Wear 436–437 (2019) 202947

Contents lists available at ScienceDirect

Wear
journal homepage: www.elsevier.com/locate/wear

Evolution in microstructure and compression behaviour of a metallic T

sintered friction material after braking
I. Serrano-Munoza,b,∗, J. Rapontchombob, V. Magnierb, F. Brunelb, S. Kossmanc, P. Dufrénoyb,∗∗
a
Bundesanstalt für Materialforschung und -prüfung (BAM), Unter den Eichen 87, 12205, Berlin, Germany
b
Univ. Lille, CNRS FRE 2016 - LaMcube - Laboratoire de Mécanique, Multiphysique, Multiéchelle, F-59000, Lille, France
c
Arts et Métiers ParisTech, MSMP, Lille, 59800, France

ARTICLE INFO ABSTRACT

Keywords: Due to the complexity of friction materials, the characterization of the tribological properties is prioritised over
Semi-metallic sintered material the bulk material properties even though the tribology is expected to be influenced by the material behaviour.
Braking load history The extent of this relationship is still unknown and further knowledge is required to account for the load history
Uniaxial compression tests and evolution of the bulk properties. With this view, the compression behaviour and microstructure of a semi-
Digital image correlation
metallic friction material with reduced formulation were investigated before and after a braking program. The
(DIC)
Scanning electron microscopy
thermal loading was monitored with inserted thermocouples. Uniaxial compression tests coupled with Digital
(SEM) Image Correlation (DIC) show significant changes in the worn material, which develops a compression behaviour
Evolution of bulk properties similar to that of a tri-layered material. The microstructural analysis indicates microcracking of the metallic
matrix and carbon diffusion in the Fe-phase. The thermal loading was found to be the key parameter controlling
both the friction behaviour and evolution of the material properties. The expected effects of material evolution
on the contact uniformity, durability and tribology are discussed.

1. Introduction the contact surface (sub-surface material) has a significant influence on

the mechanisms controlling the generation of wear sources in friction
Friction materials are at the heart of defining the performance of materials (e.g., erosion, cracks or hot-spots) [2–10]. Regarding the bulk
braking systems. Because these materials are crucial to the transpor- properties, the studies published to date mainly focus on the mechan-
tation safety, they have to fulfil numerous requirements involving dif- isms involved in disc cracking. For cast-iron [11,12] and steel [13–16]
ferent fields: a constant level of friction whatever the conditions of use materials, it was shown that the cracking is induced by phase trans-
to ensure the stopping distance, a low rate of wear to decrease the life- formations occurring during the braking. In Refs. [15,16], the trans-
cycle costs, high mechanical strength for integrity, and limited noise formation from austenite to martensite/bainite is observed in steel discs
and vibration, just to name a few. The extensive requirements have led that endured cycles of high thermal loading (800–1200 °C) followed by
to the development of very complex materials containing many dif- rapid cooling. On the other hand, very few studies deal with the bulk-
ferent component types [1,2]. The use of a metal-matrix composite properties evolution of the friction material facing the disc [7,17]. In
(MMC) is needed to withstand the high temperature resulting from the particular, reference [17] reported a strong modification in the me-
dissipation of very high energies (e.g., in the railway industry). In ad- chanical properties of a semi-metallic material after a high-energy
dition, fillers (e.g., ceramic and graphite particles) need to be added to braking sequence was applied (e.g., the stiffness decreased from
fulfil the wear and friction requirements. In the past, this complexity 5800 MPa in the virgin material to 1100 MPa in the worn one).
(material and conditions of use) has led to empirical developments The above studies indicate that the evolution of friction materials
mostly reliant on trial-error iterations. during braking has a multilevel nature (i.e., changes in the bulk and
The investigation of the friction level and wear rates along with the subsurface affect the contact surface). The further advancement of
characterization of the contact surface (usually limited to a depth of friction materials requires in-depth understanding of how the evolution
100 μm) provides knowledge of some key mechanisms involved in the in bulk properties influences the braking performance. The present
tribology. It has been reported that the material immediately beneath study approaches the challenge by investigating the mechanisms

∗
Corresponding author. Bundesanstalt für Materialforschung und -prüfung (BAM), Unter den Eichen 87, 12205, Berlin, Germany.
∗∗
Corresponding author.
E-mail addresses: [email protected] (I. Serrano-Munoz), [email protected] (P. Dufrénoy).

https://doi.org/10.1016/j.wear.2019.202947
Received 17 January 2019; Received in revised form 22 May 2019; Accepted 25 June 2019
Available online 03 July 2019
0043-1648/ © 2019 Elsevier B.V. All rights reserved.
I. Serrano-Munoz, et al. Wear 436–437 (2019) 202947

involved in the evolution of the compression behaviour. Moreover, a (ESRF), Grenoble, France) was performed to gain qualitative knowledge
detailed characterization of the mechanical properties will lead to more about the porosity. The investigated sample was a cylinder of 5 mm in
realistic numerical simulations, ultimately reducing the amount of diameter and 18 mm in height (where the height is parallel to the
parametric studies to be performed in the test-bench. The usefulness of compacting direction). The scanning was performed at 110 KeV with a
thermomechanical modelling is shown, for example, in Refs. [17–23]. voxel size of 3.425 μm3. The images were reconstructed from projec-
These studies used numerical simulations to investigate the influence of tions by applying Paganin's phase retrieval algorithm [24] and sub-
brake-pad design on the development of contact pressure localisations sequent filtered back-projection reconstruction for parallel beam. The
and subsequent formation of hot spots. SCT images correspond to the mesoscopic scale length.
Given that the multi-loading and multi-physics nature of the braking In the reconstructed images (Fig. 2a), the two types of graphite
process generally obscures the elucidation of mechanisms, the present particles and Ceramic 1 particles exhibit similar dark-grey levels.
study utilizes a sintered friction material with reduced formulation Ceramic 2 particles are shown in light-grey and the metallic matrix is
(only nine components) and the mechanical characterization is limited close to white (grey-scale values of 220–255). Close observation of the
to uniaxial compression tests. The braking program is carried out using graphite particles indicates that these particles contain some crack-like
a real-scale dynamometer, where the temperature level of the friction intrinsic porosity (Fig. 2b). It is likely that the intra-porosity only occurs
material is monitored with inserted thermocouples. The friction beha- within Graphite 2 particles, as this type of porosity is not observed
viour and differences in compressive properties and microstructure are within the smallest Graphite 1 particles. The SCT images also show
examined. The mechanisms involved in the evolution of the compres- some interface-porosity (considered as a discontinuity, Fig. 2c) between
sion behaviour are discussed along with the expected influence of these the metallic matrix and the graphite/ceramic particles, as well as be-
mechanisms on the braking performance. tween individual graphite/ceramic particles. It is, therefore, probable
that the porosity distribution throughout this friction material is com-
2. Material and methods plex and heterogeneous. Moreover, it is likely that this mesoscale por-
osity only accounts for a portion of the 14% measured by the Archi-
2.1. Material medes method. Nevertheless, a detailed study of the nanoporosity
content within the metallic matrix is outside of the scope of this study.
The simplified formulation (reduced to only nine components) is
given in Table 1. This material can be considered as a FeCu-based Metal
Matrix Composite (MMC) that contains particles of graphite, ceramics, 2.2. Methods
and a small percentage of other fillers, which are included in the me-
tallic matrix percentage. 2.2.1. Braking set-up and protocol
The manufacturing process consists of three main steps: (i) pre- A real-scale railway dynamometer, replicating those used in TGV
paration of the raw powders, including weighing and mixing, (ii) high-speed trains, was used to apply realistic braking loads to the
compacting of the powders (up to 1000 tons along the Z axis) using a friction pair. It was equipped with a solid disc (non-ventilated) made of
shaped mould to consolidate and densify the model (also known as a 28CrMoV5 steel (Fig. 3a) and brake pads made of the friction material
green compact), and (iii) sintering in a protective-atmosphere furnace described in Section 2.1. The brake pads and disc were brought into
for 8 h at 1000 °C. The final geometry of the brake-pad is sur- contact by a pneumatic caliper actuated with a pressure-control system.
face = 200 cm2 and height = 22.5 mm. Note that the compacting and In addition to the standard output data usually acquired during typical
in-service loads are applied along the same direction. laboratory experiments (e.g., actuator pressure, torque and rotation
The wide range in the size of particles leads to a complex multiscale speed), thermal monitoring was carried out to record the temperatures
material. As shown in Fig. 1a, only Graphite 2 particles can be observed induced in the friction pair. The pad thermocouples were inserted from
at the macroscopic scale length (i.e., with the naked eye). In addition to the back side of each brake pad at two different depths (3 and 6 mm,
Graphite 2 particles, Graphite 1 (also shown in black but with smaller measured from the virgin contact surface). The thermocouple locations
sizes) and the ceramic particles are distinguishable at the mesoscale are shown in Fig. 3b. Regarding the disc, three rubbing thermocouples
(Fig. 1b). Finally, the two principal phases of the metallic matrix (ferrite were used (see Fig. 3a), as well as an infrared camera positioned in
iron and copper) can be observed at the microscopic scale length front of the disc surface to record the occurrence of thermal localisa-
(Fig. 1c and d). Note that some pearlite colonies are observed in Fig. 1c tions.
and d. These colonies most probably result from the diffusion of carbon, The test program was specifically designed to investigate the level of
which has originated from the graphite particles, into the Fe matrix thermal loading induced by an increasing energy dissipation during
during the sintering process. braking and follows the guidelines of the Union Internationale des
The density of the virgin material was measured using the Chemins de fer (UIC, International Union of Railways is the translation
Archimedes method, which compares the sample weight when mea- in English). An initial S0 braking sequence, not noted in Fig. 3c, was
sured in dry air and saturated in ethanol, giving information about the performed prior to the program. This initial braking was performed
open porosity. For the studied material, the porosity content is 14% with the same normal force used in S1V0. Once 80% of the virgin
(mass percentage). Synchrotron Computed Tomography (SCT, carried contact surface was worn away by friction (the verification was done
out at ID19 beamline of the European Synchrotron Radiation Facility visually), the program shown in Fig. 3c was started. This braking pro-
gram consisted of three blocks (series) of low, medium and high braking
Table 1 energy (S1, S2 and S3). The increment in braking energy was achieved
Friction material composition. by increasing the inertial mass to be stopped (see fifth column in
Main Components Mass (% ) Size Fig. 3c). The braking force applied on the pad (defined as normal force
in Fig. 3c) was also increased in each series. Before the starting of each
Metallic Matrix (MM) 70 Sintering powder series, five low-energy braking sequences were carried out as a
< 100 μm
breaking-in procedure, meant to resettle the contact surfaces for the
Graphites 1 & 2 20 100–600 μm
400–1100 μm
following series. These series were subdivided into five braking se-
Ceramics 1 & 2 10 100–200 μm quences, where the speed (equivalent to the speed of an in-service
100–250 μm train) was progressively increased. Each braking sequence is noted as S
(i)V(i).

2
I. Serrano-Munoz, et al. Wear 436–437 (2019) 202947

Fig. 1. Microstructure of the studied friction material: (a) Light-optical image at the macroscopic scale length. (b) Backscattered Electron (BSE) image at the
mesoscopic scale length (c) Secondary Electron (SE) image of the metallic matrix at the microscopic scale length and (d) same as previous image but in BSE mode.
Here, the pearlite colonies are almost indistinguishable. The surface was etched with Nital reagent for 5 s.

Fig. 2. (a) 2D slice (transverse plane) of the reconstructed synchrotron CT data corresponding to the material mesoscopic scale length. (b) Detail of a Graphite 2
particle showing some intrinsic porosity. (c) Detail of the metallic matrix, where the porosity/discontinuity between the matrix and the graphite/ceramics particles in
addition to some porosity at the interface between individual graphite/ceramic particles can be observed.

2.2.2. Uniaxial compression tests During the compression tests, two optical fibres illuminated the
Compression tests required the extraction of cubic samples ( 20 × sprayed surface while a digital camera (16-bit digitisation, 2048×2048
20 × 20 mm3, a volume considered large enough to be representative of pixel array) recorded macroscopic images of the frontal surface (
the macroscopic behaviour) from the central region of the brake pad 20×20 mm2) every 500 ms (see Fig. 4b). The pixel size was 100 μm.
(see Fig. 4a). A testing machine equipped with a 50 kN load cell was YaDICs free software [25] was used to calculate the displacement fields.
used with a loading rate of 0.01 mm/s. The loading/unloading direction This software uses the entire image as a sampling grid (i.e., total sam-
corresponds to the in-service braking direction. No important barrelling pling), and uses a bicubic interpolation for the grey level evaluation at
effects were observed during the experiments and, therefore, the cross non-integer coordinates. The Optical Flow elastic transform based on
section dimension was kept constant at 20×20 mm2. The top and the Finite Element Model (OF-FEM) kinematics method is used with a
bottom surfaces were ground to ensure parallel alignment with a mis- discretized pixel subset of 14×14 pixels. The OF-FEM method reduces
alignment error lower than 20 μm. In the case of the virgin material, measurement uncertainties by ensuring the continuity of the found
two conventional strain gauges were also glued to the lateral surfaces displacement field over the Region Of Interest (ROI, see Fig. 4b) [26].
for redundancy. The front surface was ground using P500, P1200, The strain was computed from the displacement field using an in-house
P2500 and P4000 grit paper (without water) and subsequently polished Python script. For the sake of simplicity, the study of the mechanical
using 3 μm diamond paste and 0.04 μm colloidal silica suspension be- behaviour was limited to the normal strain ( zz ). A good agreement was
fore etching the surface with Nital reagent for no longer than 5 s. This found between the DIC stress-strain curves and those obtained using
surface was sprayed with black paint to create the background before gauges (although the DIC results tended to exhibit slightly lower strain
applying an airbrushed white speckle pattern. More details about the values, >10%) and no significant out-of-plane effects were observed
uniaxial experiment can be found in Ref. [17]. [23].

3
I. Serrano-Munoz, et al. Wear 436–437 (2019) 202947

Fig. 3. (a) Image of the real-scale dynamometer set-up. (b) Localisation of the pad-inserted thermocouples located at 3 and 6 mm depth from the contact surface. (c)
Table summarising the program carried out during the braking test.

thermocouples (located on the disc and denoted in the figure by filled

bars) and the 3 mm inserted thermocouples (located on the pad and
denoted in the figure by empty bars). Some precautions must be taken
when analysing these measurements. First, the disc-rubbing thermo-
couple results correspond to the mean of the measurements at the inner,
medium and outer radius (see Fig. 3a). Apart from the fact that this
provides an average value over the radius and the circumference, the
main downside of these measurements is the fact is the contact con-
tinuity between the thermocouples and the disc is interrupted when the
disc surface is not flat. This situation is observed during high-energy
braking sequences, where thermo-mechanical effects usually lead to the
formation of hotspots and undulating distortions [8]. Also, these per-
turbations are more significant with increasing speeds. Therefore, the
values from the rubbing thermocouples have to be considered with
caution, particularly in the case of high speed braking, where the
thermocouples tend to measure values that are lower than the actual
surface temperatures.
Wear of the brake-pads induces a reduction in the distance between
the contact surface and the inserted thermocouples. This effect becomes
more intense with each subsequent braking series. Nevertheless, it is
considered that, for similar heating levels, the reduction of thickness
does not have a great impact on the temperature values because the
thermal diffusivity of the sintered metallic matrix is known to be high.
Fig. 5a shows that the pad temperature is close to that of the disc at
low speeds, with increasing difference at higher speeds. These mea-
sured differences could be due to the contact interruptions between the
Fig. 4. (a) Image of a 20 × 20 × 20 mm3 cube, where a speckle pattern has been rubbing thermocouples and the disc, but also due to the fact that the
applied to the front surface and two strain gauges have been glued to the lateral disc temperature measurements are the mean of measurements at dif-
surfaces. (b) Image of the experimental configuration of the compression test ferent radius lengths. The difference of measured temperature between
showing the ROI used for the computation of the vertical (Z) displacement field the inner and outer radii is, thus, not accounted for and could be
(expressed in units of pixel). comparatively high. Nevertheless, a thermal gap between the disc and
pad surfaces is often reported in braking applications. This is because
the contact is permanent for the pad but not for the disc, which can be
3. Results considered as an open thermodynamic system with alternating heating
and cooling cycles. Therefore, higher temperatures are usually expected
3.1. Braking experiment at the pad contact surface. Fig. 5a also shows increasing temperature
levels, which correspond to increasing dissipated energies. Moreover,
3.1.1. Thermal measurements similar heating levels are observed between series of different speed
Fig. 5a shows the mean temperature obtained from the rubbing and inertial mass when the dissipated energy and friction power have a

4
I. Serrano-Munoz, et al. Wear 436–437 (2019) 202947

Fig. 5. (a) Maximum temperatures in the disc (mean value of the rubbing thermocouple measurements) and pad (mean value of the measurements from the inserted
thermocouples at 3 mm) for the three braking series. (b) 3 mm inserted thermocouples measurements for the three braking series at maximum speed (V5). (c) 6 mm
inserted thermocouple measurements for the three braking series at maximum speed (V5). The legend for (b) and (c) is shown within the schematic brake-pad at the
top of the graphs.

Fig. 6. (a) Evolution of COF during braking for the three series. (b) COF average values (with error bars) for all braking series.

similar magnitude (e.g., S1V5, S2V4 and S3V3). destruction of one of the 3 mm inserted thermocouples (which occurred
This point is confirmed by the small temperature difference ( during S3V5, see bottom graph of Fig. 5b).
100 °C) observed between the recordings of 3 and 6 mm inserted ther- Fig. 5b shows the temperature recordings at maximum speed (V5)
mocouples (Fig. 5b and c). Note that the wear reduction led to the for the thermocouples located at 3 mm depth from the contact surface.

5
I. Serrano-Munoz, et al. Wear 436–437 (2019) 202947

The highest maximum temperature was reached during the S3 series (

1000 °C), which corresponds to the highest dissipated energy and fric-
tion power. The S3V5 results exhibit a plateau stage that can be related
to saturation. Also, significant temperature variations are observed
among the thermocouple locations for all series. These variations
probably correspond to macroscopic discontinuities at the contact
surface. It has been observed that the dynamic process of heating-di-
latation-wear induced by friction leads to losses of contact uniformity
for brake-pad surfaces that are as extensive as the ones studied in this
work [18,27]. Similar contact alterations are observed in the case of
6 mm inserted thermocouples (Fig. 5c), although to a lower extend.
Note that these levels of temperature, and corresponding thermo-
mechanical stress, are able to induce material changes such as phase
transformation or damage [19].

3.1.2. Friction
The instantaneous coefficient of friction ( µa ) and the mean friction
coefficient ( µm ), noted as COF hereafter, was calculated following the
UIC standard [28]. µa is calculated for any instant of braking by the
ratio of the tangential force (Ft ) to the normal contact (Fb , µa = Ft / Fb ).
The tangential force is calculated by the ratio between the torque and
the mean of the contact radius. The normal force is obtained from the
pneumatic actuator pressure. The evolution of the COF during all
braking sequences is shown in Fig. 6a. The fluctuations in friction
Fig. 7. (a) Image of the contact surface of the right half-pad after the braking
during braking are higher in S1 as well as in the low speed sequences of
program. The red dashed square indicates the extraction location of the worn
S2 and S3 series. Such steep growths in the COF at low speed (e.g., compression cube. (b) SEM (SE) image of the contact surface showing a com-
S1V1) are often observed when the evolution in friction mechanisms pacted 3rd body and some graphite particles (in black) on the left.
promotes adhesion effects [6]. The average COF in each braking se-
quence is shown in Fig. 6b, where two main tendencies can be ob-
15 MPa, and 20 MPa. For the sake of simplicity, only the stress-strain
served. First, a stabilizing trend is observed over the series, with S3
curves at 20 MPa are shown in Fig. 8b (in this figure the strains are
exhibiting the lowest differences in COF. This stabilizing trend over the
shown in negative values while stresses are kept positive).
series indicates that the 3rd body is evolving. Given the high level of
Remarkably, the stress-strain curves shown in Fig. 8b have strong
thermal loading, it is likely that during S3 the tribological circuit is
similarities with those reported in the literature for ceramics and rocks
strongly activated, with various wear mechanisms occurring simulta-
[29,30]. Also, the yield point at which non-linear behaviour similar to
neously (i.e, generation of wear sources from material damage and/or
plastic deformation was observed to first occur at 10 MPa in the virgin
adhesion, debris accumulation as well as ruptures in the 3rd body).
material. However, considerably micro-plasticity (due to the presence
Secondly, the friction level tends to be similar in sequences with
of graphite and ceramics inducing stress concentration in the metallic
equivalent levels of energy dissipation (e.g., S1V5, S2V4 and S3V3).
matrix) is expected to occur in the metallic matrix even at the lowest
stress levels.
3.1.3. Surface analysis
Fig. 7a shows the contact surface for the right half-pad at the end of
the braking program. This half-pad exhibits a macrocrack on the top- 3.2.1. Virgin material
left region which seems to be caused by the presence of a 3 mm depth The virgin stress-strain curve (shown in blue in Fig. 8b) exhibits four
thermocouple. The main part of the contact surface is covered by a main stages:
thick layer of 3rd body. In general, the examination of the contact At B1, the curve has an initial nearly perfect linear elasticity fol-
surface indicates that the tribological circuit was fully developed when lowed by yielding (B2 stage, with a zz = −0.27% ) induced by the
the program ended. The analysis at high magnification shows a com- first loading at the 15–20 MPa stress level.
pacted tribo-layer on the right of Fig. 7b, and some of the components
(mostly graphites) lying on the left, where there is also a dent indicating • B3 stage corresponds to the switching from loading to unloading and
the development of detachments. The red dashed square in Fig. 7a in- exhibits an apparent increase in stiffness.
dicates the location from where the compression cube was carved out. • At the end of unloading (B4), the stiffness is 8% lower than in the
In addition, note that the disc exhibits a non-uniform deposit of 3rd forward B5 loading.
body material which spreads over the entire contact surface. Moreover, • The forward loading within the hysteresis loop (B5) is also nearly
the infrared images confirm that the deposit heterogeneity on the disc is linear, with a slightly lower stiffness than that observed at B1.
induced by the occurrence of hotspots.
Note that the loading/unloading cycles exhibit a strain stays in
3.2. Uniaxial loading-unloading tests hysteresis loop behaviour ( zz = −0.33%) and the residual strain at the
minimum stress level during B is zz -0.2%.
Fig. 8a shows the uniaxial compression protocol followed during the The DIC zz strain map (computed for a B5 loop-loading branch
testing of the worn material. The virgin protocol is identical but the pre- between 1 and 20 MPa, Fig. 8c) indicates a heterogeneous macroscopic
load is established at min = 0.1 MPa instead of min = 1 MPa. The tests deformation. The lowest strains ( zz -1.2%) are concentrated within
start with a loading-unloading sequence at max = 3 MPa, (five cycles and around Graphite 2 particles [31], leading to the formation of de-
are applied, = 2 MPa) before complete release of the compression formation bands. In this case the bands are more pronounced at the
load. 2–3 min are necessary in-between loading/unloading cycles to bottom of the sample, but the examination of other samples of the same
change both the loading and the image recording parameters. The virgin material extracted from different locations over the brake-pad
maximum load is subsequently increased to max = 5 MPa, 10 MPa, indicates a random formation of the bands depending on the

6
I. Serrano-Munoz, et al. Wear 436–437 (2019) 202947

Fig. 8. (a) Graph showing the different stages of the uniaxial loading/unloading protocol (compression loads are shown in positive values). (b) Stress-strain curves
obtained from the DIC results for both virgin (in blue) and worn (in red) materials at 20 MPa stress level. (c) DIC map of the normal strain ( zz , expressed in % ) for the
virgin material. The map corresponds to a loading loop between 1 and 20 MPa (indicated by two red dots in (a)). (d) DIC map of the normal strain ( zz , expressed in % )
for the worn material. The top of the map corresponds to the contact surface. (For interpretation of the references to colour in this figure legend, the reader is referred
to the Web version of this article.)

positioning of Graphite 2 particles. The formation of deformation bands smaller than in the virgin material and the loading/unloading branches
has been reported in the literature for a cellular Al alloy [32] as well as do not follow the exact same paths between cycles; in other words, the
for a low-density porous copper material [33]. In addition, the effect of strain stays in hysteresis loop behaviour is less noticeable in the worn
strain localisation induced by the presence of graphite particles has material. The amount of yielding and residual strains ( zz -0.25%
been observed in cast iron materials [34,35], although in this case the and zz -0.2%, respectively) seems to be similar for both materials.
strain localisation occurs at the microscopic scale length. Note that the occurrence of the non-linear settling down phase (ob-
served in A1 and A5) is usually related to the presence of damage. In the
case of cracking media such as brittle rock, this phenomenon is linked
3.3. Worn material to the closure of microcracks [36].
The DIC zz strain map of the worn material (Fig. 8d) also indicates a
The worn stress-strain curve (shown in red in Fig. 8b) exhibits five heterogeneous macroscopic deformation with a marked strain band
main stages with three main differences compared to the virgin mate- occurring at 4 mm from the contact surface. In this case, the presence
rial: of such large deformation bands cannot be exclusively correlated to the
presence of Graphite 2 particles. The bottom region of this sample ex-
• At A1, the curve has an initial settling down phase followed by a hibits strain localisation occurring at/around Graphite 2 particles, si-
nearly linear region at higher load levels. milar to the one observed in the virgin material.
• Yielding is also observed to occur during the first loading at the
15–20 MPa stress level (A2 stage, with a zz = −0.23%). Note
that, during an actual braking situation, the mean normal load is not 3.3.1. Highly strained deformation band (HSDB)
expected to reach these high values (15–20 MPa), although it is The effects on the mechanical response induced by the presence of
likely that these values could occur locally due to the material this band has been studied in detail by comparing the DIC zz strain
heterogeneity. maps computed for loop-unloading branches at different stress levels
• A3 stage corresponds to the switching from loading to unloading (Fig. 9). An additional DIC computation is performed between the
and also exhibits a relatively increased stiffness. stages noted as 10 min and 2 s after the sample is fully unloaded
• At the end of the unloading (A4), the stiffness is similar to the (Fig. 9a). At 3-1 MPa (Fig. 9b), the mean strain unloading values are
settling down phase of the forward loading A1. close to 0.1%, although a small band showing 0.35% values is observed
• The forward loading within the hysteresis loop (A5) is non-linear. on the right middle of the image. At 5-1 MPa (Fig. 9c), the strain lo-
calisation only takes place within the region identified as HSDB. The
In this case, the general stiffness is lower and therefore higher va- highest strain values observed in this region are close to 1.0 %. At 10-
lues of strain are induced ( zz -0.53% is observed within the loading/ 1 MPa, 15-1 MPa and 20-1 MPa (Fig. 9d–f), the strain localisation is
unloading cycles). Moreover, the area within the hysteresis loop is progressively extended to the bottom region of the sample.

7
I. Serrano-Munoz, et al. Wear 436–437 (2019) 202947

Fig. 9. (a) Graph showing the different stages of the uniaxial loading/unloading protocol. The red dots indicate the images used for the computation of zz strain
maps: (b) at 3-1 MPa, (c) 5-1 MPa, (d) 10-1 MPa, (e) 15-1 MPa, (f) 20-1 MPa, and (g) between an image taken 10 min after sample unloading and one taken
immediately after sample unloading.

Interestingly, a time-dependent unloading effect is observed in Fig. 9g, 3.3.4. Microcracking

mainly occurring within the HSDB region. The strain values occurring The visual evaluation of the SEM images does not allow a thorough
in this region are close to 0.1%, and the mean zz calculated from the investigation of the microcrack density. Nevertheless, it is sufficient to
image yields a negligible unloading of 0.016%. Finally, note that the determine an increased content of microcracks within the middle layer
regions exhibiting the highest strain levels (for instance, red in Fig. 9f) (Fig. 11). No macrocrack is observed, although the possibility of a bulk
during the loop-unloading correspond to the regions with the lowest 3D linkage between microcraks must be considered. The content of
strain levels during the loop-loading (blue in Fig. 8d). These results microcracks within the top and bottom layers seems similar, although
indicate that the compression behaviour of the worn material is mainly this statement is conjectural. Note that the virgin material also exhibits
controlled by the HSDB region where the strain localisation is maximal. some microcracking after the uniaxial compression test. The orientation
of cracks seems random, as horizontal, oblique and vertical cracks can
3.3.2. Layered bulk behaviour be indistinctly observed. These microcracks tend to occur at the thin
The use of layered ROIs (Fig. 10a) allows the study of the local walls of metallic matrix laying between large Graphite 2 particles. Fi-
mechanical behaviour at three different depths (noted as top, middle nally, some out-of-plane features can be observed in (Fig. 11). These
and bottom). The stress-strain curves obtained from the local DIC out-of-plane features are also observed within the top and bottom layers
computations are shown in Fig. 10b. The top layer behaviour shows an although they are less noticeable.
initial and relatively small settling down phase followed by a nearly
linear region at higher load levels. No plastic deformation and very 3.3.5. Hardness
small residual strain ( zz -0.05%) are observed. The hysteresis can be Vickers hardness tests were performed using an instrumented in-
considered as non-existent. The behaviour of the middle layer is very dentation machine (with about 10 points per layer and up to 120 N of
similar to the worn curve shown in Fig. 8b but with higher strains multi-cycling). The reported values correspond to the Martens hard-
observed within the loading/unloading cycles ( zz -0.8% against zz ness, computed from the maximum displacement of the load-displace-
-0.42%). The hysteresis seems to be slightly increased. This result ment curves (see Table 2). The highest values (+30%) are observed in
additionally confirm the above conclusion about the HSDB region the top layer. The hardness of middle and bottom layers is similar to
controlling the macroscopic mechanical behaviour of the worn mate- that of the virgin material. These results are in good agreement with the
rial. By contrast, the behaviour of the bottom layer is quasi-linear and increased content of carbon observed in the Fe-phase of the top layer.
exhibits lower stiffness when compared to the virgin material. This The high dispersion of these measurements is related to the material
reduction in the stiffness could be explained by the development of heterogeneity, which gives a different response depending on which
microcracks. Further, the bottom stays in hysteresis loop behaviour is less phase/particles are being indented, as well as on the arrangement of
noticeable when compared to that of the virgin material. these phases/particles within the volume.

3.3.3. Microstructural changes 4. Discussion

The analysis of the microstructure indicates a carburization-like
effect occurring in the Fe-phase. In the top layer, the Fe-phase is 4.1. Evolution of the compression behaviour related to the microstructure
completely covered by pearlite (Fig. 10c-c’). In the middle layer, the
colonies of pearlite are homogeneously distributed throughout the Fe- Fig. 12 summarizes the main findings concerning the evolution of
phase (Fig. 10d). Finally, the microstructure in the bottom (Fig. 10e) the compression behaviour and microstructure. The virgin material
layer is similar to that observed in the virgin material (Fig. 1c and d). exhibits a complex mechanical behaviour where Graphite 2 particles,

8
I. Serrano-Munoz, et al. Wear 436–437 (2019) 202947

Fig. 10. (a) Light-optical image of the worn sample showing the ROI locations used in the DIC computations of the top, middle and bottom layers. (b) Stress-strain
curves of the top, middle and bottom layers at 20 MPa stress level. The virgin curve is added in blue for comparison purposes. (c) SE image and (c’) BSE image of the
microstructure within the top layer. (d) SE image of the microstructure within the middle Layer. (e) SE image of the microstructure within the bottom layer. See
Fig. 1c and d for comparison with the virgin material. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of
this article.)

Fig. 11. (a) SE image of the mesostructure within the middle layer. Charging effects are observed on some ceramic particles. (b) Detail of the microcracks. See Fig. 1b
for comparison with the virgin material.

9
I. Serrano-Munoz, et al. Wear 436–437 (2019) 202947

Table 2 As is the case in geological materials, the pronounced non-linearity

Martens hardness in GPa (standard deviation) of the virgin and worn materials. (Fig. 10b) exhibited by the middle layer can be related to a higher
The worn material is subdivided into the three regions identified in Fig. 10a. content of microcracks. Moreover, it has been reported [30] that an
Virgin Material Worn Material increased microcrack content leads to the increase of the area enclosed
within the hysteresis loop. Also, the reduced reversibility in the hys-
Top Middle Bottom teresis loop of the middle layer (i.e., the stays in hysteresis loop behaviour
is less noticeable) could be linked to enhanced crack closure effects
0.30(0.21) 0.46(0.31) 0.31(0.18) 0.32(0.20)
such as higher friction between the crack lips. On the other hand, it is
probable that the time-dependent recovery observed in Fig. 9g is con-
trolled by the intrinsic hysteretic behaviour of Graphite 2 particles ra-
which are softer than the matrix, induce strain localisation in the form ther than by crack closure effects. Even though we were not able to
of macroscopic deformation bands (Fig. 12a, shown in green and blue). confirm this point, the possibility of 3D macrocracks resulting from the
After the compression, some microcracking can be distinguish on the coalescence of microcracks should be contemplated (Fig. 12d). Note
SEM images (Fig. 12c). Additional compression tests, not shown here that the compression curve of the bottom layer (Fig. 10b) also suggests
for the sake of simplicity, indicate that the Graphites 2 particles control an increased microcrack density when compared to the virgin material.
the formation of hysteresis (the graphite material is known for having The combination of mechanical loading (shear and compression)
intrinsic hysteretic behaviour [37]). Nonetheless, there are other factors with constrained thermal expansions and microstructural transforma-
(e.g., energy dissipation induced by friction between Graphite 2 parti- tions could explain the higher microcrack density observed in the
cles and the metallic matrix walls) that might also contribute, in some middle layer. Nevertheless, the Fe-phase microstructural transforma-
extent, to the hysteresis loop. tion (i.e., higher content of C within the lattice structure combined with
The mechanical response of the worn material is even more complex austenitization) seems to be the key parameter controlling the increase
than that of the virgin material. The changes induced by the braking in microcrack density. This is because the depth of the middle layer
program led to the formation of a tri-layered-like material. Strain lo- mainly corresponds to the range of temperature (800-700 °C) where the
calisation is also observed, but there are variations depending on the Fe-phase starts to transition from ferrite to austenite (Fig. 12d). This
layer (Fig. 12b). The carbon content is higher in the top layer (Fig. 12d, transition would induce a boundary effect, which promotes the locali-
temperatures as high as 800–1000 °C). This is consistent with the fact sation of stresses. In any case, the elucidation of the mechanisms
that the austenitization occurring at 800–1000 °C enhances carbon leading to the increase of microcracks observed in the middle layer
diffusion (the solubility of C is greater in austenite compared to ferrite needs further research mostly concerning material science (e.g., the use
[38]). Thus, we consider that the mechanical changes observed in the of diffraction techniques for phase analysis).
top layer (increased Young's modulus and hardness) are mainly induced Finally, given that the virgin material is considerably porous (see
by increased carbon content. Furthermore, the lack of yielding observed Section 2.1) and its tendency to microcrack, the residual stresses in-
in the top layer (Fig. 10b) would be due to a carbon-induced increase in duced by Fe-phase transformation as well as the different dilatation
the elastic limit ( y ) of the Fe-phase. coefficient of components are considered to play a minor role in the

Fig. 12. (a) Schematic illustration of the strain localisation occurring in the virgin material (b) and the worn material when tested in a loop-loading branch between 1
and 20 MPa. (c) Schematic illustration of the microcracking observed in the virgin material at the end of the compression test. (d) Schematic illustration of the
microcracking and level of carbon content observed in the worn material at the end of the compression test. The arrow indicates that the carbon content in the Fe-
phase decreases along the depth. The temperature gradient is an estimation based on the thermocouple results.

10
I. Serrano-Munoz, et al. Wear 436–437 (2019) 202947

evolution of the compressive behaviour. Ideally, the evolution of material properties should contribute to the
further stabilization of the COF and 3rd body. Further investigation on
4.2. Evolution of the friction behaviour related to the contact surface and the subject requires the design of highly instrumented dynamometers
sub-surface material accounting for the monitoring of other parameters (e.g., vibrations), the
implementation of multi-loading tests (e.g., to investigate the effect of
At low temperature (S1V1-S1V3, S2V1-S2V2, S3V1-S3V2, see Fig. 5 shear loads), as well as a detailed characterization of the 3rd body
and Fig. 6b) the COF values are higher and more variable when com- composition (e.g., the use of Focused Ion Beam (FIB-SEM) to map the
pared to those occurring at high temperature (S1V4-S1V5, S2V3-S2V5, chemical composition and microstructure over the contact surface).
S3V3-S3V5). These results correlate well with the fact that the friction
tends to be more abrasive at low temperature and more adhesive at
high temperature [39]. They also provide a demonstration of the role of 5. Conclusions
temperature level in COF stabilization. For instance, a significant sta-
bilization can be observed during the S1V4 braking sequence. During The changes in bulk properties induced by a braking program were
this sequence, a temperature higher than 700 °C is reached for the first investigated using a simplified semi-metallic friction material. The use of
time at the contact surface (more than 600 °C is registered by the inserted thermocouples enabled the monitoring of the thermal loading
thermocouples at 3 mm, Fig. 6b). As discussed in Section 4.1, ferrite to while the characterization of the compressive and microstructural
austenite transformation starts at temperatures above 700 °C. At this properties was carried out using compression tests coupled with DIC
temperature, the copper is able to endure higher levels of plastic and SEM analysis. The following conclusions can be drawn:
yielding. It is generally accepted that the increase of copper ductility
contributes to the development of a tribolayer that facilitates the sta- 1. The thermal loading plays a dominant role in the evolution of the
bilization of the COF [40]. Thus, opposing mechanisms develop during friction behaviour and material properties.
the braking program: the hardness of the sub-surface material increases 2. The stiffness of the worn material is lower, compared to that of the
due to carbon diffusion while the copper contributes to an increase in virgin material, and the evolution in the material properties leads to
the ductility of the tribolayer. The effect that an increasing sub-surface a compression behaviour similar to that of a tri-layered material.
hardness has on the ductility of the tribolayer has yet to be determined. 3. The changes in the compression behaviour of the worn material are
However, an increase in the brake-pad hardness usually leads to an related to the microstructure:
increase in the wear rates of the counter-face (i.e., the disc). This aspect (a) The increased microcrack density at the central portion of the
could not be investigated in the present study because it would require brake-pad thickness leads to the formation of a deformation
long-term tests, but the results of endurance tests conducted by the band containing, in turn, the maximum (during unloading) or
industrial partner confirms this trend. minimum (during loading) strain values.
(b) An increase in the carbon content (in the form of perlite co-
4.3. Limitations and open questions lonies) is observed in the Fe-phase. The increase in carbon
content is particularly important in the top layer, which exhibits
The combination of thermal monitoring and bulk properties character- the highest stiffness and hardness.
isation has led to a comprehensive understanding of the mechanisms in- 4. The results indicate that the material evolution has deleterious ef-
volved in the evolution of the compression behaviour. Nevertheless, so far fects on the contact quality (spatial variation of the mechanical
we can only conjecture about the influence of the material evolution on the behaviour over the brake-pad cross section), durability (localisation
braking performance. Some of the most relevant issues concern: of damage favouring material fragmentation) and tribology (spatial
variation of the 3rd body composition).
4.3.1. Contact uniformity
The results shown in Fig. 5 indicate the development of different Even if the tribological aspects are to be prioritised, the bulk be-
thermal gradients depending on the pad location. A spatially hetero- haviour should also be considered when developing new formulations.
geneous mechanical behaviour can be therefore expected over the Understanding the interactions between components at the micro-
contact surface. The spatial variation of the mechanical properties structural level gives valuable insight into possible ways to improve the
usually leads to a reduced contact uniformity and subsequent thermal material formulation. For example, the results presented in this work
localisations. In this work, however, the mechanical heterogeneity indicate that the increase in carbon content has negative effects. Thus,
could be neutralized by the reduced stiffness of the middle layer. new formulations ensuring reduction in carbon diffusion can be en-
visaged. Hence, further investigation of the evolution of bulk material
4.3.2. Durability properties with special emphasis on the quantification of the sub-
The damage localisation observed in the middle layer can be con- sequent effects on the braking performance is recommended.
sidered as highly detrimental to the brake-pad durability. It would be
interesting to investigate whether this damage could eventually lead to
the final failure of the brake-pad. Acknowledgments

4.3.3. Vibrations The present investigation was supported by the International

A reduction in vibrations could be expected given the decrease in Campus on Safety and Inter-modality in Transportation, France; the
stiffness (Fig. 8b) and the dominant role of Graphite 2 particles in the Hauts-de-France's Regional Delegation for Research and Technology,
middle layer (Fig. 9). However, a deleterious effect resulting from the France; the French Ministry of Higher Education and Research; the
increase in stiffness and hardness of the top layer should be considered. French National Research Agency (ANR COMATCO); and the French
National Center for Scientific Research (CNRS). The authors gratefully
4.3.4. Tribology acknowledge these institutions for their support. FLERTEX SINTER
The spatial heterogeneity of the thermal loading most likely affects Company is acknowledged for supplying the studied material and
the 3rd body composition. As is the case for contact uniformity, one lending the dynamometer where the experiments were performed.
could argue that a spatially heterogeneous 3rd body can hamper the Special gratitude is expressed to Tyler Oesch for his thoughtful advice
stabilization of COF and even increase wear rates by lowering the co- and guidance.
hesiveness of the layer.

11
I. Serrano-Munoz, et al. Wear 436–437 (2019) 202947

References presence of wear, Wear 312 (1) (2014) 11–20 https://doi.org/10.1016/j.wear.

2014.01.008.
[21] A. Heussaff, L. Dubar, T. Tison, M. Watremez, R. Nunes, A methodology for the
[1] P.J. Blau, Compositions, Functions, and Testing of Friction Brake Materials and modelling of the variability of brake lining surfaces, Wear 289 (2012) 145–159
Their Additives, Tech. rep. Oak Ridge National Lab., TN (US), 2001, http://www. https://doi.org/10.1016/j.wear.2012.04.002.
osti.gov/scitech/biblio/788356. [22] V. Magnier, D.N. Ramasami, J. Brunel, P. Dufrénoy, T. Chancelier, History effect on
[2] W. Österle, A. Dmitriev, Functionality of conventional brake friction materials – squeal with a mesoscopic approach to friction materials, Tribol. Int. 115 (2017)
perceptions from findings observed at different length scales, Wear 271 (9–10) 600–607 https://doi.org/10.1016/j.triboint.2017.06.031.
(2011) 2198–2207 18th International Conference on Wear of Materials https://doi. [23] M. Sutton, J. Yan, V. Tiwari, H. Schreier, J. Orteu, The effect of out-of-plane motion
org/10.1016/j.wear.2010.11.035. on 2D and 3D digital image correlation measurements, Optic Laser. Eng. 46 (10)
[3] M. Eriksson, F. Bergman, S. Jacobson, On the nature of tribological contact in au- (2008) 746–757 https://doi.org/10.1016/j.optlaseng.2008.05.005.
tomotive brakes, Wear 252 (1) (2002) 26–36 https://doi.org/10.1016/S0043- [24] D. Paganin, S.C. Mayo, T.E. Gureyev, P.R. Miller, S.W. Wilkins, Simultaneous phase
1648(01)00849-3. and amplitude extraction from a single defocused image of a homogeneous object,
[4] W. Österle, M. Griepentrog, T. Gross, I. Urban, Chemical and microstructural J. Microsc. 206 (1) (2002) 33–40, https://doi.org/10.1046/j.1365-2818.2002.
changes induced by friction and wear of brakes, Wear 251 (1) (2001) 1469–1476 01010.x arXiv:https://onlinelibrary.wiley.com/doi/pdf/10.1046/j.1365-2818.
13th International Conference on Wear of Materials https://doi.org/10.1016/ 2002.01010.x.
S0043-1648(01)00785-2. [25] U. Lille:, Yadics Software. URL http://yadics.univ-lille1.fr/wordpress/.
[5] A.-L. Cristol-Bulthé, Y. Desplanques, G. Degallaix, Y. Berthier, Mechanical and [26] F. Hild, S. Roux, Digital Image Correlation. Optical Methods for Solid Mechanics: a
chemical investigation of the temperature influence on the tribological mechanisms Full-Field Approach, Wiley-VCH, Berlin (Germany), 2012http://www.wiley-vch.
occurring in omc/cast iron friction contact, Wear 264 (9) (2008) 815–825 papers de/de/fachgebiete/ingenieurwesen/optical-methods-for-solid-mechanics-978-3-
presented at the 12th Nordic Symposium on Tribology, NORDTRIB 2006 https:// 527-41111-5.
doi.org/10.1016/j.wear.2006.12.080. [27] R. Mann, Experiments and Thermomechanical Modelling of Braking Application &
[6] Y. Desplanques, O. Roussette, G. Degallaix, R. Copin, Y. Berthier, Analysis of tri- Friction Material Characterization with Loading History Effect, Ph.D. thesis
bological behaviour of pad–disc contact in railway braking: Part 1. laboratory test University of Lille, 2017.
development, compromises between actual and simulated tribological triplets, [28] Frein - frein à disques et leur utilisation - conditions générales pour la certification
Wear 262 (5) (2007) 582–591 https://doi.org/10.1016/j.wear.2006.07.004. de garnitures de frein .URL https://uic.org/IMG/pdf/f541x3_anl_i_07-18.pdf.
[7] Y. Xiao, Z. Zhang, P. Yao, K. Fan, H. Zhou, T. Gong, L. Zhao, M. Deng, Mechanical [29] G. Bruno, M. Kachanov, Microstructure–property connections for porous ceramics:
and tribological behaviors of copper metal matrix composites for brake pads used in the possibilities offered by micromechanics, J. Am. Ceram. Soc. 99 (12) (2016)
high-speed trains, Tribol. Int. 119 (2018) 585–592 https://doi.org/10.1016/j. 3829–3852, https://doi.org/10.1111/jace.14624 arXiv:https://onlinelibrary.wiley.
triboint.2017.11.038. com/doi/pdf/10.1111/jace.14624.
[8] S. Panier, P. Dufrénoy, D. Weichert, An experimental investigation of hot spots in [30] E. David, N. Brantut, A. Schubnel, R. Zimmerman, Sliding crack model for non-
railway disc brakes, Wear 256 (7) (2004) 764–773 https://doi.org/10.1016/S0043- linearity and hysteresis in the uniaxial stress–strain curve of rock, Int. J. Rock Mech.
1648(03)00459-9. Min. Sci. 52 (2012) 9–17 https://doi.org/10.1016/j.ijrmms.2012.02.001.
[9] M. Kumar, X. Boidin, Y. Desplanques, J. Bijwe, Influence of various metallic fillers [31] Ruddy Mann, Vincent Magnier, Itziar Serrano-Munoz, Jean-Francois Brunel,
in friction materials on hot-spot appearance during stop braking, Wear 270 (5) Florent Brunel, Philippe Dufrenoy, Michele Henrion, Non-linear mechanical beha-
(2011) 371–381 https://doi.org/10.1016/j.wear.2010.11.009. vior of a sintered material for braking application using digital image correlation,
[10] D. Kolluri, X. Boidin, Y. Desplanques, G. Degallaix, A. Ghosh, M. Kumar, J. Bijwe, Mech. Ind. 18 (6) (2017) 601, https://doi.org/10.1051/meca/2017031 https://doi.
Effect of Natural Graphite Particle size in friction materials on thermal localisation org/10.1051/meca/2017031.
phenomenon during stop-braking, Wear 268 (11–12) (2010) 1472–1482 https:// [32] H. Bart-Smith, A.-F. Bastawros, D. Mumm, A. Evans, D. Sypeck, H. Wadley,
doi.org/10.1016/j.wear.2010.02.024. Compressive deformation and yielding mechanisms in cellular Al alloys determined
[11] T.J. Mackin, S.C. Noe, K. Ball, B. Bedell, D. Bim-Merle, M. Bingaman, D. Bomleny, using X-ray tomography and surface strain mapping, Acta Mater. 46 (10) (1998)
G. Chemlir, D. Clayton, H. Evans, R. Gau, J. Hart, J. Karney, B. Kiple, R. Kaluga, 3583–3592 https://doi.org/10.1016/S1359-6454(98)00025-1.
P. Kung, A. Law, D. Lim, R. Merema, B. Miller, T. Miller, T. Nielson, T. O'Shea, [33] M. Hakamada, Y. Asao, T. Kuromura, Y. Chen, H. Kusuda, M. Mabuchi, Density
M. Olson, H. Padilla, B. Penner, C. Penny, R. Peterson, V. Polidoro, A. Raghu, dependence of the compressive properties of porous copper over a wide density
B. Resor, B. Robinson, D. Schambach, B. Snyder, E. Tom, R. Tschantz, B. Walker, range, Acta Mater. 55 (7) (2007) 2291–2299 https://doi.org/10.1016/j.actamat.
K. Wasielewski, T. Webb, S. Wise, R. Yang, R. Zimmerman, Thermal cracking in disc 2006.11.024.
brakes, Eng. Fail. Anal. 9 (1) (2002) 63–76 https://doi.org/10.1016/S1350- [34] T. Sjögren, P.E. Persson, P. Vomacka, Analysing the deformation behaviour of
6307(00)00037-6. compacted graphite cast irons using digital image correlation techniques, Science
[12] F. Bagnoli, F. Dolce, M. Bernabei, Thermal fatigue cracks of fire fighting vehicles and Processing of Cast Iron IX, Vol. 457 of Key Engineering Materials, Trans Tech
gray iron brake discs, Eng. Fail. Anal. 16 (1) (2009) 152–163 https://doi.org/10. Publications, 2011, pp. 470–475 10.4028/www.scientific.net/KEM.457.470.
1016/j.engfailanal.2008.01.009. [35] J. Pina, S. Shafqat, V. Kouznetsova, J. Hoefnagels, M. Geers, Microstructural study
[13] M. Boniardi, F. D'Errico, C. Tagliabue, G. Gotti, G. Perricone, Failure analysis of a of the mechanical response of compacted graphite iron: an experimental and nu-
motorcycle brake disc, Eng. Fail. Anal. 13 (6) (2006) 933–945 https://doi.org/10. merical approach, Mater. Sci. Eng., A 658 (2016) 439–449 https://doi.org/10.
1016/j.engfailanal.2005.04.012. 1016/j.msea.2016.02.017.
[14] Z. Yang, J. Han, W. Li, Z. Li, L. Pan, X. Shi, Analyzing the mechanisms of fatigue [36] C.H. Scholz, Experimental study of the fracturing process in brittle rock, J. Geophys.
crack initiation and propagation in CRH EMU brake discs, Eng. Fail. Anal. 34 (2013) Res. 73 (4) (1968) 1447–1454, https://doi.org/10.1029/JB073i004p01447
121–128 https://doi.org/10.1016/j.engfailanal.2013.07.004. arXiv:https://agupubs.onlinelibrary.wiley.com/doi/pdf/10.1029/
[15] H. Kasem, J. Brunel, P. Dufrénoy, M. Siroux, B. Desmet, Thermal levels and sub- JB073i004p01447.
surface damage induced by the occurrence of hot spots during high-energy braking, [37] M. Barsoum, A. Murugaiah, S. Kalidindi, T. Zhen, Y. Gogotsi, Kink bands, nonlinear
Wear 270 (5) (2011) 355–364 https://doi.org/10.1016/j.wear.2010.11.007. elasticity and nanoindentations in graphite, Carbon 42 (8) (2004) 1435–1445
[16] Z. Li, J. Han, Z. Yang, W. Li, Analyzing the mechanisms of thermal fatigue and https://doi.org/10.1016/j.carbon.2003.12.090.
phase change of steel used in brake discs, Eng. Fail. Anal. 57 (2015) 202–218 [38] D.E. Jiang, E.A. Carter, Carbon dissolution and diffusion in ferrite and austenite
https://doi.org/10.1016/j.engfailanal.2015.07.002. from first principles, Phys. Rev. B 67 (2003) 214103, https://doi.org/10.1103/
[17] R. Mann, V. Magnier, J.-F. Brunel, F. Brunel, P. Dufrénoy, M. Henrion, Relation PhysRevB.67.214103 https://link.aps.org/doi/10.1103/PhysRevB.67.214103.
between mechanical behavior and microstructure of a sintered material for braking [39] Y. Desplanques, G. Degallaix, R. Copin, Y. Berthier, A tribometer for the study of
application, Wear 386–387 (2017) 1–16 https://doi.org/10.1016/j.wear.2017.05. materials under railway braking conditions, in: G. Dalmaz, A. Lubrecht, D. Dowson,
013. M. Priest (Eds.), Tribology Research: from Model Experiment to Industrial Problem,
[18] P. Dufrénoy, Two-/three-dimensional hybrid model of the thermomechanical be- Vol. 39 of Tribology Series, Elsevier, 2001, pp. 381–391 https://doi.org/10.1016/
haviour of disc brakes, Proc. Inst. Mech. Eng. - Part F J. Rail Rapid Transit 218 (1) S0167-8922(01)80123-1.
(2004) 17–30, https://doi.org/10.1243/095440904322804402 arXiv:https://doi. [40] M. Kumar, J. Bijwe, Role of different metallic fillers in non-asbestos organic (nao)
org/10.1243/095440904322804402. friction composites for controlling sensitivity of coefficient of friction to load and
[19] K. Lee, J.R. Barber, Frictionally excited thermoelastic instability in automotive disk speed, Tribol. Int. 43 (5–6) (2010) 965–974 special Issue on Second International
brakes, J. Tribol. 115 (4) (1993) 607–614, https://doi.org/10.1115/1.2921683 Conference on Advanced Tribology (iCAT2008) https://doi.org/10.1016/j.triboint.
https://doi.org/10.1115/1.2921683. 2009.12.062.
[20] M. Graf, G.-P. Ostermeyer, Efficient computation of thermoelastic instabilities in the

12