SANDIA REPORT
SAND2001-3031
Unlimited Release
Printed September 2001
Mitigation of Chatter Instabilities in
Milling by Active Structural Control
Jeffrey L. Dohner, James P. Lauffer, Terry D. Hinnerichs, Natarajan Shankar,
Mark Regelbrugge, Chi-Man Kwan and Roger Xu, Bill Winterbauer, Keith Bridger
Prepared by
Sandia National Laboratories
Albuquerque, New Mexico 87185 and Livermore, California 94550
Sandia is a multiprogram laboratory operated by Sandia Corporation,
a Lockheed Martin Company, for the United States Department of
Energy under Contract DE-AC04-94AL85000.
Approved for public release; further dissemination unlimited.
Issued by Sandia National Laboratories, operated for the United States Department
of Energy by Sandia Corporation.
NOTICE: This report was prepared as an account of work sponsored by an agency
of the United States Government. Neither the United States Government, nor any
agency thereof, nor any of their employees, nor any of their contractors,
subcontractors, or their employees, make any warranty, express or implied, or
assume any legal liability or responsibility for the accuracy, completeness, or
usefulness of any information, apparatus, product, or process disclosed, or represent
that its use would not infringe privately owned rights. Reference herein to any
specific commercial product, process, or service by trade name, trademark,
manufacturer, or otherwise, does not necessarily constitute or imply its endorsement,
recommendation, or favoring by the United States Government, any agency thereof,
or any of their contractors or subcontractors. The views and opinions expressed
herein do not necessarily state or reflect those of the United States Government, any
agency thereof, or any of their contractors.
Printed in the United States of America. This report has been reproduced directly
from the best available copy.
Available to DOE and DOE contractors from
U.S. Department of Energy
Office of Scientific and Technical Information
P.O. Box 62
Oak Ridge, TN 37831
Telephone: (865)576-8401
Facsimile: (865)576-5728
reports@adonis.osti.gov
E-Mail:
Online ordering: http://www.doe.gov/bridge
Available to the public from
U.S. Department of Commerce
National Technical Information Service
5285 Port Royal Rd
Springfield, VA 22161
Telephone: (800)553-6847
Facsimile: (703)605-6900
orders@ntis.fedworld.gov
E-Mail:
Online order: http://www.ntis.gov/ordering.htm
SAND2001-3031
Unlimited Release
Printed September 2001
Mitigation of Chatter Instabilities in Milling
by Active Structural Control
This work was performed by a research team led by Lockheed Martin and funded by DARPA.
Team members and their organizations are given below.
Jeffrey L. Dohner
Mark Regelbrugge (technical lead)
MEMS Device Technologies Dept.
Rhombus Consultants Group, Inc.
2565 Leghorn St., Mountain View, CA 94043
James P. Lauffer
Engr. Mechanics, Modeling & Simulation Dept.
Chi-Man Kwan and Roger Xu
Terry D. Hinnerichs
Intelligent Automation, Inc.
7519 Standish Place, Rockville, MD 20855
Solid Mechanics Engineering Dept.
Bill Winterbauer
Sandia National Laboratories
P.O. Box 5800, Albuquerque, NM 87185-1080
Ingersoll Milling Machine Company,
707 Fulton Avenue, Rockford, IL 61103
Natarajan Shankar (project lead)
Keith Bridger
Lockheed Martin Space Systems Company
3251 Hanover St., Palo Alto, CA 94304
Active Signal Technology, Inc.
13027-A Beaver Dam Road
Cockeysville, MD 21030
Abstract
This report documents how active structural control was used to significantly
enhance the metal removal rate of a milling machine. An active structural control
system integrates actuators, sensors, a control law and a processor into a structure
for the purpose of improving the dynamic characteristics of the structure. Sensors
measure motion, and the control law, implemented in the processor, relates this
motion to actuator forces. Closed-loop dynamics can be enhanced by proper control law design.
Actuators and sensors were imbedded within a milling machine for the purpose of
modifying dynamics in such a way that mechanical energy, produced during cutting, was absorbed. This limited the on-set of instabilities and allowed for greater
depths of cut. Up to an order of magnitude improvement in metal removal rate was
achieved using this system.
Although demonstrations were very successful, the development of an industrial
prototype awaits improvements in the technology. In particular, simpler system
designs that assure controllability and observability and control algorithms that
allow for adaptability need to be developed.
3
Intentionally Left Blank
Acknowledgements
The authors would like to thank the following individuals for their support. Dennis Bray, Ingersoll Milling Machine Company, contributed a
tremendous amount of insight and understanding to this problem.
Without his support, this project would not have been successful. Bob
Winfough, (formerly at Ingersoll), was instrumental in helping team
members understand the details behind the regenerative chatter problem. David Martinez, Sandia National Laboratories, was also a great
supporter of this project. His support in times of uncertainty was
greatly appreciated. Others who gave significantly to this project were
Leonard Haynes, IAI, James Handrock, Brian Driessen, Jim Redmond,
and Karen Archibeque, Sandia Laboratories.
Thanks also goes to DARPA for their funding and programmatic guidance, and to the USAF for their efforts as contract managers. Special
thanks goes to Lockheed Martin Space Systems Co. for their Request
for Loan of Personnel, RFL #99-36.
5
Intentionally Left Blank
Table of Contents:
Abstract................................................................................................................... 3
Acknowledgments.................................................................................................. 5
Table of Contents................................................................................................... 7
List of Figures........................................................................................................ 7
Mitigation of Chatter Instabilities in Milling
by Active Vibration Control
Introduction.......................................................................................................... 9
Hardware Design.................................................................................................. 9
Characterization.................................................................................................. 12
Control Design.................................................................................................... 14
Results................................................................................................................. 17
Conclusions.......................................................................................................... 20
References............................................................................................................. 21
List of Figures:
Figure 1. Hardware Configuration....................................................................... 11
Figure 2. Modified Tool and Smart Spindle Unit.................................................. 12
Figure 3. Maximum Singular Value of FRFs Showing the Location of the Tool
Mode................................................................................................. 13
Figure 4. Maximum Singular Value of FRFs for Rotating and Stationary
Spindles............................................................................................ 13
Figure 5. Controller Function in Block Diagram Form........................................ 14
Figure 6. Frequency Response Function, Magnitude and Phase, With and Without
Control.............................................................................................. 15
Figure 7. Nyquist Diagram of Loop Gain............................................................. 16
Figure 8a. Unstable and Stable Dynamic Response (time domain), Strain
Response for a 0.01 mm Depth of Cut at 3600 rpm.......................... 18
Figure 8b. Power Spectral Densities of Unstable and Stable Dynamics Responses
Shown in Figure 8a............................................................................ 18
Figure 9. Stability Limits of a Typical Machine Tool........................................... 19
Figure 10. Enhancement in Stability Due to Active Control................................. 20
7
Intentionally Left Blank
I
Milling machines are designed for a set
of tools with diameters similar to the
diameter of the spindle. When a tool
has a diameter that is much less than
the diameter of the spindle, a mechanical impedance mismatch occurs
between the tool and the spindle, and
mechanical energy can be trapped
within the tool (i.e. a tool mode). During cutting, these tool modes can
become unstable. Larger diameter tools
usually do not experience this phenomenon due to the power limitations of the
machine. Other researchers have developed methods to avoid these instabilities by varying spindle speed (see
Jemielniak and Widota, 1984,Altintas
and Chan, 1992, and CRAC, 1992).
However, this often requires moving to
higher or lower spindle speeds where
MMRR is decreased (i.e. lower spindle
speeds) or where tool wear is increased
(i.e. high spindle speeds).
ntroduction
The following is a discussion of the
design, analysis, and testing performed
to demonstrate the value of using active
vibration control to enhance the productive capacity of a milling machine.
Maximum Metal Removal Rate
(MMRR) is a quantitative measure of
the productive capacity of a machine
tool. The MMRR is limited by several
factors including the onset of machining instabilities that are a function of
the vibratory modes of both the
machine and the tool (see Merritt,
1965, Tlusty and Ismail, 1983, and Shi
and Tobias, 1984).
By altering the dynamic characteristics
of these modes, instabilities could be
mitigated and MMRR improved. Alteration could be achieved by physically
redesigning or modifying the structure
of the machine; however, an alternative
approach is to use an active control system to alter the system dynamics. In an
active control system, actuators, sensors, computers and software replace
mechanical components to provide the
desired dynamic response characteristics.
Actively changing the dynamics of a
machine can enhance MMRR without
the need for changing spindle speeds.
The following is a discussion of an
alternate approach - the development of
hardware and software constructed to
enhance the MMRR of a machine tool
using an active structural control system.
Examples of the use of active systems
to alter dynamics during machining are
relatively few, see Comstock, Tse, and
Lemon, 1969, Sadek and Tobias, 1973,
Hong-Yeon Hwang, Jun-Ho Oh, and
Kwang-Joon Kim, 1988, and Shiraishi,
Yamanaka, and Fujita, 1990, General
Dynamics Team 1998, and Redmond,
Barney, and Smith, 1999. In general,
most of these papers address turning,
not milling.
H
ardware Design
An illustration of the hardware
designed and constructed to demonstrate the utility of active control is
shown in Figure 1. Much of this hardware was constructed and assembled at
the Ingersoll Milling Machine Company (Rockford IL).
9
As shown in Figure 1, vibration is
sensed at the root of a rotating tool by
strain gages that are arranged in half
bridge configurations to sense bending
in two lateral directions. Excitation
voltages were supplied to the half
bridges using commercial electronics.
Power is supplied to these electronics
via magnetic coupling between rotating
and stationary wires.
For our application, the control law
operates on stationary strain signals
and feeds back activation signals to the
actuators. Control laws were designed
to absorb energy from the rotating tool
thereby, reducing entrapped energy.
This absorption of energy increases the
stability of the cutting process and
improves MMRR. The control law is
defined by programmable logic.
A telemetry system is used to transmit
strain data from the rotating spindle to
stationary receivers. Both the rotating
and stationary electronics were fabricated by the Wireless Data Corporation
(see Wireless Data).
As shown in Figure 1, the controller
produces four voltage signals that drive
a set of four power amplifiers. These
power amplifiers drive stacks of electrostrictive material (PMN) embedded
within the housing of the machine.
These stacks (produced by Lockheed
Martin Corporation and integrated into
the housing by Active Signal Inc.) produce force against a non-rotating portion of the machine called the cartridge.
The spindle moves with the center line
of the cartridge and floats on a hydrostatic bearing. Thus, forces on the cartridge produce motions in the tool.
Motions corresponding to tool bending
can be sensed by the strain gages and
fed back into the control system.
Strain is measured in a coordinate system that rotates with the shaft,
( x, y, z ) , however, actuation occurs in
a coordinate system that is stationary
with the machine, ( X, Y, Z ) . Therefore, the strain data must be translated
from rotating to stationary coordinates.
To do this, the angular position of the
spindle is measured using a decoder.
Decoder and strain-gage-bridge voltages are fed into anti-aliasing filters,
analog to digital converters (A/D), and
a processor for the computation of this
transformation. The anti-aliasing filters, the A/Ds, and the processor are
part of a component called a controller.
Hardware, consisting of the spindle, the
tool holder, the cartridge, the actuators,
part of the telemetry system, and much
of the surrounding housing was given
the special name - the Smart Spindle
Unit (SSU). A photograph of the SSU
is shown in Figure 2. Separate from the
SSU are the power amplifiers, the controller, and the telemetry package
receiver.
A controller is a hardware component
with the ability to capture voltage signals, combine them in accordance with
a defined mathematical relationship,
and output the result as another set of
voltage signals. Intelligent Automation
Inc. (Rockville, MD) designed and fabricated the controller discussed in this
paper.
10
hydrostatic bearing
Z
z
electrostrictive
actuators
ω
y
X
Y
x
telemetry
power supply
and receiver
antenna
(stationary)
cartridge
ball bearings
spindle
(rotating)
telemetry package
and bridge electronics
(rotating)
strain gages
at tool root
power amplifiers
cutting tool
cutting
excitation
controller
Figure 1. Hardware Configuration
11
telemetry receiver
cross section of Smart Spindle Unit (SSU)
decoder
electrostrictive
actuator housing
telemetry
power supply
and receiver
antenna
(stationary)
strain gages
at tool root
cutting tool
Figure 2. Modified Tool and Smart Spindle Unit
C
of the Maximum Singular Values
(MSVs) of these FRFs. The MSVs
give a bound of the FRF response of
the system. From the MSVs the
modes of the system can be identified. The first “tool” mode occurred
at about 800Hz, however, as shown
in this figure, it participated little in
the response. Thus, a state space
model derived from the measured
FRFs would not be controllable or
observable.
haracterization
Without the benefit of previous
design data, the SSU design relied
heavily on the use of numerical analysis (see Dohner et. al., 1997).
Although this allowed for enough
insight to complete an initial design,
a full characterization of dynamics
was required upon fabrication.
Initial experimental analysis of the
SSU showed that system dynamics
were not controllable or observable
(see Kwakernaak and Sivan, 1972 for
an explanation of controllability and
observability). Frequency Response
Functions (FRFs) were measured
between voltage inputs to the power
amplifiers and tool strain responses
in stationary coordinates. Initial measurements were made with the spindle at rest (0 rpm). Figure 3, is a plot
The reason the system was neither
controllable nor observable was
because of an anti-resonance in the
FRFs between the actuators and
strain sensors. The frequency of the
anti-resonance occurred at virtually
the same frequency as the fundamental mode of the tool. The anti-resonance was due to modal cancellation
between rigid-body modes of the cartridge/spindle system.
12
MSV (volts/volts)
tool mode response
in modified system
unmodified system
modified system
1.00
0.10
region of tool mode
in unmodified system
0.01
0
500
1000
2000
1500
frequency (Hz)
Figure 3. Maximum Singular Value of FRFs Showing the Location of the Tool Mode
Controllability and observability
could have been achieved by shifting
the resonant frequencies of the cartridge and spindle to frequencies
above the fundamental frequency of
the tool; however, to do this would
have required a complete redesign of
the SSU, and such modifications
were beyond available budget and
time.
option 1) fabricate a longer more
flexible tool with a
lower fundamental frequency, or
option 2) modify the existing tool
to lower its fundamental
frequency.
Due to budget constraints, the second
option was chosen. A mass was
added close to the end of the existing
tool to move the fundamental tool
mode away from any rigid-body anti-
MSV (volt/volts)
Therefore, two options were available:
rotating spindle
stationary spindle
1.00
0.10
0.01
0
500
1000
1500
2000
frequency (Hz)
Figure 4. Maximum Singular Value of FRFs for Rotating and Stationary Spindles
13
A/D antialiasing
Translate from the rotating
coordinate system ( x, y, z )
to the stationary coordinate
system
voltage from
decoder
(spindle location)
voltages from
receiver
(strain signals)
( X , Y, Z )
D/A
voltages
into power
amplifiers
-1
control law
x c ( k + 1 ) = Ac xc ( k ) + B c y ( k )
u ( k ) = Cc xc ( k )
-1
y( k)
u(k)
Figure 5. Controller Function in Block Diagram Form
resonance. Subsequent FRFs are
shown in Figure 3. As shown, the
tool mode (now clearly visible) was
shifted from 800 Hz down to 453 Hz.
The resulting realization (state space
model) of the modified system was
both controllable and observable.
mode appears to be invariant with
respect to rotation speed. Therefore,
the same control law can be used
regardless of the spindle rotation
speed.
Additional measurements were performed to determine how FRFs varied with rotation. In Figure 4 the
maximum singular value of the actuator to strain FRFs for both rotating
(3000 RPM) and stationary conditions are shown. The main difference
between these plots is the presence of
harmonics at multiples of the rotational speed of the spindle. These
harmonics are artifacts due to bearing
inputs, out-of-roundness, and balancing that could have been reduced by
using more ensemble averages when
estimating the FRFs.
C
ontrol Design
Control design was performed as a
two step process.
step 1) The production of a
reduced order realization
of dynamics, and
step 2) The design of a robust controller.
Figure 5 shows controller logic.
Three voltage signals are fed into the
controller - two voltage signals from
the receiver and a voltage signal from
the decoder. These signals are passed
through anti-aliasing filters and are
More importantly, Figure 4 shows
that the dynamics of the fundamental
14
then sampled. The result is a numerical data train representing tool strain,
in rotating coordinates, ( x, y, z ) , and
spindle location. This data is combined to calculate tool strain in the
stationary coordinate system
( X, Y, Z ) (as discussed above).
u ( k ) = Cc xc ( k )
where A c ∈ ℜ
is the con2xn
is the
controller output matrix, and n is the
number of states in the controller (see
Kwakernaak and Sivan 1972). For
this application, the control law was
designed to absorb energy from the
system; consequently, closed-loop
tool dynamics are more heavily
damped than open-loop tool dynamics. The state, input and output matrices are chosen in such a way that this
form of energy absorption occurs.
εX ( k )
εY ( k )
stationary, strain data in the X and Y
planes. This vector is used by the
control law to compute the outputs of
the controller. The control law takes
the form
Using the control law and the data
train, y ( k ), y ( k – 1 ), y ( k – 2 )… , the output vector,
x c ( k + 1 ) = A c x c ( k ) + B c y ( k ) (1.a)
MSV (volts/volts)
nx2
troller input matrix, C c ∈ ℜ
where ε X ( k ) and ε Y ( k ) is sampled,
1.000
uncontrolled response
controlled response
0.100
0.010
Magnitude
0.001
0
(degrees)
is the controller
state matrix, B c ∈ ℜ
At sample time k , stationary strain
data is given in vector form by
y(k) =
nxn
(1.b)
-500
-1000
Phase
-1500
0
500
1000
1500
2000
frequency (Hz)
Figure 6. Frequency Response Function, Magnitude and Phase, With and
Without Control
15
u(k) =
u1 ( k )
u2 ( k )
tions. The algorithm used in this
effort was the Eigensystem Realization Algorithm with Direct Correlations, ERA/DC, (see Juang, 1994).
Neglecting any direct feed through
effects, this algorithm produces a
realization of the form
,
can be calculated. The output vector
data train is converted into two analog voltage signals by a set of digital
to analog converters (D/A). Because
of the cruciform configuration of the
SSU, actuators on either side of the
cartridge were assumed to move the
same amount; therefore, voltage signals into the power amplifiers can be
formed by splitting each D/A output
voltage and changing the sign on one
of the signals.
x ( k + 1 ) = Ax ( k ) + Bu ( k )
(2.a)
y ( k ) = Cx ( k )
(2.b)
where A ∈ ℜ
nxn
matrix, B ∈ ℜ
is the plant state
nx2
is the plant input
2xn
matrix, and C ∈ ℜ
is the plant
output matrix. Again, n is the number of states.
In order to choose a controller state,
input, and output matrix that will
damp tool motion, a mathematical
The control law, equation 1a,b, is a
realization of dynamics from u ( k ) to
mathematical relationship from y ( k )
y ( k ) must be produced. This realization is often referred to as the plant.
A variety of algorithms can be used
to produce a plant realization from
measured frequency response func-
to u ( k ) . The plant, equation 2a,b, is
a mathematical relationship from
u ( k ) to y ( k ) . Initially, a Linear Quadratic Gaussian (LQG) approach (see
> 90 deg.
1 phase margin
427Hz
Notice: The Loop Gain is greater
than 1.0 only for frequencies around
the fundamental tool frequency.
0
-1
475Hz
-2
-2
-1
1
0
real part
2
imaginary part
imaginary part
2
4
2
0
-2
-4
-2
0
2
4
real part
Figure 7. Nyquist Diagram of Loop Gain
16
6
8
Kc s + α
where Θ ( s ) = -----------------------------------------.
2
2
s + 2ζ c ω c s + ω c
Notice that this is a Positive Real (PR)
control law even through the plant was
not PR. Nevertheless, because it looks
unimodal in any one direction, the PR
control law was adequate to produce
high levels of performance with sufficient levels of robustness.
Kawakernak and Sivan) was used to
determine the state, input and output
matrices of the control law. As with
most uses of LQG, the weighting
matrices used to define performance
were manipulated to shape the loop
until a sufficient balance between performance and robustness was achieved.
Surprisingly, in doing this, it was found
that LQG produced high levels of both
robustness and performance. This
occurred even for low order plant realizations ( n = 4 ) .
To better understand this, notice that
for ω ≈ ω c and moderate values of K s
and K c the loop gain, Θ ( s )Γ ( s ) , is
greater than 1.0 only for frequencies
near to ω . At all other frequencies the
closed loop system is gain stabilized.
This has a significant influence on the
Nyquist diagram of the loop gain. Figure 7 shows the Nyquist diagram of the
loop gain for a typical control law. The
Nyquist diagram contains a single lobe
that occurs near the fundamental frequency of the tool. The rotation and
size of this lobe is controlled by the
parameters K c , ζ c , and α . LQG
selects these parameters such that the
lobe is always deep in the right hand
plane of the Nyquist diagram. This
gives high loop gains and therefore
good performance while also maintaining robustness (over 90 degrees of
phase margin at high gain margins).
Thus, for this plant, LQG was able to
produce a robust, high performance,
low order control law. The final control
law contained only four states.
After examination of the control law
and plant, a better understanding of
control dynamics was developed. Figure 6 shows a Bode plot of the plant
with and without control. Notice that
there is one dominant mode in this system at 450Hz. This is the tool mode.
This tool mode has a peak response
that is almost an order of magnitude
greater than the response of any other
mode in the system. Therefore, the
plant can be approximated as a second
order system cascaded with all pass
dynamics (see Oppenheim, A.V., Schafer, R.W., 1975). Considering symmetry and neglecting cross coupling, the
plant can be approximated by
H ( s ) = C ( Is – A )
–1
B =
Γ(s) 0
0 Γ(s)
φ ( s )Ks
-,
where Γ ( s ) = ----------------------------------------2
2
s + 2 ζω s + ω
n
n
φ ( s ) = 1 and s is the Laplace transform variable.
The LQG approach produced controllers that can be approximated by
G(s)
= C c ( Is – A c )
–1
Bc =
Θ(s)
0
R
esults
0
Θ(s)
Chatter instabilities occur during cutting due to dynamic feedback between
17
4
(volts)
2
0
-2
with Control off (Chatter)
-4
0
5
10
15
time (seconds)
4
(volts)
2
0
-2
with Control on (no Chatter)
-4
0
5
time (seconds)
10
15
Figure 8a. Unstable and Stable Dynamic Response (time domain),
Strain Response for a 0.01 mm Depth of Cut at 3600 rpm.
10
10
2
volts
------------- Hz -
10
10
10
0
cross over frequency
near tool mode resonance
chatter instability
-2
with Control off
with Control on
-4
-6
-8
0
200
400
600
800
1000
frequency (Hz)
Figure 8b. Power Spectral Densities of Unstable and Stable Dynamic
Responses Shown in Figure 8a
18
maximum stable
depth of cut
process damping effect
lobing effect
unstable region
stable region
n min
nmax
spindle speed
Figure 9. Stability Limits of a Typical Machine Tool
tool inserts. A cutting tool can have a
number of inserts. As an insert cuts
through metal it lays down a pattern,
and this pattern affects the cut of the
next insert on the tool. This interaction
creates a dynamic feedback path
between successive cuts, and, as in
many feedback systems, can lead to
instability. During cutting, energy is
pumped into well coupled modes. If the
pattern on the part is great enough that
energy gain is not balanced by energy
loss, energy storage will grow; thereby,
producing the dynamic instability
known as regenerative chatter.
Figure 8a,b show the response of the
strain gages for a tool in chatter and a
tool not in chatter. Notice that chatter
can produce over an order of magnitude change in the dynamic response of
the tool for a 0.01 mm depth of cut at
3600 rpm spindle speed.
Figure 9 is a cartoon of the stability
limits of a hypothetical machine and
tool. The area below the curve is stable
and the area above the curve is unstable
(unstable is hatched). Notice, that at
low spindle speeds, large depths of cut
can be taken; however, metal removal
rate, MRR, is low due to low rotational
speeds. At higher rotational speed, lobing exists. This lobing represents
regions of stability. Operating within
these lobes will produce high MRRs;
however, for many poorly thermally
conducting materials, this will also
result in high tool wear caused by elevated temperatures. Therefore, most
machine tools operate at intermediate
spindle speeds between spindle speeds
Cutting instabilities can pump enough
vibrational energy into the tool to eject
the insert from the part. At ejection, the
forces on the tool are relieved and the
insert bounces back into the metal. This
ejection and reimmersion creates a
non-linear dynamic limit cycle process
that results in severe vibration in the
machine and poor surface finish.
19
maximum stable depth of cut (mm)
0.45
0.4
0.4
0.35
0.3
0.3
0.25
0.2
0.2
Enhancement in Stability
Due to Active Control
0.15
0.1
0.1
0.05
0.00 0
0
500
1000
1000
2000
2000
1500
n min
2500
3000
3000
3500
spindle speed (rpm)
4000
4000
4500
n max
Figure 10. Enhancement in Stability Due to Active Control
n min and n max . For the machine discussed in this paper, nmin ≈ 700rpm
C
and n max ≈ 4000rpm .
onclusions
The stability limits of the machine
shown in Figure 2 were determined for
control on and control off. Per ANSI/
ASME standards, chip loading for full
immersion cutting was held to 0.1mm/
insert. Figure 10 shows the change in
machine stability due to control. As
shown, an order of magnitude increase
in the maximum stable depth of cut
occurred. Active vibration control significantly increased the cutting performance of this machine tool.
This project demonstrated that active
structural control can be used to
increase the MMRR of a milling
machine by more than an order of magnitude. Although these results are very
promising, there are still practical limitations to this technology. In particular,
better methods of design, and control
are required.
The design of machines that can properly leverage the use of active control
is an evolving area of study. These
machines must be designed to allow for
the full observability and controllability of the vibration modes of interest.
For this effort, the tool was altered to
overcome this problem, however, in a
Other cutting tests (quarter and half
immersion tests) demonstrated
improved MMRR with lower levels of
performance. For these tests, the maximum stable depth of cut increased by
factors of 4 to 5.
20
CRAC, Chatter Recognition and Control System, 1992, MLI Manufacturing Laboratories, Inc. Gainesville,
Florida, U.S. Patent No. 5,170,358.
more mature system, this type of alteration would not be acceptable.
Because actuation and sensing
occurred in two separate coordinate
systems, one rotating and the other stationary, the control system was needlessly complex. This complex control
system, necessary in a proof of concept
experiment could be significantly simplified in a production system; such as
its packaging in a compact tool holder
configuration (see General Dynamics,
1998).
Dohner, J.L., Hinnerichs, T.D., Lauffer,
J.P., Kwan, C.M., Regelbrugge, M.E.,
and Shankar, N., 1997, Proceedings
of SPIE, Smart Structures and Materials, Industrial and Commercial
Applications of Smart Structures
Technologies, March.
General Dynamics, Advanced Technology Systems, Guilford Center, 1998
"Precision Machining - Active Structural Control Milling Concept Demonstration System", Air Force
Research Laboratory Report, AFRLML-WP-TR-1998-4118, June.
Active control changes the dynamics of
the machine such that chatter instabilities occur at much higher depths of cut.
At present, this requires the intervention of an operator. However, theoretically, the machine could be given
sufficient intelligence to make these
changes on its own.
Hong-Yeon Hwang, Jun-Ho Oh, and
Kwang-Joon Kim, 1988, "Modeling
and Adaptive Pole Assignment Control in Turning", International Journal of Machine Tool Design Research,
vol. 29, no. 2, pp. 275-285.
R
Jemielniak, K., and Widota, A., 1984,
“Suppression of Self-Excited Vibration by the Spindle Speed Variation
Method”, International Journal of
Machine Tool Design Res., vol. 24,
no. 3, pp 207-214,
eferences
Altintas, T. and Chan, P.K., 1992, "In
Process Detection and Suppression of
Chatter in Milling", International
Journal of Machine Tools Manufacturing, vol. 32, no. 3, pp329-347.
Juang, Jer-Nan,. 1994, Applied System
Identification, Englewood Cliffs,
N.J., Prentice Hall.
Comstock, T.R., Tse, F.S., Lemon, J.R.,
1969, "Application of Controlled
Mechanical Impedance for Reducing
Machine Tool Vibrations", Journal of
Engineering for Industry, Transactions of the ASME, pp. 1057-1062.
Kwakernaak, H., and Sivan, R., 1972,
Linear Optimal Control Systems,
Wiley-Interscience, New York, N.Y.
Merritt, H.E., 1965, "Theory of SelfExcited Machine-Tool Chatter, Con21
tribution to Machine-Tool Chatter,
Research--1", Journal for Engineering for Industry, Transactions for the
ASME, pp. 447-453, November.
Oppenheim, A.V., Schafer, R.W., 1975,
Digital Signal Processing, PrenticeHall, Inc. Englewood Cliffs, NF.
Redmond, J., Barney, P., and Smith, D.,
1999 “A Biaxial Actively Damped
Boring Bar for Chatter Mitigation”,
The International Journal for Manufacturing Science and Production,
Vol. 2, No. 1, pp. 1-16.
Sadek, M.M., Tobias, S.A., 1973,
"Reduction of Machine Tool Vibration", American Society of Mechanical Engineering, Applied Mechanics
Division, v1, for Meeting, Cincinnati,
Ohio, pp. 128-172, September.
Shi, H.M. and Tobias, S.A., 1984,
"Theory of Finite Amplitude Machine
Tool Instability", International Journal of Machine Tool Design Research,
vol. 24, No.1. pp. 45-69.
Shiraishi, M., Yamanaka, K., and
Fujita, H., 1991 "Optimal Control of
Chatter in Turning", International.
Journal of Machine Tools in Manufacturing, vol. 31, No. 1, pp.31-43.
Tlusty, J., and Ismail, F., 1983, "Special
Aspects of Chatter in Milling", Journal of Vibration, Acoustics, Stress,
and Reliability in Design, Transactions of ASME, vol. 105, pp. 24-32,
January.
Wireless Data Corporation, 620 Clyde
Avenue, Mountain View, CA 94043
22
D
istribution List
MS9405 8700 R. H. Stulen
MS9402 8724 K. Wilson
MS9403 8723 J. Wang
MS9161 8726 Er-Ping Chen
MS9042 8727 J. L. Handrock
MS9042 8727 J.P. Lauffer (10)
Bill Winterbauer
Ingersoll Milling Machine Company
707 Fulton Avenue
Rockford, IL 61103
Keith Bridger
Active Signal Technology, Inc.
13027-A Beaver Dam Road
Cockeysville, MD 21030
MS1080 1749 J.F. Jakubczak II
MS1080 1749 J. J. Allen
MS1080 1749 J.L. Dohner (10)
Dr. Jer-Nan Juang
NASA Langley Research Center
Mail Stop: 297
Hampton, VA 23665-5225
MS0892 1715 S.J. Martin
MS0847 9120 H.S. Morgan
MS0847 9124 D.R. Martinez
MS0847 9124 J.M. Redmond
MS0847 9124 D.J. Segalman
MS0847 9124 C.R. Dohrmann
MS0847 9124 R.V. Field
MS0847 9124 D.W. Lobitz
MS0847 9126 R. May
MS0847 9126 T.D. Hinnerichs (5)
Richard Pappa
NASA Langley Research Center
Mail Stop: 230
Hampton, VA 23665-5225
Dr. Roy R. Craig
University of Texas at Austin
ASE-EM Department
Austin, TX 78712-1085
Dr. Robert E. Skelton
MS0835 9140 J. M. McGlaun
MS0828 9100 T.C. Bickel
MS0824 9130 J. L. Moya
MS0824 9110 A. C. Ratzel
Prof. of Aeronautical & Astronautical Engineering
323 Grissom Hall
West Lafayette, IN 47907
MS9018 8945-1 Central Tech. Files
MS0899 9616 Tech. Library (2)
MS0612 9612 Review & Approval
Desk (1) For DOE/OST I
MS0557 9125 T.J. Baca
MS0501 2338 Ming Lau
MS0501 2338 P.S. Barney
Natarajan Shankar
Lockheed Martin Space Systems Company
3251 Hanover St.
Palo Alto, CA 94304
Mark Regelbrugge
Rhombus Consultants Group, Inc.
2565 Leghorn St.
Mountain View, CA 94043
Chi-Man Kwan and Roger Xu
Intelligent Automation, Inc.
2 Research Place, Suite 202
Rockville, MD 20850
23
Intentionally Left Blank
- NOTES -