International Journal of Machine Tools & Manufacture 44 (2004) 933–944

www.elsevier.com/locate/ijmactool

A comparison of orthogonal cutting data from experiments

with three different finite element models
Halil Bil
, S. Engin Kılıç, A. Erman Tekkaya
Department of Mechanical Engineering, Middle East Technical University, Ankara, Turkey

Received 6 October 2003; received in revised form 6 January 2004; accepted 21 January 2004

Abstract

The aim of this study is to compare various simulation models of orthogonal cutting process with each other as well as with the
results of various experiments. Commercial implicit finite element codes MSC.Marc, Deform2D and the explicit code Thirdwave
AdvantEdge have been used. In simulations, a rigid tool is advanced incrementally into the deformable workpiece which is reme-
shed whenever needed. In simulations with MSC.Marc and Thirdwave AdvantEdge, there is no separation criterion defined since
chip formation is assumed to be due to plastic flow, therefore, the chip is formed by continuously remeshing the workpiece. How-
ever, in simulations with Deform2D, the Cockroft–Latham damage criterion is used and elements, which exceed the predefined
damage value, are erased via remeshing. Besides this different modeling of separation, the three codes also apply different friction
models and material data extrapolation schemes. Estimated cutting and thrust forces, shear angles, chip thicknesses and contact
lengths on the rake face by three codes are compared with experiments performed in this study and with experimental results sup-
plied in literature. In addition, effects of friction factor, different remeshing criteria, and threshold tool penetration value on the
results are examined. As a result, it has been found that although individual parameters may match with experimental results, all
models failed to achieve a satisfactory correlation with all measured process parameters. It is suggested that this is due to the
poor modeling of separation.
# 2004 Elsevier Ltd. All rights reserved.

Keywords: Finite element method; Orthogonal metal cutting; Remeshing; Damage; Friction

1. Introduction and Iwata et al. [7] who analyzed the steady-state

orthogonal cutting. Chip formation and separation
Cutting is one of the most important and common from the workpiece was first attempted by Strenkowski
manufacturing processes in industry. Simplified analyti- and Carroll [8]. Chip formation was achieved by vari-
cal methods have been developed, for example, by ous separation criteria like distance tolerance criterion
Merchant [1] who introduced the concept of shear [9], strain energy density criterion [10,11], and fracture-
angle and by Lee and Shaffer [2] who proposed an ana- mechanics based criterion [12]. To deal with large
lytical model using slip-line field theory. Later, more element distortion in metal cutting simulation, Shih
complicated models [3–5] have been developed which and Yang [13] and Shih [14,15] developed a mesh
also consider the effects of friction, work-hardening, rezoning technique. Ceretti et al. [16] developed a cut-
and temperature. These successful models gave useful ting model by deleting elements having reached a
insight into the mechanics of cutting. critical value of accumulated damage.
In recent years, finite element analysis has become In almost all early finite element models developed,
the main tool for simulating metal cutting processes. non-commercial finite element codes have been
Early analyses were made by Usui and Shirakashi [6] employed. However, application of commercial finite
element codes is preferable for industrial utilization.


Corresponding author. Tel.: +90-312-2105263; fax: +90-312-
Recently, some commercial finite element codes offer
2101266. the modeling of cutting. As these codes are based on
E-mail address: hbil@metu.edu.tr (H. Bil). different algorithms, it is not known if and by what
0890-6955/$ - see front matter # 2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijmachtools.2004.01.016
934 H. Bil et al. / International Journal of Machine Tools & Manufacture 44 (2004) 933–944

degree results from each of them will differ. The aim of

this study is to evaluate these available models for
machining.
The commercial codes MSC.Marc, Deform2D and
Thirdwave AdvantEdge have been used to create a
coupled thermo-mechanical finite element model of
plane-strain orthogonal metal cutting operations.
Material is modeled as elastic–plastic, with flow stress
being dependent on strain, strain rate and temperature.
The friction between the tool and chip is of shear type
Fig. 2. General geometry of the finite element model with
for MSC.Marc and Deform2D, whereas it is of MSC.Marc.
Coulomb type for Thirdwave AdvantEdge. In simula-
tion with MSC.Marc and Thirdwave AdvantEdge, no
damage or failure criteria are defined assuming the for-
mation of chip due to plastic flow. However, in simula-
tions with Deform2D, the Cockroft–Latham damage
criterion has been used, although the plastic flow
model is also available in this code.
Experiments have been performed to verify the simu-
lation results. Purely orthogonal cutting operations can
be done on a shaping machine with a single point cut-
ting tool or on a lathe by cutting a hollow cylinder
with a large diameter and small wall thickness from the Fig. 3. General geometry of the finite element model with
Deform2D.
end. In this study, the latter method is used. Cutting
and thrust forces are measured by means of a dyna-
mometer. In addition, contact length and chip thick- boundary conditions; however, MSC.Marc cannot han-
ness are measured and shear plane angles are dle this when global remeshing is enabled.
calculated from the measured chip thicknesses. The finite element model of Deform2D is exactly the
same as MSC.Marc except for the satisfaction of the
2. Finite element models condition on the left and bottom boundaries (Fig. 3)
since in Deform2D, they are defined as boundary con-
The finite element model is composed of a deform- ditions and not satisfied by means of some geometric
able workpiece and a rigid tool. The tool penetrates entities.
through the workpiece at a constant speed and con- On the other hand, AdvantEdge is an automated
stant feed rate. The model assumes plane-strain con-
program and it is enough to input process parameters
dition since generally depth of cut is much greater than
to make a two-dimensional simulation of orthogonal
the feed rate.
cutting operation. The boundary conditions are hidden
The workpiece is discretized by bilinear four-noded
to the user. Fig. 4 shows the model of Thirdwave
quadrilateral elements in MSC.Marc and in Deform2D.
However, Thirdwave AdvantEdge uses six-noded AdvantEdge.
quadratic triangular elements by default. Fig. 1(a),(b) Friction between the chip and tool interface is mod-
shows the elements used. eled as constant shear in MSC.Marc and Deform2D.
Left and bottom boundary nodes of the workpiece On the other hand, Thirdwave AdvantEdge uses the
are fixed in both x and y directions by gluing them to a Coulomb friction model.
rigid curve (Fig. 2). This could also be done by defining
3. Workpiece and tool material modeling

The workpiece material used for the plane-strain

orthogonal metal cutting simulation is C15 steel. Its
flow curve is represented by several tabulated data
(MSC.Marc database), which depends on strain, strain
rate and temperature. These data represent the material
flow curve at three different strain rates (1.6, 8 and 40
1/s) and 13 different temperatures (20, 100, 200, 300,
v
Fig. 1. Elements used in numerical modeling. (a) By MSC.Marc and 400, 500, 600, 700, 800, 900, 1000, 1100, 1200 C),
Deform2D. (b) By Thirdwave AdvantEdge. while strain changes between 0 and 1. A sample flow
 H. Bil et al. / International Journal of Machine Tools & Manufacture 44 (2004) 933–944 935

Fig. 6. Flow curves at different strain rates and at a constant tem-

v
perature of 500 C.

Fig. 6 shows three flow curves at different strain

Fig. 4. Thirdwave AdvantEdge model. rates and at a constant temperature, which gives an
idea of error introduced due to improper flow stress
curve at 40 1/s strain rate can be seen in Fig. 5. Fur- evaluation at elevated strain rates.
ther details are given in Ref. [26]. AdvantEdge, finally, uses an analytical formulation
In the cutting process, the deformation at the cutting for material modeling. In a typical machining event, in
zone takes place at elevated temperatures and strain the primary and secondary shear zones very high strain
rates. For example, in the simulations performed, the rates are achieved, while the remainder of the work-
v
temperature reaches above 500 C and the strain rates piece deforms at moderate or low strain rates. In order
are in the order of 104 1/s. Therefore, extrapolation is to account for this, Thirdwave AdvantEdge incorpo-
necessary. Given flow curves are not extrapolated with rates a stepwise variation of the rate sensitivity
MSC.Marc but evaluated at the limit of the available exponent:
data range. Deform2D, on the other hand, extrapolates  
e_ p 1=m1
using the last two data points. In this case, when the
¼ rf ðep Þ  1 þ p
r ; if e_ p  e_ pt ð1Þ
e_ 0
strain rate exceeds 40 1/s, some error is introduced due
to improper flow stress evaluation. This extrapolation    
e_ p 1=m2 e_ t 1=m1
is a weak point of the finite element models used in this r p

¼ rf ðe Þ  1 þ p  1þ p ; if e_ p  e_ pt
e_ 0 e_ 0
study.
ð2Þ
where r
is the effective von Mises stress, rf is the flow
stress, ep is the accumulated plastic strain, e_ p0 is a refer-
ence plastic strain rate, m1 and m2 are low and high
strain-rate sensitivity exponents, respectively, and e_ t is
the threshold strain rate which separates the two
regimes. In calculations, a local Newton–Raphson itera-
tion is used to compute e_ p0 according to the low-rate
equation, and switches to the high rate equation if the
result lies above e_ t .
rf, which is used in Eqs. (1) and (2) is given as:
 
ep 1=n
rf ¼ r0  wðTÞ  1 þ p ð3Þ
e0
where T is the current temperature, r0 is the initial
yield stress at the reference temperature T0, ep0 is the
reference plastic strain, n is the hardening exponent
Fig. 5. Workpiece flow curve for strain rate of 40 (1/s). and w(T) is the thermal softening factor.
936 H. Bil et al. / International Journal of Machine Tools & Manufacture 44 (2004) 933–944

In the present study, it is assumed that the tool is the separation of material when the tool advances
not plastifying. Hence, it is considered as rigid. Heat through the workpiece. In Deform2D, a damage cri-
can be transferred to the tool only from the workpiece. terion is used for material separation. Therefore, ele-
ments at the tool tip are erased during remeshing when
they have reached a critical damage value. A normal-
4. Chip formation ized Cockroft–Latham damage criterion has been used.
The critical damage value has been taken as 0.2 for
In recent years, formation of a continuous chip is
C15 steel. It should be re-emphasized that Deform2D
assumed to be due to plastic flow and to simulate chip
formation, the workpiece is continuously remeshed, offers the option of remeshing as well. But in this
e.g. [17–20]. The present study uses the same solution study, the aim was also to investigate the separation
for chip formation in simulations with MSC.Marc and model.
Thirdwave AdvantEdge. Thirdwave AdvantEdge also uses remeshing to form
In the model analyzed with MSC.Marc, the work- the chip. It remeshes the workpiece periodically to
piece is remeshed whenever a predefined threshold refine large elements, remesh distorted elements, and
value of tool penetration occurs. Therefore, a new coarsen small elements. There is no user control for the
boundary for the workpiece at the tool–chip interface remeshing process.
is determined and the workpiece is remeshed according
to it forming the chip. The default threshold value of
tool penetration is two times the contact tolerance 5. Heat transfer
value, which is by default 0.05 times the minimum
element edge length. In addition, a penetration check The sources of heat generation are the plastic work
can be selected to be performed at every iteration or at in the cutting zone and friction at the tool–chip inter-
the end of increments. face. At the same time, workpiece loses heat to the
Although a varying threshold penetration value environment due to convection and to tool due to con-
(default, 0.0001 and 0.0005 mm was used where the
duction.
average element edge length is 0.01 mm), which is used
The rate of specific volumetric flux due to plastic
as a trigger for remeshing the workpiece, does not
work is given by the formula:
introduce a significant difference, the time of pen-
etration check has a strong effect on the results. Cut- fjW_ P
ting force results, obtained when the penetration check q_ ¼ ð4Þ
q
is done at every iteration, is almost half of those
where W _ P is the rate of plastic work, j is the fraction
obtained when the check is done at the end of incre-
ments (Fig. 7). In this study, the penetration check is of plastic work converted into heat which is taken as
done at the end of increments. 0.9, f is the mechanical equivalent of heat (taken as 1.0
The same simulations have also been done with by default) and q is the density of workpiece material
Deform2D. The basic difference between the models of (7.85 g/cm3).
MSC.Marc and Deform2D is the assumption made for The rate of heat generated due to friction is given by
the formula:
_ ¼ Ffr  vr  f
Q ð5Þ
where Ffr is the friction force, vr is the relative sliding
velocity between tool and chip, and f is the mechanical
equivalent of heat (f ¼ 1:0).
The generated heat due to friction is given to each of
the two contacting bodies, which are chip and tool in
this case, by equal proportions.
The workpiece loses heat to the environment due to
convection according to the formula:
qh ¼ h  ðTw  T0 Þ ð6Þ
where h is the convection heat transfer coefficient of
v
the workpiece (0.4 N/(mm C)), Tw is the workpiece
Fig. 7. Comparison of cutting force results obtained with different surface temperature, and T0 is the ambient temperature
v
penetration check times. (20 C).
 H. Bil et al. / International Journal of Machine Tools & Manufacture 44 (2004) 933–944 937

6. Experiments of the process variables. In this study, such three mod-

els are used as a preliminary validation of the results
In this study, experiments were done by cutting a obtained from finite element simulations.
hollow cylinder from the end on a lathe. In this experi- Slip-line models by Lee and Shaffer [2], Kudo [21],
ment, the diameter of the workpiece should be rela- and Dewhurst [22] give an allowable range for the
tively larger than the depth of the cut (wall thickness of possible solutions of the relation between the shear
the tube) to satisfy orthogonal cutting condition. plane angle, /, the rake angle of the tool, c and the
Workpiece material was C15 steel with the inner and chip–tool friction angle, k. In this study, diagrams have
outer diameters being 56 and 53.1 mm, respectively. been drawn with (/ – c) against k according to Hill’s
Therefore, depth of the cut is 1.45 mm. Cutting tools [23] overstressing criterion as suggested by Childs [24].
were made of high speed steel and had rake angles of Fig. 8(a),(b) shows allowable ranges for the rake angles
v v v
20 and 25 , and 5 clearance angle. Cutting speed was v v
of 20 and 25 , respectively.
125 rev/min. Fig. 8 also shows the simulation results from
A lathe tool dynamometer was used to measure cut- MSC.Marc. All the results are between the upper and
ting and thrust forces. Calibration of the device was lower bounds suggested by Kudo and Dewhurst,
made before the cutting experiments by applying
respectively. The graph also shows that, Lee and
known weights progressively.
Shaffer model, which is relatively simpler than the oth-
In addition to the forces, the thickness of cut chips
ers, gives good predictions. Especially for the rake
and the contact length on the rake face were measured v
angle of 20 , simulated results are almost coincident
by means of a tool maker’s microscope. Shear angles
with the suggested solution of Lee and Shaffer.
were calculated from the measured chip thicknesses.

7.2. Chip geometry

7. Results
Table 2 shows the experimental results for the thick-
Finite element simulations were carried out for four ness of the cut chips and contact lengths between the
cases of experiments composed of two different rake
angles and two different feed rates with MSC.Marc,
Deform2D and Thirdwave AdvantEdge separately to
examine the effects of different parameters. Table 1
shows the different parameters of experiments.
In all simulations, it is made sure that steady-state
has been reached and some more data are collected
after that time. Therefore, all the results presented in
this work were gathered under steady-state conditions.
From the simulations, variables like stresses, strains,
strain rates and temperature distributions can be
obtained. However, all these are very difficult to mea-
sure experimentally. On the other hand, cutting and
thrust forces, chip thickness and contact length can be
measured relatively easily. Also experimental results
from the literature are compared with the simulation
results.
7.1. Analytical comparison

Although the closed form analytical solutions of cut-

ting processes are impossible, there are analytical mod-
els defining allowable ranges for the possible solutions

Table 1
Varied parameters in the experiments

Rake angles (deg) Feed rates (mm/rev)

0.05 0.1
20 Exp. 1 Exp. 2
Fig. 8. Allowable slip-line model solutions by Hill’s overstressing
25 Exp. 3 Exp. 4 v v
criterion. (a) For rake angle of 20 . (b) For rake angle of 25 .
938 H. Bil et al. / International Journal of Machine Tools & Manufacture 44 (2004) 933–944

Table 2 ness deviates from the experimentally measured values.

Experimental results of chip geometry parameters Contact lengths, on the other hand, are very poorly
c (deg) f (mm) tc (mm) / (deg) lc estimated. But here it should be taken into account
20 0.05 0.10 29.54 0.85
that the measurement may be overestimated up to 50%
0.1 0.27 21.73 1.10 due to the primitive measurement method.
25 0.05 0.12 24.62 1.00 Table 4 shows the results from Deform2D. Chip
0.1 0.21 28.24 0.90 thicknesses are not very well estimated but again the
results are best when the friction factor is 0.7. Since
chip thicknesses are poorly estimated, the shear angles,
chip and rake face of the tool, where c is the rake which are calculated from them, are also not in agree-
angle, f is the feed rate (uncut chip thickness), tc is the ment with the experiments. On the other hand, strain-
chip thickness, / is the shear angle and lc is the contact rate distribution gives good results only for the case
v
length. Shear angles are calculated from the measured where rake angle is 25 . Contact lengths are again very
chip thicknesses by using Eq. (7). However, experimen- poorly estimated. Another point, which is worth men-
tally measured contact length results should be con- tioning here, is that the results from Deform2D are not
sidered carefully since the measurements were taken on consistent between each other and they are not stabi-
v
a conventional lathe and not on a quick-stop machine. lized. For example, for a rake angle of 20 and feed
There may be up to 50% error in the measurements. rate of 0.05 mm/rev, best result of shear angle from
rc  cosc strain-rate distribution is got when the friction factor is
tanu ¼ ð7Þ v
0.4. However, for a rake angle of 25 and feed rate of
1  rc  sinc
0.05 mm/rev, the same shear angle is best estimated
where rc is the chip thickness ratio. when the friction factor is 0.7.
Simulations of these experiments were carried out Table 5 shows the results from Thirdwave Advant-
with the three codes at different friction factors and Edge. The results of both chip thickness and shear
friction coefficients. angle are in very good agreement with the experiments.
Table 3 shows the results from MSC.Marc. Compar- As is mentioned before, this program uses the Coulomb
ing these results with the experimental results given in friction model, and the chip geometry results are best
Table 2, it can be concluded that for m ¼ 0:7 estimation estimated when the friction coefficient is taken as 0.5.
of chip thickness is best. Shear angle was found by two In view of the above results, it can be concluded
ways. Firstly, since in the shear zone, strain rate is sig- that, the three commercial code estimates the chip
nificantly larger than the remaining parts of the work- thickness, and shear angle quite good if an appropriate
piece, plotting the equivalent strain-rate distribution friction factor or coefficient is used. However, the esti-
supplies the location of shear zone and hence shear mation of contact length between the chip and rake
angle. Secondly, it can be calculated from the estimated face of the tool is not in agreement with the experi-
chip thicknesses by using Eq. (7). Results show that, ments for all the codes. However, contact length results
although both approaches are in good agreement, the should be considered carefully since implemented
strain-rate distribution approach gives a better esti- experimental measurement technique may lead to
mation of the shear angle, since estimated chip thick- errors up to 50%.

Table 3
Chip geometry results obtained by MSC.Marc

c (deg) f (mm) tc (mm) / (strain rate) (deg)

m ¼ 0:2 m ¼ 0:4 m ¼ 0:7 m ¼ 0:2 m ¼ 0:4 m ¼ 0:7
20 0.05 0.08 0.09 0.09 31.62 28.01 28.45
0.1 0.16 0.18 0.22 30.77 30.97 25.61
25 0.05 0.07 0.08 0.11 34.83 32.39 27.73
0.1 0.15 0.18 0.22 33.63 29.01 27.39

/ (chip thickness) (deg) lc

m ¼ 0:2 m ¼ 0:4 m ¼ 0:7 m ¼ 0:2 m ¼ 0:4 m ¼ 0:7
20 0.05 37.65 32.66 32.45 0.06 0.08 0.12
0.1 37.20 32.45 26.83 0.09 0.11 0.22
25 0.05 40.60 36.22 28.38 0.05 0.05 0.30
0.1 40.07 34.32 27.68 0.10 0.15 0.22
 H. Bil et al. / International Journal of Machine Tools & Manufacture 44 (2004) 933–944 939

Table 4
Chip geometry results obtained by Deform2D

c (deg) f (mm) tc (mm) / (strain rate) (deg)

m ¼ 0:2 m ¼ 0:4 m ¼ 0:7 m ¼ 0:2 m ¼ 0:4 m ¼ 0:7
20 0.05 0.07 0.08 0.09 32.74 29.60 33.69
0.1 0.12 0.13 0.14 38.74 37.41 33.69
25 0.05 – 0.07 0.08 – 29.93 26.57
0.1 0.12 0.12 0.13 33.42 31.36 29.28

/ (chip thickness) (deg) lc

m ¼ 0:2 m ¼ 0:4 m ¼ 0:7 m ¼ 0:2 m ¼ 0:4 m ¼ 0:7
20 0.05 44.45 39.06 34.68 0.06 0.07 0.08
0.1 48.27 43.28 42.99 0.11 0.13 0.14
25 0.05 – 44.65 37.59 – 0.06 0.09
0.1 51.25 49.38 46.26 0.12 0.11 0.12

Fig. 9(a)–(c) shows simulated chip geometry from measuring the angle between the horizontal and the
MSC.Marc, Deform2D and Thirdwave AdvantEdge, line connecting the two nodes with the highest equiva-
respectively. All the pictures were drawn for the same lent plastic strain-rate values on the free surface of the
tool stroke, so that, they are comparable with each workpiece and the tool tip, is plotted for comparison
other. Cutting conditions are as the following: rake purposes. Numerical values are already given in
v v
angle is 20 , clearance angle is 5 , feed rate is 0.1 mm/ Tables 3–5. This comparison reveals that, MSC.Marc
rev, depth of cut is 1.45 mm and cutting speed is 125 and Thirdwave AdvantEdge again supply similar
rev/min. Contour lines indicate the equivalent plastic results, whereas Deform2D simulates a larger shear
angle although the difference is small.
strain-rate distribution.
Fig. 10, on the other hand, shows simulated chip
geometries from three codes when the rake angle has 7.3. Cutting forces
v
been changed to 25 , while the other cutting conditions
Fig. 11 shows the effect of friction on cutting force
were kept constant. obtained by MSC.Marc, in comparison with experi-
Figs. 9 and 10 show that, MSC.Marc and Thirdwave mental results. It can be stated that, contrary to the
AdvantEdge give almost the same chip shapes. On the chip geometry variables, cutting forces are best esti-
other hand, Deform2D computes a more curled chip. mated when the frictional shear factor is small. When
From contact length measurements, it can be seen that the coefficient of shear friction is taken as 0.1, which is
Deform2D simulates smaller values than the other two the smallest value used in the simulations, cutting
commercial codes, which is the reason of more curled forces are estimated with a maximum error of 4%.
chip. On the figures, equivalent plastic strain-rate dis- Fig. 12 shows estimated cutting forces using
tribution, from which shear angle can be estimated by Deform2D. First of all it should be noted that, the

Table 5
Chip geometry results obtained by Thirdwave AdvantEdge

c (deg) f (mm) tc (mm) / (strain rate) (deg)

l ¼ 0:2 l ¼ 0:4 l ¼ 0:5 l ¼ 0:2 l ¼ 0:4 l ¼ 0:5
20 0.05 0.09 0.103 0.11 27.73 27.39 27.78
0.1 0.18 0.211 0.23 31.04 26.39 26.15
25 0.05 0.09 0.10 0.10 32.21 31.18 26.50
0.1 0.18 0.20 0.21 32.86 29.22 30.25

/ (chip thickness) (deg) lc

l ¼ 0:2 l ¼ 0:4 l ¼ 0:5 l ¼ 0:2 l ¼ 0:4 l ¼ 0:5
20 0.05 32.92 28.59 26.82 0.07 0.08 0.13
0.1 32.02 27.97 25.89 0.13 0.17 0.24
25 0.05 34.72 30.71 29.53 0.06 0.06 0.12
0.1 33.93 30.24 28.35 0.13 0.15 0.17
940 H. Bil et al. / International Journal of Machine Tools & Manufacture 44 (2004) 933–944

v
Fig. 9. Chip geometries for rake angle of 20 : (a) MSC.Marc, (b)
Deform2D, (c) Thirdwave AdvantEdge (m ¼ 0:7 for MSC.Marc and
Deform2D and l ¼ 0:5 for Thirdwave AdvantEdge). v
Fig. 10. Chip geometries for rake angle of 25 : (a) MSC.Marc, (b)
Deform2D, (c) Thirdwave AdvantEdge (m ¼ 0:7 for MSC.Marc and
v Deform2D and l ¼ 0:5 for Thirdwave AdvantEdge).
simulation, where rake angle is 25 and feed rate is
0.05 mm/rev could not be completed due to repeated
values obtained by the two other codes. The reason for
software crash. It can be seen that, changing friction
this is that Thirdwave AdvantEdge makes use of the
factor does not affect the results as much as it does in Coulomb friction law, which supplies regardless of the
the simulations of MSC.Marc. However, again, cutting magnitude of interface pressure a frictional stress that
forces are best estimated when the friction factor is 0.1. is proportional to this pressure. However, the pressure
Fig. 13 shows cutting force results from Thirdwave values in these examples are very high, yielding fric-
AdvantEdge. These force values are higher than the tional shear stresses which are larger than the shear
 H. Bil et al. / International Journal of Machine Tools & Manufacture 44 (2004) 933–944 941

Fig. 11. Verification of predicted cutting force results obtained by Fig. 14. Comparison of cutting force results from all codes. m¼ 0:7
MSC.Marc. for MSC.Marc and Deform2D and l¼ 0:5 for Thirdwave Advan-
tEdge.

7.4. Thrust forces

From Fig. 15, it can be seen that, for small friction

factors the tool appears to be drawn into the work-
piece. The reason for this is that, for small friction
forces (referring to the sketch on Fig. 15), the vertical
component of normal force, (Fn)y, exceeds the normal
component of friction force, (Ff)y, hence the tool is
pushed into the workpiece. When the friction factor is
taken as 0.7, the results of thrust force switches to the
Fig. 12. Verification of predicted cutting force results obtained by
positive values. However, there are still significant
Deform2D.
errors in the results.
Fig. 16 shows the simulated thrust force results
obtained using Deform2D. It is seen that the same
negative values also exists here. However, the thrust
forces never switch to the positive values, even when
the friction factor is increased to 0.7. This means that
(Fn)y is always computed as larger than (Ff)y.

Fig. 13. Verification of predicted cutting force results obtained by

Thirdwave AdvantEdge.

strength of the material. Hence, the supplied frictional

stress is not realistic for the friction coefficient selected.
A quantitative discussion of this phenomenon will be
supplied in Section 7.4.
Fig. 14 shows the comparison of cutting force results
from all the three commercial codes as well as experi-
mental results. For MSC.Marc and Deform2D, the
results corresponding to a friction factor of m ¼ 0:7
have been selected based on the best correlation of the
shear angle with experiments. In a similar way, for
Thirdwave AdvantEdge, l ¼ 0:5 has been selected.
Obviously, Deform2D supplies the best cutting forces,
whereas Thirdwave AdvantEdge deviates at most due Fig. 15. Effect of friction factor on the thrust force results obtained
to the friction model as expected. by MSC.Marc.
942 H. Bil et al. / International Journal of Machine Tools & Manufacture 44 (2004) 933–944

Fig. 16. Effect of friction factor on the thrust force results obtained
by Deform2D.

Thrust force results from Thirdwave AdvantEdge

can be seen on Fig. 17. These are the best estimated
results for the thrust forces among the three commer-
cial codes. This code uses Coulomb friction model, in Fig. 18. Comparison of contact normal pressure and shear yield
strength of the material along contact length on the rake face.
which frictional stress on the rake face is calculated
from the normal stress acting on the same surface and
not from the shear yield strength of the material 7.5. Results by Movahhedy and Altıntaş
(initially k 175 MPa). However, since the normal
stress is very high, larger friction forces than the ones Since the experimental technique applied in this
obtained using other two codes result. This is revealed study was not accurate in determining the contact
in Fig. 18 where the contact normal pressure and shear lengths, it has been decided to perform a comparison
yield strength of the material along the rake face of the with the study of Movahhedy and Altıntaş [25] in
tool during the cutting operation are plotted. It can be which a quick-stop mechanism has been used for mea-
seen that the resulting frictional stress is even higher suring the contact length accurately. Since the general
than the shear yield strength, which is physically not behavior of the three codes is already known from the
possible. Therefore, (Ff)y becomes larger than (Fn)y, comparisons made before, only MSC.Marc has been
which means a positive thrust force. In Fig. 17, it can used for these comparison.
also be seen that thrust forces are best estimated when Movahhedy and Altıntaş [25] did orthogonal metal
the friction coefficient is taken as 0.5. But even than the cutting experiments with low carbon steel for the verifi-
results are not very accurate. cation of their finite element simulations obtained by a
Comparison of all codes according to the results
commercial code in which they have measured cutting
obtained for thrust force are given in Fig. 19. Again,
force, thrust force, shear plane angle, contact length
here, the friction coefficients and factors are selected to
and maximum temperature reached.
have best agreement with the experimental shear angle
The cutting conditions and experimental results as
values. It is seen that Deform2D is not able to predict
well as the simulation results of Movahhedy and Altın-
the direction of the thrust force. Although the direction
is true for the other two codes, there are still significant taş can be found in Table 6, where c is the rake angle,
errors.

Fig. 19. Comparison of thrust force results from all codes. m¼ 0:7
Fig. 17. Effect of friction factor on the thrust force results obtained for MSC.Marc and Deform2D and l¼ 0:5 for Thirdwave Advant-
by Thirdwave AdvantEdge. Edge.
 H. Bil et al. / International Journal of Machine Tools & Manufacture 44 (2004) 933–944 943

Table 6
Comparison with the results of Movahhedy and Altıntaş [25]
v
c (deg) Vc (m/min) f (mm/rev) dc (mm) Fc (N) Ft (N) / (deg) lc (mm) Tmax ( C)
Experiment Movahhedy, 0 150 0.1 1 174 83 18.8 0.6 590
Altıntaş
Simulation Movahhedy, 0 150 0.1 1 207 96 22 0.55 571
Altıntaş
Simulation (m ¼ 0:1) 0 150 0.1 1 174 21 22.1 0.168 535
Bil, Kılıç, Tekkaya
Simulation (m ¼ 0:7) 0 150 0.1 1 225 85 17.7 0.45 547
Bil, Kılıç, Tekkaya

Vc is the cutting speed, f is the feed rate, dc is the depth Plain damage models for chip separation are not
of cut, Fc is the cutting force, Ft is the thrust force, / is appropriate for machining purposes. Although the
the shear angle, lc is the contact length and Tmax is the remeshing model gives better results, it is based on the
maximum temperature. misconception of the crack generation in the material
Table 6 also shows the results of this study predicted at the tool tip. This can be explained by Fig. 20.
by MSC.Marc, where friction factor was taken as 0.1 Firstly, penetration of material into the tool occurs
(based on having the same cutting force with experi- which triggers the remeshing step. A new mesh is
ments of Movahhedy and Altıntaş) and material is C15 defined with a new surface, which is bigger than the
steel. It can be seen that other than the cutting force, former surface for a constant volume. This is only
process parameter estimations are not in good agree- possible by introducing a crack. This crack exists inde-
ment with experimental results. Temperature prediction pendently of the element size used. In addition, the
has an error about 9% which can be accepted as a direction of this crack is not in the direction of shear-
good estimation. ing. Both observations are the source for the poor
On the other hand, with a friction factor of 0.7, agreement with experiments in all parameters. A new
which is proved to supply better results according to sound chip separation model is necessary.
the comparisons with experiments performed in this In a typical metal cutting process, very high strain
study, thrust force prediction becomes quite precise rates (about 104 1/s) are achieved. However, the typi-
while the cutting force is overestimated. Shear angle cal available material data are valid up to strain rates
was simulated with an error of about 5%. Temperature of 40 (1/s). Despite the fact that the extrapolation of
prediction becomes even better due to higher heat gen- flow curves by three orders of magnitude of strain rate
eration at the chip–tool interface because of higher fric- is performed by completely different procedures in the
tion factor. The error is about 7%. three analyzed codes, no significant effect has been
observed on the results. This may be due to the simul-
taneous effect of increased flow stress due to high strain
8. Conclusions rates and softening due to the high temperatures
reached at the cutting zone. Still, it should be noted
This study reveals the effects of friction, chip forma- that none of the well known shear angle relationships
tion model and material data in the simulation of
orthogonal cutting.
The friction parameter affects the simulation results
drastically, but tuning this parameter yields good
agreement only for some variables in the range. A
smaller friction parameter leads to good results for cut-
ting force, whereas other variables (such as thrust force
and shear angle) are computed more accurately with
larger friction parameters. Therefore, the accuracy of a
simulation must be assessed by examining all predict-
able process parameters. Furthermore, tuning of fric-
tion parameter should be based on the shear angle
variable since it defines the mechanics of the process.
Finally, the plain Coulomb friction model is not appro-
priate for machining purposes since it supplies friction
stresses, which are larger than the shear yield strength
of the material, at the tool–chip interface. Fig. 20. Crack generation during a remeshing step.
944 H. Bil et al. / International Journal of Machine Tools & Manufacture 44 (2004) 933–944

are material dependent; hence, the real effect of [11] M. Usta, Finite element analysis of orthogonal metal cutting
material behavior can only be seen in the parameters operations, Ph.D. Thesis, Mechanical Engineering Department
of Middle East Technical University, Ankara, Turkey, 1999.
such as temperatures, forces and contact length. How- [12] J. Hashemi, A.A. Tseng, P.C. Chou, Finite element modeling of
ever, to asses these, first of all a reliable chip separation continuous and segmental chip formation in high speed orthog-
model must be developed. onal machining, Journal of Materials Engineering and Perform-
ance 3 (6) (1994) 712–721.
[13] A.J. Shih, H.T.Y. Yang, Experimental and finite element predic-
Acknowledgements tions of residual stresses due to orthogonal metal cutting, Inter-
national Journal of Numerical Methods in Engineering 36 (1993)
The authors want to thank Ercenk AKTAY from 1487–1507.
FİGES Company in Turkey for making the commer- [14] A.J. Shih, Finite element analysis of orthogonal metal cutting
mechanics, International Journal of Machine Tools Manufactur-
cial code Thirdwave AdvantEdge available and Doruk
ing 36 (1996) 255–273.
MERDOL from British Columbia University in [15] A.J. Shih, Finite element simulation of orthogonal metal cutting,
Canada for the experimental results supplied. Journal of Engineering for Industry 117 (1995) 84–93.
[16] E. Ceretti, P. Fallbohmer, W.T. Wu, T. Altan, Application of
2D FEM to chip formation in orthogonal cutting, Journal of
References Materials Processing Technology 59 (1996) 169–180.
[17] H. Borouchaki, A. Cherouat, P. Laug, K. Saanouni, Adaptive
[1] M.E. Merchant, Mechanics of the metal cutting process, Journal remeshing for ductile fracture prediction, Comptes Rendus
of Applied Physics 16 (1945) 267–318. Mecanique 330 (2002) 709–716.
[2] E.H. Lee, B.W. Shaffer, The theory of plasticity applied to a [18] G.S. Sekhon, J.L. Chenot, Numerical simulation of continuous
problem of machining, ASME Journal of Applied Mechanics 73 chip formation during non-steady orthogonal cutting, Engineer-
(1951) 405–413. ing Computations 10 (1993) 31–48.
[3] R.G. Fenton, P.L.B. Oxley, Mechanics of orthogonal machining: [19] M. Baker, J. Rosler, C. Siemers, A finite element model of high
allowing for the effects of strain-rate and temperature on the speed metal cutting with adiabatic shearing, Computers and
tool–chip friction, Proceedings of Institution of Mechanical Structures 80 (2002) 495–513.
Engineers 183 (1969) 417–438. [20] F. Klocke, S. Hoppe, Simulation of the metal cutting process—
[4] G. Boothroyd, J.A. Bailey, Effects of strain-rate and temperature reliability and optimization, International Journal of Production
in orthogonal metal cutting, Journal of Mechanical Engineering Engineering and Computers 4 (2003) 43–52.
Science 8 (1966) 264. [21] H. Kudo, Some new slip-line solutions for two dimensional
[5] V. Gorianis, S. Kobayashi, Strains and strain-rate distributions steady state machining, International Journal of Mechanical Sci-
in orthogonal metal cutting, Annals of CIRP 15 (1967) 425. ences 7 (1965) 43–55.
[6] E. Usui, T. Shirakashi, Mechanics of machining from descriptive [22] P. Dewhurst, On the non-uniqueness of the machining process,
to predictive theory, on the art of cutting metals, ASME PED 7 Proceeding of Royal Society A360 (1978) 587–610.
(1982) 13–35. [23] R. Hill, The mechanics of machining: a new approach, Journal
[7] K. Iwata, K. Osakada, Y. Terasaka, Process modeling of orthog- of Mechanics and Physics of Solids 3 (1954) 47–53.
onal cutting by the rigid-plastic finite element method, Journal of [24] T.H.C. Childs, Numerical experiments on material properties
Engineering Materials and Technology 106 (1984) 132–138. and machining parameters influencing normal contact stress
[8] J.S. Strenkowski, J.T. Carroll, A finite element model of orthog- between chip and tool, University of Leeds School of Mechan-
onal metal cutting, Journal of Engineering for Industry 107 ical Engineering, CIRP Winter Meeting, February, 2003.
(1985) 347–354. [25] M.R. Movahhedy, ALE Simulation of chip formation in orthog-
[9] K. Komvopoulos, S.A. Erpenbeck, Finite element modeling of onal metal cutting process, Ph.D. Thesis, Supervised by
orthogonal metal cutting, Journal of Engineering for Industry Y. Altıntaş, Department of Mechanical Engineering, The Uni-
113 (1991) 253–267. versity of British Columbia, Vancouver, BC, Canada, 2000.
[10] Z.C. Lin, S.Y. Lin, A coupled finite element model of thermo- [26] H. Bil, Simulation of orthogonal metal cutting by finite element
elastic–plastic large deformation for orthogonal cutting, Journal analysis, MSc. Thesis, Mechanical Engineering Department of
of Engineering Material Technology 114 (1992) 218–226. Middle East Technical University, Ankara, Turkey, 2003.