Finite Elements in Analysis and Design 134 (2017) 27–40

Contents lists available at ScienceDirect

Finite Elements in Analysis and Design

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

Finite element modelling of 3D orthogonal cutting experimental tests with MARK

the Coupled Eulerian-Lagrangian (CEL) formulation
⁎
F. Ducobu , E. Rivière-Lorphèvre, E. Filippi
University of Mons (UMONS), Faculty of Engineering (FPMs), Machine Design and Production Engineering Lab, 20 Place du Parc, B-7000 Mons, Belgium

A R T I C L E I N F O A BS T RAC T

Keywords: Most of the current finite element models of cutting concerns the 2D plane strain orthogonal cutting
Coupled Eulerian-Lagrangian (CEL) configuration, which, although of valuable interest to study the fundamental phenomena of the process, is
Finite element still far from most practical cutting operations. The 3D models on the other side usually concern a 2D tool path
Lateral expansion with a cutting edge that is not straight any more. The step just after 2D orthogonal cutting is almost not
3D orthogonal cutting
addressed; it is the 3D orthogonal cutting. Based on an experimental reference of Ti6Al4V orthogonal cutting,
Ti6Al4V
this paper introduces a 3D finite element Coupled Eulerian-Lagrangian (CEL) model of orthogonal cutting that
faithfully reproduces the experimental operation and is verified by comparison with it. Such a model and a
comparison are not available in the current literature. A comparison is performed with a 2D orthogonal cutting
model as well. The forces, the chip thickness and the lateral expansion of the chip were accurately modelled. The
results showed that it is necessary to set the width of the workpiece large enough in order to get close to the
plane strain assumption valid in the experimental configuration. A width of the workpiece equal to the uncut
chip thickness constitutes a good compromise between the quality of the results and the number of nodes of the
model. The Eulerian formalism of the workpiece mesh allowed to reduce its height and therefore the number of
nodes without affecting the quality of the results. Moreover, this study showed that cubic elements should be
adopted, as well as the addition of a range of elements that stays full of void and is parallel to the cutting plane at
the boundary of the model to avoid influencing the results.

1. Introduction were reported in the literature. The geometrical configuration can be

rather complex, depending on the process studied. Turning and drilling
Numerical modelling with the finite element method to study metal gather most of the publications up to now and some of them are
cutting was introduced at the beginning of the seventies [1]. Due to its presented in the following literature review. When the model concerns
high complexity and the large amount of phenomena it involves, the turning, the workpiece is nearly always simplified to a parallelepipedal
process is mostly studied in orthogonal cutting in order to mainly block and the movement of the tool is linear, which is similar to a
reduce the geometrical difficulties and the number of degrees of shaping operation. The curvature of the workpiece is ignored as its
freedom of the models. The physical coupled phenomena (large strains, radius is assumed large enough for the short modelled cutting length.
strain rates, at high temperatures and temperatures gradients, friction, Guo and Liu [4] presented a 3D model of hardened AISI 52100 steel
and so on) must still however be considered and addressed, which with a preformed chip as initial condition to help reaching the steady-
leads to many publications. For more details, Arrazola et al. [2] present state. The model of Aurich and Bil [5] was close to the previous one but
a review of metal machining modelling since 1998, while the review of it focused on the formation of segmented chip; no experimental
Van Luttervelt et al. [3] concerns previous modelling works. Their great validation was performed. Li and Shih [6] developed a 3D model for
majority deals with 2D orthogonal cutting. While this configuration is turning two grades of titanium with a focus on the chip curl. Arrazola
really useful to understand the fundamental phenomena of cutting, it is and Özel [7] developed a 3D model of hard turning of AISI 4340 steel
far from the practical industrial applications and therefore industrial with an hybrid procedure involving a remeshing step followed by the
needs. As highlighted by Arrazola et al. [2], it is now time to move from transfer of the current results into a 3D Arbitray Lagrangian-Eulerian
2D modelling to 3D in order to answer the demands of the industry. (ALE) model with Eulerian boundaries to model the steady-state of the
3D finite element models of cutting do however already exist and process. The influence of the tool edge geometry was studied by Özel

⁎
Corresponding author.
E-mail address: [email protected] (F. Ducobu).

http://dx.doi.org/10.1016/j.finel.2017.05.010
Received 16 February 2017; Received in revised form 22 May 2017; Accepted 28 May 2017
Available online 03 June 2017
0168-874X/ © 2017 Elsevier B.V. All rights reserved.

https://freepaper.me/t/461202 : ‫خودت ترجمه کن‬

F. Ducobu et al. Finite Elements in Analysis and Design 134 (2017) 27–40

Fig. 1. Experimental test sequence (a) Beginning of the cut: the tool enters in the workpiece, (b) Cutting, (c) End of cut: the tool leaves the workpiece, adapted from Ducobu et al. [21].

[8] with a 3D model of turning AISI 4340 steel. Abaqus/Explicit and cutting edge and/or a tool path that is not straight any more. The
Deform 3D software were compared by Özel et al. [9] to machine modelled cutting operation is exactly the same as the experimental
Inconel 718. Özel and Ulutan [10] predicted the residual stresses with a tests, which has not previously been seen in the literature. The 3D
3D finite element model of turning Ti6Al4V and IN100 alloys. For model will be introduced before being compared to the experimental
drilling, Guo and Dornfeld [11] presented a 3D finite element model of results and to the results of the same model but in 2D. The 3D model
drilling in 304L stainless steel to study burr formation. A more recent will then be optimized in order to decrease the computing time and
work of Abouridouane and al. [12] dealt with multi-phase material with make it more useful in an industrial context. Then, the sensitivity of the
a 3D two-phase drilling model of C45 steel. Modelling of milling results to the width of the workpiece will be studied. The 3D resulting
introduces either the complex geometry of the tool, as Soo et al. [13] model will finally be applied to two other uncut chip thicknesses values
did with a 3D finite element model ball nose end-milling in Inconel to show its predictive capacity and be compared to the experimental
718, either the consecutive action of several teeth, as in the model of reference.
Pantalé et al. [1] in 42CrMo4 steel.
All the works listed so far have the complex geometry of a 3D
operation and are consequently rather different of the simplified 2D 2. Experimental reference
orthogonal cutting configuration. A few works however deal with an
intermediary configuration, which is the direct evolution of 2D The experimental setup introduced in [21] corresponds to a 3D
orthogonal cutting: 3D orthogonal cutting and/or oblique cutting with orthogonal cutting configuration. It consists in removing the upper
a straight cutting edge. Fang and Zeng [14] presented a 3D finite layer of the workpiece by a linear displacement of the workpiece
element model of oblique cutting in ANSI 1045 steel; no experimental (Fig. 1) towards the fixed tool. The configuration is similar to that of a
validation was provided. Pantalé et al. [15,16,1] developed a 3D ALE planing operation. Such a setup provides strictly orthogonal cutting
finite element model of orthogonal and oblique 42CrMo4 cutting. They conditions in 3D that can be directly compared to the modelling as it
highlighted a lateral expansion of the chip equal to 14% of the corresponds to the same operation. The machined feature is a tenon of
workpiece width. This lateral expansion is linked to a defect observed 1 mm width situated at the top of a cylindrical part (Fig. 2) fixed in the
on the machined surface: the side flow, which is a material flow in the spindle (not rotating) of a Deckel-Maho DMU-80 T milling machine.
direction perpendicular to the chip flow [17]. The material sticks then The ratio between the uncut chip thickness, h, and the width of cut, b, is
to the machined surface and has a negative impact on its quality. It was of at least 10 for the chosen cutting conditions of this work (width of
studied experimentally by Kishawy and Elbestawi [17] in hard turning cut is 1 mm and the largest uncut chip thickness value is 0.1 mm),
of AISI 4615 steel. They observed that the fraction of side flow on feed which makes it suitable for the validation of 2D plane strain orthogonal
increases when feed decreases, that an increase in cutting speed cutting models.
increases the temperature which facilitates the side flow and that the The cutting conditions are summarized in Table 1. The machined
side flow increases with the tool nose radius and the tool wear. Kishawy material is the Ti6Al4V titanium alloy. Each cutting condition was
et al. [18] studied this side flow numerically in a 3D finite element repeated three times. The value of the cutting speed is limited to the
model of multi pass hard turning of AISI 52100 steel. In the cutting maximal feed rate of the machine, 30 m/min. That value was adopted
conditions of the model, it was estimated between 2% and 10% of the to stay in the recommended cutting speed range of the tool manufac-
feed (the feed was 0.1 mm/rev and the depth of cut was 0.25 mm) turer [22]. The WC/Co tool is coated with TiN PVD. The forces were
depending on the value of the tool nose radius. Attention was not paid measured with a dynamometer Kistler 9257B. Their root mean square
to the side flow and/or the lateral expansion of the chip in the models (RMS) values are given in Table 2. Chips were collected and embedded
other than that of Pantalé et al. [15,16,1] and Kishawy et al. [18], to be observed with an optical microscope after being polished. All of
although it is a significant difference between 2D and 3D models. them were continuous. The average value (on 25 measurements for
Illoul and Lorong [19] applied another numerical method than the each uncut chip thickness) of their thickness values are provided in
finite element to model 3D cutting. They used the Constrained Natural Table 2.
Element Method (CNEM) for 3D orthogonal and oblique cutting. Their The lateral expansions of the chips, b′, are provided in Table 3. The
objective was to show the capacity of the method to solve that kind of lateral expansion value is computed by the difference between the
problems. No information was provided on the quality of the results
and they were not compared to an experimental reference.
This overview of 3D numerical modelling of cutting showed that few
models focused on the step just after 2D orthogonal cutting towards an
industry-relevant 3D operation: 3D orthogonal cutting. The numerical
results in this configuration were moreover almost not validated
experimentally.
This paper performs this transition from 2D to 3D orthogonal
cutting on the basis of a 2D Coupled Eulerian-Lagrangian (CEL) finite
element model previously validated experimentally [20]. Lateral ex-
pansion of the chip is one of the phenomena to be addressed and that
makes the difference with a 2D configuration. Although not yet of
industrial interest, 3D orthogonal cutting is the preamble to oblique
cutting, itself introducing industry-relevant complex 3D cutting with a
Fig. 2. Sample to be machined, dimensions in mm [21].

28

https://freepaper.me/t/461202 : ‫خودت ترجمه کن‬

F. Ducobu et al. Finite Elements in Analysis and Design 134 (2017) 27–40

Table 1
Cutting conditions and tool geometry of the experimental reference [21].

Cutting speed, Vc (m/min) 30

Uncut chip thickness, h (μm) 40, 60, 100
Rake angle, γ (°) 15
Clearance angle, α (°) 2
Cutting edge radius, r (μm) 20
Width of the cutting edge (mm) 6

Table 2
Experimental cutting (CF) and feed (FF) forces average RMS values, chip thickness (h′)
average values [21].

h (μm) RMS CF (N/mm) RMS FF (N/mm) h′ (μm)

40 86 ± 2 40 ± 1 59 ± 5
60 113 ± 2 44 ± 1 80 ± 4
100 174 ± 2 50 ± 1 135 ± 6

Fig. 3. Schematic initial geometry and boundary conditions of the 3D model in a view
Table 3 perpendicular to the cutting plane XY (r: cutting edge radius, h: uncut chip thickness, Hv:
Experimental lateral expansion of the chips (b′), bx′ : ratio of b′ to the length x (the uncut height of the area initially filled with void).
chip thickness, h, or the width of the workpiece, b). The experimental width of the
workpiece, b = 978 ± 10.7 μm . above the workpiece in which the chip will develop, Hv, depends on the
h
value of the uncut chip thickness in order to limit the number of nodes
b′ bb′ bh′
(μm) (μm) (% b) (% h) of the model. It is chosen to allow the chip to form entirely in the
Eulerian mesh.
40 0.3 ± 6 0 1 The workipece is meshed with 8-node brick square elements of
60 1 21
12.6 ± 15 5 μm×5 μm in the cutting plane XY. The tool is more coarsely meshed
100 39.1 ± 16 4 39
except near the edge radius where the elements length is 5 μm in the
cutting plane. Fig. 4 presents the 3D initial conditions of the model, as
measured width of the chip and the measured width of the tenon, b. For well as the initial mesh with 13 elements in the width of the model for
each uncut chip thickness, 10 measurements were performed. Two an uncut chip thickness of 60 μm. Half of the experimental workpiece
indicators, bb′ and bh′, are introduced to quantify the magnitude of the width is modelled to take advantage of the symmetry of the configura-
b′ tion. One of the two external faces parallel to the cutting plane is
lateral expansion of the chip: bb′ = in % of the width of the workpiece
b
b′
considered as a plane of symmetry. No constraint on the degrees of
b and bh′ = h in % of the uncut chip thickness h. They are introduced to freedom of the second face is applied in order to allow the free side of
show the magnitude of the lateral expansion by comparison to the the chip to deform (and potentially expand) freely. The number of
width of the tenon, b, and the uncut chip thickness, h. elements out of the cutting plane (in the Z direction, Fig. 4) is a
Standard deviation values on the experimental lateral expansion of parameter that will be further studied and discussed. It depends on the
the chips, b′, are high: the variations of the width of the chips are number of elements composing the workpiece and the elements added
significant compared to its lateral expansion. When h=40 μm and next to it (and initially filled with void) that will receive the material
60 μm, the standard deviation is larger than the average value, mean- resulting of the lateral expansion of the chip and the side flow of the
ing that the width of the chips is smaller than the width of the tenon on workpiece. Except for one particular case, the elements length in this
some measured points. When comparing with the numerical works
from the literature, the lateral expansion, in the experimental reference
of this study, represents a smaller fraction of the width of the tenon
(bb′ = 0–4% versus ≈14 % in the literature [16]) and a larger fraction of
the uncut chip thickness (bh′ = 1–39% versus 2–10% in the literature
[18]). The evolution of bh′ is moreover the opposite of the trend
reported in the literature [18]: it decreases when the uncut chip
thickness decreases.

3. Finite element model

The 3D CEL finite element model introduced in this work is based

on the 2D model experimentally validated in [20]. The tool is described
by the Lagrangian formalism and the Ti6Al4V workpiece by the
Eulerian one. This avoids the usual severe mesh distortions in a
Lagrangian workpiece due to the large strains occurring during the
cut. It is however necessary to mesh the volume where the chip will
form contrary to the Lagrangian formulation, which increases the
number of nodes of the model. Fig. 3 shows schematically the
dimensions and the boundary conditions of the model. The cutting
duration sets the workpiece length. It is chosen to have the lowest
number of nodes but to form a chip long enough to be measured and to Fig. 4. In a 3D view: initial conditions and mesh of the 3D model for h = 60 μm and a
reach the steady-state for the cutting forces. The height of the area given width of the model.

29

https://freepaper.me/t/461202 : ‫خودت ترجمه کن‬

F. Ducobu et al. Finite Elements in Analysis and Design 134 (2017) 27–40

Table 4 Table 5
Materials properties of the numerical model [25–28]. Number of elements and nodes and value of Hv for the 2D model and the 3D models with
1 element of 5 μm in the width of the workpiece (narrow model: 3 elements in the width
JC constitutive model A (MPa) 997.9 of the model, other models: 4 elements in the width of the model) for h=60 μm.
B (MPa) 653.1
C 0.0198 Model Elements Nodes Hv (μm)
m 0.7
n 0.45 2D 8800 17,978 300
ε̇0 1 1 narrow 26,625 36,332 300
Troom (K) 298 1 35,500 45,415 300
Tmelt (K) 1878 1, 50 μm 35,500 45,415 300
Young's modulus, E (GPa) Ti6Al4V 113.8
Carbide 800
Density, ρ (kg/m3) Ti6Al4V 4430 This means that the number of elements next to the initial workpiece is
Carbide 15,000
not large enough to allow the chip to expand freely outside of the
Conductivity, k (W/mK) Ti6Al4V 7.3
Carbide 46 cutting plane. This cannot be allowed and it is necessary to increase the
Expansion, α (K−1) Ti6Al4V 8.6 e−6 width of the initial empty space modelled next to the workpiece.
Carbide 4.7 e−6 One element is then added next to the workpiece, the workpiece
Specific heat, cp (J/KgK) Ti6Al4V 580 width is still 1 element and the total width of the model is 4 elements
Carbide 203
Friction coefficient 0.2
(model called “1”). Figs. 5 (c) and 6 (c) present the chip for the model
Friction energy to heat (%) 100 with 3 elements next to the workpiece. No material goes outside of the
Inelastic heat fraction 0.9 mesh any more although a lateral expansion of the chip is noticed. The
Heat partition to workpiece (%) 50 maximal temperature is found in the secondary shear zone. This is in
accordance with the expected chip formation process.
In Fig. 7 (a) and (b), it is however noticed that no material is found
direction is the same, 5 μm, as in the cutting plane. The elements are
in the added elements and that almost no material is in the second
thus cube-shaped. This is performed to avoid influencing the results
layer of elements. The previous model with 2 elements next to the
with the shape of the elements.
workpiece (model called “1 narrow”) should therefore have given
The well-known, in metal cutting modelling, Johnson-Cook con-
results in which no material flows outside of the mesh. From this
stitutive model [23] is adopted to describe the behaviour of the
observation, it is concluded that the material entering in the elements
workpiece material, the Ti6Al4V titanium alloy:
at the boundary of the Eulerian mesh is attracted outside of the mesh.
ε̇ ⎞⎛ ⎡ T − Troom ⎤ ⎞
m
n ⎛
To avoid this phenomenon, it is necessary to add a supplementary
σ = (A + B ε ) ⎜1 + C ln ⎟ ⎜⎜1 − ⎢ ⎥ ⎟⎟ element that will stay full of void for the whole computation.
⎝ ε0̇ ⎠ ⎝ ⎣ Tmelt − Troom ⎦ ⎠ (1)
To increase the width of the workpiece for it to be closer to the
Parameters linked to the material properties are A, B, C, m and n. Tmelt experimental workpiece, the length of the elements is increased in the Z
and Troom are, respectively, the melting and the room temperatures. ε̇0 direction (perpendicular to the cutting plane) and it is checked if the
is the reference strain rate. In accordance with the work of Ducobu shape of the elements alters the results. Figs. 5(d) and 6(d) show the
et al. [24] on the choice of the Johnson-Cook set of parameters, the set only model in this study with an element length in the Z direction that
of parameters adopted in this study comes from Seo et al. [25]; it is is not 5 μm (model called “1, 50 μm”). The length adopted is 50 μm (so
provided in Table 4. The tool material is tungsten carbide and it is 10 times larger than the edges of the elements in the cutting plane) and
described by a linear elastic model. In accordance with the experi- 3 elements are added next to the workpiece. When visually comparing
mental study of Rech et al. [26] performed in the same conditions of the chip morphology to that of the previous model in Fig. 6 (c) (the only
this work (same cutting speed, dry cutting, Ti6Al4V machined material, difference between both models is thus the elements length outside of
WC/Co cutting tool with TiN coating), friction at the tool – chip the cutting plane), it is clear that the shape of the elements influences
interface is modelled as Coulomb's friction. Concerning the thermal the results. The lateral expansion of the chip, b′, increases dramatically
aspects, the faces of the model are adiabatic. The materials properties with the parallelepipedal elements and such a size is questionable.
of the numerical model are given in Table 4. The cutting conditions are From a general point of view, the lateral expansion of the chip
the same as for the experimental reference (Table 1). mainly occurs in the area in contact with the tool. This is due to the fact
that the material flows easily laterally when it is forced to deform by the
4. Comparison of the 3D orthogonal cutting model with the tool as nothing constraints its displacement in this direction. To
experimental and numerical references compare the order of magnitude of the lateral expansion of the chip
with the experimental values, it is estimated in the numerical model
The 3D model is compared to the experimental reference and the from the views provided in Fig. 7. The experimental values were
2D model in the same cutting conditions. The uncut chip thickness of measured on the whole chip width and consequently with a lateral
60 μm is adopted. The width of the 3D model (in the Z direction) is as expansion on both sides of the chip. To compare the same experimental
small as possible to reduce the number of nodes: 1 element for the and numerical expansions, both full chips needs to be considered.
workpiece and 2 elements next to it (model called “1 narrow”); the total Consequently, the numerical values are doubled to get the expansion
width of the model is thus 3 elements. The number of elements and on both sides of the chip as in the experiments.
nodes of both 2D and 3D models are given in Table 5, as well as the Table 6 summarizes the results. For the model with 50 μm elements
value of Hv. In all the models, the tool is as wide as the Eulerian mesh. length, the value is large, as well as the difference with the experimental
Both chips are plotted in a view perpendicular to the cutting plane value. For the model with cubic elements, the lateral expansion is
in Fig. 5 (a) and (b); temperature contours are shown. In such views closer to the reference. The experiments showed that the lateral
perpendicular to the cutting plane, temperature contours are selected expansion is very small by comparison to the width of the chip but
to control if the high temperature area is in the second shear zone. This that it is significant by comparison to the uncut chip thickness. The
will contribute to check if the fundamental aspects of metal cutting are ratio of the uncut chip thickness to the width of the tenon in the
correctly modelled. Fig. 6 shows a 3D view of the chips. For the 3D experiments is very large and the plane strain assumption can be
model (Fig. 6 (b)), it is clearly seen that a fraction of the material adopted, as in the 2D model. In both 3D models, this ratio is smaller
expands so much in the Z direction that it goes outside of the mesh. and it does not allow to adopt the plane strain assumption. As the

30

https://freepaper.me/t/461202 : ‫خودت ترجمه کن‬

F. Ducobu et al. Finite Elements in Analysis and Design 134 (2017) 27–40

Fig. 5. Temperature contours (in K) of the numerical chips (a) 2D model, (b) 3D model with 1 element of 5 μm in the width of the workpiece (3 elements in the width of the model), (c)
3D model with 1 element of 5 μm in the width of the workpiece (4 elements in the width of the model) and (d) 3D model with 1 element of 50 μm in the width of the workpiece (4
elements in the width of the model) for h=60 μm after 600 μs of cutting in a view perpendicular to the cutting plane.

width of the tenon (i.e. the workpiece) is very small in the models, it contours are different in the plane of symmetry and the lateral free
was expected to have b′b values larger than in the experiments, which is plane, confirming the advantage of the 3D configuration to better
noted in Table 6. The values of b′h can on the contrary be compared represent the actual cutting process. The uncertainty on the plane in
while keeping in mind that the plane strain conditions are not satisfied which the experimental value was measured needs to be pointed out.
in the 3D models. Only a qualitative comparison is therefore wise. The Indeed, as mentioned in the experimental reference section, the
value of b′h for the model with elements of 5 μm in the width, very close experimental chip thickness was measured on an embedded chip after
to the experimental one, indicates more that the order of magnitude of polishing. It is therefore reasonable to take the mean of both numerical
the lateral expansion is similar, than the lateral expansion is accurately values as the experimental thickness was measured somewhere be-
modelled. The very large value of b′h when elements are 50 μm long in tween the lateral face of the chip and its middle plane.
the width indicates that it is largely overestimated which was suspected The thicknesses are provided in Table 7. The 2D model gives a
from the chip morphology. The tendency of elongated elements to thicker chip than the 3D models. This was expected as no lateral
increase the lateral expansion is then confirmed. Cubic (“neutral”) expansion is allowed in 2D. In 3D and in the experiments, a lateral
elements will therefore be adopted for the whole study and are expansion of the chip is allowed which leads to a thinner chip; the
recommended to avoid any influence of the elements shape on the volume of the chip is conserved so a wider chip is thinner. When
results. Moreover when a predictive models is sought. comparing with the experimental chip, the 3D values are closer than
In the 3D numerical models, the measurement of the chip thickness the 2D one, in accordance with the previous observation. The chip
is not trivial as its value is different depending on the distance of the thickness values of both 3D chips are the same but their standard
plane in which it is measured with the plane of symmetry (Fig. 7). The deviations are different. It is smaller for the cubic elements. This means
adopted method consists in taking the mean of two thicknesses: the that the thickness of the chip in both planes of measurement are closer
first one in the plane of symmetry (Figs. 8 (b) and (d)) and the second than with the parallelepipedal elements. This was expected when
one in the plane of the lateral face of the workpiece (Figs. 8 (a) and (c)); looking at the lateral expansion of the chip with elongated elements
3 values are measured in each plane. For the 2D model, the thickness in Figs. 6 and 7.
is, of course, the same in both planes. In the 3D model, the temperature All the chips are continuous (Fig. 5). Forces (cutting force, CF, in

Fig. 6. Temperature contours (in K) of the numerical chips (a) 2D model, (b) 3D model with 1 element of 5 μm in the width of the workpiece (3 elements in the width of the model), (c)
3D model with 1 element of 5 μm in the width of the workpiece (4 elements in the width of the model) and (d) 3D model with 1 element of 50 μm in the width of the workpiece (4
elements in the width of the model) for h=60 μm after 600 μs of cutting in a 3D view showing the lateral chip expansion.

31

https://freepaper.me/t/461202 : ‫خودت ترجمه کن‬

F. Ducobu et al. Finite Elements in Analysis and Design 134 (2017) 27–40

Fig. 7. Numerical chips (a) and (b) 3D model with 1 element of 5 μm in the width of the workpiece (4 elements in the width of the model), (c) and (d) 3D model with 1 element of 50 μm
in the width of the workpiece (4 elements in the width of the model) for h=60 μm after 600 μs of cutting. (a) and (c) are seen from the end of the workpiece, (b) and (d) are seen from the
top of the workpiece.

Table 6 Table 7
Lateral expansion of the chip of the 3D models with 1 element in the width of the RMS cutting forces and chip thickness summary of the 3D models with 1 element in the
workpiece for h = 60 μm when comparing with the experimental reference, bx′ : ratio of b′ width of the workpiece for h=60 μm when comparing with the 2D model and the
to the length x. experimental reference, Δx : difference with the experimental values.

Case b′ bb′ bh′ Case CF ΔCF FF ΔFF h′ Δh′

Nber el. (μm) (% b) (% h) Nber el. (N/mm) (%) (N/mm) (%) (μm) (%)

Exp. 12.6 ± 15 1 21 Exp. 113 ± 2 – 44 ± 1 – 80 ± 4 –

1 11.4 ± 3 114 19 2D 114 −1 34 22 99 ± 0 −24
1, 50 μm 84.9 ± 31 85 142 1 123 −8 25 44 66 ± 2 18
1, 50 μm 109 4 30 33 66 ± 7 18

the X direction, feed force, FF, in the Y direction) with a nearly

constant value when the steady-state of the cutting process is reached model and the RMS experimental values. Both 2D and 3D evolutions
are therefore expected. Fig. 9 shows the forces evolutions obtained with are very close and confirm that the steady-state is reached for the
the 3D model with 1 element of 5 μm. They are compared to the 2D forces. The increase of the 3D forces is however slightly slower. This is

Fig. 8. Temperature contours (in K) of the numerical chips in 2 views perpendicular to the cutting plane (a) and (b) 2D model, (c) and (d) 3D model with 1 element of 5 μm in the width
of the workpiece (4 elements in the width of the model) for h=60 μm after 600 μs of cutting. (a) and (c) are seen from the free plane side, (b) and (d) are seen from the plane of symmetry
side.

32

https://freepaper.me/t/461202 : ‫خودت ترجمه کن‬

F. Ducobu et al. Finite Elements in Analysis and Design 134 (2017) 27–40

Table 8
RMS cutting forces and chip thickness summary of the 3D model with a low workpiece
and 1 element of 5 μm in its width for h=60 μm when comparing with the references and
the 3D model with the full workpiece, Δx: difference with the experimental values.

Case CF ΔCF FF ΔFF h′ Δh′

Nber el. (N/mm) (%) (N/mm) (%) (μm) (%)

Exp. 113 ± 2 – 44 ± 1 – 80 ± 4 –
2D 114 −1 34 22 99 ± 0 −24
1 123 −8 25 44 66 ± 2 18
1 low 122 −8 25 44 66 ± 2 18

although it is underestimated contrary to the 2D model. As shown by

the b′b indicator, the lateral expansion takes however too much
Fig. 9. Typical forces evolutions for h=60 μm. importance in the current 3D models. The width of the workpiece
should therefore be increased in order to get closer to the plane strain
explained by the lateral expansion of the chip which lengthens the time assumption and the conditions of the experimental tests. This will
needed for the stable contact length between the tool and the chip to dramatically increase the number of nodes of the model and it is
establish. Another difference with the 2D model is that the cutting force proposed to reduce them in the current model before increasing its
in 3D is slightly higher and the feed force in 3D is slightly lower, width.
increasing the difference with the experimental reference. This is
confirmed by the RMS values. The RMS cutting force in 3D is still
5. Reduction of the number of Eulerian nodes
accurately modelled while the feed force in 3D is more far from the
reference. The change in the forces is explained by the change in the
In order to decrease the number of nodes in the model, the height of
chip morphology. As it gets wider and thinner, the horizontal force
the area initially filled with void, Hv, will now be reduced. The chips
needed to form it becomes larger and the vertical one becomes smaller.
produced are continuous in the adopted cutting conditions. So, when
In conclusion of this comparison, the model developed produces a
the cutting regime is reached, the chip morphology and the forces
chip with a 3D morphology and reproduces faithfully the experimental
hardly change with time. It would consequently make sense to only
tests while taking advantage of the symmetry of the configuration. The
model the fraction of the chip close to the primary and secondary shear
lateral expansion of the chip is now modelled. The cutting force is
zones. A value of 60 μm =h for Hv is adopted (model called “1 low”) and
accurately estimated and the difference with the experimental refer-
the results are compared to the model with the initial Hv value of
ence is larger for the feed force. The RMS forces values are closer to the
300 μm. This leads to the reduction of the number of nodes to 21,175
experimental reference for the 2D model than the 3D one. The chip
(45,415 with the full workpiece) and of elements to 16,300 (35,500
thickness is lower than the 2D model which is explained by the
with the full workpiece). The reduction of the height of the workpiece
modelling of the lateral expansion in 3D. As the volume of the chip
will lead to a fraction of the material to flow out of the Eulerian mesh.
is the same, a wider chip ends in a smaller thickness. The 3D numerical
This is classic in Eulerian meshes such as in computational fluid
thickness is closer to the experimental reference than the 2D one,
dynamics and no difference should arise from it.

Fig. 10. Temperature contours (in K) of the numerical chips in 2 views perpendicular to the cutting plane (a) and (b) 3D model with 1 element of 5 μm in the width of the workpiece (4
elements in the width of the model), (c) and (d) low 3D model with 1 element of 5 μm in the width of the workpiece (4 elements in the width of the model) for h=60 μm after 600 μs of
cutting. (a) and (c) are seen from the free plane side, (b) and (d) are seen from the plane of symmetry side.

33

https://freepaper.me/t/461202 : ‫خودت ترجمه کن‬

F. Ducobu et al. Finite Elements in Analysis and Design 134 (2017) 27–40

Fig. 11. Numerical chips (a) and (b) 3D model with 1 element of 5 μm in the width of the workpiece (4 elements in the width of the model), (c) and (d) low 3D model with 1 element of
5 μm in the width of the workpiece (4 elements in the width of the model) for h=60 μm after 600 μs of cutting. (a) and (c) are seen from the end of the workpiece, (b) and (d) are seen
from the top of the workpiece.

Table 9 The chips morphologies of both models are qualitatively close in

Lateral expansion of the chip of the 3D model with a low workpiece and 1 element of Fig. 10. In the cutting plane (Fig. 10 (d)), the results obtained with the
5 μm in its width for h=60 μm when comparing with the experimental reference and the low workpiece are similar to that of a 2D plane strain Arbitrary
3D model with the full workpiece, bx′ : ratio of b′ to the length x.
Lagrangian-Eulerian (ALE) model with Eulerian boundaries
Case b′ bb′ bh′ [27,29,26,30], but it has the advantage of being in a 3D configuration
Nber el. (μm) (% b) (% h) identical to the experiments and to keep the history of the material in
the machined surface. As in the previous models, the maximal
Exp. 12.6 ± 15 1 21
temperature is in the secondary shear zone as expected. The thickness
1 11.4 ± 3 114 19
1 low 10.8 ± 3 108 18
values of both chips are equal as well as their standard deviation
(Table 8) and their lateral expansion (Fig. 11 and Table 9). Similarly,
both cutting and feed forces are equal.
Table 10 This 3D model with a low workpiece can therefore be used to model
Number of elements and nodes for the low 3D models with an increasing number of the 3D orthogonal cutting operation of the experiments if the full chip
elements of 5 μm in the width of the workpiece for h=60 μm. geometry is not requested. It provides the same results as a model with
Model Elements Nodes Hv
a high workpiece but with the benefit of significantly reducing the
number of nodes (by more than 50%).
3 low 16,300 21,175 60
5 low 32,600 38,115 60
6 low 48,900 55,055 60 6. Sensitivity of the results to the width of the workpiece
7 low 52,975 59,290 60
9 low 69,275 76,230 60
In order to come closer to the experimental workpiece, it is
24 low 122,250 131,285 60
necessary to increase the width of the workpiece in the model. To
limit the rise of the number of nodes, the low workpiece model
configuration will be adopted. The number of elements in the width

Fig. 12. Temperature contours (in K) of the numerical chips (a) low 3D model with 3 elements of 5 μm in the width of the workpiece, (b) low 3D model with 5 elements of 5 μm in the
width of the workpiece and (c) low 3D model with 6 elements of 5 μm in the width of the workpiece for h=60 μm after 600 μs of cutting in a view perpendicular to the cutting plane.

34

https://freepaper.me/t/461202 : ‫خودت ترجمه کن‬

F. Ducobu et al. Finite Elements in Analysis and Design 134 (2017) 27–40

Fig. 13. Temperature contours (in K) of the numerical chips (a) low 3D model with 7 elements of 5 μm in the width of the workpiece, (b) low 3D model with 9 elements of 5 μm in the
width of the workpiece and (c) low 3D model with 24 elements of 5 μm in the width of the workpiece for h=60 μm after 600 μs of cutting in a view perpendicular to the cutting plane.

Table 11 width.
RMS cutting forces and chip thickness summary of the low 3D models for an increasing Figs. 12 and 13 present the 6 chips obtained after 600 μs of
number of elements of 5 μm in the width of the workpiece for h=60 μm when comparing
simulation. They seem rather close and the high temperature area is
with the 2D model and the experimental reference, Δx: difference with the experimental
values. still the secondary shear zone. It seems however that the rigidity of the
chip increases with the number of elements in the width. When
Case CF ΔCF FF ΔFF h′ Δh′ comparing the two extreme cases in Fig. 12 (a) and Fig. 13 (c), it is
Nber el. (N/mm) (%) (N/mm) (%) (μm) (%) indeed noticed that the contact length is longer for the wider chip of
Fig. 13 (c). The thickness of the chip seems to increase as well.
Exp. 113 ± 2 – 44 ± 1 – 80 ± 4 –
2D 114 −1 34 22 99 ± 0 −24 The measured values of the chip thickness (Table 11) confirm the
3 low 126 −11 28 36 74 ± 1 8 visual observations. It increases with the width of the workpiece. As the
5 low 124 −10 29 33 77 ± 2 4 thickness was lower than the experimental reference for the workpiece
6 low 124 −9 30 31 78 ± 3 3 with 1 element in its width, it comes closer to the reference but it then
7 low 122 −8 31 30 79 ± 4 1
9 low 120 32 28 83 ± 4
oversteps it for the two wider workpieces. The larger value is still
−6 −4
24 low 114 −1 33 25 94 ± 9 −18 however lower than that of the 2D model. The forces values are
influenced by the chip morphology (as in the previous section),
showing that the phenomena are correctly captured by the model.
Table 12 When the width increases, the cutting force decreases and the feed
Lateral expansion of the chip of the low 3D models for an increasing number of elements force increases towards the reference values. For the wider workpiece,
of 5 μm in the width of the workpiece for h=60 μm when comparing with the
the cutting force is nearly equal to the reference value, while the feed
experimental reference, bx′ : ratio of b′ to the length x.
force is still underestimated by 25%, which is similar to the results of
Case b′ bb′ bh′ the 2D model. As the 3D model comes closer to the plane strain
Nber el. (μm) (% b) (% h) conditions, its results come closer to that of the 2D model.
Concerning the lateral expansion of the chip, the values shown in
Exp. 12.6 ± 15 1 21
Table 12 exhibit globally a decrease towards the experimental values.
3 low 32.4 ± 10 108 54
5 low 42.3 ± 17 85 70 The difference with the reference for the wider chip is still noticeable
6 low 38.3 ± 16 64 64 but the evolution is clear. The standard deviations are large. Except for
7 low 41.2 ± 14 59 69 the wider workpiece, their values are however close to that of the
9 low 37.1 ± 20 41 62 experiments. Figs. 14 and 15 and particularly the (b), (d) and (f) cases
24 low 16.3 ± 39 7 27
clearly show why the standard deviation values of h′ and b′ increase
with the width of the workpiece when looking at the chip morphology.
of the workpiece goes from 3 to 24 (i.e. a modelled width from 15 μm to It is clearly seen that the thickness of the chip is not constant through
120 μm). Table 10 presents the number of elements and nodes of the its width: it is thicker in the plane of symmetry and decreases towards
six models considered. It is interesting to highlight the benefit of using the lateral free plane. This confirms that it is necessary to measure the
a low workpiece for the number of nodes in the model. For example, thickness in both planes as the experimental measurement plane
when 6 elements are in the width of the workpiece, the number of position is unknown. For the lateral expansion, it is larger for the zone
nodes of the model (55,055) is only slightly higher than the model with of the chip in contact with the tool as previously observed.
the full height of the workpiece (45,415) but with only 1 element in the In conclusion, the global tendencies are clear for the results when
the width of the workpiece increases: except for the chip thickness, they

35

https://freepaper.me/t/461202 : ‫خودت ترجمه کن‬

F. Ducobu et al. Finite Elements in Analysis and Design 134 (2017) 27–40

Fig. 14. Numerical chips (a) and (b) low 3D model with 3 elements of 5 μm in the width of the workpiece, (c) and (d) low 3D model with 5 elements of 5 μm in the width of the
workpiece, (e) and (f) low 3D model with 6 elements of 5 μm in the width of the workpiece for h=60 μm after 600 μs of cutting. (a), (c) and (e) are seen from the end of the workpiece, (b),
(d) and (f) are seen from the top of the workpiece.

Fig. 15. Numerical chips (a) and (b) low 3D model with 7 elements of 5 μm in the width of the workpiece, (c) and (d) low 3D model with 9 elements of 5 μm in the width of the
workpiece, (e) and (f) low 3D model with 24 elements of 5 μm in the width of the workpiece for h=60 μm after 600 μs of cutting. (a), (c) and (e) are seen from the end of the workpiece,
(b), (d) and (f) are seen from the top of the workpiece.

all get closer to the experimental and numerical references. Increasing nodes it would require (roughly estimated at 470,000) is too large and
the width of the workpiece leads to a model closer to the plane strain it is currently not realistic to develop such a model.
assumption valid in the experiments and the 2D model. The evolution In order to obtain results with a model useful in an industrial
of the results for the 3D models makes therefore sense. The next step context, a compromise between the quality of the results and the
would consequently be to continue the increase of the width of the number of elements is required. The model with 6 elements in the
workpiece to reach that of the experimental workpiece. The number of width of the workpiece is a good candidate. The forces are rather well

36

https://freepaper.me/t/461202 : ‫خودت ترجمه کن‬

F. Ducobu et al. Finite Elements in Analysis and Design 134 (2017) 27–40

Fig. 16. Velocity contours (in mm/s) (a) direction –X, (b) direction Y, (c) direction Z and (d) equivalent plastic strains.

Fig. 19. Evolution of the temperature across the tool after 600 μs of cutting.
Fig. 17. Temperature contours (in K) of the tool after 600 μs of cutting.

Table 13
Number of elements and nodes and value of Hv for the 2D and 3D models for h=40 μm
and h=100 μm.

Model Elements Nodes Hv

40 μm, 2D 6,875 14,126 220

40 μm, 3D 32,750 37,697 40
100 μm, 2D 10,075 20,590 320
100 μm, 3D 97,500 105,903 100

7. Focus on some aspects of the chip formation

The model with 6 elements in the width of the workpiece for an

uncut chip thickness of 60 μm is studied more deeply in this section
with a focus on chip morphology and the lateral expansion of the chip.
Velocity contours are plotted in Figs. 16 (a)–(c). They contribute to
Fig. 18. Evolution of the temperature along the tool in the plane of symmetry after
600 μs of cutting. understand the chip shape and its lateral expansion. For the nodes
velocity in the cutting speed direction (Fig. 16 (a), direction –X), all the
estimated and the chip morphology, although not the closest to the nodes of the chip go in the same direction as the tool as it pushes them.
reference, is globally well reproduced. Moreover, the computed work- In the vertical direction (Fig. 16 (b), direction Y), all the nodes of the
piece has a width equal to half of the uncut chip thickness. It then chip have a movement in the upper direction. This means that the chip
provides information on a chip with a width equal to the uncut chip grows. In the third direction, outside of the cutting plane (Fig. 16 (c),
thickness value thanks to the symmetry of the configuration. This direction Z), the nodes velocity is positive near the cutting radius. This
provides an objective criterion to select the width of the workpiece for means that these nodes are pushed laterally outside of the cutting
the 3D model. plane, confirming the lateral expansion of the chip. Fig. 16 (d) shows

37

https://freepaper.me/t/461202 : ‫خودت ترجمه کن‬

F. Ducobu et al. Finite Elements in Analysis and Design 134 (2017) 27–40

Fig. 20. Temperature contours (in K) of the numerical chips (a) 2D model for h=40 μm, (b) low 3D model with 4 element of 5 μm in the width of the workpiece for h=40 μm, (c) 2D
model for h=100 μm and (d) low 3D model with 10 element of 5 μm in the width of the workpiece for h=100 μm after 600 μs of cutting in a view perpendicular to the cutting plane.

Table 14 cutting radius. Maximal temperatures are measured in the plane of

RMS cutting forces and chip thickness summary of the 3D models for h=40 μm and symmetry of the model, as expected. This is confirmed by the
h=100 μm when comparing with the 2D models and the experimental references, Δx: temperature evolution along (Fig. 18) and across (Fig. 19) the tool.
difference with the experimental values.
The temperature increases when getting closer to the edge radius and
Case CF ΔCF FF ΔFF h′ Δh′ closer to the plane of symmetry.
(N/mm) (%) (N/mm) (%) (μm) (%)

40, exp. 86 ± 2 – 40 ± 1 – 59 ± 5 – 8. Validation of the approach at the two other uncut chip
40, 2D 84 2 36 9 67 ± 1 −14 thicknesses of the experiments
40, 3D 93 −8 33 19 50 ± 2 15
100, exp. 174 ± 2 – 50 ± 1 – 135 ± 6 –
100, 2D 171 2 31 38 165 ± 1 −22 The 3D model of orthogonal cutting developed in this work has
100, 3D 184 −6 27 47 128 ± 7 5 considered only one cutting condition so far. For it to be useful in an
industrial context and to be predictive, it has to be able to handle other
cutting conditions. This section is intended to assure that results in
Table 15 other cutting conditions than that adopted to develop the model are
Lateral expansion of the chip of the 3D models for h=40 μm and h=100 μm when
still in accordance with the experimental reference.
comparing with the experimental references, bx′ : ratio of b′ to the length x.
The model with 6 elements in the width of the workpiece has been
Case b′ bb′ bh′ selected to model the orthogonal cutting with an uncut chip thickness
Nber el. (μm) (% b) (% h) of 60 μm. This corresponds to a model with a workpiece width equal to
half of the uncut chip thickness (30 μm). The 3D modelling of
40 μm exp. 0.3 ± 6 0 1 orthogonal cutting will now be extended to the two other uncut chip
40 μm 3D 27.2 ± 11 68 45
100 μm exp. 4 39
thicknesses of the experimental reference, h=40 μm and h=100 μm.
39.1 ± 16
100 μm 3D 66.5 ± 17 66 66 The 3D model configuration with the low workpiece is kept to reduce
the number of nodes. The same method as for h=60 μm is adopted and
the height of the area initially filled with void, Hv, is set to the uncut
the equivalent plastic strains. They are larger for the material near the chip thickness value. The number of nodes and elements are compared
cutting radius and the rake face. The plastic strains are even more large to that of the corresponding 2D models in Table 13.
for the material that has been laterally expanded. This has an influence Fig. 20 compares the chips morphologies between the 2D and the
on the quality of the machined surface: a fraction of this highly 3D models for the two uncut chip thicknesses. In all the cases, the
deformed material is on the exterior side of the workpiece. maximal temperatures area is in the secondary shear zone, showing
Fig. 17 presents the temperature contours of the tool. The CEL that the chip formation is correctly modelled. As expected from the
formulation in Abaqus v6.14 does not allow to take into account the results at h=60 μm and Fig. 20, the chip thicknesses values of the 3D
thermal transfers between the Eulerian (the workpiece) and the models given in Table 14 are smaller than for the 2D models. They are
Lagrangian (the tool) parts. Consequently, the increase of temperature closer to the experimental reference for the 3D models than the 2D
at the interface between both parts is due to the friction energy ones. The difference with the reference is similar to that at h=60 μm for
converted into heat The temperature on the tool is lower than what h=100 μm, while it is larger at h=40 μm.
could be expected and the time to reach its temperature steady state is Concerning the cutting forces RMS values (Table 14), although they
longer. The results presented on the tool temperatures has therefore to are close to the reference, the difference is larger than for the 2D
be analysed with that information in mind and should be considered as models as previously observed. The same is noted for the feed forces
a qualitative tendency and not quantitative temperatures values. The RMS values for which the difference with the experimental reference is
maximal temperature area is in the secondary shear zone and in the larger. These results are in accordance with that at h=60 μm.

38

https://freepaper.me/t/461202 : ‫خودت ترجمه کن‬

F. Ducobu et al. Finite Elements in Analysis and Design 134 (2017) 27–40

Fig. 21. Numerical chips (a) and (b) low 3D model with 4 element of 5 μm in the width of the workpiece for h=40 μm, (c) and (d) low 3D model with 10 element of 5 μm in the width of
the workpiece for h=100 μm after 600 μs of cutting. (a) and (c) are seen from the end of the workpiece, (b) and (d) are seen from the top of the workpiece.

Table 16 load of the supercomputer, 10 μs of simulation are computed on an

Computation time variations for 10 μs of simulation. Intel i7-5700 HQ 3.48 GHz processor for each model. The very small
variation of the stable time increment for a CEL model [20] guarantee
Model Time variation (%)
the representativeness of this short simulation time.
2D −73 The computation times evolve as expected: the computation
1 narrow −55 increases with the width of the workpiece (for a same uncut chip
1 −44 thickness and a same workpiece height: models “1 low” to “24 low”)
1, 50 μm −33
and with the height of the workpiece (for a varying uncut chip
1 low −68
3 low −38 thickness: models “6 low”, “40, 3D” and “100, 3D” or at the same
5 low −7 uncut chip thickness but with a low or a complete workpiece: models
6 low 0 “1” and “1 low’). The aspect ratio of the elements influences the
7 low 20 computation time: cube-shaped elements lead to a lower computation
9 low 32
time than parallelepipedal elements for the same number of elements
24 low 153
40 μm, 2D −78 and nodes (models ”1”’ and “1, 50 μm”). Finally, it is significantly lower
40 μm, 3D −36 for 2D models as their number of nodes is smaller.
100 μm, 2D −72
100 μm, 3D 65 10. Conclusions

A 3D model of orthogonal cutting has been introduced. It faithfully

The lateral chip expansion values (Table 15) are estimated by the
reproduces the experimental tests, with a thinner workpiece. The true
same method as at h=60 μm (Fig. 21). The numerical values are again
numerical reproduction of a 3D orthogonal cutting operation with its
larger than for the experiments. This was expected due to the same
experimental validation has not been carried out previously in the
configuration, still not entirely satisfying the plane strain assumption
literature. It is a first step towards the modelling of an industry-
valid in the experiments. An increasing tendency with the uncut chip
relevant operation. This work allowed to model more accurately
thickness was experimentally observed for b′h . It is as well for the
orthogonal cutting experiments by including phenomena occurring
modelling but the increase is less significant. All the indicators are
outside of the cutting plane, such as the lateral expansion of the chip.
therefore globally accurately captured for the 3D models by compar-
The main highlights of the study are:
ison with the experimental reference and the developed 3D model can

•
be validated to model the 3D orthogonal cutting of Ti6Al4V.
It is recommended to adopt cube-shaped elements for the Eulerian
mesh to avoid influencing the results. The machined material was
9. Computation time indeed attracted in the elongated direction of the elements. It is also
recommended to leave a row of elements in which the material will
The Abaqus simulations are run on a supercomputer, Dragon1. This not flow at the boundary of the Eulerian mesh to avoid an attraction
contributes to reduce the computation time but it becomes more effect of the material out of the mesh.
difficult to compare the computation times between two simulations. • The forces were accurately modelled and particularly the cutting
Indeed, for the same simulation, the computation time varies greatly force, which shows a difference with the experimental values of less
(of more than ± 20%) depending on the load of the supercomputer. As than 10%. An increase of the width of the workpiece brought the
this load cannot be controlled, it is not possible to obtain reliable and forces values closer to the reference. This is due to the uncut chip
h
reproducible computation times. To give an order of magnitude, for the thickness to chip width ratio, b , that gets closer to the plane strain
model with 6 elements in the width of the workpiece for an uncut chip assumption valid in the experiments and the 2D model.
thickness of 60 μm, around 13 h are needed to compute 600 μs with 4 • The lateral chip expansion was modelled and is the main difference
Intel SandyBridge 2.60 GHz processors. by comparison to a 2D model. Nodes velocity and equivalent plastic
Table 16 presents the variation of the computation time for the strains contributed to explain the chip morphology and its lateral
different models of this study. These variations are obtained by taking expansion. This lateral expansion of the chip however took too much
the model with 6 elements in the width of the workpiece for an uncut importance by comparison to the width of the workpiece and the
chip thickness of 60 μm as a reference. To avoid the influence of the uncut chip thickness but its value decreased with a wider workpiece,

39

https://freepaper.me/t/461202 : ‫خودت ترجمه کن‬

F. Ducobu et al. Finite Elements in Analysis and Design 134 (2017) 27–40

which is closer to the plane strain assumption. [9] T. Özel, I. Llanos, J. Soriano, P.-J. Arrazola, 3D finite element modelling of chip

• The evolutions of the model results with the width of the workpiece
formation process dfor machining Inconel 718: comparison of fe software predic-
tions, Mach. Sci. Technol. 15 (2011) 21–46.
showed that it is necessary to increase it to reduce the difference [10] T. Özel, D. Ulutan, Prediction of machining induced residual stresses in turning of
with the experimental reference. Increasing the width of the work- titanium and nickel based alloys with experiments and finite element simulations,
CIRP Ann. - Manuf. Technol. 61 (2012) 547–550.
piece however increased significantly the computation time due to [11] Y. Guo, D. Dornfeld, Finite element modeling of burr formation process in Drilling
the rise of the number of nodes of the model. 304 stainless steel, J. Manuf. Sci. Eng. 122 (2002) 612–619.
• To keep a reasonable number of nodes in the 3D model, it was [12] M. Abouridouane, F. Klocke, D. Lung, O. Adams, A new 3D multiphase FE model
for micro cutting ferritic-pearlitic carbon steels, CIRP Ann. - Manuf. Technol. 61
proposed to adopt a workpiece width equal to half of the uncut chip (2012) 71–74.
thickness. This gives results for a workpiece of a width equal to the [13] S.L. Soo, R.C. Dewes, D.K. Aspinwall, 3D FE modelling of high-speed ball nose end
uncut chip thickness as only half of it is modelled by taking milling, Int. J. Adv. Manuf. Technol. 50 (2010) 871–882.
[14] G. Fang, P. Zeng, Three-dimensional thermo-elastic-plastic coupled FEM simula-
advantage of the symmetry of the configuration. With an height of
tions for metal oblique cutting processes, J. Mater. Process. Technol. 168 (2005)
the domain in which the chip can form equal to the uncut chip 42–48.
thickness value, this ended in a model exploitable in an industrial [15] O. Pantalé, Modélisation et Simulation Tridimensionnelles de la Coupe des Métaux,
context and providing results with a satisfactory accuracy in 3D (Ph.D. thesis), Université de Bordeaux I, 1996.
[16] O. Pantalé, R. Rakotomalala, M. Touratier, An ALE three-dimensional model of
orthogonal cutting. orthogonal and oblique metal cutting processes, Int. J. Form. Process. 1 (1998)
371–388.
Acknowledgements [17] H. Kishawy, M. Elbestawi, Effects of process parameters on material side flow
during hard turning, CIRP Ann. - Manuf. Technol. 39 (1999) 1017–1030.
[18] H. Kishawy, A. Haglund, M. Balazinski, Modelling of material side flow in hard
Computational resources have been provided by the supercomput- turning, CIRP Ann. - Manuf. Technol. 55 (2006) 85–88.
ing facilities of the University of Mons (Dragon1/UMONS) and the [19] L. Illoul, P. Lorong, On some aspects of the CNEM implementation in 3D in order
to simulate high speed machining or shearing, Comput. Struct. 89 (2011) 940–958.
Consortium des Équipements de Calcul Intensif (CÉCI), funded by the [20] F. Ducobu, E. Rivière-Lorphèvre, E. Filippi, Application of the Coupled Eulerian-
Fonds de la Recherche Scientifique de Belgique (F.R.S.-FNRS) under Lagrangian (CEL) method to the modeling of orthogonal cutting, Eur. J. Mech. - A/
Grant No. 2.5020.11. François Ducobu gratefully acknowledges Solids 59 (2016) 58–66.
[21] F. Ducobu, E. Rivière-Lorphèvre, E. Filippi, Experimental contribution to the study
Sébastien Kozlowskyj, System Manager of Dragon1, for his help and
of the Ti6Al4V chip formation in orthogonal cutting on a milling machine, Int. J.
support to run Abaqus on the cluster. Mater. Form. 8 (2015) 455–468.
[22] SECO TOOLS, Turning catalog and technical guide 2012, SECO TOOLS AB, 2011.
[23] G. Johnson, W. Cook, A constitutive model and data for metals subjected to large
References
strains, high strain rates and high temperatures, in: Proceedings of the Seventh
International Symposium on Ballistics, The Hague, The Netherlands, 1983, pp.
[1] O. Pantalé, J.-L. Bacaria, O. Dalverny, R. Rakotomalala, S. Caperaa, 2D and 3D 541–547.
numerical models of metal cutting with damage effects, Comput. Methods Appl. [24] F. Ducobu, E. Rivière-Lorphèvre, E. Filippi, On the importance of the choice of the
Mech. Eng. 193 (2004) 4383–4399. parameters of the Johnson-Cook constitutive model and their influence on the
[2] P. Arrazola, T. Özel, D. Umbrello, M. Davies, I. Jawahir, Recent advances in results of a Ti6Al4V orthogonal cutting model, Int. J. Mech. Sci. 122 (2017)
modelling of metal machining processes, CIRP Ann. - Manuf. Technol. 62 (2013) 143–155.
695–718. [25] S. Seo, O. Min, H. Yang, Constitutive equation for Ti-6Al-4V at high temperatures
[3] C.V. Luttervelt, T. Childs, I. Jawahir, F. Klocke, P. Venuvinod, The State-of-the-Art measured using the SHPB technique, Int. J. Impact Eng. 31 (2005) 735–754.
of modelling in machining processes, CIRP Ann. - Manuf. Technol. 47 (1998) [26] J. Rech, P. Arrazola, C. Claudin, C. Courbon, F. Pusavec, J. Kopac, Characterisation
587–626. of friction and heat partition coefficients at the tool-work material interface in
[4] Y. Guo, C. Liu, 3D FEA modeling of hard turning, J. Manuf. Sci. Eng. 124 (2002) cutting, CIRP Ann. - Manuf. Technol. 62 (2013) 79–82.
189–199. [27] T. Özel, E. Zeren, Numerical modelling of meso-scale finish machining with finite
[5] J. Aurich, H. Bil, 3D finite element modelling of segmented chip formation, Ann. edge radius tools, Int. J. Mach. Mach. Mater. 2 (2007) 451–768.
CIRP 55 (2006) 47–50. [28] M. Nasr, E.-G. Ng, M. Elbestawi, Effects of workpiece thermal properties on
[6] R. Li, A.J. Shih, Finite element modeling of 3D turning of titanium, Int. J. Adv. machining-induced residual stresses - thermal softening and conductivity,
Manuf. Technol. 50 (2005) 825–833. Proceedings of the Institution of Mechanical Engineers, Part B: Journal of
[7] P. Arrazola, T. Özel, Numerical modelling of 3d hard turning using arbitrary Engineering Manufacture, 221, 2007, pp. 1387–1400.
lagrangian eulerian finite element method, Int. J. Mach. Mach. Mater. 3 (2008) [29] P. Arrazola, T. Özel, Investigations on the effects of friction modeling in finite
238–249. element simulation of machining, Int. J. Mech. Sci. 52 (2010) 31–42.
[8] T. Özel, Computational modelling of 3D turning: influence of edge micro-geometry [30] F. Ducobu, P.-J. Arrazola, E. Rivière-Lorphèvre, E. Filippi, Finite element predic-
on forces, stresses, friction and tool wear in PcBN tooling, J. Mater. Process. tion of the tool wear influence in Ti6Al4V machining, Procedia CIRP 31 (2015)
Technol. 209 (2009) 5167–5177. 124–129.

40

https://freepaper.me/t/461202 : ‫خودت ترجمه کن‬