You are on page 1of 23

Journal of Sound and Vibration 331 (2012) 4668–4690

Contents lists available at SciVerse ScienceDirect

Journal of Sound and Vibration


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

Stability improvement and regenerative chatter suppression in


nonlinear milling process via tunable vibration absorber
Hamed Moradi a, Firooz Bakhtiari-Nejad b,n, Mohammad R. Movahhedy a,
Gholamreza Vossoughi a
a
Centre of Excellence in Design, Robotics and Automation (CEDRA), School of Mechanical Engineering, Sharif University of Technology, Tehran, Iran
b
Department of Mechanical Engineering, Amirkabir University of Technology, Tehran, Iran

a r t i c l e in f o abstract

Article history: In this paper, a tunable vibration absorber set (TVAs) is designed to suppress regenerative
Received 14 October 2011 chatter in milling process (as a semi-active controller). An extended dynamic model of the
Received in revised form peripheral milling with closed form expressions for the nonlinear cutting forces is
17 May 2012
presented. The extension part of the cutting tool is modeled as an Euler–Bernoulli beam
Accepted 20 May 2012
with in plane lateral vibrations (x–y directions). Tunable vibration absorbers in x–y
Handling Editor: L.N. Virgin
Available online 23 June 2012 directions are composed of mass, spring and dashpot elements. In the presence of
regenerative chatter, coupled dynamics of the system (including the beam and x–y
absorbers) is described through nonlinear delay differential equations. Using an optimal
algorithm, optimum values of the absorbers’ position and their springs’ stiffness in both x–y
directions are determined such that the cutting tool vibration is minimized. Results are
compared for both linear and nonlinear models. According to the results obtained, absorber
set acts effectively in chatter suppression over a wide range of chatter frequencies. Stability
limits are obtained and compared with two different approaches: a trial and error based
algorithm and semi-discretization method. It is shown that in the case of self-excited
vibrations, the optimum absorber improves the process stability. Therefore, larger values of
depth of cut and consequently more material removal rate (MRR) can be achieved without
moving to unstable conditions.
& 2012 Elsevier Ltd. All rights reserved.

1. Introduction

Peripheral milling is extensively used in manufacturing processes, especially in aerospace industries where end mills
are used for milling of wing parts and engine components. The generation of complex shapes with high quality for various
types of materials is the main advantage of milling in contrast to other machining processes. During the milling process,
the occurrence of self-excited vibrations or chatter may cause reduction in material removal rate (MRR), damage to the
tool and spindle bearing or may result in poor dimensional accuracy and surface finish of the work-piece.
Regeneration and mode coupling are two well known mechanisms that cause machine tool chatter. In most machining
processes, the regenerative type is found to be the main mode of chatter. It occurs when the oscillation in the chip thickness, at
one pass of the tool, leaves waves on the machined surfaces that are regenerated in the subsequent passes of cut. Under such
conditions, the dynamics of the cutting system is defined through a coupled set of delayed differential equations. Depending on
the phase difference between the inner and outer waves and dynamic gain of the system, the cutting process can be unstable
with exponentially growing oscillation in chip load and consequently large forces and vibration amplitudes [1,2].

n
Corresponding author. Tel.: þ 98 21 64543417; fax: þ98 21 66419736.
E-mail addresses: hamedmoradi@asme.org (H. Moradi), baktiari@aut.ac.ir (F. Bakhtiari-Nejad).

0022-460X/$ - see front matter & 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.jsv.2012.05.032
H. Moradi et al. / Journal of Sound and Vibration 331 (2012) 4668–4690 4669

The recent trend toward high speed and high precision cutting processes requires more accurate modeling; particularly
for the purpose of chatter predictions. Modeling and analysis of the complex geometry and relative tool/work-piece
motion in milling process was carried out. Thereafter, mechanistic approach was extensively used for governing dynamic
equations of various milling processes and the prediction of cutting forces, deflection of machine components and form
errors [3,4]. Alternatively, the mechanics of cutting is used for determination of the milling force coefficients. In this
method, after determination of the flank forces, friction and shear angles from the orthogonal cutting test and
transforming to the oblique cutting conditions; cutting forces are computed [5].
In contrast to the single point metal cutting, a rotating cutter having multiple teeth is used in the milling process. As the
cutter rotates, the directions of the forces vary, and the system is periodic at tooth passing intervals when the cutter has a
uniform pitch (or at spindle intervals for variable tooth spacing). For linear models of the milling process, chatter stability
has been investigated in the frequency and discrete time domains. Analytical solution for the direct prediction of stability
lobes in the frequency domain [6] and prediction of chatter stability in high-speed milling with considering multimode
dynamics have been studied [7].
For direct inclusion of periodically varying system parameters in stability analysis of the milling process, various
methods in discrete time domain have been used. Time finite element analysis (TFEA) to investigate limit cycles and
bifurcation behavior [8]; semi-discretization method (SDM) and full-discretization method (FDM) for determination of
stability lobes diagram [9–13]; and TFEA for simultaneous predictions of stability and surface location error have been
used [14]. In addition, analytical predictions of chatter stability for variable pitch and variable helix tools based on SDM
and TFEA [15]; and stability analysis with TEFA and Chebyshev collocation method based on harmonic balance approach
have been investigated [16]. Also, a comparison between frequency and semi-discrete time solutions of the chatter
stability [17]; and a fuzzy stability analysis for such different domains (frequency and time) have been presented [18].
Recently, nonlinear modeling of the chatter phenomenon has received more attention, because the stability theories
based on linear dynamics cannot predict stability limits in correlation with those of obtained experimentally. Delayed
nonlinear models with square and cubic polynomial terms related to cutting force, structural stiffness and a power-law
function for cutting forces have been considered [19–22]. In other works, visco-elastic and hysteresis effects arisen from
the interaction of cutting force and chip thickness, variable friction, thermo-mechanical effect and intermittent
engagement of the cutting tool have been identified as other sources of nonlinearity in machining processes [23].
Various passive and active control approaches have been used to suppress regenerative chatter. Among the passive
techniques, tunable vibration absorbers and tuned visco-elastic dampers for turning [24,25], boring [26,27] and milling
processes have been used [28,29]. Also, the change and selection of spindle speed for the chatter avoidance have been
investigated [30–33]. For active vibration control, vibration absorbers [34] and model reference adaptive control to achieve
constant cutting forces have been used, e.g., [33,35]. In addition, implementation of active vibration control systems for
chatter suppression [36,37], sliding mode and robust control of regenerative chatter have been proposed [38,39].
However in the majority of previous works, a single degree of freedom model (sdof) has been used for the cutting tool.
During the milling process, for the cases where a relatively long extension part is used, this assumption may not be
sufficiently accurate. This is because the extension link is a continuous system and contains infinite number of vibration
modes. So, when the control procedure is developed to suppress only the dominating mode, the rest of the structural
modes may be excited, resulting in a spill-over problem [26,27,36].
In this paper, a tunable vibration absorber set (TVAs) is designed to suppress regenerative chatter in milling process in
which the extension part is modeled as an Euler–Bernoulli beam (with in-plane lateral vibrations in x–y directions). An
extended dynamic model of the peripheral milling with closed form expressions for the nonlinear cutting forces is
presented. Coupled dynamics of the system, including the beam and x–y absorbers, is described through nonlinear delay
differential equations caused by regenerative chatter. Developing an algorithm based on mode summation method,
optimum values of the absorber position and spring stiffness (in both x–y directions) are determined such that the cutting
tool vibration is minimized. Stability limits are obtained and compared for two different approaches: a trial and error
based algorithm and semi-discretization method (as was used in e.g., [9–12,17]). Improvement in the stability lobes
diagram of the process after implementation of the optimum absorber set is investigated. Results are compared for both
linear and nonlinear models of cutting forces.

2. Dynamics of the peripheral milling process with nonlinear cutting forces

2.1. Linear model of the cutting forces

Fig. 1 shows the geometry of the milling process and the helical milling cutter with damping and stiffness elements in
x–y directions. The dynamics of milling chatter with linear cutting forces was developed by Altintas [40]. For this in-plane
2D modeling, cutting force in the axial direction is assumed to be negligible with respect to other components. The
immersion angle is measured clockwise from the y-axis and the axial (a) and radial (w) depths of cut are constant.
Assuming the bottom end of one flute as the reference immersion angle f, the bottom end points of other flutes are
described at the angles:
fj ð0Þ ¼ f þ jfp ; j ¼ 0,1,. . .,ðN1Þ; fp ¼ 2p=N (1)
4670 H. Moradi et al. / Journal of Sound and Vibration 331 (2012) 4668–4690

Fig. 1. (a) dynamics of the milling process, (b) dynamic chip thickness in regenerative chatter (Source: [17,40]).

where fp is the cutter pitch angle and N is the number of cutter teeth. Considering the lag angle at an arbitrary axial depth
of cut z, the immersion angle of flute j is expressed as [40]:
2z tanb
fj ðzÞ ¼ f þjfp  ; j ¼ 0,1,. . .,ðN1Þ (2)
D
where b and D are helix angle and diameter of the cutter, respectively. Considering the cutting coefficients contributed by
the shearing and edge actions in tangential (Ktc, Kte) and radial (Krc, Kre) directions, acting cutting forces on a differential
flute element with height dz are expressed as:
dF t,j ðfj ,zÞ ¼ ½K tc hj ðfj ðzÞÞ þ K te  dz

dF r,j ðfj ,zÞ ¼ ½K rc hj ðfj ðzÞÞ þK re  dz (3)

where the chip thickness is:


hj ðf,zÞ ¼ cf sinfj ðzÞ (4)

and cf is the feed per tooth per revolution. According Fig. 1, elemental forces in feed (x) and normal (y) directions are
expressed in terms of tangential and radial forces as:
dF x,j ðfj ,zÞ ¼ dF t,j cosfj ðzÞdF r,j sinfj ðzÞ

dF y,j ðfj ,zÞ ¼ þ dF t,j sinfj ðzÞdF r,j cosfj ðzÞ (5)

Substituting Eqs. (3) and (4) in Eq. (5) and integrating analytically along the cut portion of the flute j, yields the closed
form solution for cutting forces (for more details, see [40]). Aligning the flute j¼0 at f ¼ 0 in the beginning of the
algorithm, the cutting forces contributed by all flutes are calculated and integrated digitally to obtain the total
instantaneous forces on the cutter at immersion f as:
X
N 1 X
N 1
F x ðfÞ ¼ F xj ; F y ðfÞ ¼ F yj (6)
j¼0 j¼0

2.2. Nonlinear model of the cutting forces with chatter effect

Since the stability theories based on linear dynamics cannot predict stability limits in correlation with those of obtained
experimentally, nonlinear modeling of the milling process is of great importance. Instead of approximating the cutting
forces as a power-law function of the third degree Taylor series (in terms of chip thickness) [2,20], the cutting forces are
expressed as a complete third-order polynomial function of the chip thickness as:
3 2
dF t,j ðfj ,zÞ ¼ ½x1 hj ðfj ðzÞÞ þ x2 hj ðfj ðzÞÞþ x3 hj ðfj ðzÞÞ þ x4  dz
3 2
dF r,j ðfj ,zÞ ¼ ½Z1 hj ðfj ðzÞÞ þ Z2 hj ðfj ðzÞÞ þ Z3 hj ðfj ðzÞÞ þ Z4  dz (7)

where cutting force coefficients xi, Zi, i ¼1y4 are found directly from experimental force signals [41]. In the presence of
regenerative chatter, the variable total chip thickness is expressed as (instead of Eq. (4)):
hðfj ðzÞÞ ¼ ½cf sinfj ðzÞ þ vj,0 vj gðfj ðzÞÞ (8)
H. Moradi et al. / Journal of Sound and Vibration 331 (2012) 4668–4690 4671

cf sinfj ðzÞ is the static part of the chip thickness caused by the rigid body motion of the cutter and vj,0 vj is the dynamic
part; produced due to vibrations of the tool at the present (vj) and previous (vj,0) tooth periods. g(fj(z)) is a unit step
function determining whether the tooth is in or out of cut and is described in terms of the start (fst) and exit immersion
(fex) angles of the cutter as:
(
1 fst o fj ðzÞ o fex
gðfj ðzÞÞ ¼ (9)
0 fst 4 fj ðzÞ or fex o fj ðzÞ

where fj(z) is given by Eq. (2). Since static part of the chip thickness (cf sinfj ) has no effect on the dynamic chip load
regeneration mechanism, and according to the coordinates shown in Fig. 1, Eq. (8) is rewritten as:
hðfj ðzÞÞ ¼ ½Dx sinfj ðzÞ þ Dy cosfj ðzÞgðfj ðzÞÞ (10)

where
Dx ¼ xðtÞxðttÞ; Dy ¼ yðtÞyðttÞ; t ¼ 2p=ðN OÞ (11)
[x(t),y(t)] and [x(t  t),y(t  t)] represent dynamic displacements of the cutter at the present and previous tooth periods and
t is the delay time of milling process; where O is the spindle angular speed in (rad/s). According to the formulation
discussed in previous section, the total tangential (Ft) and radial (Fr) forces are described through a third order polynomial
function of the chip thickness as:
3 2
F t ¼ x1 h þ x2 h þ x3 hþ x4
3 2
F r ¼ Z1 h þ Z2 h þ Z3 h þ Z4 (12)
Substituting Eqs. (10)–(12) into Eq. (5) results in cutting forces in x–y directions as:
F x ¼ fL1 ½ðDx sin fÞ3 þðDy cos fÞ3  þ L2 ½ðDx sin fÞ2 þ ðDy cos fÞ2  þ L3 ½Dx sin f þ Dy cos f
þ 3L1 Dx2 Dy sin2 f cos f þ 3L1 DxDy2 sin f cos2 f þ 2L2 Dx Dy sin f cos f þ L4 g (13)
n
F y ¼ þ L01 ½ðDx sin fÞ3 þ ðDy cos fÞ3  þ L02 ½ðDx sin fÞ2 þðDy cos fÞ2  þ L03 ½Dx sin f þ Dy cos f

þ 3L01 Dx2 Dy sin2 f cos f þ 3L01 DxDy2 sin f cos2 f þ2L02 DxDy sin f cos f þ L04 g (14)
where
Li ¼ xi cosf þ Zi sinf, L0i ¼ xi sinfZi cosf, i ¼ 1,2,3,4 (15)
Eqs. (13) and (14) can be expressed in the matrix form as:
( ) ( )
Fx Dx, Dy, Dx2 , Dy2 , Dx3 , Dy3
¼ ½HðfÞ (16)
Fy Dx, Dy, Dx2 , Dy2 , Dx3 , Dy3

Since angular position varies with time (fj(t) ¼ O t), Eq. (16) is rewritten as:
~
fFðtÞg ¼ ½HðtÞf Dtg (17)
~
Similar to the milling forces, since ½HðtÞ is periodic at tooth passing frequency, it can be expanded into Fourier series as:
X1 Z t
~
½HðtÞ ¼ ~ r eirNOt ; ½H
½H ~ r  ¼ ð1=tÞ ~
½HðtÞeirNOt
dt (18)
r ¼ 1 0

The immersion conditions and the number of teeth in cut determines the number of harmonics (r) of the tooth passing
~
frequency (NO) that must be considered for an accurate reconstruction of ½HðtÞ. Since in this paper the immersion cut is
not low, the average components of the Fourier series expansion (r ¼0) is sufficient [17]. In this case:
Z t
~ 0  ¼ ð1=tÞ
½H ~
½HðtÞ dt (19)
0

~ 0  is valid for gj(fj) ¼1, it equals the average value of ½H


Since ½H ~ ðtÞ at cutter pitch angle as:
Z fex
~ 0  ¼ ð1=f Þ
½H ~ fÞ df
½Hð (20)
p
fst

Since the average cutting force per tooth period is independent of the helix angle, ½H ~ 0  is valid for helical end mills as
well as those with zero helix angle [17]. Considering Eq. (16) and carrying out the integral of Eq. (20) on Fx and Fy given by
Eqs. (13) and (14), yields:
N  
Fx ¼  a1 Dx3 þ b1 Dy3 þ a2 Dx2 þ b2 Dy2 þ a3 Dx þ b3 Dyþ 3g1 Dx2 Dy þ3g2 DxDy2 þ 2g3 DxDy þ g4
2p
N  0 
Fy ¼ þ a Dx3 þ b01 Dy3 þ a02 Dx2 þ b02 Dy2 þ a03 Dx þ b03 Dy þ 3g01 Dx2 Dyþ 3g02 DxDy2 þ2g03 DxDyþ g04 (21)
2p 1
4672 H. Moradi et al. / Journal of Sound and Vibration 331 (2012) 4668–4690

where for the half immersion up-milling (as the case study considered in this paper) with fst ¼0 and fex ¼ p/2, coefficients
of Eq. (21) are determined as:
         
a1 ¼ 14 x1 þ 34 pZ1 b1 ¼ 14 Z1 þ 34 px1 a2 ¼ 13 x2 þ2Z2 b2 ¼ 13 Z2 þ 2x2 a3 ¼ 12 x3 þ 12 pZ3
       
b3 ¼ 12 Z3 þ 12 px3 g1 ¼ 14 Z1 þ 14 px1 g2 ¼ 14 x1 þ 14 pZ1 g3 ¼ 13 Z2 þ x2 g4 ¼ x4 þ Z4
  0         (22)
1 3 1
a1 ¼ 4 Z1 þ 4 px1 b1 ¼ 4 x1  4 pZ1
0 3 1
a2 ¼ 3 Z2 þ2x2
0
b02 ¼ 13 x2 2Z2 a03 ¼ 12 Z3 þ 12 px3
0 1
 1
 1
 1
 1
 1
 1
 
b3 ¼ 2 x3  2 pZ3 g1 ¼ 4 x1  4 pZ1 g2 ¼ 4 Z1 þ 4 px1
0 0
g3 ¼ 3 x2 Z2
0
g4 ¼ Z4 þ x4
0

It should be mentioned that the procedure explained for the nonlinear force modeling can be similarly extended to any
other machining condition in which the cutting force signals are measured experimentally.

3. Formulation of the problem with tunable vibration absorber (TVA)

In this section, the formulation of tunable vibration absorbers (TVAs) for the milling process with extension part of
cutting tool (modeled as an Euler–Bernoulli beam) is explained. Extension parts may be essentially used in cases where the
access to the work-piece is difficult. Since the relatively long extension part makes it very susceptible to chatter and the
requirement on surface quality is usually stringent, chatter suppression is particularly critical in this process.
In addition, the process of absorber design relies on the fact that the model parameters such as mass, stiffness and
damping are accurate. But, both the nature of milling process as a dynamic system and the existence of movable elements
of machine tool and work-piece make it difficult to obtain accurate results from a sdof system. Structures made of several
beams are common in engineering fields. Each beam constitutes of an infinite number of degrees of freedom, and the mode
summation method is used to model it as a system consisting of a finite number of more significant degrees of freedom.
For general forced vibration of a one-dimensional structure, a force per unit length f(z,t) at any location of z is
considered. If normal modes of the structure ci(z) are known, its deflection at any point z can be expressed as:
X
Wðz,tÞ ¼ si ðtÞci ðzÞ (23)
i

where the generalized modal coordinate si(t) must satisfy the following equation [42]:
Z
1
s€ i ðtÞ þ 2zi oi s_ i ðtÞ þ o2i si ðtÞ ¼ f ðz,tÞci ðzÞ dz (24)
mi
and the generalized modal mass mi is defined as
Z
mi ¼ r0 c2i ðzÞ dz (25)
0
The damping ratio zi is a small positive number with common values of zi r0.05 for slightly damped structures and r is
the mass per unit length of the beam. For a cantilever Euler–Bernoulli beam, the eigenvector at any mode is given as [42]:
  coshli þ cosli
ci ðzÞ ¼ C i coshli zcosli zmi ðsinhli zsinli zÞ ; mi ¼ (26)
sinhli þ sinli
and the natural frequencies of a cantilever beam are defined as
oi ¼ l2i ðEI=r0 l4 Þ1=2 (27)

where EI is flexural rigidity, l is the beam length. After applying the boundary conditions for a cantilever beam, li’s can be
determined from the following characteristic equation:
cosli l coshli l þ 1 ¼ 0
li are constants taking the values of l1 ¼1.87, l2 ¼4.69, l3 ¼7.85R
for the first three natural frequencies. Normal modes ci(z)
are orthogonal functions that satisfy the orthogonally condition ( (Ci ci(z))2 dz¼1); which yields Ci ¼1. If instead of distributed
load f(z,t), a concentrated force F(z0,t) is applied at some point z¼z0 along the beam length, Eq. (24) can be written as
1
s€ i ðtÞ þ 2zi oi s_ i ðtÞ þ o2i si ðtÞ ¼ Fðz0 ,tÞci ðz0 Þ (28)
mi
For the in-plane lateral vibrations of the extension part (as a two-dimensional structure), the formulation presented
above is extended. Instead of the deflection W(z,t) given by Eq. (23), two deflections x(z,t) and y(z,t) in x–y directions exist
that are described as (while z is the coordinate along the beam length):
8 P
< xðz,tÞ ¼ pi ðtÞci ðzÞ
>
i
P (29)
>
: yðz,tÞ ¼ qi ðtÞci ðzÞ
i

where pi(t) and qi(t) are the generalized coordinates for 2D structure (instead of one coordinate si(t) for one-dimensional
problem). Therefore, Eq. (28) is extended for 2D problem in terms of coordinates pi(t) and qi(t) as follows (where
H. Moradi et al. / Journal of Sound and Vibration 331 (2012) 4668–4690 4673

concentrated force F(z0,t) in Eq. (28) is replaced with the related cutting forces):
8
< p€ i ðtÞ þ 2zxi oxi p_ i ðtÞ þ o2xi pi ðtÞ ¼ M
1 ~ c ðdÞ
F x ðd,tÞ ~
i
2 1 ~ ~ (30)
: q€ ðtÞ þ2z o q_ ðtÞ þ o q ðtÞ ¼ F y ðd,tÞc ðdÞ
i yi yi i yi i M i

where M¼mi is the mass of extension part, (zx, zy) and (ox, oy) are damping ratio and natural frequencies of the milling
~
cutter in x–y directions.F x ðd,tÞ ~
and F y ðd,tÞ are the cutting forces in x–y directions that in the presence of regenerative
chatter are given by Eq. (21). It should be mentioned that according to the formulation presented for the cutting forces in
previous section, Fx and Fy are distributed along the cutting portion (along the axial depth of cut, a). However, for simplicity
of the formulation, it is assumed that these forces are exerted concentrative at the middle of the axial depth of cut, i.e.,
d~ ¼ la=2 in Eq. (30), where l is the length of extension part including the cutting portion. The accuracy of such
assumption, using concentrated forces instead of distributed ones, is shown in results section. In the presence of
regenerative chatter, delayed terms Dx and Dy in cutting forces of Eq. (21) and at the z ¼ d~ are described as:
X
~
Dx9d~ ¼ xðtÞxðttÞ9d~ ¼ ci ðdÞ ½pi ðtÞpi ðttÞ
i
X
~
Dy9d~ ¼ yðtÞyðttÞ9d~ ¼ ci ðdÞ ½qi ðtÞqi ðttÞ (31)
i

For the case of cutting forces with cubic nonlinearity, i.e., x2 ¼ x4 ¼ Z2 ¼ Z4 ¼0 in Eq. (12) and consequently a2 ¼
a02 ¼ b2 ¼ b02 ¼ g3 ¼ g03 ¼ g4 ¼ g04 ¼ 0 in Eq. (22), substituting Eqs. (21) and (31) in Eq. (30) and doing some simplifications yields:
8
~ P
>
> p€ ðtÞ þ2zxi oxi p_ i ðtÞ þ o2xi pi ðtÞ þ N2cpiM
ðdÞ
Lx ðDp~ j , Dq~ j , Dr~ j , Ds~ j Þ ¼ 0
>
< i j
~ P
(32)
>
> 2 N ci ðdÞ
: q€ i ðtÞ þ2zyi oyi q_ i ðtÞ þ oyi qi ðtÞ 2pM
> Ly ðDp~ j , Dq~ j , Dr~ j , Ds~ j Þ ¼ 0
j

where

Lx ¼ a1 Dp~ 3j þ b1 Dq~ 3j þ a3 Dp~ j þ b3 Dq~ j þ3g1 Dr~ j þ3g2 Ds~ j

Ly ¼ a01 Dp~ 3j þ b01 Dq~ 3j þ a03 Dp~ j þ b03 Dq~ j þ3g01 Dr~ j þ3g02 Ds~ j (33)

~ Dp ;
Dp~ j ¼ cj ðdÞ: ~ Dp2 :Dq
Dr~ j ¼ c3j ðdÞ:
j j j

~ Dq ;
Dq~ j ¼ cj ðdÞ: ~ Dq2 :Dp
Ds~ j ¼ c3j ðdÞ: (34)
j j j

and
Dpj ¼ pj ðtÞpj ðttÞ
Dqj ¼ qj ðtÞqj ðttÞ (35)

3.1. Application of the vibration absorber for the cutting tool

As shown in Fig. 2, extension part of the cutting tool is modeled as a cantilever beam which has in-plane lateral
vibrations. To suppress chatter vibrations, two tunable vibration absorbers (consisting of mass, spring and damper) are
implemented in x–y directions. In practice, a shell-type coverage can be provided around the extension part of the cutting
tool. The spindle shaft is embedded into this shell-type coverage with a similar mechanism which is used for the tool-
holder part. For this purpose, the lateral undesirable cutting forces exerted on the extension part are transmitted to the
absorber through a mechanism consisting of an inner bushing, a journal bearing and a fixed internal shell, respectively
(Fig. 2). The absorber consisting of a mass mx, two springs with stiffness of kx and two dampers of cx slides through a
groove created on the internal shell coverage. On the other hand, at the end of the absorber, rollers are used that allows the
absorber moves along an external shell coverage; which itself is fixed to the body of machine. It should be mentioned that
during regenerative chatter with high vibration amplitudes, undesirable forces on the cutting tool can be transmitted to
the absorbers through the mentioned mechanism. Similarly, this configuration is used for the absorber in y-direction.
To make the modeling more general, it is assumed that spring and damper elements of absorbers are located at (z¼ ax,
z¼ ay) and (z¼bx, z ¼by) from the clamped support, respectively. However, in simulation results and as shown in Fig. 2, the
distance between spring and damper elements is considered zero (as in many practical cases). So, the distance between the
absorbers and the clamped support (ax, ay) are selected as design parameters. As discussed, cutting forces in x–y directions
are exerted at z ¼ d~ ¼ la=2 where l is the extension length and a is the axial depth of cut. Denoting the x–y absorbers
4674 H. Moradi et al. / Journal of Sound and Vibration 331 (2012) 4668–4690

Fig. 2. Schematic of the extension part of the milling process with attached tunable vibration absorbers in x–y directions.

displacements as u(t) and w(t), Eq. (32) is modified to (in the presence of absorbers):
8 ~ P
>
> p€ i ðtÞ þ 2zxi oxi p_ i ðtÞ þ o2xi pi ðtÞ þ N2cpiMðdÞ
~ ~ ~ ~ 1 _ _
>
> j Lx ðDp j , Dq j , Dr j , Ds j Þ ¼ M fkx ½xðax ,tÞuðtÞci ðax Þc x ½xðbx ,tÞuðtÞci ðbx Þg
>
> P
<€ 2
~
Nci ðdÞ 1
q i ðtÞ þ2zyi oyi q_ i ðtÞ þ oyi qi ðtÞ 2pM ~ ~ ~ ~ _ _
j Ly ðDp j , Dq j , Dr j , Ds j Þ ¼ M fky ½yðay ,tÞwðtÞci ðay Þcy ½yðby ,tÞwðtÞci ðby Þg (36)
>
> mx uðtÞ€ _
uðtÞ _
xðb ,tÞ ,tÞ 0
>
> þc x ½ x þk x ½uðtÞxða x ¼
>
>
: my wðtÞ€ þ cy ½wðtÞ_ _ y ,tÞ þ ky ½wðtÞyðay ,tÞ ¼ 0
yðb

expanding
8 P
< xðax ,tÞ ¼ pi ðtÞci ðax Þ
>
i
P (37)
>
: yðay ,tÞ ¼ qi ðtÞci ðay Þ
i

Eq. (36) is rewritten as:


8 ~ P
>
> p€ i ðtÞ þ 2zxi oxi p_ i ðtÞ þ oxi 2 pi ðtÞ þ N2cpiM
ðdÞ
Lx ðDp~ j , Dq~ j , Dr~ j , Ds~ j Þ
>
>
>
> j
>
> ( " # " # )
>
>
>
> P P
>
> 1
¼ M kx pl ðtÞcl ðax ÞuðtÞ ci ðax Þcx p_ l ðtÞcl ðbx ÞuðtÞ _ ci ðbx Þ
>
>
>
> l i
>
>
>
> ~ P
N c ðdÞ
> 2
< q€ i ðtÞ þ 2zyi oyi q_ i ðtÞ þ oyi qi ðtÞ 2piM Ly ðDp~ j , Dq~ j , Dr~ j , Ds~ j Þ
j
( " # " # ) (38)
>
>
>
> P P
>
> ¼ 1
ky ql ðtÞcl ðay ÞwðtÞ ci ðay Þcy q_ l ðtÞcl ðby ÞwðtÞ
_ ci ðby Þ
>
> M
>
> l l
>
> P P
>
> € þ d~ x ½uðtÞ
_
>
> uðtÞ p_ l ðtÞcl ðbx Þþ dx ½uðtÞ pl ðtÞcl ðax Þ ¼ 0
>
> i l
>
> P P
>
>
>
: € ðtÞ þ d~ y ½w
w _ ðtÞ q_ l ðtÞ cl ðby Þ þ dy ½w ðtÞ ql ðtÞ cl ðay Þ ¼ 0
l l

where

dx ¼ kx =mx , d~ x ¼ cx =mx , dy ¼ ky =my , d~ y ¼ cy =my (39)


H. Moradi et al. / Journal of Sound and Vibration 331 (2012) 4668–4690 4675

Under regenerative chatter conditions (with the chatter frequency of oc), normal modes of the beam and absorbers’
displacement can be written in the exponential form as:
pi ¼ pi eJoc t , qi ¼ qi eJoc t , u ¼ ueJoc t , w ¼ weJoc t (40)
where J denotes the imaginary unit. Substituting Eq. (40) in Eq. (38) yields coupled dynamics of the system as:
8
~ P P P
>
> pi ðo2xi o2c þ 2zxi oxi oc JÞ þ N2cpiM
ðdÞ
Lx ðDp~ j , Dq~ j , Dr~ j , Ds~ j Þ ¼ M1 fkx ½ pl cl ðax Þuci ðax ÞJcx oc ½ pl cl ðbx Þuci ðbx Þg
>
>
>
> j l i
>
>
>
> Nci ðdÞ~ P P P
>
> q ðo 2
o 2
þ 2z o o JÞ L ð Dp~ , D ~
q , Dr~ , Ds~ Þ ¼ 1
fk ½ q c ða Þw c ða ÞJc o ½ ql cl ðby Þwci ðby Þg
< i yi c yi yi c 2pM y j j j j M y l l y i y y c
j l l
P P (41)
>
> uðdx o2c þJd~ x oc ÞJd~ x oc pl cl ðbx Þdx pl cl ðax Þ ¼ 0
>
>
>
> l l
>
> P P
>
> 2 ~ ~
>
>
:
wð dy o c þ Jd o
y c ÞJ d o
y c q c ðb
l l y Þd y ql cl ðay Þ ¼ 0
l l

and as given by Eq. (33), the terms Dpj ¼ pj pnj and Dqj ¼ qj qnj exist in functions Lx and Ly . pj or qj (and similarly those
with subscripts i and l) are related to the values at time t while pnj and qnj are related to the normal modes at time t  t. The
coupled system of Eq. (41) is the vibration equations for the combined dynamic system (two equation set with index
i, j, l ¼1,2,y for the beam lateral vibrations in x–y directions and two equations for the vibration absorbers). In the presence
of regenerative chatter, this set of equations can be solved numerically via SIMULINK Toolbox of MATLAB. Under regenerative
chatter condition, and using Eq. (33) through (35), Eq. (41) is written in the general matrix form as [E(t,t)]{W (t, t t)}¼0. For
s number of modes, [E(t,t)] is a square matrix of order (4s þ2) which is constituted numerically and
fWðt,ttÞg ¼ fp1 ðtÞ. . .ps ðtÞ q1 ðtÞ. . .qs ðtÞ p1 ðttÞ. . .ps ðttÞ q1 ðttÞ. . .qs ðttÞ uðtÞ wðtÞgT
In the absence of regenerative chatter (Dpj ¼ Dqj ¼ 0); the rank of [E(t)] is (2s þ2) because 2s delay elements including
pi n and qi n are vanished from [E(t,t)] and {W (t,t  t)}. Under such conditions without time delay terms, if the determinant
of [E(t)] is set to zero, the (2s þ2) natural frequencies of the whole system including the beam and absorber are obtained.
To achieve the generalized coordinates pi , qi , the amplitude of the absorbers mass displacement u and w, Eq. (41) which
includes (4s þ2) coupled equations must be solved simultaneously.

4. Stability analysis of the linear and nonlinear absorbed systems using semi-discretization method (SDM)

As discussed in the literature, in discrete time domain, various methods such as time finite element analysis (TFEA) and
Chebyshev collocation method [8,14–16]; semi-discretization method (SDM) and full-discretization method (FDM) have
been used for stability analysis of the milling process [9–13]. Also, a comparison between frequency and semi-discrete
time solutions of the chatter stability has been presented [17].
In this section, semi-discretization method (SDM) is used for stability analysis of the absorbed milling system described
by Eq. (38). First, this formulation is accomplished for the linear model and then is modified for the nonlinear model of
cutting forces. SDM technique is often used in computational fluid mechanics for solving partial differential equations
(PDEs). In this approach, the PDE is discretized along the spatial coordinates while the time coordinate are kept unchanged.
Similarly, this method was applied for the analysis of delay differential equations (DDEs); where the time delayed terms
are approximated by a piecewise constant function while the current time terms are kept unchanged [10,11]. Therefore,
DDE is approximated by a series of ordinary differential equations (ODEs). Following SDM as used in [9–11,17], and
defining the state vector as
_ Tð4s þ 4Þ1
q ¼ ½p1 . . .ps . . .q1 . . .qs u w p_ 1 . . .p_ s q_ 1 . . .q_ s u_ w (42)

where s is the number of modes used in mode summation approach, Eq. (38) is represented in the state-space
0
configuration as (without considering nonlinear terms; i.e., in Eq. (33), Lx ¼ a3 Dp~ j þ b3 Dq~ j and Ly ¼ a03 Dp~ j þ b3 Dq~ j ):
fq_ ðtÞg ¼ ½AðtÞfqðtÞg þ½BðtÞfqðttÞg (43)
where [A(t)] and [B(t)] are square matrices of order 4s þ4 and expressed as:
" # " #
½0 ½I ½0 ½0
½AðtÞ ¼ ; ½BðtÞ ¼ (44)
½R½x2n  ½S½2fn xn  ½R ½0

where each matrix component in [A(t)] and [B(t)] is a square matrix of order 2s þ2; [I] is the identity matrix; ½x2n  is a
diagonal matrix with components of o2x1 . . .o2xs , o2y1 . . .o2ys , dx , dy and [2fnxn] is a diagonal matrix with components of
2zx ox1 . . .2zx oxs ,2zy oy1 . . .2zy oys , d~ x , d~ y (as given by Eqs. (27) and (39)) and:
2 3
½R11 ss ½R12 ss ½R13 s1 ½0s1
6 ða0 =b Þ½R  ½R22 ss ½0s1 ½R24 s1 7
6 3 3 12 ss 7
½R  ¼ 6 6 ðdx M=kx Þ½R13 T
7;
7
4 ½01 s 0 0 5
T
½01s ðdy M=ky Þ½R24  0 0
4676 H. Moradi et al. / Journal of Sound and Vibration 331 (2012) 4668–4690

2 3
½S11 ss ½0ss ½S13 s1 ½0s1
6 ½0ss ½S22 ss ½0s1 ½S24 s1 7
6 7
½S ¼ 6
6 ðd~ x M=cx Þ½S13 T
7
7 (45)
4 ½01s 0 0 5
½01s ðd~ y M=cy Þ½S24 T 0 0

where
2 2
3
 kMx c1 ðax ÞN ~
^ a3 c2 ðdÞ ... ... ~ c ðdÞ
^ a3 c ðdÞ
 kMx c1 ðax Þcs ðax ÞN ~
1 1 s
6 7
6 2 ^ a3 c2 ðdÞ
 kMx c2 ðax ÞN ~ ... ... 7
6 2 7
½R11  ¼ 6 7
6
4 & ... 7
5
2 ~
^ a3 c2 ðdÞ
sym  kMx cs ðax ÞN s

2 k k
3
2
 My c1 ðay ÞN ~
^ b0 c2 ðdÞ ... ... ~ c ðdÞ
^ b0 c ðdÞ
 My c1 ðay Þcs ðay ÞN ~
3 1 3 1 s
6 7
6 k 2 ~
^ b c ðdÞ
 My c2 ðay ÞN
0 2
... ... 7
6 3 2 7
½R22  ¼ 6 7
6
4 & ... 7
5
k
sym  My ~
cs ðay ÞN^ b03 c2s ðdÞ
2

2 3
~
c21 ðdÞ ... ... ~ c ðdÞ
c1 ðdÞ s
~
2 3 2 3
6 7   c1 ðax Þ   c1 ðay Þ
6 ~
c2 ðdÞ ... ... 7 kx k
6 7 6 ^ 7 y 6 7
4 ^ 5
^ 2
½R12  ¼ N b3 6 7; ½R13  ¼ 4 5; ½R24  ¼ (46)
6
4 & ... 7
5
M M
cs ðax Þ cs ðay Þ
sym ~
c2 ðdÞ s

and
2 3
c21 ðbx Þ ... ... c1 ðbx Þcs ðbx Þ 2 3
6 7 c1 ðbx Þ
 c 6 c22 ðbx Þ ... ... 7 c
x 6 7 6 7
½S11  ¼  6 7; ½S13  ¼
x
4 ^ 5
M 6
4 & ... 7
5
M
cs ðbx Þ
sym c2s ðbx Þ
2 3
c21 ðby Þ ... ... c1 ðby Þcs ðby Þ 2 3
6 7 c1 ðby Þ
c 6 c22 ðby Þ ... ... 7 c
y 6 7 6 7
½S22  ¼  6 7; ½S24  ¼
y
4 ^ 5 (47)
M 6
4 & ... 7
5
M
cs ðby Þ
sym c2s ðby Þ

N^ ¼ ðN=2pMÞ and (X)T stands for the transpose of the quantity (X). Arbitrary current time tn and delay time t are divided
into the n and i number of discrete time intervals Dt as tn ¼ nDt and t ¼iDt. For simplicity, the value of {q(tn)} at the current
time tn is expressed as {qn}and at the time tn  t as {q(tn  t)} ¼{q[(n  i) Dt]} ¼{qn  i}. For small enough sampling interval
time Dt, the value of {q(tn  t)}can be approximated by averaging the values of two successive intervals as [9,17]:

 


  Dt ½qðt n tÞ þ qðt n t þ DtÞ fqni g þ fqni þ 1 g
qðtn tÞ ffi q tn t þ ffi ffi (48)
2 2 2
Consequently, Eq. (43) is described at discrete time intervals as:
    1    
q_ n ðtÞ ¼ ½An  qn ðtÞ þ ½Bn ð qni þ qni þ 1 Þ (49)
2
At the small time intervals Dt, the solution of Eq. (49) including the homogeneous ({qhn(t)}) and particular ({qpn(t)})
parts is obtained as:
1
qn ðtÞ ¼ expð½An ðttn ÞÞfD0 g  ½An 1 ½Bn ðfqni g þ fqni þ 1 gÞ (50)
|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl} 2
|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
fqhn ðtÞg
fqpn ðtÞg

where the initial conditions determine{D0}. Solving Eq. (50) at the discrete time t ¼tn, yields {D0} as:
  1    
fD0 g ¼ qn þ ½An 1 ½Bn ð qni þ qni þ 1 Þ (51)
2
Due to the validity of the solution at the time intervals Dt¼tn þ 1–tn, according to Eqs. (50) and (51), the states {q(tn þ 1)}
are expressed as:
    1    
qn þ 1 ¼ expð½An DtÞ qn þ ðexpð½An DtÞ½IÞ½An 1 ½Bn ð qni þ qni þ 1 Þ (52)
2
H. Moradi et al. / Journal of Sound and Vibration 331 (2012) 4668–4690 4677

Therefore, the series of equations are expressed at discrete time intervals as:
fvn þ 1 g ¼ ½Cn fvn g (53)

where {vn} is a [(2s þ2)  (iþ1)]  1 vector and [Cn] is a square matrix of order (2s þ2)  (iþ 1) as:
h iT
fvn g ¼ fqn g fqn1 g fqn2 g    fqni þ 1 g fqni g (54)

and
2 1 1
3
expð½An DtÞ ½0 ... ½0 ½  ½ Þ½An 1
2 ðexpð An DtÞ I
½  ½ Þ½An 1 ½Bn 
2 ðexpð An DtÞ I
6 7
6 ½I ½0 ... ½0 ½0 ½0 7
6 7
6 ½0 ½I ... ½0 ½0 ½0 7
6 7
½Cn  ¼ 6 7 (55)
6 ^ ^ & ^ ^ ^ 7
6 7
6 ½0 ½0 ... ½I ½0 ½0 7
4 5
½0 ½0 ... ½0 ½I ½0

As it is seen in SDM, the original DDE of Eq. (43) is approximated by a series of ODE’s while the terms {q(t)} and fq_ ðtÞg
are unchanged. The delayed term {q(t  t)} is approximated by a linear function of time and the periodic matrix [B(t)] is
approximated by a piecewise constant function [9,10]. Therefore, at the time intervals Dt, the time varying milling process
can be simulated by solving the discrete set of recursive Eq. (52). Previous terms including {qn},{qn  i} and {qn  i þ 1} are
required for this solution. It is sufficient to solve the equations at i number of time intervals, because the process is periodic
at tooth pass intervals t. Therefore, the modified Eq. (53) is expressed at i number of intervals within the tooth period t as:
fvn þ i g ¼ ½Ci . . .½C2 ½C1 fvn g ¼ ½Ufvn g (56)

Stability of the process is directly investigated through the monodromy operator ½U. According to the Floquet theory,
dynamics of Eq. (43) is stable if any of the eigenvalues of monodromy operator ½U (its characteristic multipliers) has a
modulus less than one [43].
To investigate the stability of the nonlinear model, Eq. (38) is linearized through linearization of the nonlinear terms
Lx ðD p~ j , D q~ j , D r~ j , D s~ j Þ and Ly ðD p~ j , D q~ j , D r~ j , D s~ j Þ of Eq. (33). For this purpose, these terms are linearized around the
operating point ðD _ pj, D_ q j Þ; given by Eq. (35). Accordingly, the formulation is exactly the same as that presented for the
linear model, but the matrix [A(t)] of Eq. (44) is modified to [A*(t)] as (for the nonlinear model linearized
aroundðD _ pj, D_ q j Þ):
~
½An ðtÞ ¼ ½AðtÞ þ ½AðtÞ (57)
where the matrix [A(t)] is found for the linear model as given by Eq. (44) and;
" #
½0 ½0
~
½AðtÞ ¼ ~ (58)
½R ½0

2 3
½R~ 11 ss ½R~ 12 ss ½0s1 ½0s1
6 ~ 7
6 ½R  ½R~ 22  ½0s1 ½0s1 7
~ ¼ 6 21 ss
½R
ss 7 (59)
6 ½0 ½01s 0 0 7
4 1s 5
½01s ½01s 0 0

where
2 3
c41 ðdÞ 1 p 1 , D_
~ Y ðD_ q1Þ ... ... c1 ðdÞ s p s , D_
~ Ys ðD_
~ c3 ðdÞ qsÞ
6 7
6 c2 ðdÞY2 ðD_p 2 , D_ 7
4 ~
^6 ^ q2Þ ... ... 7
½R~ 11  ¼ 3N 6 7
6
4 ^ ... & ... 7
5
~ c ðdÞ
c31 ðdÞ s 1 p 1 , D_
~ Y ðD_ q1Þ ... ^ c4s ðdÞ p s , D_
~ Ys ðD_ qsÞ
2 3
c41 ðdÞ 1 p 1 , D_
~ Y ðD_ q1Þ ... ... c1 ðdÞ s p s , D_
~ Ys ðD_
~ c3 ðdÞ qsÞ
6 7
~ ^
6
6 ^ ~ Y ðD_
c42 ðdÞ 2 p 2 , D_
q2Þ ... ... 7
7
½R 12  ¼ 3N 6 7 (60)
6
4 ^ ... & ... 7
5
p 1 , D_
~ Y ðD_
~ c ðdÞ cs ðdÞYs ðD_p s , D_
4 ~
c31 ðdÞ s 1 q1Þ ... ^ qsÞ

p g , D_
and Yg ðD_ q g Þ ¼ a1 D_ p g D_
p g þ2g1 D_ q g þ g2 D_
2 2
qg ; g ¼ 1,2,. . .s

p g , D_
Yg ðD_ q g Þ ¼ g1 D_ p g D_
p g þ 2g2 D_ q g þ b1 D_
2 2
qg ; g ¼ 1,2,. . .s (61)

½R~ 21  equals to ½R~ 11  while the coefficients a1, g1 and g2 in function Yg of Eq. (60) are replaced with a01 , g01 and g02 .
Similarly, ½R~ 22  equals to ½R~ 12  while the coefficients b1, g1 and g2 in function Yg of Eq. (61) are replaced with b1 , g01 and g02
0
4678 H. Moradi et al. / Journal of Sound and Vibration 331 (2012) 4668–4690

(these coefficients are given by Eq. (22)). In the proposed SDM and during the progress of algorithm, average values of the
_
delayed terms at the operating point ðD _ p j , D q j Þ and current time tn are computed as:

1n o 1n o
D_
p j 9n ¼ Dpj 9n þ Dpj 9n1 ; D_
q j 9n ¼ Dqj 9n þ Dqj 9n1 ; j ¼ 1,2. . .s (62)
2 2

The stability analysis of the nonlinear model is the same as that of discussed for the linear model. But in all formulation
presented for the linear model, the modified coefficient matrix [A*(t)] (given by Eq. (57)) must be used instead of [A(t)]
(given by Eq. (44)).

5. Simulation, results and discussion

For simulation of the problem, realistic parameters of the extension part, modeled as a cantilever Euler–Bernoulli beam,
is given in Table 1.1. Also, related parameters for the cutting forces with cubic nonlinearity [41], and initial values of
absorbers parameters are given in Tables 1.1 and 1.2. Here, as a case study, half immersion up-milling with fst ¼0 and
fex ¼ p/2, four teeth N ¼4, spindle speed O ¼4000 rev/min, helix angle b ¼ 30, radial width of cut wc ¼ 2.5 mm, feed rate
cf ¼0.2 mm/rev-tooth and three number of modes s ¼3 are considered.
It should be mentioned that the similar simulation results can be presented for any other machining condition; while
cutting forces may be considered as a complete third order polynomial function of the chip thickness. Also, absorber
parameters given in Table 1.1 are the initial values selected to start the simulation. For this input data, the first three
natural frequencies of the system without absorber are given in Table 2. Since the third natural frequency is so high, it is
too far from the common values of the chatter frequency and tool passing frequency in machining. Therefore, for
simulation results, using two number of lower modes is sufficient (s ¼2).
However, it should be mentioned that for relatively small extension parts, only its first natural frequency is in order of
chatter frequency (second and higher natural frequencies are too far from the chatter frequency). In that case and as it was
observed in the majority of previous works, using a lumped sdof model (s ¼1) leads to the acceptable results.
Under the conditions mentioned above, the milling process (Eq. (38)) is simulated in the time domain via SIMULINK Toolbox
of MATLAB (all over the paper, initial functional conditions for two first modal coordinates are considered as
(p1(0)¼q1(0)¼0.1 mm) and (p2(0)¼q2(0)¼ 0). The critical depth cut at which the system changes from the stable condition
to the unstable one is found as alim ¼ 4:83 mm. For a ¼ 4:5 mm o alim , time response of the tool tip for the cases with
distributed cutting forces along the cut portion and with concentrated cutting forces at the middle of the axial depth of cut, i.e.,
d~ ¼ la=2 is shown in Fig. 3. As it is shown, there is a good accordance between them. Therefore, as discussed in Section 3, and
for simplicity analysis of the problem, it is assumed that the cutting forces are exerted concentrative at d~ ¼ la=2.

Table 1.1
Realistic parameters of the system for cutting conditions, extension part and absorbers.

Cutting conditions Extension part Absorbers


0
Start & exit immersion fst ¼ 0 Mass per unit length & mass r ¼1.44 kg/m Spring stiffness (initial kx ¼20 kN/m
angles density value)
fex ¼ p=2 r^ ¼ 8500 kg=m3 ky ¼ 40 kN/m
Cutter teeth number N¼4 Young modulus E ¼210 GN/m2 Mass mx ¼ 0:1 kg
my ¼ 0:1 kg

Spindle speed O ¼ 4000 rev/min Length l ¼0.25 m Dampers cx ¼ 20 N:s=m


cy ¼ 20 N:s=m

Helix angle b ¼30 Number of modes s¼2 Operating range of 0o ax o 0:22 m


position 0 o ay o 0:22 m

Feed rate per tooth per cf ¼ 0:2 mm=rev-tooth Inside & outside diameters Di ¼3 mm Position (initial value) Do ¼ 0:1 m
revolution ay ¼ 0:2 m

Do ¼ 15 mm

Table 1.2
Realistic parameters of the system for cutting force coefficients and milling process.

Cutting force coefficients x1 ¼ 4790 N=mm3 ; Z1 ¼ 128 N=mm3 ; x3 ¼ 1167 N=mm; Z3 ¼ 350 N=mm;
xi ¼ Zi ¼ 0, i¼2,4
Milling process k~ x ¼ 9 kN=m, k~ y ¼ 50 kN=m, c~ x ¼ 10 N:s=m, c~ y ¼ 10 N:s=m
specifications
H. Moradi et al. / Journal of Sound and Vibration 331 (2012) 4668–4690 4679

Table 2
Natural frequencies of the extension part of cutting tool without/with absorber.

Natural frequencies (rad=s)

o1x o2x o3x o1y o2y o3y

Without absorber on1x ¼1570 on2x ¼ 9520 on3x ¼ 26,660 on1y ¼ 2420 on2y ¼ 9550 on2y ¼ 26,670

With absorber ðabest ,kbest Þ1x,y 1850 9544 26,665 2750 9558 26,672
ðabest ,kbest Þ2x,y 1610 9531 26,662 3760 9580 26,678
ðabest ,kbest Þ3x,y 1586 9528 26,664 2492 9555 26,673

Fig. 3. Tool tip vibrations in (a) x direction and (b) y direction after implementation of the initial arbitrary absorber; with exerted concentrated cutting
forces at d~ ¼ la=2 (blue circles) and distributed cutting forces along the axial depth of cut (black dots); while a ¼ 4:5 mmo alim . (For interpretation of
the references to color in this figure legend, the reader is referred to the web version of this article.).

5.1. Structure of the optimal algorithm

In this paper, an optimal trial and error based algorithm is used to find the optimum specifications of the absorbers
such that the cutting tool vibration is minimized. Absorbers’ position and springs’ stiffness values are considered as control
parameters (while the values of absorbers’ mass and damper are kept constant). For simplicity, the distance between
4680 H. Moradi et al. / Journal of Sound and Vibration 331 (2012) 4668–4690

springs and dampers are assumed to be zero (ax ¼bx, ay ¼by). Therefore, springs position (ax, ay) denote absorbers’ position
as the control parameters.
For developing the algorithm, the extension part of cutting tool is diffracted into a finite number of elements (e.g., 50
elements) in z directions. Then, the absorbers are moved along the extension part by being placed at the end nodes of successive
elements (in both x–y directions). The midpoint deflection of all beam elements is computed. This procedure is repeated for
other next nodes as the absorbers are fixed there (50 nodes). Finally, among all the values for the absorbers position, their best
positions are those that the corresponding computed midpoint deflections of elements minimize the following index:
X
50
2 2
w¼ ðx^ s þ y^ s Þ (63)
s¼1

where x^ s and y^ s are the midpoint deflection of the sth-beam element after using the absorbers. It should be mentioned that
the proposed search for the best values of the absorbers’ position can be done for the best values of the springs’ stiffness
(while other parameters are kept constant). Moreover, this procedure can be performed with an optimum search on both
values of springs’ stiffness and absorbers’ position, simultaneously.

5.2. Optimum design of absorbers’ parameters for various chatter frequencies

In this section, optimum values of absorbers’ positions for various springs’ stiffness are found such that cutting tool
deflection is minimized (index given by Eq. (63)). This investigation is carried out for three cases of chatter frequencies:

Fig. 4. Absorber best position for (a) ax versus 1 o kx o 10 kN=m, (b) ax versus 10 o kx o 100 kN=m, (c) ay versus1 o ky o 10 kN=m and (d) ay
versus10 o ky o 100 kN=m; while oc ¼ 1550 rad=s  on1x ; and a ¼ 6 mm 4alim (corresponding to an unstable condition).
H. Moradi et al. / Journal of Sound and Vibration 331 (2012) 4668–4690 4681

oc ¼ 1550 rad=s  on1x , oc ¼ 2400 rad=s  on1y and on1x o oc ¼ 1800 rad=s o on1y while absorbers’ mass is kept constant
as given in Table 1.1 (mx ¼my ¼0.1) and a ¼ 6 mm 4alim (corresponding to an unstable condition). During the milling
process, since the dominant frequencies are the first natural frequencies in x–y directions, the investigation is done
around them.
To find optimum values of absorbers’ specifications, spring stiffness values are changed in the region 1okx o100 kN/m
and 1 oky o100 kN/m while allowable absorbers’ position is 0 oax o0.22 m and 0 oay o0.22 m (away from the cut
portion). Fig. 4 shows the best values of absorbers’ position for various values of springs’ stiffness while
oc ¼ 1550 rad=s  on1x . As it is shown, for both absorbers in x–y directions, their best position is constant for small
values of springs’ stiffness (for kx o10 kN/m and ky o10 kN/m). Similarly, their position is insensitive to the very large
values of springs’ stiffness (for kx 4100 kN/m and ky 4100 kN/m); that is not shown here. This is because, for very small
values of springs’ stiffness, corresponding simple natural frequencies of the absorbers (oxabsorber ¼ ðkx =mx Þ1=2 and
oy  absorber ¼ (ky/my)1/2) are too far from the first natural frequency of the extension part (Table 2). Therefore, for this
case (very small values of springs’ stiffness), using absorbers does not lead to significant reduction in cutting tool
vibrations (absorbers cannot store a considerable amount of undesirable energy which causes chatter vibration). On the
other hand, for very large values of the springs’ stiffness, the springs resist against deflections under transmitted external
forces; like rigid parts. Consequently, the undesirable motion of the cutting tool is not appropriately transmitted through
the absorbers. Therefore, for next simulation results, the operating medium range of 10 okx , ky o100 kN/m is used.
Similarly, Figs. 5 and 6 show the absorbers’ best position for various values of springs’ stiffness while
oc ¼ 2400 rad=s  on1y and on1x o oc ¼ 1800 rad=s o on1y , respectively.

Fig. 5. Absorber best position for (a) ax versus 10 o kx o 100 kN=m, (b) ay versus 10 o ky o 100 kN=m; while oc ¼ 2400 rad=s  on1y and a ¼ 6 mm4 alim .
4682 H. Moradi et al. / Journal of Sound and Vibration 331 (2012) 4668–4690

Fig. 6. Absorber best position for (a) ax versus 10 o kx o 100 kN=m, (b) ay versus 10 o ky o 100 kN=m; while on1x o oc ¼ 1800 rad=s o on1y and
a ¼ 6 mm 4alim .

Moreover, comparing Fig. 4b and d reveals that for the case of oc ¼ 1550 rad=s  on1x , best position of the vibration
absorber in x-direction is more sensitive to the springs’ stiffness values rather that of in y-direction. This is because for the
chatter frequency around the first natural frequency of the beam in x-direction (oc  on1x ), the related vibration absorber is
more active in chatter suppression.
According to Fig. 4b (oc ¼ 1550 rad=s  on1x ), for small and large values of kx, the absorber in x-direction must be
located at the end of the extension part (ax ¼0.22 m) while for the medium values of kx, it must be placed around the
middle of the extension part (ax ¼0.16 m). For other values of kx, the best position of x-absorber changes transitionally
from the middle to the end of the extension part. According to Fig. 4d, for a wide range of ky, absorber in y-direction must
be placed around the middle of the extension part (ay ¼ 0.09 m). For a limited range of 15 o ky o28 kN=m, it must be
located at the end of extension part (ay ¼0.21 m) and gradually moves toward the middle of the extension part
(ay ¼0.09 m).
According to Fig. 5a (oc ¼ 2400 rad=s  on1y ), for small and large values of kx, the absorber in x-direction must be
located around the middle of the extension part (ax ¼0.09 m) while for the medium values of kx, it must be placed at the
end of the extension part (ax ¼0.21 m). For other values of kx, the absorber in x-direction must be moved from the end to
the middle of the extension part. According to Fig. 5b, for medium and larger values of ky, the absorber in y-direction must
be placed around the middle of the extension part (ay ¼0.09 m) and at the end of the extension part (ay ¼0.21 m),
respectively.
H. Moradi et al. / Journal of Sound and Vibration 331 (2012) 4668–4690 4683

According to Fig. 6a and b (on1x o oc ¼ 1800 rad=s o on1y ), for a limited ranges of springs’ stiffness (65okx o 70 kN/
m,76okx o88 kN/m and 40 oky o55 kN/m), best positions of the absorbers in x–y directions change sensitively. However
under this condition, where chatter frequency is away from the first natural frequencies of the extension part, best
positions of the absorbers are fixed for a wide range of spring’s stiffness. As Fig. 6 shows, for a wide range of springs’

Table 3
Optimum values for the absorbers’ position and springs’ stiffness and the corresponding reduction in tool tip vibrations of nonlinear and linear models
(after using the absorber).

oc ¼ 1550 rad=s  on1x oc ¼ 2400 rad=s  on1y on1x o oc ¼ 1800 rad=s o on1y

kbest ðkN=mÞ x-direction 68 72 85


y-direction 80 95 44
abest (m) x-direction 0.205 0.11 0.17
y-direction 0.09 0.19 0.1
Reduction in tool tip vibration Nonlinear model 55 44 12
Linear model 71 57 23

Fig. 7. Tool tip vibrations in x-direction (a) without and (b) with optimum absorber for the nonlinear (blue circles) and linear (black dots) models with
oc ¼ 1550 rad=s  on1x and a ¼ 6 mm4 alim . (For interpretation of the references to color in this figure legend, the reader is referred to the web version of
this article.)
4684 H. Moradi et al. / Journal of Sound and Vibration 331 (2012) 4668–4690

stiffness, absorbers in x–y directions must be placed around the end (ax ¼0.185 m) and middle (ay ¼0.1 m) of the extension
part, respectively.

5.3. The effect of optimum absorber on chatter suppression

Returning to the results of previous section for determination of absorbers’ best position for various springs’ stiffness,
optimum set of parameters is found such that cutting tool vibrations become minimized (Eq. (63)); while a ¼ 6 mm4 alim .
For instance, for the case of oc ¼ 1550 rad=s  on1x , among the results given by Fig. 4b and d, optimum set of absorbers’
position and springs’ stiffness are found as kx, best ¼68 kN/m, ky, best ¼80 kN/m, ax, best ¼0.205 m and ay, best ¼0.09 m (as
given in Table 3). After implementation of the optimum absorber with using these parameters, tool tip vibration is reduced
about 55 times (in comparison to the system without absorber). Similarly, for other cases of chatter frequencies
oc ¼ 2400 rad=s  on1y and on1x o oc ¼ 1800 rad=s o on1y , optimum values of absorbers’ parameters and the amount of
reduction in tool tip vibrations are given in Table 3. For these three cases of design, first three natural frequencies of the
system after implementation the related optimum absorber set are given in Table 2.
Optimum absorbers designed for the nonlinear model of cutting forces are implemented on the linear model (where
x1 ¼ Z1 ¼0). The related reductions in vibration amplitude of the tool tip are given in Table 3. As it is shown, more

Fig. 8. Tool tip vibrations in y-direction (a) without and (b) with optimum absorber for the nonlinear (blue circles) and linear (black dots) models with
oc ¼ 1550 rad=s  on1x and a ¼ 6 mm 4alim . (For interpretation of the references to color in this figure legend, the reader is referred to the web version of
this article.)
H. Moradi et al. / Journal of Sound and Vibration 331 (2012) 4668–4690 4685

reductions are obtained when the optimum absorbers are implemented on the linear model. Therefore, designing the
absorbers for the nonlinear model also guarantees the chatter suppression for the linear model (this investigation is done,
because using linear models is more common in industry). It should be mentioned that similar results can be presented for
any other chatter frequency.
Figs. 7 and 8 show the tool tip vibrations in x and y directions after implementation of the optimum absorber kx,
best ¼68 kN/m, ky, best ¼ 80 kN/m, ax, best ¼0.205 m and ay, best ¼0.09 m under the chatter conditions with
oc ¼ 1550 rad=s  on1x and a large value of depth of cut a ¼ 6 mm4 alim . Results are presented when the optimum
absorbers are implemented on linear and nonlinear models. As it is shown, chatter suppression is guaranteed for both
models, especially with more rapid reduction in linear model. Fig. 9 shows the vibration amplitude of the extension part of
cutting tool (under steady-state condition) before and after implementation of the optimum vibration absorber with
oc ¼ 1550 rad=s  on1x and a ¼ 6 mm 4alim ; for linear and nonlinear models. As it is shown in Figs. 7–9, optimum
vibration absorber works efficiently in chatter suppression and guarantees the system stability for larger values of depth
of cut.

5.4. The effect of optimum absorber on frequency response of the system

To investigate the effect of chatter frequency on the optimum position of the absorbers, other parameters are kept fixed
while a ¼ 6 mm4 alim (corresponding to an unstable condition). For instance, optimum values of the absorbers’ stiffness
as kx, best ¼68 kN/m and ky, best ¼ 80 kN/m are used (Table 3). This set is chosen because according to the previous results,
the chatter vibration is more dominant in the vicinity of the first natural frequency of the extension part

Fig. 9. Extension part vibrations (a) for nonlinear model without absorber in x (black circles) and y (blue dots) directions; (b) in x-direction and (c) in
y-direction after using optimum absorber for nonlinear (black circles) and linear (blue stars); while oc ¼ 1550 rad=s  on1x and a ¼ 6 mm4 alim .
(For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
4686 H. Moradi et al. / Journal of Sound and Vibration 331 (2012) 4668–4690

(oc ¼ 1550 rad=s  on1x ). Using these values, chatter frequency is changed over a wide range of values. For each value,
absorbers’ best positions are determined such that the index w given by Eq. (63) is minimized. Fig. 10 shows the optimum
position for absorbers in x–y directions for chatter frequencies of 0 o oc o12000 rad/s (including the first and second
natural frequencies of the extension part).
According to Fig. 10 (0 o oc o 12,000 rad=s), best position of the absorber in x-direction shows more sensitivity to the
variation in the chatter frequency (rather than that of in y-direction). With the chosen values of kx,best ¼ 68 kN=m and
ky,best ¼ 80 kN=m, corresponding natural frequencies of the absorbers (oxabsorber ¼ ðkx =mx Þ1=2 and oyabsorber ¼ ðky =my Þ1=2 )
are found close to the first natural frequency of the extension part in x-direction (on1x in Table 2). Consequently, for various
chatter frequencies, the absorber in x-direction must be moved to more different points along the extension part to
suppress regenerative chatter (Fig. 10a). But as it is shown in Fig. 10b, generally two best positions exist for the absorber in
y-direction. First, for the range of oc o 2000 rad=s and 4000 o oc o9600 rad=s, the absorber must be placed at the middle
of the extension part (ay ¼0.09 m) and second for the range of 2000 o oc o4000 rad=s and 9600 o oc o 12,000 rad=s, it
must be located around the end of the extension part (ay ¼0.19 m).
After determination of absorbers’ best positions, as shown in Fig. 10, and using the stiffness values of kx,best ¼ 68 kN=m
and ky,best ¼ 80 kN=m, frequency response of the tool tip in x–y directions without and with optimum absorber is shown in
Fig. 11 (for linear and nonlinear models). As it is shown, using optimum absorber set guarantees the chatter suppression
over a wide range of chatter frequencies for both linear and nonlinear models.

Fig. 10. Absorber optimum position in (a) x-direction and (b) y-direction for chatter frequencies 0 o oc o 12,000 rad=s while kx,best ¼ 68 kN=m and
ky,best ¼ 80 kN=m while a ¼ 6 mm 4alim .
H. Moradi et al. / Journal of Sound and Vibration 331 (2012) 4668–4690 4687

Fig. 11. Tool tip vibration amplitude (a) for nonlinear model without absorber in x (black circles) and y (blue dots) directions, (b) in x-direction and (c)
y-direction for linear (blue stars) and nonlinear (black circles) after implementation of the optimum absorber; while kx,best ¼ 68 kN=m; ky,best ¼ 80 kN=m
and a ¼ 6 mm 4alim . (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Fig. 12. Stability limits of the milling process obtained by trial and error based algorithm for linear model without absorber (black stars), nonlinear
model with optimum absorber (red triangles) and linear model with optimum absorber (blue squares); while kx ¼ 68 kN=m and ky ¼ 80 kN=m,
ax ¼ 0.205 m and ay ¼ 0.09 m. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
4688 H. Moradi et al. / Journal of Sound and Vibration 331 (2012) 4668–4690

5.5. The effect of optimum absorber on stability lobes diagram

Stability of any machine-tool system can be shown in the diagrams in which the effects of axial depth cut, cutting
velocity and other machining parameters on stability are analyzed. In stability lobes diagram, the margins between stable
and unstable conditions are shown with several lobes for various conditions of axial depth of cut and spindle speed, e.g.,
[1,6,17].
In this section, to investigate the milling stability, instead of classical approaches used for determination of stability
lobes [6,17,40], first a trial and error based algorithm is developed in MATLAB via SIMULINK Toolbox to analyze Eq. (41).
For various spindle speeds and consequently chatter frequencies, dynamics of the system without and with absorber is
simulated. Axial depth of cut is increased gradually until reaching unstable conditions. The corresponding value is
considered as the critical axial depth of cut. Stability limits for the linear model without absorber, linear and nonlinear
models with optimum absorber is shown in Fig. 12 (while parameters of the optimum absorber are chosen as
kx ¼ 68 kN=m and ky ¼ 80 kN=m, ax ¼0.205 m and ay ¼0.09 m, as for the case oc  on1x ). It should be noticed that for the
linear and nonlinear models, two different formulations are considered for the cutting forces as first and third-order

Fig. 13. Stability limits of the milling process obtained from trial and error based algorithm (blue squares and red triangles) and semi-discretization
method (black circles) for (a) linear model and (b) nonlinear model after implementation of the optimum absorber; while kx ¼ 68 kN=m and
ky ¼ 80 kN=m, ax ¼0.205 m and ay ¼ 0.09 m. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of
this article.)
H. Moradi et al. / Journal of Sound and Vibration 331 (2012) 4668–4690 4689

polynomial functions of chip thickness, respectively (Eqs. (3) and (7)). So, coefficients Lx and Ly of Eq. (33) and
consequently of Eq. (41) differ for two models. Therefore, stability boundaries obtained for linear and nonlinear models are
different.
As the second approach, semi-discretization method (SDM) is applied for stability analysis of the milling process (with
the procedure explained in Section 4). Fig. 13 shows the stability lobes diagram obtained from the trial and error based
algorithm and SDM for both linear and nonlinear models (after implementation of the absorber). As is it shown, there is a
good agreement between stability boundaries determined from two different approaches.
In addition, as Figs. 12 and 13 show, using optimum absorbers improves the stability lobes diagram through increasing
the critical value of depth of cut. Therefore, larger values of axial depth of cut and consequently higher material removal
rate (MRR) can be achieved after applying the optimum absorber (without moving to unstable conditions). Moreover, as it
is shown, implementation of the absorbers is more effective in increasing the stability limits for the linear model of cutting
forces (than the nonlinear model). Therefore, designing the absorbers for the nonlinear model, as accomplished in this
paper, also guarantees the chatter suppression and stability improvement for the linear model.

6. Conclusions

In this paper, the effect of tunable vibration absorber set (TVAs) in chatter suppression of the milling process with linear
and nonlinear models of cutting forces is investigated. The extension part of cutting tool is modeled as an Euler–Bernoulli
beam with in-plane lateral vibrations (in x–y directions). An extended dynamic model of the peripheral milling with closed
form expressions for the nonlinear cutting forces is presented. In the presence of regenerative chatter, coupled dynamics of
the system (including the beam and x–y absorbers) is described through nonlinear delay differential equations. Two
tunable vibration absorbers composed of mass, spring and damper elements are attached to the extension part (in x–y
directions). For various machining conditions, it is desired to determine optimum values of absorbers’ position and springs’
stiffness, as the control parameters, such that cutting tool vibrations is minimized (while the mass and damper of
absorbers are kept constant).
Optimum design of absorbers’ parameters is determined for various chatter frequencies through a developed optimal
based algorithm. The effect of optimum absorber set on chatter suppression, frequency response of the system and
stability lobes diagram is investigated. For various springs’ stiffness and chatter frequencies, best position of absorbers
along the extension part is determined. According to results obtained, as the chatter frequency approaches one of the
natural frequencies of the extension part, vibration absorbers acts more efficiently in chatter suppression. Also, using
optimum absorber set guarantees the chatter suppression over a wide range of chatter frequencies.
Moreover, stability limits are obtained by two different approaches: a trial and error based algorithm and semi-
discretization method. Results show a good agreement between two approaches. It is shown that after implementation of
the optimum absorbers, stability lobes diagram is improved. Therefore, with keeping stable conditions for a given spindle
speed, larger values of depth of cut and consequently higher material removal rate (MRR) can be achieved. Although TVA’s
are designed for the nonlinear model, their implementation on the linear model leads to the more stable conditions.
Intuitive clarity, simple design, reduction in hardware and development cost are the great advantages of the proposed
TVAs design, as a semi-active control approach.

References

[1] H.E. Merrit, Theory of self excited machine-tool chatter: contribution to machine-tool chatter research-1, Transactions of ASME Journal of Engineering
for Industry B87 (4) (1965) 447–454.
[2] J. Tlusty, F. Ismail, Basic nonlinearity in machining chatter, CIRP Annals, Manufacturing Technology 30 (1) (1981) 229–304.
[3] F. Koenigsberger, A.J.P. Sabberwal, An investigation into the cutting force pulsations during milling operations, International Journal of Machine Tool
Design and Research 1 (1966) 15–33.
[4] E. Budak, Analytical models for high performance milling, part I: Cutting forces, deflections and tolerance integrity, International Journal of Machine
Tools and Manufacture 46 (2006) 1478–1488.
[5] E. Budak, Y. Altintas, E.J.A Armarego, Prediction of milling force coefficients from orthogonal cutting data, Transactions of the ASME Journal of
Manufacturing Science and Engineering 118 (1996) 216–224.
[6] Y. Altintas, E. Budak, Analytical prediction of stability lobes in milling, Annals of the CIRP 44 (1) (1995) 357–362.
[7] W.X. Tang, Q.H. Song, S.Q. Yu, S.S. Sun, B.B. Li, B. Du, X. Ai, Prediction of chatter stability in high-speed finishing end milling considering multi-mode
dynamics, Journal of Materials Processing Technology 209 (2009) 2585–2591.
[8] B.P. Mann, P.V. Bayly, M.A. Davies, J.E. Halley, Limit cycles, bifurcations, and accuracy of the milling process, Journal of Sound and Vibration 277 (2004)
31–48.
[9] T. Insperger, G. Stepan, Updated semi-discretization method for periodic delay-differential equations with discrete delay, International Journal for
Numerical Methods in Engineering 61 (2004) 117–141.
[10] T. Insperger, G. Stépán, Semi-discretization for Time-delay Systems-stability and Engineering Applications, Springer, N.Y, 2011.
[11] J. Gradisek, M. Kalveram, T. Insperger, K. Weinert, G. Stepan, E. Govekar, I. Grabec, On stability prediction for milling, International Journal of Machine
Tools and Manufacture 45 (2005) 769–781.
[12] X.H. Long, B. Balachandran, Stability analysis for milling process, Journal of Nonlinear Dynamics 49 (2007) 349–359.
[13] Y. Ding, L.M. Zhu, X.J. Zhang, H. Ding, A full-discretization method for prediction of milling stability, International Journal of Machine Tools and
Manufacture 50 (2010) 502–509.
[14] B.P. Mann, K.A. Young, T.L. Schmitz, D.N. Dilley, Simultaneous stability and surface location error predictions in milling, ASME Transactions Journal of
Manufacturing Science and Engineering 127 (2005) 446–453.
4690 H. Moradi et al. / Journal of Sound and Vibration 331 (2012) 4668–4690

[15] N.D. Sims, B.P. Mann, S. Huyanan, Analytical prediction of chatter stability for variable pitch and variable helix milling tools, Journal of Sound and
Vibration 317 (2008) 664–686.
[16] O.A. Bobrenkov, F.A. Khasawneh, E.A. Butcher, B.P Mann, Analysis of milling dynamics for simultaneously engaged cutting teeth, Journal of Sound and
Vibration 329 (2010) 585–606.
[17] Y. Altintas, G. Stepan, D. Merdol, Z. Dombovari, Chatter stability of milling in frequency and discrete time domain, CIRP Journal of Manufacturing
Science and Technology 1 (2008) 35–44.
[18] N.D. Sims, G. Manson, B.P. Mann, Fuzzy stability analysis of regenerative chatter in milling, Journal of Sound and Vibration 329 (2010) 1025–1041.
[19] N.H Hanna, S.A. Tobias, A theory of nonlinear regenerative chatter, Transactions of ASME Journal of Engineering for Industry 96 (1974) 247–255.
[20] L. Vela-Martı́nez, J.C. Jáuregui-Correa, O.M. González-Brambila, G. Herrera-Ruiz, A. Lozano-Guzmán, Instability conditions due to structural
nonlinearities in regenerative chatter, Journal of Nonlinear Dynamics 56 (2009) 415–427.
[21] N. Deshpande, M.S. Fofana, Nonlinear regenerative chatter in turning, Journal of Computer Integrated Manufacturing 17 (2001) 107–112.
[22] G. Stepan, T. Insperger, R. Szalai, Delay, parametric excitation, and the nonlinear dynamics of cutting process, International Journal of Bifurcation and
Chaos 15 (9) (2005) 2783–2798.
[23] M. Wiercigroch, E. Budak, Sources of nonlinearities, chatter generation and suppression in metal cutting, Philosophical Transactions of the Royal
Society of London 359 (2001) 663–693.
[24] Y.S. Tarng, J.Y. Kao, E.C. Lee, Chatter suppressions in turning operations with a tuned vibration absorber, Journal of Materials Processing Technology
105 (2000) 55–60.
[25] H. Moradi, F. Bakhtiari-Nejad, M. R. Movahhedy, Using a vibration absorber to suppress chatter vibration in turning process with a worn tool, in:
Proceedings of ASME 2009 Des. Eng. Tech. Conf. & Comp. and Inf. in Eng. Conf., DETC2009-86100, California, USA, 2009.
[26] H. Moradi, F. Bakhtiari Nejad, M.R. Movahhedy, Tunable vibration absorber design to suppress vibrations: an application in boring manufacturing
process, Journal of Sound and Vibration 318 (2008) 93–108.
[27] M.H. Miguelez, L. Rubio, J.A. Loya, J. Fernandez-Saez, Improvement of chatter stability in boring operations with passive vibration absorbers,
International Journal of Mechanical Sciences 52 (2010) 1376–1384.
[28] K.J. Liu, K.E. Rouch, Optimal passive vibration control of cutting process stability in milling, Journal of Materials Processing Technology 28 (1991)
285–294.
[29] A. Rashid, C.M. Nicolescu, Design and implementation of tuned viscoelastic dampers for vibration control in milling, International Journal of Machine
Tools and Manufacture 48 (2008) 1036–1053.
[30] Y.S. Liao, Y.C. Young, A new online spindle speed regulation strategy for chatter control, International Journal of Machine Tools and Manufacture 36 (5)
(1996) 651–660.
[31] M. Zatarain, I. Bediaga, J. Munoa, R. Lizarralde, Stability of milling processes with continuous spindle speed variation: analysis in the frequency and
time domains, and experimental correlation, CIRP Annals—Manufacturing Technology 57 (1) (2008) 379–384.
[32] S. Seguy, T. Insperger, L. Arnaud, G. Dessein, G. Peigne, On the stability of high-speed milling with spindle speed variation, International Journal of
Advanced Manufacturing Technology 48 (2010) 883–895.
[33] K.H. Hajikolaei, H. Moradi, G.R. Vossoughi, M.R. Movahhedy, Spindle speed variation and adaptive force regulation to suppress regenerative chatter
in turning process, Journal of Manufacturing Process 12 (2) (2010) 106–115.
[34] J.R. Pratt, A.H. Nayfeh, Chatter control and stability analysis of a cantilever boring bar under regenerative cutting conditions, Philosophical
Transaction of the Royal Society of London, part A 359 (2001) 759–792.
[35] B.K. Fussell, K. Srinivasan, Adaptive control of force in end milling operations—an evaluation of available algorithms, Journal of Manufacturing
Systems 10 (1) (1991) 8–20.
[36] C. Mei, Active regenerative chatter suppression during boring manufacturing process, Journal of Robotics and Computer Integrated Manufacturing 21
(2005) 153–158.
[37] E. Abele, H. Hanselka, F. Haase, D. Schlote, A. Schiffler, Development and design of an active work piece holder driven by piezo actuators, Production
Engineering Research Development 2 (2008) 437–442.
[38] H. Moradi, M.R. Movahhedy, G.R. Vossoughi, Sliding mode control of machining chatter in the presence of tool wear and parametric uncertainties,
Journal of Vibration and Control 16 (2) (2010) 231–251.
[39] H. Moradi, M.R. Movahhedy, G.R. Vossoughi, Robust control strategy for suppression of regenerative chatter in turning, Journal of Manufacturing
Process 11 (2) (2009) 55–65.
[40] Y. Altintas, Manufacturing Automation: Metal Cutting Mechanics, Machine Tool Vibrations, and CNC Design, Cambridge University Press, 2000.
[41] H. Moradi, M.R. Movahhedy, G.R. Vossoughi, Linear and nonlinear model of cutting forces in peripheral milling: a comparison between 2D and 3D
models, ASME 2010 International Mechanical Engineering Cong. and Experimental, IMECE (2010) 2010–38641. British Columbia, Canada.
[42] L. Meirovitch, Principles and Techniques of Vibrations, Prentice Hall, New Jersey, 2000.
[43] L.N. Virgin, Introduction to Experimental Nonlinear Dynamics, Cambridge University Press, UK, 2000.

You might also like