Dynamic Characteristics of a

Tool Holder Shank in Lathe

Carry Andersson
Lennart Kisswani
Department of Mechanical Engineering
University of Karlskrona/Ronneby
Karlskrona, Sweden
2000
Thesis submitted for completion of Master of Science in Mechanical
Engineering with emphasis on Structural Mechanics at the Department of
Mechanical Engineering, University of Karlskrona/Ronneby, Karlskrona,
Sweden.
Abstract:
Dynamic characteristics of a Tool Holder Shank were determined by using
the Finite Element Method and Experimental Modal Analysis. Good
agreement between FE-model and the experimental model was obtained for
the Tool Holder Shank itself. With the Tool Holder Shank in Lathe the
agreement was less good. Reasons for this are discussed. Suggestions for
further work on including actuators to reduce vibrations are given.
Keywords:
Dynamic Characteristics, Tool Holder Shank, Lathe, Turning, Actuators, FEmodel, Modal Analysis, Experimental Verification.

Acknowledgements
This work was carried out at the Department of Mechanical Engineering,
University of Karlskrona/Ronneby, Karlskrona, Sweden, under the
supervision of Dr Mats Walter. We wish to express our sincere appreciation
for his guidance throughout the work.
The work was initiated in July 1999 as a co-operation project between the
Department of Telecommunications and Signal Processing and the
Department of Mechanical Engineering at the University of
Karlskrona/Ronneby.
We also wish to express our sincere appreciation to Dr Lars Hkansson at
the Department of Telecommunications and Signal Processing who acted as
a co-supervisor, as well as the whole department for their help and support
throughout the work.
Finally, we would like to thank our colleagues in the Master of Science
programme and all the members of the Department of Mechanical
Engineering for valuable discussions and support.
Karlskrona, February 2000
Carry Andersson
Lennart Kisswani

Contents
1. Notation

2 Introduction
2.1 Background
2.2 Project Description
2.3 Method

7
7
8
8

3 Theories
3.1 Turning
3.1.1 The Turning Operation
3.1.2 Undesired Vibrations and Noise
3.2 Actuators
3.2.1 Piezoelectric Actuators
3.3 Dynamics
3.3.1 Multiple Degree of Freedom (MDOF) Systems
3.3.2 Natural Frequencies and Mode Shapes
3.4 Modal Analysis
3.4.1 Mathematical Models
3.4.2 Experimental Modal Analysis
3.5 Eigenvalues and Eigenvectors in I-DEAS
3.5.1 Lanczos Formulation
3.5.2 Lanczos Algorithm
3.6 Signal Processing and Vibration Testing
3.6.1 Transfer Functions used in Vibration Measurement
3.6.2 Coherence
3.7 Correlation
3.7.1 Comparison of Natural Frequencies
3.7.2 Comparison of Mode Shapes

10
10
10
10
12
13
14
14
15
15
16
16
17
17
17
19
19
20
21
21
22

4 Theoretical Models
4.1 Elements
4.1.1 Beam Elements
4.1.2 Shell Elements
4.1.3 Solid Elements
4.2. Boundary Conditions
4.2.1 Free-Free
4.3 The Tool Holder Shank

24
24
24
24
24
25
25
25

4.4 The Tool Holder Shank Free-Free

4.5 The Tool Holder and Tool Holder Shank Free-Free
4.5.1 With 40 mm Hang Out
4.5.2 With 60 mm Hang Out
4.6 The Tool Holder and Tool Holder Shank in Concrete Lump
4.6.1 With 40 mm Hang Out
4.6.2 With 60 mm Hang Out
4.7 The Tool Holder and Tool Holder Shank in Lathe
4.7.1 With 40 mm Hang Out
4.7.2 With 60 mm Hang Out

26
27
27
28
28
28
29
29
29
29

5 Experimental Procedures
5.2 Excitation Methods and Requirements
5.3 Equipment in All Experimental Measurements
5.4 The Tool Holder Shank Free-Free
5.5 The Tool Holder and Tool Holder Shank Free-Free
5.6 The Tool Holder and Tool Holder Shank in Concrete Lump
5.6.1 Measurements with Accelerometers
5.6.2 Measurements with Laser Vibrometer
5.7 The Tool Holder and Tool Holder Shank in Lathe

30
30
31
31
32
34
34
35
36

6 Results

38

7 Conclusions
7.1 Complexity of Problems
7.2 The Continuation of the Analysis of the Tool Holder Shank

41
41
45

8 References

46

Appendix A

47

Appendix B

49

Appendix C

52

Appendix D

55

1. Notation
I
K
M
U
K
F
H
K
L
M
X
r
u
x
c
f
i
j
k
m
n
s
t
u
v
x
y
z

Sxx

Identity matrix
Stiffness matrix [N/m]
Mass matrix [kg]
Mass normalised eigenvectors
Shifted stiffness matrix [N/m]
Input force signal
Transfer function
Stiffness matrix [N/m]
Length [m]
Moment [Nm]
Output signal
Starting vector
Mode shape vector [m]
Displacement vector [m]
Damping constant
Frequency [Hz]
Number
Imaginary unit, number
Stiffness [N/m]
Mass [kg]
Number
Transform variable, feed rate [mm/rev]
Time [s]
Eigenvector
Velocity [m/s]
Displacement [m], co-ordinate direction
Co-ordinate direction
Co-ordinate direction
Coherence function
Eigenvalue
Eigenvalues of shifted problem
Eigenvalue shift
Natural frequency [Hz]
Power spectral density of the signal x [dB]

Sxy
Syx
Syy

Cross spectral density of the signal x and the signal y [dB]

Cross spectral density of the signal y and the signal x [dB]
Power spectral density of the signal y [dB]

Indices
T
t
n
i
j
x
y
dr
lin

Transpose
Tightening
Number
Number
Imaginary unit, number
Signal, input
Signal, output
Damped natural frequency
Linear

Abbreviations
EMA
FEM
FRF
MDOF
SVI

Experimental Modal Analysis

Finite Element Method
Frequency Response Function
Multiple Degree of Freedom
Simultaneous Vector Iteration

2 Introduction
2.1 Background
The turning process is one of the most important manufacturing processes
in the industry today. The evolution of the turning process has been ongoing
in the last centuries. It can be seen as a part of the industrial evolution.
Every day, we use turned products, from the hardware to deliver water to
kitchen faucets, cars and planes for transportation, furniture and so on, see
figure 2.1. Over the past thirty years, new technologies have made a
dramatic impact on the turning process. The new technologies have made it
possible to produce goods of better quality, faster and at lower costs.
Examples are the numeric controlled machines that make it possible to
manufacture a complex work piece in one single operation.
In modern turning industry one of most important factors to obtain high
productivity and high quality is production safety. Factors that affect the
production safety are among others durability of the lathe, tool life and also
the machinability of the workpiece material. The machinability, or
cuttability, can be quantified from the cutting power or cutting forces that
the work piece give rise to, how easy chip breakage is obtained, the work
piece material influence of the tool life and finally the surface finish of the
machined product. These different aspects usually interact, e.g. a work
piece where it is easy to obtain fair surface finish, usually gives rise to high
cutting forces. There are many process behaviours that affect the surface
finish, e.g. the ability to get build-up edges or layers and the cutting forces
themselves. Both the mean values of the cutting forces and the dynamic
properties influence the surface finish. The dynamic characteristics give rise
to vibrations in the lathe and the work piece.

Figure 2.1. Examples of different turned products.

Generally in turning operations there will be some level of relative dynamic

motion between cutting tool and work piece. Energy from the chip
formation process excites the mechanical modes of the machine-tool
system. Modes of the work piece may also influence tool vibration. The
dynamic properties of the excitation, i.e. the chip formation process are
correlated to the material properties and the geometry of the work piece.
The vibrations may lead to unwanted noise, degraded surface finish and
reduced tool life.

2.2 Project Description

A common way to avoid vibration problem is to operate the turning process
under-critical and thereby reduce the excitation of the modes of the
machine-tool system and/or the work piece. However, under-critical turning
process will often lead to high manufacturing and operation time. One way
to reduce the vibration problem associated with critical machining is to use
actuators, which automatically modifies the structural response of the tool
holder shank, which leads to further reduction of the dynamic motion
between cutting tool and work piece. This leads to turning products with
higher cutting data that results in a reduction of the production time and
hence competitive products.
The aim of this work is to determine the natural frequencies, mode shapes
and damping ratios for a Tool Holder Shank. This is a first step in
investigating if it would be possible to use piezoelectric actuators to reduce
vibrations during the cutting process.

2.3 Method
The first step is to develop a mathematical model using the Finite Element
Method (FE-model) of the problem to get an understanding of the dynamics
involved. To verify the FE-model an Experimental Modal Analysis (EMA)
is done. Based on the experimental results the FE-model is adjusted and
when the FE-model and the experimental model correlates accurately the
FE-model of the tool holder shank is accepted.

The agreement between FE-models and the experimental models gives

opportunity to use the FE-models for more extensive studies without further
extensive experimental measurements.

3 Theories
3.1 Turning
3.1.1 The Turning Operation
Turning [2] in its simplest form is a machining process used to generate
external, cylindrical surfaces by removing material by a cutting tool. The
primary motion is rotation. The feed motion is normally a rectilinear
movement, see figure 3.1. When machining, the work piece is
perpendicular to the cutting tool.

Primary motion

Longitudinal feed
Transverse feed
Figure 3.1. The external longitudinal turning process [2].
3.1.2 Undesired Vibrations and Noise
Undesired vibrations are a serious problem that affects and deteriorates the
surface finish of the work piece. It also affects the dimensional accuracy of
the work piece, and reduces tool/machine lifetime. Also vibrations can be
initiated in machine tools by component defects, unbalanced parts, poor
assembly etc.
One way to represent the vibration energy in a cutting process [6] is to plot
the cutting speed as a function of frequency, see figure 3.2.

10

Figure 3.2. Typical spectral density estimate of the dynamic response for a
tool holder shank in the primary direction during a continuous
cutting operation in SS0727-02 with constant feed rate s = 0.5
mm/rev and cutting speed v = 50 400 m/min [6].
To show that the natural frequencies are constant at each feed rate and
cutting speed the spectral density is plotted for cutting speed from 50 - 400
m/min. Figure 3.2 also shows the dynamic response in the primary direction
during a continuous cutting operation.
Other problems in machining are plastic deformation and friction in the
contact between the cutting tool and the work piece. This contact generates
heat, which increases the temperature of both components, see figure 3.3.
The elevated temperature of the tool reduces its wear resistance and
changes both the geometry and the size of the turned component. If the
vibrations are controlled, higher cutting data can be used and the time
varying loads on the cutting tool are decreased. The result is longer
tool/machine lifetime.

11

Figure 3.3. Temperature distribution in the cutting area [2].

If the working environment is considered, noise is frequently introduced by
dynamic motion between the cutting tool and the work piece. By improving
the dynamic stiffness of the machine structure, the problem of relative
dynamic motion between cutting tool and work piece may be partially
solved.

3.2

Actuators

Reducing the noise and the undesired vibrations is clearly a very important
goal. This can be achieved by using actuators, which automatically modifies
the structural response of a mechanical system. The actuator is used to
enhance the performance of a structural system by inducing a favourable
structural deformation according to the applied voltage. It develops
cancelling force to reduce the vibration level and acoustic noise level. There
are several types of actuators that could be used to solve vibration problems
in mechanical systems, for example magnetostrictive [1] and piezoelectric
actuators [11]. Typical application areas for actuators are flexible robotics,
aircraft, marine hulls, vibrating machinery and modal testing. In control
applications piezoelectric actuators are usually preferred due to their
superior linearity and simplicity to mount on the structures to be controlled.
In our application the selection of actuator type is limited to the

12

piezoelectric actuators by two facts; the actuator mount and the linearity
request.
3.2.1

Piezoelectric Actuators

The patch actuator [10] is a piezoelectric actuator containing a high

efficiency piezoceramic element within a rugged, electrically insulated
housing, see figure 3.4. It has an integral cable and is attached to the
structure with a two-part epoxy or similar adhesive.

Figure 3.4. Patch actuators [10].

Actuators convert electrical signals like voltages or charges into mechanical
displacements or forces. As for sensors, a reasonably linear relationship
between input signal and movement is required the actuators can be divided
in three main groups.

Axial actuators

Transversal actuators

Flexural actuators

Axial and transversal actuators have high stiffness and are optimised for
small movements and high forces. Flexural actuators cover the applications
where larger movements are required. However, they have a restricted
ability withstand large stress, and are usually used on light structures.
Advantages of piezoelectric actuators is their low profile, low mass and
high output, it makes the actuator ideal for controlling adaptive structures.

13

3.3

Dynamics

3.3.1 Multiple Degree of Freedom (MDOF) Systems

Figure 3.5 illustrates an undamped system with n translation degrees of
freedom.

k1

m1

k2

k n 1

m2

x1

kn

mn 1

x2

mn
xn

x n 1

Figure 3.5. Undamped MDOF-system.

The forces from the springs acting on each mass are determined through a
free body diagram, see figure 3.6.

k1 x1

m1

k 2 ( x2 x1 )
x1

m2

k n1 ( xn1 xn2 )

x2

mn 1

k n ( xn xn1 )

mn

x n 1

xn

Figure 3.6. Free body diagram for the undamped MDOF-system.

Using Newtons second law equations of motion for each mass can be
written as
m1 x1 x1 (k1 k 2 ) x 2 k 2 0
m 2 x2 x 2 (k 2 k 3 ) k 2 x1 k 3 x 3 0

(3.1)

m n 1 xn 1 x n 1 (k n 1 k n ) k n x n k n 1 x n 1 0
m n xn k n x n 1 k n x n 0

14

Equation 3.1 consists of n coupled second order differential equation and

can be written in matrix form as
Mx Kx 0

(3.2)

where M is the mass matrix and K is the stiffness matrix, and x is the
displacement vector. Dots indicate time derivatives.
3.3.2

Natural Frequencies and Mode Shapes

Assuming harmonic motion for the spring-mass system [7] the

displacements are expressed as
x(t ) ue jt

(3.3)

where u is a vector of amplitudes and is the natural angular frequency.

Substituting this into the equations of motion yields
( 2 M K )ue jt 0

(3.4)

Noting that the scalar ejt 0 for any value of t equation 3.4 yields that
(M 2 K )u 0

(3.5)

Equation (3.5) is a generalised eigenvalue problem and has non trivial

solutions if
det( 2 M K ) 0

(3.6)

This is called the characteristic equation of the system and has solutions
12 , 22 ,..., n2 , known as the eigenvalues. The eigenvalues give the natural
frequencies. For the eigenvalues the corresponding eigenvectors
u1 , u 2 , ..., u n can be obtained. The eigenvectors describe the mode shapes of
the dynamic system.

3.4 Modal Analysis

Modal analysis theory [8] refers to classical vibration theory that explains,
theoretically, the existence of natural frequencies, damping ratios and mode
shapes for linear systems. This theory includes lumped-parameter or
discrete models as well as continuos models. The theory also includes real

15

normal modes as well as complex modes of vibration as possible solutions

for the modal parameters.
Modal analysis is used to help design the structural system for noise and
vibration applications.
The modes of a structure or system can be estimated from two different
approaches:

Mathematical Models

Experimental Modal Analysis

3.4.1 Mathematical Models

In its most basic form, a mathematical model simplifies a structure by
breaking it up into masses and springs. This process can be done by the
simple lumped mass and lumped spring approach. This modelling process
reduces the complicated structure into many mass/spring systems. The
eigenvalue problem can then be solved to get the frequency and mode shape
of each mode for the assumed mass and stiffness distribution. When the
modes are established, calculations can be done to determine how the
structure will respond to various dynamic inputs. One of the forced
response runs that is often used is to input a unit force with variable
frequency at one point while monitoring the response as a function of
frequency (FRF) at several important locations.
3.4.2 Experimental Modal Analysis
Experimental Modal Analysis (EMA) is the interpretation of test data
collected from a vibrating structure [8]. It involves determining modal
parameters as frequencies, damping ratios, and mode shapes of a linear
time-invariant system. A common reason for EMA is verification of results
from FE-models.

16

3.5 Eigenvalues and Eigenvectors in I-DEAS

For FE-calculations of the natural frequencies and modeshapes I-DEAS
System Dynamics Analysis software is used. There are several ways of
determining the eigenvalues and eigenvectors when calculating the natural
frequencies and modeshapes. In our case the Lanczos method [5] is used
due to its many advantages in comparison to the other methods, Guyan, and
SVI.
3.5.1

Lanczos Formulation

The purpose of the Lanczos method is to compute a relatively few

eigenvalue and eigenvector pairs for a model defined by a large number of
degrees of freedom. By using restart solutions coupled with frequency
shifting, a large number of modes can be determined efficiently by
obtaining a few at a time.
The technique can be used to solve certain large, sparse, symmetric
eigenproblems. The method involves partial tridiagonalisations of the given
matrix. Information about extremal eigenvalues tends to emerge long before
the tridiagonalisation is complete. This makes the Lanczos algorithm
particularly useful in situations where a few of the largest or smallest
eigenvalues are desired.
3.5.2

Lanczos Algorithm

The Lanczos algorithm [5] has the ability to compute the smaller
eigenvalues of a matrix without any factorisation. However, they will not be
approximated accurately until nearly all eigenvalues have been determined.
Consequently, it is necessary to apply Lanczos factorisation to an inverted
form of the matrix.
The Lanczos algorithm uses a shift-and-invert procedure to converge
quickly to the eigenvalues closest to the shift. The eigenvectors of the
original and shifted problem are the same. The general eigenproblem is
Kx Mx

(3.7)

17

Frequency shifting selectively extract modes by shifting the analysis, so that

mode extraction begins at higher modes or higher frequencies. Frequency
shifting shifts the eigenvalue problem so that the lowest eigenvalues are
closest to the frequency shift point. Applying a shift, the algorithm works
with the equation
Kx Mx

(3.8)

K K M

(3.9)

where

Since the algorithm works with the inverse of K the spectrum of the
original eigenproblem is related to the spectrum of the shifted problem by

(3.10)

The Lanczos method constructs an orthogonal set of vectors, known as

Lanczos vectors, for use in the Rayleigh-Ritz [3] procedure. Given a
starting vector r, these basic methods generates a sequence of vectors

1
1
1
r, K Mr , K M 2r,........, K M jr

(3.11)

during j iterations. These vectors are referred to as the Krylov sequence [5].
The sequence converges (as j infinity) to the eigenvector corresponding
to the eigenvalue closest to the shift.
The basic difference between the Lanczos method and the other two
methods (Guyan, and SVI) is that the information contained in each
successive vector of the Krylov sequence is used to obtain the best
approximation to the wanted eigenvectors instead of using only the last
vector in the sequence. In other words, the Lanczos algorithm is equivalent
to obtaining the Rayleigh-Ritz approximation with the vectors in the Krylov
sequence as the trial vectors.
This method involves supplementing the Krylov sequence with an
orthogonalisation process with respect to the other vectors. The result is a
set of M-orthonormal vectors (the Lanczos vectors) that is used in the
Rayleigh-Ritz procedure to reduce the dimension of the eigenproblem. This
leads to a standard eigenproblem with a tri-diagonal matrix.

18

3.6 Signal Processing and Vibration Testing

3.6.1

Transfer Functions used in Vibration Measurement

Usually vibration is measured in terms of motion and therefore the

corresponding frequency response function may be presented in terms of
displacement, velocity or acceleration, see table 3.1.
Table 3.1. Transfer functions.
Response
Measurement

Transfer
Function

Inverse Transfer
Function

Acceleration

Accelerance

Apparent mass

Velocity

Mobility

Impedance

Displacement

Receptance

Dynamic stiffness

The three transfer functions given in Table 3.1 are related to each other by
simple multiplications of the transform variable s, since this corresponds to
differentiation. Thus with the receptance transfer function denoted by
X ( s)
1
H ( s)
2
F ( s)
ms cs k

(3.12)

the mobility transfer function becomes

sX ( s)
s
sH ( s)
2
F ( s)
ms cs k

(3.13)

because sX(s) is the transform of the velocity. Similarly, s 2 X ( s) is the

transform of the acceleration and the accelerance transfer function becomes

s 2 X ( s)
s2
s 2 H ( s)
F ( s)
ms 2 cs k

19

(3.14)

Each of these also defines the corresponding frequency response function

by substituting s j dr .
For example, the response measurements done in Chapter 5 are made by
using accelerometers and by using laser vibrometer. The accelerometers
measure the acceleration in a certain point giving the corresponding
accelerance transfer function. The laser vibrometer on the other hand
measures the velocity in a certain point giving the corresponding mobility
transfer function. To be able to compare the results the transform variable s
has to be used according to equation 3.13 and 3.14. The accelerance transfer
function, Equation 3.14, could by divided by s or the mobility transfer
function, Equation 3.13, could be multiplied by s.
3.6.2 Coherence
The coherence function, denoted by 2, is defined to be the ratio of the two
values of the frequency response function H(j), calculated from the
equations below:
S yx ( ) H ( j ) S yy ( )

(3.15)

S xx ( ) H ( j ) S xy ( )

(3.16)

where Syx( ) and Sxy( ) is the cross-spectral density [9] of the signal x(t)
and the signal y(t), Sxx( ) and Syy( ) is the power spectral density [9] of
each signal. This yields

S xy ( )

(3.17)

S xx ( ) S yy ( )

The coherence function gives a measure of how well the output signal
corresponds linearly to the input signal at a specific frequency. The value of
the coherence function always lies between zero and unity, see Figure 3.7.
If the output signal can be explained linearly from the input signal the
coherence is unity. If the output signal is statistically independent or
nonlinearly related to the input signal the coherence is zero.
The coherence is taken as an indication of how accurate the measurement
process is over a given range of frequencies.

20

Figure 3.7. Example of a coherence function.

Values close to unity should occur near the structures resonance
frequencies. At resonance the signals are large and less affected by noise.
Usually data with a coherence of less than 0.9 indicates that the test should
be redone.

3.7 Correlation
There are several ways of determining how closely the experimental and
analytical models correspond. Several comparison techniques are presented
in [8]. In general comparison techniques are to a greater or lesser extent
based on the orthogonality conditions:
U T MU I

and

U T KU n2

(3.18)

where U represents the mass normalised eigenvectors, either from test or

analysis, and I and [ n2 ] are the diagonal identity and eigenvalue matrices,
respectively.
3.7.1

Comparison of Natural Frequencies

One way to compare the natural frequencies for a predicted and

experimental model is to plot the experimental values against the predicted
values [4]. In this way it is possible to see the degree of correlation between

21

the two sets of results. The points plotted should lie on or close to a straight
line, see figure 3.8.

700

Measured Frequency [Hz]

600
500
400
300
200
100
Predicted Frequency [Hz]
0

100

200

300

400

500

600

700

Figure 3.8. Example of comparison of measured and predicted natural

frequencies.
3.7.2 Comparison of Mode Shapes
One way to compare the mode shapes for a predicted and experimental
model is to plot the predicted and experimental nodal amplitudes as
indicated in figure 3.9. The individual points on this plot relate to specific
co-ordinates on the model and should lie close to a straight line passing
through the origin.

22

1.0
Experimental

Predicted
1.0

-1.0

Mode1
Mode2
Mode3
-1.0

Figure 3.9. Example of comparison of measured and predicted mode

shapes.

23

4 Theoretical Models
4.1 Elements
One of the most important choices when making a finite element model is
which element type to use. There is three different types presented here.
The beam element, shell element and solid element. All finite element
modelling is performed in I-DEAS.
4.1.1 Beam Elements
Beam elements can be used to make a very efficient finite element model to
predict overall deflection and bending moments but will not be able to
predict the local stress concentrations at the point of application of a load or
at joints. One of the advantages of using beam elements is that the computer
time is relatively short since it has fewer degrees of freedom. A beam has
three transitional degrees of freedom and three rotational degrees of
freedom. There are four types of beam elements in I-DEAS: linear,
parabolic, curved, and tapered.
4.1.2 Shell Elements
Shell elements can be effectively used for structures with relatively thin
walls such as sheet metal parts, discs, plates, etc.
4.1.3 Solid Elements
The most general elements are the solid elements. The disadvantage is that
more elements and nodes are usually required, and the model will therefore
be more time demanding to solve, compared to for example a beam element
model. However, many devices designed by engineers have highly threedimensional geometry, which makes it necessary to use three-dimensional
elements.

24

4.2. Boundary Conditions

It is very important to understand how the boundary condition is defined.
The boundary condition includes loads, temperature restraints, and
displacement restraints. It is also very important that especially the
experimental set-up is well defined and experimental repeatable. Exact
definition of the boundary conditions may be problematic, but nevertheless,
tests should be considered to prove the repeatability of the installation.
4.2.1

Free-Free

For a structure to be really free, it should be suspended in the air, free in

space with no holding points whatsoever. Such a situation is commonly
designated as free-free. The simulation of free-free conditions is easy to
achieve. It suffices to suspend or support the structure using very flexible
springs so that the rigid body resonance frequencies are far away from the
frequency range of interest. In this work rubber bands were used, see figure
5.2.

4.3 The Tool Holder Shank

The tool holder shank that is used for the experimental procedures has a
cross section of 32x25 mm. The geometry of the tool holder shank is
presented in figure 4.1.

25

Figure 4.1. The geometry of the tool holder shank.

4.4 The Tool Holder Shank Free-Free

The first FE-model is built up with beam elements. It is important to get an
apprehension of which frequencies the tool holder shank involves. The tool
holder shank is built up with 7 elements and 8 nodes, see figure 4.2. Five
elements with 32x25 mm, one with 36x26 mm and one with 30x12 mm
cross section are used. The FE-model has 1 lumped mass of 0.020 kg. The
mass of the FE-model is 1.26 kg and for the real tool holder shank 1.1 kg.
The difference in mass between the FE-model and the experimental model
depends among other things on the difficulty to achieve equal mass moment
of inertia between the two models. Free-free boundary conditions for the
tool holder shank are used and the eigenvalue problem is solved by the
Lanczos method. The natural frequencies are presented in chapter 6 and the
modeshapes are presented in Appendix A. A solid model for the tool holder
shank is also created and presented in Appendix C.

26

Figure 4.2. The FE-model for the tool holder shank with beam elements
and lumped mass.

4.5 The Tool Holder and Tool Holder Shank FreeFree

4.5.1

With 40 mm Hang Out

The FE-model is built up with beam elements. The tool holder shank is
built up with 5 elements and 6 nodes and 2 lumped masses, see figure 4.3.
Three elements with 32x25 mm, one with 36x26 mm and one with 30x12
mm cross section are used. The FE-model has two lumped masses of 0.04
kg and 0.08 kg. To simulate the boundary conditions the FE-model is
clamped at 40 mm hang out. The eigenvalue problem is solved by the
Lanczos method. Results are presented in chapter 6.

Figure 4.3. The FE-model for the tool holder and tool holder shank freefree with beam elements and lumped masses.

27

4.5.2 With 60 mm Hang Out

The FE-model is built up with beam elements. The tool holder shank is
built up with 6 elements and 7 nodes and 2 lumped masses, see figure 4.4.
Four elements of 32x25 mm, one with 36x26 and one with 30x12 mm cross
section are used. The lumped masses are 0.02 kg and 0.01 kg. To simulate
the boundary conditions the FE-model is clamped at 60 mm hang out. The
eigenvalue problem is solved by the Lanczos method. Results are presented
in chapter 6.

Figure 4.4. The FE-model for the tool holder and tool holder shank freefree with beam elements and lumped masses.

4.6 The Tool Holder and Tool Holder Shank in

Concrete Lump
4.6.1 With 40 mm Hang Out
The FE-model is built up with beam elements. The tool holder shank is
built up with 5 elements and 6 nodes and 2 lumped masses, see figure 4.3.
Three elements of 32x25 mm, one with 36x26 and one with 30x12 mm
cross section are used. The lumped masses in this case are 0.28 kg and 0.25
kg. To simulate the boundary conditions the FE-model is clamped at 40 mm
hang out. The eigenvalue problem is solved by the Lanczos method. Results
are presented in chapter 6.

28

4.6.2

With 60 mm Hang Out

The FE-model is built up with beam elements. The tool holder shank is
built up with 6 elements and 7 nodes and 2 lumped mass, see figure 4.4.
Four elements of 32x25 mm, one with 36x26 and one with 30x12 mm cross
section are used. The lumped masses in this case are 0.35 kg and 0.3 kg. To
simulate the boundary conditions the FE-model is clamped at 60 mm hang
out. The eigenvalue problem is solved by the Lanczos method. Results are
presented in chapter 6.

4.7 The Tool Holder and Tool Holder Shank in

Lathe
4.7.1 With 40 mm Hang Out
The FE-model is built up with beam elements. The tool holder shank is
built up with 5 elements and 6 nodes and 2 lumped masses, see figure 4.3.
Three elements of 32x25 mm, one with 36x26 and one with 30x12 mm
cross section are used. The lumped masses in this case are 0.3 kg and 0.2
kg. To simulate the boundary conditions the FE-model is clamped at 40 mm
hang out. The eigenvalue problem is solved by the Lanczos method. Results
are presented in chapter 6.
4.7.2

With 60 mm Hang Out

The FE-model is built up with beam elements. The tool holder shank is
built up with 6 elements and 7 nodes and 2 lumped masses, see figure 4.4.
Four elements of 32x25 mm, one with 36x26 and one with 30x12 mm cross
section are used. The lumped masses in this case are 0.2 kg and 0.1 kg. To
simulate the boundary conditions the FE-model is clamped at 60 mm hang
out. The eigenvalue problem is solved by the Lanczos method. Results are
presented in chapter 6.

29

5 Experimental Procedures
5.2 Excitation Methods and Requirements
The choice of experimental excitation method depends on several factors,
such as the geometry and shape of the structure. It took several tests to
establish a method to use for our application. The structure could be excited
either by an impulse hammer or by using a shaker connected with an
impedance head by a flexible drive rod, see figure 5.1. The impulse hammer
gives rise to an impulse signal when exciting the structure with a distinct
shock. The shaker gives rise to a vibrating excitation signal that is
transferred into the structure by an impedance head. The impedance head is
a force transducer and an accelerometer in one; it measures the acceleration
and the input force in the same point, the so-called driving point.

b)

a)

Figure 5.1. Two excitation methods, a) impulse hammer excitation and b)

shaker excitation.
The requirements for the excitation method were that the frequency range
amount to 12.8 kHz and that sufficient energy are still in the system at least
up to 10 kHz. These requirements were thought to be fulfilled by both
excitation methods at the beginning, but after several tests it came clear that
the shaker excitation only has sufficient energy up to approximately 4 kHz.
The problems when measuring with the shaker is that it is very important to
mount the impedance head in the centre of the tool holder shank to avoid

30

undesired movements and moments into the structure. Another problem is

the mass spring system of the impedance head that could affect the natural
frequencies of the system. These facts excluded the shaker as excitation
method.

5.3 Equipment in All Experimental Measurements

The following equipment is used, Hewlett Packard measuring system HP3565S, impulse hammer PCB with force sensor PCB model 086C03, nickel
tip and I-DEAS test software v.5.

5.4 The Tool Holder Shank Free-Free

The following equipment is used, tool holder shank, two accelerometers
Dytran model 3220A, Loctite fast glue as adhesive for mounting the
accelerometers, rubber bands and hoop.
The tool holder shank is mounted with rubber bands on the hoop to isolate
it from disturbance from the surroundings, see figure 5.2.

Figure 5.2. Experimental set-up for the tool holder shank free-free.
The accelerometers are mounted in two directions (wide and narrow) at the
planar surfaces at the front end of the tool holder shank, see figure 5.3.
Excitation is done with the impulse hammer in the points of interest, and
perpendicular to the surface of the tool holder shank.

31

Accelerometer 1

*
*

*
*

*
*

*
*

*
*

Accelerometer 2

Figure 5.3. Measuring points and accelerometers placements.

Nodes and trace lines are created in I-DEAS test Modal preparation for the
measuring points, see figure 5.4a and b. The two nodes at the front end side,
where the surface is leaning are connected to a local co-ordinate system.
The other nodes are connected to the global co-ordinate system.

Figure 5.4a. Tracelines for the tool holder shank, seen from above.

Figure 5.4b. Tracelines for the tool holder shank, seen from beside.

5.5 The Tool Holder and Tool Holder Shank FreeFree

The following equipment is used: tool holder, tool holder shank, two
accelerometers Dytran model 3220A, Loctite fast glue as adhesive for
mounting the accelerometers, rubber bands and hoop.

32

The tool holder shank is attached to the tool holder with two screws with a
tightening moment of 100 Nm and treated as one structure. The structure is
mounted with rubber bands on the hoop to isolate it from disturbance from
the surroundings, see figure 5.5.

Figure 5.5. Experimental set-up for the tool holder and tool holder shank
free-free.
The measurement is done twice for different mounting lengths of the tool
holder shank. First for 40 mm, second for 60 mm.
The accelerometers are mounted as in figure 5.3. Excitation is done with the
impulse hammer in the points of interest, and perpendicular to the surface
of the tool holder shank. The two nodes at the front end side, where the
surface is leaning are connected to a local co-ordinate system. The other
nodes are connected to the global co-ordinate system, see figure 5.6 and
figure 5.7.

Figure 5.6a. Tracelines for the tool holder shank. 40 mm hang out , seen
from above.

33

Figure 5.6b. Tracelines for the tool holder shank, 40 mm hang out, seen
from beside.

Figure 5.7a. Tracelines for the tool holder shank, 60 mm hang out, seen
from above.

Figure 5.7b. Tracelines for the tool holder shank, 60 mm hang out, seen
from beside.

5.6 The Tool Holder and Tool Holder Shank in

Concrete Lump
5.6.1 Measurements with Accelerometers
The following equipment is used, tool holder, tool holder shank, tool holder
attachment, two accelerometers Dytran model 3220A, Loctite fast glue as
adhesive for mounting the accelerometers and maintaining their physical
properties and fastening elements.
The tool holder shank is attached to the tool holder with two screws. The
complete structure is then fastened into the concrete lump with fastening
elements trying to simulate the lathe to minimise the disturbances acting on

34

the tool holder shank when in the lathe. The accelerometers are mounted as
in figure 5.3. Excitation is done with the impulse hammer in the points of
interest, and perpendicular to the surface of the tool holder shank. Nodes,
trace lines and co-ordinate systems are created as in section 5.5. The
measurement is done twice for different mounting lengths of the tool holder
shank. First for 40 mm, second for 60 mm. The tightening moments Mt that
is used for tightening the fastening elements is chosen to 100, 120 and 140
Nm, see Figure 5.8.

120 Nm

140 Nm

100 Nm
Figure 5.8. Experimental set-up for the tool holder and tool holder shank in
concrete lump.
5.6.2

Measurements with Laser Vibrometer

The following equipment is used: tool holder, tool holder shank, tool holder
attachment, oscilloscope, laser vibrometer VS-100 with power supply and
fastening elements.
The tool holder shank is attached to the tool holder with two screws. The
complete structure is then fastened into the concrete lump with fastening
elements trying to simulate the lathe to minimise the disturbances acting on
the tool holder shank when in the lathe. The laser beam is directed into the
tool holder shank measuring the velocity when an impulse is made with the
impulse hammer. The distance from the laser vibrometer to the structure is
830 mm, which is one of the optimum distances recommended from the
manual of the laser vibrometer, see figure 5.9. The measuring points are the

35

same as in section 5.5. Excitation is done with the impulse hammer in the
points of interest, and perpendicular to the surface of the tool holder shank.
Nodes, trace lines and co-ordinate systems are created as in section 5.5. The
measurement is done twice for different mounting lengths of the tool holder
shank. First for 40 mm, second for 60 mm.

Oscilloscope

Tool holder
shank

Power
supply

Laser
Vibrometer

Figure 5.9. Experimental set-up for the tool holder and tool holder shank in
concrete lump.

5.7 The Tool Holder and Tool Holder Shank in

Lathe
The following equipment is used, tool holder, tool holder shank, lathe, two
accelerometers Dytran model 3220A, Loctite fast glue as adhesive for
mounting the accelerometers and maintaining their physical properties,
oscilloscope, laser vibrometer VS-100 with power supply.
The tool holder shank is placed in the tool holder and fixed into the lathe.
Measurements are made first with two accelerometers and second with the
laser vibrometer for the mounting lengths of 40 and 60 mm for the tool
holder shank, see figure 5.10 and figure 5.11. The distance used for the
Laser vibrometer in this case is 220 mm, recommended from the manual.

36

Figure 5.10. Experimental set-up for the tool holder and tool holder shank
in lathe with accelerometers.

Figure 5.11. Experimental set-up for the tool holder and tool holder shank
in lathe with laser vibrometer.

37

6 Results
The modes that are interesting are the bending modes in the wide and
narrow direction, see figure 4.1. The theoretical and experimental results of
the frequencies for the tool holder and tool holder shank are presented in
tables 6.1 6.8 below. The mode shapes for the tool holder shank free-free
are presented in Appendix A, the comparison of natural frequencies and
modeshapes are presented in Appendix B.
Table 6.1. Results for the tool holder shank free-free with accelerometers.
Mode
Number

FE-Results beam
elements

Experimental
Results

Frequency (Hz)

Frequency (Hz)

Difference
between FE.
and Exp.

Damping ratio
(%)

(%)

4503

4503

0.29

5567

5479

1.6

0.39

10905

10646

2.4

0.39

Table 6.2. Results for the tool holder and tool holder shank free-free, 40
mm hang out with accelerometers.
Mode
Number

FE-Results beam
elements

Experimental
Results

Frequency (Hz)

Frequency (Hz)

Difference
between FE.
and Exp.

Damping ratio
(%)

(%)

2956

2798

5.6

0.27

3568

3685

3.3

0.80

10484

11112

6.0

0.47

38

Table 6.3. Results for the tool holder and tool holder shank free-free, 60
mm hang out.
Mode
Number

FE-Results beam
elements

Experimental
Results

Frequency (Hz)

Frequency (Hz)

Difference
between FE.
and Exp.

Damping ratio
(%)

(%)

2385

2362

1.0

0.33

2927

3080

5.2

0.28

10015

9165

9.3

0.37

Table 6.4. Results for the tool holder and tool holder shank in concrete
lump, 40 mm hang out with accelerometers.
Mode
Number

FE-Results beam
elements

Experimental
Results

Frequency (Hz)

Frequency (Hz)

Difference
between FE.
and Exp.

Damping ratio
(%)

(%)

1789

1856

3.7

1.69

2145

2072

3.5

0.81

Table 6.5. Results for the tool holder and tool holder shank in concrete
lump, 60 mm hang out with accelerometers.
Mode
Number

FE-Results beam
elements

Experimental
Results

Frequency (Hz)

Frequency (Hz)

Difference
between FE.
and Exp.

Damping ratio
(%)

(%)

1109

1096

1.2

0.95

1365

1340

1.9

0.86

39

Table 6.6. Results for the tool holder and tool holder shank in lathe, 40 mm
hang out with accelerometers.
Mode
Number

FE-Results beam
elements

Experimental
Results

Frequency (Hz)

Frequency (Hz)

Difference
between FE.
and Exp.

Damping ratio
(%)

(%)

1868

1776

5.2

0.82

2239

2336

4.3

0.94

Table 6.7. Results for the tool holder and tool holder shank in lathe, 60 mm
hang out with accelerometers.
Mode
Number

FE-Results beam
elements

Experimental
Results

Frequency (Hz)

Frequency (Hz)

Difference
between FE.
and Exp.

Damping ratio
(%)

(%)

1497

1264

18.4

0.62

1903

2032

6.8

0.58

Table 6.8. Results for the tool holder and tool holder shank in concrete
lump and lathe, 40 and 60 mm hang out with laser vibrometer.
Experimental Results
Frequency (Hz)
Mode
Number

40 mm hang out in
concrete lump

60 mm hang out in
concrete lump

40 mm hang out in
lathe

60 mm hang out in
lathe

1712

1216

1712

1232

1760

1592

2080

2112

40

7 Conclusions
The aim of this work was to determine the natural frequencies, mode shapes
and damping ratios for a tool holder shank. This is considered to be a
suitable first step in investigating if it would be possible to use piezoelectric
actuators to reduce vibrations during the cutting process. Finite element
models to describe the tool holder shank have been suggested. Agreement,
disagreement, and possible reasons for disagreement between results
predicted by these models and by experimental measurements are
discussed below.
There is no single right way to perform a vibration test. In almost every case
the support, the excitation equipment or the transducers will influence the
dynamic behaviour of a structure under an experimental test. It is very
important to realise these influences, understand them and design the test to
minimise their effects on the dynamic behaviour of the structure. All modal
tests involve a degree of compromise. Almost all methods for applying the
structural excitation will have some unwanted modification on the
structure. Similarly almost all the response measurement transducers and
support fixtures will have unwanted influence on the structure.

7.1 Complexity of Problems

It is very important to get the right set-up for the experimental procedure to
have good agreement between the experiment and the FE-model. It is also
very important to get the right boundary conditions. The problem is to
define how the tool holder and tool holder shank are fixed together.
Consideration of the support of the structure under test is an important part
of the test set-up. The support conditions should be well defined and
experimentally repeatable. It is almost impossible to achieve a complete
grounded structure in practice. A grounded structure will have some
movement at the grounding point, usually rotation.

41

The agreement between the experimental and the FE-calculated frequencies

for the tool holder shank free-free is good. Also for the tool holder and tool
holder shank free-free there is good agreement.
The modeshapes of the experimental model and the FE-model for the tool
holder shank free-free agree well, except for the nodes beside the measuring
points were the accelerometers are mounted, because no excitation is made
in these points.
The agreement between the experimental and the FE-calculated frequencies
for the tool holder and tool holder shank in concrete lump and lathe is not
so good. The problem is to decide which frequencies really are frequencies
belonging to the tool holder shank. A reason among others is that the
stiffness between the tool holder and the tool holder shank is much lower
than the stiffness of the tool holder and the tool holder shank itself.
Another problem is that the distance L1 and L2 are different, see figure 7.1.
When the tool holder shank moves in the negative y-direction the length is
L1. When it moves in the positive y-direction the length is L2. That means
that we have two linear systems. One for length L1, and one for length L2.
Together they build a non-linear system depending on its position. When it
moves in the positive y-direction the length varies from L1 to L2, which
means that the movement between L1 and L2 is non-linear depending on its
position, see figure 7.1 and 7.2. This will result in two different frequencies
during its movements from its negative to its positive direction. This
problem will also come up in the z-direction (perpendicular to the paper).
This is one reason for the less good agreement between the experimental
and the theoretical frequencies and mode shapes.

Figure 7.1. The tool holder and tool holder shank with length L1 and L2.

42

Figure 7.2. The tool holder and tool holder shank bending in the negative
and positive y-direction.
The tightening moment Mt also affects the frequencies. The moment that is
used in the experiment is 100, 120 and 140 Nm for the five bolts, see figure
5.6. A comparison is done between two different tightening moments Mt.
ADF number 32 is tightened with 100, 120 and 140 Nm and the ADF
number 48 is tightened with 60, 80 and 140 Nm, see figure 7.3. The
tightening tool has also a source of error of about 5-10 %.

Figure 7.3. FRF for different tightening moments Mt.

The lathe has a lot of parts that also have natural frequencies. The
attachment for the tool holder, as the screws on the attachment for the tool

43

holder, also has natural frequencies. A great problem is to decide which

frequencies that really are frequencies belonging to the tool holder shank.

Figure 7.4. The lathe with the tool holder and tool holder shank.
Accelerometers build in new mass/spring systems that could affect the
natural frequencies of the tool holder shank.
The problems mentioned makes it difficult to determine the natural
frequencies for the tool holder and tool holder shank in concrete lump and
in lathe. In Appendix D FRFs from measurements with the laser vibrometer
for the tool holder shank in concrete lump is presented. One reason for the
poor measurements can be the problem for the laser to register the high
velocities for frequencies from 5000 Hz and up.
Getting correct modal data from relatively complex structures involves
many difficulties. A minority of measurement points may probably not
indicate some of the mode shapes. But increasing the number of locations
not only increase the time of experiment but also induce more errors, such
as noise, improper positioning of the impulse hammer and exciter, signal
truncation error due to selection of window, and so forth.

44

7.2 The Continuation of the Analysis of the Tool

Holder Shank
With the basic knowledge of the dynamic characteristics of the tool holder
shank obtained in this work the next step would be to include actuators in
the model. The actuators can perhaps be modelled as forces/pressures on
the two surfaces, according to figure 7.5. The voltage to the actuators could
be controlled from for example a Matlab-Simulink routine. By modelling
the tool holder shank in Femlab the whole system could be simulated in the
same software. Doing this would give an idea of how much the dynamic
behaviour of the tool holder shank can be influenced by actuators.

Figure 7.5. Simulation of the actuators.

45

8 References
1. Andersson, P., Anordning fr vibrationsassisterad bearbetning i
supportsvarv, MSc Thesis, Lund, 1990.
2. Bjrklund, S., Hgeryd, L., Lenner, L., Modern Produktions Teknik Del
1, Liber Utbildning, 1993.
3. Craig, Jr., Roy, R., Structural Dynamics, an Introduction to Computer
Methods, John Wiley & Sons, 1981.
4. Ewins, D.J., Modal Testing: Theory and Practice, Research Studies
Press Ltd., 1984.
5. Golub, G.H., Van Loan, C.F., Matrix computations, The Johns Hopkins
University Press, 1989.
6. Hkansson, L., Adaptive Active Control of Machine Tool Vibration in
a Lathe, PhD Thesis, KFS AB Lund, 1999.
7. Inman, D.J., Engineering Vibration, Prentice-Hall, Inc., 1996.
8. Maia, Silva, He, Lieven, Lin, Skingle, To, Urgueira, Theoretical and
Experimental Modal Analysis, Research Studies Press Ltd, 1997.
9. Newland, D.E., Random Vibrations, Spectral Wavelet Analysis,
Longman Singapore Publishers Pte Ltd, 1993.
10. PCB Piezotronics, AVC Instrumentation,
http://www.pcb.com/products/avc/avc713a01.html, 1999-07-12
11. Philips, Piezoelectric Ceramics Properties And Applications, N. V.
Philips Gloeilampenfabrieken, 1991.

46

Appendix A

Figure A.1. The first mode for the tool holder shank free-free with beam
elements.

Figure A.2. The second mode for the tool holder shank free-free with beam
elements.

Figure A.3. The third mode for the tool holder shank free-free with beam
elements.

47

Figure A.4. The first experimental mode for the tool holder shank free-free.

Figure A.5. The second experimental mode for the tool holder shank freefree.

Figure A.6. The third experimental mode for the tool holder shank freefree.

48

Appendix B
12000

Measured Frequency [Hz]

10000

8000

6000

4000

2000

2000

4000
6000
8000
Predicted Frequency [Hz]

10000

12000

Figure B.1. Measured versus predicted natural frequencies for the tool
holder shank free-free.

12000

Measured Frequency [Hz]

10000

8000

6000

4000

2000

2000

4000
6000
8000
Predicted Frequency [Hz]

10000

12000

Figure B.2. Measured versus predicted natural frequencies for the tool
holder and tool holder shank free-free, 40 mm hangout.

12000

Measured Frequency [Hz]

10000

8000

6000

4000

2000

2000

4000
6000
8000
Predicted Frequency [Hz]

10000

12000

Figure B.3. Measured versus predicted natural frequencies for the tool
holder and tool holder shank free-free, 60 mm hangout.

49

Figure B.4. Traceline and nodes for the tool holder shank free-free, seen
from above.
Mode 1

Experimental

N1

1.0
N7
N2

Predicted
1.0

N5

-1.0
N4

N3

-1.0

N6

Figure B.5. Comparison of mode shape 1 for the tool holder shank freefree, experimental versus predicted.
Mode 2

N1
Experimental

1.0
N7

N2
N5
-1.0

N6

N3

Predicted
1.0

N4

-1.0

Figure B.6. Comparison of mode shape 2 for the tool holder shank freefree, experimental versus predicted.

50

Mode 3

Experimental

1.0

N5

N4
-1.0

N7

N1

Predicted
1.0

N2

N3

-1.0

Figure B.7. Comparison of mode shape 3 for the tool holder shank freefree, experimental versus predicted.

51

Appendix C
The FE-model for the Tool Holder Shank Free-Free with Solid
Elements
The tool holder shank is modelled as a solid part spliced into two parts, see
figure C.1. The reason for splicing the solid model in two parts is to mesh
the two different parts with different mapped mesh.
Part 2
Part 1

Figure C.1. The tool holder shank modelled in two parts.

Part 1 is meshed with 23 elements in longitudinal direction, 8 elements in
the wide direction and 5 elements in the narrow direction, see Figure C.2.
Part 2 is meshed with 4 elements in the longitudinal, 10 elements in the
wide and 6 elements in the narrow direction. The total number of elements
is 1240 and the total number of nodes is 1735.

52

Figure C.2. The mesh of the tool holder shank in longitudinal, wide and
narrow direction.
The mass of the tool holder shank in the FE-model is 1.06 kg. The
boundary conditions for the tool holder shank is free-free and the
eigenvalue problem is solved by the Lanczos method. Results see Figure
C.3-C.5.

Figure C.3. The first mode for the tool holder shank free-free with solid
elements.

Figure C.4. The second mode for the tool holder shank free-free with solid
elements.

53

Figure C.5. The third mode for the tool holder shank free-free with solid
elements.

54

Appendix D

Figure D.1. Coherence function for the FRF below.

Figure D.2. Accelerance FRF for the tool holder and tool holder shank in
concrete lump.

55

Figure D.3. Coherence function for the FRF below.

Figure D.4. Accelerance FRF for the tool holder and tool holder shank in
lathe.

56