. The obtained maximum values, Dmax are 0.24 for
by shot peening is presented in Fig. 7. It shows a qualitative the friction coefficient = 0.01 and 0.44 for = 0.2. This is in
good agreement with the previous works of [11,12].
Fig. 9 shows the change of the maximum surface damage
value Dmax with the friction coefficient, and Fig. 10 shows the
change of the surface residual stress with the friction coefficient.
In order to support the principal originality of this paper and
to study the influence of the proposed damage model on the
obtained residual stress profiles, a supplementary calculation has
been made by elastic-plastic model without damage. The com-
parison of the obtained residual stresses is reported in Fig. 11.

6. Discussion

The calculated results show the feasibility of the proposed


shot peening finite elements simulations. The obtained shot
peening residual stress and plastic deformation profiles, by
finite elements are in good agreement with the X-ray diffraction
analysed profiles. The proposed model has a complementary
Fig. 9. Change of the maximal value of the superficial damage with the friction
coefficient ().
advantage to evaluate the superficial damage values. This is very
interesting to predict correctly the high cycle fatigue resistance
of the shot peened part, as proposed in the works of [11,12],
which take into account the following initial shot peened surface
properties: (i) the compressive residual stress profile, (ii) the
plastic deformation profile and (iii) the superficial damage.
The obtained values of the maximum surface damage are situ-
ated in the border of the shot indentation zone. This is physically
coherent with the previous observations made by Fathallah et al.
[6]. Fig. 12 shows a cross section of the iso-damage values. The
maximum damage values show that the superficial shot peening
cracks will have most probably moustache shape as observed
previously [11,12,36]. Fig. 9 shows that the Dmax value and the
extremely damaged zone increase as the coefficient of friction
increases.
Fig. 10 shows that the amplitude of the surface compressive
residual stress decreases when the friction coefficient increases.
Fig. 10. Change of the surface residual stress with the friction coefficient (). It reaches a slight positive value for the case of friction coefficient
180 M. Frija et al. / Materials Science and Engineering A 426 (2006) 173180

Fig. 12. Cross section of the iso-damage values (case: S230, V = 52 m/s, coverage 100%, friction coefficient = 0.2).

equal to 0.3. This is in agreement with the model proposed by [11] N. Sidhom, A. Laamouri, R. Fathallah, C. Braham, H.P. Lieurade, Int.
Fathallah et al. [16]. J. Fatigue 27 (2005) 729745.
The comparison of the residual stress profiles obtained with [12] R. Fathallah, A. Laamouri, H. Sidhom, C. Braham, Int. J. Fatigue 26
(10) (October 2004) 10531067.
and without damage is physically coherent. It shows a slight [13] J. Zarka, J. Casier, Mechanics Today, vol. 6, Pergamon Press, Nemat-
residual stress relaxation in the first outer layers of the compres- Nasser, 1979.
sive zone (0.1 mm thick). The maximal residual stress amplitude [14] H. Guechichi, L. Castex, J. Frelat, G. Inglebert, Impact surface treatment,
decreases from 1378 to 1313 MPa. This is in coherence with the vols. 1112, Elsevier, London, 1986.
results obtained on shot peened ductile steel [6]. [15] M.T. Khabou, L. Castex, G. Inglebert, Eur. J. Mech., A/Solids 9 (1989)
537549.
[16] R. Fathallah, G. Inglebert, L. Castex, Mater. Sci. Technol. 14 (1998)
7. Conclusion 631639.
[17] S.T.S. Al-Hassani, Mechanical aspects of residual stress development in
shot Peening, in: Proceedings of the First International Conference on
A numerical simulation of the shot peening process using
Shot Peening, 1981, pp. 583602.
finite element method is presented. The principal hypotheses [18] H. Wohlfahrt, The influence of peening conditions on the resulting dis-
are detailed. The shot is supposed to be a rigid sphere. The tribution of residual stress, in: Proceedings of the Second International
mechanical behaviour of the subjected material is assumed to be Conference on Shot peening, Chicago, USA, 1984, pp. 316331.
elastic plastic coupled with damage, using an integrated form of [19] S. Kyriacou, Shot peening mechanics, a theorical study, in: Proceedings
of the Sixth International Conference on Shot Peening ICSP6, 1996.
the Lematre and Chaboche model. The shot peening loading is
[20] K. Schiffner, C. Droste gen. Helling, J. Comput. Struct. 72 (1999)
simulated by a static indentation, obtained by an energically 329340.
equivalence with the dynamic impact. The feasibility of the [21] S.M. Meo et, R. Vignjevic, J. Adv. Eng. Software 34 (2003) 569575.
model has been demonstrated by an application in the case of a [22] Y.F. Al-Obaied, The automated three-dimensional analysis of steel plate
standard shot peening on a Ni base super-alloy Waspaloy. The to shot peening mechanics, in: Proceedings of the Second International
Conference of Shot Peening, 1984, pp. 595606.
obtained results are in a good agreement with the experimen-
[23] S.A. Meguid, G. Shagal, J.C. Stranart, J. Daly, Finite Element Anal.
tal observations. It leads to predict, in addition to, the induced Des. 31 (1999) 179191.
residual stress and plastic deformation profiles, the superficial [24] S.T.S. Al-Hassani, Numerical simulation of multiple shot impact, in: Pro-
damage. This work will be very interesting to optimise as cor- ceedings of the 7th International Conference on Shot Peening (ICSP7),
rectly as possible, the shot peening conditions parameters. Institute of Precision Mechanics, 1999.
[25] M. Guagliano, J. Mater. Process. Technol. 110 (2001) 277286.
[26] S.A. Meguid, G. Shagal, J.C. Stranart, Int. J. Impact Eng. 27 (2) (Febru-
References ary 2002) 119134.
[27] G.H. Majzoobi, R. Azizi, N.A. Alavi, J. Mater. Process. Technol. (2005).
[1] Society of Automotive Engineers Handbook, SAE Publications, 1964. [28] J. Lematre, J.L. Chaboche, Mecanique des materiaux solides, Dunod,
[2] Y. Ochi, K. Masaki, T. Matsumura, T. Sekino, Int. J. Fatigue 23 (2001) 2eme Edition, ISBN 2 10 005662X, 2002.
441448. [29] W. Johnson, Impact Strength of Materials, Edward Arnold, London,
[3] S. Tekeli, Mater. Lett. 57 (2002) 604608. 1972.
[4] H.O. Fuchs, Mechanical Engineers Handbook, Wiley, New York, 1986. [30] SAE J444, Cast Shot and Grit Size Specifications for Peening and Clean-
[5] A.M. Eleiche, M. Megahed, N.M. Add-Allah, J. Mater. Process. Technol. ing, SAE Standards, May 1993.
113 (2001) 502508. [31] SAE J443, Procedures for using standard shot peening test strip, SAE
[6] R. Fathallah, H. Sidhom, C. Braham, L. Castex, Mater. Sci. Technol. 19 standard, 1961 (revised September 2004).
(8) (August 2003) 10501056. [32] R. Fathallah, Modelisation du Procede de Grenaillage: Incidence des
[7] B. Gentil, M. Desvignes, L. Castex, Mater. Tech. (Decembre 1987) Billes et Taux de Recouvrement, These de Doctorat de lENSAM, Paris,
493497. 1994.
[8] M. Desvignes, Influence du grenaillage de precontrainte sur la tenue en [33] W. Cao, R. Fathallah, L. Castex, Mater. Sci. Technol. 11 (1995) 967973.
service de lacier 35CD4, These de Doctorat de lENSAM, 1987. [34] Hibbit, Karlsson and Sorensen Inc, ABAQUS/Standard Users manual,
[9] R. Ahmed, Etude de la creation des fissures en surface de lacier 35CD4 vol. II, version 6.4.1, 2003.
lors du grenaillage de precontrainte, memoire de DEA mecanique [35] R.W. Hertzberg, Deformation and Fracture Mechanics of Engineering
appliquee a la construction, ENSAM Paris, 1987. Materials, New York, 1996, p. 17, ISBN 0-471-01214-9.
[10] L. Wagner, Mater. Sci. Eng. A263 (1999) 210216. [36] V. Fridrici, S. Fouvry, Ph. Kapsa, Wear 250 (2001) 642649.

You might also like