Advanced FRF based determination

of structural inertia properties

2
LM8 InternationaI | [email protected] | www.lmsintl.com
Introduction
The challenge for todays engineers is not only
to develop a better product, but to bring it to the
market quickly. In order to meet this challenge it is
important to establish a close relationship between
simulation data and experimental results, and rigid
body characteristics are keys to speeding up the
modeling and simulation process. Because it is not
always possible to accurately calculate the rigid
body properties of a complex structure, being able
to estimate them from measured data and use
them to validate and update a fnite element model
provides the relationship we need.
This application note outlines the characteristics
of rigid body properties and their importance. It
describes how they can be estimated from modal
measurement data and demonstrates how these
estimates compare with calculated FE data.
Why inertia properties are needed
The determination of inertia properties; the center of
gravity and the moments of inertia of a structure, is
important if engineers are to meet the challenges of an
effective design and development cycle. For a simple
structure, these inertia properties can be calculated
mathematically. For a complex structure however,
constructed from several components composed of
different materials, such as a car body, a truck, or a
washing machine, a fnite element model is required. To
generate these models in itself, is not a trivial task.
Rigid body properties which describe the inertia
properties of a structure can be used to complete or
update a fnite element model and serve as input for a
simulation model. Rigid body modes which represent the
mass behavior of a structure are an essential component
of a realistic modal model over the low frequency range.
An understanding of inertia properties is essential when
considering the assembly of a complex structure from
its constituent components, such as the motor, gear box,
drum, damper, of a washing machine. In order to predict
the kinematic and dynamic behavior of the structure,
a fnite element model for each of these components
would be required. The creation of all these fnite element
models is expensive, so in this case it is both effective
and practical to carry out a modal test on the components
to generate the rigid body properties and rigid body
modes that can then be used for a multi-body dynamic
calculation.
Rigid body properties and rigid body modes are also
applicable in the context of coupling a small rigid
component to a complex fnite element model, such
as coupling an exhaust to a car body for instance. The
rigid body modes of the small component though rarely
available from a classical experimental modal test are vital
for the accuracy of a modal coupling calculation (modal
synthesized method).
A further example of the important application of rigid
body characteristics relates to vehicle engine and body
mount systems. In order to optimize the powertrain
mount, the vehicle is divided into three main subsystems;
the powertrain, the body and the suspension. The
powertrain itself is modeled as a rigid body system with
external excitations, and its mounting system is modeled
by its dynamic stiffness curves over the frequency range
of interest. The body and suspension are represented
by their measured FRF model. The model for the whole
vehicle is then constructed by combining the three
subsystem models using the substructuring method. In
considering the powertrain as a rigid body system, inertia
properties will be required to build up the equations of
motion.
3
LM8 InternationaI | [email protected] | www.lmsintl.com
Experimental methods to
determine inertia properties
PenduIum teat
Traditionally, inertia properties were obtained
experimentally by a somewhat laborious pendulum test
that involved suspending the structure freely in different
directions. This is not only diffcult in the case of a
complex structure such as a car body, it is also very time
consuming.
FRF baaed method
An effective and accurate alternative to the pendulum test
is a modal test. Inertia characteristics are determined by
calculating rigid body properties based on the measured
frequency response functions and mass line. The modal
test can be either impact hammer or shaker excitation
on the structure in free-free conditions. In most cases
6 excitation locations in single direction and between 8
and 12 response locations in 3 directions are suffcient.
The test involves no special equipment and only limited
measurement effort. The new LMS Test.Lab Rigid
Body Properties workbook uses FRF measurements to
determine inertia properties.
How to determine rigid body properties
Two essential requirements for this process are an
accurate geometrical wire-frame model and the weight of
the structure in kilograms. The structure then needs to be
suspended in free-free conditions and frequency response
functions measured using either hammer or shaker
excitation.
Mass line methods are used to compensate for the
possible poor quality of the measured FRFs at low
frequencies, and to reduce the least square error. A
frequency band which best represents the mass line is
selected.
Rigid body properties can then be calculated with a
least square solution over this selected frequency band.
Based on the defned geometry the resulting rigid body
properties (center of gravity and moment of inertia), rigid
body modes are synthesized using user defned frequency
and damping values. The results can be validated either
by observing the animation of rigid body modes or by
comparing the rigid body properties to those of a fnite
element model.
1heoreticaI background
A mass-spring model of single degree of freedom is shown
in Figure 1.
ground
c k
m
f(t)
x(t)
Figure 1: Mass-spring model of single degree of freedom
The equation of motion in the frequency domain is
expressed by:

(m
2
+ jc + k) x() = F ()
in displacement
format and


(m+
c
j

2
) x () = F () in acceleration format.
It can also be expressed as inertance (acceleration/force)
by:


x ()
F ()
=
1
(m+
c
j

2
)
.
The frequency response functions (compliance, mobility
and inertance) are shown in Figure 1. At low frequencies,
below the resonance frequency, i.e.

( <<
n
) , the
inertance is approximately equal to the asymptote shown
in the lower graph in Figure 2. This asymptote is called
the stiffness line and is approximately equal to

2
k
.
At higher frequencies, above the resonance frequency,
i.e.

( >>
n
), the inertance is approximately equal to
the asymptote shown in the lower graph in Figure 1. This
asymptote is called the mass line and is approximately
equal to

1
m
.
For a single degree of freedom model, i.e. a rigid body
structure, the resonance frequency depends on the
mounting of the structure and the mass line that lies
above this resonance frequency. At the resonance
frequency, the structure vibrates as rigid body.
4
LM8 InternationaI | [email protected] | www.lmsintl.com
Maaa Iine methoda
At frequencies above the rigid body modes, the inertance
level becomes constant and is related to the mass value.
Determination of the mass line is required in order to
calculate the rigid body properties.
In practice however, a structure does not behave as a
purely rigid body; fexible modes are also present during
any classical modal test. In addition, the manner in which
the structure is suspended infuences the rigid body
frequencies. It is of fundamental importance that the
structure is suspended in free-free conditions to ensure
that the rigid body motion has negligible effect on the
fexible modes.
Figure 3: Frequency response function used for
different mass line method
Figure 2: Different formats of a frequency response functions for a single
degree of freedom system. Compliance (upper), mobility (middle) and
inertance (lower)
Thus the identifcation of the mass line may not be easy,
but three methods, that depend on the characteristics of
the FRFs, can be used to distinguish it.
Unchanged FRFa
When rigid body modes and fexible modes are suffciently
separated, for an inertance FRF, the curve between the
rigid body modes and the frst fexible mode appears as a
horizontal line. The frequency band this represents can
be used for the mass line method to extract rigid body
properties. The solid line (red curve) in Figure 3 shows
rigid body modes that are quite separate from the fexible
)modes.
Corrected FRFa
When insuffcient bandwidth exists between the rigid
body and the fexible modes, the measured FRFs must be
corrected by subtracting the fexible modes. Firstly the
fexible modes are estimated from measured FRFs using
modal parameter estimators such as LMS PolyMAX for
pole estimation and least square frequency domain (LSFD)
for mode shape estimation. Then FRFs are synthesized
from these modes (without lower residual) in the selected
frequency band and subtracted from the measured FRF.
This results in corrected FRFs in which the mass line of
the rigid body modes is no longer affected by the fexible
modes. Such a corrected FRF is shown by the dashed line
(green curve) in Figure 3.
Lower reaiduaI
In the situation where accurate measured FRFs are not
available in the frequency range between the rigid body
and the fexible modes, lower residual terms can be used.
Lower residual terms are estimated from the measured
FRFs using modal parameter estimators such as LMS
PolyMAX for pole estimation and least square frequency
domain (LSFD) solver for mode shape, upper and lower
residual term estimation. Lower residual terms represent
the infuence of the modes below the frequency band
used for modal parameter estimation. In the case where
the frequency band used for modal analysis extends only
to the frst fexible modes, they therefore represent the
infuence of the rigid body modes.
5
LM8 InternationaI | [email protected] | www.lmsintl.com
More explanation on the mass line methods can be found
in reference [4].
8yntheaia of rigid body modea
Rigid body modes are necessary to have a complete
modal model for either simulation or predictive purposes.
The preferred solution is of course to measure them,
but obtaining good quality data at the low frequencies,
where rigid body modes occur, is not easy. The main
reason for this is due to the diffculties in exciting these
low frequencies resulting in noisy FRFs in this range.
In addition, the frequency resolution used to identify
fexible modes is often insuffcient to accurately estimate
rigid body modes. An alternative is to synthesize mode
shapes based on geometrical coordinates and the inertia
properties and then add the rigid body mode shapes to
the modal model.
A free-free structure has six rigid body modes, and for
simplifcation, 3 translational and 3 rotational modes
around the principal axes of inertia are synthesized. The
residues depend on the total mass of the structure, the
principal axes of inertia and the principal inertia values.
Each mode has a mode shape vector and participation
factor, a resonance frequency and damping value. The
residue is expressed as the product of mode shape and
participation factor. The equation of motion of a single
degree of freedom system can be expressed as:

m x + c x + kx = F
Transforming into Laplace domain (s) this yields:


1 m
s
2
+ c
s
m
+
k
m
=
X
F


1 m
s
2
+ 2
0
s +
0
2
=
X
F
where

0
= k m= 2 f
0
with the natural frequency and


=
c
2m k m
.
The denominator has two poles:


s
1
=
0
+ j
0
1
2
s
2
=
0
j
0
1
2
.
Partial fraction expansion yields:


A
s
0
+ j
0
1
2
( )
+
B
s
0
j
0
1
2
( )
=
1 m
s
2
+ 2
0
s +
0
2


A = B =
j
2m
0
1
2

X
F
=
R
i
s
+
R
i
*
s
*

The displacement residue of a single degree of freedom
system thus has an amplitude:

1
2m
0
1
2
and a phase of -90.
Rotational rigid body modes are derived similarly, except
that the rotational inertia should be translated to a linear
mass. With the angular displacement and the moment,
the angular equation of motion of a free-free inertia is


M
=
1
I
and

M
=
1
I
2

X r
x
F .r
F
=
1
I
2

X
F
=
r
F
r
x
I
2
Replacing

1
m
by

r
F
r
x
I
in the residue for a translational
mode gives us the residue for a rotational mode.

R
i
=
r
F
r
x
2I
0
1
2
.
Figure 4 show the distances

r
F
and

r
X
used in the
moment applied around the Z axis (perpendicular to the X
and Y axes) and a reference point in the X direction.
F

r
F

r
X
Y
X

r
Y
Figure 4: Distance rX and rF used for rotational residues

LM8 InternationaI | [email protected] | www.lmsintl.com

The frequency and critical damping ratio can be arbitrarily
chosen. If the damping of the rigid body mode is chosen
to be small so that

1
2
is approximately 1, the
residues become:


R
translational
=
1
2m
0
=
1
2m 2 f
0
R
rotational
=
r
F
r
X
2I
0
=
r
0
r
2I 2 f
0
Choosing a rigid body frequency that is too small can give
rise to numerical problems, but if the frequency is too
high, it will infuence the calculated fexible modes, after
a coupling calculation for example. Acceptable values for
the rigid body frequencies lie between 1/10 and 1/100
below the frst fexible mode and above 1/1000 of the
highest fexible mode.
The calculated inertia properties can be controlled by
synthesizing FRFs (inertance) for each rigid body mode
on a signifcant DOF as response and the DOF used to
determine the participation factor as reference. In the
upper frequency range, the FRF should have a value given
by:


X
F
=
1
m
for a translational inertia and


X
F
=
r
F
r
X
I
for a rotational inertia.
More detailed explanations on this can be found in
reference [2]
How to guarantee the accuracy
of inertia properties
The quality of experimental rigid body properties depends
on the accuracy of the measurements. The points given
below provide some practical considerations that can be
implemented to realise the best possible results.
Suspend the structure in free-free boundary conditions.
Use a very soft suspension that allows rigid body
movements in all directions; translational and rotational.
Measure the location of the measurement points as
accurately as possible.

Calibrate the transducer as accurately as possible.

Aim for high measurement accuracy in the lower
frequency range. The frequency resolution must be
very high, so use a small frequency step and a long
measurement time. In general, at least 5 frequency
lines in the selected frequency band used for rigid body
estimation will be suffcient.
Perform the modal FRF test using a single reference
excitation.
Use a hammer for excitation (impact test) given the
number of excitation points.
Use a frequency range of excitation that includes the
frst fexible mode.
Because the frequency range of interest is low, use an
appropriate hammer and hammer tip (soft rubber).
Typically 6 different excitation DOFs are required (at
least one time on each direction X, Y, and Z)
Between 8 and 12 response points in 3 directions are
required.
Validate the consistency of the measured FRFs,
such as correct calibration, correct direction and
sign of accelerometer, before starting the rigid body
calculation.
Choose the most appropriate mass line method based
on the available FRFs.
Synthesize the rigid body modes with a low frequency
(close to zero) and damping for further use in modeling
and simulation.
LMS Test.Lab Rigid body
properties calculator
ImpIementation
To use the LMS Test.Lab Rigid body properties calculator
requires a set of FRFs between excitation DOFs and
response DOFs. A frequency band, which represents
the mass line, is selected between the rigid body modes
and fexible modes (as shown in Figure 5). These mass
line values are used for calculating rigid body properties
according to the selected mass line method. All the
spectral lines of the frequency band are used as input
for a global (least square) solution. More detailed
explanations of the kinematic and dynamic equations
used are available in reference [4].

Figure 5: Frequency band of the FRFs represent the mass line which is between the rigid body modes and the frst fexible/deformation mode.
7
LM8 InternationaI | [email protected] | www.lmsintl.com
Workaheet ow
The LMS Test.Lab rigid body properties calculator uses
a frequency response function (FRF) to extract essential
inertia properties such as the center of gravity, moments
of inertia and principal axes of inertia. As with other
members of Test.Lab product family, the software guides
you through the different steps required to calculate rigid
body properties and synthesize rigid body modes.
In the frst Data Selection worksheet a selection of
FRFs is made for the analysis. Individual FRFs and a sum
of FRFs can be visualized and inspected in a dedicated
2D display. The three mass line methods described
above are available in addition to tools such as, rigid body
correlation and ODS animation that provide immediate
and frm data validation feedback before performing the
analysis.
In a second Calculate worksheet, only one click
is required to extract the rigid body properties and
synthesize rigid body modes with respect to the defned
geometry. The frequency and damping values are defned
by the user. The center of gravity and the principal
direction can also be visualised on the geometry model.
Assessing the performance of the
rigid body properties calculator on an
academic case
VaIidation
A representative academic model of an engine is
used to validate the performance of the calculator.
This structure was selected for a number of reasons;
the analytical inertia properties are easy to calculate,
the structure itself however is not trivially simple and
it exhibits low frequency fexible modes which allow
evaluation of the different mass line methods. Figure 6
shows the geometry of the structure.
FRFs were measured by impact testing. In total there
were 8 excitation points represented in Figure 7 and 10
response points measured in 3 directions represented in
Figure 8. In both cases the arrows represent the direction
of the references and responses.
Figure 6: Geometry of the representative academic structure of an engine.
Figure 7: Excitation directions on the representative academic structure
of the engine.
Figure 8: Response directions on the representative academic structure
of the engine.
8
LM8 InternationaI | [email protected] | www.lmsintl.com
Figure 10: Excitation directions of impact tests on the frame structure.
Figure 9 shows the sum of FRFs (red, solid) and some of
the individual FRFs (dotted).
A comparison of the results of the three mass line
methods is shown in Table I. Compared to the FE results,
all 3 methods give an excellent prediction, the lower
residual method being the closest.
Assessing the performance of the
LMS Test.Lab rigid body properties
calculator on real life applications
Frame
In this investigation a more realistic frame structure
is analysed. The inertia derived from an ANSYS fnite
element model with 2272 elements and 2347 nodes
was used as the reference for the experimental rigid
body analysis. An appropriate test setup was confgured
to measure 408 FRFs with 12 excitation DOFs and 32
response DOFs as shown in Figure 10 and Figure 11.
The frame structure is excited at the four corner nodes in
3 directions.
Figure 11: Response directions of impact testes on the frame structure.
Figure 9: Sum of FRFs (red, solid)
and some typical FRFs (dotted)
of the representative academic
structure of the engine
Table I : Comparison of the results of mass line methods and FE model
results.
9
LM8 InternationaI | [email protected] | www.lmsintl.com
Figure 12 shows the sum of all the FRFs (in red) and other
signifcant FRFs. The rigid body modes are suffciently
separate from the frst fexible modes, and the 25 Hz to
35 Hz frequency band is used for the experimental rigid
body properties calculation.
As the results in Table II show, all the mass line methods
are close to the fnite element result (ANSYS). More
details on this test case can be found in reference [4].
This test case also serves to illustrate the infuence of the
number of references and responses used in the modal
test. An assessment of these factors can be made using
the concept of rigid body correlation.
Rigid body correlation measures the correlation between
the operational defection shape and the rigid body
modes in a selected band. While it is obvious that a
minimum number of references and responses are
required for any modal test, the quality of the measured
FRFs also infuences the quality of the estimated rigid
body properties. When there are more than 6 references
and 24 responses (8 response locations in 3 directions),
conditions defned as the minimum requirement, the
quality of the FRFs can be evaluated using rigid body
correlation. Rigid body correlation for input should be
higher than 90% and that for the output should be higher
than 98%. The column with 6 ref 34 resp (6 references
and 34 responses, highlighted in yellow) in the Table III
and the column with 12 ref 27 resp (12 refenences and
27 responses, highlighted in yellow) in the Table IV show
that the optimum values of the rigid body correlation
are indeed dependent on number of references and
responses. The column with 6 ref 27 resp (6 referenced
and 27 responses, highlighted in green) in the Table V
shows the combination of the two.
Figure 12: Sum of FRFs (red) and some signifcant FRFs (dotted).
Table II: Comparison of the results of the three mass line methods and the
ANSYS results on the frame structure.
Table III: Infuence of reducing the number of reference on the frame
structure.
10
LM8 InternationaI | [email protected] | www.lmsintl.com
hxhauat
A second real life test case concerns an exhaust system.
The purpose of this exercise was to include rigid body
modes in a modal coupling calculation in which the
exhaust is coupled to a car body. In this particular case
a detailed fnite element model of the exhaust system
was available which allowed comparison between the
experimental results and the fnite element results. It
should be recalled that even though rigid body modes
are of primary importance concerning the accuracy of
a modal coupling calculation (modal synthesis method),
these rigid body modes are not usually available from
a classical experimental modal test due to their low
resonance frequency.
The exhaust system was suspended in free-free
conditions, and a modal test carried out using a total of 10
excitation points and 10 response points in 3 directions.
The excitation points were chosen such that each point
excited at least 1 rigid body mode. All directions were
excited with a hammer and the excitations points were
distributed evenly over the exhaust system. To avoid mass
loading effects generated by modifcation of the setup,
the locations of the accelerometers were not moved
between measurements, only the measure directions were
changed.
Figure 13 shows the sum of all the FRFs (in red) and one
particular FRF. This clearly shows that a mass line is not
easy to distinguish. In this situation the best policy is
to use the Lower residual method. The poor quality of
the mass line is due to the suspension of the test setup;
it was not soft enough and so infuences the rigid body
movement of the structure on all the directions.
It is clearly shown in Table VI that the lower residual
method gives better results than the other two methods,
especially for the moment of inertia.
A study made on the infuence of number of reference
and response on the rigid body calculations for this
case study is described in reference [6]. This reference
also highlights the importance of obtaining good quality
measurements, assessing infuences such as the modal
parameter estimation method, the frequency band
for modal analysis, the frequency band for rigid body
properties calculation, the precision of the geometry
wireframe and others.
Table IV: Infuence of reducing the number of response on the frame
structure.
Table V: Infuence of reducing both number of reference and response on
the frame structure.
Figure 13: Sum of all the FRFs (red) and a typical FRF (blue) of the exhaust.
Table VI: Comparison of the FE model results and the three mass line meth-
ods on the exhaust.
11
LM8 InternationaI | [email protected] | www.lmsintl.com
Rigid body modes of the exhaust
12
LM8 InternationaI | [email protected] | www.lmsintl.com
Conclusion
This application note explains importance of and the
concepts behind rigid body properties. It proves that
based on classical FRF measurement, with only limited
measurement effort, it is possible to perform a fast and
accurate analysis which uses a least squares solution
to determine inertia properties together with rigid body
modes for simulation purpose such as modal-based
substructuring. Several mass line methods are presented
to compensate for the effects of the frst fexible mode
and eliminate the effects of poor quality FRFs. A number
of integrated validation tools, rigid body correlation,
animation and coloring feedback, help the user to validate
data used for rigid body properties calculation. The LMS
Test.Lab Rigid body calculator workbook is suitable for a
wide range of applications such as determination of the
center of gravity of
a truck cab
spoiler of a formula one
the drum of a washing machine
a satellite
a tractor
a tank
a helicopter
and others
Applications can also be extended to a simulation purpose
such as multi-body dynamic simulation, modal coupling
and modal-based modifcation.

References
[1] J. Toivola, O. Nuutila. Comparison of three methods
for determining rigid body inertia properties from
frequency response functions. Tampere University of
Technology, P.O. Box 589, SF-33101 Tampere, Finland.
[2] LMS international. How to add rigid body modes to
an existing modal model in CADA-X. LMS international
consulting reports, Leuven, Belgium, 1991.
[3] H. Okuzumi. Identifcation of the rigid body
characteristics of a powerplant by using experimental
obtained transfer functions. Central engineering
laboratories, Nissan motor Co., Ltd., June 1991.
[4] W. Leurs, L. Gielen, M. Burghmans, B. Dierckx.
Calculation of rigid body properties from FRF data:
pratical implementation and test cases. Proc. of the 15th
International Modal Analysis Conference, Tokyo, Japan,
1997.
[5] LMS international. LMS Test.Lab Modal Analysis
Workbook Modal Rigid Body Manual, Rev 7B, LMS
international, Leuven, Belgium, 2006.
[6] LMS international. Measuring rigid body properties
of an exhaust system: A comparative analysis. LMS
international, Leuven, Belgium, 1998.
[7] R. Madjlesi, A. Khajepour, F. Ismail, M. Wybenga, B.
Rice, J. Mihalic. Optimization of engine mounting systems
using experimental FRF vehicle model. Int. J. Vehicle
Noise and Vibration, Vol. 1, Nos. _ 2005.
LMS INTERNATIONAL
Researchpark Z1, Interleuvenlaan 68
B-3001 Leuven [Belgium]
T +32 16 384 200 | F +32 16 384 350
[email protected] | www.lmsintl.com
Worldwide For the address of your local representative, please
visit www.lmsintl.com/lmsworldwide
LMS is an engineering innovation
partner for companies in the automotive,
aerospace and other advanced
manufacturing industries. LMS enables its
customers to get better products faster
to market, and to turn superior process
effciency to their strategic competitive
advantage. LMS offers a unique
combination of virtual simulation software,
testing systems and engineering services.
LMS is focused on the mission
critical performance attributes in key
manufacturing industries, including
structural integrity, system dynamics,
handling, safety, reliability, comfort and
sound quality. Through our technology,
people and over 25 years of experience,
LMS has become the partner of
choice for most of the leading discrete
manufacturing companies worldwide.
LMS is certifed to ISO9001:2000 quality
standards and operates through a network
of subsidiaries and representatives in key
locations around the world.