Você está na página 1de 13

See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/319464738

Finite element model for modal analysis of


engine-transmission unit: numerical and
experimental...

Conference Paper · September 2017

CITATIONS READS

0 52

7 authors, including:

Patrick Langer Steffen Marburg


Technische Universität München Technische Universität München
15 PUBLICATIONS 6 CITATIONS 232 PUBLICATIONS 1,903 CITATIONS

SEE PROFILE SEE PROFILE

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

Pianissimo - Development of Noise Optimized Stage Elevator View project

Noise propagation and radiation of an PMSM at mean value free radial field stimulation View project

All content following this page was uploaded by Patrick Langer on 04 September 2017.

The user has requested enhancement of the downloaded file.


Finite element model for modal analysis of engine-transmission
unit: numerical and experimental investigations
P. Langer1 , C. Jelich1 , A. Hoppe1 , A. Schneider1 , C. Guist2 , K. Sepahvand1 and S. Marburg1
1
Chair of Vibroacoustics of Vehicles and Machines
Department of Mechanical Engineering
Technical University of Munich, 85748 Garching, Germany
2
BMW Group

ABSTRACT
The scope of this paper is to enhance the reliability of finite element models by taking uncer-
tain parameters of an underlying structure into account. Such parameters can be system-related
(e.g. geometry, material behavior, boundary conditions and mesh density) or influenced by gen-
eral assumptions in the process of modeling. Here, an engine-transmission assembly is studied.
This investigation develops instructions on how detailed a numerical model must be built in order
to get satisfactory results for each component. For this purpose, the system’s eigenfrequencies
are computed numerically and compared to experimental results from a laser Doppler vibrometer
(LDV). Within this contribution, finite element models based on computer aided design (CAD) and
computer tomography (CT) data are chosen. By choosing an engine-transmission assembly with
unusually high manufacturing deviations, the effect of geometry variations is highlighted. Further-
more, after developing sufficient knowledge of the dynamic behavior and the simulation accuracy
for single components, the parts are bolted together and the complexity of the problem is increased.
This work presents some limitations of finite element modeling, which can be useful to find the most
accurate concept for modeling the real dynamic behavior.
Keywords: structural vibration, experimental modal analysis, parameter identification, uncer-
tainty quantification, finite element method, I-INCE Classification of Subjects Number(s): 47.3

1. INTRODUCTION

Due to improvements in computing capabilities over the past decades, the finite element method
(FEM) has become one of the main tools used by engineers in the design process. Additionally,
FEM is of crucial importance in industry for reducing the number of experimental examinations.
However, a major issue with this method is that every finite element model possesses some degree
of uncertainty due to topological and material parameters, initial and boundary conditions, forcing
terms, etc. The results from finite element models can be trusted only if they are validated with
experimental or analytical results. For this, it is highly recommended to take uncertainties into
account.
Sargent [1] has given general recommendations regarding model uncertainties that are impor-
tant for model creation and the development process. These recommendations include ideas about
how to represent results together with their associated uncertainties and important steps in the
model validation process. Studying the discrepancy between model-based simulations and phys-
ical reality is described as uncertainty quantification (UQ) [2]. Kompella and Bernhard [3] have
introduced an approach to determine the uncertainty in a production line. Their findings emphasize
the importance of uncertainties when producing the final product. They measured the variation
of two structure-borne and two airborne paths of 57 apparently identical Isuzu pickup trucks. A
loudspeaker and an impact hammer were used for measuring the airborne and structure-borne
1
Email: p.langer@tum.de

3695
transmission paths. For quantifying the uncertainties the results were compared with a reference
truck. This method has been applied in various practical engineering problems involving uncer-
tainty and is well explained in literature, see [4–8]. Hills et al.[9] have compared the measurement
variability of audio-frequency responses of a hatchback vehicle model based on a three-door (411
vehicles) and five-door (403 vehicles) derivative and a mid-sized family five-door car (316 vehicles).
In summary, the frequency response functions (FRFs) varied by approximately 5 − 15 dB over the
frequency range 0 − 1.000 Hz for the structure-borne and airborne paths.
Especially for structures with screwed joining areas, a further possibility of uncertainty in virtual
modeling is included. Since virtual models employ assumptions and simplifications to depict the
real structure’s behavior, the impact of these have to be analyzed, particularly when facing contact
problems in joining areas. Here, a distinction must be drawn between expanded joints and those
joints where the joint area is small compared to the pressure cone. Therefore, the location of the
joints with respect to the vibration system is crucial. Geisler [10] laid out a synopsis of contact
segment elements or contact elements utilizing the segment-to-segment formulation. Thin-Layer-
and Zero-Thickness-elements were listed as being capable of offering implementation possibilities
for a wide range of constitutive relations. Neher [11] used beam elements as a simplified model for
bolts together with a Multi-Point-Constraint at the joint interface with a diameter three times larger
than the fastener’s diameter. Their simple models showed accurate results for the stiffness of large
clamping lengths, but overestimated this quantity for small clamping lengths. The authors examine
the influence of one simple modeling strategy of the pressure cone at the contact zone in such a
way that the modal superposition is still possible. Uncertainties in contacting surfaces and dynamic
problems of bolted joints are further described in [12–14].
This paper presents a comparison of the numerical and experimental results for monolithic and
assembled complex structures. The Software Abaqus/CAE was used as pre- and postprocessor.
This paper is organized as follows: After the introduction, a review of the investigated finite
element models is given. Afterwards, the measurement setup of the experimental modal analysis
is described. A comparison of numerical and experimental results and the material parameter
identification method are presented. In the end, a short summary and a conclusion is given.

2. MODEL DESCRIPTION

A crank case and an oil sump from an engine-transmission assembly are considered. To determine
the influence of the geometry on the eigenfrequencies, two different finite element model types
are investigated. TYPE 1 is a finite element model based on computer-aided design data and
TYPE 2 is a model based on a geometry scan by a computer tomography (CT). By choosing an
engine-transmission assembly with unusual high manufacturing deviations, the effect of geometry
variations is highlighted. Fig. 1(a) and Fig. 1(b) show the finite element models built using the
computer aided-design data (TYPE 1). There are two selected components of the complete engine-
transmission assembly which are shown in Fig. 1(c). The model TYPE 1 is meshed with quadratic
tetrahedral elements (C3D10) with an average edge length of 7 mm. The total number of elements
of the crank case is 892.998 and of the oil sump it is 89.082. The percentage of distorted elements in
the finite element models is 0.02 % of the total number of elements. Model TYPE 2 is meshed with
linear tetrahedral elements (C3D8) with an average edge length of 2 mm. The shorter element edge
length is sufficient to accurately interpolate the geometry points gained by the CT scan. The total
number of elements of the crank case and of the oil sump are 12 million and 2.105.498, respectively.
The percentage of distorted elements in the finite element model of the crank case is 0.02 % from
the total number of elements. For the oil sump no distorted elements are generated. Langer
et al. [15] show that the influence of having this amount of distorted elements on the calculated
eigenfrequencies is negligible.
The specimens are made of aluminum, for which linear elastic material behavior, i.e. Hooke’s
Law, is assumed. The material properties are determined experimentally. In all simulations, ideal
free-free boundary conditions are used.

3696
(a) (b)

(c)

Figure 1 – Test samples; (a) finite element model of the crank case (TYPE 1); (b) finite element
model of the oil sump (TYPE 1); (c) finite element model of the engine-transmission assembly
(TYPE 1)

3. MEASUREMENT SETUP

The experimental results highly depend on the precision of the measurement equipment and the
apparatus used in the experiment. Thus the technique used for measuring the excitation force and
the structural response should be chosen carefully. Within this investigation two contact measure-
ment setups are used: Method A and Method B.
A common way to excite the structure is to use an acoustic source and measure the sound pres-
sure throughout the structure. In this case, a force response function is achieved [16]. This method,
however, is limited in the excitation energy and thus can only be used for small structures. Another
possibility is the excitation via a magnetic field in which a force is applied on the structure [17]. This
method is limited to conductive specimens.
For these reasons, the authors use an excitation by impact hammer and shaker. Regarding the
impact hammer, a very short pulse duration, a perpendicular impact on the surface and a sufficient
resolution of the pulse signal must be ensured. An excitation of the structure by an electrodynamic
shaker is avoided if possible, because the connection of the structure with a bending flexible bar
tunes the dynamic behavior of the structure, which influences the measurements. However, to de-
termine the mode shapes, measuring with a shaker is more convenient whereas the best method for
measuring the resonance frequencies is excitation with an impact hammer. To connect these reso-
nance frequencies with the correct deflection shape, having data from shaker measurements is very
useful. A review of different measurement techniques is given by Ewins [18]. The deflection shape
is measured by a laser Doppler vibrometer in front of the components of the engine-transmission
assembly.
Fig. 2 shows one measuring setup for the oil sump: elastic strings (1), Laser Doppler Vibrometer
(2), oil sump (3), shaker (4) and vibration-insulated table (5). Handing over the frequency response
functions (FRFs) to ME’Scope [19] as a postprocessing tool, the eigenfrequencies and the associ-
ated mode shapes are identified.

3697
2
1

4 5

Figure 2 – Experimental setup: elastic strings (1), Laser Doppler Vibrometer (2), oil sump (3),
shaker (4) and vibration-insulated table (5).

Approximately the same setup is used for the experimental modal analysis of the crank case.
More details on experimental modal analysis are given by Langer et al. in [20].

4. RESULTS

5. IDENTIFICATION OF MATERIAL PARAMETERS

The material parameters the Young’s modulus E and the Poisson’s ratio ν are determined using
ultrasonic measurements. The sound velocities of longitudinal (cl ) and transversal (ct ) waves in an
elastic continuum are given by Eq. (1) and Eq. (2), respectively [21]:

E · (1 − ν)
cl = (1)
 · (1 + ν) · (1 − 2ν)

and: 
E
ct = . (2)
2 · (1 + ν)
A transformation of the two equations gives the Young’s modulus E :

cl 2 ·  · (1 + ν) · (1 − 2ν)
E(cl ) = (3)
(1 − ν)

E(ct ) = 2 · ct 2 ·  · (1 + ν) (4)
Assuming equality of Eqs. (3) and (4), ν is obtained as:

2ct 2 − cl 2
ν= . (5)
2 · (ct 2 − cl 2 )

Given the longitudinal and transversal sound velocities cl and ct , the Young’s modulus E and the
Poisson’s ratio ν can be easily obtained from Eqs.(3),(4) or(5). The ultrasonic measurements are
performed using the ultrasonic module UT/Mate from Vogt ultrasonics and transducers for longitu-
dinal and transversal waves with a frequency of 5 MHz. All measurements are taken in reflection
(pulse-echo) mode on specimens with plane-parallel grinded surfaces. On each specimen, ten
individual measurements are taken for both longitudinal and transversal sound velocities.
The density was measured with hydrostatic weighing using Archimedes’ principle with a Mettler
Toledo 204AG scale and a resolution of 0.1 mg. The mass of the specimen in air and water was
measured with the hydrostatic weighing method using Archimedes’ principle with a Mettler Toledo
204AG 0.1 mg scale. Therefore, each specimen was first weighed in air. The mass mwater was
determined by weighing the specimen in distilled water using a special weighing frame. In this
case, the weight of the specimen is reduced by the weight of the displaced water. Given the two
weights mwater and mair and the density of water water and air air at the temperature of weighing, the
density  of the specimen can be calculated by Eq.(6) without knowing the volume of the specimen.

3698
Therefore, this method is very useful for determining the density of irregularly-shaped specimens:

mair · (water − air )


= + air (6)
mair − mwater
The uncertainty of the Young’s modulus was calculated with error propagation of density Eq.(6) as
well as the ultrasonic measurement (Eqs.(3),(4) and (5)) considering a systematic error and the
scattering of individual measurements. Table 1 shows the mean values and the uncertainties of
material parameters of the oil sump and the crank case.

Table 1 – Mean value with uncertainty of material parameters; E: Young’s modulus; ν: Poisson’s
ratio; : density

Crank case Oil sump


E [GPa] 72.53 ± 1.71 % 70.99 ± 1.83 %
ν [−] 0.323 ± 0.009 % 0.343 ± 0.008 %
 [g/cm3 ] 2.67388 ± 0.0271 % 2.68077 ± 0.0006 %

6. COMPARISON OF EXPERIMENT AND SIMULATION

In this section, a comparison between measurements and numerical results is presented and dis-
cussed. The relative deviations εP of the first natural frequencies is calculated by
 
fN
εP = − 1 · 100%, (7)
fLDV

where fN is the solution of the finite element model (TYPE1 or TYPE2) and fLDV represents the
eigenfrequency measured by a Laser Doppler Vibrometer.
Fig. 3 shows the relative deviation of the first six obtained natural frequencies from the experi-
mental modal analysis (EMA) of the oil sump. For this purpose, simulation results of two models
based on a computer tomography scan and a computer-aided design are compared. The material
parameters used in the numerical modeling process, such as Young’s modulus E , Poisson ratio ν
and density , are constant. Therefore, the presented deviation arises solely due to the geometry
variations. Relating these deviations to the most sensitive input parameter, the Young’s modulus
E , the input values of E are varied towards the upper and lower limits calculated by the afore-
mentioned parameter identification method. Regarding the virtual model based on CT scans, the
numerical results deviate by about 2 % from experimental results. In contrast, the finite element
model based on CAD data has a deviation of up to 7 %. This difference is explained by the un-
usually high deviation in shape and geometry between the CAD model and the CT model from a
different real manufactured structure. As mentioned before, the experimental structure is chosen
such that the deviation in shape and geometry is as high as possible to highlight the effect of these
deviations.

3699
Deviation fromexperimental results–oilƐƵŵƉ
6

relativ deviation [%]


2

-2

-4

-6

-8

-10
eigenfrequencies of EMA 269 Hz 450 Hz 727 Hz 905 Hz 1189 Hz 1342 Hz
numerical model based on CT 2͘6 2͘4 2͘8 2͘0 2͘8 4͘0
numerical model based on CAD -7͘0 -7͘0 -4͘0 -5͘3 -6͘6 -5͘1

Figure 3 – The first six natural frequencies of the oil sump; CT: numerical model based on
computer tomography scan; CAD: numerical model based on computer aided design data;
EMA: experimental modal analysis; error indicator: due to the measurement uncertainty of the
Young’s modulus E.

To quantify the uncertainty in geometry, Fig. 6(a) shows a match between the shape of the
computer aided-design model and the computer tomography scan of the real physical structure.
Comparing both models, a maximum deviation of 1.4 mm is observed in the highlighted radii. The
deviation at the surfaces is mostly ±0.4 mm.
Fig. 4 shows the mode shapes associated with the first six natural frequencies. In each mode
shape, areas with high mode shape deviations also feature high geometry deviations. In other
words, geometry deviations between the numerical model and the experimental structure does
have a high influence on the accuarcy of the numerically calculated eigenfrequencies and mode
shapes of the real structure. The same results were concluded by Langer et al. [22] in a structural
dynamics investigation of simple beam structure. Therefore, it is a challenging task to build up an
accurate numerical model without optimizing the material parameters in a non-physical approach.
This especially holds for thin-walled monolithic structures.

1st 2nd

3rd 4th

5th 6th

Figure 4 – Finite element model: The first six mode shapes of the oil sump

To quantify the geometrical influences in the case of stiffer structures, a crank case was consid-
ered. Fig. 5 shows the relative deviation of the first six natural frequencies obtained from exper-
imental results. The numerical models were built up from CAD and CT data. The finite element

3700
model based on CAD data has a lower deviation of the first six eigenfrequencies. The maximum
deviations of the numerical models based on CAD and CT are 1.2 % and −1.4 %, respectively. This
indicates that either, the geometrical uncertainties are much lower compared to the thin-walled oil
sump or the geometrical uncertainties have lesser impact on the numerical analysis.

Deviation to experimental resultsͲĐƌĂŶŬĐĂƐĞ


3

relative deviation [%]


1

-1

-2

ĞŝŐĞŶĨƌĞƋƵĞŶĐŝĞƐŽĨD -3 849 Hz 1144 Hz 1788 Hz 1926 Hz 2002 Hz 2037 Hz


numerical model based on CT 1.ϰ -0.ϲ 0.9 0.03 0.7 0.ϴ
numerical model based on CAD 0.6 -0.ϰ 0.ϱ -1.2 -0.ϰ -0.ϱ

Figure 5 – The first six natural frequencies of the crank case; CT: numerical model based on
computer tomography scan; CAD: numerical model based on computer aided design data;
EMA: experimental modal analysis.

In support of this statement, Fig. 6(b) shows a match between the shape of the computer-aided
design model and the computer tomography scan for the crank case. The maximum and minimum
deviations, with 0.9 mm and −1.0 mm are much lower compared to the oil sump. For the CAD model
of the crank case, large areas of the geometry fit the real structure more accurately. In addition,
Fig. 7 shows the mode shapes of the crank case. Areas with high mode shape deflections feature
are much lower deviation comparing the finite element model based on CAD data and the geometry
of the real structure.

[mm]
0.9
0.8

[mm]
0.6
1.4
1.2 0.4
0.9
0.2
0.6
0.0
0.3

0.0 -0.2

-0.3
-0.4
-0.6
-0.8
-0.9

-1.3 -1.0

(a) (b)

Figure 6 – Geometry check; (a) oil sump: Matching of the shapes from computer-aided design
model to the computer tomography scan of the real structure; (b) crank case: Comparison
of geometry of computer-aided design model to the computer tomography scan of the real
structure.

3701
1st 2nd 3rd

4th 5th 6th

Figure 7 – Finite element model: The first six mode shapes of the crank case.

7. COMPARISON OF DIFFERENT EXPERIMENTAL SETUPS

In this section, a comparison between different measurement setups used in the experimental
modal analysis is presented.
The results gained by using the impulse hammer and the shaker are compared to analyze
the influence of the shaker’s additional coupled mass. Fig. 8 shows the results of the first seven
eigenfrequencies determined with an impulse hammer and an electrodynamic shaker as excitation
in the experimental setup. The position of the shaker is shown in Fig. 2. Note that this position has
a high geometrical deflection, especially regarding the first mode.
The results show that using this measurement setup for the oil sump, the influence of the shaker
is negligible. Given the measurement results of the crank case, the authors come to the same
conclusion.
Comparison of experimental results with different excitation strategies
1.4E+03

1.2E+03

1.0E+03
impuls hammer
f [Hz]

8.0E+02

6.0E+02 Ɛhaker

4.0E+02

2.0E+02

1.0E+00
1 2 3 4 5 6
number of eigenfrequencies [Hz]

Figure 8 – The first seven eigenfrequencies determined with different measurement setups;
excitation strategy 1: impact hammer; excitation strategy 2: electrodynamic shaker

8. ASSEMBLY

For this investigation, the real engine components and the finite element models based on CAD
data are used. Therefore, the influence of the geometrical deviation on the eigenfrequencies of the
complex assembled structures can be shown.
The procedure for quantifying uncertainties and validated finite element models is described in
the previous sections. In order to get a highly-accurate assembled engine-transmission unit, each
model is numerically analyzed and compared with experimental results. The material parameters
are identified with the previously explained method. Taking material uncertainties into account, the
Young’s modulus is varied between its upper and lower bounds to get the minimal deviation of the
first six eigenfrequencies compared to the experimental results. Regarding the tapped joints, the
following kinematic constraints are used. According to the ABAQUS documentation [23], kinematic
coupling constraints are based on degree of freedom elimination. They limit the motion of a group
of nodes lying on two different surfaces.
The connection between the different components is modeled via tie constraints which are ap-
plied in the approximated high pressure zone at the contact surface as marked by (1) in Figure 9.
These constraints are defined using a surface-to-surface approach. The high-pressure zone is lo-
cated between the contact surfaces close to the screw joints (1). Its area is assumed to extend to
twice the thread diameter. Note that this is an empirical value commonly used in the industry for a
linear contact modeling. No other contact boundaries were defined on the joint surface.

3702
The screw mass is concentrated in a single point mass which is located on the screw axis at
a reference point (2). This point is connected with the screw cylinder by a surface-based coupling
definition, namely the continuum distributed coupling method (3). This distributes the motion and
the mass of the coupling nodes to the reference node. Regarding the mass distribution, various
weighting methods are provided. For element-based surfaces the weight factor is calculated using,
ri
wi = 1 − , (8)
r0
where wi is the weight factor at coupling node i, ri is the coupling node radial distance from the
reference node, and r0 is the distance to the furthest coupling node [23].
Langer et al. [24] have introduced an approach to determine the sensitivity of the screw tighten-
ing torque to the eigenfrequency of assembled structures. For the screw tightening torque of 30 Nm,
a converged physical model is determined, since it does not cause an increase in the eigenfrequen-
cies with higher tightening torque. Therefore, experimental results featuring this screwing setup will
be compared with numerical solutions. Table 2 shows the numerical and experimental results of
the first three global modes of the engine-transmission unit and their deviation.

Table 2 – The first three eigenfrequencies of global modes determined with simulation and
experiment; εP : relative deviation between numerical and experimental results.

FE model [Hz] Experiment [Hz] εP [%]


1st 542 562 −3.64
2nd 569 593 −4.07
3rd 804 836 −3.93

With regard to 80 tapped joints, their relatively strong nonlinear behavior, the approximation
of the high pressure contact zone and the high geometry uncertainties of the oil sump, the global
dynamic behavior is calculated in good agreement to experimental results. However, in comparison
to the real assembly, the numerical solution underestimates the stiffness. The total mass of the finite
element model and the physical model are approximately the same.
Fig. 10(a), 10(b) and 10(c) show the first three global mode shapes. The first and second modes
feature a small deflection around the Z axis whereas the third mode shape features a bending
around the Z axis and a torsion around the X axis. In comparison to the uncertainties due to
the modeled contact surfaces or some radii of the geometry the standing strucutral wavelength of
the global mode shapes are large. Therefore, these points can explain the reason for the small
deviation of the linearized numerical results from the experimental modal analysis.

Figure 9 – Modeling strategy of the connections between defined structures. (1) kinematic
coupling in the high pressure zone at the contact surface, (2) point mass to approximate the
screw mass, (3) continuum distributed coupling for distributing the mass to each coupling node
to the screw cylinder surface.

3703
=
=

<
<

;
;

(a) (b)

<

(c)

Figure 10 – (a) first global mode; eigenfrequency: 542 Hz; (b) second global mode; eigenfre-
quency: 569 Hz; (c) third global mode; eigenfrequency: 804 Hz

In order to estimate the influence of structural damping, Fig. 11 shows the averaged frequency
response function (FRF) over the total number of measuring points on the engine-transmission unit.
Since the amplitude around the resonance frequencies falls very steep, the engine-transmission
assembly only features low damping behavior.

50
DPSOLWXGH in db[db= 0 IRU1·10-6 m/s /N]

45

40

35

30

25

20

15

10

0
500 600 700 800 900
frequency [Hz]

Figure 11 – Experimental results: averaged frequency response function (FRF) over the total
number of measuring points of the engine-transmission unit.

9. CONCLUSION

This paper presents a method to quantify uncertainties in finite element models containing complex
monolithic and bolted parts using the example of an engine-transmission unit. For the numeri-
cal part, finite element models based on computer-aided design and computer tomography data
were generated. The material parameters were identified using ultrasonic measurements and a
hydrostatic weighing method making use of the Archimedes’ principle.
For calculating the system’s first natural frequencies of global mode shapes the authors con-
cluded:

3704
(I) Seeking accurate numerical results, the selected geometric tolerances used in the automotive
production line of thin walled should be very small. The deviation, especially on the radii and the
thickness, highly influences the accuracy of numerical analyses. Regarding structures without thin-
walled parts, the sensitivity of geometrical uncertainties is much lower. Numerical analyses of these
structures are thus more accurate.
(II) Geometrical and material uncertainties have a higher impact facing monolithic structures
instead of larger models with bolted connections.
(III) The presented modeling strategy of the contact areas is very efficient. With an averaged
deviation of 4 % for the first eigenfrequencies to experimental results, the finite element model is
very accurate.
(IV) Since the experimental modal analysis identified weak structural damping of the engine-
transmission unit, the impact of structural damping on the investigated eigenfrequencies can be
neglected in the finite element model.
(V) Numerical results based on linearized finite element models of large assembled structures
are nonetheless in good agreement with the highly nonlinear structural behavior of the correspond-
ing real structure.
Beyond that, the presented material identification method gives highly accurate results of the
Young’s modulus, the density and the Poisson’s ratio featuring uncertainties of < ±2.3 %, < ±0.171 %,
and < ±3.2 %, respectively. Using two different excitation strategies, i.a. an impulse hammer and
a shaker, it was identified that the dynamic shaker mass has no significant impact on the deter-
mined eigenfrequencies. Therefore, general statements saying that a dynamic shaker influences
the measured resonance frequencies do not hold.

ACKNOWLEDGMENTS

Special thanks go to the Bavarian Research Foundation for the financial support.

REFERENCES

[1] Sargent G. Verification and Validation of Simulation Models. In: Winter Simulation Confer-
ences. Orlando; 2005. p. 53–59.

[2] Ghanem RG. Uncertainty Quantification in Computational and Prediction Science. Interna-
tional Journal for Numerical Methods in Engineering. 2009;80:671âĂŞ672.

[3] M S Kompella RJB. Measurement of statistical variation of structural–acoustic characteristics


of automotive vehicles; 1993. .

[4] Li R, Ghanem R. Adaptive polynomial chaos expansions applied to statistics of extremes in


nonlinear random vibration. Probabilistic Engineering Mechanics. 1998;13(2):125–136.

[5] Lucor D, Su CH, Karniadakis GE. Generalized polynomial chaos and random oscillators.
International Journal for Numerical Methods in Engineering. 2004;60(3):571–596.

[6] Sepahvand K, Marburg S, Hardtke HJ. Stochastic free vibration of orthotropic plates using
generalized polynomial chaos expansion. Journal of Sound and Vibration. 2012;331(1):167–
179.

[7] Marburg S, Beer HJ, Gier J, Hardtke HJ. Experimental verification of structural–acoustic mod-
elling and design optimization. Journal of Sound and Vibration. 2002;252(4):591–615.

[8] Sephavand K, Marburg S. On construction of uncertain material parameter using generalized


polynomial chaos expansion from experimental data. In: IUTAM Symposium on Multiscale
Problems in Stochastic Mechanics. Karlsruhe; 2012. p. 4–17.

3705
View publication stats

[9] Hills E, Mace BR, Ferguson NS. Acoustic response variability in automotive vehicles. Journal
of Sound and Vibration. 2009;321:286–304.

[10] Geisler J. Numerische und experimentelle Untersuchungen zum dynamischen Verhalten von
Strukturen mit Fügestellen [Dissertation]. Universität Erlangen-Nürnberg; 2010.

[11] Neher J. Rechnerische und experimentelle Untersuchungen der Schallabstrahlung bei


Fahrzeuggetrieben [Dissertation]. Technische Universität Illmenau; 2012.

[12] Bittner U. Strukturakustische Optimierung von Axialkolbeneinheiten. Karlsruhe: KIT Scientific


Publishing; 2012.

[13] Ibrahim RA, Pettit CL. Uncertainties and dynamic problems of bolted joints and other fasteners.
Journal of Sound and Vibration. 2005;279:857âĂŞ936.

[14] Li WL. A new methode for structural model updating and joint stiffness identification. Mechan-
ical Systems and Signal Processing. 2002;16(1):155âĂŞ167.

[15] Langer P, Sepahvand K, Krause M, Marburg S. Simple vibroacoustic systems - Influence of


uncertainties in simulation and experiment. In: Pavic G, editor. Proceedings of NOVEM 2015.
Dubrovnik, Croatia; 2015. .

[16] Wu J, Moslehy FA. On modal testing using speaker for excitation. In: 13th International Modal
Analysis Conference; 1995. p. 24.

[17] Sodano HA. Non-Contact eddy current excitation method for vibration testing. In: Experimen-
tal Mechanics; 2006. p. 627âĂŞ635.

[18] Ewins DJ. Modal testing: theory and practice. University of Sussex, UK;.

[19] Vibrant Technology I. ME’scope Version 5.0 Operating Manual; 2009.

[20] Langer P, Sepahvand K, Krause M, Marburg S. Simple vibroacoustic systems - Influence of


uncertainties in simulation and experiment. In: Pavic G, editor. Proceedings of NOVEM 2015;
2015. .

[21] Krautkrämer J, Krautkrämer H. Werkstoffprüfung mit Ultraschall, 5. Auflage. Berlin: Springer-


Verlag; 1986.

[22] Langer P, Sepahvand K, Marburg S. Uncertainty quantification in analytical and finite element
beam models using experimental data. In: Cunha A, Caetano E, Ribeiro P, Müller G, editors.
Proceedings of the 9th International Conference on Structural Dynamics, EURODYN 2014.
Porto, Portugal; 2014. .

[23] Abaqus User and Theory Manual 6.10-3;.

[24] Langer P, Guist C, Marburg S. Efficient modeling in vibroacoustic systems with considera-
tion of uncertainties and comparison with experimental measurements. In: Pavic G, editor.
Proceedings of Internoise 2016; 2016. .

3706

Você também pode gostar