Download as pdf or txt
Download as pdf or txt
You are on page 1of 12

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/252157845

Boring Bar Chatter Control

Article  in  Proceedings of SPIE - The International Society for Optical Engineering · January 1998

CITATIONS READS

3 646

2 authors:

Jon R. Pratt Ali H. Nayfeh


National Institute of Standards and Technology Virginia Polytechnic Institute and State University
137 PUBLICATIONS   2,720 CITATIONS    824 PUBLICATIONS   34,532 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Geometrical determination of an added mass using M/NEMS resonators View project

Nonlinear Behavior of Systems View project

All content following this page was uploaded by Jon R. Pratt on 26 September 2014.

The user has requested enhancement of the downloaded file.


BORING BAR CHATTER CONTROL
J. Pratt*
National Institute of Standards and Technology
Manufacturing Engineering Laboratory
Gaithersburg, MD 20899, USA

A. H. Nayfeh
Department of Engineering Science and Mechanics
Virginia Polytechnic Institute and State University
Blacksburg, VA 24061-0219, USA

ABSTRACT. Boring bars for single-point turning on NOMENCLATURE


a lathe are particularly susceptible to chatter and have f,w natural frequencies in Hz and radians/sec
been the subject of numerous studies. Chatter is, in where subscript implies coordinate
general, caused by instability. Clearly, the cutting pro- ( damping ratio
cess can be limited to regions of known stable operation. power amplifier gain
K1
However, this severely constrains the machine-tool op- actuator electromechanical gain
K2
eration and causes a decrease in productivity. The more sensor gain
Ks
aggressive approach is to attack the stability problem Z(s) actuator electrical impedance
directly through application of vibration control. Here, R actuator electrical resistance
we demonstrate a new biaxial vibration control system L actuator electrical inductance
for boring bars. We present the experimentally deter- d(s) Laplace transformed disturbance
mined modal properties of the VPI Smart Tool and A(s) Laplace transformed accelerance
demonstrate how these properties may be used to de- modal factor
hn
velop models suitable for chatter stability analysis, sim- angle between modal coordinate and
1/JJ
ulation, and development of feedback compensation. A normal to machined surface
phenomenological chatter model that captures much of stiffness coefficient
the rich dynamic character observed during experiments damping coefficient
is presented. We introduce the notion that the mean cutting stiffness
cutting force changes direction as the width of cut in- angle between mean cutting force and
creases due to the finite nose radius of the tool. This normal to machined surface
phenomenon is used to explain the progression from T period of one workpiece revolution
chatter that is dominated by motions normal to the ma- w width of cut
chined surface at small widths of cut to chatter that is dP differential cutting force
dominated by motions tangential to the machined sur- structural nonlinearity
fn(Xn)
face at large widths of cut. We show experimental evi- cutting force nonlinearity
gn(Xn)
dence to support our assertion that a biaxial actuation overlap factor
f-Ln
scheme is necessary to combat the tendency of the tool influence coefficients
/n,V,TJ
to chatter in both directions. C(s) compensator

1 INTRODUCTION
Key 'Vords:
Boring bars are tools used in single-point turning op-
Chatter, control, self-excited oscillations, machining. erations to create cylindrical inner surfaces. A typical
boring bar operation on a two-axes engine lathe is il-
lustrated in Fig. 1. Self-excited vibrations known as
*The work presented here was completed in the Department
of Engineering SciPnce and Mechanics, Virginia Polytechnic In- chatter become increasingly problematic in boring due
stitute and State University, Blacksburg, VA 24061-0219, USA. to the tool overhang. Chatter vibrations have an ad-

215
verse effect on the surface finish and accuracy and must ing control in both principal directions of vibration.
be avoided. Generally, the rate of material removal is
reduced to maintain stability, but this causes a sub-
stantial decrease in productivity.
2 MODAL PROPERTIES
Many investigators have sought means to increase the
stability of machining operations. It is generally ac- The VPI Smart Tool bolts directly to the cross-slide of
cepted that stiff, highly damped tools have a lower ten- a two-axis engine lathe. The layout of components in
dency to chatter. Rivin and Kang [1] substantially in- relation to the lathe is illustrated in Fig. 2.
creased the damping of a lathe tool by using a sandwich
of steel plates and a hard rubber viscoelastic material The control system is comprised of two actuator/sensor
to form a laminated clamping device to hold the tool. pairs; one pair to control motions in the direction tan-
Kelson and Hsueh [2] also achieved an increase in sta- gent to the machined surface, and the other pair to con-
bility by redesigning the tool holder for added stiffness trol motions in the direction normal to the machined
and damping. Tobias [3] cites a number of instances surface. The system is configured so that the tool is
where passive vibration absorbers of various configu- much stiffer in the tangential than in the normal direc-
rations have been applied with success. He describes tion, because the corresponding Terfenol-D actuator is
applications using a Lanchester absorber, a dynamic vi- coupled to the tool through a 10/32 stud in the former
bration absorber, and an impact absorber. Researchers case and through a thin, tensioned wire "stinger" in the
have also found that the cutting speed can be modu- latter case.
lated to enhance stability [4, 5].
We seek two types of information regarding the dynam-
~achtigal, Klein, and Maddux [6] patented a synthe- ics of the Smart Tool. First, to model chatter, we need
sis circuit to counteract the cutting forces and pro- information about the natural frequencies and damping
posed using it with a reaction mass actuator to control ratios of the two lowest Smart Tool vibrational modes.
boring bar chatter. Matsubara, Yamamoto, and Mizu- Second, in order to model and develop the active vi-
moto [7] proposed an active piezoelectric damper for a bration control, we must characterize the frequency
boring bar. Recently, Tewani, Rouch, and Walcott [8] response of the combined actuator-tool-sensor system,
demonstrated chatter control of a boring bar by using which includes the actuator electrical impedance prop-
what they call an active dynamic absorber. The con- erties.
trol scheme uses a piezoelectric, reaction-mass actuator
mounted inside the boring bar and has been patented
[9]. All of these control systems are unidirectional and
control vibrations in a principal modal direction of the 2.1 Linear Frequencies and Damping
tool.
The linear frequencies and damping ratios of the Smart
In this paper, we discuss the dynamic characteristics Tool were identified using impact testing and circle fits
of the VPI Smart Tool system. This system is de- of the resulting frequency-response functions or FRF's.
signed to sense and control tool vibrations both nor- For our purposes, it was sufficient for chatter modeling
mal and tangential to the machined surface, which is to obtain information about only the first transverse
a natural extension of previous unidirectional research vibration mode in the directions normal and tangential
efforts. We describe here how experimental identifi- to the machined surface.
cation techniques have been used to develop models
of the Smart-Tool dynamics for control development We find, as did Salje [14] with his boring bar experi-
and chatter modeling. Chatter signatures are presented ments, that the modal characteristics of the tool sys-
that clearly indicate complex motions both normal and tem vary greatly depending on the clamping condi-
tangential to the machined surface. A phenomenolog- tions. The dynamic properties were seen to vary from
ical model based on nonlinear regenerative chatter is ft = 491 Hz and (t ~ 0.025 and fn = 364 Hz and
offered to explain this behavior, and it is shown to pos- (n ~ 0.02 on one occasion to ft = 524 Hz and (t ~ 0.03
sess good qualitative agreement with observed boring and fn = 365 Hz and (n ~ 0.02 on another. Clearly,
bar vibrations. We then briefly review our active con- the tangential coordinate shows the greatest variation.
trol scheme, details of which are reported elsewhere This may be due to the rigid mounting of the actuator
[10, 11, 12]. Finally, we present experimental results in this direction, which made the system highly sensi-
from performance testing of the VPI Smart Tool. These tive to the actuator alignment within the supporting
experimental results reinforce the importance of apply- fixture.

216
2.2 Plant Response Function
In order to design a compensator, we formulate a single- Table 1: Estimated Properties of the Tangential Plant
input, single-output problem in terms of the plant re- Response Function.
sponse function. We identify this function by develop-
Mode Hz damping ratio modal factor
ing a model for the FRF between the voltage input to
1 523.4 0.028 0.018
an actuator and the voltage output by its correspond-
2 750 0.1 -0.0013
ing accelerometer.
3 2535 0.05 0.223
4 3000 0.05 0.223
The normal and tangential vibration-control loops are
aligned with the principal axes of the tool cross section
so that we treat them as uncoupled systems in the ab-
sence of cutting forces. For this paper, we develop a
model for the tangential control system as an example expansion of the cross-accelerance so that its Laplace
of the procedure. transform A(s) is

The actuator is provided with a random voltage ex-


citation and the corresponding output acceleration is
sensed as a voltage from the accelerometer signal con-
ditioner. The actuator power amplifier is set at a
where Wn, (n, and hn are the frequency, damping ratio,
midrange gain and the random voltage input is selected and modal factors, respectively, of the nth mode.
to avoid amplifier and actuator saturation. Both the
input and output voltages are measured using a digi-
\Ve assume that the frequencies and dampings asso-
tal frequency analyzer that computes the correspond-
ciated with the structural response are global proper-
ing FRF, shown as the dashed line in the Bode plot of
ties that can be determined using FRF's obtained any-
Fig. 3.
where along the structure. Since the resonant peaks are
fairly well separated, we use single-degree-of-freedom
The phase indicates first-order dynamics, as evidenced
techniques to obtain estimates of the natural frequen-
by the roll off from 180° to approximately 90° between
cies, damping ratios, and modal factors of the dominant
0 and approximately 300 Hz. This is to be expected
modes directly from the plant response function. The
since the actuator is magnetostrictive, and its electrical
results of such an analysis for the tangential control
impedance should resemble that of an inductive load.
system are summarized in Table 1.
Between 450 and 600 Hz the response exhibits a res-
onance peak accompanied by a 180° shift in phase.
Clearly, this mode is observable and controllable. How- The second-mode parameters are the result of trial and
ever, we note evidence of a noncollocated response in error curve fits where the frequency, damping ratio, and
the continued phase roll off near the first antiresonance. modal factor were varied in order to produce the ap-
In a collocated actuator-sensor pair, the input is in propriate total phase response. The second mode is
phase with each modal response and there is phase re- not observable from the magnitude response and was
covery associated with each antiresonance. not apparent in impact tests of the structure; however,
its effect on the phase of the plant response function is
Based on the preceding observations, we construct a dramatic.
tentative block diagram for our control system in Fig. 4,
where K1 is the power amplifier gain, K 2 is the elec- Next, the actuator's electrical properties L and R are
tromechanical gain of the actuator, the d( s) are un- determined. The resistance R is measured and found to
known cutting force disturbances, K 8 is the gain due be 5.7 Ohms. The manufacturer provides a plot of the
to the accelerometer sensitivity and amplification of its electrical impedance as a function of frequency from
signal conditioner, Z(s) = (R + Ls)- 1 is the actuator which we estimate an L of approximately 6 mH. We
electrical impedance, A(s) is the Laplace transform of note that our model does not account for back emf or
the cross-accelerance between the transduction "force" eddy current losses.
input at the actuator location and the acceleration mea-
sured at the sensor location, and C (s) is a compensator The force generated by the actuator is assumed to be
to be determined. proportional to the current by a constant electrome-
chanical gain factor K 2 and the actuator power am-
Following standard procedures [13], we assume a modal plifier is simply characterized as a constant gain K 1 .
model for the structural response in a partial-fraction Taking into account the sensor gain Ks, we obtain the

217
plant transfer function the cross spectra and coherences between these signals
for some typical chatter signatures obtained when the
(2) tool frequencies and dampings are ft = 491 Hz and
(t :::::: 0.025 and fn = 364 Hz and (n :::::: 0.02. In this
series of tests, the tool tends to chatter more easily,
and the effect of positive damping due to the cutting
Rather than determine each of the gains separately, we process seems to be much less.
lump them together as a single parameter K. Setting
K = 48 and letting s = iw, where w is the frequency of Chatter is observed at spindle rotational speeds as low
excitation, we plot the model response as the solid line
as 90 rpm and for widths of cut as small as 0.002 11 ,
in Fig. 3.
as seen in Fig. 6. We note that for the shallow cut of
0.002 11 , chatter tends to be localized in a direction nor-
3 CHATTER ID AND MODEL- mal to the machined surface, whereas for the deeper cut
of 0.03", chatter is oriented more along a line tangent
ING to the machined surface, as is evident in Fig. 7. In both
cases, the fundamental harmonic of the motion is seen
In this section we report some experimental observa- to be linearly coherent, suggesting that mode coupling
tions of chatter obtained using the VPI Smart Tool. is present.
Based on the observed tool responses, we develop a phe-
nomenological model of chatter using a reduced-order,
nonlinear dynamic model for the tool and cutting force 3.2 Chatter Modeling
interactions.
The dynamic response of coupled machine-tool systems
has long been recognized for its complexity. Salje [14]
3.1 Chatter Signatures examined experimentally the problem of tools with two
We begin by considering the response of the VPI Smart orthogonally oriented degrees of freedom and revealed
Tool without feedback control and investigate boring rich dynamics. Kuchma [16] considered boring bars
bar chatter for light, finish style cuts at the minimum with non-circular cross-sections and found that the sta-
feedrate of the lathe, which is 0.0024 11 per revolution bility is strongly dependent on the orientation of the
(ipr ). Chatter signatures are obtained for a variety of plane of symmetry with respect to the normal to the
spindle speeds at nominal widths of cut of w 0 = 0.002" machined surface. Tlusty [17] developed the first sim-
and w 0 = 0.03". ple model for mode-coupled chatter, whereas Tobias
[3] suggested that regeneration must also be considered
In Fig. 5 the amplitude of the fundamental harmonic in these systems. A variety of other models have been
of the tool's acceleration tangent to the machined sur- suggested and studied for orthogonal cutting conditions
face is plotted as a function of the workpiece rotational [18, 19, 20].
speed in revolutions per minute (rpm) for the case when
wo = 0.03". For the data in this plot, the first lat- Recall that, in the experiments, we have observed
eral vibration modes have frequencies and dampings jumps from static to dynamic cutting conditions, or
of ft = 524 Hz and (t :::::: 0.03 and fn = 365 Hz and the so-called subcritical-type instability typical of non-
(n :::::: 0.02, respectively, as determined by impact test- linear dynamic systems. Similar behavior was reported
ing and circle fits of the resulting frequency-response by Hooke and Tobias [21] for single-point turning, and
functions. was observed and modeled by Hanna and Tobias [22]
for face milling. Hanna and Tobias [22] used a single-
Two observations can be made from the results in degree-of-freedom model with both cutting-force and
Fig. 5. First, the presence of stable cutting for all structural nonlinearities to account for the jump behav-
workpiece rotational speeds below 165 rpm indicates ior. This model was recently the subject of a thorough
that the cutting forces are dependent on the mean cut- bifurcation analysis that revealed the local supercritical
ting velocity and contribute a positive damping effect and global sub critical nature of its stability [23, 24, 25].
at low speeds, in agreement with Tlusty's results for Shi and Tobias [26] reproduced the theoretical results of
aluminum [15]. Second, a subcritical-type instability Hanna and Tobias [22] by dropping the structural non-
occurs between 165 and 195 rpm, where both chatter linearity and considering the effect of the tool leaving
and stable cutting conditions coexist. the cut, a condition known as "multiple regenerative
chatter".
Next, we consider the time histories and autospectra
of the tangential and normal accelerations as well as All of the aforementioned behaviors seem evident to

218
varying degrees in our experiments. To further com- two-degree-of-freedom coupled machine-tool systems.
plicate matters, it appears from our experimental data Summing forces in Fig. 9, we obtain the following lin-
that the cutting geometry varies as the width of cut earized equations of motion expressed in terms of the
is changed. Consider the sketches of Fig. 8 where we modal coordinates:
recreate the chip flows that were observed to occur over
mi'1 +c1i1 +>.1x1 = -dPcos(8-1/J1) (3)
the cutting insert for two extreme cutting conditions.
1
mi'z + cziz + AzXz = -dP cos ( 27r + 1/J1 - 8) (4)
For the case of a light cut, such as that characterized
by the chatter signature of Fig. 6, the chip tends to where x 1 and x 2 are the tool modal coordinates, re-
flow directly from the work in a direction approaching spectively, m is the modal mass, c1 and c2 are linear
the normal to the machined surface. The differential viscous damping coefficients, >. 1 and >. 2 are linear stiff-
cutting force dP lies approximately within the x 1 - x 2 ness coefficients, and dP is the differential regenerative
plane, as sketched in Fig. 8 (b). cutting force
(5)
In contrast, in a more substantial cut, such as that Here, k 8 is the cutting force coefficient associated with
characterized by the chatter of Fig. 7, the chip tends the workpiece material, x is the displacement compo-
to flow from the work in a direction normal to the so- nent normal to the machined surface, and Xr = x(t- T)
called transient surface. The differential cutting force is the displacement during the previous workpiece rev-
dP appears to change orientation and now lies within olution, w is the width of cut, and T is the period of
the xz - x 3 plane, as is sketched in Fig. 8 (a). the revolution.

The seeming change in the orientation of the plane in We neglect the effect of the tensioned wire stinger and
which cutting might be characterized as nearly "or- assume that the stiffness of the tool in the x 1 direction
thogonal" is due to the finite nose radius of the insert. is that of a simple cantilever beam of length l, modulus
Clearly, the effective tool lead angle can vary from 0° E, and moment of inertia I about its midplane so that
to ;::::; 90° as the chip width is decreased below a value
>. 1 -_ 3EI (6 )
corresponding to the radius of the insert (0.01" in this [3
case). Similar observations regarding the chip flow in
oblique, multi-edged cutting geometries were originally
reported by Stabler[27]. Dividing through by the modal mass m, we obtain
2
To begin our modeling, we consider first the condi- i'1 +2(lwdl +wix1 1
= -dPw cos(8-1j; 1) (7)
tions prevalent during the light cutting scenario, be- >.1
2
cause this will define the lower bound of stable cut
i'z + 2(zwziz + w~xz = dP~~ sin(() -1/Jl) (8)
widths. For the case of a plunge type of cut, where
the tool feeds directly into the tube wall along the x 1 where ( 1 and ( 2 are linear viscous damping ratios, and
axis, the schematic diagram of Fig. 9 is appropriate. w1 and w 2 are the natural frequencies.
We consider this cutting geometry to provide a crude
approximation of the case of a very shallow cut. From Fig. 9 we find that
x = x1 cos (1/J 1)- x 2 sin (1/Jl) (9)
We note that the bar is essentially symmetric, so that
the modal mass is the same for each of the first trans- so that
verse modes, but that the stiffnesses differ owing to the
dP = k 8 w[cos ('l/;l)(x1- Xlr)-
manner in which the boring bar is supported by the
actuators. \Ve assume that the principal modal coor- sin ('l/;1) (xz - Xzr )] (10)
dinates are rotated with respect to the normal to the
machined surface X, as indicated by the angle 1/J 1, in an
effort to reconcile the fact that, in actuality, this is an Let /'1 = wi~ cos(() -1j;1), 1'2 = -wi~ sin (8 -1/Jl),
oblique cutting geometry and not an orthogonal plunge TJ = cos (1/J1), and v = - sin (1j;1), then the equations of
type of cut. motion can be rewritten as
i'1 + 2(1w1i1 + wix =
3.3 Linear Stability Boundary -1'1w[ry(x1- X1r) + v(xz- Xzr)] (11)
The analysis to determine the stability boundary i'z + 2(zwzil + w~x =
closely parallels those of Tlusty [15] and Tobias[3] for -!'zw[ry(xl - X1 r) + v(xz - Xzr )] (12)

219
motions. Taking our cue from Hanna and Tobias [22],
we include both structural and cutting force nonlinear-
To analyze the stability of this system, we assume zero ities and modify the linear model as follows:
initial conditions, take the Laplace transform, and ob-
tain i"1 + 2(1w1i1 + 26i1 + wf[x1 + h(xl)]
= -~1 W{ 77 [(X1 - J.11X1 T) + gl(X1 - J.11X1 T)]
(s 2 +2C1w1s+w12 )X1(s) = (17)
+v [(xz - J.izXzr) + gz(Xz - J.12X2r)]}
-11w[77(l- e-sr)Xl(s) + v(l- e-sr)X2(s)] (13)
(s 2 + 2(2wzs + w~)Xz(s) =
i2 + 2(zwzi2 + 26iz + wnxz + h(xz)] =
-1zw[77(l- e-sr)X1(s) + v(l- e-sr)Xz(s)] (14)
-/zW{ 77 [(x1 - J.11X1 T) + gl(x1 - J.11X1 T)]
Then, solving equation(14) for X 2 (s) yields +v [(xz - J.izXzr) + gz(Xz - J.izXzr)]} (18)

where fn (xn) is a function describing the structural


nonlinearity in the nth modal coordinate, 9n (xn -
J.inXnr) is a function describing the regenerative cutting
Substituting equation (15) into equation (13), we find force nonlinearity in each coordinate, ~n is the damp-
the characteristic equation ing due to the cutting process in each coordinate, In
is an influence coefficient describing the component of
[s 2 + 2(1w1 s + wf + 1 1w77(l - e-sr)][s 2 + 2(zwzs the differential cutting force projected into each coor-
+w~ + 12wv(l- e-sr)]- 'Y112 w 2 77v(l- e-sr) 2 = 0 (16) dinate, 77 is a coefficient describing the degree to which
motions in the x 1 direction contribute to the differen-
To obtain the familiar lobed stability boundary, we as- tial cutting force, v is a coefficient describing the degree
sume s = iwc, separate the real and imaginary terms, to which motions in the x 2 direction contribute to the
and solve the resulting pair of equations for w and We differential cutting force, and J.ln describes the degree
as a function of the delay T. of cut overlap in each direction.

We now consider the predicted cutting stability of the To obtain the simulated tool accelerations of Fig. 12,
Smart Tool when the modal properties are h = 491 Hz we let 6/w1 + (1 = 0.2, 6 = 0, ~/1 =cos (7r/3)wn~,
and ( 1 = 0.025 and h = 364 Hz and ( 2 = 0.02.
11 = cos (7r /6)wf ~~ , 77 = 0.495, v = -0.24, J.11 = 1,
J.lz = 1, and assume the following functional relations
The boring bar itself may be approximated as a circu-
for the structural and cutting force nonlinearities:
lar cylinder of length 9" (length of bar measured from
if x 1 < 0
~x1
tool tip to clamped end) and diameter 1". We recall
that the moment of inertia for a circular cross section h(xl) = { otherwise
is I = 7r / 4( d/2) 4 , then A. 1 = 6060 psi for a modulus
for steel of E = 30000 kpsi. A represenative value for h(x 2 ) =lOx~+ 300x~
the cutting stiffness of aluminum is ks = 145000 psi 2 3
9n (xn- /lnXnr) = 5. 7(xn- J.inXnr ) - 3700(xn- J.inXnr )
according to Kalpakjian [28].

We consider two cases. The first case is for a two-


degree-of-freedom system where 'lj;1 = 15°. In the sec- The regenerative cutting force nonlinearity is identical
ond case, 'lj;1 = 0° and the stability reduces to that of to that used by Hanna and Tobias [22]. The clamp-
a single-degree-of-freedom system. The resulting sta- ing in the x 1 direction was observed to be qualitatively
bility lobes are plotted for a range of spindle speeds "high" in the time traces of the impact response of the
around 170 rpm in Fig. 10. We see from the plot that tool measured during stable cutting at lower speeds;
the stability is affected by the presence of the second hence, the very substantial effect of process damping
degree of freedom. The limit width of cut is in rea- incorporated by the large value of 6. The overlap co-
sonable agreement with that observed experimentally, efficient J.1 1 = 1 was selected to capture the fact that
though cuts below 0.002" were beyond the accuracy of motions in the x 1 direction modulate the undeformed
the lathe used. chip cross-section in a complex fashion that depends on
the previous tool position; on the other hand, motions
in the x 2 direction can modulate the chip thickness via
3.4 Chatter Simulation
a penetration effect labeled type B chatter by Tobias
To simulate the chatter motions during the heavier cut, [3]. The sign of the coefficient 1 2 was chosen to be neg-
we must include some nonlinearity to produce bounded ative based on the tendency of the chatter frequency to

220
be lower than w 2 , which would not be the case if 1 2 > 0. control loop is active. The cutting parameters are set
Perhaps the most unconventional feature of the model for a spindle speed of 170 rpm, a feedrate of 0.0024 ipr,
is the bilinear structural stiffness represented by the and a depth of cut of 0.002". Initially, the control is
function h (xl). This function was chosen to produce only active in the vertical direction. The cut appears
the unsymmetric tool motions that were evident in the to be stable. A disturbance is introduced by hitting the
time traces at this width of cut and speed. We hypoth- bar in the horizontal direction with a hammer. Clearly,
esize that the tool does not so much cut the workpiece the disturbance grows due to regeneration. After the
in the x 1 direction as plow into it, giving rise to the chatter reaches steady state, control is added in the hor-
near impact style oscillations observed in this coordi- izontal direction, thereby stabilizing the system. An-
nate during the experiments. other disturbance is applied but the control maintains
stability.

4 ACTIVE VIBRATION CON- Finally, we demonstrate biaxial or two-mode control to


TROL quench large-amplitude chatter in Fig. 14. The cut-
ting conditions are set for a spindle speed of 230 rpm,
To achieve effective chatter control, we must minimize a feedrate of 0.0024 ipr, and a depth of cut of 0.03".
the tool vibratory response to arbitrary disturbance Large-amplitude chatter is allowed to develop before
forces. In terms of closed-loop performance goals, this applying feedback control, which stabilizes the system.
translates into an output regulation problem. How- \Ve observe a dramatic improvement in the surface fin-
ever, there is little to be gained by controlling all of the ish.
states, or modes, present in our plant response func-
tions because the chatter vibration at the tool tip is
clearly dominated by the first mode in each case.
6 CONCLUSIONS
It is well known that second-order compensators can
enhance the damping of targeted structural modes We have presented the experimentally determined
in a fashion analogous to a mechanical vibration ab- modal properties of the VPI Smart Tool and demon-
sorber [30, 31]. We show in references [10, 29] that strated how these properties may be used to develop
an absorber-style approach can be quite effective for models suitable for chatter stability analysis, simula-
chatter control even in the presence of nonlinear struc- tion, and development of feedback compensation. A
tural and cutting forces. Furthermore, in reference [12] phenomenological chatter model that captures much of
we develop a compensator synthesis scheme and show the rich dynamic character observed during the exper-
that, for the Smart Tool system with Terfenol-D ac- iments was presented. We introduced the notion that
tuators, a feedback controller can be designed that is the mean cutting force changes direction as the width
entirely analogous to a vibration absorber by using a of cut increases due to the finite nose radius of the tool.
compensator of the form This phenomenon leads to a progression from chatter
that is dominated by motions normal to the machined
surface at small widths of cut, to chatter that is domi-
nated by motions tangential to the machined surface at
large widths of cut. We show experimental evidence to
where g is the feedback gain, We and (c are the fre- support our assertion that a biaxial actuation scheme is
quency and damping ratio associated with the second- necessary to combat the tendency of the tool to chatter
order dynamics of the compensator, and Tc is the time in both directions. At present, the nonlinear coefficients
constant associated with its first-order dynamics. The of the phenomenological model are determined in an ad
performance predicted for such a scheme when applied hoc fashion based on comparisons between simulation
to the tangential control system is repeated here in and experiment. In the future, we will seek a system-
Fig. 11. The response is seen to have the form one atic approach for the determination of these coefficients
would expect if a highly-damped mechanical vibration based on a regression analysis of the chatter time series.
absorber were appended to the system of Fig. 12.

5 PERFORMANCE TESTING 7 ACKNOWLEDGMENT


In Fig. 13 we show evidence of a jump-type instability This work was supported by the Army Research Office
in the single-mode control. The time history of hor- under Grant No. DAAH04-96-1-0021 and the National
izontal vibrations is recorded when only the vertical Science Foundation under Grant No. CMS-9423774.

221
References [13] Ewins, D. J., Modal Testing: Theory and Practice,
Research Studies Press, Ltd., Letchworth, Eng-
[1] Rivin, E. I. and Kang, H., "Improvement of Ma- land, 1986.
chining Conditions for Slender Parts by Tuned
Dynamic Stiffness of Tool," Int. J. Mach. Tools [14] Salje, E., "Self-Excited Vibrations of Systems with
Manuf., Vol. 29, 1989, pp. 361-376. Two Degrees of Freedom," Trans. ASME, Vol. 78,
1956, pp. 737-748.
[2] Kelson, Z. Y. and Hsueh, W. C., "Suppression of
[15] Tlusty, J., "Machine Dynamics," R.I. King (ed.),
Chatter in Inner-Diameter Cutting," JSME Int.
in Handbook of High Speed Machining Technology,
J. Series C: Dyn., Cont. Robot., Design, Manuf.,
Chapman and Hall, New York, 1985, pp. 48-153.
Vol. 39, 1996, pp. 25-33.
[16] Kuchma, L. K., "Boring Bars with Improved Re-
[3] Tobias, S. A., Machine Tool Vibration, Wiley, New
sistance to Vibration," Eng. Digest, Vol. 18, 1957,
York, 1965.
pp. 68-70.
[4] Sexton, J. S., Milne, R. D., and Stone, B. J., "A [17] Koeingsberger, F. and Tlusty, J ., Machine Tool
Stability Analysis of Single-Point Machining with Structures, Vol. 1, Pergamon Press, Oxford, 1970.
Varying Spindle Speed," Appl. Math. Modelling,
Vol. 1, 1977, pp. 310-318. [18] Grabec, 1., "Chaotic Dynamics of the Cutting Pro-
cess," Int. J. Mach. Tools Manu/., Vol. 28, 1988,
[5] Takemura, T., Kitamura T., Hoshi, T., and pp. 19-32.
Okushima, K., "Active Suppression of Chatter by
Programmed Variation of Spindle Speed," Ann. [19] Wu, D. W. and Liu, C. R., "An Analytical Model
C.I.R.P., Vol. 23, 1974, pp. 121-122. of Cutting Dynamics. Part 1: Model Building,"
ASME J. Eng. Indust., Vol. 107, 1985, pp. 107-
[6] Nachtigal, C. L., Klein, R. G., and Maddux, K. 111.
C., Apparatus for Controlling Vibrational Chatter
in a Machine- Tool Utilizing an Updated Synthesis [20] Lin, J. S. and Weng, C. 1., "A Nonlinear Dynamic
Circuit, U.S. Pat. No. 3,967,515, 1976. Model of Cutting," Int. J. Mach. Tools Manuf.,
Vol. 30, 1990, pp. 53-64.
[7] Matsubara, T., Yamamoto, H., and Mizumoto, H.,
[21] Hooke, C. J., and Tobias, S. A., "Finite Ampli-
"Chatter Suppression by Using Piezoelectric Ac-
tude Instability-A ~ew Type of Chatter," in Proc.
tive Damper," ASME Design Eng. Div. DE-Vol.
of the 4th International M. T.D.R. Conf., Manch-
18-2, ASME, New York, NY, 1989, pp. 79-83.
ester, 1964, pp. 97-109.
[8] Tewani, S. G., Rouch, K. E., and Walcott, B. L.,
[22] Hanna, N. H., and Tobias, S. A., "A Theory of
"A Study of Cutting Process Stability of a Boring
1.\"onlinear Regenerative Chatter," ASME J. Eng.
Bar with Active Dynamic Absorber," Int. J. Mach.
Indust., Vol. 96, 1974, pp. 247-255.
Tools Manfu., Vol. 35, 1995, pp. 91-108.
[23] Nayfeh, A. H., Chin, C.-M., and Pratt, J.,
[9] Rouch, et al, Active Vibration Control Device, U.S. "Perturbation Methods in Nonlinear Dynamics-
Pat. No. 5,170,103, 1992. Applications to Machining Dynamics," to appear
[10] Pratt, J. R. and Nayfeh, A. H., "Active Vibration ASME J. Manuf. Sci. Eng., 1997.
Control for Chatter Suppression," in Proc. of the [24] Pratt, J. and :'-J"ayfeh, A. H., "Experimental Stabil-
38th AIAA Structures, Structural Dynamics, and ity of a Nonlinear Time-Delay System," in Proc. of
Materials Conf., Kissimmee, FL, AIAA Paper No. the 37th AIAA Structures, Structural Dynamics,
97-1210, 1997. and Materials Conf., AIAA Paper No. 96-1643,
Salt Lake City, UT, 1996.
[11] Pratt, J. R. and Nayfeh, A. H., "Boring Bar Chat-
ter Control Using A Two-Axes Active Vibration [25] 1.\"ayfeh, A. H., Chin, C.-M., and Pratt, J., "Appli-
Absorber Scheme," in Proc. of NoiseCon97, Book cations of Perturbation Methods to Tool Chatter
2, 1997, pp. 313-324. Dynamics," F. C. Moon (ed.), in Dynamics and
Chaos in Manufacturing Processes, Wiley, New
[12] Pratt, J. R. and Nayfeh, A. H., "Experimental Sys- York, in press.
tem Identification and Active Vibration Control of
a Smart Machine Tool," to appear in Proc. of the [26] Shi, M. and Tobias, S. A., "Theory of Finite Am-
Eleventh Symposium on Structural Dynamics and plitude Machine Tool Instability," Int. J. Mach.
Controls, Blacksburg, VA, 1997. Tool Design Res., Vol. 24, 1984, pp. 45-69.

222
[27] Stabler, G. V., "The Chip Flow Law and its Conse-
quences," in Advances in Machine Tool Design and
Research, Oxford: Pergamon Press, Ltd., 1964, p.
243.

[28] Kalpakjian, S., Manufacturing Engineering and


Technology, Second Edition, Addison-Wesley, New
York, 1992.

[29] Nayfeh, A. H. and Pratt, J. R., "Chatter Iden-


tification and Control for a Boring Process", in
Proc. of IUTAM Symposium on New Applications
of Nonlinear and Chaotic Dynamics in Mechanics,
Ithaca, New York, 1997.

[30] Juang, J.-N. and Phan, M., "Robust Controller


Designs for Second-Order Dynamic Systems: A
Virtual Passive Approach," J. Guid., Cont., Dyn.,
Vol. 15, 1992, pp. 1192-1198.

[31] Goh, C. J. and Caughey, T. K., "On the Stabil-


ity Problem Caused by Finite Actuator Dynamics
in the Control of Large Space Structures," Int. J.
Cont., Vol. 41, 1985, pp. 787-802.

+Y
Figure 2: A schematic showing the location of the ac-

1z
+X
tuators and sensors along the boring bar.

180

0-

·180
-~-
-
.

\ '

~~---·--o
-360
-'--~~

000 1000.00 2000.00 3000.00

Figure 1: The lathe and boring bar showing the tool


overhang. Figure 3: Bode plot of the tangential plant response
function. The dashed line is the experimentally mea-
sured response while the solid line is obtained from the
model.

d(s)

Reference Sensor
Input Output

Figure 4: Block diagram.

223
14

..
g's

50 90 130 170 210 250


'1'III

(a) (b)
Figure 5: Bifurcation diagram showing the amplitude of
the fundamental harmonic of chatter for various spindle
Figure 8: Chip flow over the cutting insert (a) heavy
speeds when w = 0.03" and feedrate = 0.0024 ipr.
cutting and (b) light cutting.

~::~ 1 1.02 1.04


t:l [ .:J 0 500 1000 1500 2000

:::EB
time (sees) freq(Hz)

03 ol ~424Hz
(d)
"0 -20 .
1ii
-40 I
1 1.02 1.04 o~'-=so-:::o:"1!t~ooo~t~5oo~'=2oo':::o::---'
time (sees) freq (Hz)

i~ll~ ,
0 500 1000 1500 2000
freq. (Hz)
1 ] i·:iiii 0 500 1000 1500 2000
freq. (Hz)
Figure 9: Two-degree-of-freedom cutting system.

Figure 6: Time traces, autospectra, cross-spectrum,


and coherence when w = 0.002", speed = 90 rpm, and 0.0026
feedrate = 0.0024 ipr.

Single-degree-of-freedom model

~:~~ !~I ITW


0.0024

1 1.02
time (sec)
1.04 0 500 1000 1500 2000
Iraq (Hz)
w
100~·
:~LG·
0.0022

J
(c)

~ ~ ·.v
·z·.·
olmvvmvmvvvvvmml 0
o
-20
... . .
.
73

. • A Two-degree-of-freedom model
·100'-------:-:-:-----::-c--'
1 1.02 1.04 0 500 1000 1500 2000
time (seco) freq(Hz)
0.002-----

2.80 2.84 2.88 2.92


1/'t

Figure 7: Time traces, autospectra, cross-spectrum, Figure 10: Stability lobes for the two-degree-of-freedom
and coherence when w = 0.03", speed = 210 rpm, and cutting system
feedrate =0.0024 ipr.

224
201

J 2 3 5 6 8 10 11

1500

-100

-200
0

500 1000 1500


I : ".-----.-'
2:fr----r---" >~:~.----.----,-
·~ :~..I ,. "
- 2 ~L.4--=2L.5--=2L.6--=2=-7--:2L.8--:2"c:-9-'--:30"c----,3J-.1_-:32:':----,33:':-___J34

Figure 13: Time traces of the tool accelerations normal


Figure 11: Compensated tangential plant response
function. to the machined surface; initially, only tangential con-
trol is active; then, in addition, control is applied in the
normal direction to quench the chatter and eliminate
the jump-type instability; w = 0.002", speed = 170
rpm, and feedrate = 0.0024 ipr.

~:~E:1;lJ:X.iJJ
1
.· •. 1.02
time (sees)
1.04
"C ~50

-100
0
.
500 1000 1500 2000
freq (Hz)

i]llll i'llG
1 1.02
time (sees)
1.04 -10oo:--::5-:.-:oo~1=oo7o -=1-="5oo:-::=2:"o"'=oo="
freq (Hz)
10
time (sees)
12 14 16 18 20

•Jl I i·Yjg U
0 500 1000 1500 2000
freq. (Hz)
0 500 1000 1500 2000
freq. (Hz)

10 12 14 16 18 20
Figure 12: Simulated time traces, autospectra, cross- time (sees)

spectrum, and coherence of the normal and tangential


tool accelerations when w = 0.03", speed = 210 rpm, Figure 14: Time traces of the normal and tangential
and feedrate = 0.0024 ipr. tool accelerations; w = 0.03", speed = 210 rpm, and
feedrate = 0.0024 ipr.

225

View publication stats

You might also like