Chatter Stability of Machining Operations

Dedicated to S.A. Tobias and J. Tlusty

Yusuf Altintas, (lead)

Professor, Fellow ASME
The University of British Columbia
Department of Mechanical Engineering
Manufacturing Automation Laboratory (MAL)
2054-6250 Applied Science Lane
Vancouver BC V6T 1Z4 Canada
[email protected]

Gabor Stepan,
Professor of Applied Mechanics, Budapest University of Technology and Economics
[email protected]

Erhan Budak,
Professor, Faculty of Engineering and Natural Sciences, Sabanci University, Turkey
[email protected]

Tony Schmitz,
Professor, Fellow ASME
University of Tennessee, Knoxville
Joint Faculty, Oak Ridge National Laboratory
[email protected]

Zekai Murat Kilic,

Lecturer, University of Manchester,
[email protected]

Abstract
This paper reviews the dynamics of machining and chatter stability research since the first stability
laws were introduced by Tlusty and Tobias in the 1950s. The paper aims to introduce the
fundamentals of dynamic machining and chatter stability, as well as the state of the art and research
challenges, to readers who are new to the area. First, the unified dynamic models of mode coupling
and regenerative chatter are introduced. The chatter stability laws in both the frequency and time
domains are presented. The dynamic models of intermittent cutting, such as milling, are presented
and their stability solutions are derived by considering the time-periodic behavior. The
complexities contributed by highly intermittent cutting, which leads to additional stability pockets,
and the contribution of the tool’s flank face to process damping are explained. The stability of
parallel machining operations is explained. The design of variable pitch and serrated cutting tools
to suppress chatter is presented. The paper concludes with current challenges in chatter stability of

MANU-19-1385 Altintas p.1

machining which remains to be the main obstacle in increasing the productivity and quality of
manufactured parts.

1. Introduction
Chatter continues to be the main limitation in increasing material removal rates, productivity,
surface quality, and dimensional accuracy of machined parts. F. W. Taylor, who was considered
to be the initiator of manufacturing engineering, declared that chatter was the “most obscure and
delicate of all problems facing the machinist’’ in his 1906 ASME article [1]. The early
investigations on chatter mechanisms were conducted by Arnold [2], who classified machining
vibrations as forced and self-induced types. He conducted several turning experiments at varying
speeds and feeds and reported the influence of cutting conditions and tool wear on the shape of the
vibration waves and stability. Doi and Kato presented the effect of time lag in the thrust force
relative to the chip thickness variation as the source of chatter instability in turning [3]. The first
scientific stability laws were independently presented at almost at the same time by Tobias et al.
[4] and Tlusty et al. [5] in the 1950s. Tlusty presented an absolute stability law that predicted the
critical depth of cut proportional to the machine’s dynamic stiffness and inversely proportional to
the material’s cutting force coefficient in orthogonal cutting regardless of the spindle speed. Tobias
presented a similar stability law, but included the effect of spindle speed, i.e., the regenerative time
delay, which led to the stability pockets or “lobes” in orthogonal cutting. The dynamics of cutting
and chatter stability models were reviewed by Tlusty [6, 7] and Altintas and Weck [8]. General
literature reviews on the sources of nonlinearities in the dynamics of cutting [9] and chatter [10]
were also presented. Altintas et al. presented the frequency and time domain chatter stability
prediction laws [11].

The chatter theories have been applied to the analysis of machine tools to improve their dynamic
stiffness through design modifications or by adding passive and active dampers. Machine tools are
dynamically tested; the critical mode shapes which affect their chatter stability are identified and
modified to strengthen their stiffness by machine tool designers. Merritt converted the stability
equations of Tlusty and Tobias into a closed-loop system with unity and delayed feedback to
consider regeneration and solved the stability using the Nyquist criterion [12]. Tlusty showed that
the productivity gain is proportional to the dynamic stiffness improvements at the tool-workpiece
contact zone in the direction of chip generation [13]. The dynamics of various metal cutting and
grinding operations have been modeled by several researchers and their stability was solved by
applying the one-dimensional, frequency domain stability laws of Tlusty and Tobias. Tlusty
pioneered the development of high-speed milling machines by stating that the product of spindle
speed and the number of teeth on the cutter (i.e., the tooth passing frequency) should be matched
with the first bending mode of the spindle to operate the machine at the highest speed (first, or
rightmost) lobe which leads to highest material removal rate [7]. High speed-high power spindles
were developed to remove more than 95% of the material from aluminum slabs to produce
monolithic, lightweight parts for the aerospace industry [14, 15]. As the application domains have
grown, the limitation of Tobias’s and Tlusty’s one-dimensional stability theories have also been
investigated. It was observed that as the spindle speed is reduced relative to the natural frequency
of the machine, the stability of the process increases due to process damping. Tobias attributed this
process damping effect to the penetration of the cutting edge into the wavy cut surface [16] which
is still studied at the present. Process damping significantly depends on the ratio of the cutting
speed to the chatter frequency caused by the dominant mode. In the presence of low-frequency

MANU-19-1385 Altintas p.2

modes for multi-mode systems, the process damping effect may not be observed at low speeds as
would be expected. Munoa et al. [17] noted that if the tooth passing frequency is low relative to
the natural frequency of the mode, the process damping may stabilize the cutting process. On the
other hand, if the tooth passing frequency is several times higher than the natural frequency, the
mode can hardly create chatter problems. Tunc et al. [18] showed that when the high-frequency
mode is suppressed by the process damping effect, the chatter mode shifts to the low-frequency
mode even if the low-frequency mode is more rigid. This is critically important when machining
thin-wall parts.

Sridhar [19] showed that milling operations have coupled dynamics in two directions with periodic
coefficients; therefore, the stability of such multi-point machining applications cannot be solved
by the one-dimensional chatter theories of Tlusty and Tobias. Minis and Yanushevsky [20, 21]
applied Flouquet theory to solve the stability of two dimensional, periodic milling operations in
the frequency domain. Budak and Altintas determined the stability limit by considering the
harmonics of tooth passing frequencies in the eigenvalues [22, 23], as well as a direct solution of
stability analytically by considering the averages of directional factors [24] in the frequency
domain. The research group of Stepan presented a semi-discrete time-domain stability solution of
a single [25] and two-directional milling operations [26]. The research in chatter stability
continues due to its complexity and application to the design of machine tools and cutting tools,
as well as their chatter-free and productive use in machining operations.

The aim of this paper is not a general review of the chatter literature, but rather the presentation of
mathematical modeling of the dynamic cutting process and its stability solution in the frequency
and time domains; see Sections 2 and 3. The mechanism behind the appearance of stability islands
for highly intermittent cutting operations at high spindle speeds is explained in Sections 2.3 and
3.3. The design of special-purpose cutters to suppress chatter is presented with the aid of stability
models in Section 4, and the complexities of parallel machining dynamics are discussed in Section
5. The paper is intended to provide strong foundations in dynamic cutting and chatter stability
while explaining the current research challenges including improved accuracy (Section 6). The
paper is concluded in Section 7 by pointing out the future challenges in chatter research. The paper
is dedicated to the two pioneering scientists, the late Professors Tobias and Tlusty, who contributed
immensely to chatter stability literature that led to the present high-speed machine tool and high-
performance machining technologies.

2. Chatter Stability of Orthogonal Cutting

A general diagram of a single point cutting system, where the tool edge is orthogonal to the cutting
speed vc and the chip thickness varies normal to the velocity in the direction r , is shown in Figure
1a. The system is assumed to have multiple flexibilities in directions xi with an inclination angle
of i from the velocity direction. The tangential ( Ft ) and thrust ( Fr ) forces act in the directions
of cutting velocity (vc ) and its normal ( r ) , respectively. The resultant cutting force
( Fc  Ft 2  Fr 2 ) has an orientation angle of  with the velocity direction. The resultant cutting
force excites the flexibilities leading to vibrations in directions xi . When the vibrations generated
during the previous and current passes are projected into the direction of chip thickness ( r ) , the

MANU-19-1385 Altintas p.3

chatter type is considered to be regenerative and it is most commonly observed in production. Even
if the previous pass is neglected, the vibrations at the current pass in two orthogonal directions
affect the cutting forces which lead to the mode coupling chatter. The regenerative and mode
coupling chatter mechanisms are presented first, followed by the process damping mechanism to
consider the effect of the tool’s flank contact with the wavy surface finish.
2.1 Chatter Stability Model in the Frequency Domain
The chip thickness h(t ) varies as a function of the present vibration amplitude r (t ) and vibrations
left on the cut surface during the previous pass r (t  T ) [4, 5]:

h(t )  h0  hd (t )  h0   r (t )  r (t  T )  , (1)

where the delay T is the time between the two passes, h0 is the commanded (static) chip thickness
which corresponds to the feed per revolution in orthogonal turning (Figure 1a), and ℎ𝑑 is the
dynamic (time-varying) chip thickness. The chip thickness (h(t )) produces variable tangential
( Ft ) and normal ( Fr ) cutting forces which form the resultant cutting force ( Fc ) as:

Ft (t )  Kt ah(t ) , Fr (t )  K r Ft (t )  K r Kt ah(t )
, (2)
Fc (t )  Ft 2 (t )  Fr 2 (t )  K c ah(t )  K c  Kt 1  K r2

where Kt is the tangential cutting force coefficient, Kr  Fr / Ft  tan  is the radial to tangential
force ratio, and a is the depth of cut. The edge components of the cutting forces, which depend
only on the depth of cut, are neglected since they don’t contribute to the regeneration of chip
thickness. The cutting force coefficients are either identified from the orthogonal to oblique cutting
transformation using Merchant’s thin shear plane model [27, 28] or mechanistically evaluated from
cutting tests by considering the chip thickness and thermal softening effects [29]. The resultant
cutting force excites each spring i attached to the tool by projection into the corresponding mode’s
direction:

Figure 1 Dynamics of orthogonal cutting system. a) Regenerative mechanism. b)

Process damping mechanism

MANU-19-1385 Altintas p.4

 Fi (t )  Fc (t ) cos(i   ) . (3)

Each mode with a transfer function of (ii ( s)) will cause vibration xi ( s) in the Laplace domain
( s ) as:

xi ( s)  ii ( s) Fi ( s) . (4)

By projecting and superposing all the vibrations in the chip thickness (r) direction, vibration in
this direction is calculated according to:
l
r ( s )   o ( s ) Fc ( s )   o ( s )   sin i cos i     i ( s)  , (5)
i 1

where  o ( s) is the oriented transfer function between the vibrations in the chip thickness and
resultant cutting force directions. By neglecting the static chip thickness ( h0 ), which does not
affect stability, the dynamic or regenerative chip thickness contributed by the vibrations is
evaluated as:

hd (t )  r (t )    r (t )  r (t  T )  , (6)

where the time delay is T  60 / n for the spindle speed n [rev/min]. Substituting the dynamic
chip thickness into the resultant cutting force (Eq. (2)) yields the dynamic cutting force in the
Laplace domain as:

Fc ( s)   K c ar ( s)   K c a(1  e  sT ) 0 ( s) Fc ( s). (7)

The characteristic equation of the system is:

1  Kc a(1  e sT ) 0 ( s)  0 , (8)

which has an infinite number of roots due to the delay term e  sT . By substituting 𝑠 = 𝑖𝜔𝑐 for the
critical stability condition, the characteristic equation of the dynamic cutting system in the
frequency domain becomes:
1 1

1  Λ 0 (ic )  0  Λ  aK c 1  eicT   
 0 (ic ) G0 (c )  iH 0 (c )
, (9)

where Go and H o are the real and imaginary parts of oriented frequency response function, or FRF
(  0 ). Expanding Eq. (9) gives:

1
Λ
G0 (c )  iH 0 (c )
 
 aK c 1  e icT  aK c 1  cos cT   i sin cT 
(10)
1  aK c 
G0 1  cos cT   H 0 sin cT   iaK c G0 sin cT  H 0 1  cos cT   0

and equating the imaginary part to zero leads to the critical spindle speed ( n ) :

MANU-19-1385 Altintas p.5

 H0 sin cT
G0 sin cT  H 0 1  cos cT   0  tan  
G0 cos cT  1
 T 
tan  tan  c  3   cT  2  3  (2kl  1)  2  kl  1, 2,.. (11)
 2 2
60 c
Speed: n[rev / min]   60
T (2kl  1)  2

By also equating the real part to zero and substituting H 0 / G0  sin cT /  cos cT  1 , the critical
depth of cut alim is determined to be:

1 1
alim   . (12)
 H ( )  2 KcG0 (c )
KcG0 (c ) 1  cos cT   0 c sin cT 
 G0 (c ) 

The solution leads to a positive depth of cut only when the real part of the oriented FRF (G0 (c ))
is negative. Tobias considered the entire negative region of the real part of the FRF which leads
to the critical depth of cut and corresponding spindle speeds for each integer number of vibration
waves ( kl ) generated within the time delay (i.e., spindle period) T that led to the stability lobes
[4]. Tlusty considered the worst case by taking the minimum of G0 (c ) which led to the minimum
depth of cut regardless of the spindle speed [5]:
1
amin  . (13)
2Kc min G0 (c )

Case 1: Mode coupling with zero time delay ( r (t  T )  0 )

The eigenvalue (Eq.(9)) in mode coupling does not contain the delay term and the system must
have at least two flexibilities which affect the chip thickness through the excitation by the cutting
force. Consider a case with two flexibilities which are orthogonal to each other, but oriented from
the velocity and chip thickness directions, i.e. 2  1   / 2 . The characteristic equation (9)
becomes:

 o ( s)  sin 1 cos 1    1 ( s)  cos 1 sin 1     2 ( s)

1
1  Λ 0 (ic )  0  Λ  aK c 
G0 (c )  iH 0 (c )
1
alim 
K c G0 (c )  iH 0 (c ) 

Since alim is a real number, the imaginary part of the oriented FRF must have a zero crossing, i.e.,
H 0 (mc )  0 at the mode coupling chatter frequency mc . Mode coupling chatter is independent
of speed and has the limiting depth of cut:

MANU-19-1385 Altintas p.6

 1
alim  , (14)
K c G0 (mc )

which leads to about two times more stable depth than regenerative chatter. The orientation of the
cutting force and modes alter the stability as indicated by Eq. (14). If the imaginary part of the
oriented FRF does not cross zero, mode coupling would not occur. Ismail et al. showed that the
mode coupling stability can be improved by selecting the ratio of modes through the modification
of end mill stiffness [30].
Case 2: Regenerative chatter stability with process damping
Consider that the system is flexible only in the chip direction ( r ) and the force is also in the same
direction for simplicity to explain the process damping mechanism (   / 2,    / 2) . The
clearance face and edge radius of the tool have time-varying ploughing contact with the presently
cut wavy surface (Figure 1b) which generates a damping force, aC p r / vc , proportional to the
process damping constant of the material ( C p [ N / m] ) and the ratio of the vibration velocity (𝑟̇ =
𝑑𝑟/𝑑𝑡) to the cutting speed (vc ) as described by Das and Tobias [31]. Assuming a structure with
stiffness ( k ) , a viscous damping coefficient ( c ) , and a mass ( m ) with a corresponding natural
frequency n  k / m and damping ratio   c / (2 km ) is excited by the regenerative cutting
and process damping forces, Fr (t )  K r Kt ah(t )  aC p r / vc , the equation of motion for orthogonal
cutting is expressed as:

n2 n2  r
r  2n r  n2 r  Fr (t )   K r K t ah(t )  aC p  (15)
k k  vc 

Note that the process damping term contains vibration velocity ( r ), which increases the viscous
damping of the system. Eq. (15) is expressed in the Laplace domain as:

2 2 1
 s 2  2n s  n 2  r ( s)  n K r Kt a  h0  1  e  sT  r ( s )   n
k   k v C p sr ( s )
c
(16)
r (s) n 2 / k
( s)   2
Fr ( s) s  2n s  n 2

where ( s) is the transfer function of the system’s structural dynamics in the chip thickness
direction. Eq. (16) was expressed as a closed-loop system by Merrit [12], but without the process
damping term which is now included in Figure 2. The closed-loop transfer function of the system
is expressed as:

MANU-19-1385 Altintas p.7

 h0 + + h KK + n2 / k r
_ a (s)  2
_ s  2n s  n2
r t
+

Cp
s
V

e  sT
Figure 2 Block diagram of dynamic orthogonal cutting with process damping and regenerative
chip feedback.

Cp
1 a ( s) s
h( s ) vc
 . (17)
h0 ( s )  C 
1  a (1  e  sT ) K r K t  p s  (s)
 vc 


The characteristic equation of the system 1  a (1  e sT ) Kt  (C p / vc )s  (s) has an infinite 
number of roots ( s    ic ) due to the delay term e  sT
. By considering the critical stability
limit, where the real part of the root is zero (   0 ) and the structure’s FRF is represented by its
real and imaginary parts (   (ic )  G (c )  iH (c )  ), the characteristic equation becomes:

   C p       C   
1  K r K t alim G 1  cos cT   H  sin cT  c     i  K r K t alim G  sin cT  p    H 1  cos cT     0 .
   vc       vc   
(18)

Note that regardless of whether the system has a regenerative delay ( e  jcT ) or not such as in mode
coupling, the process damping affects the system dynamics as shown in Eq. (18). Tlusty et al. [5]
and Tobias et al. [4] neglected the process damping term (i.e., C p  0 ) to find an analytical,
frequency domain stability solution in their initial publications as described in Case 2. By forcing
both the real and imaginary parts of Eq. (18) to zero, the critical depth of cut ( alim ) and spindle
speed ( n ) are found by Tobias and Fishwick [4] as :

1 c H (c )
alim [m]  ; n[rev / min]  60    tan 1 , kl  0,1, 2,.. (19)
2 K r K t G (c ) (2kl  1)  G (c )

MANU-19-1385 Altintas p.8

The stability solution exists only when the real part of the FRF is negative [4], i.e., G(c )  0 .
Since the negative real part has the smallest value near the natural frequency (n ) of the structure,
the highest depth of cut is achieved when the spindle speed is close to the natural frequency (i.e.,
the first stability lobe kl  0 ) or at its integer (kl ) fractions or lobes [7]. Tlusty and Polacek [5]
took the minimum negative real part of the FRF [5] as:
1 2 k 1
amin    Gmin (c )  , (20)
2 K r K t Gmin (c ) K r K t 4k

which gives the absolute minimum depth of cut that is linearly proportional to the dynamic
stiffness of the machine ( 2k ) and inversely proportional to the product of the cutting force
coefficients for the selected work material ( Kr Kt ) . Tlusty’s simple formula (Eq.(20)) has been
widely used as a guide by machine tool designers and Tobias’s formula has been applied to select
the stable depth of cuts at high spindle speeds [7].
Sample stability lobes are shown for mode coupling, with and without the process damping effect,

Figure 3 Chatter stability diagrams and corresponding chatter frequency diagram for
mode coupling chatter, and regenerative chatter with and without process damping.
Simulation parameters: Cutting coefficients - Kt=1000 MPa, Kr=0.3; Modal parameters-
k1=15×106 N/m, k2=10×106N/m, n,1  150 Hz, n,2  250 Hz ,  1  0.010,  2  0.012 . The
orientation of flexibilities- 1  70deg , 2  160deg ; process damping constant - Cp=2e6
N/m; workpiece diameter = 30 mm. Simulated cases: h0  0.1mm / rev, a  2.5mm. Cases-
A: n=700 rev/min (stable, process damping zone), B: n=5000 rev/min (stable, high speed
zone), C: n=8000 rev/min (unstable, high speed zone).
in

MANU-19-1385 Altintas p.9

Figure 3. When the process damping is considered in Eq. (18), the spindle speed-dependent cutting
velocity term ( vc ) diverges from the solutions of Tobias (Eq.(19)) and Tlusty (Eq.(20)) at low
cutting speeds. In this case, the critically stable depths of cut are first analytically solved without
process damping. Later, the process damping is included and the depth of cut ( a ) at each speed (
n ) is increased until the critically stable condition is identified using the Nyquist stability criterion.
The real and imaginary parts of the characteristics equation (Eq.(18)) are computed within the
frequency range of the structure’s FRF. The process is considered to be unstable if the polar plot
of real and imaginary parts encircles the origin clockwise and is considered stable otherwise. If the
polar plot passes through the origin, the process is critically stable at that particular speed and
depth of cut [32, 33]. The process is repeated at each spindle speed to generate the stability lobes.
Alternatively, the viscous damping term is modified by including the process damping term (i.e.,
(2  aC p / vc ) , where the depth of cut ( a ) and speed ( vc ) are adopted from the process damping
free solution. The stability solution given in Eq. (19) is iteratively applied until the depth of cut
converges to a critical limit [34]. The process damping constants are usually identified
experimentally from cutting tests with or without chatter. With chatter [35], the constants are
determined by considering the energy dissipation of the vibrations due to tool penetration into
work material. The identification of process damping constants from chatter-free cutting tests is
either based on the estimation of dynamic cutting forces contributed by the tool flank contact with
the wavy surface [36] or the identification of equivalent damping ratios from the operational modal
analysis [37]. Tuysuz and Altintas recently presented an analytical model of process damping
based on contact mechanics laws [38].
2.2 Continuous time-domain analysis of turning
The stability analysis of stationary cutting processes can also be analyzed in the time domain using
the state space approach from the classical mathematical theory of differential equations. Consider
the governing equation of regenerative chatter for turning in the form of Eq.(15), where the cutting
force and the flexibility are assumed to be only in the normal direction r as in Figure 1a. By
neglecting the process damping term, the delay differential equation of the system becomes:
K r Kt
r (t )  2n r (t )  n2 r (t )  n2 a  h 0  r (t  T )  r (t )  , (21)
k
which is further simplified using the dimensionless time to be:
d d
t  n t , r (t )  r (t )  n r (t )  n r (t ), r (t )  n2 r (t ), r (t  T )  r (t  T ) . (22)
dt dt
~ ~
The dimensionless time delay T , spindle speed  and depth of cut a~ are defined as:
2 1  n KK
T  n T ,    n   , a  r t a. (23)
T 2 T n 60 f n k

The spindle speed n is measured in [rev/min], so the angular velocity of the spindle is
f  n / 60  1 / T in [Hz], and it is   2 / T  2 f in [rad/s], where the undamped natural

MANU-19-1385 Altintas p.10

frequency f n   n / 2 is in [Hz]. The equation of motion (Eq.(21)) can be simplified for the small
oscillation  (t )  r (t )  r0 where the static deformation

Kr Kt
r0  ah0  ah0
k
of the tool causes a surface location error. The resulting governing equation then assumes the form

 (t )  2 (t )   (t )  a  (t  T )  (t )  , (24)

where the three remaining parameters are: the damping ratio  ~,

, the dimensionless chip width a
and the dimensionless spindle speed  :

  2 / T   / n  f / f n ,

which is used to construct the stability charts in the parameter plane. The trial solution of the linear
delay-differential equation is the same as it is for the linear ordinary differential equations:

 (t )  e t  A cos(t )  B sin(t )  . (25)

Due to delay term, there are infinite values for  and  that satisfy Eq.(24). For the stability of
stationary cutting, all the possible  values must be negative for the exponentially decaying
vibrations. The values of  and  correspond to the characteristic exponents     i in the
Laplace domain, which corresponds to the Laplace transformation applied to Eq. (16) in the
frequency domain with the relation s  c . Accordingly, the stability boundaries can be found
~
in the (, a~) plane when Eq. (25) is substituted back into Eq. (24) with   0 . The separation of
~ ~
the coefficients of the cos(t ) and sin(t ) terms leads to two equations for the dimensionless
spindle speed () and the dimensionless depth of cut a ~ with the parameter  . This  can be
interpreted as a dimensionless chatter frequency since it relates to the chatter frequency  c
appearing in Eq. (18) according to   c / n  fc / f n along the stability boundaries when process
damping is neglected. The corresponding closed-form expressions for the critically stable
dimensionless spindle speed () and depth of cut ( ( a ) ) are:

 
 , kl  1, 2,...
1  1 2
kl  arctan 
 2  (26)
(  1)  4 
2 2 2 2 
a , 1     . 
2( 2  1) 

These curves are also called stability lobes and they are presented in Figure 4 together with their
characteristic parameters. The major characteristic parameters of the chart can be identified from
Eq.(26). For example, the lower bound for all lobes is a straight-line at the absolute limit for the

MANU-19-1385 Altintas p.11

depth of cut, below which the system is stable for any cutting speed. This value comes from Eq.
(23) and Eq. (26):
k 2 k
amin  2 (1   )  , (27)
Kr Kt Kr Kt

which is the same absolute minimum as given in Eq. (20) via the frequency domain model for
small damping ratios. At these minimum parameter points of the lobes, the angular chatter

frequency c and the corresponding “worst” spindle speeds  are:

1  2 4
c  1  2 n and   n  n , kl  1, 2,... (28)
1
kl  arctan
1 4kl  1
 1  2

Figure 4. Theoretical stability chart of turning processes in the plane of the dimensionless
spindle speed  and the dimensionless depth of cut a . The dimensionless chatter
frequency   fc / f n is greater than 1. The damping ratio is fixed at   0.05 , while
kl  1, 2,... refers to the sequence number of the lobes.

MANU-19-1385 Altintas p.12

while the stable pockets between the instability lobes can be reached close to the asymptotes of
the lobes at the spindle speeds   n / kl , (kl  1, 2,..) where the expected chatter frequency is
0

close to c  n . Note that at the intersection of the lobes, quasi-periodic chatter may also occur
0

with two frequencies involved. The dimensionless stability model in the time domain provides the
critical speeds where the depths of cut are either maximum or minimum as a function of the natural
frequency of the structure.
2.3 Discrete time-domain analysis of highly interrupted cutting
As opposed to the continuous time domain analysis, the discrete time-domain analysis can also be
represented in case of the simplified model of highly interrupted orthogonal cutting shown in
Figure 5. Consider the mechanical model of the regenerative mechanism introduced in Figure 1,
where only one vibration mode is relevant at the angle 1   / 2 , so the tool position is given by
the single coordinate x : x1  r . The cylindrical workpiece has a rotation period T , which means
that its circumference is Tvc , where vc is the cutting speed. However, the cylindrical workpiece
has a large groove of width (1   )Tvc , and it is hypothetically assumed that this groove is almost
as large as the whole circumference of the workpiece. The tool is therefore in cut for only very
short time intervals T  0 , where the dimensionless parameter  stands for the ratio of the time
spent in the cut to the time spent out of the cut. The depth of cut is still a , so during this very short
time of contact, the tool is subjected to a kind of impact load due to the cutting force applied during
the very short contact time T . Still, the regenerative mechanism appears again, since the impact
force will depend on the difference of the past and present positions of the tool during contact.

Note, that the same mechanical model can also be derived from the dynamics of milling presented
in Section 3 (see Figure 7). In that case, the rotating tool is considered to be rigid in the y direction
and flexible only in the x direction. Assume that the milling tool has only one cutting edge with
rotation period T , the cutting speed is vc again, and the circumference of the milling tool is Tvc .

Figure 5 Dynamic model of highly interrupted cutting

MANU-19-1385 Altintas p.13

Now, the workpiece is assumed to be thin with a thickness of Tvc in the cutting speed direction.
If the single cut at each rotation takes place at 1   / 2 only, the dynamic model of milling is
simplified to the one shown in Figure 5.
For these simplified conditions, the process can be solved analytically in the time domain and a
discrete time mathematical model can be constructed to determine an analytical stability chart
having a structure similar to the turning stability chart shown in Figure 4. This procedure applies
the time domain solution of the fly-over section as a lightly damped oscillator, while the short
cutting response is calculated according to the Newtonian impact theory: the position of the tool
does not change during this short cutting period, while the velocity of the tool changes abruptly
due to the impulse-like cutting force.
In accordance with the notation in Figure 5, consider that the tool leaves the workpiece at time
instants t j  jT , j  0,1,2,... At the ( j  1) time instant, the tool starts a free-flight with the initial
st

position x j 1  x(t j 1 )  x(t j  T ) and the initial velocity v j 1  x (t j 1 )  x (t j  T ) . The tool has a

damped angular natural frequency d  n 1   2 . The homogeneous ordinary differential

equation of the system during free-flight (i.e., free vibration with some initial conditions) is:

x(t )  2n x(t )  n2 x(t )  0 (29)

with the following solution

 en t   en t  
x(t )   cos(dt   )  x j 1   sin(d t )  v j 1 
 1  2   d
    
 vi  x((1   )T )  x(T ) , (30)
  en t   en t  
x(t )    n sin(dt )  x j 1   cos(dt   )  v j 1 
 1  2   1  2 
    

where the phase angle  is calculated from tan    / 1   2 . The position and the velocity of the
tool can then be calculated when it enters the workpiece again by substituting t  (1   )T  T into
Eq. (30) :

vi  x((1   )T )  x(T ) , (31)

where the negative sign in the superscript refers to the fact that these values occur at the start of
the short impact-like cutting. During this infinitesimally short cutting time of T , the variation of
the position of the tool is negligible, while the cutting force variation Fr  K r Kt a x j 1  x j has  
the linear impulse that causes the variation m v j  v j   of the tool’s linear momentum.

Accordingly, when the tool leaves the workpiece again at the time instant t j , its position and the
velocity can be calculated as:

MANU-19-1385 Altintas p.14

 x j  x j  x(T ) 

2 Kr Kt Kr Kt  (32)
v j  v j  n

a ( x j 1  x j ) T  x (T )  T an ( x (0)  x (T )) 
2

kx kx 

Substituting the values of x(0) and x(T ) from Eq. (30), a discrete time model is obtained that
describes the connection of the subsequent positions and velocities of the tool after each short
cutting period by means of the two-dimensional iteration:

 x j   A11 A12   x j 1 
    , j  1, 2,... (33)
 v j   A21 A22   v j 1 

where
e nT e nT
A11  cos(dT   ), A12  d sin(dT )
1 2

  cos( T   ) 
 nT
A21   n e sin(dT )  T n2 a 1 e  T
Kr Kt
1 2 kx 1 2 d

A22  e nT
1 2
 cos( T   )  T
d n  a  sin( T )  .
Kr Kt
kx d

The stability of the stationary highly interrupted cutting process is equivalent to the convergence
of the geometric vector series given in Eq.(33), which means that the quotients, 1,2 , that is, the
eigenvalues of the quotient matrix A in Eq.(33), must have moduli less than 1: 1, 2  1 .

In this discrete system (Eq.(33)), the counterpart of the classical chatter with angular frequency  c
appears when the following characteristic equation has the complex conjugate roots
1,2  eic  cos c  i sin c on the unit circle of the complex plane which satisfy:

det   I  A    2  ( A11  A22 )   ( A11 A22  A12 A21 )  0 . (34)

Using the same dimensionless quantities as defined in Eq.(23), the substitution of these roots into
Eq. (34) leads to the chatter boundaries a~c for the dimensionless axial depth of cut in the following
explicit form:

 2 
sinh   
 1  2   2 1   2
ac    >ac,min  , (35)
  2  
sin  1   2 
  
where the lower estimate for the absolute stable region is not as sharp as it is in case of turning,
but the formula is quite good in the high spindle speed domain as shown in Figure 6. However,
there appears another kind of vibration that does not occur in turning. This is the result of a period-
doubling (or period-2) bifurcation, when the critical characteristic root is 1  1 , while  2  1 .
It is called period-doubling because the corresponding time period of the critical chatter vibration
MANU-19-1385 Altintas p.15
is two times the time period of the spindle rotation. If this is substituted back into Eq. (34), the new
kind of instability lobes have the following closed-form expression:

 2   2 
cos  1   2   cosh   
 1  2
     >a  1  2
apd    . (36)
2  2 2  
pd,min
sin  1   
 

These stability boundaries are presented in Figure 6. Compared to the stability chart of turning in
Figure 4, the number of lobes is doubled, the classical chatter lobes become much thinner and,
although the period-doubling lobes show up between them, large stable pockets are still present at
~
  1 / k l similar to the case of turning. These stable pockets are relevant at high spindle speeds.

The chatter and period-doubling vibrations at the stability limits have a rich frequency content due
to the parametric excitation in the time-delayed system. This means that the critical self-excited
vibrations are not harmonic anymore, although they are periodic, and they still have a relevant
fundamental harmonic vibration component. The corresponding dominant frequency is usually the
lowest frequency among the many frequencies presented above the stability chart in Figure 6, but

Figure 6 Theoretical stability chart of highly interrupted cutting in the plane of the
dimensionless spindle speed  and the dimensionless (axial) depth of cut a. The
dimensionless chatter frequencies are   fc / f n . The damping ratio is fixed at   0.05 ,
  0.1 .

MANU-19-1385 Altintas p.16

not necessarily. Dombovari et al. [39] presented a simple method to identify the dominant chatter
frequency among the many harmonics.
This stability chart was constructed by Davies et al. [40, 41], and parallel to their work, by Budak
et al. [22] and Merdol et al. [42] with the multi-frequency method, by Insperger et al.[26] using
the semi-discretization method, and by Bayly et al. [43] using the time finite element method. The
experimental verification of this double-lobe structure together with the precise identification of
the rich chatter and period doubling frequencies including the relevant higher harmonics was
given, for example, by Insperger et al. [44], Mann et al.[45], and by Gradisek et al. [46].
The stability chart in Figure 6 is valid for the extreme conditions of highly interrupted cutting, but
it also provides the basic structure of the stability domains in milling processes especially for high-
speed milling with a low number of cutting edges and with low radial immersion. Still, the precise
determination of the lobe structure for general milling processes requires more sophisticated
numerical tools like the one presented in Section 3.2 where the continuous and discrete time-
domain analyses of Sections 2.2 and 2.3 are combined.

3. Chatter Stability in Multi-Point Machining

Multi-point machining operations, such as milling, are carried out with tools having multiple teeth
that periodically cut the material. The dynamics and stability of milling operations are summarized
in the following sections.
3.1 Stability of milling operations in the frequency domain
Milling cutters have multiple teeth that have intermittent engagements with the workpiece. A
diagram of milling with dynamic flexibilities in feed ( x ) and normal ( y ) directions is shown in
Figure 7. If the tooth j is at the angular immersion ( j ) which is measured clockwise from the
( y ) axis, the dynamic chip thickness in the radial direction is generated by the vibrations at the
present and previous tooth periods ( x(t )  x(t )  x(t  T ) , y (t )  y (t )  y (t  T ) ) as:

hdj (t )  x(t )sin  j  y (t ) cos  j . (37)

The dynamic chip creates tangential ( Ftj  Kt ah( j ) ) and radial ( Ftj ( j )  K r Ftj ) cutting forces at
each engaged tooth, which are projected in the feed ( x ) and normal ( y ) directions, and they are
summed to find the total force components exciting the structure [24].

 Fx (t ) 
  1  axx axy  x(t )  1
   aKt      F(t )  aKt A(t )Δ(t ) (38)
 Fy (t ) 
  2  a yx a yy  y (t )  2

MANU-19-1385 Altintas p.17

 Figure 7 Dynamics of milling

The directional factors, which exist when the cutter is engaged with the workpiece (i.e., when a
tooth is located between the start st and exit ex angles), are:
N 1 N 1
axx    g j [sin 2 j  K r (1  cos2 j )] ; axy    g j [(1  cos 2 j )  K r sin2 j )]
j 0 j 0
N 1 N 1
a yx   g j [(1  cos2 j )  K r sin 2 j ] ; a yy   g j [sin 2 j  K r (1  cos2 j )]
j 0 j 0

g j  1  st   j  ex and g j  0 otherwise

Since the cutter rotates at angular speed (rad / s ) , the immersion angle is time-varying
( j  t  ( j  1) p   p  2 / N ) . Consequently, the directional factors ( A (t ) ) are also time-
varying and periodic at the tooth passing interval (T ) , or at the tooth passing frequency
(T  2 / T ) in the frequency domain, or at cutter pitch angle intervals ( p  T  2 / N ) in the
angular domain. The periodic, dynamic cutting force (Eq.(38)) can be transformed into the
frequency domain as:
1
𝑭(𝜔) = 2 𝑎𝐾𝑡 [𝑨(𝜔)(1 − 𝑒 −𝑖𝜔𝑐𝑇 )𝒒(𝜔)] ← 𝒒(𝜔) = {𝑥(𝜔) 𝑦(𝜔)}𝑇 = 𝜱(𝑖𝜔)𝑭(𝜔), (39)

where Φ  j  is the FRF matrix of the structure in the ( xx, xy; yx, yy ) directions. The periodic
directional factors can be represented by their Fourier series components as [22]:
ex

N  axx axy  irN
A( )  Ae r
 irT t
; Ar 
2  a a yy 
e d , (40)
r  st  yx

MANU-19-1385 Altintas p.18

where N is the number of teeth on the uniform pitch cutter. The details of transformations can be
found in [33]. The resulting dynamic force for the milling system is expressed as [19]:
1
𝑭(𝜔) = 2 𝑎𝐾𝑡 [𝑨(𝜔)(1 − 𝑒 −𝑖𝜔𝑐𝑇 )𝜱(𝑖𝜔)𝑭(𝜔)] (41)

The stability condition provided in Eq. (41) is challenging to determine due to the presence of the
delay term eicT and the corresponding periodicity of the directional matrix A( ) at the unknown
tooth passing frequency T . Minis and Yanushevsky [21] applied the Floquet theory to identify
the critical stable depth of cut (alim ) and spindle speed ( n ) iteratively. Budak and Altintas
proposed two approaches: the zero-order solution [24], where only the average of directional
factors is considered; and the multi-frequency [22] solution, which includes the harmonics of the
directional factors for low immersion, highly intermittent milling operations.
Zero-Order Solution: In this case, only the average component of the directional matrix is
considered so that [24]:
ex
N  axx axy 
A0 
2  a
  yx a yy 
d . (42)
st

The dynamic milling (Eq. (41)) process then becomes time-invariant and its critical stability can
be found from the characteristic equation using:
1
   R  i I   K t a (1  e  icT )  det I ]  A 0 (i )  0  a0  2  a1  1  0 . (43)
2
From the computed real ( R ) and imaginary ( I ) parts of the eigenvalues, the critical stable
depth of cut (alim ) and spindle speed ( n ) are evaluated directly as [24]:

2 R 
alim   1   2     I
NKt R
(44)
  2 tan 1   kl 2 60
T (sec)  , n(rev / min)  , kl  0,1, 2, ..
c NT

MANU-19-1385 Altintas p.19

The analytical stability solution given in Eq. (44) is computationally inexpensive and sufficiently
accurate for most of the common milling operations found in the industry. A comparison of
experimentally validated stability lobes for the zero-order solution and time marching, numerical
simulations are shown in Figure 8 [33]. While the numerical method, which is used as an accurate
reference simulation, took more than 24 hours on a PC in 1997, the zero-order solution took less
than a second because it is a direct, analytical solution.

Figure 8 Stability lobes predicted with the zero-order solution analytically and time-domain
numerical simulations for a cutter with two circular inserts.

Multi-Frequency Solution: When the harmonics of tooth passing frequency are included in the
directional matrix A( ) , the dynamic milling equation (Eq. (39)) is expanded using Floquet
theory to obtain [22]:

F( )   Pl      c  lT 
l 
(45)
    1
F( )       A r l  Φ c  lT  F( )    aKt 1  eicT  
 r  l   2

This system has an infinite dimension, but it is truncated to few harmonics in practice. If the
intermittency is severe, such as the case for small radial depths of cut, more harmonics of the tooth
passing frequency (T ) need to be considered to obtain accurate results, which increases the size
of the eigenvalue problem. For example, if only one harmonic is used ( (r , l )  (0, 1) ), the
eigenvalue problem becomes:

 P0    A 
 0  A1   A1    c     P0  
 
P1     A1   A0   A2    c  T    
P1

(46)

 P  
 1 (61)    A1 

 A2   A0   (66)  c  T     P  
 1  ( 61)
(62)

MANU-19-1385 Altintas p.20

 Figure 9 Comparison of zero-order and multi-frequency with three harmonics stability
lobe solutions. The cutter has three teeth with a diameter of 23.6 mm used in down
milling of Al6061 material ( Kt =500 MPa, Kr =0.3) with a feed rate of 0.120
mm/rev/tooth and a radial depth of 1.256 mm. The structure is assumed to be flexible in y
(normal) direction only with k y =1.4×106 N/m, n , y =907 Hz,  y =0.017.

which leads to six eigenvalues and the eigenvalue that gives the minimum depth of cut is selected
as a solution. Since the spindle speed is needed to find the tooth passing frequency (T ) , the
solution is obtained at each given spindle speed. Typically, even with the most severe intermittent
conditions, including two to three harmonics is sufficient for stability convergence. The multi-
frequency method is able to predict the added stability pockets (i.e., period-doubling lobes) at very
high spindle speeds which are beyond the natural frequency of the structure; see Figure 9 [42]. An
efficient elimination of false eigenvalues, which improves the computational speed, was developed
by Merdol [33, 47].

3.2 Semi-discrete time-domain stability of milling operations

In the case of industrially realistic milling models, the milling tool has N  2 cutting edges and
the radial depth of cut is not negligible as it was assumed among the approximations of highly
interrupted cutting in Section 2.3. Consequently, the analytical calculation of the lobe structure is
no longer feasible. Even if a single vibration mode is considered in the x direction only as it was
for the highly interrupted cutting model, the structure of the governing equation is:

MANU-19-1385 Altintas p.21

 K t A(t )
x(t )  2n x(t )  n2 x(t )  n2 a  x(t  T )  x(t )  , (47)
kx

where the time delay is the tooth passing period T  2 /( N) , and the directional factor
A(t )  A(t  T ) is time-periodic with the same time period T as the delay (for details, see Section
3.1).

Figure 10 Discretization of the time-periodic function A(t) by piece-wise constant step-

function (a), and the discretization of the constant time delay T by time-periodic delay (b).

In this case, full discretization may be applied, where the time-periodic coefficients and the time
derivatives are all discretized with discrete time intervals. This brute force method is successful,
but, as explained in Section 2.2, the construction of a stability chart takes several hours on a
standard personal computer. However, the combination of the continuous and discrete-time
domain methods explained in Sections 2.2 and 2.3, respectively, results in a numerical method
where the time derivatives are not discretized, but only the time periodicity and the time delay are.
The basic idea of the discretization of the delay is not trivial. However, a discrete-time mapping
can be constructed similar to the quotient matrix A of highly interrupted cutting, but with a much
larger size. The elements of the matrix can be calculated for each time step in closed form;
analytical solutions of linear non-homogeneous ordinary differential equations can be used; and
the computation time of the stability analysis can be reduced radically. Then, the stability
boundaries can be reconstructed in the same way as in the case of Eq.(33). Although the size of
the quotient matrix A is large, there are several advanced and efficient numerical methods and
related routines to check whether all eigenvalues of a large matrix have moduli less than 1.
The basic semi-discretization is now introduced. While it is straightforward to approximate the
time-periodic function A(t ) in Eq.(47) by the piecewise constant function:

A(t )  A(it ) for t  [it , (i  1)it ), i  0,1,..., (n  1) , (48)

where the discrete time step is t  T / n ; see Figure 10a. It is more completed with respect to
the time delay to obtain an approximate system in an analytically manageable system of linear

MANU-19-1385 Altintas p.22

ordinary differential equations. The corresponding discretization of the time delay is represented
graphically in Figure 10b, where a specific time-periodic delay:
 (t )  t  (n  int(t / t ))t (49)

is defined that refers back to the same time instant in the past for each time step:
x(t  T )  x(t   (t ))  x(t  (t  (n  int(t / t ))t ))  x( j t  T )  x(( j  n)t )
(50)
t  [ j t , ( j  1)it ), j  0,1, 2,...

The two kinds of discretization can be carried out with the same approximation number n (number
of time steps) since the delay and the time period are the same. While the time-periodicity of the
time delay seems to be a further complication, it makes the approximate system simpler since it
refers back to the same past value within one time step. This way, the time-periodic delay-
differential equation provided in Eq. (47) can be approximated by the linear non-homogeneous
ordinary differential equation:

 aK t   aK 
x(t )  2n x(t )  n2 1  A( j t )  x(t )  n2  t A( j t )  x(( j  n) t ) (51)
 kx   kx 

for each time interval t  [ jt , ( j  1)it ), j  0,1,2,... . Although the average time delay in the
approximate system is somewhat larger than the exact delay T , the approximation is convergent
by decreasing the size of the time step t , that is, by increasing the approximation number n (see
[48] for details). Since the right hand side is piece-wise constant, the closed form solution of this
approximate system can be obtained similarly to Eq.(30) for the case of highly interrupted cutting.
This time, the discrete map expressed in Eq.(33) will have a much larger size. While the piece-
wise constant approximation of the periodic function A(t ) does not increase the size of this matrix,
the intermittent delayed state values do, since not just x (t ) and x(t  T ) , but all the discrete states
x(( j  i )t ) , ( i  0,1,..., n ) and their time derivatives have to be used as state variables at the j th
time instant t j  jt , j  0,1,2,... . This way, the size of the quotient matrix will be at least
2(n  1)  2(n  1) , and its eigenvalues must be checked with appropriate numerical method to have
absolute values less than 1.
A typical stability chart is presented in Figure 11, where the effect of the time periodicity is not as
strong as in the case of highly interrupted cutting because the milling process with higher number
of teeth (N = 4), while the radial immersion is still far from full immersion. In this case, the
presence of the period-doubling instabilities is not so characteristic and is less relevant. Moreover,
as it was proven by Szalai and Stepan [49], these are not lobes anymore, but rather unstable islands
(or unstable pockets/lenses), which are mostly surrounded by the classical chatter instability lobes.
These unstable islands can still be separated and shifted to the stable domains as fully isolated
islands as shown experimentally by Zatarain et al.[50], which discusses also further complications
when modeling helical cutting edge tools.

MANU-19-1385 Altintas p.23

 Figure 11 Stability chart of the milling process. The period-doubling lobes become unstable
islands. Feed per tooth is 0.1 mm, radial immersion is 0.02, a number of cutting edges is
N  4 , the cutting coefficients are K t  1000 MPa and Kr  0.3 , the damping ratio is
  0.05 .

3.3 Milling bifurcations

As described in Sections 3.1-3.3, in the surface regeneration process during milling, a time delay
appears in the system of second-order differential equations of motion that describe the dynamic
behavior. Specifically, the instantaneous chip thickness, which scales the cutting force, depends
not only on the commanded value and current relative vibration between the workpiece and
endmill tooth creating the new surface but also on the relative vibration one tooth period earlier
(i.e., the surface left by the previous tooth). Due to this time delay, various bifurcations (i.e., the
appearance of a qualitatively different solution as a control parameter is varied) are exhibited.
These bifurcations include: 1) secondary Hopf instability (classic chatter) which depends on the
averaged directional factors; and 2) period-n motions which consider the time-periodic directional
factors which have high harmonics at severely intermittent cutting operations.

MANU-19-1385 Altintas p.24

Drawing from techniques in nonlinear dynamics, analysis tools have been implemented to study
bifurcations in milling [51]. These include the phase-space Poincaré map, where the dynamic
trajectory of the tool (or workpiece) is presented graphically as the displacement versus velocity
and then sampled at the forcing frequency, or tooth period for milling. The character of the once-
per-tooth period samples then describes milling behavior. For stable cuts (forced vibration), the
motion is periodic with the tooth period, so the sampled points repeat and a single grouping of
points is observed. When secondary Hopf instability occurs, the motion is quasi-periodic with tool
rotation because the chatter frequency is (generally) incommensurate with the tooth passing
frequency. In this case, the once-per-tooth sampled points do not repeat and they form an elliptical
distribution. For a period-2 bifurcation (or period-doubling chatter with added lobes as described
in Sections 3.2 and 3.3), the motion repeats only once every other cycle (i.e., it is a sub-harmonic

Figure 12 Top) Time domain results for a period-2 bifurcation, i.e. chatter with added lobes.
(Bottom) Poincaré map for period-2 bifurcation. The sampled points align at a two fixed
locations for the period-2 bifurcation.
.
MANU-19-1385 Altintas p.25
of the forcing frequency). In this case, the once-per-tooth sampled points alternate between two
solutions. For period-n bifurcations, the sampled points appear at n distinct locations in the
Poincaré map.
An example period-2 bifurcation, the chatter with added lobes, is depicted in Figure 12. In the top
panel, the time domain displacement and velocity are shown individually. The inset provides a
magnified view; it is observed that the sampled points repeat every other tooth period, rather than
every period. The bottom panel shows the Poincaré map, where the sampled points appear at two
distinct locations.
A second analysis tool is the bifurcation diagram. Here, the independent variable, such as axial
depth of cut, is plotted on the horizontal axis against the once-per-tooth sampled displacement on
the vertical axis. The transition in stability behavior from stable (at low axial depths) to period-n
or secondary Hopf instability (at higher axial depths) is then directly observed. This diagram
represents the information from multiple Poincaré maps over a range of, for example, axial depths,
all at a single spindle speed. A stable cut appears as a single point (i.e., the sampled points repeat
when only forced vibration is present). A period-2 bifurcation, on the other hand, appears as a pair
of points offset from each other in the vertical direction. This represents the two collections of
once-per-tooth sampled points from the Poincaré map. A secondary Hopf bifurcation is seen as
vertical distribution of points; this represents the range of once-per-tooth sampled displacements
from the elliptical distribution of points in the Poincaré map.

Figure 13 Example bifurcation diagram. Stable behavior is observed up to an axial depth (b)
of approximately 2.5 mm. It is indicated by the single value of the sampled feed direction (x)
tool displacements. Between 2.5 mm and 4 mm, secondary Hopf chatter with added lobes
occurs (i.e., a distribution of sampled points). Between 4 mm and 5.5 mm, a period-3
bifurcation is seen (i.e., three distinct sampled points).

MANU-19-1385 Altintas p.26

An example bifurcation diagram is provided in Figure 13 [52]. The selected axial depth is listed
on the horizontal axis, while the once-per-tooth sampled points for the tool displacement are
presented on the vertical axis. All results are for a single spindle speed. It is demonstrated that
changing the system gain (axial depth) for a fixed time delay (spindle speed) can transition the
behavior from forced vibration (up to 2.5 mm) to secondary Hopf bifurcation chatter with added
lobes (between 2.5 mm and 4 mm) to period-3 bifurcation (between 4 mm and 5.5 mm).
In the literature, Davies et al. used once per revolution sampling to characterize the synchronicity
of cutting tool motions with the tool rotation and first measured period-3 tool motion (i.e., motion
that repeated with a period of three cutter revolutions) during partial radial immersion milling [53].
They followed with an analytical map that predicted a doubling of the number of optimally stable
spindle speeds when the time in cut is small.
As noted in Sections 3.1-3.3, follow-on modeling efforts included time finite element analysis [44,
45], semi-discretization [54, 55], and the multi-frequency method [22, 23, 42], which were used
to produce milling stability charts that predicted both stable and bifurcation behavior. Time
marching simulation has also been implemented to study milling bifurcations [56-58]. To aid in
the analysis of the simulation results, a new metric was described that automatically differentiates
between stable and unstable behavior of different types for time-domain simulation of the milling
processes [59, 60]. The approach was based on periodic sampling of milling signals at once per
tooth period (harmonic sampling) and integer multiples of the tooth period (subharmonic
sampling). The geometric accuracy of parts machined under both stable and period-2 bifurcations
was also predicted by time-domain simulation and verified experimentally [61].

3.4 Chatter stability of multi-point tools with lateral, torsional, and axial flexibilities

Any flexible direction that affects the regenerative chip thickness must be considered in the
equation of motion with time delay. For example, twist drills have lateral ( x, y ) and torsional ( )
flexibilities that create the axial ( z ) vibrations which change the regenerative chip thickness as
shown in Figure 14. The dynamics of a twist drill with such flexibilities may be expressed as:
Mq(t )  Cq(t )  Kq(t )  F(q(t ), q(t  T )) , (52)

MANU-19-1385 Altintas p.27

 Figure 14 Dynamic flexibilities of a twist drill and milling cutter with circular
inserts.

where the vibration vector q(t )   x(t ) y (t ) z (t )  (t ) has lateral (𝑥, 𝑦), axial (𝑧) , and

 
torsional ( ) vibrations. The force vector is F(t )  Fx (t ) Fy (t ) Fz (t ) Tc (t ) where (Tc ) is
the cutting torque. The FRF in Eq. (44) becomes [62]:

  xx  xy 0 0 
  yy 0 0 
Φ( )   yx , (53)
 0 0  zz  z 
 
 0 0  z  

where  is the torsional FRF excited by the drilling torque and  z is the FRF contributed by the
torsional-axial coupling of vibrations. The eigenvalue problem becomes four-dimensional in Eq.
(44), but the frequency and semi-discrete time-domain solution methods remain the same. Boring
heads and plunge mills have the same dynamics as twist drills [63]. The same argument is valid
for workpieces where there may be cross-coupling of the structure in different directions. Milling

MANU-19-1385 Altintas p.28

tools with ball ends or circular inserts also have the same stability model but present two and three-
dimensional eigenvalue problems since regeneration may take place both in the lateral and axial
directions due to its flexibility [64-66], but no torsional-axial coupling is considered. The lead and
tilt angles of the tools in five-axis ball-end milling of curved surfaces can be optimized to increase
the stability [67, 68].

4. Suppression of chatter with non-uniform tools

The chatter stability limit can be increased using tool holders with tuned damper mechanisms [69,
70], active damping using actuators [17], or special tool geometries which create stability pockets
at the desired speeds by proper selection of the tooth spacing angles, variable helix, or serrated
cutting edges. Only the tool design methods are summarized here.
The main principle behind the variable pitch, or unequally spaced, teeth on milling tools is to alter
the delay which is responsible for the regeneration mechanism. The effectiveness of variable pitch
cutters in suppressing chatter vibrations in milling was first demonstrated by Slavicek [71]. He
assumed a rectilinear tool motion for the cutting teeth and applied the orthogonal stability theories
of Tlusty and Tobias for the irregular tooth pitch case. By assuming an alternating pitch variation,
he obtained an expression for the stability limit which was a function of the pitch variation.
Successive cutting teeth were separated by a distance lj which was different for each interval as
shown in Figure 15.

Figure 15 Rectilinear representation of variable pitch milling tool proposed by Slavicek [71].

Slavicek [71] expressed the dynamic forces on N cutting teeth as:


Fj   Kt a r  re
i j
 for j  1, N , (54)

where Fj is the dynamic cutting force on tooth j, r is the vibration amplitude, and φ𝑗 is the phase
shift or delay between the waves, i.e., the inner and outer modulations. The phase shift is different
for each cutting tooth due to the unequal distance between each successive tooth:

MANU-19-1385 Altintas p.29

 cl j
j  , (55)
vc

where c is the chatter frequency and vc is the cutting speed. This simple relation shows that the
tooth spacing and the cutting speed affect the delay, hence the tooth spacing l j can be selected to
minimize the delay φ𝑗 to increase the stability.

Opitz et al. [72] considered milling tool rotation using average directional factors in analyzing
stability with irregular tooth pitch. They also considered alternating pitch with only two different
pitch angles. Average directional factors which relate the dynamic chip thickness to the vibrations
do not accurately represent the tool rotation which causes time-varying milling dynamics. A
comprehensive study on the stability of milling tools with non-constant pitch was carried out by
Vanherck [73] using computer simulations. His study demonstrated that a certain pitch variation
pattern was effective, i.e., increased the stability limits significantly for a given milling system
over a certain cutting speed range. This fundamental idea was later used in several works for
optimal design of variable pitch tools.
Tlusty et al. [74] presented the stability of milling cutters with special geometries such as irregular
pitch or serrated edges using numerical simulations. Their numerical simulations showed the
effectiveness of irregular milling tool geometries and cutting edges in increasing stability against
chatter. Later, Altintas et al. [75] analyzed the stability of variable pitch cutters accurately using
their analytical milling stability model [24]. The analytical model considers tool rotation, time-
varying dynamics, and multiple vibration modes, hence their experimentally verified predictions
were more accurate. They considered both linear and alternating pitch variations and showed that
each variation pattern had an effective zone where chatter stability limits were increased
substantially compared to standard milling tools. Olgac et al. used the same dynamic model of
variable pitch milling operations, but proposed a parametric stability analysis [76]. These studies
mainly concentrated on the effect of pitch variation on the stability limit; they did not directly
address the cutting tool design to determine the optimal pitch variation, although they can be used
to see the effectiveness of various tool designs. Budak [77] proposed an optimization methodology
for the design of variable pitch tools considering the chatter frequency and spindle speed. The
spacing variation amount is related to the chatter waves left on the surface in order to establish the
linear pitch angle variation as  p ,  p   ,  p  2 ,.. where  p is the base pitch angle and  is
the pitch angle increment between the successive teeth. Budak [76] showed that the eigenvalue
solution in the analytical chatter stability model takes the following form for variable pitch tools:

 
N
a
Kt  1  e c j ,
 i T
 (56)
4 j 1

where T j is the tooth period which is variable due to non-constant pitch angles and a is the depth
of cut. Budak obtained the following simple equation for the critically stable depth of cut in the
case of variable pitch milling tools:

MANU-19-1385 Altintas p.30

 4  I
vp
alim  (57)
Kt S

where  I is the imaginary part of the eigenvalue and

N
S   sin cTj or S  sin 1  sin  2  sin  3  ...
j 1

represents the summation of the phase delays caused by each tooth interval. This equation implied
that S must be minimized to maximize the chatter stability limit using variable pitch end mills. The
effect of the phase variation due to the variable pitch tool on the stability gain, which is defined as
the stability limit for the variable pitch tool over the one with the standard tool, was investigated
and it was shown that for particular values of the delay  , the stability limit was maximized as
illustrated in Figure 16.

10

6
r

4
1/4 1 13/2
2

0
0 0.25 0.5 0.75 1
 (rad) x2

Figure 16 Effect of  on stability gain for a four-tooth end mill with

linear pitch variation.

Budak [77] showed that this condition could be achieved if the variation amount was selected as
follows:

   for even N
c
(58)
 ( N  1)
   for odd N
c N

where  is the spindle speed in (rad/sec),  is the pitch variation (rad), c is the chatter
frequency (rad/sec), and N is the number of teeth. Note that the additional phase delay introduced
by the optimal pitch variation is set as integer divisions of the original delay of the milling system
in a standard tool. The pitch variation should cancel the “remainder wave” for maximized stability.

MANU-19-1385 Altintas p.31

It can be seen from Eq. (58)that the required pitch variation is proportional to the spindle speed.
As discussed in [78], large pitch variations required for high speeds cause non-uniform chip loads
on the cutting edges whereas small pitch variations in the low speed zone become sensitive to
slight changes in the chatter frequency. The formulation neglected the effect of pitch variation on
the chatter frequency by using a variable pitch tool; the chatter frequency was obtained
experimentally using a standard end mill.This issue was addressed by Comak et al. [79] where the
predicted effect of pitch variation on the chatter frequency was considered when selecting the
optimal pitch angles as shown in Figure 17. As this figure illustrates, the effect of the pitch angle
variation on the chatter frequency can be significant and should, therefore, be taken into account
in the selection of angles for variable pitch tools. Also, different pitch variations may yield similar
increases in the stability limits as shown in Figure 17, which shows that small pitch angle variation
can yield significant productivity gain using the proposed optimization approach.

MANU-19-1385 Altintas p.32

Figure 17. The effect of pitch angle variation on the stable depth of cut for different spindle speeds
and on the chatter frequency [77].
Also note that the closer the tooth passing frequency is to the chatter frequency, the larger the pitch
variation will be as evident from Eq.(58). As a result, variable pitch cutters become more practical
to use at low speeds where the tooth passing frequencies are several times lower than the chatter
frequency, such as in machining titanium and nickel alloys used in the aerospace industry.

MANU-19-1385 Altintas p.33

Alternatively, or in combination with variable pitch tools, end mills with varying helix angles have
also been introduced to vary pitch angle along the tool axis [80-82]. Serration on the end mill’s
cutting edges creates pitch angle variation between the teeth at each elevation, which also disturbs
regeneration [83, 84]. A sample comparison of chatter stability lobes with and without serration
is presented by Merdol et al. [83] as shown in Figure 18. Tehranizadeh et al. [85] optimized the
shapes of serration waves to reduce the cutting forces. Their results show that optimized tools have
better chatter stability performance due to the reduction of the effective axial depth of cut which
is possible at lower feed rates.

Figure 18 Chatter stability of regular and serrated end mills, after Merdol et al. [81].

5. Stability of parallel machining operations

Parallel, or simultaneous, machining has received considerable attention in the industry due to its
potential to increase the material removal rate (MRR). However, parallel machining can be
susceptible to chatter due to the dynamic interaction between the system components unless stable
process parameters are selected. The parallel turning and milling operations shown in Figure 19
are presented here as examples.

MANU-19-1385 Altintas p.34

Figure 19 (a) Parallel turning[85] and (b) parallel milling operations [92].

5.1 Parallel turning

In parallel turning, the cutting tools may dynamically interact through: (a) the shared cutting
surface, (b) the tool holder structure, and (c) the workpiece structure. Lazoglu et al. [86] developed
a time-domain model for chatter and stability of parallel turning where the tools cut different
surfaces and interact with each other due to the flexibility of the workpiece structure. A one-
dimensional frequency domain analysis of chatter in parallel turning operations was presented by
Budak et al. [87] where tools were mounted on different turrets and cut a shared surface with
different depth of cuts as shown in Figure 19a. The dynamic cutting forces have multiple time
delays:

     
 a1   z1  t   z 2  t    
 F1  t  
     2  ,
   Kr  (59)
 2  
F t      
  a1   z2  t   z1  t      a2  a1    z2  t   z 2  t     
   2  

where Kr is the cutting force coefficient in the feed direction, z1 and z2 are the displacements of the
first and second tool in the feed direction, respectively; and τ is the workpiece rotation period. The
depth of cuts for the first and second tools are represented by a1 and a2, (a2 > a1), respectively
(Figure 20a). In this model, two regions on the shared surface can exist since the tools’ depths of
cut can be different. In the first region, both tools remove the same depth of cut a1. Therefore, the
displacement of each tool at any instant is influenced by the other tool’s displacement at the half
rotation period of the workpiece before. However, in the second region, the depth (a2-a1) is
removed by the second tool solely. Thus, the displacement of the second tool at time t is affected
by its displacement at the previous rotation period of the workpiece. The eigenvalue problem for
the marginally-stable case is formulated in the frequency domain and solved numerically. The
solution is provided for the first tool’s depth of cut over a range of spindle speed values while the

MANU-19-1385 Altintas p.35

second tool’s depth of cut is initially selected. Figure 20b shows the stability lobe diagram for the
first tool where a2 = 1.5 mm. In this case, when a1 = 0 mm (point d) chatter occurs. By increasing
a1 to 1 mm (point e), the productivity is increased and the process is stabilized. By further
increasing the a1 value to 1.5 mm (point f), chatter again occurs. Simulation results were validated
with cutting tests in [87].
Ozturk et al. [88] presented the chatter stability model for parallel turning operation with two
different tool configurations. In the first configuration, the tools were mounted on a turret. In this
case, the dynamic displacements of the tools do not interact via the shared surface but rather by
the turrets. The results showed that by employing the second tool, the stability limit of the first
tool slightly decreases compared to single tool turning. However, the overall material removal rate
of the parallel process is almost doubled due to having a second tool. For the second configuration,
cutting a shared surface (similar to [87]), they showed that tools with identical natural frequencies
have the lowest stability limit. Similar behavior was reported by Reith et al. [89] who included the
effect of the tool holder using a non-proportional damping model. Brecher et al. [90] studied the
effect of the radial orientation of the tools when they cut a shared surface with identical depths of
cut (see Figure 20b). It was demonstrated that by properly setting the tool’s orientation, the stability
boundaries can be shifted upwards. For highly flexible workpieces, it is essential to model the true
geometry of the insert due to the dominance of the cutting forces in the radial direction as
demonstrated by Azvar et al. [91]. They concluded that when cutting a shared surface of a flexible
workpiece, tools with identical insert geometry demonstrate a higher stability limit. Also, for the
case of the parallel turning of a flexible workpiece from different surfaces, the tool which is closer
to the free end of the workpiece should have a smaller nose radius and side edge cutting angle to
obtain higher stability.

Figure 20 (a) parallel turning configuration, (b) stability lobe diagram for a1 versus spindle
speed (n) when a2= 1.5 mm and (c) time domain simulation for different points f (unstable), e
(stable) and d (unstable).(From [85])

MANU-19-1385 Altintas p.36

 5.2 Parallel milling
Two milling tools cut the workpiece simultaneously in parallel milling operations where the
stability limit can be increased due to the cancellation of the dynamic cutting forces. Shamoto et
al. [92] introduced a technique to eliminate the chatter vibrations in simultaneous face milling of
flexible parts by imposing a phase shift resulting from different rotational speeds of milling tools.
Budak et al. [93] developed the frequency domain stability model for parallel milling systems. The
dynamic forces are evaluated analytically to form the characteristic force function:

 Fx1   Fx1 
F  F 
 y1  ict 1  y1 
 e   CPM    DM   OTF     eict , (60)
 Fx 2  4  Fx 2 
 Fy 2   Fy 2 

where [CPM] is the cutting parameters matrix, [DM] is the delay matrix which includes the phase
delays for the milling tools, and [OTF] is the oriented transfer function matrix. Equation (60)
results in an eigenvalue problem that provides the experimentally proven stable depths for different
spindle speeds; see Figure 21.

Figure 21 Predicted stability lobe diagram and the experimental results.

6. Current Challenges in machining dynamics

Chatter is still the main obstacle to achieve precision and productivity for machining and grinding
operations. Chatter can also be found in other operations, such as rolling [94] and sawing [95]
operations. The prediction of chatter-free cutting conditions primarily depends on the
identification of machine’s structural dynamics, cutting force and process damping coefficients,

MANU-19-1385 Altintas p.37

and tool-workpiece engagement conditions. Each subject has its own nonlinearities and
uncertainties.
The structural dynamics (FRF) as reflected at the tool point (i.e., the tool point receptance) is highly
critical in predicting the stability charts. While modal analysis may be applied to measure
receptances, this can pose a significant obstacle when equipment or expertise is not available in
the production facility. Schmitz and Donaldson first presented the receptance coupling
substructure analysis (RCSA) approach to predict tool point receptances by joining models and
measurements of the tool, holder, spindle, and machine through appropriate connection parameters
[96]. Follow on work by Schmitz et al. includes a three-component model [97], at-speed
predictions [98], and improved spindle receptance identification [98]. Others have considered the
flute geometry [99], the asymmetric dynamics of rotating tools [100], and the elastic coupling of
tool and holders with the spindle [101]. Furthermore, the FRF of the machine may change as a
function of speed [102], cutting loads, and the position of the machine during machining [103,
104]. Additionally, machining with robots or machines with parallel kinematic configurations have
pose-dependent dynamics; therefore, their stability varies along the tool path [105, 106]. As the
material is removed during machining, parts with thin-walled structures have continuously varying
dynamics which need to be updated along the tool path [107]. Another challenge is the use of
multiple spindles to cut a single part simultaneously as presented in Section 5. While one spindle
may mill the part, another spindle may carry out grinding, turning, or drilling operation on the
same part when using multi-functional machine tools [108]. The cross-talk between the spindles
and multiple delays introduce challenges when modeling the process dynamics and its stability,
such as in the case of mill-turn operations [109]. There are some heavy duty machining
applications at low speeds that excite the low frequency modes of the large components of the
machine tools. These low frequency modes shift as the position of the machine vary. Iglesias et al.
[110] proposed position and feed direction dependent stability charts in planning heavy duty
cutting of steel alloys within the operating envelope of large machine tools.
Cutting tools have rounded and chamfered cutting edges that are not consistently manufactured.
The force required to shear away the chip is split into the ploughing and cutting components,
depending on the relative sizes of the cutting edge radius and the chip [111]. Furthermore, the wear
changes the contact between the wavy cut surface and the tool’s flank face that contributes to the
process damping [32]. As a result, the cutting force and process damping coefficients have
significant uncertainties. Since the cutting force coefficient acts as a gain and the process damping
coefficient improves the stability, uncertainties in these parameters affects the accuracy of chatter
predictions. For example, the process damping is challenging to model in twist drills where the
cutting speed starts from zero at the chisel tip and increases towards the periphery [112].
When the cutter shape is irregular as in form tools, serrated end mills, and multi-functional tools
which may include drilling, boring, and chamfering in a single operation, the cutter-part
engagement conditions may change along the tool axis [113, 114]. As a result, the directional
coefficient matrix, as well as the process dynamics, may vary significantly as in the case of
threading [115] and gear machining with multiple teeth [116]. Efficient modeling and the stability
solution for such tools and operations are still needed for high-performance machining. A recent

MANU-19-1385 Altintas p.38

metrology solution is structured light scanning, which can be used to measure the actual cutting
edge geometry [117].

7. Conclusion
This paper presents the fundamentals of chatter stability laws developed in the frequency and
discrete-time domains. It is shown that the dynamics of the machining process need to be modeled
by considering the interaction between the machining process and the structural dynamics of both
the machine and the machined part. The dynamics can be stationary with time-invariant
coefficients, such as in turning, or time-periodic with time-varying delays as in the case of turn-
milling operations. Although the understanding of stability for dynamic machining has increased
significantly during last six decades, the accuracy of stability predictions still suffers due to
measurement uncertainties, nonlinearities in the machine structure and process, time-varying
dynamics of machine tools and parts, and ploughing-based material removal by a worn tool or
chamfered cutting edge. The integration of physics-based off-line chatter prediction models needs
to be combined with on-line, adaptive learning and tuning techniques for robust chatter detection
and avoidance. Complex processes such as gear shaping with form tools, the stability of machining
composite metals, and additively manufactured parts with non-uniform material properties remain
to be studied.
Acknowledgments: The authors acknowledge the Natural Sciences and Engineering Council of
Canada Grants (IRCPJ 260683-18 and CANRIMT NETGP 479639-15); USA National Science
Foundation Grant (CMMI-1561221); TUBITAK-Turkey Grant Numbers: 105M032, 108M340,
110M522, 217M210; Hungarian National Research, Development and Innovation Fund
(TUDFO/51757/2019-ITM Thematic Excellence Program; and the Research Excellence Program
under Grant no. KKP133846.
References
[1] Taylor, F. W., 1906, "On the art of cutting metals," Transactions, The American Society of
Mechanical Engineers, 28, pp. 70-350.
[2] Arnold, R. N., 1946, "Cutting tools research: report of subcommittee on carbide tools: the
mechanism of tool vibration in the cutting of steel," P I Mech Eng B-J Eng, 154/1, pp. 261-284.
[3] Doi, S., and Kato, S., 1956, "Chatter vibration of lathe tools," Trans. ASME 78, p. 1127.
[4] Tobias, S., and Fishwick, W., 1958, "Theory of regenerative machine tool chatter," The
engineer, 205/7, pp. 199-203.
[5] Tlusty, J., and Polacek, M., 1963, "The stability of the machine tool against self excited
vibration in machining.," Proceedings of the ASME International Research in Production
Engineering.Pittsburgh, pp. 465-474.
[6] Tlusty, J., 1978, "Analysis of the state of research in cutting dynamics," Annals of the CIRP,
27/2, pp. 583-589.
[7] Tlusty, J., 1986, "Dynamics of High-Speed Milling," ASME Journal of Engineering for
Industry, 108(2), pp. 59-67.
[8] Altintas, Y., and Weck, M., 2004, "Chatter stability of metal cutting and grinding," Cirp
Annals-Manufacturing Technology, 53(2), pp. 619-642.
[9] Wiercigroch, M., and Budak, E., 2001, "Sources of nonlinearities, chatter generation and
suppression in metal cutting," Philos T R Soc A, 359(1781), pp. 663-693.

MANU-19-1385 Altintas p.39

[10] Quintana, G., and Ciurana, J., 2011, "Chatter in machining processes: A review,"
International Journal of Machine Tools and Manufacture, 51/5, pp. 363-376.
[11] Altintas, Y., Stepan, G., Merdol, D., and Dombovari, Z., 2008, "Chatter stability of milling
in frequency and discrete time domain," CIRP Journal of Manufacturing Science and
Technology, 1/1, pp. 35-44.
[12] Merritt, H. E., 1965, "Theory of Self-Excited Machine-Tool Chatter: Contribution to
Machine-Tool Chatter Research—," Journal of Engineering for Industry, 87/4, pp. 447-454.
[13] Koenigsberger, F., and Tlusty, J., 1970, Structures of machine tools, Pergamon Press,
Oxford.
[14] Smith, S., and Tlusty, J., 1997, "Current trends in high-speed machining," ASME Journal of
Manufacturing Science and Engineering, 119(4B), pp. 664-666.
[15] Abele, E., Altintas, Y., and Brecher, C., 2010, "Machine tool spindle units," Cirp Annals-
Manufacturing Technology, 59/2, pp. 781-802.
[16] Sweeney, G., and Tobias, S., 1969, "Survey of basic machine tool chatter research,"
International Journal of Machine Tool Design, 9/3, pp. 217-238.
[17] Munoa, J., Beudaert, X., Dombovari, Z., Altintas, Y., Budak, E., Brecher, C., and Stepan,
G., 2016, "Chatter suppression techniques in metal cutting," Cirp Annals-Manufacturing
Technology, 65(2), pp. 785-808.
[18] Tunc, L. T., Mohammadi, Y., and Budak, E., 2018, "Destabilizing effect of low frequency
modes on process damped stability of multi-mode milling systems," Mechanical Systems and
Signal Processing, 111, pp. 423-441.
[19] Sridhar, R., Hohn, R. E., and Long, G. W., 1968, "A Stability Algorithm for the General
Milling Process: Contribution to Machine Tool Chatter Research," ASME Journal of
Engineering for Industry, 90/2, pp. 330-334.
[20] Minis, I., Yanushevsky, R., Tembo, A., and Hocken, R., 1990, "Analysis of Linear and
Nonlinear Chatter in Milling," CIRP Annals, 39/1, pp. 459-462.
[21] Minis, I., and Yanushevsky, R., 1993, "A New Theoretical Approach for the Prediction of
Machine Tool Chatter in Milling," Journal of Engineering for Industry, 115/1, pp. 1-8.
[22] Budak, E., and Altintas, Y., 1998, "Analytical prediction of chatter stability in milling - Part
1: General formulation," Transcation of ASME, Journal of Dynamic Systems Measurement and
Control, 120/1, pp. 22-30.
[23] Budak, E., and Altintas, Y., 1998, "Analytical prediction of chatter stability in milling - Part
II: Application of the general formulation to common milling systems," Transactions of the
ASME, Journal of Dynamic Systems Measurement and Control, 120/1, pp. 31-36.
[24] Altintas, Y., and Budak, E., 1995, "Analytical Prediction of Stability Lobes in Milling,"
CIRP Annals, 44/1, pp. 357-362.
[25] Stépán, G., 1998, "Delay-differential equation models for machine tool chatter," Dynamics
and chaos in manufacturing processes, Wiley, New York, pp. 165-192.
[26] Insperger, T., and Stépán, G., 2000, "Stability of the milling process," Periodica
Polytechnica, Mechanical Engineering, 44/1, pp. 47-57.
[27] Merchant, M. E., 1945, "Mechanics of the metal cutting process. I. Orthogonal cutting and a
type 2 chip," Journal of applied physics, 16(5), pp. 267-275.
[28] Budak, E., Altintas, Y., and Armarego, E. J. A., 1996, "Prediction of milling force
coefficients from orthogonal cutting data," Transactions of ASME Journal of Manufacturing
Science and Engineering, 118/2, pp. 216-224.

MANU-19-1385 Altintas p.40

[29] Ehmann, K., Kapoor, S., DeVor, R., and Lazoglu, I., 1997, "Machining process modeling: a
review," ASME Journal of Manufacturing Science, 119(4B), pp. 655-663.
[30] Ismail, F., and Vadari, V., 1990, "Machining chatter of end mills with unequal modes,"
ASME Journal of Engineering for Industry, 112(3), pp. 229-235.
[31] Das, M. K., Tobias, S.A., , 1967, "The relation between the static and the dynamic cutting of
metals," International Journal of Machine Tool Design and Research, 7(2), pp. 63-68.
[32] Eynian, M., and Altintas, Y., 2009, "Chatter Stability of General Turning Operations With
Process Damping," Transactions of ASME Journal of Manufacturing Science and Engineering,
131(4), p. 041005.
[33] Altintas, Y., 2012, Manufacturing Automation: Metal Cutting Mechanics, Machine Tool
Vibrations, and CNC Design, Cambridge University Press, Cambridge.
[34] Tyler, C. T., and Schmitz, T. L., 2013, "Analytical process damping stability prediction,"
Journal of Manufacturing Processes, 15(1), pp. 69-76.
[35] Budak, E., and Tunc, L. T., 2010, "Identification and modeling of process damping in
turning and milling using a new approach," CIRP Annals, 59(1), pp. 403-408.
[36] Altintas, Y., Eynian, M., and Onozuka, H., 2008, "Identification of dynamic cutting force
coefficients and chatter stability with process damping," Cirp Annals-Manufacturing
Technology, 57(1), pp. 371-374.
[37] Ahmadi, K., and Altintas, Y., 2014, "Identification of Machining Process Damping Using
Output-Only Modal Analysis," Journal of Manufacturing Science and Engineering-Transactions
of the Asme, 136(5).
[38] Tuysuz, O., and Altintas, Y., 2019, "Analytical Modeling of Process Damping in
Machining," Trans. ASME Journal of Manufacturing Science and Engineering, pp. 1-52.
[39] Dombovari, Z., Iglesias, A., Zatarain, M., and Insperger, T., 2011, "Prediction of multiple
dominant chatter frequencies in milling processes," International Journal of Machine Tools and
Manufacture, 51(6), pp. 457-464.
[40] Davies, M. A., Balachandran, B., 2000, "Impact Dynamics in the Milling of Thin-Walled
Structures," Nonlinear Dynamics, 22(4), p. 17.
[41] Davies, M. A., Pratt, J. R., Dutterer, B., and Burns, T. J., 2002, "Stability Prediction for Low
Radial Immersion Milling," Journal of Manufacturing Science and Engineering, 124(2), pp. 217-
225.
[42] Merdol, S. D., and Altintas, Y., 2004, "Multi Frequency Solution of Chatter Stability for
Low Immersion Milling," Journal of Manufacturing Science and Engineering, 126(3), pp. 459-
466.
[43] Bayly, P. V., Halley, J. E., Mann, B. P., and Davies, M. A., 2003, "Stability of Interrupted
Cutting by Temporal Finite Element Analysis," Journal of Manufacturing Science and
Engineering, 125(2), pp. 220-225.
[44] Insperger, T., Mann, B. P., Stépán, G., and Bayly, P. V., 2003, "Stability of up-milling and
down-milling, part 1: alternative analytical methods," International Journal of Machine Tools
and Manufacture, 43(1), pp. 25-34.
[45] Mann, B. P., Insperger, T., Bayly, P. V., and Stépán, G., 2003, "Stability of up-milling and
down-milling, part 2: experimental verification," International Journal of Machine Tools and
Manufacture, 43(1), pp. 35-40.
[46] Gradišek, J., Kalveram, M., Insperger, T., Weinert, K., Stépán, G., Govekar, E., and Grabec,
I., 2005, "On stability prediction for milling," International Journal of Machine Tools and
Manufacture, 45(7), pp. 769-781.

MANU-19-1385 Altintas p.41

[47] Merdol, D., 2008, "Virtual three-axis milling process simulation and optimization," Ph.D. ,
University of British Columbia.
[48] Insperger, T., Stepan, G., 2011, Semi-Discretization for Time-Delay Systems, Springer,
New York.
[49] Szalai, R., Stepan, G., 2006, "Lobes and lenses in the stability chart of interrupted turning.
Journal of Computational and Nonlinear Dynamics," Transactions of ASME, Journal of
Computational and Nonlinear Dynamics, 1, p. 6.
[50] Zatarain, M., Muñoa, J., Peigné, G., and Insperger, T., 2006, "Analysis of the Influence of
Mill Helix Angle on Chatter Stability," CIRP Annals, 55(1), pp. 365-368.
[51] Moon FC, K.-N. T., 2001, "Nonlinear models for complex dynamics in cutting materials.,"
Philosophical Transactions of the Royal Society London A(359), pp. 695-711.
[52] Honeycutt, A., and Schmitz, T. L., 2018, "Milling Bifurcations: A Review of Literature and
Experiment," Journal of Manufacturing Science and Engineering, 140(12), pp. 120801-120801-
120819.
[53] Davies, M. A., Dutterer, B., Pratt, J. R., Schaut, A. J., and Bryan, J. B., 1998, "On the
Dynamics of High-Speed Milling with Long, Slender Endmills," CIRP Annals, 47(1), pp. 55-60.
[54] Insperger, T., and Stépán, G., 2004, "Vibration Frequencies in High-Speed Milling
Processes or a Positive Answer to Davies, Pratt, Dutterer and Burns," Journal of Manufacturing
Science and Engineering, 126(3), pp. 481-487.
[55] Insperger, T., and Stépán, G., 2004, "Updated semi-discretization method for periodic
delay-differential equations with discrete delay," International Journal for Numerical Methods in
Engineering, 61(1), pp. 117-141.
[56] Zhao, M. X., and Balachandran, B., 2001, "Dynamics and stability of milling process,"
International Journal of Solids and Structures, 38(10), pp. 2233-2248.
[57] Honeycutt, A., and Schmitz, T., 2016, "A Numerical and Experimental Investigation of
Period-n Bifurcations in Milling," Journal of Manufacturing Science and Engineering, 139(1),
pp. 011003-011003-011011.
[58] Honeycutt, A., and Schmitz, T. L., 2016, "Experimental Validation of Period-n Bifurcations
in Milling," Procedia Manufacturing, 5, pp. 362-374.
[59] Honeycutt, A., and Schmitz, T., 2016, "A new metric for automated stability identification
in time domain milling simulation," ASME Journal of Manufacturing Science and Engineering,,
138/7, p. 074501.
[60] Honeycutt, A., and Schmitz, T. L., 2016, "Milling Stability Interrogation by Subharmonic
Sampling," Journal of Manufacturing Science and Engineering, 139(4), pp. 041009-041009-
041009.
[61] Honeycutt, A., and Schmitz, T. L., 2017, "A Study of Milling Surface Quality during
Period-2 Bifurcations," Procedia Manufacturing, 10, pp. 183-193.
[62] Roukema, J. C., and 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 & Manufacture, 47(9), pp. 1455-1473.
[63] Altintas, Y., and Ko, J. H. J. C. A.-M. T., 2006, "Chatter stability of plunge milling," 55(1),
pp. 361-364.
[64] Altintas, Y., 2001, "Analytical Prediction of Three Dimensional Chatter Stability in
Milling," JSME International Journal Series C Mechanical Systems, Machine Elements and
Manufacturing, 44(3), pp. 717-723.

MANU-19-1385 Altintas p.42

[65] Altintas, Y., Shamoto, E., Lee, P., and Budak, E., 1999, "Analytical prediction of stability
lobes in ball end milling," Journal of Manufacturing Science and Engineering-Transactions of
the Asme, 121(4), pp. 586-592.
[66] Jensen, S. A., and Shin, Y. C., 1999, "Stability Analysis in Face Milling Operations, Part 1:
Theory of Stability Lobe Prediction," Journal of Manufacturing Science and Engineering,
121(4), pp. 600-605.
[67] Ozturk, E., Budak, E., and Engineering, 2010, "Dynamics and stability of five-axis ball-end
milling," Trans ASME Journal of Manufacturing Science and Engineering, 132/2, p. 021003.
[68] Chao, S., and Altintas, Y., 2016, "Chatter free tool orientations in 5-axis ball-end milling,"
International Journal of Machine Tools & Manufacture, 106, pp. 89-97.
[69] Sims, N. D., 2007, "Vibration absorbers for chatter suppression: A new analytical tuning
methodology," Journal of Sound and Vibration, 301(3-5), pp. 592-607.
[70] Yang, Y., Munoa, J., and Altintas, Y., 2010, "Optimization of multiple tuned mass dampers
to suppress machine tool chatter," International Journal of Machine Tools & Manufacture, 50(9),
pp. 834-842.
[71] Slavicek, J., 1965, "The effect of irregular tooth pitch on stability in milling," 6th Machine
Tool Design and Research Conference, pp.15-22.
[72] Opitz, H., "Improvement of the dynamic stability of the milling process by irregular tooth
pitch," Proc. Proc. 7th Int. MTDR Conference pp. 213-227.
[73] Vanherck, P., "Increasing Milling Machine Productivity by Use of Cutters with Non
Constant Cutting Edge Pitch," Proc. Proceedings of the 8th MTDR Conference, pp. 947-960.
[74] Tlusty, J., Ismail, F., and Zaton, W., "Use of special milling cutters against chatter," Proc.
NAMRC, pp. 408-415.
[75] Altintas, Y., Engin, S., and Budak, E., 1999, "Analytical stability prediction and design of
variable pitch cutters," Journal of Manufacturing Science and Engineering-Transactions of the
Asme, 121(2), pp. 173-178.
[76] Olgac, N., and Sipahi, R., 2007, "Dynamics and Stability of Variable-pitch Milling," Journal
of Vibration and Control, 13(7), pp. 1031-1043.
[77] Budak, E., 2003, "An Analytical Design Method for Milling Cutters With Nonconstant
Pitch to Increase Stability, Part I: Theory," Journal of Manufacturing Science and Engineering,
125(1), pp. 29-34.
[78] Iglesias, A., Dombovari, Z., Gonzalez, G., Munoa, J., and Stepan, G., 2019, "Optimum
selection of variable pitch for chatter suppression in face milling operations," Materials, 12(1), p.
112.
[79] Comak, A., and Budak, E., 2017, "Modeling dynamics and stability of variable pitch and
helix milling tools for development of a design method to maximize chatter stability," Precision
Engineering, 47, pp. 459-468.
[80] Sims, N. D., Mann, B., and Huyanan, S., 2008, "Analytical prediction of chatter stability for
variable pitch and variable helix milling tools," Journal of Sound and Vibration, 317(3), pp. 664-
686.
[81] Turner, S., Merdol, D., Altintas, Y., and Ridgway, K., 2007, "Modelling of the stability of
variable helix end mills," International Journal of Machine Tools and Manufacture, 47(9), pp.
1410-1416.
[82] Dombovari, Z., and Stépán, G., 2012, "The Effect of Helix Angle Variation on Milling
Stability," Trans. ASME Journal of Manufacturing Science and Engineering, 134/5, p. 051015.

MANU-19-1385 Altintas p.43

[83] Merdol, S. D., and Altintas, Y., 2004, "Mechanics and dynamics of serrated cylindrical and
tapered end mills," Journal of Manufacturing Science and Engineering-Transactions of the
Asme, 126(2), pp. 317-326.
[84] Dombovari, Z., Altintas, Y., and Stepan, G., 2010, "The effect of serration on mechanics
and stability of milling cutters," International Journal of Machine Tools & Manufacture, 50(6),
pp. 511-520.
[85] Tehranizadeh, F., Koca, R., and Budak, E., 2019, "Investigating effects of serration
geometry on milling forces and chatter stability for their optimal selection," International Journal
of Machine Tools Manufacture, p. 103425.
[86] Lazoglu, I., Vogler, M., Kapoor, S. G., and DeVor, R. E., 1998, "Dynamics of the
simultaneous turning process," Trans. Am. Res. Inst. SME, pp. 135-140.
[87] Budak, E., and Ozturk, E., 2011, "Dynamics and stability of parallel turning operations,"
CIRP Annals, 60(1), pp. 383-386.
[88] Ozturk, E., Comak, A., and Budak, E., 2016, "Tuning of tool dynamics for increased
stability of parallel (simultaneous) turning processes," Journal of Sound and Vibration, 360, pp.
17-30.
[89] Reith, M. J., Bachrathy, D., and Stepan, G., 2016, "Optimal Detuning of a Parallel Turning
System—Theory and Experiments," Trans ASME Journal of Dynamic Systems, Measurement,
and Control, 139/1(1), pp. 014503-014503-014507.
[90] Brecher, C., Epple, A., Neus, S., and Fey, M., 2015, "Optimal process parameters for
parallel turning operations on shared cutting surfaces," International Journal of Machine Tools
and Manufacture, 95, pp. 13-19.
[91] Azvar, M., and Budak, E., 2017, "Multi-dimensional chatter stability for enhanced
productivity in different parallel turning strategies," International Journal of Machine Tools and
Manufacture, 123, pp. 116-128.
[92] Shamoto, E., Mori, T., Sencer, B., Suzuki, N., and Hino, R., 2013, "Suppression of
regenerative chatter vibration in multiple milling utilizing speed difference method–Analysis of
double-sided milling and its generalization to multiple milling operations," Precision
Engineering, 37/3, pp. 580-589.
[93] Budak, E., Comak, A., and Ozturk, E., 2013, "Stability and high performance machining
conditions in simultaneous milling," CIRP Annals, 62/1, pp. 403-406.
[94] Yun, I. S., Wilson, W. R. D., and Ehmann, K. F., 1998, "Review of chatter studies in cold
rolling," International Journal of Machine Tools & Manufacture, 38(12), pp. 1499-1530.
[95] Thaler, T., Krese, B., and Govekar, E., 2015, "Stability diagrams and chatter avoidance in
horizontal band sawing," Cirp Annals-Manufacturing Technology, 64(1), pp. 81-84.
[96] Schmitz, T., and Donaldson, R., 2000, "Predicting High-Speed Machining Dynamics by
Substructure Analysis," Annals of the CIRP, 49/1, pp. 303-308.
[97] Schmitz, T., and Duncan, G. S., 2005, " Three-Component Receptance Coupling
Substructure Analysis for Tool Point Dynamics Prediction, ," Trans. ASME , Journal of
Manufacturing Science and Engineering, 127/4, pp. 781-790.
[98] Cheng, C.-H., Duncan, G.S., and Schmitz, T., , 2007, "Rotating Tool Point Frequency
Response Prediction using RCSA,," Machining Science and Technology, 11(3), pp. 433-446.
[99] Ozsahin, O., and Altintas, Y., 2015, "Prediction of frequency response function (FRF) of
asymmetric tools from the analytical coupling of spindle and beam models of holder and tool,"
International Journal of Machine Tools & Manufacture, 92, pp. 31-40.

MANU-19-1385 Altintas p.44

[100] Comak, A., Ozsahin, O., and Altintas, Y., 2016, "Stability of Milling Operations With
Asymmetric Cutter Dynamics in Rotating Coordinates," Journal of Manufacturing Science and
Engineering-Transactions of the Asme, 138(8).
[101] Namazi, M., Altintas, Y., Abe, T., and Rajapakse, N., 2007, "Modeling and identification
of tool holder-spindle interface dynamics," International Journal of Machine Tools &
Manufacture, 47(9), pp. 1333-1341.
[102] Postel, M., Özsahin, O., and Altintas, Y., 2018, "High speed tooltip FRF predictions of
arbitrary tool-holder combinations based on operational spindle identification," International
Journal of Machine Tools & Manufacture, 129, pp. 48-60.
[103] Ozsahin, O., Budak, E., and Ozguven, H. N., 2015, "In-process tool point FRF
identification under operational conditions using inverse stability solution," International Journal
of Machine Tools and Manufacture, 89, pp. 64-73.
[104] Law, M., Phani, A. S., and Altintas, Y., 2013, "Position-Dependent Multibody Dynamic
Modeling of Machine Tools Based on Improved Reduced Order Models," Trans. ASME Journal
of Manufacturing Science and Engineering, 135/2, p. 021008.
[105] Cordes, M., Hintze, W., and Altintas, Y., 2019, "Chatter stability in robotic milling,"
Robotics and Computer-Integrated Manufacturing, 55, pp. 11-18.
[106] Law, M., Ihlenfeldt, S., Wabner, M., Altintas, Y., and Neugebauer, R., 2013, "Position-
dependent dynamics and stability of serial-parallel kinematic machines," CIRP Annals-
Manufacturing Technology, 62/1, pp. 375-378.
[107] Tuysuz, O., and Altintas, Y., 2018, "Time-Domain Modeling of Varying Dynamic
Characteristics in Thin-Wall Machining Using Perturbation and Reduced-Order Substructuring
Methods," Trans. ASME Journal of Manufacturing Science, 140(1), p. 011015.
[108] Moriwaki, T., 2008, "Multi-functional machine tool," CIRP Annals, 57/2, pp. 736-749.
[109] Comak, A., and Altintas, Y., 2018, "Dynamics and Stability of Turn-Milling Operations
With Varying Time Delay in Discrete Time Domain," Transactions of the ASME, Journal of
Manufacturing Science and Engineering, 140/10.
[110] Iglesias, A., Munoa, J., and Ciurana, J., 2014, "Optimisation of face milling operations
with structural chatter using a stability model based process planning methodology," The
International Journal of Advanced Manufacturing Technology, 70(1-4), pp. 559-571.
[111] Jin, X., and Altintas, Y., 2013, "Chatter Stability Model of Micro-Milling With Process
Damping," Journal of Manufacturing Science and Engineering-Transactions of the Asme,
135(3).
[112] Ahmadi, K., and Altintas, Y., 2013, "Stability of lateral, torsional and axial vibrations in
drilling," International Journal of Machine Tools & Manufacture, 68, pp. 63-74.
[113] Wan, M., Kilic, Z. M., and Altintas, Y., 2015, "Mechanics and Dynamics of
Multifunctional Tools," Journal of Manufacturing Science and Engineering-Transactions of the
Asme, 137(1).
[114] Kilic, Z. M., and Altintas, Y., 2016, "Generalized mechanics and dynamics of metal
cutting operations for unified simulations," International Journal of Machine Tools &
Manufacture, 104, pp. 1-13.
[115] Khoshdarregi, M. R., and Altintas, Y., 2018, "Dynamics of Multipoint Thread Turning-
Part I: General Formulation," Journal of Manufacturing Science and Engineering-Transactions of
the Asme, 140(6).

MANU-19-1385 Altintas p.45

[116] Katz, A., Erkorkmaz, K., Ismail, F. J. J. o. M. S., and Engineering, 2018, "Virtual Model of
Gear Shaping—Part I: Kinematics, Cutter–Workpiece Engagement, and Cutting Forces," 140(7),
p. 071007.
[117] No, T., Gomez, M., Copenhaver, R., Uribe Perez, J., Tyler, C., and Schmitz, T., 2019,
"Force and Stability Modeling for Non-standard Edge Geometry Endmills," ASME Journal of
Manufacturing Science, 141(12), p. 121002.

MANU-19-1385 Altintas p.46

 FIGURE CAPTIONS
Figure 1 Dynamics of orthogonal cutting system. a) Regenerative mechanism. b) Process
damping mechanism
Figure 2 Block diagram of dynamic orthogonal cutting with process damping and regenerative
chip feedback.
Figure 3 Chatter stability diagrams and corresponding chatter frequency diagram for mode
coupling chatter, and regenerative chatter with and without process damping. Simulation
parameters: Cutting coefficients - Kt=1000 MPa, Kr=0.3; Modal parameters- k1=15×106 N/m,
k2=10×106N/m, , . The orientation of flexibilities- , ; process damping constant - Cp=2e6
N/m; workpiece diameter = 30 mm. Simulated cases: Cases- A: n=700 rev/min (stable, process
damping zone), B: n=5000 rev/min (stable, high speed zone), C: n=8000 rev/min (unstable, high
speed zone).
Figure 4. Theoretical stability chart of turning processes in the plane of the dimensionless spindle
speed and the dimensionless depth of cut . The dimensionless chatter frequency is greater
than 1. The damping ratio is fixed at , while refers to the sequence number of the lobes.
Figure 5 Dynamic model of highly interrupted cutting
Figure 6 Theoretical stability chart of highly interrupted cutting in the plane of the dimensionless
spindle speed and the dimensionless (axial) depth of cut The dimensionless chatter
frequencies are . The damping ratio is fixed at , .
Figure 7 Dynamics of milling
Figure 8 Stability lobes predicted with the zero-order solution analytically and time-domain
numerical simulations for a cutter with two circular inserts.
Figure 9 Comparison of zero-order and multi-frequency with three harmonics stability lobe
solutions. The cutter has three teeth with a diameter of 23.6 mm used in down milling of Al6061
material ( =500 MPa, =0.3) with a feed rate of 0.120 mm/rev/tooth and a radial depth of 1.256
mm. The structure is assumed to be flexible in y (normal) direction only with =1.4×106 N/m,
=907 Hz, =0.017.
Figure 10 Discretization of the time-periodic function A(t) by piece-wise constant step-function
(a), and the discretization of the constant time delay T by time-periodic delay (b).
Figure 11 Stability chart of the milling process. The period-doubling lobes become unstable
islands. Feed per tooth is 0.1 mm, radial immersion is 0.02, a number of cutting edges is , the
cutting coefficients are MPa and , the damping ratio is .
Figure 12 Top) Time-domain results for a period-2 bifurcation, i.e. chatter with added lobes.
(Bottom) Poincaré map for period-2 bifurcation. The sampled points align at a two fixed
locations for the period-2 bifurcation.

MANU-19-1385 Altintas p.47

Figure 13 Example bifurcation diagram. Stable behavior is observed up to an axial depth (b) of
approximately 2.5 mm. It is indicated by the single value of the sampled feed direction (x) tool
displacements. Between 2.5 mm and 4 mm, secondary Hopf chatter with added lobes occurs
(i.e., a distribution of sampled points). Between 4 mm and 5.5 mm, a period-3 bifurcation is seen
(i.e., three distinct sampled points).
Figure 14 Dynamic flexibilities of a twist drill and milling cutter with circular inserts.
Figure 15 Rectilinear representation of variable pitch milling tool proposed by Slavicek [71].
Figure 16 Effect of on stability gain for a four-tooth end mill with linear pitch variation.
Figure 17. The effect of pitch angle variation on the stable depth of cut for different spindle
speeds and on the chatter frequency [77].
Figure 18 Chatter stability of regular and serrated end mills, after Merdol et al. [81].
Figure 19 (a) Parallel turning [85] and (b) parallel milling operations [92].
Figure 20 (a) parallel turning configuration, (b) stability lobe diagram for a1 versus spindle speed
(n) when a2= 1.5 mm and (c) time-domain simulation for different points f (unstable), e (stable)
and d (unstable).(From [85])
Figure 21 Predicted stability lobe diagram and the experimental results

MANU-19-1385 Altintas p.48