Journal of Manufacturing Processes 26 (2017) 446458

Contents lists available at ScienceDirect

Journal of Manufacturing Processes

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

Finite element simulation and analysis of serrated chip formation

during highspeed machining of AA7075T651 alloy
Walid Jomaa a,b, , Oussama Mechri a , Julie Lvesque a , Victor Songmene b,c ,
Philippe Bocher b,c , Augustin Gakwaya a,c
a
Department of Mechanical Engineering, Universit Laval, 1065 Avenue de la Mdecine, Qubec, G1 V 0A6, Canada
b
Department of Mechanical Engineering, cole de Technologie Suprieure, 1100 Rue Notre-Dame O., Montreal, H3C 1K3, Canada
c
Aluminium Research Centre-REGAL, Universit Laval, 1065, avenue de la Mdecine, Qubec, Qubec, G1V 0A6, Canada

a r t i c l e i n f o a b s t r a c t

Article history: Highspeed machining (HSM) is widely used in the manufacturing of monolithic aluminum components
Received 25 July 2016 for automotive and aeronautical industries. However, previous research studies on HSM of high strength
Received in revised form 24 January 2017 aluminum alloys have shown that serrated and/or elemental chips can form at critical cutting conditions,
Accepted 16 February 2017
impacting the machining stability and nal parts quality. Hence, understanding the physical mechanisms
governing the chip serration is essential to improve HSM part quality especially when machining high
Keywords:
strength aluminum alloys. In the present work, this was achieved by developing a 2D nite element mod-
Finite element analysis
elling (FEM), based on a lagrangian approach, for simulating and analysing the serrated chip formation
High speed machining
AA7075T651 alloy
during HSM of the AA7075-T651 alloy. The FEM was developed using Abaqus/Explicit v6.13 software. The
Serrated chip JohnsonCook (JC) constitutive equation combined with a damage criterion implemented into Abaqus
Adiabatic shearing was used to account for the shear localization during the serrated chip formation. The proposed nite
Shear band spacing element model was validated using experimental data obtained upon high speed orthogonal machining.
The results showed that the serrated chip morphology was accurately predicted over a large range of cut-
ting speed. In particular, the nite element model captured properly the fact that the chip segmentation
intensity increases with cutting speed. Furthermore, physical phenomena governing the serrated chip
formation were highlighted and discussed in depth using nite element numerical data and an analytical
modelling of chip serration.
2017 The Society of Manufacturing Engineers. Published by Elsevier Ltd. All rights reserved.

1. Introduction the machined parts. Thus, to enhance the HSM of aeronautical

Under specic conditions, the machining of metallic materials
often generates serrated chip type, also known as segmented/saw-
toothed chips. There are two cases in which the serrated chip
formation can take place. The rst case happens when hardtocut
materials such as hardened steels, super-alloys, and titanium alloys
are machined at relatively low cutting speeds (less than 70 m/min)
[1]. The second case is when ductile materials are machined with
cutting speed higher than a critical value [2]. Research studies [35]
raised evidence of such serrated chip formation during the HSM
of high strength aluminum alloys, such as 2xxx and 7xxx series.
The serrated chip formation not only impairs the stability of the
machining processes, but can also affects the surface integrity of
Corresponding author at: Universit Laval, Dpartement Gnie Mcanique,
Pavillon Adrien-Pouliot, 1065 Avenue de la Mdecine, Qubec, G1V 0A6, QC, Canada.
E-mail address: walid.jomaa.1@etsmtl.net (W. Jomaa).
aluminum parts, it is imperatively required to understand the phys-
ical mechanisms governing the serrated chip formation in order to
guarantee stable machining conditions.
Various theories and models were developed in order to under-
stand the serrated chip formation origins. On one hand, according
to Astakhov [6], segmented chip formation is attributed to high
variations in stress and plastic deformation, and to the resulting
temperature rise in the cutting zone. Later, Liyao et al. [7] stated that
segmented chip formation was mainly attributed to adiabatic shear
instability/catastrophic strain localization. On the other hand, Vyas
and Shaw [8] suggested that the fracture/crack propagation mech-
anism is at the root of the chip generation. The location where the
crack may initiate is also a point of contention. For Poulachon et al.
[9], the crack initiates on the free surface of the work material and
propagates to the tool tip. However, other researchers [2,6] argued
that the crack actually occurs at the tool tip and propagates partway
to the free workpiece surface.

http://dx.doi.org/10.1016/j.jmapro.2017.02.015
1526-6125/ 2017 The Society of Manufacturing Engineers. Published by Elsevier Ltd. All rights reserved.
 W. Jomaa et al. / Journal of Manufacturing Processes 26 (2017) 446458 447

Most of FEM based studies on serrated/discontinuous chip 2.1.1. Mechanical analysis

formation have mainly focused on steels and titanium alloys During the mechanical analysis step, the displacement, the
[5,1015]. However, only few studies [35] have concerns with strain, and the stress elds are calculated at every integration point
aluminum alloys. Mabrouki et al. [3] studied numerically and of an element in the model. These elds satisfy the linear momen-
experimentally the dry cutting of aluminium alloy A2024-T351, tum equation written at any point of the continuum medium as the
but only the cutting forces and chip segmentation frequencies were local conservation equation:
used to validate the developed FEM model. Other chip segmenta-
tion characteristics such as morphology features (shear band and + fd = u
div() (1)
pick/valley heights) were not considered. Later, Atlati et al. [4] have where  is the material density, u is the acceleration,  is the Cauchy
introduced a segmentation intensity ratio (SIR), which is dened stress tensor and fd is the body force.
as the ratio of the plastic deformation into the shear band to the The weak variational form associated with the equation Eq. (1)
one found outside the shear band. However, this ratio does not is derived from the principle of virtual work as follows:
describe properly the chip segmentation during the machining of     
A2024 alloy as two extreme cutting speeds tested (60 m/min and  : DdV + fd udV + tudS+ tC udS uudV = 0
1000 m/min) are associated with the same SIR but with very dif- V V SF SC V
ferent segmentation features and frequencies so that limiting its (2)
applicability to a large speed range. More recently, Ye et al. [5] pro-
where V denotes the volume occupied by the body, SF and SC the
posed an empirical model to dene the critical cutting speed for the
surfaces of the volume V where the surface forces t and the contact
onset of serrated chip ow for the high speed machining of ductile
surface forces tC are applied. u is the virtual velocity, and D is the
and hard-to-cut metallic materials. Their model needs extensive
virtual strain rate. The Eq. (2) is strongly nonlinear (as the stress
numerical sensitivity analysis of the effects of material properties
and strain depend on temperature and hardening parameters) and
and cutting parameters on the critical cutting speed. Among the
is numerically solved using the Newton-Raphsons method.
materials tested was the non-heat treated aluminum 7075 alloy.
However, the effects of cutting speed on the adiabatic shear banding
2.1.2. Thermal analysis
mechanism and state variables evolution have been neglected.
The steadystate thermal energy governing the orthogonal
In view of the foregoing literature review, the present study
machining process is derived from the principle of conservation
aims to develop a nite element model for simulating and analysing
of energy, which for mainly heat conduction dominated problem,
the serrated chip formation during the HSM of age hardened alu-
is as written below:
minum alloy AA7075-T651. The serrated chip mechanisms will be
discussed based on the evolution of the critical state variables and e =  : divq +  (3)
the thermo-mechanical loading calculated numerically using the
proposed nite element model. This was done by using (a) a JC where e denotes the internal energy variation,  the density of
constitutive equation combined with a damage evolution criterion
internal heat production,  : is the mechanical power dissipated
to describe the material behaviour; (b) a Lagrangian formulation to
into heat, divq is the heat dissipated by conduction. Assuming
simulate the serrated chip formation; and (c) adequate managing
Fourier law of heat conduction to apply:
of the element ditortion issues using a distortion control function

with the proper hourglass mode in ABAQUS/Explicit v6.13 software.
q = k.grad(T ) (4)
In addition, the accuracy of the proposed nite element model will
be conrmed by comparing the obtained numerical data with the Applying the entropy inequality and taking account of damage
analytical solution of the shear band spacing parameter. dissipated energy, one can show that the nal energy equation for
coupled thermo-visco-elasto-plastic behaviour with damage can
be written as [16]:
2. FEM of serrated chip formation 
div(K grad(T )) +  +  : p X :
Rr + Y D Cv T
In this paper, as a Lagrangian formulation is adopted, the  
mesh follows the material during deformation and there is no  X R Y
+T : e + +
: : r + : D = 0 (5)
need for a predened chip form. The workpiece is modeled as T T T T
an elastoplastic material with damage, whereas the tool is mod-
assuming
eled as a rigid body. This approach will allow as to predict the
serrated chip geometry and to understand the shear localization Rpl =  : p X :
Rr + Y D
phenomenon accompanying the serrated chip formation of the
AA7075-T651 alloy.  X R Y
+T [ : e + +
: : r + : D] (6)
T T T T

2.1. Governing equations of the coupled thermomechanical
analysis
In metal cutting, the chip formation process involves large
plastic deformation at high temperature and strain rate, where
the stress and temperature elds mutually inuence each other.
Hence, the problem is solved by applying a fully coupled
thermomechanical simulation algorithm. For simplicity, the gov-
where  : p represent the mechanical power induced by plastic
represents the isotropic
deformation and dissipated into heat, X :
hardening, Rr the kinematic hardening, Y D the isotropic ductile
damage, and e the elastic strain rate.
The nal heat equation, including the mechanical dissipation in
viscoplastic ow, hardening, and damage can be written as [16]


erning equations will be briey presented in the following Cv T = div(K grad(T )) +  + Rpl (7)
sections. For more details about the nite element method applied where T is the temperature (in Kelvin), K is the thermal conductiv-
to thermo-elasto-visco-plastic material behaviour with damage, ity, Cv is the thermal capacity,  is the internal heat source, and Rpl
readers can refer to [16]. is the internal heat generated by the mechanical load.
448 W. Jomaa et al. / Journal of Manufacturing Processes 26 (2017) 446458

Table 1
JC material constants for the AA7075-T651 alloy [22,23].

Plasticity parameters A [MPa] B [MPa] n C m 0 [s1 ] T0 [C] Tm [C]

527 575 0.72 0.017 1.61 1 25 635
Failure parameters D1 D2 D3 D4 D5
0.110 0.572 3.446 0.016 1.099

The weak variational form associated with the differential equa- By applying the central difference integration rule, the velocity
tion Eq. (7) is dened as follows: u and the acceleration u are approximated as follows [19]:
  
ti + ti+1
T
T Cv T dV T
T Rpl dV T qf dS+
ui+1/2 = ui1/2 + ui
V V ST
2
   (12)


T,iT (KT,i )dV T
T dV + T
T qdS = 0 (8)
ui+1 = ui+1 + ti+1 ui+1/2
V V Sq
where the superscript i refers to the increment number.
where T denotes the virtual temperature eld satisfying the ther- Compared to the implicit scheme, the dynamic explicit scheme
mal boundary conditions and q is the applied heat ux. qf is an allows a signicant reduction in the computation simulation time.
unknown heat ux corresponding to a known temperature at the However, this is achieved at the price of the model stability where
surface boundary ST . The temperature eld is integrated using the the time increments are very small, making the integration scheme
Galerkins method [17]. conditionally stable in time. The critical time-step for stability can
be approximated by the following formula:
2.1.3. Thermo-mechanical coupling
The weak variational forms associated with the mechanical Lmin
t = (13)
and thermal equilibrium equations (Eqs. (2) and (8)) constitute s
the system of equations dening the coupled thermomechanical where Lmin is the smallest representative element size in the nite
problem. On one hand, these equations are strongly coupled as the element model and s is the dilatational wave speed in the work
stress tensor ,
in the mechanical equilibrium equation, depends material and given by:
on the temperature T, via the constitutive equation (Eq. (15)). On
the other hand, the total heat ux q is the sum of the heat uxes due E
s = (14)
to the mechanical inelastic work (qp ) and the friction work at the 
tool/work material contacts (qf ). qp and qf are dened as follows: where E denotes the Young modulus and  the density of the work
material.
qp = p  : p (9)

and 2.3. Workpiece material modeling

qf = ff f f
p (10) The thermomechanical behaviour of the AA7075-T651 alloy
where p denotes a factor dening the amount of the plastic work is modeled using the JC constitutive equation [20]. This equation
describes the ow stress with a multiplicative formula of strain,
converted to heat (equal to 0.9 for metal), p the plastic strain rate,
strain rate, and temperature as follows:
f the frictional work converted into heat (assumed equal to 1), f
the friction stress given by the equation Eq. (21), ff the fraction of p T T0 m
 = [A + B(p )n ][1 + C ln( )][1 ( ) ] (15)
the thermal energy conducted into the chip. It the present study, 0 Tm T0
ff = 0.63 based on previous work [17].
where denotes the equivalent ow stress, p the equivalent plastic
Conclusively, in metal cutting process, the thermomechanical
strain, p the equivalent strain rate, 0 the reference strain rate, T the
coupling system is strongly non-linear, involving large displace-
temperature of the work material, Tmelt the melting temperature, T0
ment, large strain, and large rotation. Hence, this problem can be
the room temperature. The material constants can be interpreted as
resolved only with advanced numerical methods such as the Finite
follows: A is the yield strength coefcient, B the hardening modulus,
element method. In Abaqus, this is done by the linearization of the
C the strain rate sensitivity coefcient, n the hardening coefcient,
equations Eqs. (2) and (8) based on the Newton-Raphsons method.
and m the thermal softening coefcient. The values of the JC mate-
The thermal and the mechanical boundary conditions needed for
rial constants for the 7075-T651 alloy are given in Table 1.
this analysis will be presented later in Section 2.5.
The FEM simulation of the serrated chip formation is performed
by implementing a damage model which can describe the material
2.2. Numerical time integration scheme
behaviour when damage occurs. In the present study, the JC failure
model, which is based on the equivalent plastic strain at failure f ,
In the present work, the dynamic explicit scheme was used
is used as a criterion for damage initiation. f is dened as follows:
as the time integration procedure. The stresses and strains are
predicted by solving an uncoupled equation system based on infor- p p T T0
f = [D1 + D2 exp(D3 )][1 + D4 ln( )][1 D5 ( )] (16)
mation from the previous time steps. At each node, the weak form of  0 Tmelt T0
the mechanical equilibrium equations can be discretized as follows
where  p denotes the pressure stress and  the von Mises stress,
[18]:
D1 to D5 damage constants (see Table 1).
M un + Fnint (un ) = Fnext (tn ) (11) The damage initiation threshold is modelled in Abaqus/Explicit
v6.13 [21] software according to a cumulative damage law:
where un and un are, respectively, the nodal acceleration and dis-

placement at time tn , M is the mass matrice, and Fnint and Fnext are D= (17)
the internal and external forces at the node n. f
 W. Jomaa et al. / Journal of Manufacturing Processes 26 (2017) 446458 449

Fig. 1. Geometry of the model and boundary conditions.

where is the increment of the equivalent plastic strain. an empirical models developed based on experimental data from
In FEM involving damage phenomenon, once the damage ini- [24]. The developed limit model is given as:
tiates (D = 1), the ow stress will be no longer solely governed by
the constitutive equation, however, it will be affected by the dam- limit = 252 0.016V + 767.06t + 11.20
2588.42t 2 10.05t
2.36
2 (22)
age evolution and a strong mesh dependency of the state variables
occurs due to severe strain localization [3]. In practice, there are and the  model is given as:
different ways to describe the evolution of the ow stress after the
 = 1.09 0.00039V 5.74t 0.02
+ 1.57E 7V 2 + 12.62t 2 + 0.092t
(23)
damage takes place. The damage evolution criterion is dened by
a scalar variable denoted here d (varying between 0 and 1) which
where V(m/min) is the cutting speed, t (mm) is the uncut chip
can be described either by a linear or exponential function of the
thickness, and
( ) is the rake angle.
equivalent plastic strain. In the present model, an exponential evo-
As the Eqs. (22) and (23) are empirical, it is worth to recall that
lution of damage which describes a Mode II fracture behaviour was
the applicability of these models is limited to the following cutting
adopted, assuming the damage occurs in a shear stress dominated
parameters ranges: cutting speed range of 3611539 m/min, uncut
deformation mode during serrated chip formation. In the uncut
chip thickness range of 0.010.31 mm, and rake angle range of 8
chip zones (zone 1 in Fig. 1), the applied exponential function is
+8 .
as follow:
 up

d = 1 exp( dup ) (18) 2.5. Finite element, meshing, and boundary conditions
0 GfII
The geometry of the nite element model for the orthogo-
where up is the equivalent plastic displacement. For plane strain
nal machining process, assuming plane strain conditions is shown
state, the fracture energy GfII for Mode II fracture is given as [3]:
in Fig. 1. Plain strain four-node quadrilateral elements of type
  CPE4RT are used to model both the workpiece and the cutting
1 2
GfII = 2
KIIC (19) tool. High-density mesh is dened around the tool tip and the
E
uncut chip zones. The mesh of the work material is partitioned
where is the Poissons ratio. KIIC is the fracture toughness for the into three zones as follows: the uncut chip material (zone 1), the
failure mode II. KIIC is equatl to 40 MPa(m)1/2 for the AA7075-T651. machiningaffected zone (zone 2), and the remained part of the
Once the damage occurs, the actual ow stress  d is estimated by: workpiece far from the cutting zone (zone 3). The mesh in zone
1 is characterized by an inclination angle of 45 and minimum
d = (1 d)  (20) element height of 10 m in order to capture the shear localiza-
tion phenomenon and to facilitate the chip ow. Zone 2 is meshed
2.4. Contact modeling with uniform parallel elements (average height of 10 m) over
100 m depth in order to capture the machiningaffected zone.
The friction at the toolchip contact interface is described by Both zones 1 and 2 were meshed using structured meshing func-
Zorev model. Whether the friction is sticking or sliding, it depends tion while zone 3 was meshed using free meshing function available
upon the normal stress acting on the tool-chip interface as follow: in Abaqus/Explicit v6.13 software.
The mechanical boundary conditions applied to the FEM are
Nf , < limit shown in Fig. 1. For the work piece, the nodes along the left (AB) and
f = (21)
limit , limit the bottom (BC) edges are constrained to zero displacement in the
y-direction. However, they are given a velocity in the x-direction
where limit is the shear stress at the tool/chip interface,  is the equal to the cutting speed. For cutting tool, the edge FG and GH are
apparent friction coefcient, and Nf is the friction normal stress constrained to zero displacement in both x and y-directions (xed
along the contact length. Both limit and  were determined using tool).
450 W. Jomaa et al. / Journal of Manufacturing Processes 26 (2017) 446458

Fig. 2. Comparison between (a) experimental and predicted chips using (b) stiffness, (c) relax stiffness, and (d) enhanced hourglass modes for V = 1144 mm/min and
t = 0.15 mm/rev.

The thermal boundary conditions are as follows: 3.1. Machining tests

The studied material is the high strength aluminum alloy

T = T0 on AB, BC, FG, GH

AA7075T651 which is well accepted in manufacturing several

T automotive and aeronautical structural parts. Orthogonal dry
k = hconv (T T0 ) on AD, DC, EF, EH (24)
n machining tests were conducted on a three-axis MAZAK CNC

machinetool using ISO TNMA160408 uncoated carbide inserts.

k T = hint (T T0 ) on chip toolinterface
n
where T0 denotes the room temperature, hconv (equal to 20 W/m2 K)
the heat transfer coefcient between the environment and the free
surfaces of the work piece, chip, and cutting tool.
Heat transfer is allowed at the tool/chip contact interface. To
reach steady state rapidly, a high convection heat transfer coef-
cient is selected (hint = 100 kW/m2 K).
The physical properties of the workpiece and the cutting tool insert
are presented in Table 2. A disk shaped workpiece with a diame-
ter of 70 mm and a width of 4.3 mm was used in the experiments.
Three cutting conditions are tested, including cutting speeds 156,
650, and 1144 m/min while the uncut chip thickness was kept
xed (0.15 mm) for all machining tests. An Optical microscope was
used for analysing the chip morphology and Kistler type 9255B
dynamometer was used for cutting forces measurements.
3.2. Element distortion control method

3. Model calibration and validation
The proposed nite element model is calibrated using experi-
mental chip morphology. Then, the calibrated model is validated
using experimental data obtained during orthogonal machining
tests.
The element type selected (CPE4RT) is a 4nodes isoparamet-
ric solid element with bilinear displacement and temperature. This
element uses a reduced integration scheme with hourglass control
in order to get rid of hourglassing modes resulting from the use of
one integration point. Hence, as shear banding occurs during ser-
rated chip formation, the use of element type CPE4RT with reduced
 W. Jomaa et al. / Journal of Manufacturing Processes 26 (2017) 446458 451

Table 2
Physical properties of the AA7075-T651 alloy and the cutting insert (WC Co)
[2426].

Property Work material Cutting insert

Density,(kg/m3 ) 2810 11900

Elastic modulus,E(GPa) 71.7 612
Poisson s coefcient,v 0.33 0.22
Specic heat,Cp (J/kgK) Cp = 0.786 T + 839.2 Cp = 0.12T + 334.01
Thermal K = 0.0002 T2 + 0.1366T + 128.63K = 0.042T + 35.95
conductivity,K(W/mK)
Thermal 25.2 4.9
expansion,(m/mK)
Melting 908
temperature,Tmelt (K)

integration can lead to a zero energy modes; which, in turn, can

result in uncontrolled distortion of the mesh, known as the hour-
glass effects [27,28]. If this happens, the numerical solution may Fig. 3. Effect of hourglass modes on machining forces for V = 1144 mm/min and
quickly degrade. t = 0.15 mm/rev.
In Abaqus/Explicit 6.13, the following formulations related to
the hourglass control treatment are available:

- Stiffness method denes hourglass control that is strictly elastic,

- Relax stiffness method uses the integral viscoelastic form to
dene the anti-hourglass forces,
- Enhanced hourglass control is based on the assumed enhanced
strain method.

For the chosen element, it is recommended to use either the arti-

cial stiffness method or the articial damping method as hourglass
control scheme. Finite element sensitivity analysis was carried
out in order to select the best Hourglass model for accurately
predicting the segmented chip formation. Fig. 2 displays orthogo-
nal machining simulation results for the same cutting conditions
and using various hourglass modes. As expected, the Enhanced
hourglass control method produces continuous chip that makes it
unsuitable for the proposed modelling. Stiffness and relax stiffness
methods produces serrated chip in agreement with the experi-
ment. Particularly, relax stiffness method is able to reproduce the
chip curvature more accurately than the stiffness method. Over-
all, cutting forces are less sensitive to hourglass modes, as depicted
in Fig. 3. A good agreement between the predicted thrust forces
and the experimental forces is obtained, while the cutting forces
appear to be underestimated by the proposed model. In all cases
the hourglass energy was seen to be less than 6% of the total strain
energy. To summarise, the relax stiffness method was selected to
control the hourglass effect since it provides good prediction of
the chip morphology and a reasonable estimation of the cutting
forces as compared to other tested methods. The following sec-
tion addresses a numerical analysis of the phenomena governing
the thermomechanical material behaviour during the serration of
the chip.
Fig. 4. Effect of cutting speed on (a) cutting force and (b) thrust force for uncut chip
3.3. Cutting forces thickness t = 0.15 mm/rev.

The average cutting and thrust forces are displayed in

Fig. 4. Measured forces decrease as the cutting speed increases.
This decrease is signicant when cutting speed increases from
156 m/min to 650 m/min (thrust force being more affected, 48%,
than the cutting force, 18%). For the cutting speed range of
6501144 m/min, no signicant variations is observed for both
force components, with only 7.5% and, 22% reduction for cut-
ting and thrust forces, respectively. The observed trend can be
explained by the fact that as the cutting speed increases, the cutting
temperature increases leading to material softening, and thereafter
machining forces reduction [29]. Moreover, it has been shown that
at low cutting speeds, build-up edge (BUE) forms [30], ploughing
and seizure effects increase [31], leading to signicant increase in
thrust force when machining AA7075-T651 alloy. When increasing
the cutting speed, BUE disappears and suitable friction conditions at
the tool/chip interface take place, leading to low cutting and thrust
forces [25].
The predicted cutting and thrust forces are in agreement
with experimental values except for the lowest cutting speed
(156 m/min) where the predicted thrust force is signicantly
underestimated (Fig. 4b). It is worth to note that the proposed FEM
do not consider the BUE formation phenomenon, providing a possi-
452 W. Jomaa et al. / Journal of Manufacturing Processes 26 (2017) 446458

Fig. 5. Comparison between experimental and predicted chip morphology for cutting speeds (a) 156 m/min, (b) 650 m/min, and (c) 1144 m/min.

Fig. 6. Geometrical characteristics of the segmented chip.

 W. Jomaa et al. / Journal of Manufacturing Processes 26 (2017) 446458 453

Fig. 7. Comparison of FEM and experimental chip parameters, (a) height of the chip H, (b) height of the continuous portion h, (c) shear band spacing S, and (d) average shear
angle .

ble reason for the underestimation of the thrust force at low cutting
speed.

3.4. Chip morphology

The validation of the proposed FEM is also carried out through

a comparison between predicted and experimental chip mor-
phology. Fig. 5 depicts the experimental and simulated chip
morphologies under the three tested cutting speeds. As the cutting
speed increases, the chip becomes thinner, the chip segmentation
intensify, and then the chip curvature decreases. The latter is most
probably related to the decrease in the coefcient of friction as the
cutting speed increases. Hence, the proposed FEM captures well the
observed experimental trends.
The chip morphology can be described using four parameters
as illustrated in Fig. 6: the chip thickness H, the height of the con-
tinuous portion of the chip h, and the shear band spacing S. Fig. 5
compares the predicted and the experimental chip parameters for
different cutting speeds. As the cutting speed increases, the peak
(H) and valley (h) decreases as shown in Fig. 7a and b, respectively.
The shear band spacing (S) seems to be less sensitive to the change
of cutting speed as compared to H and h (see Fig. 7c). Conversely,
the shear angle increases with the cutting speed (Fig. 7d). This is
in agreement with the theory of machining which states that as
shear angle increases the machining forces decreases (Fig. 4) and
the machining becomes easier. Also, Fig. 5 shows good agreement
between predictions and experimental chip parameters. In order to
check the capability of the proposed FEM to estimate the chip seg-
mentation intensity, the predicted and experimental peak to valley
ratio (h/H), which is a good criterion for estimating the segmenta-
tion intensity, is calculated. The rule of thumb is the lower is h/H
ratio, the higher is the segmentation intensity. Fig. 8 shows that
the ratio (h/H) decreases as the cutting speed increases, showing
that the chip segmentation intensity is sensitive to cutting speed
Fig. 8. Comparison of FEM and experimental average chip segmentation intensity
(h/H).
variation. This was not the case when machining hardened steels
AISI 52100 [1] and AISI 1550 [32]. In these previous works [1,32] for
which the ratio (h/H) is constant (0.4) over a large range of tested
cutting speeds. In the present study, the predicted chip segmen-
tation intensity parameter is in agreement with the experimental
data. The discrepancy observed at low cutting speed (156 m/min)
is may be related to the BUE formation when machining aluminum
alloys at low cutting speed.
In the present study, based on the chip morphology (Fig. 5),
the critical cutting speed for the onset of chip serration seems to
be lower than 156 m/min for the age hardened 7075-T651 alloy,
while it was found to be higher than 270 m/min for the non-aged
7075 alloy based on data from [5]. This is in agreement with previ-
ous studies [1] attesting that the harder is the material the higher
is its capability to produce serrated chips. Moreover, the present
machining tests were performed under negative rake angle (5 )
while those in [5] were performed with 7 , which may explain the
454 W. Jomaa et al. / Journal of Manufacturing Processes 26 (2017) 446458
higher critical cutting speed value in their case. Rake angle is one
of the most important machining parameters inuencing the chip
formation process. In particular, Lee [33] demonstrated that nega-
tive rake angle are more favourable to produce serrated chips than
positive rake angles. The obtained results are in agreement with
ndings in [5] suggesting that the critical cutting speed relative to
the onset of chip segmentation has a strong dependence on the
work material and machining conditions.
Conclusively, the predicted results in terms of cutting forces and
chip morphology were very close to those obtained experimentally
and follow the same trends. Therefore, the proposed FEM model
can be realistically used to study the physical mechanisms involved
during serrated chip formation.
4. Numerical analysis of serrated chip formation
prim ) prim a
4.1. Mechanics of shear localization
The equivalent plastic strain, temperature, and shear damage
distributions during serrated chip formation for the highest cut-
ting speed (1144 m/min), displayed in Fig. 9, shows that serrated
chips are formed following three steps. Initially, a plastic strain
eld is localised on the tool tip (Fig. 9a). As the cutting progresses,
the equivalent plastic strain eld continuously intensify, forming
a straight line toward the free surface of the chip (Fig. 9b). At this
)
stage, the heat induced by the plastic deformation increases glob-
ally in the chip, resulting in temperature raise (Fig. 9e). Finally,
when the strain localization and temperature reach certain values,
a narrower localized shear band forms and a new segment of chip
is formed by sliding over the uncut work material (Figs. 9c and
f).
prim denotes the
Regarding the very short time in which the above mechanisms
are taking place, this deformation process is considered as an adia-
batic process. Fig. 9gi, which display the damage evolution during
prim V cos
cos(
)
deformation at the last stage of chip formation, conrm that the
prim = (27)
serrated chip was formed under a local drop of the material ow
+ t) sin  H cos(
)
stress, resulting in catastrophic failure (Fig. 9i). This catastrophic
prim is the shear strain, dened by the following equation:
strain localization takes place when thermal softening compensate
the work hardening induced by strain and high local strain rates
[1]. This behavior was also reported in previous study on hardened
prim = (28)
steel [2,14] which stated that the primary instability within the pri-
mary shear zone during serrated chip formation is the initiation of
adiabatic shearing at the tool tip and its propagation partway to the
free surface of the work material.
An interesting phenomenon is observed when the cutting speed
increases: strain rate signicantly increases in a local region of
the chip (Fig. 10b), the highest temperatures are concentrated
in narrower zones (Fig. 10c), then the equivalent stresses sud-
denly drop (Fig. 10d), favouring the appearance of an adiabatic
shear band. Conversely, when machining at low cutting speed, the
cutting mechanism is no longer governed by the thermal soft-
ening as the maximum temperature is relatively low and heat
is able to propagate over a large zone, avoiding shear localiza-
tion. Looking back at the results of the machining forces (Fig. 4),
prim m
one can argue that the thermal softening phenomenon is the
0
main reason for reducing the machining forces, as it decreases the
material strength in the primary shear zone. In turn, this results
0 are the reference temperature and the reference
in a signicant decrease of the chip bending moment, therefore
decreasing the contact length and friction coefcient as stated in
[34].
4.2. Prediction of the shear band spacing: FEM versus analytical
solution
In this section, the analytical modeling of the shear localization
phenomenon is discussed with emphasize on comparing analyti-
cal data with those numerically predicted using the proposed FEM.
Originally observed in ballistic experiments, the shear localiza-
tion phenomenon has aroused the curiosity of many researchers
to develop analytical tools for analysing chip serration observed
in machining processes. The developed models mainly focused on
predicting the onset of the shear localization and serrated chip
characteristics. The shear band spacing (S) is one of the most the-
oretically studied chip parameter as it can help understanding the
mechanisms of chip serration. Molinari et al. [35] modeled the shear
band spacing parameter (S) based on a visco-thermo-plastic con-
stitutive equation. The Molinaris model is dened as [35]:
14
Tprim 2
m2 KCp (1 ) T0
S = 2
T0
(25)
(1 1 2 3 2
m ) (
where Tprim denotes the temperature at the primary shear zone.
In the present study, the authors adopt the Oxley model [36] for
estimating Tprim . Although it is originally developed for continu-
ous chips, this model has shown good accuracy for serrated chip
formation [37]. Tprim is given by [36]:
 
1 prim cos
Tprim = T0 +  (26)
Cp sincos (
where  is the Taylor-Quinny coefcient, representing the amount
of deformation energy converted to heat (equal to 0.9 in this study)
and  is the fraction of energy dissipated into the workpiece. In this
study,  is set equal to 0.1, assuming a deformation process driven
by adiabatic shearing at the primary shear zone.
strain rate, calculated using an analytical model for serrated chip
formation developed by Duan et al. [38]:
(P cos
and

M H 2 + t 2 2Ht sin
Ht cos
where P is the distance between two consecutive saw-tooth (Pitch);
 is the slope angle of the sawtooth (Fig. 6), M is the height of the
sawtooth. t is the primary shear zone thickness and is assumed
equal to 0.1 for the 7075-T651 alloy [39].
Finally, 0 , a, and m are, respectively, the yield stress, the mate-
rial softening coefcient, and the strain rate sensitivity. In the
present work, these material constants are calibrated using regres-
sion analysis on dynamic tests data of 7075-T651 alloy from [40].
In developing the shear band spacing model (Eq. (25)), Molinari
et al. [35] have used the following visco-thermo-plastic constitutive
equation:
Tprim
prim = 0 (1 + a )( ) (29)
T0
where T0 and
shear strain rate, which are, respectively, set equal to 25 C and
1 s1 . The calibration results are as follow: 0 = 217.5MPa, a = 0.053,
and m = 0.1.
It is worth noting that Molinari et al. [35] also used Eq. (29) to
calculate prim . Nevertheless, in the present study, the material con-
stants are calibrated using dynamic tests data limited to the strain
of 0.3 and the strain rate of 3.1 103 s1 which are lower than those
encountered during machining. Thus, to have good accuracy in esti-
mating the actual shear stress at the primary shear zone prim , the
 W. Jomaa et al. / Journal of Manufacturing Processes 26 (2017) 446458 455

Fig. 9. Sequences of a chip serration: (a)(c) equivalent plastic strain, (d)(f) temperature distributions, and (g)(i) shear damage for V = 1144 mm/min and t = 0.15 mm/rev.

following equation developed based on the mechanics of serrated As shown in Fig. 11, the predicted shear band spacing parameter
chip formation was adopted [38]: S are in agreement with experiments. In particular, the analytical
and the FEM solutions are very close, except at the highest cutting
sin [FC sin (
) FT sin (
)] sin  speed where the FEM overestimates the parameter S. The results
prim =   (30)
H M/2 w also show that the use of the proper chip formation model and
material constants is essential for the prediction of the shear band
where is the angle between the primary and the secondary shear spacing using Eq. (25). Given that the material properties were
zones (Fig. 6), FC and FT are the cutting and the thrust forces, respec- derived from the high strain rate and elevated temperature data
tively. by Lee et al. [40], the results here show that the shear band spac-
Table 3 lists the measured chip parameters values along with ing parameter must cater for the variation in strain, strain-rate,
the state variables needed for the calculation of S. Some other com- and temperature in order to obtain good correlation between the
plementary chip dimensions and cutting force values needed for theoretical and experimental results. Furthermore, in the present
the analytical model were already presented in Sections 3.3 and work, the thermal properties are considered depended on the cut-
3.4. ting temperature (Table 2). This was not the case in some previous
456 W. Jomaa et al. / Journal of Manufacturing Processes 26 (2017) 446458

Fig. 10. State variable distributions along (a) path AB for (b) strain rate, (c) temperature, and (d) Von-Mises stress at different cutting speeds and xed uncut chip thickness
t = 0.15 mm.

Table 3
Data for computing the shear band spacing parameter S.

Measured chip parameters

V [m/min] M [mm] P [mm]  [ ] [ ] [ ]

156 0.124 0.072 61.5 58.7 56.7
650 0.131 0.122 69.7 54.1 54.7
1144 0.126 0.070 75.7 47.5 47.0

Predcited state variables

V [m/min] Analytical model FEM

 1
  

prim [%] 5

prim 10 Sec Tprim [ C] prim [MPa]
prim [%]
prim 105 Sec 1 Tprim [ C] prim [MPa]

156 966 1.27 332.4 391.6 189 0.49 261.3 344.1

650 1077 1.61 322.9 417.3 207 2.48 340.3 278.1
1144 1138 1.92 327.8 457.8 271 6.23 324.8 266.1

works, such as [1], where the shear band spacing parameter was
calculated using xed shear angle and cutting temperature along
with thermal properties taken independent of the temperature,
resulting in an overestimation of S by about 50% for a hardened
steel.
The discrepancy between the analytical and the FEM state vari-
able values used in the prediction of S is an interesting aspect that
has to be considered. The highest discrepancy is observed in the
shear strain (Table 3) for which the analytical model prediction is
about 5 times higher than the FEM one. For the shear strain rate and
shear stress, the highest discrepancy is observed at the highest cut-
ting speed (1144 m/min). At moderate and low cutting speeds, the
shear strain rate, the shear stress, and the temperature predicted
with the analytical and the FEM are in the same order of magni-
tude. However, the analytical temperature decreases with cutting
speed while the numerical temperature increases. The inverse is
true for the shear stress. On one hand, these results conrm the
agreement between the analytical and FEM prediction at low and
moderate cutting speed, as shear band spacing parameter is inde-
pendent of the shear strain and only depends on shear strain rate,
temperature, and shear stress (Eq. (25)). On the other hand, the
above results suggest that different set of state variable values can
lead to an equivalent thermomechanical response of the mate-
rial, resulting in similar shear band spacing values. Thus, there is
no unique solution to the problem. This issue will be discussed in
future works.
Based on the above discussions, it is clear that any attempt to
establish a criterion for the onset of chip serration or adiabatic shear
bending should be imperatively accounted for the actual levels of
the state variables (strain, strain rate, ow stress, and temperature)
 W. Jomaa et al. / Journal of Manufacturing Processes 26 (2017) 446458
Fig. 11. Numerical versus analytical shear band spacing parameter S.
into the primary shear zone. This is out of the scope of the present
study since the main objective was to develop and to validate the
nite element model in order to predict chip serration under dif-
ferent cutting conditions. Future work will be addressed to this
issue by using a large machining data involving other relevant cut-
ting parameters such as cutting tool rake angle and cutting edge
preparation.
5. Conclusions
In this study, a numerical investigation of serrated chip forma-
tion during high speed machining of aluminum alloy AA7075-T651
is carried out. A FEM model is developed using Abaqus/Explicit 6.13
and JC constitutive and damage equations. Based on the obtained
results, the following conclusions can be drawn:
- The proposed FEM model with a Lagrangian formulation was able
to simulate accurately the serrated chip formation in HSM of
AA7075-T651 alloy,
- Simulations results were able to capture the experimental fact
that the chip segmentation intensity increases when the cutting
speed increases,
- Serrated chips are mainly formed by shear strain localisation at
the tool tip and its propagation partway to the free surface of the
work material,
- The decrease of the machining forces observed under segmented
chip formation can be related to the material softening effects
taking place locally in the narrow primary shear zone of the chip,
- As for hard-to-cut materials, chip segmentation of AA7075-T651
alloy occurs under adiabatic shear banding process at high cutting
speeds.
- The FEM and the analytical predictions of the shear band spacing
parameter were in good agreement in spite of the difference in
the calculated state variables, revealing the non-uniqueness of
the solution.
Acknowledgments
This research was supported partially by the National Science
and Engineering Research Council of Canada (NSERC) and Hroux-
Devtek Inc. A part of the research presented in this paper was also
nanced by the Fonds de recherche du Qubec Nature et Tech-
nologies by the intermediary of the Aluminium Research Centre
REGAL.
References
[1] Morehead MD, Huang Y, Luo J. Chip morphology characterization and modeling
in machining hardened 52100 steels. Mach Sci Tech 2007;11:33554.
[2] Barry J, Byrne G. The mechanisms of chip formation in machining hardened
steels. J Manuf Sci Eng 2002;124:52835.
457
[3] Mabrouki T, Girardin F, Asad M, Rigal J-F. Numerical and experimental study of
dry cutting for an aeronautic aluminium alloy (A2024-T351). Int J Mach Tools
Manuf 2008;48:118797.
[4] Atlati S, Haddag B, Nouari M, Zenasni M. Analysis of a new segmentation
intensity ratio SIR to characterize the chip segmentation process in machining
ductile metals. Int J Mach Tools Manuf 2011;51:687700.
[5] Ye G, Chen Y, Xue S, Dai L. Critical cutting speed for onset of serrated chip ow
in high speed machining. Int J Mach Tools Manuf 2014;86:1833.
[6] Astakhov VP. Tribology of Metal Cutting. 1st ed. Elsevier; 2006.
[7] Liyao G, Minjie W, Chunzheng D. On adiabatic shear localized fracture during
serrated chip evolution in high speed machining of hardened AISI 1045 steel.
Int J Mech Sci 2013;75:28898.
[8] Vyas A, Shaw M. Mechanics of saw-tooth chip formation in metal cutting. J
Manuf Sci Eng 1999;121:16372.
[9] Poulachon G, Moisan A, Jawahir I. Tool-wear mechanisms in hard turning with
polycrystalline cubic boron nitride tools. Wear 2001;250:57686.
[10] Calamaz M, Coupard D, Girot F. A new material model for 2D numerical sim-
ulation of serrated chip formation when machining titanium alloy Ti6Al4 V.
Int J Mach Tools Manuf 2008;48:27588.
[11] Ceretti E, Lucchi M, Altan T. FEM simulation of orthogonal cutting: serrated chip
formation. J Mater Process Tech 1999;95:1726.
[12] Ducobu F, Rivire-Lorphvre E, Filippi E. Numerical contribution to the com-
prehension of saw-toothed Ti6Al4 V chip formation in orthogonal cutting. Int J
Mech Sci 2014;81:7787.
[13] Hua J, Shivpuri R. Prediction of chip morphology and segmentation during the
machining of titanium alloys. J Mater Process Tech 2004;150:12433.
[14] Jomaa W, Ben Fredj N, Zaghbani I, Songmene V. Non-conventional turining of
hardened AISI D2 tool steel. Int J Adv Mach Form Op 2011;3:93126.
[15] Sima M, zel T. Modied material constitutive models for serrated chip forma-
tion simulations and experimental validation in machining of titanium alloy
Ti6Al4 V. Int J Mach Tools Manuf 2010;50:94360.
[16] Saanouni K, Lestriez P, Labergre C. 2D adaptive FE simulations in nite thermo-
elasto-viscoplasticity with ductile damage: application to orthogonal metal
cutting by chip formation and breaking. Int J Damage Mech 2009;20(1):2361.
[17] Komanduri R, Hou ZB. Thermal modeling of the metal cutting process: part
Itemperature rise distribution due to shear plane heat source. Int J Mech Sci
2000;42:171552.
[18] Vaz Jr M, Owen D, Kalhori V, Lundblad M, Lindgren L-E. Modelling and simula-
tion of machining processes. Arch Comput Meth Eng 2007;14:173204.
[19] Lorong P, Yvonnet J, Cofgnal G, Cohen S. Contribution of computational
mechanics in numerical simulation of machining and blanking: state-of-the-
art. Arch Comput Meth Eng 2006;13:4590.
[20] Johnson GR, Cook WH. Fracture characteristics of three metals subjected to
various strains, strain rates, temperatures and pressures. Eng Fract Mech
1985;21:3148.
[21] ABAQUS/Explicit, version 6.13 users manual, in: Simulia, 2013.
[22] Brar N, Joshi V, Harris B. Constitutive model constants for Al7075-t651 and
Al7075-t6. In: Shock compression of condensed matter 2009: Proceedings of
the American Physical Society Topical Group on Shock Compression of Con-
densed Matter. AIP Publishing; 2009. p. 9458.
[23] Li S, Hou B. Material behavior modeling in machining simulation of 7075-T651
aluminum alloy. J Eng Mat Tech 2013;136:011001.
[24] Daoud M, Jomaa W, Chatelain J, Bouzid A. A machining-based methodology to
identify material constitutive law for nite element simulation. Int J Adv Manuf
Tech 2015;77:201933.
[25] Fang N, Wu Q. The effects of chamfered and honed tool edge geometry in
machining of three aluminum alloys. Int J Mach Tools Manuf 2005;45:117887.
[26] Yen Y-C, Jain A, Altan T. A nite element analysis of orthogonal machining using
different tool edge geometries. J Mater Process Tech 2004;146:7281.
[27] Flanagan D, Belytschko T. A uniform strain hexahedron and quadrilateral with
orthogonal hourglass control. Int J Num Meth Eng 1981;17:679706.
[28] Haddag B, Atlati S, Nouari M, Znasni M. Finite element formulation effect in
three-dimensional modeling of a chip formation during machining. Inter J
Mater Form 2010;3:52730.
[29] Jomaa W, Songmene V, Bocher P. An investigation of machining-induced resid-
ual stresses and microstructure of induction-hardened AISI 4340 steel. Mater
Manuf Processes 2016;31:83844.
[30] Jomaa W, Songmene V, Bocher P. Surface nish and residual stresses induced
by orthogonal dry machining of AA7075-T651. Materials 2014;7:160324.
[31] Kilic DS, Raman S. Observations of the tool-chip boundary conditions in turning
of aluminum alloys. Wear 2007;262:889904.
[32] Elbestawi MA, Srivastava AK, El-Wardany TI. A model for chip formation during
machining of hardened steel. CIRP J Manuf Sci Technol 1996;45:716.
[33] Lee D. The effect of cutting speed on chip formation under orthogonal machin-
ing. J Eng Indus 1985;107:5563.
[34] Jomaa W, Daoud M, Songmene V, Bocher P, Chtelain J-F. Identication
and validation of Marusichs constitutive law for nite element modeling
of high speed machining. ASME 2014 International Mechanical Engineering
Congress and Exposition: American Society of Mechanical Engineers 2014.
V02AT02A057V002AT002A057.
[35] Molinari A, Musquar C, Sutter G. Adiabatic shear banding in high speed machin-
ing of Ti-6Al-4V: experiments and modeling. Int J Plast 2002;18:44359.
[36] Oxley P, Hastings W. Predicting the strain rate in the zone of intense shear in
which the chip is formed in machining from the dynamic ow stress proper-
ties of the work material and the cutting conditions. Proc R Soc Lond Ser A
1977;356:395410.
458 W. Jomaa et al. / Journal of Manufacturing Processes 26 (2017) 446458

[37] Jomaa W, Songmene V, Bocher P. An hybrid approach based on machining and
dynamic tests data for the identication of material constitutive equations. J
Mater Eng Perform 2016;25:101027.
[38] Duan C, Wang M, Pang J, Li G. A calculational model of shear strain and strain
rate within shear band in a serrated chip formed during high speed machining.
J Mater Process Tech 2006;178:2747.
[39] W. Jomaa, Contributions to Understanding the High Speed Machining Effects
on Aeronautic Part Surface Integrity, Ph.D. in: mechanical engineering, Canada:
cole de technologie suprieure, 2015, pp. 240.
[40] Lee W-S, Sue W-C, Lin C-F, Wu C-J. The strain rate and temperature dependence
of the dynamic impact properties of 7075 aluminum alloy. J Mater Process Tech
2000;100:11622.