Professional Documents
Culture Documents
Tribology International
journal homepage: www.elsevier.com/locate/triboint
Institute of Structural Engineering, Poznan University of Technology, Piotrowo 5, 60-965 Poznan, Poland
National Engineering School of Metz, Laboratory of Mechanics, Biomechanics, Polymers and Structures, 1 route dArsLaquenexy, 57000 Metz, France
c
University of Lorraine,Laboratory for the Study of Microstructures and Mechanics of Materials, Ile du Saulcy, 57045 Metz, France
d
Department of Civil Engineering, American University of Sharjah, P. O. Box 26666, Sharjah, U.A.E
b
art ic l e i nf o
a b s t r a c t
Article history:
Received 28 August 2015
Received in revised form
26 October 2015
Accepted 30 October 2015
Available online 15 November 2015
A proper denition of the real dynamic contact behavior is very important in numerical simulation of
different dynamic processes such as machining, high speed cutting, impact and perforation of structures.
This paper presents an optimal analysis for accurate estimation of the dynamic friction coefcient using a
tribometer device. A three-dimensional nite element (FE) model is developed to better understand the
test set-up and its associated method to accurately dene the friction coefcient under dynamic loading.
The aspect of the experimental methodology currently used to dene the dynamic friction law and its
main drawbacks are identied and discussed. Based on the FE analysis, a new methodology is proposed
to estimate the dynamic friction by introducing the concept of a correction factor that can be used to
correct the old experimental data. This correction factor depends on the initial projectile velocity and
pressure for any friction coefcient value.
& 2015 Elsevier Ltd. All rights reserved.
Keywords:
Dynamic friction coefcient
Tribometer
Impact
Finite element
1. Introduction
When the friction phenomenon has to be investigated, a large
variety of apparatus is employed. In the domain of low and
moderate sliding speeds, classical instrumentations such as the
pin-disk system are mainly used [1,2]. However, when high or
very high sliding velocity is considered, the development of specic devices is often necessary. For the tests reliability, the local
conditions of pressure and sliding speed must be achieved with a
simple set of geometrical elements and several tribosystems to
achieve a convenient access of the different measures. To attain
this goal, some techniques use, for example, a pendulum-type test
device to apply a single pass, while others utilize a modied torsional Kolsky's bar apparatus or direct impact of plates for applying high normal pressure [35]. Although friction is a basic interaction phenomenon between two bodies, the experimental results
such as the evolution of friction forces and wear mechanisms are
not only strongly dependent on the tribological conditions [69]
but also on the instrument itself. For each proposed devices, it is
then necessary to study the exact conditions in which the different
measurements are carried out.
It must be mentioned that the friction coefcient is an important
parameter that is frequently used in numerical simulations but
n
Corresponding author.
http://dx.doi.org/10.1016/j.triboint.2015.10.039
0301-679X/& 2015 Elsevier Ltd. All rights reserved.
2. Experimental setup
In order to analyze the dynamic friction process between two
surfaces under a high sliding velocity combined with a high normal pressure, a specic setup is used [2022]. This setup (Fig. 1)
can cover a wide range of sliding speeds; up to120 m/s with a
maximum normal pressure close to 200 MPa.
In this setup, the friction is produced between specimen B and
the two plates A. The two parts (plates A) are symmetrically xed
into the dynamometer ring, see Fig. 1a. The difference between the
thicknesses of the plates and the width of the specimen B denes a
compression thickness notied as L which imposes the normal
pressure P, between dynamometer ring and specimen B, see
Fig. 1b. Different widths of the pads are chosen to vary L and thus
the normal pressure. The gas gun is used to propel the projectile in
the launch tube and impose the impact velocity. After impact
between the projectile and the specimen B, the latter slides across
the dynamometer ring. The strain waves transmitted to the load
sensor are measured using two strain gauges symmetrically glued
on two thin walls of the load sensor [2022]. This technique is
based on the well know process of the Split Hopkinson Pressure
Bar (SHPB) [35]. For each test using strain gauges under dynamic
loading as in this paper, the location of the point of measurements
are crucial and must be studied precisely using for example an
inverse method to avoid errors or disturbances [23,24].
The two most important parameters during the experimental
test are: the applied normal pressure, P and the initial impact
velocity, V 0 . The initial impact velocity is measured using laser
sensors and is dependent on the initial pressure in the gas gun.
Fig. 2 shows sample results of two signals measured during a typical
experiment with an initial velocity equal to 23 m/s and a pressure of
49 MPa [23]. The strains are measured on two opposite sides of the
load sensor on its outer surfaces. Additionally taking into account
these two parameters and the geometry of the device the dynamic
friction coefcient, t , can be calculated as follows [23]:
F T t
FN
where F T and F N are the friction and normal forces which are
dened as follows:
F T t E U AG U ~ t and F N P U AC
Here E is the Young modulus, AG is the cross section of the part with
glued gauges, P is applied normal pressure and AC is the area of
contact. Their corresponding values used in the preliminary analysis
are listed in Table 1. The strain level
t used to calculate the
friction force (Eq. (2)) is taken as the readings average of the two
strain gauges [23].
To have a better understanding of this set-up and the method
associated to dene the friction coefcient under dynamic loading,
a 3D numerical model has been built as will be explained next.
0,0010
Gauge 1
Strain signal (-)
87
Gauge 2
0,0005
0,0000
-0,0005
-0,0010
0
0,001
0,002
Time (s)
0,003
0,004
Fig. 2. Experimental example of strain signals in two gauges for 23 m/s initial
velocity and 49 MPa pressure [23].
Fig. 1. Detailed description of the device used for dynamic friction measurements, denition of the force sensor zone.
88
E (MPa)
AC (mm2)
AG (mm2)
210,000
120
20
Interference fit:
0.02-0.4 mm on both sides
Fig. 3. Model geometry and mesh conguration for the pretention analysis.
200
160
120
y = 359,3x + 30,5
80
40
0
0
0,2
0,4
Pad thickness (mm)
0,6
89
The walls of the force sensors were discretized using continuum shell element to predict properly the compressive and
bending behavior of the walls. The transmitted bar head (place
close to the force sensor), see Figs. 1 and 5 were discretized using
continuum C3D8R brick element and the rest of the transmitted
bar (5 m) were discretized with shell elements to have real non
reection boundary conditions as in experiments. The projectile
has 85,004 C3D8R nite elements. The transmitted bar is composed of 91,000 C3D8R nite elements whereas the force sensor
was modeled using 1248 SC8R and 64 of SC6R type including all
the four thin walls where the strain measuring is done. The SC8R
element is 8-node hexahedron, general-purpose, nite membrane
strains continuum shell element and SC6R is 6-node triangular inplane continuum shell wedge, general-purpose, nite membrane
strains continuum shell element. Using this type of discretization
helps obtain accurate strain measurements and correct bending
behavior of this exible structure, see Figs. 1 and 5.
As discussed previously the local strains were recorded only on
the outher surface and middle of the piece no.1, points B in Fig. 5.
The goal is to validate the experimental analysisdescribed in [23]
and to demonstrate that by considering only these two readings
the friction cofcient may not be dened correctly. This analysiswas done using several cases where the friction coefcient
Projectile
Specimen
Ring
Piece no. 2
(Zone 2)
A - Side
B - Middle
Piece no. 1
(Zone 1)
A - Side
B - Middle
Tube
Fig. 5. 3D FE model used to estimate the dynamic friction measurement device.
90
0,4
Input 0,3
0,3
0,2
0,1
0
0,0005
0,001
Time (s)
0,0015
0,002
Fig. 6. Verication of the friction coefcient values for the case of V 0 23 m/s and
P 49 MPa using numerical simulations.
Fig. 7. Strain history in all considered measuring points using the numerical model.
0,5
Friction coefficient [-]
Friction-old
Friction-new
0,4
Average0,303
0,3
Input 0,3
0,2
Average0,202
0,1
0
0
0,0005
0,001
Time [s]
0,0015
0,002
~ t
i1
i t
t zone1 t zone2 t
The results using Eq. (4) are reported in Fig. 9 which shows 83%
of the friction value is obtained from zone 1 and 17% from zone 2.
Therefore, for this specic case, an error of 30% is observed when
comparing the assumed value with the numerical prediction using
the original (old) analysis reported in [23].
The main goal of this research is to nd a solution as how to
correct the previous experimental results knowing the limitation
of the experimental setup. One of these limitations, is the difculty
to add gauges/sensors on the inner side of the wall to record the
pieces bending (the projectile is ying through the inner hole).
However, the results obtained by the present numerical analysis
for different cases of impact will be utilized to introduce the
concept of a correction factor for the friction coefcient. This
correction factor could be used with the old methodology to
estimate the correct value of friction coefcient.
Although several combinations of normal pressures and impact
velocities were performed experimentally, only four cases are
discussed in this paper and compared with numerical simulations.
The studied cases include four different combinations of normal
pressures and initial impact velocities as listed in Table 3.
The next sub-sections will explain how to calculate and use the
correction factor to correct the friction coefcient obtained
experimentally.
4.1. Correction factor for friction coefcient
The concept of a correction factor (CF) for the coefcient of
friction that is calculated from the numerical simulations using the
original methodology is introduced in this section. Thus, the friction correction factor is calculated as the ratio between the friction
coefcient values obtained using the new methodology and the
ones obtained using the original methodology.
It was observed from the previously veried results of case 1 in
Table 3 that the friction coefcient was consistently underestimated, and a correction factor of 1.5 could be deduced. This
constant value was used to correct the old methodology to obtain
experimental friction coefcients. The correction factor of case
1 was veried fora range of friction inputs between 0.0 and 1.0 as
illustrated in Fig. 10. The variation of the friction coefcient
throughout the impact time step using the old methodology is
shown in Fig. 10a, whereas, the results of the corrected old
methodology (i.e., after considering the CF) are presented in
Fig. 10b.
It can be seen from Fig. 10 that using a constant CF 1.5 accurate corrections of the old methodology could be predicted
Table 2
The data for friction coefcient calculation new methodology.
AG (mm2)
210,000
120
80
0,4
AC (mm2)
E (MPa)
zone 1
0,3
0,2
83%
0,1
0
0
0,0005
0,001 0,0015
Time (s)
0,002
91
Case
1
2
3
4
49
168
42
151
23
17
75
76
0,4
zone 2
0,3
0,2
17%
0,1
0
0
0,0005
0,001 0,0015
Time (s)
Fig. 9. Decomposition of friction coefcient measurements (FE results) into zone 1 and zone 2.
0,002
92
Friction coefficient
Multiplied by CF
(CF=1,5)
2
1,5
1,5
0,5
0,5
fric 0,0
fric 0,1
fric 0,2
fric 0,3
fric 0,7
fric 1,0
-0,5
-0,5
0
0,0005
0,001
0,0015
0,002
Time (s)
0,0005
0,001
0,0015
Time (s)
0,002
Fig. 10. Friction coefcient variations over the dynamic time step using based on numerical simulations (a) old methodology and (b) after correction.
Table 4
Friction coefcient results predicted using the corrected old methodology, comparison with the original predictions.
imposed
old
new
Error between new and
assumed values (%)
0
0.0049
0.007
100
0.1
0.073
0.109
8.26
0.2
0.139
0.210
4.60
0.3
0.202
0.303
1.13
0.7
0.458
0.688
1.82
1
0.649
0.974
2.68
Table 5
Correction factors for all cases.
Correction factor CF
1
2
3
4
1.5
1.9
1.3
2.15
Case
2,5
2
1,5
1
case 1 and 2
case 3 and 4
0,5
0
0
50
100
150
Normal pressure, P (MPa)
200
Fig. 11. Variation of the correction factor with normal pressure applied using
numerical simulations.
Table 6
Corrected friction coefcients from experiments.
Case
1
2
3
4
0.23
0.2
0.15
0.1
0.35
0.36
0.19
0.22
0,4
0,3
y = -0,0027x + 0,4085
0,2
0,1
0
0
20
40
60
80
Initial impact velocity, V0 (m/s)
0,4
93
y = 8E-05x + 0,3459
0,3
y = 0,0003x + 0,1784
0,2
case 3 and 4
0,1
case 1 and 2
0
0
50
100
150
Initial normal pressure, P (MPa)
200
35
Sliding velocity, VS (m/s)
Fig. 12. Variation of corrected friction coefcient with initial (a) impact velocity and (b) normal pressure based on corrected experiments.
33
32
0
0,2
0,4
0,6
0,8
Friction coefficient, (-)
case-2-vel-23
-10
-20
-30
difference in pressure is 119MPa
between case-1 and case-2-vel-23
-40
Fig. 13. FE results showing the inuence of the friction coefcient on the maximum
sliding velocity V S max for an initial impact velocity V 0 23 m/s.
case-1
0,0005
0,001
Time (s)
0,0015
0,002
0
-10
fric 0,0
fric 0,2
-20
fric 0,3
fric 0,7
-30
-40
fric 1,0
case-3
case-4
-70
-80
-90
-100
0
0,0005
0,001 0,0015
Time (s)
0,002
0
-20
case-2
-40
Limit line - after this line the
friction process is ended
-60
case-3
case-4
case-1
-80
case-2-vel-23
-100
0
0,001
0,002
Time (s)
0,0005
0,001
Time (s)
0,0015
0,002
Fig. 14. Variation of the sliding velocity dependence on friction coefcient for
case-1 using FE analysis.
-60
0,003
0,004
Fig. 15. FE sliding velocity for the four cases considered using a constant friction
coefcient of 0.3.
94
Acknowledgment
Table 7
Friction analysis data to be used for a proposed friction law.
Sliding velocity (m/s)
45.5
159.5
22
0.35
0.36
83
0.19
0.22
5. Conclusions
The friction parameters calculated for the experimental cases
considered in the present analysis are presented in Table 7. Taking
into account the variation of the sliding velocity during the friction
process, a friction law can be proposed and utilized in dynamic
friction analysis. It considers an average sliding velocity which is
generally different from the sliding velocity assumed by the initial
impact velocity. It is clear that the sliding velocity is an important
and fundamental parameter as compared with the initial normal
pressure. It should be noted that increasing the sliding velocity
decrease the friction coefcient value. Opposite relation is found
concerning the initial normal pressure.
The present paper discussed in details many aspects of the
current inaccurate experimental methodology used to dene the
dynamic friction law. The main problems of this old method were
identied and discussed. A new methodology was, therefore,
proposed to dene the dynamic friction coefcient accurately for
further experiments with dynamometer ring. The new methodology was developed numerically using nite element simulations of the real experiments. Caution should also be considered in
nite element modeling to account for the evolution of the real
contact area. In fact, the contact area evolves with the normal
pressure and cannot be assumed constant. It was demonstrated
numerically that the simple analysis approach reported in [2023]
is not enough to estimate precisely the friction coefcient under
dynamic loading for different normal pressures. Several assumptions used in those old studies were not consistent as how to
dene the average strain using the dynamometer and the sliding
velocity during the process. In this paper, a precise analysis based
on an inverse method was developed to estimate the correct
values and to correct the previous experiments as reported in [23].
The crucial outcome is the idea of introducing a correction factor
to correct the old methodology in order to dene the dynamic
friction coefcient accurately. In general, this correction factor was
found to be dependent on initial pressure and impact velocity and
independent of friction coefcient.
References
[1] Nair RP, Grifn D, Randall NX. The use of the pin-on-disk tribology test
method to study three unique industrial applications. Wear 2009;267:8237.
[2] So H. Characteristics of wear results tested by pin-on-disc at moderate to high
speeds. Tribol Int 1996;29:41523.
[3] Espinosa HD, Patanella A, Fischer M. A novel dynamic friction experiment
using a modied Kolsky bar apparatus. Exp Mech 2000;40:13853.
[4] Rajagopalan S, Irfan MA, Prakash V. Novel experimental techniques for
investigating time resolved high speed friction. Wear 1999;225:122237.
[5] Rajagopalan S, Prakash V. A modied torsional Kolsky bar for investigating
dynamic friction. Exp Mech 1999;39:295303.
[6] Bowden FP, Freitag EH. The friction of solids at very high speeds. I. Metal on
metal; II. Metal on diamond. Proc R Soc Lond Ser Math Phys Sci 1958:35067.
[7] Archard JF. Contact and rubbing of at surfaces. J Appl Phys 1953;24:9818.
[8] Bowden FP, Moore AJW, Tabor D. The ploughing and adhesion of sliding
metals. J Appl Phys 1943;14:8091.
[9] Greenwood JA, Tabor D. Deformation properties of friction junctions. Proc Phys
Soc Sect B 1955;68:609.
[10] Lodygowski T, Rusinek A, Jankowiak T, Sumelka W. Selected topics of high
speed machining analysis. Eng Trans 2012;60:6996.
[11] Cristol-Bulthe AL, Desplanques Y, Degallaix G. Coupling between friction
physical mechanisms and transient thermal phenomena involved in pad-disc
contact during railway braking. Wear 2007;263:123042.
[12] Gellert EP, Cimpoeru SJ, Woodward RL. A study of the effect of target thickness
on the ballistic perforation of glass-bre-reinforced plastic composites. Int J
Impact Eng 2000;24:44556.
[13] Iwamoto Takeshi, Yokoyama Takashi. Effects of radial inertia and end friction
in specimen geometry in split Hopkinson pressure bar tests: a computational
study. Mech Mater 2012;51:97109.
[14] Zhong WZ, Rusinek A, Jankowiak T, Abed F, Bernier R, Sutter G. Inuence of
interfacial friction an dspecimen conguration in split Hopkinson pressure bar
system. Tribol Int 2015;90:114.
[15] Sutter G, Philippon S, Garcin F. Dynamic analysis of the interaction between an
abradable material and a titanium alloy. Wear 2006;261:68692.
[16] Mandard R, Witz JF, Boidin X, Fabis J, Desplanques Y, Meriaux J. Interacting
force estimation during blade/seal rubs. Tribol Int 2015;82:50413.
[17] Rusinek A, Zaera R, Forquin P, Klepaczko JR. Effect of the plastic behavior and
boundary conditions coupled with elastic wave propagation on the collapse
site of a crashbox, ThinWalled Structures. Thin-walled Struc 2008;46:1143
63.
[18] Sanjeev NK, Vinayak M, Suresh H. Effect of coefcient of friction in nite
element modeling of friction stir welding and its importance in manufacturing
process modeling applications. Int J Appl Sci Eng Res 2014;3(4):75562.
[19] Lovell MR, Deng Z. Characterization of interfacial friction in coated sheet
steels: inuence of stamping process parameters and wear mechanisms. Tribol Int 2002;35:8595.
[20] Sutter G, Philippon S, Molinari A. An experimental investigation of dry friction
for a large range of sliding velocities. Matr Tech 2004:337.
[21] Sutter G, Ranc N. Flash temperature measurement during dry friction process
at high sliding speed. Wear 2010;268:123742.
[22] List G, Sutter G, Arnoux JJ. Analysis of the high speed sliding interaction
between titanium alloy and tantalum. Wear 2013;301(12):66370.
[23] Sutter G, List G, Arnoux JJ, Rusinek A. Finite element simulation for analysing
experimental friction tests under severe conditions. Finite Elem Anal Des
2014;85:508.
[24] Rusinek A, Cheriguene R, Baumer A, Klepaczko JR, Larour P. Dynamic behaviour of high-strength sheet steel in dynamic tension, experimental and
numerical analyses. Strain Anal Eng Des 2008;43:3753.
[25] Dassault Systemes Simulia Corp, Abaqus 6.11 analysis user's manual. Providence (RI): DassaultSystemesSimulia Corperation; 2011.