CIRP Annals - Manufacturing Technology 63 (2014) 585605

Contents lists available at ScienceDirect

CIRP Annals - Manufacturing Technology

jou rnal homep age : ht t p: // ees .e lse vi er. com/ci rp/ def a ult . asp

Virtual process systems for part machining operations

Y. Altintas (1)a,*, P. Kersting b, D. Biermann (2)b, E. Budak (1)c, B. Denkena (1)d,
I. Lazoglu (2)e
a

Manufacturing Automation Laboratory, University of British Columbia, Canada

Institute of Machining Technology (ISF), Technische Universitat, Dortmund, Germany
c
Manufacturing Research Laboratory, Sabanci University, Istanbul, Turkey
d
Institute of Production Engineering and Machine Tools (IFW), Leibniz Universitat, Hannover, Germany
e
Manufacturing and Automation Research Center, Koc University, Istanbul, Turkey
b

A R T I C L E I N F O

A B S T R A C T

Keywords:
Virtual
Machining
CAM

This paper presents an overview of recent developments in simulating machining and grinding processes
along the NC tool path in virtual environments. The evaluations of cutterpart-geometry intersection
algorithms are reviewed, and are used to predict cutting forces, torque, power, and the possibility of
having chatter and other machining process states along the tool path. The trajectory generation of CNC
systems is included in predicting the effective feeds. The NC program is automatically optimized by
respecting the physical limits of the machine tool and cutting operation. Samples of industrial turning,
milling and grinding applications are presented. The paper concludes with the present and future
challenges to achieving a more accurate and efcient virtual machining process simulation and
optimization system.
2014 CIRP.

1. Introduction
The current trend is to develop digital models of the manufacturing chain from conceptual design to engineering analysis and
manufacturing processes. Conceptual design has been practiced
since the 1960s with the introduction of Computer Aided Design
(CAD) and Computer Aided Manufacturing (CAM) methods.
Engineering analysis has also accompanied design via Computer
Aided Engineering (CAE) tools such as Finite Element (FE) Analysis.
The concept of digital machines has also been widely implemented in
industry by utilizing computer graphics and animation technologies.
Machine tools are designed using solid models with integrated FE
analysis systems that predict the mode shapes and their dynamic
stiffness at the cutting toolworkpiece interface, which leads to a
prediction of the machines maximum material removal limits
during design [5]. The geometric removal of material on a machine
tool is graphically simulated to check the collision and kinematic
correctness of the tool path. Virtual geometric simulations of the
material removal and machine tool motions are now commonly used
in industry. The dynamics of the servo drives, trajectory generation,
tool change and part handling mechanisms are simulated in virtual
environments [15]. The interaction between the manufacturing
processes and machine tools has also been analyzed using digital
models as presented in [1,5,31]. However, the virtual machining of
parts by considering the physics of the manufacturing processes has
recently been evolving, and the progress being made in this eld is
subject of this keynote paper.

* Corresponding author.
http://dx.doi.org/10.1016/j.cirp.2014.05.007
0007-8506/ 2014 CIRP.

The virtual machining concept is illustrated in Fig. 1. The CAD

model of the part is used to generate NC programs in a CAM
environment where the process planners design tool path
strategies and select cutting conditions based on their experience.
The NC program is tried on a physical machine, and if the process
is found to be faulty, the trial and error cycle between the CAM and
physical machining steps is repeated until a satisfactory result is
obtained. The aim of the virtual machining is to reduce or even
eliminate physical trials by simulating the physical operations in
digital environments ahead of costly production as introduced by
Altintas in 1991 [13]. There has been progress toward virtual
machining by simulating the cutting forces and optimizing the feed
along the tool path in three-axis peripheral [76,131] and three[74,91] to ve-axis ball-end milling of dies and molds [104].
The virtual machining system requires sound mathematical
models of metal cutting and grinding processes, the dynamics of
machine kinematics and CNC servo drives, and cutterpart
geometry engagement conditions along the tool path. The
mechanics [20,82] and dynamics of cutting [16], drilling [123]
and grinding [34,121] processes have been investigated for almost
a century, and the progress in their mathematical modeling has
been reported in the cited keynote papers and will not be repeated
here. This paper presents the integration of cutting and grinding
process models into CAM systems for the simulation of part
machining operations in virtual environments.
Henceforth, the paper is organized as follows. The identication
methods for toolworkpiece engagement conditions along the tool
path are summarized in Section 2. The computationally efcient
mathematical modeling of metal-cutting and grinding process
mechanics that are relevant to virtual machining are presented in

[(Fig._1)TD$IG]

586

Y. Altintas et al. / CIRP Annals - Manufacturing Technology 63 (2014) 585605

Fig. 1. Architecture of a virtual machining system (UBC MAL).

Section 3. The kinematics and dynamics of machines that govern the

relative motion between the tool and workpiece are given in Section
4. The optimization criteria for NC programs are given in Section 5
with the presentation of industrial applications in Section 6. The
paper concludes by highlighting the current research challenges that
need to be resolved before fully utilizing manufacturing process
simulation and optimization tools in CAM environments.
2. Toolworkpiece-engagement identication algorithms
Machining process simulation and optimization requires the
geometric modeling of the engagement of the cutter with the
workpiece at discrete intervals along the tool path [116,117]. The
cutterworkpiece engagement (CWE) will lead to the variation in
chip thickness, and axial and radial depth of cut which are needed to
evaluate force [11,27,115], torque, power, vibration [98,99] and other
process states along the tool path [68,134]. Various geometric
modeling techniques are known in the literature for the description of
the engagement between a tool and a workpiece, which are reviewed
as follows.

However, the computation of the intersection between the

represented surfaces (Fig. 2d) is a difcult and computationally
time-consuming task.
In contrast to the B-rep technique, the CSG (Constructive Solid
Geometry) representation allows an easy description of the
composition of individual components [72]. The idea of the CSG
technique is to combine solid objects, e.g., spheres, cones, cuboids,
using Boolean operations, like union, difference or intersection as
[(Fig._3)TD$IG]shown in Fig. 3 [1,128].

Fig. 3. Example of a CSG-based composition: combining two primitives (here: cube

and sphere) using set operations: union, difference, and intersection respectively
(ISF).

2.1. Solid-model-based systems

2.2. Wire-frame-based systems

Solid modeling techniques, such as Constructive Solid Geometry (CSG) or Boundary Representation (B-Rep), are used to model
three-dimensional objects [108]. These techniques were designed
in the mid-1960s when CAD/CAM-systems required models
containing the geometric dimensions of the parts.
The method of describing solid objects by their boundaries, i.e.,
surface patches, edges and vertices, is called B-rep [26]. It supports
various mathematical descriptions [77] such as Bezier, Spline, or
NURBS (NonUniform Rational B-Splines) techniques [72]. B-rep offers
design exibility and high reproducibility of free-form surfaces [26],
and allows a continuous and accurate representation of the sweep
[(Fig._2)TD$IG]volume [130] of a moving cutter envelope as shown in Fig. 2 [57,133].

The shape of a 3D object can also be represented by points and

lines. These models do not provide any information about the
inside and outside of the component, but allow a fast and simple
visualization of the components as shown in Fig. 4, although not as
[(Fig._4)TD$IG]smooth as in solid model representation of parts.

Fig. 4. Example of a wire-frame-based system: The tool and the workpiece are
depicted by lines (ISF).

2.3. Voxel-, dexel-, and Z-buffer-based systems

Fig. 2. Example of a boundary representation of a tool-workpiece engagement. (a)

Initial (P1) and nal (P2) conguration of the cutter. (b) Sweep volume of the moving
cutter envelope. (c) Raw stock material and generated sweep volume. (d) Result of the
Boolean operation of the sweep volume and the raw stock material [133].

Modeling techniques based on z-buffer, dexels or voxels are

discrete representations of objects [72]. Using a voxel-based
system, the volume is approximated using small, uniform cuboids
(Fig. 5) which are called voxels (volume element/volumetric
pixel). Voxels are either lled with a material or kept empty. Since
the number of voxels (n) depends on the resolution by O(n3) [128],
the drawback of this easy-to-implement modeling technique is a
high demand of memory and computation time when the
resolution of the model is increased.

[(Fig._7)TD$IG]

Y. Altintas et al. / CIRP Annals - Manufacturing Technology 63 (2014) 585605

[(Fig._5)TD$IG]

587

Fig. 7. Concept of the dexel-based technique (ISF).

2.4. Point-based methods

Fig. 5. Example of a milling simulation using a voxel-based workpiece model [119].

The basic idea of dexel (depth element)-based systems is to

discretize an object not by using cuboids, but parallel line
segments, which are arranged on a regular grid. These line
segments can have different lengths dened by their start and
end points. If the elements are only used in one direction and
share the same start value [73,134], the method is called zbuffer technique (Fig. 6), which is named after the rendering
technique in graphic processor units [129]. This approach is
very easy to implement and has a low computation time.
However, only convex shapes (without undercuts) can be
modeled, and each line has to be replaced by a list of line
segments
[128].
[(Fig._6)TD$IG]

Point-based methods discretize an object using single points

(Fig. 8 [85]). For example, the basis for element-free techniques,
like SPH (Smooth Particle Hydrodynamics [93]), or methods using
background meshes, e.g., MPM (Material Point Method [19]) are
used. These techniques are adapted from computer science in
order to overcome the disadvantages of nite element models, e.g.,
a strong distortion of nite elements resulting in a low computation accuracy or high computational costs for re-meshing [19].
2.5. Analytical methods
Besides discretization and solid modeling techniques, analytical
approaches have also been used in calculating toolworkpiece
engagement. Such approaches mainly depend on the approximation of the workpiece surface information by use of the cutter
location (CL) le itself as shown in [126]. The cutting parameters
such as step over, cutting depth, lead and tilt angles are calculated
using the analytical difference between consecutive CL points, tool
axis vector and the approximated workpiece surface (Fig. 9).
Although such methods are fast, their application is limited to
special machining cases such as ball end milling of blades due to
the geometrical denition requirements for the features on the
part.
2.6. Discussions

Fig. 6. Example of modeling the surface of a honing tool [83] using z-buffer method
(ISF).

In order to improve the accuracy of the z-buffer model, three

dexel models each aligned along one of the Cartesian axes can
be used (Fig. 7). In contrast to the voxel-based technique, the dexel
model has less memory demand of O(n2).

The choice of the models for the workpiece and the tool (Fig. 8)
depends on the focus of the simulation, the required accuracy of
the results and the demand on the computation time. Additionally,
if particular simulation systems or commercial software are used,
the kind of model is generally predetermined. The approach for
calculating the material removal process and for the tool
workpiece-engagement identication depends on the chosen
models. Generally, the machining processes are simulated using
a time [133] or displacement discretization of the tool path
[97,115] (Fig. 10).

[(Fig._8)TD$IG]

Fig. 8. Different possibilities to model the workpiece and the tool (ISF).

[(Fig._9)TD$IG]

588

Y. Altintas et al. / CIRP Annals - Manufacturing Technology 63 (2014) 585605

equation:
0

B
C
B W0 C
C \ Tn
Cn B
Bn1
C
@[ A
Ti

(1)

i1

where W0 is the model of the stock material and Ti is the envelope

of the tool model at the position of the nth cut (Fig. 12 [118]). Since
Boolean operations can be dened easily and accurately, the CSG
technique is a popular model for high-precision engagement
[(Fig._12)TD$IG]calculations [85].

Fig. 12. Geometric model of the current chip shape Cn based on the CSG technique
(ISF).
Fig. 9. Analytical calculation of cutting parameters [105].

[(Fig._10)TD$IG]

The visualization of the material removal on the part is also an

essential feature in virtual machining. Although a CSG representation of the workpiece can be updated in constant time, it is
computationally costly to directly render this model using raytracing techniques [115]. Therefore, an alternative visualization
model is often used in combination with the CSG model [118].
3. Metal cutting process models

Fig. 10. Discretization of the tool movement [133]. (a) Tool at different NC positions.
(b) Approximated sweep surface.

Since discrete models (voxel-, dexel-, z-buffer-based systems)

are easy to implement, they are most commonly used for the
representation of tools or workpieces [24,25,28,56]. The simplest
and fastest approach is the z-buffer technique, which is, for
example, used in (Fig. 6) to model the honing process [83].
However, z-buffer models are insufcient for use in the ve-axis
machining of free-form surfaces, where dexel- or voxel-based
systems are preferred (Fig. 11). Due to the discrete sampling, these
modeling techniques are subject to aliasing errors, which can be
reduced by increasing the number of elements (dexel or voxel) but
at the expense of higher computation loads and larger memory
demands, i.e. O(n3) for voxel and O(n2) for dexel boards. If a more
precise model is needed, solid modeling techniques such as CSG or
B-rep
methods, should be used [92].
[(Fig._1)TD$IG]

Fig. 11. Example of dexel-based workpiece and a CSG-based tool model. Simulation
of the (a) NC-milling process, (b) NC-grinding, (c) drill grinding (ISF).

Using the CSG technique, the material removal process and the
current chip shape Cn can be easily described by the following

The main objective in simulating part machining operations in a

virtual environment is to predict the maximum cutting forces,
torque, power, vibration amplitudes and dimensional surface
errors left on the part in short computational time windows. As a
result, the micro-mechanics of cutting which aims to predict the
temperature, stress, residual stresses and strain distribution in
toolchip interface zones are not considered in the virtual
machining of parts. However, the micro-analysis at the cutting
zone is still possible by freezing the tool at a particular location and
simulating the process in detail if needed. The details of micromechanics models, which are not used by NC programmers but by
process design specialists, can be found in the past CIRP keynote
papers [20,82].
3.1. Mechanics of orthogonal cutting
The macro mechanics of orthogonal cutting can be simplied by
assuming a thin shear plane with an average shear angle (fc) and
shear yield stress (ts) as shown in Fig. 13 [4]. The sticking and
sliding friction zones between the chip and rake face are simplied
by having an average Coulomb friction coefcient (ma). The
fundamental orthogonal cutting parameters (fc, ts, ma) can be
obtained from orthogonal cutting tests and stored in a material
data base as an experimentally calibrated function of uncut chip
thickness (h), cutting velocity (V) and tool rake angle (gr) [4].
Alternatively, the orthogonal cutting parameters can be predicted
from micro-metal cutting mechanics methods such as Finite
Element or slip line eld analysis [136,137]. The chip thickness (h)
has a static part (hcs) which is dependent on the tool geometry and
kinematics of the cutting operation, and a dynamic part (hcd) due to
regenerative vibrations [9,16,37].
Since a number of metal cutting operations such as turning,
drilling, and milling occur in machining a part on CNC machining
centers, the mechanics of cutting for a variety of tool geometries
must be handled by the virtual machining system. The cutting edge

[(Fig._13)TD$IG]

[(Fig._14)TD$IG]

Y. Altintas et al. / CIRP Annals - Manufacturing Technology 63 (2014) 585605

589

Fig. 13. Mechanics of orthogonal cutting with thin shear plane [4].

geometry is rst dened according to ISO 3002 [81] standards. The

normal rake, cutting edge and oblique angles, which are needed in
modeling the mechanics of cutting, are evaluated from the tool
geometry and cutting velocity direction [84]. The tool reference
plane (Pr) is oriented perpendicular to the primary motion (cutting
velocity) vector and parallel to the tool axis (Fig. 14a). The tool
cutting edge plane (Ps) is tangential to the cutting edge at the
selected point and perpendicular to Pr. The cutting velocity vector
(v0) is transformed to design coordinates (Frame D) after the insert
is oriented on the cutter body (Fig. 14b) [84]. The tool with an
oblique angle of cut (ls) and the normal rake angle (gn) is dened in
the oblique cutting mechanics frame (Fig. 14) as a function of the
orientation of the cutting edge with respect to the cutting velocity
direction and the cutter body. The details of geometric transformations can be found in [84].
The generalized mechanics model requires the evaluation of the
friction force on the rake face (Fu), which is aligned with the chip
ow angle (h) and the normal force (Fv) described in chip ow
coordinates, as shown in Fig. 15. A differential cutting edge that
produces a chip with an area (dAc) and length (dS) creates the
following friction and normal forces:
dF u kdz K uc kdzdAc kdz K ue dSkdz
dF v kdz K vc kdzdAc kdz K ve dSkdz

Fig. 14. Tool-in-hand planes and motion directions (adapted from ISO 3002) and the
denition of normal rake angle [81].

[(Fig._15)TD$IG]

(2)

where Kuc and Kvc are the friction and normal cutting force
coefcients, and Kue and Kve are the edge force coefcients in
oblique cutting for each insert and differential segment (k) with dz
height. The chip area and width are evaluated as
dAcj kdz h j kdz  dS j kdz

and

dS j kdz

dz
sin kr j

where hj (kdz) is the local chip thickness cut by tooth j. Cutting

coefcients in friction and normal directions can be evaluated
using orthogonal to oblique transformation methods such as
proposed by Armarego et al. [21] as follows:
q

t s 1  tan2 h sin2 bn

q sin ba ;
cos2 fn bn  g n tan2 h sin2 bn
q
t s 1  tan2 h sin2 bn
q cos ba :

cos ls sin fn cos2 fn bn  g n tan2 h sin2 bn

K uc

cos ls sin fn

K vc

(3)

Fig. 15. Mechanics of oblique cutting [84].

[(Fig._16)TD$IG]

Y. Altintas et al. / CIRP Annals - Manufacturing Technology 63 (2014) 585605

590

The shear stress (ts), shear angle (fn), average friction angle
(ba = tan1 ma) and edge coefcients (Kue and Kve) are evaluated
from the orthogonal cutting model as described in [21]. The
projection of the friction angle on the cutting edge normal plane is
shown as (Pn), is bn = tan1(ms cos h), where h is the chip ow angle
which can be assumed to be equal to oblique angle h  ls using the
Stabler rule [48] for simplistic force analysis. Alternatively, the
cutting force coefcients (Kuc, Kvc) can be mechanistically
calibrated from dedicated cutting tests conducted with each tool
geometry. Both methods are widely used in creating a cutting force
coefcient data base for various materials. The edge force
coefcients (Kue, Kve) are highly dependent on the radius of the
cutting edge and ank wear. They need to be identied either
experimentally, or by using Finite Element [20,137] or slip line
eld [136] models. The thermo-mechanical behavior of the
material can be modeled using the JohnsonCook material model
where strain, strain rate and temperature effects can be considered
in Finite Element or slip line eld models when predicting the
cutting force coefcients [38,101]. A sample mechanistic cutting
force coefcient identied from the ow stress and friction
parameters of a material as a nonlinear function of chip thickness
(h) and cutting edge radius (r) from Finite Element and slip line
eld simulations is given as [17]:
d

K t h; r K t1 h K t2 h; r at h t bt h t r qt :

[(Fig._17)TD$IG]

Fig. 16. Turning operation [9].

[(Fig._18)TD$IG]

Fig. 17. Multiple teeth boring [84].

(4)

Once the differential friction (dFu) and normal (dFv) forces on

the rake face are identied from the material properties and
cutting edge geometry (Eqs. (2) and (3)), they can be integrated
along the width of cut (b) to nd the total cutting (Fu, Fv) forces on
the rake face of the cutting edge.
3.2. Transformations to machining coordinates
In a virtual machine tool, the chip thickness must be calculated
as a function of feed, tool geometry, the kinematics of the
machining operation, and relative vibrations between the cutting
tool and workpiece. There are several coordinate systems in the
process chain, but the fundamental base is the rake face of the tool
where the chip leaves the part.
The rake face coordinates (u, v) are transformed into cutting
edge coordinates (x, y, z-RTA coordinate system I) from Fig. 15 as
[84]:
fx

z gTI T1U f u

v gT

(5)

T 1U

3
cosg n cosh
sing n
4 sinls sinh cosls sing n cosh cosls cosg n 5:
cosls sinh sinls sing n cosh sinls cosg n

The RTA coordinates dene the cutting edge, which can be

transformed to a machine coordinate system (0) as a function of
each machining operation:


T0R  TRI
(6)
f x y z c gT0
f x y z gTI
0 Rt 0
where Rt is the moment arm for the force in y direction creating the
torque in direction c. The transformation matrices TOR and TR1 are
unique for each machining operation. TOR matrix is represented as:
2
3
sin c j cos c j 0
4
(7)
TOR cos c j
sin c j
0 5;
0
0
1
where the location angle (cj) and TR1 depend on the cutting tool
orientation for each operation as shown in Figs. 1618.
The chip thickness is measured perpendicular to the cutting
edge and dened by the following general expression:
h j t; kdz c j sin k


r kdz

 sin f j t; kdz

ecd
j t; kdz

 Dqt

(8)

Fig. 18. Drilling operation [9].

where cj is the feed rate for tooth j. The engagement angle time
varies with the time in milling (fj = cj) and the constant for
turning (fj = p/2) operations.

Y. Altintas et al. / CIRP Annals - Manufacturing Technology 63 (2014) 585605

3.3.2. Boring heads and drilling operations

Drilling and boring heads have multiple cutting edges but with
identical kinematics as shown in [22,23]. The tool rotates and
moves axially along the hole axis, and the corresponding
transformation parameters are set as shown in Figs. 17 and 18
[9,84].

The vibrations in lateral (x, y) and axial (z) directions are dened
in machine coordinates as [9]:
qt f xt

zt gT

yt

(9)

The effect of regenerative vibrations Dq qt  qt  T is

considered by orienting them into the direction of the chip
thickness with the transformation vector:
e j t; kdz f 1

0 gT0R  TR1 :

(10)

The cutting forces on the rake face (Eq. (2)) are transformed to
RTA coordinates using Eq. (5) which have cutting, ploughing and
process damping parts as follows [59]:
ed
dF j t; kdz dFcj t; kdz dFes
j t; kdz dF j t; kdz ;
|{z} |{z} |{z}
cutting

Ft; kdz

Fx

Fy

Fz

Tc

T

ploughing
N X
K
X

process damping

(11)

591

3.3.3. Milling operations

Milling cutters have multiple teeth but the direction of feed is
perpendicular to the plane formed by the spindle (Z8) and normal
to the feed (X8) directions, as shown in Fig. 19. If the cutter has an
arbitrary geometry with different inserts along each ute as in Fig.
20, the cutter is subdivided into small differential disk elements
[7,10]. The contributions of all differential cutter elements to forces
are calculated and summed digitally to nd the total forces, torque
and power contributed by the total cutter engaged with the
workpiece
[60,61].
[(Fig._19)TD$IG]

gt; kdzdF j t; kdz;

j1 k1

where g (t, kdz) = 1 when the cutting edge is in cut and g (t, kdz) = 0
otherwise. The cutting component dF cj of the forces is divided into
cd
stationary dF cs
j and dynamic dF j parts as:
t; kdz dFcd
t; kdz;
dFcj t; kdz dFcs
j 
 j
T0R  TRI
cs
dF j t; kdz
T 1U
0 Rt kdz 0


K uc kdz
c j sinf j t; kdz dz;
K vc kdz


T0R  TRI
t;
kdz

T
dFcd
j
0 Rt kdz 0 1U


K uc kdz
ecd
j t; kdzDqt dz:
K vc kdz

(12)

The ploughing forces are given as:



T0R  TRI
dFes
t;
kdz

T1U 
j

0 Rt kdz 0
1
K ue kdz
dz
K ve kdz sinkr kdz

(13)

[(Fig._20)TD$IG]

Fig. 19. Milling operation [9].

Process damping forces are found as [9]:

dFed
j t; kdz

T0R  TRI
0 Rt kdz


0

e pd
K s p Lw 2
j t; kdz
dz
P
sinkr kdz
4vc

(14)

where P f 1 m 0 gT and m is the Coulomb friction coefcient.

Ksp is the material-dependent contact force coefcient; Lw is ank
wear land width of the tool and vc is the cutting speed. The total
dynamic forces that affect the stability of cutting become:
Fd t Fcd t Fed t:

(15)

The effects of run-out, variable pitch and helix, and serrated

edge can also be added to the overall cutting forces [59,95].
3.3. Specic machining operations
Fig. 20. Cutters with arbitrary geometry [10].

The generalized force models can be adapted to various chip

removal operations by simply transforming the forces from RTA
(cutting edge) coordinates to the coordinates of the specic
machining operation. Only the operation and tool-geometry
specic engagement angle (c) and transformation matrices
(TR1,e) are needed as described in Figs. 1618 [9].
3.3.1. Single point turning operations
Single point turning and boring operations have the same
geometry, and the fundamental parameter is the side cutting edge
angle kr as shown in Fig. 16. The rotational matrices TOR and TRI
are automatically transformed to predict cutting forces in turning.

An example of simulated milling forces predicted by considering the thermo-mechanical properties of the material is shown in
Fig. 21 [38]. Instead of simulating the periodic states such as
milling forces and torque at discrete time intervals, Altintas et al.
[11] argued that the maximum force, torque, power and
dimensional surface errors are the most crucial information for
process planners. The maximum values of the process states must
not exceed the machine limits or the tolerance of the part. They
developed analytical, closed form formulas to predict the
maximum and minimum values of process states at each discrete

[(Fig._21)TD$IG]

592

[(Fig._2)TD$IG]

Y. Altintas et al. / CIRP Annals - Manufacturing Technology 63 (2014) 585605

Fig. 22. Time domain milling simulation. (a) Visualization of surface location errors
during the simulation of the machining process. (b) Analyzing the chip shape using
scanning rays (ISF).

Fig. 21. Comparison of simulated and measured forces using the thermomechanical model [38].

tool position along the tool path [98,99]. The closed form solutions
give the exact values of maximum states directly and accurately
thus avoiding the time marching solutions. They argued that a time
marching simulation can still be conducted if the process needs to
be interrogated at an unsafe tool path position.

[(Fig._23)TD$IG]

and analysis of the chip geometry if the cutterpart engagement

conditions do not change along the path [117].
Alternatively, the delayed differential equations can be solved
at discrete time intervals using the semi-discretization method
[14], which leads to a prediction of chatter stability, cutting forces,
dimensional surface errors and vibrations along the tool path [59].
The process damping reduces the vibration amplitudes at lower
speeds and increases the stability demonstrated by Budak et al.
[41,42]. Although the depth of cut is the same, the vibration
amplitudes are reduced signicantly at the lower cutting speed
due to the process damping effect, as in Fig. 23. The process
damping, which affects the prediction of surface quality at lower
speeds [124,125], can be easily considered in time-domain
simulation models, as shown in Eq. (11).

3.4. Time-domain simulation of machining processes

As the tool travels along the tool path, the cutterpart
engagement boundaries are evaluated at each discrete path
segment by one of the methods presented in Section 2. The
cutterworkpiece engagement maps contain the axial depth of cut
and width of cut along the cutting edge engaged with the material.
There are various approaches in simulating the process states such
as forces and vibrations. The time marching methods revolve
the spindle and move the tool along the tool path at discrete
time intervals and simulate the process states as a function of time
[114]. The process equations (Eq. (11)) are solved at each
time interval as demonstrated in boring [22,23,88], ve axis
milling [51,89,109]. Such methods are computationally costly, and
ne sampling intervals (i.e. 0.2 ms to capture 5000 Hz vibration
marks) are needed in order to simulate the maximum values of
forces in periodic processes like milling.
However, there have been efforts to improve the computational
speed of time domain simulations. Surmann et al. [118] developed
a time-domain simulation system for the milling process, in which
it is not necessary to explicitly solve the process equations. By
modeling the tool and the workpiece using CSG (Section 2.1), an
implicit model of the chip is obtained for each tooth feed. The
process forces are then calculated by analyzing this chip shape
using scanning rays for the determination of the undeformed chip
thickness (Fig. 45b), and subsequently applying an empirical force
model. In order to predict the tool vibrations, the process forces are
applied to a system of single-degree-of-freedom oscillators, which
is used to describe the dynamic behavior of the cutting tool [118].
This approach was extended and applied for the modeling of
workpiece vibrations by Kersting et al. [86,87]. The resulting tool
and workpiece deections are added to a CSG model of the tool
workpiece engagement that generates the chip shape in the
subsequent tooth period in order to consider the regenerative
effect [135]. The numerical simulation thereby provides a
prediction of cutting forces, relative vibrations between the tool
and workpiece, and dimensional form errors left on the nish
surface as shown in Fig. 22 [29,30]. The computational speed of
this approach can be further increased by avoiding the remodeling

Fig. 23. Vibration amplitude under process damping effect [41].

3.5. Grinding operations

Grinding processes can be simulated similarly to metal cutting
by evaluating the engagement geometry between the grinding
wheel and the workpiece along the tool path as presented in
Section 2. An overview of the grinding process modeling and
simulation techniques is given by Tonshoff et al. [121], Brinksmeier
et al. [34] and Brecher et al. [31]. Besides analytical methods,
process simulation models based on geometric-kinematic
approaches, nite element analysis and molecular dynamics are
widely used.
In contrast to the previously described metal cutting processes,
the cutting edges are not specied in abrasive processes due to the
large amount of grains and the variation of their sizes, shapes,
distribution and orientations. Unlike in metal cutting where the
cutting edges are dened and have a line contact with the material,
the grinding wheel has a contact surface which can be quite complex
in multi-axis machining where free-form surfaces are generated

[(Fig._25)TD$IG]

Y. Altintas et al. / CIRP Annals - Manufacturing Technology 63 (2014) 585605

593

[122]. In addition to the workpiece material properties, wear and

deections of the wheel strongly inuence the grinding process [78].
In analytical approaches, which can be used for the simulation
of process forces and operation states such as the resulting
maximum height of prole in surface grinding, the tool and
workpiece material properties are described by empirically
determined constants [121]. The specic normal force F 0n , per
unit width of cut (ap), can be expressed for surface grinding
processes (Fig. 24a) as [34]:

[(Fig._24)TD$IG]

F 0n cw p cgw

e1
vf
e3
ae2
e deq ;
vc

(16)

Fig. 25. Geometric-kinematic simulation of a at surface grinding process. (a)

Distribution of the grains on the grinding tool and the topography of the ground
surface. (b) Dexel-based representation of the workpiece with a dexel distance of
3 mm (ISF).

values of grinding processes [24]. In order to calculate the process

forces acting on the tool, two different methods can be applied [113].
In the rst method the engaged chip cross sectional areas Acu have to
be summed and multiplied by an empirically determined specic
cutting force coefcient kc,sim [24] (Fig. 24):
Ft kc;sim

N
X
Acu;i t:

(17)

i1

For the second force model, the undeformed chip thicknesses

have to be determined within the simulation system [113]. The
force model for each grain can be expressed as:
F nkc;sim bd0

Fig. 24. Engagement situation of tool and workpiece in (a) longitudinal surface
grinding (IFW) and (b) NC grinding (ISF).

where ae is the depth of cut, deq is the equivalent grinding wheel

diameter which reects the wheel engagement, and vc and vf are the
cutting speed and the feed velocity, respectively. Additionally, the
coefcients cwp for the workpiece material, cgw for the grinding wheel
and e1, e2 and e3 for the process parameters have to be determined
empirically from experiments. An extended version of these models
has been applied to other grinding operations such as internal or
external grinding, which require the contact area between the wheel
and the workpiece [34]. However, the applicability of analytical
models to grinding operations with varying contact conditions (e.g.
prole grinding [55,88,132]) is limited.
Geometric-kinematic simulation models offer a exible solution
for the determination of the tool engagement with the workpiece in
such grinding operations (Fig. 24b)[132]. The basic shapes of the
grinding tools can be represented with solid modeling techniques
(e.g. CSG) as described in Section 2. The contact area or the material
removal rate can be calculated by intersecting the tool and the
workpiece models [34,112]. Similar to the analytical approaches,
process forces can be calculated using empirical methods, and the
inuence of the distribution and shapes of the grains are considered
by the calibration of empirical constants.
Instead of an implicit representation of the cutting grains, the
geometry of the individual grains on the grinding tool can be
modeled [28,113]. These grains can be generated based on statistical
distributions [34,113] or measured tool topographies [78]. The
extended model can be used to estimate the topography of the
workpiece after nishing (Fig. 25a) or to determine characteristic

0 1mc;sim

d
d0

(18)

where n is the normal direction of the cutting grain face, b the

width of a considered segment of the cutting edge and d the
undeformed chip thicknesses. The coefcients kc,sim and mc,sim have
to be separately determined empirically for the calculation of
normal and tangential forces.
For example, by applying the simulation to drill-grinding
processes, the material removal per individual grain was analyzed
in order to allow improvement of the tool layout and thereby to
avoid material clogging [28]. The same basic approach was utilized
for simulating the NC grinding of free-formed surfaces [113]. By
analyzing the chip shape for each cutting grain, it is possible to
calculate the process forces, which can be used to adapt the tool path
for keeping the grinding forces at a desired level. For the denition
and calibration of the process force model, the geometric-kinematic
approach was coupled with a nite element simulation [112].
However, these extended simulation models are computationally costly due to small simulation time steps which are necessary
to evaluate the changing process conditions during the rotation of
the grinding tool. Moreover, high resolutions of the workpiece
models are necessary due to the small contact area of the
individual grains (Fig. 26).

[(Fig._26)TD$IG]

Fig. 26. Simulated single grain engagement (ISF).

[(Fig._28)TD$IG]

Y. Altintas et al. / CIRP Annals - Manufacturing Technology 63 (2014) 585605

594

4. Machine dynamic models

The dynamics of the machine tool are modeled by the rigid body
kinematics, CNC system and structural dynamics of the multi-axis
machine system. The NC tool path is followed by the machine at a
feed velocity governed by the kinematics and control of the
machine used by the CNC system. Any change in the feed along the
tool path affects the chip load, and hence the cutting process.
4.1. CNC and trajectory generation models of machines
A detailed review of CNC systems, their architecture and
functional modules has been presented in previous CIRP keynote
papers [15,107]. The virtual design and simulation of CNC systems
[66], which include trajectory generation [63], drive dynamics [64]
and servo control [65] have also been presented in detail by
Altintass research group [139,140]. Here, only the kinematics of
multi-axis machines and trajectory generation are briey summarized since they are most relevant to the prediction of chip
thickness, forces and vibrations in the virtual machining of a part.
NC programs contain the tangential feed velocity, tool tip and
orientations on the part coordinate (P) system as shown in Fig. 27.
The tool motions are transformed into axes commands in the
machine tool coordinate system (M) by the inverse kinematics of the
machine tool. The tangential feed commanded by the NC program
has to go through acceleration, constant feed and deceleration stages
without saturating the torque limits of each drive, while preserving
jerk continuity in order to have the smooth velocity prole as shown
in Fig. 28 [4]. The feed varies during the acceleration and
deceleration stages, and may not even reach the programmed
values in the 3- to 5-axis machining of sculptured surfaces if the path
segments are short and the drives do not have high accelerations (i.e.
torque) [111]. For example, if the jerk is constant (J0), the
acceleration will linearly increase, stay constant, and linearly
decrease when the feed reaches a desired, cruising velocity. The
machine will start decelerating by following the reverse phases of
acceleration. As a result, the feed will have seven distinct variation
zones within one NC block in the part program as follows [4]:
8
>
J t2
>
>
fs 0
>
>
2
>
>
>
>
f 1 At
>
>
>
>
J0 t2
>
>
>
< f 2 At  2
f t
f3 f4
>
>
>
J t2
>
>
f4  0
>
>
2
>
>
>
f 5  At
>
>
>
>
>
J t2
>
: f 6  At 0
2

zone h1i
h2i

Fig. 28. Trajectory generation prole with jerk continuity [63].

Correct feed must be calculated by considering the trajectory

prole of the CNC so that the chip loads, and hence the cutting
loads, are estimated accordingly. If the jerk is a second order
continuous system, the feed changes in nine distinct zones during
the machining of one NC block, which has to be considered in
predicting the process in virtual environments [4]. Including the
effect of the servo dynamics will not drastically improve the
evaluation of feedrates, i.e. chip loads, and hence it can be
disregarded for virtual machining process simulations.
4.2. Stability inspection of machining processes

h3i
h4i
h5i
h6i
h7i

[(Fig._27)TD$IG]

(19)

Process planners must select chatter-free global cutting speeds

and cutting conditions in preparing NC programs. However, since
the tool-part engagement conditions vary continuously along the
tool path, the virtual machining system must alert the planner to
the possibility of having chatter at specic locations. Instead of
predicting the chatter-free stability lobes, virtual machining must
inspect the presence of chatter and change the spindle speed, if
possible, or alert the planner to modify the process.
The transfer function of the structural dynamics of the machine
measured at the toolworkpiece contact zone can be constructed
in Laplace domain in the machine coordinates as:
2
3
Fxx s Fxy s Fxz s
(20)
F Ft Fw p 4
Fyy s Fyz s 5:
sym:
Fzz s
The corresponding relative vibrations between the tool and
workpiece can be predicted by applying the total cutting force (Eq.
(11)) on the machine at the toolworkpiece interface:
q31 F33 F31

(21)

where q31 represents the relative vibrations between the tool and
workpiece:
q qt qw p

Fig. 27. Kinematics of ve-axis machine tool [141].

(22)

The vibrations (Eq. (21)) and total dynamic cutting forces

(Eq. (11)) cannot be directly solved in time domain due to the

Y. Altintas et al. / CIRP Annals - Manufacturing Technology 63 (2014) 585605

regenerative term Dqt qt  qt  T with the time delay

(t  T) in the dynamic cutting force (Eq. (12)). The process can
either be solved in time domain using numerical integration
methods [100], or by converting the delayed dynamic equation
into an analytical, semi-discrete form [14,80]. However, it is more
practical to check whether the cutting system is stable or unstable
at the particular tool path position in the frequency domain as
follows.
The critical stability of the system can be determined by setting
the displacements and forces as:
qt q
vc eivc t ;

vc eivc t
Ft F

(23)

where vc is the chatter frequency and q and F are the amplitudes of

displacement and force vectors, respectively. Similarly, the delayed
displacement vector can be expressed as:
qt  T qe
ivc t eivc T :

(24)

Some of the machining operations, such as milling and boring

with an uneven number of inserts on the cutter, have time periodic
coefcients in their dynamic cutting forces (Eq. (11)). By averaging
the periodic coefcients in tooth passing or spindle periods [6], the
dynamic cutting force with regenerative and process damping
parts that affect the chatter stability [9,138] can be described as:
Ft A0 1  expivc T C0 dzqe
ivc t

(25)

where the averaged directional matrix (A0) is given as [35,36]

A0 V

1=V
0

N
X
j1

Z 2p X
N
1
f tdt
f cdc:
2p 0 j1

(26)

Since the cutter is engaged only when it is between entry (fst)

and exit (fex) angles (fst < c < fex), Eq. (26) can be simplied as:
8
9
2
3
f
Zex X
< K rc kdz = e t; kdz
N X
K
1
j
4
5  dc
T 01 K tc kdz
A0
:
; sinkr kdz
2p j1 k1
K
kdz
ac
2fst
3
8
9
f
Zex
K rc kdz = e c; kdz
<
K
X
N
N 1
j
6
7
A
dc5

4 T 01 K tc kdz
:
; sinkr kdz
2p k1
2p 2
K ac kdz
f

595

CAM systems [40,45,46]. However, since Eq. (30) is speed dependent

due to process damping, and the depth of cut is already assigned, the
stability is solved using the Nyquist criterion [68,69]. For a specic
spindle speed, the critical stable depth of cut is identied by
increasing the depth of cut a K  dz until the cutting system
becomes critically stable. If the depth of cut identied from the
cutterpart engagement module of the virtual machining system is
greater than the predicted critical depth, a chatter alert is sent to the
process planner at the analyzed tool path location [11].
The structural dynamics of the system may vary along the tool
path either due to the kinematic conguration of the machine tool,
structural changes in the part or removal of the mass from the thin
walled parts. It is possible to attach specic FRFs at each tool location
along the NC tool path in virtual environments [12,30], and to check
the corresponding chatter stability. However, if the dynamics
change rapidly, the part machining needs to be simulated in time
domain, and the FRF of the structure must be updated as the chips
are removed from the part. Budak et al. [52] proposed an FRF
updating scheme based on a matrix inversion method [102]. The FRF
of the nal part shape after machining is predicted with the FE
method and used as a core model. The mass [M], structural damping
[H] and stiffness [K] matrices are identied a priori. Instead of
removing machined elements from the part, un-machined elements
with their incremental mass [DM], damping [DH], and stiffness [DK]
are added at each tool path location and the FRF is updated as:

1
g  K DK  v2 M DM iH DH

(31)

where [g] is the receptance matrix and v is the vibration frequency

in rad/s. The method was applied in virtual, ve-axis ball-end
milling
of turbine blades as shown Fig. 29.
[(Fig._29)TD$IG]

(27)

st

Similarly, the process damping matrix (C0) in Eq. (25) is

evaluated by averaging as [59]:

Z fex X
N X
K 
e j t; kdz
1
K s p L2w
dc
T01 P
2p fst j1 k1 4vc
sinkr kdz
"Z
#
K
fex K L2
e j c; kdz
N X
N K s p L2w
sp w
dc

T01 P
C

2p k1 fst
2p 4vc
4vc
sinkr kdz
C0

(28)
Fig. 29. FE-based structural modication. (a) FE mesh of workpiece and stock, (b)
element determination [52].

The generalized A0 and C0 matrices become constant and time

invariant, but dependent only on the geometry of the tool and the
kinematics (A, C) of each operation as in the case of general force
models. Although considering time varying, periodic coefcients in
multi-frequency [35,96] or semi-discrete [14,80] solutions leads to
more accurate stability solutions when the radial depth of cuts are
small, they are computationally more costly to use in virtual
machining applications. By substituting Eq. (21) into Eq. (25), the
stability of machining operations is reduced to the following,
generalized eigenvalue problem:
h
i
n
o
ivc t 0
I  1  eivc T A0 C0 Fvc dz Fe
(29)
which leads to the following characteristic equation:

n
h
i
o
det I  1  eivc T A0 C Fvc dz 0:

It is common to use multi-functional machines such as mill-turns

in industry. While the simulation of the process forces, power and
torque of such operations can be simply done by superposing process
states, their chatter stability inspection differs due to the coupling
between them. A parallel turning system shown in Fig. 30 has the
following
coupled dynamics [39,49]:
[(Fig._30)TD$IG]

(30)
Fig. 30. Geometry of simultaneous turning and milling [49].

The chatter-free critical depth of cut and spindle speeds can be

directly predicted in frequency domain for milling [3,6], turning
[103] and drilling [110,120]; the form errors can be predicted and
constrained for process planning during NC tool path generation in

z1 t
z2 t


z 1 iv c t
e ;
z2

F 1 t
F 2 t


F 1 iv c t
e
F2

(32)

[(Fig._3)TD$IG]

Y. Altintas et al. / CIRP Annals - Manufacturing Technology 63 (2014) 585605

596

The stability analysis leads to optimal metal removal conditions

without violating the chatter limits of both tools used in the
operation, see Fig. 31. Since multi-functional machines are more
widely used, the chatter stability of combined turning, milling,
drilling and grinding operations must be further developed to
detect the abnormalities along the tool path in the virtual
machining
environment.
[(Fig._31)TD$IG]

Fig. 31. Stability diagram for a parallel turning operation with two tools.
Experimental results are shown by markers [39].

5. Virtual optimization of machining operations

Fig. 33. Flow chart of the optimization procedure [UBC MAL].

Machining parameters such as speeds, feedrates and depths

of cut need to be selected properly in order to improve the
process efciency and quality of the nished product. An
optimization strategy is proposed to determine the most
efcient machining parameters based on the process physics
[99]. The machining constraints include the cutting forces, chip
thickness, spindle torque-power, form errors on the workpiece
and system stability. The objective of the optimization is to
maximize the Material Removal Rate (MRR), which is dened as
[97]:
MRR a B c n N

[(Fig._32)TD$IG]

(33)

where a is the axial depth of cut, B is the radial width of cut, c is the
feed per tooth, n is the spindle speed, and N is the number of utes
on the milling cutter, as depicted in Fig. 32.

Fig. 32. Milling process with design variables [99].

The optimization strategy is divided into a pre-process and

post-process optimization as illustrated in Fig. 33.

stability. Therefore, the objective function of the pre-process

optimization reduces to [99]:
f ob j a B c:

(34)

5.1.1. Chatter stability constraint

Selection of the depth of cut, width of cut and spindle speed
should lead to stable cutting conditions, which is determined by
the stability lobe. Fig. 34(a) shows the stability lobe with a constant
width of cut (B), and Fig. 34(b) shows the stability lobe with a
constant depth of cut (a) when the width of cut is normalized with
respect to the tool diameter. By locating the optimum spindle
speed from these two gures, the design space shown as Fig. 34(c)
is formed by extracting a maximum stable depth and width of cut
at a constant spindle speed [98].
Stability limits increase drastically with the effect of process
damping at low cutting speeds [8]. Although its accurate
mathematical modeling is still an ongoing research topic, the
cutting edge geometry, ank wear, work materials contact
resistance, vibration frequency and cutting speed highly affect
the stability with process damping [138] as shown in Fig. 35 [124].
The inclusion of process damping in selecting a chatter-free,
optimal depth of cut is therefore crucial in machining thermal
resistant alloys in the aerospace industry.
5.1.2. Machine tool torque/power constraint
In order to avoid spindle stall, the torque/power limit of the
machine tool should not be exceeded. Both torque and power are
dependent on tangential cutting force Ft (w) as [97]:
8
>
>
<

5.1. Pre-process optimization

D
Torque f F t Nm
2


Dpn
>
>
1:34102  103 hp
: Power F t
60

Pre-process optimization provides an upper bound of the

cutting parameters to obtain an efcient machining operation
during the process planning stage. Since the cutter is chosen before
selecting the cutting conditions and generating the tool path, the
number of teeth of the tool (N) is known prior to the optimization.
The spindle speed (n) is selected based on tool life and chatter

where D is in (m), n is in (rev/min), and Ft is in (N). An analytical

solution of the tangential force for a cylindrical end mill is derived
as a function of an angular position w [99].
The critical angular positions tmin and tmax corresponding to
global minimum and maximum tangential forces are obtained. In
order to prevent the violation of torque/power constraints, the

(35)

[(Fig._34)TD$IG]

Y. Altintas et al. / CIRP Annals - Manufacturing Technology 63 (2014) 585605

597

cumulative average torque and power, which are dened as:

qcav

v"
#
u
u qt 2 qt 2
max
min
t

:
2
q fTorque;

(36)

Powerg

must not exceed the machine tool torque/power limitations of the

machine tool at spindle speed n:
Torquecav T mo n;

Powercav Pmo n

(37)

where Tmo and Pmo are machine tool torque and power curves
provided
by a manufacturer as shown in Fig. 36.
[(Fig._36)TD$IG]

Fig. 36. Torque-power characteristics of a machine tool [99].

5.1.3. Chip thinning constraint

When the width of cut is smaller than the radius of the tool (i.e.
b < 0.5), the maximum chip thickness does not reach the
commanded feed per tooth. A very small chip thickness leads to
a low material removal rate and to ploughing of the cutting edge on
the workpiece instead of cutting. The chip limit must be
maintained by manipulating the feed rate as the width of cut
varies. The feedrate cmax, which generates the desired maximum
chip load hmax, is calculated as:

cmax

8
9
hmax
< p
i f 0 < b 0:5 =
:
2 b1  b
:
;
hmax otherwise

(38)

The constraints are identied ahead of optimizing the NC

programs as the tool traverses along the tool path.
5.2. Post-process optimization

Fig. 34. Chatter stability lobes for constant axial depth of cut, constant radial depth
of cut and constant speed [99].

[(Fig._35)TD$IG]

Once the NC program is generated, the width and depth of cut,

i.e., the NC tool path, are dened. However, since the part geometry
varies along the tool path, the desired depth of cut and width of cut
cannot be maintained, which leads to a variation in the cutting
process. The cutterpart engagement boundaries along the tool
path are identied at desired displacements along the tool path,
and the feed and speed are optimized by respecting the physical
constraints of the process as outlined in the previous section. The
constraints include the maximum chip load, the resultant force on
the tool, form errors left on the part, and the torque/power
characteristics of the machine tool.
5.2.1. Feedrate optimization
The majority of the process output is linearly dependent on the
feedrate in the following form, provided that the cutting force
coefcients are independent of the feedrate:

Fig. 35. Effect of cutting parameters on process damping and stability in turning (MRL).

F q Aq0 Aq1 c
q
x; y; z; t; r; trq; pwr

(39)

Y. Altintas et al. / CIRP Annals - Manufacturing Technology 63 (2014) 585605

598

rates due to the continuously varying workpiece geometry, and

result in the saturation of axis motors and undesired feed marks on
the nished surface; therefore, optimized feed rates are ltered by
considering the trajectory generation parameters of the CNC
system as described in the machine dynamics section. It is
particularly important to achieve smooth feed changes to avoid
marks on the nish surface without violating the jerk and
acceleration limits of the drives [67,111].

where subscripts x, y, z, t, r are the cutting forces in the feed,

perpendicular to feed, axial, tangential, and radial directions; and
trq, pwr are the spindle torque and power, respectively. By
numerically solving the angular position qmax at which the process
output becomes the maximum dened as Fq,max, the maximum
allowable feedrate is solved as:
q
cmax

F q;max  Aq0 qmax

:
Aq1 qmax

(40)

5.2.4. Tool path optimization

Tool path selection is one of the critical parameters in planning
machining processes. In the commercial CAM packages, tool paths
are selected from standard path libraries by considering only the
geometry of the part. New approaches are needed to generate
optimized tool-paths for complex free-form surfaces. In recent
studies [90,94], the physical relationship between the mean
resultant forces, cycle times and scallop heights were introduced.
Since the three critical process outputs conict with each other, the
optimal path strategy was identied by using the objective
weighting algorithm. Moreover, the method allows for observation
of the trade-off between each criterion, and determination of the
corresponding toolpath for each solution. A sample threedimensional free-form surface generated with the optimized tool
[(Fig._38)TD$IG]path strategy is shown in Fig. 38.

The maximum resultant force in the xy plane is a quadratic

function of the federate [99], shown as:
q
res
res
2
F res Ares
(41)
0 A1 c A2 c :
The maximum feedrate is obtained by solving the root of
Eq. (41) when the resultant force is equal to the user-dened
maximum Fres,max.
Considering the form errors in milling, the maximum allowable
d fl
feedrate for a user-dened deection constraint dmax is solved as:
d fl
c1 c2  c1
cmax

dmax  d1
d2  d1

(42)

where c1, c2 are two arbitrary feed rates, and d1, d2 are the
corresponding calculated deections.
5.2.2. Feedrate and spindle speed optimization
The optimization of spindle speed is based on the constraints of
the torque-power characteristics of the machine and the stability
limit. For a xed depth of cut, the feasible spindle speed is in the
range that results in a stable cutting condition as:


n nmax;i ; n nmin;i
(43)
where nmin,i and nmax,i are the lower and upper spindle speeds
under a critically stable condition for the ith lobe. The graphical
representation of the design space considering the nonlinear
torque and power constrains is shown in Fig. 37. The spindle speed
and feedrate corresponding to the optimum solution (maximum
MRR) are obtained at the limits of the constraints, which is marked
with
a star in the Fig. 37.
[(Fig._37)TD$IG]

Fig. 38. Simulated and machined free-form surface and optimized tool path in 3D.

[(Fig._39)TD$IG]

The cycle times can be considerably long in the nish

machining of highly exible gas turbine blades [47]. The structural
dynamics of the blade varies both, in the axial and along the tool
path as shown in Fig. 39. The blade is far more exible than the
tool, and it becomes more rigid as the tool moves from the tip to the

Fig. 37. Graphical representation of design space [98].

5.2.3. CL le updating
The original cutter location (CL) le is used to obtain the cutter
workpiece engagement along the tool path, then the feedrate and
spindle speed are adjusted based on the user-dened constraints.
The optimization process might generate highly uctuating feed

Fig. 39. Stability of nishing and multi-level nishing.

[(Fig._41)TD$IG]

Y. Altintas et al. / CIRP Annals - Manufacturing Technology 63 (2014) 585605

599

hub of the blade. As a result, the stability limits and natural

frequencies vary along the blade surface as the cutting tool
removes the material [52]. A multi-level cutting strategy for blade
nishing is found to be more feasible in a virtual machining
environment, as opposed to the commonly used constant cutting
depth strategy (see Fig. 39). The blade surface is divided into four
segments along the axial direction. For each segment the spindle
speed and cutting depth are selected according to the largest
stability pocket of the corresponding segment. By changing the
cutting depth in the multi-level nishing strategy, the cycle time is
decreased from 35 minutes to 19 minutes while still respecting the
part quality requirements.

6. Virtual part machining examples

Typical virtual machining software is shown in Fig. 40. The
system reads the standard CL le from a CAM system, parses the NC
program and separates the operations by identifying tool change
commands, and either simulates or optimizes the process by
predicting/optimizing the maximum forces, torque, power, form
errors and chip loads along the tool path. The system automatically
breaks large tool paths into smaller segments by respecting the
CNCs trajectory generation prole and gives a new CL le with
optimized feed commands. The original NC tool path geometry is
unchanged. There are few commercially available virtual machining systems, and they all operate in a similar fashion except that
their process models and optimization strategies may differ. The
prediction accuracy of virtual machining systems is most affected
by the cutting force coefcients, the mathematical model of the
physical processes and cutterworkpiece engagement conditions
[50]. Any error in these three modules will reect directly on the
accuracy of the virtual prediction and optimization of NC
programs. The cutting force coefcients can be mechanistically
calibrated for each cutterwork material couple to minimize their
effect on the prediction accuracy for the mass production of parts,
but generalized orthogonal-to-oblique cutting transformation
methods would be less costly for small batch production [2]. It
is ideal to digitize the cutterpart engagement conditions at feed
rate increments so that any small change in part geometry can be
detected along the tool path. However, a densely digitized tool
path may unnecessarily increase the computation time and make
the
[(Fig._40)TD$IG] virtual machining system unfeasible to use in production.
Fig. 41. Virtual machining of a stamping die, predicted maximum cutting forces
with MACHPROTM and measured forces along the tool path
(Courtesy of Sandvik Coromant and MAL Inc.).

Fig. 40. A sample view of a virtual machining system.

(Courtesy of Sandvik Coromant and MAL Inc.).

6.1. Metal cutting applications

A sample three-axis milling of a stamping die with circular
inserts is shown in Fig. 41. The accuracy of the simulation system
was veried by comparing the predicted and measured maximum
cutting forces along the tool path. The process had an experience of

severe chip thinning at sharp curvatures, which was eliminated

while decreasing the machining time by 25% with the feed
regulation. Virtual machining is most crucial in manufacturing
very costly aerospace parts, since their trial- and error-based
optimization may be cost prohibitive. A sample optimization of a
jet engine impeller is shown before and after virtual machining in
Fig. 42. The cutter was a taper helical ball end mill, whose pitch
angles are designed to maximize the chatter-free depth of cuts at
the desired cutting speeds [18,43,44]. The virtual process
optimization parameters included maximum stress at the tool
shank to avoid tool failure, chip load and torque limit on the
spindle and feed drives [47,70,71].
Virtual machining with process forces and deections can be
complemented with visualization by updating the part geometry
using solid modeling methods. The result of simulating the
peripheral milling of a turbine blade with modeled workpiece and
tool vibrations is shown in Fig. 43a. Both spherical end mill and
workpiece geometries were modeled by CSG where the workpiece
and tool deections were applied by altering the CSG model of the
tool for each tooth feed. This allows the visualization of the
vibration marks on the nish surface by rendering with ray-tracing

[(Fig._42)TD$IG]

600

[(Fig._4)TD$IG]

Y. Altintas et al. / CIRP Annals - Manufacturing Technology 63 (2014) 585605

Fig. 42. Five axis ank milling of a jet engine impeller before and after optimizing
with virtual machining system. The cycle time is reduced by 62% and the surface
nish is improved by 8.4 fold.
(Courtesy of Pratt & Whitney Canada).

[(Fig._43)TD$IG]

Fig. 44. Toolpath and the broken tool (MRL).

[(Fig._45)TD$IG]

indicated that the axial cutting force increased from 150 N to 450 N
as seen in Fig. 44, where the tool was digging into the material due
to a negative effective lead angle [58]. The problem was solved by
re-orienting the tool axis vector that minimized the indentation of
the tool into the material [126,127]. The effect of tool orientation,
i.e. lead and tilt angles, on the process mechanics is quite important

Fig. 43. Simulated surface structures. (a) Simulation of the machining of a turbine
blade taking workpiece vibrations into account [29]. (b) Simulation surface location
errors, (c) simulation of structures generated during face milling [32].

techniques (Fig. 43b) [29,118]. For the modeling of the surfaces

generated during face milling, another modeling technique has to
be applied. Since the tool is constantly engaged with the
workpiece, an altered CSG model for each tooth feed is not
sufcient. For the simulation of this process, the geometry of the
cutting edge can be approximated by an open traverse line, whose
points are adjusted corresponding to the tool deections. By
connecting the lines from subsequent time steps, a triangulated
sweep surface is obtained, which can be projected directly onto a
height eld in order to visualize the surface structure (Fig. 43c).
Another approach allows the direct modeling and visualization of
the surface location error without the need for ray-tracing or
triangulation techniques. It is based on the combination of the CSG
technique for the determination of the process forces and the
dynamic behavior, and a dexel-based workpiece model for the
visualization of surface location errors. Each time a dexel is cut, the
intersection points can be displaced by the projected vibration
amplitude. This applied displacement also corresponds directly to
the surface location error (Fig. 43b). The combination of the two
different workpiece models also permits the identication of
constant engagement situations for the optimization technique
described in Section 3.4. Alternatively, it is possible to estimate the
errors left on the nish surface by correlating the measured forces
and stiffness of the system in a virtual environment as well [54].
A problematic tool path for roughing of an integrally bladed
rotor, which leads to frequent breakage of a serrated tool at its ball
end, is shown in Fig. 44. The multi-axis virtual machining software

Fig. 45. Roughing toolpath and kinematic proles for a sample pass, resultant
cutting forces and cycle time comparison [67].

Y. Altintas et al. / CIRP Annals - Manufacturing Technology 63 (2014) 585605

as presented by Ozturk et al. [106] who reported that the lead angle
of about +10 degrees would be most suitable as demonstrated in
this application.
Once the optimal feed is identied from the process, a further
reduction in machining time can be achieved by optimizing the
trajectory proles in three- to ve-axis machining of free-form
surfaces. The CNC systems may not be able to achieve the
commanded feeds when the paths have sharp curvatures,
demanding high torque and acceleration from feed drives. It is
possible to reduce the machining cycle time by re-shaping the
trajectory commands along the curved paths without violating the
feed drive limits and process-imposed federate along the tool path.
Erkorkmaz et al. [67] and Sencer et al. [111] demonstrated that up
to a 65% reduction can be achieved in the machining of free-form
parts via trajectory optimization. The tool path for a free-form
surface shown in Fig. 45 was generated with 0.04 mm tolerance,
leading to 11,186 CL points for the complete operation. Processing
by the solid modeler, applying the B-rep method, took 482 s and
consumed 1.2 GB of memory. In feed planning, the resultant force
limit was set to 250 N. A Mori Seiki NMV 5000DCG machining
center was used in the cutting tests. The drives velocity,
acceleration and jerk limits were identied by inspecting the
corresponding registers in the CNC. The process was rst
optimized by varying the feed while keeping the cutting force at
250 N [62]. At the second step, the feed was optimized again by
respecting the acceleration and jerk limits of the drives, which led
to a further 17% reduction in machining time [67].
6.2. Grinding applications
When grinding complex parts such as airfoils and blade
retention slots as shown in Fig. 46, the complex wheel/workpiece
geometry and multi-axis motion result in variable wheel
workpiece contact. All grinding process parameters such as depth
of cut, wheelspeed, workspeed, forces and temperature will vary
along
the wheel axis and grinding path. Optimizing such grinding
[(Fig._46)TD$IG]

601

processes requires detailed knowledge about the distributions of

these process parameters. A virtual environment is developed to
simulate and optimize ve-axis grinding with complex wheel and
work geometries [75]. Boolean intersection algorithms are used to
determine the shape and geometrical characteristics of the wheel
workpiece contacts at any instant. The contact geometry is then
analyzed as a stack of discrete sub-wheels for which the physical
process parameters such as forces, power and temperature are
predicted using the contact data. Grinding examples show that the
grinding process parameters vary signicantly along the wheel
axis at any instant and along the grinding path. The grinding
process is far from optimal if a constant workspeed is used. Multiconstraint optimization is then applied to optimize the workspeed
to reduce cycle time while maintaining grinding forces, power and
temperature below the specied limits as shown in Fig. 46. The
optimization leads to about a 40% cycle time reduction, while the
maximum power is reduced by >30% and the maximum force
reduced by about 35% [75].
A good example of a virtual grinding application is the
manufacturing of tungsten carbide end mills. Due to the hardness
of tungsten carbide, cutting tools are manufactured using deep
grinding processes with diamond or CBN-grains. Because of the
large grinding forces, the exible end mill which has varying
stiffness along its axis, experiences static deections and vibrations. The deections lead to geometrical errors and poor cutting
edge grinding quality which need to be optimized through virtual
simulations. The virtual model considers the kinematics of the 8axis tool grinding machine, the discrete removal of material and
the tool-grinding wheel engagement zone, which are used to
[(Fig._47)TD$IG]predict the grinding process [53] (Fig. 47).

Fig. 47. Calculated material removal rate Qwa(i) (mm3/s) and (b) calculated
equivalent chip thickness heq (mm) for a tool grinding process (vft = 30 mm/min,
d = 10 mm, R = 62.0 mm) [53].

Fig. 46. Grinding of jet engine impellers with tapered ball ended tools (a) and form
grinding (b) of turbine blade retention slots.
(Courtesy of United Technologies Research Center).

The workpiece discretization is based on dexel grids in all three

coordinate directions. The cylindrical grinding wheel is approximated by a non-rotating polyhedron which is reduced to a pie
segment. During each simulation step, the grinding wheel moves
in a step-wise linear motion, and the intersection of the sweep
volume and workpiece is calculated to estimate grinding forces. To
map values to different areas, the grinding tool is divided into slices
along the tool axis. By dividing the removed volume of each dexel

602

Y. Altintas et al. / CIRP Annals - Manufacturing Technology 63 (2014) 585605

at discrete simulation time steps and mapping it to the

corresponding slice, the distribution of the material removal rate
Qwa (Fig. 47a) and equivalent chip thickness heq (Fig. 47b) can be
determined for each element. A constant cutting speed is assumed,
which is reasonable due to the cylindrical shape of the grinding
wheel.
To parameterize an empirical grinding force model, at
grinding experiments using a cylindrical grinding wheel and a
rectangular workpiece have been carried out on a Walter
Helitronic Power tool grinding machine. The resulting forces for
different depths of cut have to be related to geometric parameters,
which can both be evaluated in the 3D dexel simulation. The factor
heq/dlg relates the equivalent chip thickness heq to the local contact
length dlg for each segment of the contact area. This dimensionless
value replaces the factors for feed speed, cutting speed and
grinding wheel diameter, which are commonly used in grinding
force models [121].
The grinding forces are applied on the position-dependent
structural model of the ground end mill at each simulation step.
Since a deection directly affects the contact conditions and
therefore the acting grinding forces, the coupled processstructure
interaction is handled with an iterative model. The algorithm
determines the deformation of the workpiece and machine
structure during grinding at a static equilibrium of the grinding
force, and the spring-back force of the workpiece until the
deformations calculated in the previous and current steps
converge within a dened tolerance. Subsequently, the material
is removed, the current shape is dened and the calculation step is
repeated. Once the deections are predicted by the virtual grinding
model, the original tool path is modied to compensate them. The
iterative simulation process is repeated until the predicted
workpiece geometry and the ground end mill match with the
desired geometry within its specied tolerance as shown in Fig. 48.
The deections of the ground end mill are reduced from 163 mm to
35
[(Fig._48)TD$IG] mm as shown in [53,55].

Fig. 48. Comparison of resulting workpiece cross-sections with uncompensated NCprogram (left) and compensated tool path (right) [53].

For optimizing the grinding of wear-resistant coated freeformed surfaces on machining centers using abrasive mounted
points, a simulation system based on multi-dexel boards and CSGtechniques is used [113]. Due to the inhomogeneus thickness of the
thermally sprayed coating and the varying contact areas between
the grinding tool and the free-formed workpiece surface (Fig. 49),
[(Fig._49)TD$IG]the grinding forces vary along the NC path. These variations can

Fig. 49. Process simulation of grinding of wear-resistant coated free-formed

surfaces on machining centers using abrasive mounted points (ISF).

lead to shape and dimensional errors as well as surface defects. In

order to counteract these effects, the NC path can be adjusted
based on the simulation results. The process force model uses the
engagements of the individual grains, as described in Section 3.5. In
addition to the calculated forces, the generated surface roughness
can be approximated within the simulation system.
7. Uncertainties in virtual machining
Virtual machining is based on the mathematical modeling of
part-tool engagement geometry, machining process physics,
structural and rigid body kinematic motion of the machine and
work material properties. The accuracy of the prediction will be
dependent not only on how well the mathematical models
approximate the complex machining process physics, but also
on how accurately the input parameters are entered to the virtual
machining system. The accuracy of mathematical models have
been well reported in previous dedicated metal cutting [2,20],
grinding [79,121], chatter stability [16] and virtual machine tool
[5] key note articles. The effects of uncertainties in the key input
parameters are summarized as follows.
7.1.1. Geometric variations
Assuming that the part is accurately placed on the xture and
machine, the key source of uncertainties originate from casting,
forging and semi-nish tolerances. The unexpected extra stock left
on the part results in an under-prediction with the same
proportion, i.e. 5% extra stock leads to 5% less magnitudes in
predicted force, power and torque. The mesh size in the cutter
workpiece engagement affects the prediction results similar to the
forging errors.
7.1.2. Cutting process models
The key uncertainties originate from the cutting force
coefcients, which depend on the tool geometry (rake angle, helix
angle, nose radius, cutting edge radius and inclination angle), and
material properties such as temperature-dependent hardness and
ow stress, tool wear and lubrication (i.e. Coulomb friction). Errors
in the cutting force coefcient affect the cutting forces, torque and
power in the same proportion. Changes in the materials yield
shear stress directly and linearly affect the cutting force coefcient.
For example, if the tangential cutting force coefcient for AISI steel
is 1500 MPa with 10% uncertainty, the prediction of cutting forces,
torque and power will also have the same uncertainty (10%). The
effect of uncertainties in tool geometry varies. While the effect of rake
and helix angles is about 2% per degree, the cutting edge radius and
grain size of the grinding wheel have a major effect, especially in
nish machining where the chip loads can have the same magnitude.
It is recommended that the cutting force coefcients are calibrated
against the edge radius, grain size and lubricant for each material as
explained in [121,139].
7.1.3. Machine dynamic models
The forced and chatter vibrations between the tool center point
and workpiece depend on the relative dynamic stiffness between
the two and the process forces. The chatter stability is linearly
proportional to the dynamic stiffness (2k ). If the measurement of
the dynamic stiffness has an error of +10%, the chatter-free depth of
cut will be underestimated by 10%. If the natural frequencies of
the machine tool structure varies along the tool path due to
kinematic and cross coupling effects [33], the stability pockets will
also shift. One Hz error in the natural frequency measurement will
result in a chatter-free spindle speed shift of 60 rev/min/N where N
is the number of teeth on the tool. Often, the structural dynamics of
thin-walled parts change signicantly during machining, which
need to be carefully mapped to the tool path as discussed in the
dynamic section.

Y. Altintas et al. / CIRP Annals - Manufacturing Technology 63 (2014) 585605

603

7.1.4. CNC models

Acknowledgement

The positioning errors of CNC models have a negligibly small

effect on process simulation in virtual machining. However, the
chip loads in cutting and grinding are direct functions of feeds
along the tool path. If the tool path is curved, especially in threeand ve-axis machining of free-form surfaces, the feed changes
continuously as a function of path curvature, acceleration and jerk
settings of the CNC (i.e. trajectory generation). Although the
trajectory generation settings of the CNC are considered in virtual
machining [11,67,99], any uncertainty in the acceleration and jerk
have a third and second order polynomial relationship with the
feed, hence the forces as explained in the paper. Typically, 1015%
errors may originate from the approximated trajectory models of
the CNC systems in virtual machining of free-form surfaces.

The following co-workers have assisted in preparing the

manuscript: Murat Kilic, Dr. Xioliang Jin and Dr. Doruk Merdol
(MAL, University of British Columbia), Dr. Volker Bo (IFW, Leibniz
Universitat Hannover), Dr. Taner Tunc (MRL, Sabanci University),
Dr. G. Guo (United Technologies Research Center), Mikael
Lundblad (Sandvik Coromant Engineering Center).

8. Conclusion
The modeling of machining operations has been evolving as an
important engineering tool to simulate operation physics ahead of
costly production trials of parts used in industry.
Virtual machining technology has three important components
which affect the accuracy of prediction, computational speed and
visualization of the operation by the process planners.
The rst step is the evaluation of cutterworkpiece engagement
conditions along the tool path. The tool and workpiece geometries
are extracted from the CAM systems, and various solid modeling
methods have been developed with varying computational
efciency, accuracy and visualization. The accuracy of virtual
machining is directly related to the identication of the cutter
workpiece engagement conditions. Accurate engagement predictions lead to impractically long simulation times for industry, and
rough engagement predictions lead to inaccuracies in process
simulations. Further research is needed to improve both the
accuracy and computational efciency of cutterworkpiece
engagement conditions.
The mathematical models of metal cutting and grinding
processes must be well developed with computational efciency
and accurate physics. There have been major advances in
predicting the micro-mechanics of cutting (i.e. stress, temperature,
white layer characteristics of the nish surface, thermo-mechanical behavior of the material) as well as the macro-mechanics of
cutting (i.e. forces, torque, power, vibrations, part deformation
errors). While the micro-mechanics models are used to design the
process and to analyze the operation at specic part locations, the
macro-mechanics are used to optimize the general machining
operation in CAM environments. There are still challenges in
achieving computationally efcient but accurate modeling of
cutting forces, structural deformations and chatter stability when
complex cutting tools are used in NC programs.
The third component of the virtual machining system is to
model the rigid body kinematic motion of the machine and its
position-dependent structural dynamics. The tangential feed
velocity between the workpiece and tool varies as a function of
the machines kinematic and CNC conguration, which has to be
included in predicting the chip loads along the tool path. The
structural dynamics of both the machine and workpiece may also
change along the tool path, and they need to be incorporated into
the virtual machining system in order to predict deections and
vibration marks imprinted on the nish surface of the part.
Furthermore, machine tools have volumetric errors which are
reected on the part. The volumetric errors of the machine tool
must be integrated into virtual machining systems to predict these
metrology errors.
Although a signicant amount of research still must be
conducted to achieve highly efcient and accurate virtual
machining systems, the current know-how is already useful in
minimizing scrap rates and maximizing production efciency in
industry.

References
[1] Abele E, Altintas Y, Brecher C (2010) Machine Tool Spinde Units. CIRP Annals
59(2):781802.
[2] Altintas Y (2000) Modeling Approaches and Software for Predicting the
Performance of Milling Operations at MAL-UBC. International Journal of
Machining Science and Technology 4(3):445478.
[3] Altintas Y (2001) Analytical Prediction of Three Dimensional Chatter Stability
in Milling. Japan Society of Mechanical Engineers. International Journal Series
C: Mechanical Systems Machine Elements and Manufacturing 44(3):717723.
[4] Altintas Y (2012) Manufacturing Automation, Cambridge University Press, UK.
[5] Altintas Y, Brecher C, Weck M, Witt S (2005) Virtual Machine Tool. CIRP Annals
54(2):115138.
[6] Altintas Y, Budak E (1995) Analytical Prediction of Stability Lobes in Milling.
CIRP Annals 44(1):357362.
[7] Altintas Y, Engin S (2001) Generalized Modeling of Mechanics and Dynamics
of Milling Cutters. CIRP Annals 50(1):2530.
[8] Altintas Y, Eyniyan M, Onozuka H (2008) Identication of Dynamic Cutting
Force Coefcients and Chatter Stability with Process Damping. CIRP Annals
57(1):371374.
[9] Altintas Y, Kilic ZM (2013) Generalized Dynamic Model of Metal Cutting
Operations. CIRP Annals 62(1):4750.
[10] Altintas Y, Lee P (1996) A General Mechanics and Dynamics Model for Helical
End Mills. CIRP Annals 45(1):5964.
[11] Altintas Y, Merdol DS (2007) Virtual High Performance Milling. CIRP Annals
55(1):8184.
[12] Altintas Y, Montgomery D, Budak E (1992) Dynamic Peripheral Plate Milling
of Flexible Structures. Transactions of ASME Journal of Engineering for Industry
114:137145.
[13] Altintas Y, Spence A (1991) End Milling Force Algorithms for CAD Systems.
CIRP Annals 40(1):3134.
[14] Altintas Y, Stepan G, Merdol D, Dombavari Z (2008) Chatter Stability of
Milling in Frequency and Discrete Time Domain. CIRP Journal of Manufacturing Science and Technology 1:3544.
[15] Altintas Y, Verl A, Brecher C, Uriarte L, Pritschow G (2011) Machine Tool Feed
Drives. CIRP Annals 60(2):779796.
[16] Altintas Y, Weck M (2004) Chatter Stability in Metal Cutting and Grinding.
CIRP Annals 53(2):619642.
[17] Altintas Y, Xiaoliang J (2011) Mechanics of Micro-Milling with Round Edge
Tools. CIRP Annals 60(1):7780.
[18] Altintas Y, Engin S, Budak E (1999) Analytical Prediction of Chatter Stability
and Design for Variable Pitch Cutters. Transactions of ASME Manufacturing and
Engineering and Science 121:173178.
[19] Ambati R, Pan X, Yuan H, Zhang X (2012) Application of Material Point Methods
for Cutting Process Simulations. Computational Materials Science 57:102110.
zel T, Umbrello D, Davies M, Jawahir IS (2013) Recent Advances
[20] Arrazola PJ, O
in Modelling of Metal Machining Processes. CIRP Annals 62(2):695718.
[21] Armarego E (2000) The Unied-Generalized Mechanics of Cutting Approach A
Step Towards a House of Predictive Performance Models for Machining Operations.
International Journal of Machining Science and Technology 4(3):319362.
[22] Atabey F, Lazoglu I, Altintas Y (2003) Mechanics of Boring Processes Part I.
International Journal of Machine Tools and Manufacture 43:463476.
[23] Atabey F, Lazoglu I, Altintas Y (2003) Mechanics of Boring Processes Part II:
Multi-Insert Boring Heads. International Journal of Machine Tools and Manufacture 43:477484.
[24] Aurich J, Kirsch B (2012) Kinematic Simulation of High-Performance Grinding
for Analysis of Chip Parameters of Single Grains. CIRP Journal of Manufacturing
Science and Technology 26(5):164174.
[25] Aurich JC, Biermann D, Blum H, Brecher C, Carstensen C, Denkena B, Klocke F,
Kroger M, Steinmann P, Weinert K (2009) Modelling and Simulation of
Process: Machine Interaction in Grinding. Production Engineering Research
and Development 3(1):111120.
[26] Benouamer M, Michelucci D (1997) Bridging the Gap Between CSG and Brep
via a Triple Ray Representation. ACM Symposium on Solid Modeling and
Applications 6879.
[27] Bergs T, Rodriquez CA, Altan T, Altintas Y (1996) Tool Path Optimization for
Finish Milling of Die and Mold Surfaces Software Development. Transactions
of the NAMRI/SME XXIV 8186.
[28] Biermann D, Feldhoff M (2012) Abrasive Points for Drill Grinding of Carbon
Fiber Reinforced Thermoset. CIRP Annals 61(1):299302.
[29] Biermann D, Kersting P, Surmann T (2010) A General Approach to Simulating
Workpiece Vibrations During Five-Axis Milling of Turbine Blades. CIRP Annals
59(1):125128.
[30] Biermann D, Surmann T, Kersting P (2013) Oscillator-Based Approach for
Modeling Process Dynamics in NC Milling with Position- and Time-Dependent Modal Parameters. Production Engineering Research and Development
7(4):417422.

604

Y. Altintas et al. / CIRP Annals - Manufacturing Technology 63 (2014) 585605

[31] Brecher C, Esser M, Witt S (2009) Interaction of Manufacturing Process and

Machine Tool. CIRP Annals 58(2):588607.
[32] Breitensprecher T, Hense R, Hauer F, Wartzack S, Biermann D, Willner K
(2012) Acquisition of Heuristic Knowledge for the Prediction of the Frictional
Behavior of Surface Structures Created by Self-Excited Tool Vibrations. Key
Engineering Materials 504-506:963968.
[33] Bringmann B, Maglie P (2009) A method for Direct Evaluation of the Dynamic
3D Path Accuracy of NC Machine Tools. CIRP Annals 58(1):343346.
[34] Brinksmeier E, Aurich JC, Govekar E, Heinzel C, Hoffmeister HW, Klocke F,
Peters J, Rentsch R, Stephenson DJ, Uhlmann E, Weinert K, Wittmann M
(2006) Advances in Modeling and Simulation of Grinding Processes. CIRP
Annals 55(2):667696.
[35] Budak E, Altintas Y (1998) Analytical Prediction of Chatter Stability in
Milling-Part I: General Formulation. Transactions of ASME Journal of Dynamic
Systems Measurement and Control 120:2230.
[36] Budak E, Altintas Y (1998) Analytical Prediction of Chatter Stability in
Milling-Part II: Application of the General Formulation to Common Milling
Systems. Transactions of ASME Journal of Dynamic Systems Measurement and
Control 120:3136.
[37] Budak E, Ozlu E (2007) Analytical Modeling of Chatter Stability in Turning and
Boring Operations: A Multi-Dimensional Approach. CIRP Annals 56(1):401
404.
[38] Budak E, Ozlu E (2008) Development of a Thermomechanical Cutting Process
Model for Machining Process Simulations. CIRP Annals 57(1):97100.
[39] Budak E, Ozturk E (2011) Dynamics and Stability of Parallel Turning Operations. CIRP Annals 60(1):383386.
[40] Budak E, Tekeli A (2005) Maximizing Chatter Free Material Removal Rate in
Milling through Optimal Selection of Axial and Radial Depth of Cut. CIRP
Annals 54(1):353356.
[41] Budak E, Tunc LT (2009) A New Method for Identication and Modeling of
Process Damping in Machining. Transactions of ASME Journal of Manufacturing
Science and Technology 131(5). 051013 (10 pages).
[42] Budak E, Tunc LT (2010) Identication and Modeling of Process Damping in
Turning and Milling Using a New Approach. CIRP Annals 59(1):403408.
[43] Budak E (2003) An Analytical Design Method for Milling Cutters with Non
Constant Pitch to Increase Stability-Part I: Theory. Transactions of ASME
Journal of Manufacturing Science and Engineering 125:2934.
[44] Budak E (2003) An Analytical Design Method for Milling Cutters with Non
Constant Pitch to Increase Stability-Part II: Application. Transactions of ASME
Journal of Manufacturing Science and Engineering 125:3538.
[45] Budak E (2006) Analytical Models for High Performance Milling-Part I:
Forces, Form Error and Tolerance Integrity. International Journal of Machine
Tools and Manufacture 46:14781488.
[46] Budak E (2006) Analytical Models for High Performance Milling-Part II:
Process Dynamics and Stability. International Journal of Machine Tools and
Manufacture 46:14891499.
[47] Budak E (2000) Improvement of Productivity and Part Quality in Milling of
Titanium Based Impellers by Chatter Suppression and Force Control. CIRP
Annals 49(1):3136.
[48] Budak E, Altintas Y, Armarego EJA (1996) Prediction of Milling Force Coefcients From Orthogonal Cutting Data. Transactions of ASME Journal of Engineering for Industries 118(2):216224.
zturk E (2013) Stability and High Performance Machining
[49] Budak E, Comak A, O
Conditions in Simultaneous Milling. CIRP Annals 62(1):403406.
[50] Budak E, Lazoglu I, Guzel BU (2004) Improving Cycle Time in Sculptured
Surface Machining Sculptured Surface Machining Through Force Modeling.
CIRP Annals 53(1):103106.
[51] Budak E, Ozturk E, Tunc LT (2009) Modeling and Simulation of 5-Axis Milling
Processes. CIRP Annals 58(1):347350.
[52] Budak E, Tunc LT, Alan S, Ozguven HN (2012) Prediction of Workpiece
Dynamics and its Effects on Chatter Stability in Milling. CIRP Annals
61(1):339342.
[53] Deichmuller M, Denkena B, Payrebrune KM, de, Kroger M, Wiedemann S,
Schroder A, Carstensen C (2013) Modeling of Process Machine Interactions in
Tool Grinding. Process Machine Interactions, Springer, Berlin, Heidelberg143176.
[54] Denkena B, Kruger M, Bachrathy D, Stepan G (2012) Model Based Reconstruction of Milled Surface Topography from Measured Cutting Forces.
International Journal of Machine Tools and Manufacture 54-55:2533.
[55] Denkena B, Turger A, Behrens L, Krawczyk T (2012) Five-Axis-Grinding With
Toric Tools: A Status Review. Transactions of ASME Journal of Manufacturing
Science and Engineering 134(5). 54001 (6 pages).
[56] Denkena B, Tracht K, Yu JH (2006) Advanced NC-Simulation based on the
Dexelmodel and the HRMC-Algorithm. Production Engineering Research and
Development 13(1):9194.
[57] Du S, Surmann T, Webber O, Weinert K (2005) Formulating Swept Proles for
Five-Axis Tool Motions. International Journal of Machine Tools and Manufacture 45(78):849861.
[58] Ozturk E, Budak E (2010) Dynamics and Stability of 5-Axis Ball-end Milling.
Transactions of ASME Journal of Manufacturing Science and Engineering 132(2).
021003 (12 pages).
[59] Eksioglu C, Kilic ZM, Altintas Y (2012) Discrete-Time Prediction of Chatter
Stability, Cutting Forces, and Surface Location Errors in Flexible Milling
Systems. Transactions of ASME Journal of Manufacturing Science and Engineering 134(6). AN061006.
[60] Engin S, Altintas Y (2001) Modeling of General Milling Operations: Part I End
Mills. International Journal of Machine Tools and Manufacture 41:21952212.
[61] Engin S, Altintas Y (2001) Modeling of General Milling Operations: Part II
Inserted Cutters. International Journal of Machine Tools and Manufacture
41:22132231.
[62] Erdim H, Lazoglu I, Ozturk B (2006) Feedrate Scheduling Strategies for Free-Form
Surfaces. International Journal of Machine Tools and Manufacture 46:747757.

[63] Erkorkmaz K, Altintas Y (2001) High Speed CNC System Design: Part I Jerk
Limited Trajectory Generation and Quintic Spline Interpolation. International
Journal of Machine Tools and Manufacture 41(9):13231345.
[64] Erkorkmaz K, Altintas Y (2001) High Speed CNC System Design: Part II
Modeling and Identication of Feed Drives. International Journal of Machine
Tool and Manufacture 41(10):14871509.
[65] Erkorkmaz K, Altintas Y (2001) High Speed CNC System Design: Part III High
Speed Tracking and Contouring Control of Feed Drives. International Journal of
Machine Tool and Manufacture 41(11):16371658.
[66] Erkorkmaz K, Altintas Y, Yeung CH (2006) Virtual Computer Numerical
Control System. CIRP Annals 55(1):399402.
[67] Erkorkmaz K, Layegh SE, Lazoglu I, Erdim H (2013) Feedrate Optimization for
Freeform Milling Considering Constraints from the Feed Drive System and
Process Mechanics. CIRP Annals 62(1):395398.
[68] Eyniyan M, Altintas Y (2009) Chatter Stability of General Turning Operations
with Process Damping. Transactions of ASME Journal of Manufacturing Science
and Engineering 131(4):110.
[69] Eyniyan M, Altintas Y (2010) Analytical Chatter Stability of Milling with
Rotating Cutter Dynamics at Process Damping Speeds. Transactions of ASME
Journal of Manufacturing Science and Engineering 132(2). 021012 (14 pages).
[70] Ferry W, Altintas Y (2008) Virtual Five Axis Milling of Impellers, Part I:
Mechanics of Five Axis Milling. Transactions of ASME Journal of Manufacturing
Science and Engineering 130(1). 011005 (11 pages).
[71] Ferry W, Altintas Y (2008) Virtual Five Axis Milling of Impellers, Part II:
Feedrate Optimization of Five Axis Milling. Transactions of ASME Journal of
Manufacturing Science and Engineering 130(1). 0110013 (13 pages).
[72] Foley J, van Dam A, Feiner S, Hughes J (1995) Computer Graphics: Principles and
Practice, Addison-Wesley.
[73] Fussell BK, Jerard RB, Hemmett JG (2003) Modeling of Cutting Geometry and
Forces for 5-Axis Sculptured Surface Machining. Computer-Aided Design
35:333346.
[74] Gaida WR, Rodriquez CA, Altan T, Altintas Y (1995) Preliminary Experiments
for Adaptive Finish Milling of Die and Mold Surfaces with Ball-nose End Mills.
Transactions of NAMRI/SME XXIII 193198.
[75] Guo G, Ranganath S, McIntosh D, Elzy A (2008) Virtual High Performance
Grinding with CBN Wheels. CIRP Annals 57(1):325328.
[76] Heo EY, Merdol D, Altintas Y (2010) High Speed Pocketing Strategy. CIRP
Journal of Manufacturing Science and Technology 3(1):17.
[77] Imani B, Elbestawi M (2001) Geometric Simulation of Ball-End Milling
Operations. Journal of Manufacturing Science And Engineering 123:177184.
[78] Inasaki I (1996) Grinding Process Simulation Based on the Wheel Topography
Measurement. Annals of the CIRP 45(1):347350.
[79] Inasaki I (1993) Abrasive Machining in the Future. CIRP Annals 42(2):723
732.
[80] Insperger T, Mann BP, Stepan G, Bayly PV (2003) Stability of Up-Milling and
Down-Milling Part 1: Alternative Analytical Methods. International Journal of
Machine Tools and Manufacture 43(1):2534.
[81] ISO 3002, Basic Quantities in Cutting and Grinding, 1982.
[82] Jawahir I, Brinksmeier E, MSaoubi R, Aspinwall D, Outeiro J, Meyer D,
Umbrello D, Jayal A (2011) Surface Integrity in Material Removal Processes:
Recent Advances. CIRP Annals 60(2):603626.
[83] Joliet R, Kansteiner M (2013) A High Resolution Surface Model for the
Simulation of Honing Processes. Advanced Materials Research 769:6976.
[84] Kaymakci M, Kilic ZM, Altintas Y (2012) Unied Cutting Force Model for
Turning, Boring, Drilling and Milling Operations. International Journal of
Machine Tools and Manufacture 54/55:3445.
[85] Kersting P, Biermann D (2014) Modeling Techniques for the Simulating
Workpiece Deections in NC Milling. CIRP Journal of Manufacturing Science
and Technology 7(1):4854.
[86] Kersting P, Biermann D (2012) Modeling Workpiece Dynamics Using Sets of
Decoupled Oscillator Models. Journal of Machining Science and Technology
16(4):564579.
[87] Kersting P, Biermann D (2009) Simulation Concept for Predicting Workpiece
Vibrations in Five-Axis Milling. Machining Science and Technology 13(2):196209.
[88] Lazoglu I, Atabey F, Altintas Y (2002) Dynamics of Boring Processes: Part III Time
Domain Modeling. Journal of Machine Tools and Manufacture 42(14):15671576.
[89] Lazoglu I, Boz Y, Erdim H (2011) Five-Axis Milling Mechanics for Complex
Free From Surfaces. CIRP Annals 60(1):117120.
[90] Lazoglu I, Manav AC, Murtezaoglu Y (2009) Tool Path Optimization for Free
Form Surface Machining. CIRP Annals 58(1):101104.
[91] Lazoglu I (2003) Sculpture Surface Machining: A Generalized Model of BallEnd Milling Force System. International Journal of Machine Tools and Manufacture 43:453462.
[92] Lee S, Ko S (2002) Development of Simulation Systems for Machining Process
Using Enhanced Z Map Model. Journal of Materials Processing Technology
6333:110.
[93] Limido J, Espinsoa C, Salaun M, Lacome J (2007) SPH Method Applied to High
Speed Cutting Modelling. International Journal of Mechanical Sciences
4(7):898908.
[94] Manav C, Bank HS, Lazoglu I (2013) Intelligent Toolpath Selection via MultiCriteria Optimization in Complex Sculptured Surface Milling. Journal of
Intelligent Manufacturing 24:349355.
[95] Merdol D, Altintas Y (2004) Mechanics and Dynamics of Serrated End Mills.
Transactions of ASME Journal of Manufacturing Science and Engineering
126(2):317326.
[96] Merdol D, Altintas Y (2004) Multi Frequency Solution of Chatter Stability for
Low Immersion Milling. Transactions of ASME Journal of Manufacturing Science
and Engineering 126(3):459466.
[97] Merdol D, Altintas Y (2008) Virtual Cutting and Optimization of Three Axis
Milling Processes. International Journal of Machine Tools and Manufacture
48(10):10631071.

Y. Altintas et al. / CIRP Annals - Manufacturing Technology 63 (2014) 585605

[98] Merdol DS, Altintas Y (2008) Virtual Simulation and Optimization of Milling
Operations: Part I Process Simulation. Transactions of ASME Journal of
Manufacturing Science and Engineering 130(5). 051005 (10 pages).
[99] Merdol DS, Altintas Y (2008) Virtual Simulation and Optimization of Milling
Operations: Part II Optimization, Feed Rate Scheduling. Transactions of
ASME Journal of Manufacturing Science and Engineering 130(5). 051004 (12
pages).
[100] Montgomery D, Altintas Y (1991) Mechanism of Cutting Force and Surface
Generation in Dynamic Milling. Transactions of ASME Journal of Engineering for
Industry 113:160168.
[101] Moufki A, Molinari A, Dudzinski D (1998) Modeling of Orthogonal Cutting
with a Temperature Dependent Friction Law. Journal of Mechanics and Physics
of Solids 46(10):21032138.
[102] Ozguven HN (1990) Structural Modications Using Frequency Response
Functions. Mechanical Systems and Signal Processing 4(1):5363.
[103] Ozlu E, Budak E (2007) Analytical Modeling of Chatter Stability in Turning
and Boring Operations. Part I: Model Development. Transactions of ASME
Journal of Manufacturing Science and Technology 129:726732.
[104] Ozturk E, Budak E (2007) Modeling of 5-Axis Milling Process. Journal of
Machining Science and Technology 11(3):287311.
[105] Ozturk E, Tunc LT, Budak E (2011) Analytical Methods for Increased Productivity in 5-Axis Ball-End Milling. International Journal of Mechatronics and
Manufacturing Systems 4(3/4):238265.
[106] Ozturk E, Tunc LT, Budak E (2009) Investigation of Lead and Tilt Angle Effects
in 5-Axis Ball-End Milling Processes. International Journal of Machine Tools
and Manufacture 49(14):10531062.
[107] Pritschow G, Altintas Y, Javone F, Koren Y, Mitsuishi M, Takata T, Van Brussel
H, Weck M, Yamazaki K (2001) Open Controller Architecture Past, Present
and Future. CIRP Annals 50(2):446463.
[108] Requicha A, Rossignac J (1992) Solid Modeling and Beyond. IEEE Computer
Graphics and Applications 12(5):3144.
[109] Roukema JC, Altintas Y (2007) Generalized Modeling of Drilling Vibrations,
Part I: Time Domain Model of Drilling Kinematics, Dynamics and Hole
Formation. International Journal of Machine Tools and Manufacture
47(9):14551473.
[110] Roukema JC, Altintas Y (2007) Generalized Modeling of Drilling Vibrations,
Part II: Chatter Stability in Frequency Domain. International Journal of Machine Tools and Manufacture 47(9):14741485.
[111] Sencer B, Altintas Y, Croft E (2008) Feed Optimization for 5 Axes CNC Machine
Tools with Drive Constraints. International Journal of Machine Tools and
Manufacture 48(78):733745.
[112] Siebrecht T, Biermann D, Ludwig H, Rausch S, Kersting P, Blum H, Rademacher
A (2014) Simulation of Grinding Processes Using Finite Element Analysis and
Geometric Simulation of Individual Grains. Production Engineering Research
and Development 8(3):345353.
[113] Siebrecht T, Rausch S, Kersting P, Biermann D (2014) Grinding Process
Simulation of Free-Formed WC-Co Hard Material Coated Surfaces on Machining Centers Using Poisson-Disk Sampled Dexel Representations. CIRP
Journal of Manufacturing Science and Technology.
[114] Smith S, Tlusty J (1991) An Overview of Modeling and Simulation of the
Milling Process. Transactions of ASME Journal of Engineering for Industry
113(2):169175.
[115] Spence A, Altintas Y (1994) A Solid Modeller Based Milling Process Simulation and Planning System. Transactions of ASME Journal of Engineering for
Industry 116:6169.
[116] Spence A, Altintas Y (1991) CAD Assisted Adaptive Control for Milling.
Transactions of ASME Journal of Dynamic Systems Measurement and Control
113:444450.
[117] Spence A, Altintas Y, Kirkpatrick D (1990) Direct Calculation of Machining
Parameters from a Solid Model. Journal of Computers in Industry 14(4):271280.
[118] Surmann T, Biermann D (2008) The Effect of Tool Vibrations on the Flank
Surface Created by Peripheral Milling. CIRP Annals 57(1):375378.

605

[119] Surmann T, Ungemach E, Zabel A, Joliet R, Schroder A (2011) Simulation of the

Temperature Distribution in NC-Milled Workpieces. Advanced Materials Research 223:222230.
[120] Tekeli A, Budak E (2005) Maximization of Chatter Free Material Removal Rate
in End Milling Using Analytical Methods. Journal of Machining Science and
Technology 9:147167.
[121] Tonshoff HK, Peters J, Inasaki T, Paul T (1992) Modelling and Simulation of
Grinding Processes. CIRP Annals 41(2):677688.
[122] Tonshoff HK, Denkena B, Bo V, Urban B (2002) Automated Finishing of Dies and
Molds. Production Engineering Research and Development in Germany 9/2:14.
[123] Tonshoff HK, Spintig W, Koenig W (1994) Machining of Holes Developments in Drilling Technology. Annals of CIRP 43/2:551561.
[124] Tunc LT, Budak E (2012) Effect of Cutting Conditions and Tool Geometry on
Process Damping in Machining. International Journal of Machine Tools and
Manufacture 57:1019.
[125] Tunc LT, Budak E (2009) Identication and Modeling of Process Damping in
Milling. Transactions of ASME Journal of Manufacturing Science and Engineering
131(5). http://dx.doi.org/10.1115/1.4000170.
[126] Tunc LT, Budak E (2009) Extraction of 5-Axis Milling Conditions from CAM
Data for Process Simulation. International Journal of Advanced Manufacturing
Technology 43(56):538550.
[127] Tuysuz O, Altintas Y, Feng HS (2012) Prediction of Cutting Forces in Three and
Five-Axis Ball-End Milling with Tool Indentation Effect. International Journal
of Machine Tools and Manufacture vol. 66:6681.
[128] Ungemach E, Surmann T, Zabel A (2008) Dynamics and Temperature Simulation in Multi-Axis Milling. Advance Materials Research 43:8996.
[129] Van Hook T (1986) Real-Time Shaded NC Milling Display. ACM SIGGRAPH
Computer Graphics 20(4):1520.
[130] Wang W, Wang K (1986) Geometric Modeling for Swept Volume of Moving
Solids. Computer Graphics and Applications 6(12):817.
[131] Weck M, Altintas Y, Beer C (1994) CAD Assisted Chatter Free NC Tool Path
Generation in Milling. International Journal of Machine Tools and Manufacture
34(6):879891.
[132] Weinert K, Blum H, Jansen T, Rademacher A (2007) Simulation Based Optimization of the NC-Shape Grinding Process with Toroid Grinding Wheels.
Production Engineering Research and Development 1(3):245252.
[133] Weinert K, Du S, Damm P, Stautner M (2004) Swept Volume Generation for
the Simulation of Machining Processes. International Journal of Machine Tools
and Manufacture 44:617628.
[134] Weinert K, Enselmann A, Friedhoff J (1997) Milling Simulation for Process
Optimization in the Field of Die and Mould Manufacturing. CIRP Annals
46(1):325328.
[135] Weinert K, Kersting P, Surmann T, Biermann D (2008) Modeling Regenerative
Workpiece Vibrations in Five-Axis Milling. Production Engineering Research
and Development 2(3):255260.
[136] Xiaoliang J, Altintas Y (2010) Slip-line Field Model of Micro-Cutting Process
with Round Tool Edge Effect. Journal of Materials Processing Technology
211:339355.
[137] Xiaoliang J, Altintas Y (2012) Prediction of Micro-Milling Forces with Finite
Element Method. Journal of Materials Processing Technology 212(3):542552.
[138] Xiaoliang J, Altintas Y (2013) Chatter Stability Model of Micro-Milling With
Process Damping. Transactions of ASME Journal of Manufacturing Science and
Engineering 135(3):031011.
[139] Yeung CH, Altintas Y, Erkorkmaz K (2006) Virtual CNC System Part I: System
Architecture. International Journal of Machine Tools and Manufacture
46(10):11071123. (9 pages).
[140] Yeung CH, Altintas Y, Erkorkmaz K (2006) Virtual CNC System Part II: High
Speed Contouring Application. International Journal of Machine Tools and
Manufacture 46(10):11241138.
[141] Yuen A, Zhang K, Altintas Y (2013) Smooth Trajectory Generation for FiveAxis Machine Tools. International Journal of Machine Tools and Manufacture
Vol. 71:1119.