Download as pdf or txt
Download as pdf or txt
You are on page 1of 11

Int. J. M a t h . Tool Des. Res.

Vol. 1, pp. 325-335.

P e r g a m o n Press 1961.

Printed in Great Britain

A CRITICAL COMPARISON OF TWO C U R R E N T THEORIES


OF MACHINE TOOL CHATTER
C. ANDREW* and S. A. TOBIASt
Summary--This paper discusses and compares two current theories of machine tool chatter, considering
differences in their basic assumptions, their theoretical structure and their conclusions. It thus clarifies
their particular fields of application and indicates the requirements of a generally applicable chatter theory
combining the sound features from both.
INTRODUCTION
AS CUTTINGmaterials improve and metal removal rates increase, the capacity of a machine
tool is being limited more and more by a vibration frequently accompanying the machining
process, known as machine tool chatter. The limitations manifest themselves principally in
defective surface finish, poor machining accuracy and greatly reduced tool life. The last
mentioned limitation is of particular importance with many of the new cutting tool materials
developed to maintain a high hardness up to elevated temperatures. These are more brittle
than conventional materials and thus more susceptible to damage from vibration.
Machine tool chatter is a self-excited vibration the amplitude of which can build up
without the presence of any oscillatory forcing agent, such as out of balance rotating parts,
intermittent cutting, etc. Remedies associated with these forced vibration problems, such as
accurate balancing, mounting on vibration isolators, etc., are of no assistance when dealing
with chatter, which arises through the machine structure and the process of chip formation
combining to form a dynamically unstable system. Up to the present, several distinct
physical phenomena have been put forward, which, separately or combined, explain the
occurrence of chatter. The most important of these, and that which is to be considered in
this paper, is the so-called regenerative effect.
In recent years two theories dealing with regenerative machine tool chatter have been
developed independently, one by Tobias and Fishwick [1] and the other by Tlusty and
Polacek [2]. Both theories have been applied to a number of machining processes with
considerable success, claiming thus, with minor reservations, acknowledgment of general
validity. In the present paper a critical comparison of these two theories is being made.
Their basic assumptions, their theoretical structure, their power of explanation and their
limitations are discussed. With this their relative status is clarified and the basic foufldations
of an integrated theory, combining the indisputable features of both theories, are being laid.
REGENERATIVE CHATTER
A schematic representation of a machine tool under cutting conditions is shown in
Fig. 1. The machine structure "surrounds" the workpiece and tool as the chip is removed.
The cutting force exerted on the structure by the cutting action produces a structural deformation, a "prising apart" of the parts of the machine holding the tool and workpiece.
This relative deflection of tool and workpiece in turn alters the cross-sectional area of the
chip being removed, and hence the magnitude of the cutting force. The mechanics of such
* James Archdale and Co. Ltd., Worcester, formerly Department of Mechanical Engineering, University
of Birmingham.
t Professor c,f Mechanical Engineering and Head of Department, University of Birmingham.
325

326

C. ANDREWand S. A. TOBIAS

machining can thus be represented by a closed loop system, with the two principal components:
(1) the dependence of the cutting force on the relative displacement between tool and
workpiece;
(2) the response of the structure to variations in the cutting force, causing tool-workpiece relative displacement.

displace- +
ment
I

ing
force

Chip formation

Machine structure

FIG. l. Schematic representation of a machine tool in operation.


Any analysis of chatter must include these two components. These being fundamental,
it is convenient to use the assumptions made for them, and the subsequent analytical manipulations, as a basis for comparison of chatter theories.
The deflections, and the forces causing them, are essentially dynamic quantities, varying
with time, of which the static or average values are special cases. Figure 2 shows schematically the formation of a chip under such dynamic conditions. Consider that during the cutting
pass to form the top surface there is some disturbance of the cutting process, due to a hard

NQMINA[= DF=.PTH

oF c

,o

l//_//A

Xo

FIG. 2. Formation of chip under dynamic conditions.


spot in the material, etc. Whatever the cause, this disturbance results in a vibration of the
machine structure to give relative tool-workpiece displacement, and hence a wave on the
surface is formed. During the next cut this wave causes a variation of the chip size, which
produces a varying cutting force. The varying cutting force will, in turn, excite the structure
to give further relative vibration between tool and workpiece and another wavy surface.
The surface wave is thus regenerated from one cut to the next.

A Critical Comparison of Two Current Theories of Machine Tool Chatter

327

The stability of such a machining process can be defined as follows: when the amplitude
of the regenerated wave is larger than that of the generating wave, the amplitude will increase
from cut to cut, i.e., the system is unstable and chatter occurs. When the amplitude decreases, the system is stable, and disturbances die out. Both theories considered here set
out to find the conditions existing at the stability boundary when waves are regenerated at
a constant amplitude.
CUTTING

FORCE

ASSUMPTIONS

As an element of the mechanical "loop" description of machining, assumptions are


required for the way the cutting force varies when cutting a chip, as in Fig. 2. Tlusty and
Polacek have simplified this relation by assuming that the amplitude of the cutting force
variation is directly proportional to the variation in chip cross-sectional area. Thus, in
Fig. 2, if the chip has a constant width, the area variation is directly proportional to the

ks:

J
Q.

/,

o_

,I \

3
O

(J

I
I
I
depth

So
Chip

I
I

Vo

Cutting speed v

FIG. 3. Variation of cutting force with (a) chip depth, and (b) cutting speed, under steady
conditions.

chip thickness, variation ds. According to the figure, ds -- x -- x0, where xo is the wave
cut on the surface in the previous cut (top surface of chip) and x is being produced at the
present time. The force variation can thus be written as:
d P = ks ds = --ks(x -- xo)

(l)

The magnitude of the total thrust variation factor* ks decides whether or not the machining
process will be stable on a given machine: ks is proportional to the chip width. For a given
structure and nominal cutting conditions there is a boundary value of the chip width above
which chatter is present and below which the process is stable. Tlusty and Polacek aim at
determining that value of ks which corresponds to this boundary value and which will be
denoted by k,,min, representing the minimum value of the total thrust variation factor at
which chatter can arise.
Tlusty and Polacek assume that the relation between the cutting force and the chip
depth, which determines ks, is the same as that arising under steady state cutting conditions.
With this assumption, ks is found by plotting the steady state cutting force P as a function
* The total thrust variation factor ks is denoted by Tlusty and Polacek as the "Koppelungskoeffizient r'"
The different nomenclature has been introduced to show more clearly the relation between the two theories.

328

C. ANDREWand S. A. TOBIAS

of the chip depth (chip thickness) for a given set of cutting conditions, and determining the
slope of this curve at the nominal depth of cut so, i.e. ks = dP/ds as shown in Fig. 3(a).
Tobias and Fishwick, on the other hand, consider chatter to be a dynamic process and
that consequently force-displacement relations cannot be read across from steady state
cutting tests. They assume that under chatter conditions the cutting force P is a function
not only of the chip thickness s but also of the feed velocity (penetration rate) r and the
rotational speed f2. From this they derive the dynamic cutting force increment

dP = ZCkl ds 4- Z
z (ks -- kl) -~
2~r d r

zek~ -- ZC (ks -- kl) ~]

(2)

where ze is the mean number of cutting edges in simultaneous contact with the workpiece,
out of a total of z cutting teeth, so is the nominal feed and f2 the nominal rotational speed
of tool or workpiece, ks is the total thrust variation factor, determined in accordance with
Fig. 3(a). k a is the speed variation factor which is determined by finding the slope of the
cutting thrust versus cutting speed curve kv and dividing it by the workpiece or tool radius R,
as shown in Fig. 3(b). Both ks and ka are "static" coefficients which can be found from steady
state cutting tests. On the other hand, the chip thickness variation factor kl is a dynamic
coefficient which can be determined only by special experiments, k s - kl = K is called
the penetration rate factor.
Tobias and Fishwick assume that the variation of the rotational speed t2 can be neglected
(Tlusty and Polacek make the same assumption) and consequently the third term in equation
(2) can be omitted. From the resulting equation they derive

d P = zekl x ( t ) - - x t - -

4- zc ~

(3)

&

where x ( t ) - x ( t T )

oo os.oo in to

represents the instantaneous variation of the chip thickness,

x i,

,o

It is clear that the expression of the dynamic thrust variation dP as derived by Tobias
and Fishwick (equation (3)) is considerably more complicated than that used by Tlusty
and Polacek (equation (1)). The following differences between these equations are to be
noted:
(I) Equation (3) includes the mean number of cutting edges in simultaneous contact,
zc. The significance of this is explained with Fig. 4, which shows two cutters removing the same total depth of material, however, one with two teeth and the
other with just one. Since the total chip area removed is the same for each case,
the total mean cutting forces will also be of the same order. However, for stability,
it is the cutting force variation due to a given relative displacement of tool and workpiece which is significant. If two teeth are in contact, a given displacement results
in twice the chip area, i.e. cutting force variation, as when one tooth is in contact.
In that case, chatter will arise at half the chip width necessary to give rise to it
for the single edge case.
(II) Equation (3) consists of two parts. The first corresponds roughly to equation (2),
representing that part of the thrust variation which is due to the chip thickness
variation only. The second part represents the thrust variation due to the variation
of the penetration rate (feed velocity), the effect of which does not enter at all in

A Critical Comparison of Two Current Theories of Machine Tool Chatter

329

the theory due to Tlusty and Polacek. This component serves to explain some
experimental facts, as will be seen later.
(III) Equation (3) contains the time of workpiece or cutter revolution T (and the rotational speed f2 ---- 2rr/T) and through the term x ( t - - [T/z]) the phasing between
the top and bottom surface waves of the chip removed (see Fig. 2). This feature
also explains certain important experimental facts, as will be shown shortly.
(IV) Equation (3) contains a dynamic constant, the chip thickness factor kl, which
must not be confused with the total thrust factor ks, contained both in equations
(4) and (2).

F3c. 4. Effect of number of teeth in cut.


STRUCTURAL ASSUMPTIONS
It is now necessary to consider the second part of the "loop" describing the mechanics
of machining, i.e., the way the structure responds to variations in cutting force and thus
causes tool-workpiece relative displacement. Both theories assume that machine structures
have sufficiently small damping for their vibration characteristics to be considered as composed of the ,;uperposition of separate motions of many structural modes. Each mode is
defined by a vibrating shape, in which the amplitude of the motion of any point on the
structure follows the amplitude-frequency relation of a single degree of freedom system.
Thus, for each mode, since the vibrating shape is fixed, the relative motion between tool
and workpiece lies along a straight line with a fixed direction, and its amplitude has single
degree of freedom amplitude and phase characteristics.
The procedure with which Tlusty and Polacek introduce the structural characteristics
into their theory will be explained with the aid of Fig. 5. The single degree of freedom system
shown in Fig. 5(a) is used to represent the relative motion between tool and workpiece
when the structure vibrates in one particular mode. The end mass m is constrained to
move in the direction X, which corresponds to the principal direction of oscillation between
workpiece anti tool in that particular mode. If the system is excited by a force P(o,) acting
in the direction X, then the response of the mass will be denoted by X(co). The variation

330

C. ANDREWand S. A. TOBIAS

of the amplitude of the response as a function of the exciting frequency oJ, for force of unit
magnitude, is shown in Fig. 5(b), this being the well-known resonance curve of a single
degree of freedom system.
It is only the component normal to the machined surface of the response X(o~) which
leaves surface marks and produces a cutting force variation dP. Let now the direction of
this normal be denoted by Y in Fig. 5(c) and let the angle between the normal Y and the
principal direction of oscillation be a. The response of the mass in the direction Y will then
be given by X(oJ) cos a. However, as a rule the cutting force P does not act in the direction
X but forms an angle/3 with the normal to the machined surface, as shown in Fig. 5(c).

P
/

/
,iD,f-'

',,'~

\ (x)
(b)

(a)

{d)

(c)

FIG. 5. Schematic representation of structural mode--after Ylusty and Polacek [2].


The cutting force component in the direction X is clearly P cos (a --/3) and this produces
in the direction Y the response Y(~o) = X(o~) cos a cos (a --/3) = F(oJ) P where
F(Go) = X(co) cos a cos (a --/3)
P
is the directional response .[unction, a graphical representation of the amplitude variation
of which being shown in Fig. 5(d).
Every mode of the structure can now be treated in a similar way. In general, the principal
direction of oscillation (direction X) will vary from mode to mode, being always determined
by some value of the angle a. Moreover, the response function X(oJ) will also vary from mode
to mode. With these a directional response function F,(o~) for each mode can be found,
and the sum of these, the overall directional response F(oJ) = ~ F~(oJ) describes the motion
i
of the structure in the direction normal to the machined surface, for a harmonic force
which acts in the direction of the cutting force P.
Tobias and Fishwick restrict their attention to the case when those modes of the structure
which can be excited by a force, acting in the direction of the cutting thrust, are well separ-

A Critical Comparison of Two Current Theories of Machine Tool Chatter

331

ated in frequency. Under these conditions, the relative motion between tool and workpiece
to cutting force variations can be described by an equivalent system
m2 + c~ + Ax -- - - d P

(4)

where m, c and A are equivalent constants determined from resonance tests. The displacement x falls in the principal direction of oscillation between tool and workpiece, and dP
is given by equation (3).
Each mode which can be excited under these conditions can be represented by such an
equivalent single degree of freedom system, and according to Tobias and Fishwick the
chatter behaviour of each mode, i.e. of each single d e g r e e o f freedom system, must be
investigated separately. In practice, the number of modes which must actually be considered is limi~:ed, those modes which are of practical importance being discovered by
investigating the actual motion of a machine during chatter.
D E R I V A T I O N OF T H E S T A B I L I T Y C O N D I T I O N
Tlusty and Polacek consider chatter as a problem of]orced vibrations, where the wave
produced during one cut excites a further wave on the succeeding cut. Considering their
assumptions se far:
Cutting force variation (equation (1)) dP = --ks(x -- xo)
Response of x to force dP

x = dP . F(~o)

Combining these equations an expression relating the amplitudes of successive cuts is


derived as

( F ~ ) k s ] x = ksxo

(5)

where the right-hand side is a "forcing" term and the left-band side a function describing
the response of x to a given force. The boundary of stability is assumed to occur when
x and x0 have equal amplitudes. Introducing this condition into equation (5) and carrying
out certain algebraic manipulations leads to an equation which determines the minimum
value of the total thrust variation]actor ksmin at which chatter can just be maintained. Since
k.s is proportional to the chip width, ks~n determines the minimum unstable chip width,
any width smaller than that being machined without chatter. Owing to the complexity
of machine tool structures, ks~n is found graphically, although this is not an essential
feature of the l:heory.
Tobias and Fishwick consider chatter to be a problem of dynamic stability. By substituting equation (3) into equation (4) they derive

mYc + (c + Z-~ " 27rK1)

~- (A + zckl) x -- zcklx (t -- T) = 0

(6)

This is a linear differential equation of second order and consequently it has a solution of
the type x = Ae nt cos oJt. I f now a certain set of cutting conditions leads to an exponent
3 > 0, then the amplitudes of the vibration x should increase exponentially with time,
indicating the cutting condition to be unstable. 3 < 0 defines decreasing amplitudes, that is,
stable cutting conditions. It follows that the threshold o] stability conditions are found by
making 3 = 0,, i.e. by substituting x = A cos oJt into equation (6). These conditions define
those values of ka, as a function of the rotational speed f~, at which the system is on the

332

C. ANDREW and S. A. TOBIAS

threshold of stability. The chip width is again taken to be proportional to the chip thickness
variation Jactor kl and consequently for each value of the non-dimensional rotational
speed f~/w0 (o~o being the natural frequency of the mode the stability of which is being
tested) a minimum value of kl, i.e. chip width, is found which can just maintain chatter.
This information is presented graphically as a stability chart.
Before discussing the final results of the two theories, attention ought to be drawn to
certain similarities between them. Since both theories are concerned solely with boundary

Fl~. 6. Phase restriction due to several teeth in cut.

conditions, it is immaterial whether the problem is treated as one of forced vibration or


dynamic stability. The essential difference between the two approaches lies in the different
initial assumptions and not in their theoretical apparatus.
These features will become clear by writing equation (6) in the form
m3~ +

(c + zc27rK~)
: + z .

q-(;~-kzckl).=zcklx(t--T)

This can now be compared directly with equation (5)

[ ~ ) -k ks] x = k,xo

(5)

which is the equivalent equation due to Tlusty and Polacek.


(I) m~ -k c3c )tx corresponds to

the difference being that the latter term includes the effect of all structural modes, the
former of only one.
(II) In equation (5) the penetration rate factor K appears as an additional damping
coefficient. When K > 0 then the additional damping is positive and so the system is made
more stable. Since K appears as divided by the rotational speed, the influence of this effect
increases with decreasing speed.
(II|) The response of the displacement x in Tlusty and Polacek's expression

I ks]x

A Critical Comparison of Two Current Theories of Machine Tool Chatter

333

to a fixed force, is a series of resonance curves, one corresponding to each mode. It is


assumed that chatter occurs at the largest resonant amplitude given by this response. However, to excite at resonance requires a fixed phase between exciting force and excited amplitude, and therefbre, in this context, a fixed phase between the two surface waves, x and x0.
Now consider a machining case with more than one tooth in contact with the workpiece,
as in Fig. 6. The phase between the two waves at the cutting edge is fixed by the time between successive edges passing a given point on the surface (Tobias and Fishwick's T)
and the chatter frequency. Thus, in this case, the phase between successive waves is not, in
general, that corresponding to the largest resonant amplitude of Tlusty and Polacek's
expression

but will only be compatible with one particular off-resonance condition, and the stability
is thereby improved. Thus, in short, by placing no restriction on the relative phase between
successive surface waves, Tlusty and Polacek limit themselves to cases of effectively single
pass machining, such as planing and shaping. In processes where often several cutting
edges are in simultaneous contact with the workpiece, such as drilling, broaching, milling
with multi-toothed cutters, etc., their conclusions are pessimistic over certain speed ranges.
C O M P A R I S O N OF C O N C L U S I O N S AND
EXPERIMENTAL VERIFICATION
The differenoes between the two theories are most clearly demonstrated by a comparison
of their conclusions. This is most conveniently carried out by plotting the final results
obtained by them in the form of stability charts, as this is normally done by Tobias and
Fishwick. Such a chart shows the minimum unstable cut width, or a coefficient proportional
to that, as a function of the rotational speed of the tool or workpiece.
Figure 7(a) shows such a stability chart as calculated by the theory of Tlusty and Polacek.
Since according to them the minimum unstable cut, that is, ks~n, is constant and independent of the rotational speed ~, the stability chart consists of a line parallel with the
abscissa, at an ordinate value corresponding to k s ~ . All ks values larger than this are
unstable, this being indicated by shading of the area lying above ksmin.
Figure 7(b) shows a stability chart as calculated by the theory of Tobias and Fishwick.
The shaded areas correspond once more to unstable cutting conditions. It can be seen that
chatter is predicted to arise in certain speed bands, which are separated from each other by
stable speed bands. Above a certain speed all speeds are supposed to be stable. The prediction of unstable speed bands interspersed with stable speed bands can be traced to be
due to the inclusion of the time between successive cuts (T/z), i.e. the phasing of the surface
waves cut in successive cuts, in their theory. It is noted also that all unstable bands have a
lower envelope which raises the unstable bands at low speeds to high values of the chip
width (or kl). This is due to the inclusion of the penetration rate coefficient in their theory
and the assumed variation of the penetration effect as a function of the rotational speed.*
The question that remains to be answered is which of these two types of charts are
encountered in practice. The answer is given in Fig. 8, which shows the experimentally
* Strictly speaking, equation (2) is valid only for small variation of ds, dr and d~L Under these conditions
the coefficientsof these terms, i.e. kl, K, k~, etc. are constant. However, in the stability chart the rotational
speed ~ varies within wide limits, and Tobias and Fishwick tacitly assume that equation (2) is valid also
in their case. This assumption is justifiable by the correspondence of experimental and theoretical results.

334

C. ANDREWand S. A. TOBIAS

determined stability chart of a face milling process, carried out on a vertical milling machine.
CONCLUSIONS

Although a comparison of Fig. 7 with Fig. 8 appears to invalidate the theory due to
Tlusty and Polacek, this impression is by no means correct. Certain omissions of their
theory, for instance the absence of the effective number of cutting edges, do not represent
errors in principles and are easily introduced. With these the theory yields not the minimum

\\
hmo

Cutter

Stability

rotational

speed

chart as derivable from

Tlusty and Polacek


x

0"5

I00

%~
~d
<

90

'_o%

7o

i .~

60

0"4

80

"6

50

~
N

4o
3C

0'3

0'2
I0

20

30

40

50

60

70

80

90

IO0 I10 120 IBO 140 150 160 170 180 190 200

Cutter speed- rev/min (z = 52,~o=46 c/s)

Stability chart

as derived

from

Tobias and Fishwick

FIC. 7. Stability charts as derivable from (a) Tlusty and Polacek [2], (b) Tobias and Fishwick [1].

value of the total thrust variation factor but klmin. It can be shown that, in practical cases
where both theories can be applied equally, this klmin value is in fact proportional to the
horizontal asymptote of the stability band envelope in Fig. 7(b) or Fig. 8. Thus, although
the theory due to Tlusty and Polacek leads to an erroneous result, the error is on the safe
side (normally increasing with decreasing speed). However, there are cases when the
theory due to Tobias and Fishwick cannot be applied at all, for instance when the natural
frequencies of two or more modes are close to each other. In cases like these the theory
due to Tlusty and Polacek is still applicable, leading to results which though erring on the
safe side are nevertheless of practical use.

A Critical Comparison of Two Current Theories of Machine Tool Chatter

335

Quite in general, it is only too obvious that the strength of the method due to Tlusty
and Polacek lie.s on the structural side of the chatter problem. Their theory takes full
account of the structural characteristics of the machine and shows how each mode contributes to the overall chatter behaviour, making a comparison of the general stability levels
of given machine structures and cutting orientation possible and indicating the optimum
lines of development of improved structures. On the other hand, the strength of the method
due to Tobias and Fishwick lies on the cutting force side of the chatter problem, giving a
more accurate representation of the dynamic cutting of metals by taking into account the
effect of the rate of penetration, the number of cutting edges in simultaneous contact and
the rotational speed.
100 m
c!

~, 80

/;; ;

'~IB Io
\1

riO

~',

---

-"

-'hm

40

J
"~: 20

ql)

C~

I
0

2O

I 1 I
40

~0

I i
BO

Cutter

I00

speed,

I
t
120

140

IGO

IaO .

200

rev/rnin

FIG. 8. Experimentally derived stability chart--Tobias [3].

It is concluded that the two theories are approximately complementary and that a
synthesis of them ought to yield a method with their respective advantages, without their
limitations. Such an integrated theory has, in actual fact, already been developed. This
theory represents a considerable advance towards the final solution of the chatter problem
but leaves nevertheless a good many problems still unsolved.
REFERENCES
[1] S. A. TOBIASand W. FmHWmK, Engineering 205, 199, 239 (1958).
S. A. TOmAS arid W. FIsHwIO, Proc. lnstn, mech. Engrs. 170, 232 (1956).
S. A. TOBIASand W. FISh, tIcK, Trans. Amer. Soc. mech. Engrs. 80, 1079 (1958).
[2] J. TLUSTYand M. POLACEI,Drittes Forschungs und Konstruktionskolloquium Werkzeugmaschinen, p. 131,
Ill. FoKoMa, Vogel Verlag, Coburg (1957).
[3] S. A. TOBIAS,t'roc. Instn. mech. Engrs. 173, 474 (1959).

You might also like