You are on page 1of 9

Journal of Materials Processing Technology 103 (2000) 267±275

Simulation of the orthogonal metal cutting process using an


arbitrary Lagrangian±Eulerian ®nite-element method
M. Movahhedy, M.S. Gadala*, Y. Altintas
Department of Mechanical Engineering, University of British Columbia, 2324 Main Mall, Vancouver, Canada V6T 1N5

Received 3 March 1999

Abstract

Two different ®nite-element formulations, the Lagrangian and the Eulerian, have been used extensively in the modeling of the orthogonal
metal cutting process. Each of these formulations has some disadvantages that make it inef®cient for modeling the cutting process. In this
paper, it is shown that a more general formulation, the arbitrary Lagrangian±Eulerian (ALE) method may be used to combine the
advantages and avoid the shortcomings of both of the previous methods. It is also shown that due to the characteristics of the cutting
process, this formulation offers the most ef®cient modeling approach. Some preliminary results of this approach are presented to
demonstrate its capabilities and potential in simulating the cutting process. # 2000 Elsevier Science S.A. All rights reserved.

Keywords: Orthogonal metal cutting process; ALE; Finite-element method

1. Introduction there are still many issues to deal with in such a numerical
simulation.
Metal cutting is one of the most common manufacturing The cutting process usually involves large deformation of
operations, and a great deal of research has been devoted in the material at very high strain rates (103±107 sÿ1) with the
understanding the mechanics of this process, with the strains being much greater than unity. The chip produced is
objective of obtaining more effective cutting tools and more in contact with tool face in a highly pressured zone causing
ef®cient manufacturing process plans. Traditionally, these sticking friction, which will transform to sliding friction
objectives have been achieved by experimentation and pro- further up the tool face. Large plastic work and high friction
totyping. In recent years, however, with the surge in com- generate an enormous amount of thermal energy in that
putational power and capacity, the focus has turned to locality, causing an increase in the temperature of the metal.
numerical simulation of the process with the objective of The strain rate and temperature have opposite effects on the
replacing costly experimentation and prototyping with properties of the material. Considering that all of these
numerical simulation. Many such simulations have appeared phenomena occur in a very tiny region around the tip of
in the literature. the tool, the complexity of the process becomes evident. In a
Despite the many simulation attempts reported in the real machining process, there are tool wear and vibrations,
literature, it is perhaps fair to say that cutting simulation also, which affect the process. The stress, strain, strain rate
has not become a common tool in industry, perhaps due to and temperature variables are all dependent on the cutting
doubts about the reliability of such results and lack of parameters such as the feed rate, and the cutting speed as
suf®cient veri®cation. Much of this is related to the inherent well as on the geometrical features of the tool such as the
complexity of the process, which probably makes it one of rake angle and the nose radius. Further complicating the case
the most challenging processes from the numerical point of is the fact that it is very dif®cult, if not impossible, to
view. Although the actual machining process is simpli®ed to accurately obtain the ¯ow model of the material under these
the most basic 2D plane-strain problem, orthogonal cutting, conditions.
Apart from the complexities of the actual cutting process,
*
Corresponding author. Tel.: ‡1-604-822-2777; fax: ‡1-604-822-2403. there are aspects that make the numerical simulation of the
E-mail address: gadala@mech.ubc.ca (M.S. Gadala) cutting process particularly dif®cult. As with the analysis of

0924-0136/00/$ ± see front matter # 2000 Elsevier Science S.A. All rights reserved.
PII: S 0 9 2 4 - 0 1 3 6 ( 0 0 ) 0 0 4 8 0 - 5
268 M. Movahhedy et al. / Journal of Materials Processing Technology 103 (2000) 267±275

other metalworking operations, large deformation of the rial in front of the tool tip, the elements are greatly distorted
material results in the distortion of the elements, causing such that sometimes the mesh has to be regenerated. More
deterioration of the results. On the other hand, unconstrained importantly, the simulation of cutting action at the tip of the
¯ow of material occurs over free boundaries, which form a tool has been traditionally achieved by separation of nodes
large area of the material boundaries. The boundary condi- in front of the tool tip, similar to what is usually done for the
tions change during the course of the analysis as the contact propagation of a self-similar crack in fracture-mechanics
between the tool and the chip is being developed. However, problems. Despite its simplicity, this node-separation
more importantly, and unlike many forming processes, there method creates many problems that adversely affect the
is cutting action in which the bulk of the material is results. First and foremost is the choice of the criterion based
branched. This action is in a sense similar to what occurs on which a node in front of the tool tip is allowed to separate.
in elastic±plastic crack-propagation problems. Simulating Many criteria have been used in the literature, ranging from
such cutting action is perhaps one of the most challenging simple geometrical or strain criteria to more complex frac-
aspects in this analysis. In this paper, the authors focus on the ture-mechanics criteria [1±13]. Even if the type of criterion
numerical aspects of the simulation of the cutting process is chosen properly, there is no physical indication as to what
and propose an alternative approach to deal with them. criterion value should be adopted. All of this makes the
separation criterion more of arbitrary nature, with great
effects on the results.
2. Numerical aspects of cutting simulation A critical look at the literature shows that whilst most
researchers have used this approach, there is no consensus
The ®nite-element method has been largely used for the on the appropriate criterion. Usui and Shirakashi [1] use a
simulation of metalworking operations. From the viewpoint geometrical criterion based on the separation of a node when
of numerical formulation, these analyses may be divided it is suf®ciently close to the tool tip. The node is split into
into two major approaches: the Lagrangian approach and the two nodes, one moving up along the rake face and the other
Eulerian approach. In the Lagrangian approach, a natural remaining on the machined surface. Komvopoulos and
approach for solid-mechanics analysis, the FE mesh consists Erpenbeck [2], Zhang and Bagchi [3], and Shih [4] use a
of material elements that cover the region of analysis similar distance criterion. Whilst these studies all agree that
exactly. These elements are attached to the material and the value of the critical distance at which the node splits is
are deformed with the deformation of the workpiece. This crucial to the success of simulation, they use different
approach is particularly very convenient when uncon- values. In [2], a distance tolerance of 0.5L is used, where
strained ¯ow of material is involved, since the FE mesh L is the side length of the element in front of the tool tip,
will accurately represent the material boundaries during the whereas in [3], a value of (0.1±0.3)L is adopted and in [4], an
course of the analysis. The Eulerian approach, on the arbitrary value based on trial runs is used.
other hand, is more suitable for ¯uid-¯ow problems Carroll and Strenkowski [5] use a separation criterion
involving a control volume. In this method, the mesh con- based on the effective plastic strain at the node nearest to the
sists of elements that are ®xed in space and cover the control cutting edge. The typical criterion values in their analysis are
volume, and the material properties are calculated at between 0.4 and 0.65 depending on the cutting conditions.
®xed spatial locations as the material ¯ows through the Xie et al. [6] use such a criterion with a critical value of 0.5,
mesh. This approach has also been used to model the large whilst Hashemi et al. [7] adopt values that ranged from 0.6
deformation of solids, mostly in metal forming analyses. to 1.5.
Due to the ®xed nature of the approach, it is most suitable for Iwata et al. [8] employ a stress-based ductile fracture
cases where there are minimal free boundaries, i.e., where criterion as the chip-formation criterion. Ceretti et al. [9] use
the boundaries of the material region are known a priori. a damage criterion. Lin and Lin [10] note that the critical
Examples of these cases are closed-die forging and extrusion value for a criterion based on the effective plastic strain will
problems. be dependent on cutting parameters such as the depth of cut,
Focusing on cutting process simulation, both approaches thus such a criterion lacks generality. They propose a
have been used extensively to model this process. In addition criterion based on the strain energy density, which is de®ned
to relative ease in formulation and implementation, what as a material constant obtained from the tensile test. Chen
makes the Lagrangian approach very attractive for this and Black [11] emphasize that the separation criterion
process is that the boundaries of the workpiece are mostly should not be treated only as a numerical parameter, but
traction-free, and even the boundaries that are not free on the as a material constant that may be measured experimentally.
tool face are changing during the course of analysis. In this They point out, however, that the critical strain energy
approach, the analysis can be started from indentation to the density determined from the uniaxial tensile test cannot
incipient stage to steady state. The chip is formed easily to be representative of the cutting operation, and that combina-
its proper thickness, and no a priori assumption is needed tions of two or more separation criteria may also be used.
about the shape of the chip. This approach, however, has Huang and Black [12] study various combinations and
some disadvantages too. With the deformation of the mate- magnitudes of separation criteria and conclude that whilst
M. Movahhedy et al. / Journal of Materials Processing Technology 103 (2000) 267±275 269

Fig. 1. Lagrangian cutting analysis with a 308 rake angle tool. Effective plastic strain contours are also shown.

the type of criterion may not be crucial, its magnitude has a To demonstrate some of the problems pertinent to the
great effect on the stress and strain distributions in the chip Lagrangian approach with the node-splitting method, the
and the machined surface. They prefer a geometrical criter- authors performed a series of simulations using the updated
ion for steady-state cutting simulation, because it is easier to Lagrangian approach and shear friction on the tool face. The
determine. material data are the same, as will be reported in Section 4 of
Imperative to using a node-splitting method is that there this paper. The nodes are separated based on a critical
has to be a pre-de®ned parting line of the nodes, in front of distance criterion, the value of which is ®ne tuned in trial
the tool tip, that will be ``unzipped'' as the tool advances. In runs to ®nd the best compromise between having a smaller
addition to this requiring on a priori assumption, it some- crack ahead of the tool and avoiding the shifting of the
times creates problems, as the parting line may be pushed parting line. Fig. 1 shows a simulation with a 308 rake angle
out of position during the course of deformation and thus the tool. It can be seen that a crack is moving ahead of the tip,
target nodes will not be in front of the tool tip at the time of and that due to shifting of the point of maximum stress
separation. The magnitude of the separation criterion has a further up on the tool, the shear plane does not emanate from
profound effect on this, and may have to be relaxed to avoid the tool tip. The inadequacy of the node-splitting approach
it. On the other extreme, too relaxed a criterion creates an for smaller rake angles is shown in Fig. 2 which illustrates a
unrealistic material gap, which moves in front of the tool tip. simulation with a 108 rake angle tool. Although the critical
This gap is commonly observed in the node-splitting distance is ®ne tuned to secure the best result, it is seen that
approach, but not in experimental studies involving ductile the machined surface is unrealistically wavy due to shifting
materials. This crack also shifts the points of maximum of the parting line during the process. A larger criterion value
stress from the tip of the tool to a point further up the tool, will alleviate this problem at the expense of having a larger
affecting the predicted shear angle. Finally, the use of crack ahead of the tool. Fig. 3 shows the normal and
parting line effectively restricts the analysis only to cutting tangential forces on the tool in Lagrangian simulation.
with sharp-edge tools. The oscillation in the forces is due to successive separation
There are also other side effects to node separation; it of nodes in front of the tool tip. Every time, a node is split,
creates unbalanced forces in the analysis domain which the forces on the tool are suddenly reduced, and then
should be treated. Also, a Lagrangian approach with node gradually increased until the next node is released.
separation does not perform well for small or negative rake The dif®culties with the Lagrangian formulation with the
angles, including problems involving tools with chamfer or a node-splitting method have prompted some researchers to
nose radius. Although most of the above problems can be look for alternative methods for simulating the cutting
alleviated by using a denser mesh in the locality of the tool action. For example, Ceretti et al. [9] used a scheme that
tip, that in itself increases computational intensity, and deletes the elements that reach a critical damage value.
heightens the need for rezoning of the mesh. Therefore, Obviously, this approach violates the fundamental law of
the problem becomes highly mesh-dependent. continuity due to loss of volume, and thus the mesh around
270 M. Movahhedy et al. / Journal of Materials Processing Technology 103 (2000) 267±275

Fig. 2. Lagrangian cutting analysis with a 108 rake angle tool. Contours of effective plastic strain are also shown.

the tip should be very ®ne to minimize this volume loss. A viable alternative for node splitting within a Lagrangian
Furthermore, the method is too sensitive to the criterion analysis is presented by simulating the cutting action as the
value, as sometimes elements in places other than those continuous indentation and ¯ow of the material around the
neighboring to the tool tip may be targeted for deletion. The tool. Based on some experimental evidence, Madhavan et al.
authors also report that after deletion of damaged elements, [14] suggested that the indentation might be the actual chip-
the domain boundaries have to be smoothed or else the formation mechanism in ductile materials. In the indentation
subsequent remeshing attempt will fail. Overall, this approach, cutting is simulated by forcing the tool into the
approach seems to be too intrusive on the part of the analyst. material in small increments, causing the ¯ow of the mate-
Modi®cation to this approach may result in better simula- rial around the tool tip. As the tool advances, the FE mesh is
tion. As is done in crack-propagation problems, and instead distorted and has to be regenerated. In [14], the remeshing is
of deleting the element completely, the stiffness character- triggered when the effective plastic strain exceeds a value of
istics of the element may be adjusted to re¯ect a zero or very 0.25 at any point in the workpiece. Sekhon and Chenot [15]
small value in the separation direction. Such modi®cation, used a similar approach in which the FE mesh is checked at
however, will still suffer from the other problems of the the end of each increment and the mesh is automatically
element-deletion method. regenerated if a distorted element is detected. Marusich and
Ortiz [16] used continuous remeshing in combination with a
mesh-smoothing algorithm in order to resolve mesh distor-
tion problems. In their approach, elements are targeted for
re®nement when their plastic power contents exceed a
prescribed tolerance.
Although the indentation approach overcomes many of
the problems pertinent to the node-splitting approach, it
comes at a high computational cost. First, the mesh has to be
very ®ne around the tool tip, or else the material will overlap
the tool in between the nodal points, and subsequently, the
remeshing of the workpiece will amount to the loss of the
overlapped material. Second, due to highly localized defor-
mation and ®neness of elements, the mesh has to be regen-
erated frequently. In exemplary simulations performed using
DEFORM-2D software and its automatic remeshing mod-
ule, the entire region had to be remeshed at intervals of less
than 10 incremental steps, and still considerable volume of
material had to be cut out due to overlapping with the tool in
Fig. 3. Normal and tangential forces in Lagrangian cutting analysis. every remeshing. Finally, once the FE mesh is regenerated,
M. Movahhedy et al. / Journal of Materials Processing Technology 103 (2000) 267±275 271

material variables have to be interpolated from the old mesh introduced initially by Zienkiewicz and Godbole [20], is
to the new mesh. Since the two meshes may be considerably used in which the elasticity of the material is neglected, and
different from each other, this can introduce errors, which thus the residual stresses in the machined surface cannot be
will eventually deteriorate the results. It should also be noted predicted.
that even when the difference between the old and the
regenerated mesh is small, the interpolation process is
approximate and cannot account for the history of deforma- 3. Arbitrary Lagrangian±Eulerian formulation
tion. Also, there will be a speci®c need to use very small
incremental steps and high computational power. As an It can be seen that both the Lagrangian and the Eulerian
example, Marusich and Ortiz [16] report that a typical methods are inef®cient in modeling some of the character-
simulation (including thermal analysis) involving 1100 istics of the cutting process. The Lagrangian performs well
six-node triangular elements required one million incre- in modeling unconstrained formation of chip boundaries and
mental steps which took 20 h of CPU time on a DEC contact length, whilst the Eulerian is the method of choice as
3000 workstation. It is obvious that the computational cost far as the material ¯ow around the tip of the tool is con-
of this approach is sometimes prohibitive. cerned. Therefore, it would be ideal if these two methods
In the Eulerian approach, on the other hand, none of the could be combined in a single analysis and each of them
problems pertinent to the Lagrangian formulation exist in used where they perform the best. A more general approach,
the modeling of the cutting action. The cutting action is the ALE method provides this opportunity. In an ALE
modeled as the ¯ow of the material around the tool. In this analysis, the FE mesh is neither attached to the material
respect, it is similar to the above-mentioned indentation. nor ®xed in space. The mesh, in general, has a motion that is
However, since the mesh is ®xed spatially, no mesh distor- independent of the material. This general approach can be
tion occurs and consequently, no remeshing is required. The reduced to Eulerian or updated Lagrangian analyses as two
density of the mesh is only dictated by the expected gra- special cases. In the following sections a scheme for the
dients of stress and strain. Thus, the Eulerian approach is ALE method and its application in cutting analysis are
more ef®cient computationally and is best for modeling the reported.
region around the tool tip, at least for ductile materials. This
approach, however, is not suitable for modeling the uncon- 3.1. ALE procedure
strained ¯ow of material on free boundaries. The fact that the
mesh is spatially ®xed means that an a priori assumption Starting from the principle of virtual work, an ALE
should be made about the shape of the chip [5,17±19] and the formulation is derived in terms of two sets of velocities,
contact conditions. Because of the ®xed mesh, the contact those of the material points v and those of the grid points v.
length and the free boundaries of the chip cannot be devel- These velocities are in general independent of each other,
oped as natural results of deformation, and adjustment of the but there exists a one-to-one mapping between the material
boundaries are necessary outside of the solution domain. and the computational (grid) domains, provided that the
Commonly, an iterative procedure is employed in which the Jacobian of the mapping function is non-zero. To satisfy this,
velocities at free boundary nodes are checked to ensure that the boundaries of the two domain should coincide, requiring
their components normal to the boundary are zero and that that:
the boundary is a streamline. This involves changing the
…v ÿ v†  n ˆ 0 (1)
boundary, adjustment of the mesh and interpolation of
variables until convergence is achieved. It should be noted where n is the normal vector at any point on the boundary.
that the Eulerian formulation contains convective terms The physical interpretation of Eq. (1) is that no normal
which make material time derivatives much more dif®cult convective velocity occurs across the boundary if the surface
to handle. History-dependent material properties have to be particles remain on the surface.
interpolated in an approximate way and material points have To establish the mapping between the two domains, the
to be traced back to the ®xed mesh points. Many attempts time derivatives of a function f (x) in the computational
have been made to do this. For example, Abo-Elkhier et al. reference system is expressed in terms of its material time
[23] used a method in which a hypothetical mesh is gener- derivative in the material reference system (x):
ated using the displaced material points and an interpolation
df
using element shape functions is performed between the f 0 ˆ f_ ‡ …v ÿ v†  (2)
hypothetical mesh and the ®xed Eulerian mesh. Such a dx
process is computationally very intensive, introduces inter- where f 0 and f_ are the computational and material time
polation errors and creates convergence problems. derivatives, respectively. Eq. (2) is important in the incre-
The above-mentioned problems make the implementation mental analysis since it gives the relationship between the
of an Eulerian method more dif®cult than the implementa- material-associated quantities and the grid-point-associated
tion of the Lagrangian method. In the Eulerian cutting quantities, and thus makes it possible to track the material
analyses appearing in the literature, the ¯ow approach, deformation history when ALE is used.
272 M. Movahhedy et al. / Journal of Materials Processing Technology 103 (2000) 267±275

Using the above equation in the expression for virtual terms for the non-symmetric contributions to the stiffness
work and after linearization of the equation, the following matrix as loading terms and moving them to the right-hand
®nal incremental ALE formulation will result (the details side of the equilibrium equations.
can be found in [21]): To implement the ALE approach, beside the usual pro-
Z   cedures in incremental plasticity problems such as stress
@dui t @dt vj t @dt vk t @dt sij t
t s_ ij ÿ t
sik t ‡ s ij t ‡… 
v k ÿ t
v k † dV integration, two special procedures are needed at the end of
t v @ xj @ xk @ xk @ t xk each increment; one for the extrapolation of material-asso-
Z  
@dt vk @dt fij t ciated properties and the other for specifying the expected
ÿ dui t fiB t ‡ …t vk ÿ t vk † t dV
tv @ xk @ xk motion for each degree of freedom of the mesh in the next
Z   incremental step.
@dt vk @dt vj @dt sij t
ÿ dui t fiS t ÿ t sik t nj t ‡…t vk ÿ t vk †t nj t dS After the solution has converged in an incremental step,
ts @ xk @ xk @ xk
Z Z the positions of nodal points are updated in such a way that
ˆ dui t fiB dt V ‡ dui t fiS dt S the material and grid points coincide at the start of the next
tv ts
step. This requires that the properties of the material nodal
where the left superscript t refers to variables at time t (the points such as stresses, strains, and reaction forces are
start of the incremental step), f B and f S denote body and mapped to the new nodal positions. Instead of geometric
surface forces, respectively, sij are components of Cauchy extrapolation, as the common approach in previous Eulerian
stress tensor, nj represents boundary normals, and dui is formulations, a rearranged form of Eq. (2) is used for
virtual displacement. transformation:
Discretization of the above equation yields an unsymme- df
trical stiffness matrix with 2N unknown variables whilst Df ˆ Df ‡ …u ÿ u†  (5)
dx
having only N equilibrium equations. Supplementary equa-
tions are needed, therefore, to solve this set of equations. where u and u stand for material and grid displacements,
Supplementary equations are provided by assigning the respectively. This equation ensures consistency of the ALE
desired grid velocities in each incremental step at the start procedure. For the mesh-motion scheme, the objective has
of that step. Any desired criterion can be used in deciding the been set to retain the regularity of the mesh throughout the
mesh motion, re¯ecting the arbitrary nature of the ALE. For analysis. The trans®nite method for mesh generation has
example, the mesh motion can be designed such that mesh been modi®ed to account for the boundary motions of a
distortion is minimized and elements retain their regular patch of elements, and is applied at the end of each incre-
shapes throughout the analysis. Without loss of generality, ment to ®nd the positions of nodal points in the next step.
the mesh motion can be expressed in terms of material This method reproduces the boundary nodes exactly, and
velocity in the following form: only the internal nodes of a patch are adjusted. Alternatively,
an approach based on isoparametric mapping may be used
vi ˆ a…i† ‡ b…i† vi (4) that can handle patches with varying element density [22].
(no sum on i), where a(i) and b(i) are arbitrary coef®cients
decided by the mesh-motion scheme. This form of expres-
sion for vi is very helpful in two respects. First, it helps to 4. Simulation example
treat the supplementary constraint equations on the element
level, thus reducing the size of the global stiffness matrix to In this section, how the ¯exibility of the ALE approach
N instead of 2N, and second, it simpli®es degeneration of the can be used to create more ef®cient simulation for metal
ALE equations to the Eulerian and updated Lagrangian cutting processes is explained. The idea is to use features of
approaches as special cases: the Eulerian approach in an area close to the tool tip, whilst
simultaneously taking advantage of features of the Lagran-
if a…i† ˆ b…i† ˆ 0; Eulerian formulation gian approach in modeling unconstrained ¯ow of the mate-
if a…i† ˆ 0 and b…i† ˆ 1; updated Lagrangian formulation rial on the free boundaries, which de®nes the shape and size
of the chip. To achieve this, a start is made by assuming an
Note that in a general ALE motion, both of the coef®cients initial chip size, but it should be noted that this assumption is
can be non-zero. not an inherent requirement of the ALE; by designing a more
As indicated above, the stiffness matrix resulting from Eq. sophisticated mesh-motion scheme, all stages of the process
(3) is non-symmetrical in general. It may be shown that can be modeled. An initial chip geometry is assumed here as
when speci®c constitutive relations are introduced, the ALE a matter of convenience, because it considerably simpli®es
formulation may be considered a logical extension to an the mesh-motion scheme, and because the steady-state of the
updated Lagrangian formulation [24]. In other words, pro- cutting process is of interest. During the analysis, part of the
blems formulated with ALE may be solved using a regular mesh is Eulerian (®xed in space), whilst the motion on free
symmetric Gauss elimination solver as the case in total or boundaries is of Lagrangian nature, i.e. it follows the
updated Lagrangian methods, by simply handling some material deformation. The rest of the domain possesses a
M. Movahhedy et al. / Journal of Materials Processing Technology 103 (2000) 267±275 273

Fig. 4. Mesh-motion scheme in ALE cutting analysis.

general ALE motion, and moves within an incremental step positions in the central region of the model and in the
at a speed determined by the mesh-motion scheme. To avoid neighborhood of the tool remain ®xed throughout the ana-
entanglement of nodes on Lagrangian boundaries, the posi- lysis (the Eulerian region). Displacement boundary condi-
tions of these nodes are also adjusted tangential to the tions are applied on the left and bottom sides of the model,
boundary, as this does not violate Eq. (1). In other words, which advance the workpiece towards the stationary tool.
the nodes on the boundaries are Lagrangian in the normal The tool is assumed to be rigid, with a rake angle of 108, a
direction and ALE in the tangential direction. The capabil- nose radius of 0.025 mm, and a cutting a depth of 0.25 mm.
ities of ALE formulation are demonstrated in the example The model consists of about 1000 linear triangular elements,
that follows. It should be noted that the present purpose is and a total movement of 1.5 mm is simulated in 1000
only to demonstrate these capabilities and a thorough cutting incremental steps.
analysis is not intended. The mesh-motion design in this simulation is shown in
An elasto-plastic analysis with linear strain-hardening is Fig. 4. It can be seen that as the material advances towards
performed on a material with a Young's modulus of 200 GPa the tool, the area between the left boundary and the ®xed
and an initial yield stress of 414 MPa. The strain-hardening area becomes smaller, whilst the area between the ®xed area
modulus is chosen to be 1 GPa. The strain rate and tem- and the right boundary becomes larger. The motion of the
perature effects on material properties are neglected in this nodal points in these regions is designed in such a way that
analysis. Fig. 4 shows the initial con®guration of the ®nite- the elements in these areas retain their regular shape. Also,
element model as well as the con®guration when the cutting the chip length increases steadily during the analysis, this
process has reached steady state. It can be seen that the nodal being achieved by expanding the elements at the chip tail.
274 M. Movahhedy et al. / Journal of Materials Processing Technology 103 (2000) 267±275

Fig. 5. Contours of effective stress in the primary and secondary deformation zone.

Again, the mesh-motion scheme retains the mesh regularity


and the elements are distributed evenly in the shear defor-
mation zone as well as in the chip tail. It should be noted that
the nodes on free boundaries are free in the direction normal
to the boundary at any point. Fig. 5 shows the formation of
the chip after 0.05, 0.5, 1.0, and 1.5 mm of tool advance into
the material. It can be seen that the chip is formed steadily. It
is also noted that the thickness and the shape of the chip are
developed in a natural unconstrained fashion, not affected by
the assumed initial shape of the chip; the chip adjusts its
thickness automatically, and without any need for the itera-
tive adjustment performed in a similar Eulerian analysis.
Fig. 5 also shows the contours of effective stress in the
cutting zone. The chip-formation process, the primary and
secondary deformation zones, the stress patterns around the
tool tip and the chip curling, are all apparent in this ®gure.
The area at the tail of the chip corresponds to the assumed
initial chip. Finally, the ¯ow of the material within the
workpiece and the chip is shown in Fig. 6, where the velocity
vectors are plotted. The velocity vectors around the tool tip Fig. 6. Velocity vectors in the ALE cutting analysis.
M. Movahhedy et al. / Journal of Materials Processing Technology 103 (2000) 267±275 275

clearly show the plastic ¯ow of the material around the nose. [5] J.T. Carroll, J.S. Strenkowski, Finite element models of orthogonal
Overall, the stress- and strain-distributions and the chip cutting with application to single point diamond turning, Int. J. Mech.
Sci. 30 (1988) 899±920.
geometry agree qualitatively with experimental and numer- [6] J.Q. Xie, A.E. Bayoumi, H.M. Zbib, A study on shear banding in chip
ical analyses reported in the literature. formation of orthogonal machining, Int. J. Mach. Tools Manuf. 36
(1996) 835±847.
[7] J. Hashemi, A.A. Tseng, P.C. Chou, Finite element modeling of
5. Conclusions segmental chip formation in high speed orthogonal cutting, J. Mater.
Eng. Perform. 3 (1994) 712±721.
An ALE ®nite-element formulation is described and then [8] K. Iwata, A. Osakada, Y. Terasaka, Process modeling of orthogonal
employed to model the metal cutting process. It is shown cutting by rigid-plastic ®nite element method, J. Eng. Mater. Technol.
106 (1984) 132±138.
that both the Lagrangian and Eulerian analyses suffer from [9] J.E. Ceretti, P. Fallbehmer, W.T. Wu, T.J. Altan, Application of 2D
shortcomings in modeling this process, and that the strength FEM on chip formation in orthogonal cutting, J. Mater. Process.
of both methods can be combined to achieve a more ef®cient Technol. 59 (1996) 169±181.
cutting simulation. The features of an ALE analysis of [10] Z.C. Lin, S.Y. Lin, A coupled ®nite element model of thermo-elastic±
cutting process are as follows: plastic large deformation for orthogonal cutting, ASME J. Eng.
Mater. Technol. 114 (1992) 218±226.
1. No node-separation criterion is necessary, and no [11] A.G. Chen, J.T. Black, FEM modeling in metal cutting, Manuf. Rev.
change in mesh topology is needed. 7 (1994) 120±133.
[12] J.M. Huang, J.T. Black, An evaluation of chip separation criteria for
2. The density of the mesh around the tool tip need not be the FEM simulation of machining, J. Manuf. Sci. Eng. 118 (1996)
very high, and thus the process is computationally 545±554.
ef®cient. [13] T. Obikawa, E. Usui, Computational machining of titanium alloy,
3. The chip formation occurs by continuous plastic ¯ow of ®nite element modeling and a few results, J. Manuf. Sci. Eng. 118
the material around the tool. (1996) 208±215.
[14] V. Madhavan, S. Chandrasekar, T.N. Farris, Mechanistic model of
4. The shape and the thickness of the chip are developed machining as an indentation process, in: D.A. Stephenson, R.
automatically in the process and there is no need for Stevenson (Eds.), Materials Issues in Machining III and the Physics
iterative adjustment of the boundaries. of Machining Process, The Mineral, Metals and Materials Society,
5. Frequent remeshing and data interpolation are avoided, Vol. 2, 1993, pp. 187±209.
as the mesh motion becomes part of the solution [15] G.S. Sekhon, J.L. Chenot, Numerical simulation of continuous chip
formation during non-steady orthogonal cutting, Eng. Comput. 10
procedure. The computational intensity of the solution is (1993) 31±48.
not affected, however, as the supplementary constraint [16] T.D. Marusich, M. Ortiz, Modeling and simulation of high-speed
equations are treated on the element level [21]. machining, Int. J. Numer. Methods Eng. 38 (1995) 3675±3694.
[17] J.S. Strenkowski, K.J. Moon, Finite element prediction of chip
An example has been presented to demonstrate the cap- geometry and tool±workpiece temperature distribution in orthogonal
abilities of the ALE approach. Further work is still under metal cutting, ASME J. Eng. Ind. 112 (1990) 313±318.
way to improve the performance of the approach in the [18] K.W. Kim, H.-C. Sin, Development of a thermo-viscoplastic cutting
simulation of metal cutting processes; the results of full scale model using ®nite element method, Int. J. Mach. Tools Manuf. 36
(1996) 379±397.
and thorough analyses of the cutting process are to be [19] J.-S. Wu, J.R. Dillon, W.-Y. Lu, Thermo-viscoplastic modeling of
presented in forthcoming publications. machining process using a mixed ®nite element method, ASME J.
Manuf. Sci. Eng. 118 (1996) 470±482.
[20] O.C. Zienkiewicz, P.N. Godbole, Flow of plastic and visco-plastic
References solids with special reference to extrusion and forming processes, Int.
J. Numer. Methods Eng. 8 (1974) 3±16.
[1] E. Usui, T. Shirakashi, Mechanics of machining-from descriptive to [21] J. Wang, M.S. Gadala, Formulation and survey of ALE method in
predictive theory, on the art of cutting metals Ð 75 years later, nonlinear solid mechanics, Finite Elements Anal. Des. 24 (1996) 253.
ASME-PED 7 (1982) 13±15. [22] M.S. Gadala, M. Movahhedy, On the mesh motion for ALE modeling
[2] K. Komvopoulos, S.A. Erpenbeck, Finite element modeling of of metal forming processes, submitted to Finite Elements in Analysis
orthogonal cutting, ASME J. Eng. Ind. 113 (1991) 253±267. and Design.
[3] B. Zhang, A. Bagchi, Finite element simulation of chip formation and [23] M. Abo-Elkhier, G.A.E. Oravas, M.A. Dokainish, A consistent
comparison with machining experiment, ASME J. Eng. Ind. 116 Eulerian formulation for large deformation analysis with reference to
(1994) 289±297. metal-extrusion process, Int. J. Non-linear Mech. 52 (1988) 37±52.
[4] A.J. Shih, Finite element simulation of orthogonal metal cutting, [24] M.S. Gadala, J. Wang, ALE formulation and its application in solid
ASME J. Eng. Ind. 117 (1995) 84±93. mechanics, Comput. Methods Appl. Mech. Eng. 167 (1998) 33±55.

You might also like