Special Issue Article

Journal of Low Frequency Noise,

Vibration and Active Control
Frequency spectrum of the human 0(0) 1–8
! The Author(s) 2017
head–neck to mechanical vibrations DOI: 10.1177/1461348417747179
journals.sagepub.com/home/lfn

Bin Yang1,2,3, Zheng Shi1, Qun Wang1, Feng Xiao1,

Tong-Tong Gu1, Yong-Kai Yin1 and Zhong-Liang Miao1

Abstract
This study is based on a real finite element human head–neck model and concentrates on its numerical vibration
characteristic. Frequency spectrum and mode shapes of the finite element model of human head–neck under mechanical
vibration have been calculated. These vibration characteristics are in good agreement with the previous studies. The
simulated fundamental frequency of 35.25 Hz is fairly similar to the published documents, and rarely reported modal
responses such as “mastication” and flipping of nasal lateral cartilages modes, however, are introduced by our three-
dimensional modal analysis. These additional modes may be of interest to surgeons or clinicians who are specialized in
temporomandibular or rhinoplasty joint disorder. Modal validation in terms of modal shapes proposes a necessity for
elaborate modeling to identify each individual part’s extra frequencies. Furthermore, it also studies the influence of
damping on resonant frequencies and biomechanical responses. It is discovered that damping has an inverse propor-
tionality between damping effect on natural frequency and that on biomechanical responses.

Keywords
Human head–neck model, frequency spectrum, mode shape, damping factor

Introduction
When subjected to mechanical vibration either by contact with vibrating structures or by action of an acoustic
field, the human head responds by vibrating in certain modes.1 The knowledge of how and why it vibrates will
make a contribution to a normal comprehension of human biomechanics. This may make it easier in evolution of
“a new medical instrument for early diagnosis of brain diseases”.2
Earlier studies on human head–neck vibration chiefly came from experimental dates that involved animals,
cadavers or volunteers. Bekesy3 investigated “the vibration response of a cadaver skull in an acoustic field and
reported that the first resonant frequency of the skull to be, 1800 Hz”. The experimental modal study, performed
by Franke1 on both skull filled with gelatin and the same an empty dry human skull, found that “the lowest
resonant frequencies were 800 Hz and 500 Hz individually”. Both Hodgson et al.4 and Sun et al.5 discovered that
“the resonant frequency of a cadaver head in their mechanical impedance analysis was approximately 300 Hz”. At
around the same time, Stalnaker and Fogle6 utilized “an electromagnetic shaker in their experimental tests of a
fresh unembalmed cadaver head and discovered that the resonant frequencies were 166 Hz and 820 Hz”. Khalil7
revealed that “the respective fundamental frequencies found were, 1385 Hz and, 1641 Hz performed by an
experimental study on two cadaver heads using an impact hammer”. Hakansson’s8 in vivo study utilizing
“skin penetrating titanium implants on patients’ temporal bone claimed that the lowest undamped natural fre-
quency were about 972 Hz”. By comparison, experimental researches on living human subjects illustrated the
frequency spectrum arranged from several Hz to 300 Hz.1,5,9–11 In spite of the bulk of worthful information

1
College of Automobile and Traffic Engineering, Nanjing Forestry University, Nanjing, China
2
State Key Laboratory of Advanced Design and Manufacturing for Vehicle Body, Hunan University, Changsha, China
3
Department of Mechanical Engineering, National University of Singapore, Singapore, Singapore

Corresponding author:
Bin Yang, College of Automobile and Traffic Engineering, Nanjing Forestry University, Longpan Road, Xuanwu District, Nanjing 210037, China.
Email: yangb123@126.com
Creative Commons CC-BY: This article is distributed under the terms of the Creative Commons Attribution 4.0 License (http://www.
creativecommons.org/licenses/by/4.0/) which permits any use, reproduction and distribution of the work without further permission
provided the original work is attributed as specified on the SAGE and Open Access pages (https://us.sagepub.com/en-us/nam/open-access-at-sage).
2 Journal of Low Frequency Noise, Vibration and Active Control 0(0)

supplied by experimental information, these tests do not only lead to affairs in ethics and morality, but also induce
plenty of research concerns such as restricted flexibility as well as biasness in experimental information owning to
nonstandardized experimental procedures and scarce subjects.
At the same time, numeric simulations that use finite element method provides a cost-effective choice to
experimental ways, with a possibility for evaluations going beyond the experimental moral limits and would
assist in more effectual equipment’s evolution. Nickell and Marcal12 had executed finite element simulations
utilizing a simplified 3D FE skull model with different boundary supports and reported that “the cardinal
head-neck frequencies for occipital, frontal and base supports respectively were 86 Hz, 68 Hz and 164 Hz”.
Therefore, similar finite element models performed by Ward and Thompson13 simulated “the fundamental fre-
quency of brain to be 23 Hz and 43 Hz”. Ruan14 developed “a 2D finite element model using a coronal sectional of
human head and discovered that the first natural frequency fell within 49 to 72 Hz”. Chu15 constructed “two 2D
midsaggital head models and discovered that the additional brain in the latter model lowered the fundamental
frequency from 286 Hz to 119 Hz”. These researches only analyzed models of the head. Recently,
Charalambopoulos2 developed “an analytical model of the skull–brain–neck system and found that the neck
introduced additional frequency in the lower frequency spectrum with the natural frequency of 595 Hz”. Hence,
Meyer et al.’s16 finite element simulations using a rigid-head-neck model demonstrated that the fundamental
frequency of a head–neck system was 3.01 Hz. Baroudi17 executed “a 3D modal analysis of an idealized cylindrical
skull–brain–CSF model using both analytical and FE methods and found that the fundamental frequency to be
26.66 Hz”.
These earlier FE studies had provided some insights undoubtedly in the human head’s dynamic character-
istics. Nevertheless, numerous simplifications and approximations had been made in the head’s geometry. In
the present work, an elaborate finite element model of human being head–neck, which involves the intracra-
nial and facial particulars like the subarachnoid space, has been built and its modal responses have been
calculated. Additionally, the simulated modal responses in terms of mode shapes and resonant frequencies are
compared with previous literature, unlike most previous studies comparing their resonant frequencies simply
and disregarded modal validation in terms of mode shapes. What is more, we likewise inquire into the
influence of damping dynamic characteristics, with the introduction of three damping factors in the complex
eigenvalue analysis, dissimilar most previous finite element studies using the traditional frequency extraction
approach.

Resonance behavior of the finite element model

Governing equation and finite element method
Vibration characteristic utilizing finite element method was executed by Abaqus 6.11 software. The dynamic
response’s governing equation is given as follows

½Mfu€ðtÞg þ ½Cfu_ ðtÞg þ ½KfuðtÞg ¼ fFðtÞg (1)

where [K], [C], and [M] are respectively the global stiffness matrices, damping, and mass; {F} is the external
_ , {u}, and {ü} are respectively the velocity, displacement, and nodal acceleration vectors.
applied load; fug
On the condition of ½C 6¼ 0 and fFg ¼ 0, the eigenvalue equation is quadratic as it contains an extra term of
ix½C
 
½K þ ix½C  x2 ½M fUg ¼ 0 (2)

where x is each eigenvector’s corresponding eigenfrequency and {U} is the amplitudes of all the masses (eigen-
vectors or mode shapes).
In this study, three global material damping factors of 0.1, 0.2, 0.4, which are within the scope of damping
factors of head–neck that is discovered in the literature,18–20 are embraced in this complex eigenvalue analysis of
freedom problem’s multi-degree. Furthermore, the biomechanical parameters like skull stresses and intracranial
pressure are assessed. It shall be mentioned that the stresses that are assessed are not due to any outside applied
Yang et al 3

Figure 1. Various components of the human head–neck segmented from CT and MRI data.

force, as a matter of fact, they are element stresses frge , which are simulated based on the following equation in
finite element method

frge ¼ ½De fege (3)

where fege is the elemental strain; ½De is the constitutive matrix of the element.

Development and validation of the human head–neck system

In the present work, the finite element human head–neck model was developed by means of the semi-automatic
method in the human head–neck complex. In the meantime, material properties were assigned to every finite
element. The human skull’s geometric information is acquired from a series of a 50th percentile volunteer’s CT
and MRI images. The brain is initially segmented into just the cerebellum and cerebrum without gray and white
matters’ further differentiation. The model is composed of the cranial skull with elaborate facial bone attributes,
cervical vertebrae, teeth, nasal lateral cartilages, nasal septal cartilage; brain components such as cerebellum and
cerebrum and the cerebrospinal fluid separating the brain and the skull (Figure 1).
An adaptive semi-automatic meshing technique is then utilized to optimize between element quality and com-
putation efficiency with Hypermesh software. Further mesh reparation is executed to get ready for 3D mesh
generation. In spite of the cavity-dominant skull’s complex geometry, it is possible to hexahedrally mesh model
utilizing isomorphism technique. The surface mesh nodes of the head–neck model are restricted with six degrees of
freedom. Before analyzing the numerical results obtained with the finite element method, the reliability of the
numerical results themselves could be guaranteed. The head–neck finite element model was validated against
intracranial pressure data of Nahum et al.’s21 cadaver experiments, with some that was validated against extra
pressure history information for Trosseille et al.’s22 long duration impact. Further, the finite element head model
was validated against relative displacement information between brain and skull supplied by Hardy et al.23
Overall results that were acquired in the validation showed improved biofidelity that was in relation to former
finite element models.

Resonance behavior of the finite element model

Associated vibration shapes and 20 resonant frequencies are acquired in the range of 35 Hz to 330 Hz (Figure 2
and Table 1). Frequency spectrum is listed in Table 1, with their first resonant frequencies at 35.25 Hz, 34.50 Hz,
34.23 Hz, and 33.38 Hz. Additionally, there is a slight variation in the resonant frequencies between both damped
free vibration scenarios and the undamped (Table 1). Within the 20 mode shapes, five various vibration shapes of
the head–neck complex are illustrated: the nasal lateral cartilages’ lateral flexion as well as the lateral translation
of the mandible, which predominate after the head–neck structure’s chief modes within its first twenty modes, are
discovered. The mode shapes for the damped vibration cases are the same as that of the undamped, apart from the
difference in biomechanical parameters such as displacement, intracranial pressure as well as skull von Mises
stresses (Table 2). Table 2 also indicates the locations of peak skull stress and peak intracranial pressure corre-
sponding to each mode.
4 Journal of Low Frequency Noise, Vibration and Active Control 0(0)

Figure 2. The displacement contour plots of the head–neck model.

Table 1. Vibration responses of the finite element head–neck model.

With damping

Without damping n ¼ 0.1 n ¼ 0.2

Eigenfrequencies Eigenfrequencies Eigenfrequencies

Mode no. (Hz) (Hz) (Hz) Mode shape description

1 35.25 34.50 34.23 Anterior-posterior extension-flexion

of the head
2 62.40 57.72 57.59 Lateral flexion of the head
3 72.51 74.56 74.43 Axial rotational of the head
4 131.35 121.00 120.78 Vertical shutting of the jaws
5 221.59 221.06 220.55 S-shaped anterior-posterior retraction
of the head–neck
6 229.64 227.51 226.93 Lateral flexion of the nasal lateral cartilages
7 237.06 237.95 237.30 Lateral motion of mandible
8 254.04 253.80 253.08 Vertical translation of the head
9 256.47 257.89 257.14 Axial rotational of the head
10 270.16 270.13 269.31 Lateral flexion of the head; Lateral
motion of mandible
11 273.31 273.25 272.55 Flipping of nasal septal cartilage
12 277.10 277.42 276.53 Vertical opening of the jaws
13 283.20 283.44 282.45 Vertical shutting of the jaws
14 295.85 295.54 294.49 Lateral motion of mandible
15 299.86 298.60 297.52 Lateral motion of mandible
16 300.97 297.80 296.70 Lateral motion of mandible
17 316.08 315.84 314.62 Lateral motion of mandible
18 317.74 317.61 316.38 Vertical translation of the head
19 322.55 321.54 320.26 Lateral flexion of the head
20 328.78 329.16 327.80 Vertical translation of the head

Discussion
As illustrated in the midsagittal view of the head–neck model’s displacement contour plots (Figure 2), the first two
resonant modes happen at 62.40 Hz and 35.25 Hz with the minimal and maximal displacements at the top of the
head–neck junction and the head individually. These probably correspond to the entire head’s axial elongation.
The third mode, which happens at 72.51 Hz, has principal displacements at the rear and front of the head–neck
complex, being similar to the head’s longitudinal elongation. It can be found that the head–neck finite element
model is more compliant to both the axial rotation about its neck in the third mode and the front-and-back flexure
in the first two modes. It is revealed in these displacement plots that the whole head–neck resonate in
collaboration in the first three modes. It is then followed by various components’ resonance in the head; the
brain resonates at the higher modes while nasal lateral cartilages and mandible resonate at the lower modes. An
interesting point to be paid attention to is that displacement contour’s more loops come into sight in the brain for
higher frequency modes, likely showing the shearing, torsional or twisting modes within the brain tissues.
Yang et al 5

Table 2. Peak skull stress and peak intracranial pressure in both undamped and damped cases.

Max intracranial pressure (MPa) Max skull von Mises stress (MPa)

Without With damping Without With damping

Mode no. damping (n ¼ 0.2) damping (n ¼ 0.2)

1 3.16E-02 4.79E-03 4.05Eþ01 2.15Eþ01

2 3.77E-02 7.88E-03 9.54Eþ01 3.65Eþ01
3 2.46E-02 1.85E-02 2.67Eþ01 2.65Eþ01
4 2.27E-03 2.50E-03 3.20Eþ01 8.38Eþ00
5 6.83E-02 5.33E-02 1.62Eþ02 1.45Eþ02
6 6.27E-04 4.19E-04 3.33Eþ00 1.12Eþ00
7 2.46E-02 2.15E-02 1.30Eþ01 1.13Eþ01
8 1.27E-01 8.65E-02 1.55Eþ01 6.43Eþ00
9 7.87E-02 7.25E-02 2.42Eþ01 2.24Eþ01
10 4.68E-02 4.87E-02 7.91Eþ00 3.81Eþ00
11 5.24E-05 6.85E-05 2.74Eþ00 2.71Eþ00
12 5.24E-02 5.14E-02 3.07Eþ01 2.96Eþ01
13 4.87E-02 4.51E-02 7.17Eþ01 4.41Eþ01
14 4.48E-02 2.78E-02 1.22Eþ02 1.10Eþ02
15 6.26E-02 5.93E-02 1.00Eþ01 6.66Eþ00
16 1.06E-01 9.50E-02 5.70Eþ01 2.62Eþ01
17 5.20E-02 5.42E-02 3.41Eþ01 1.11Eþ01
18 6.76E-02 6.14E-02 3.72Eþ00 2.26Eþ00
19 4.71E-02 4.46E-02 1.64Eþ01 4.25Eþ00
20 7.82E-02 6.90E-02 2.69Eþ01 7.92Eþ00

It appears that rotational traumatic brain injury because of “shearing of brain tissues occurs mainly in the higher
frequency modes while translational brain injury occurs in the lower frequency modes”.24
Most of the published documents showed frequency spectrum of human head using traditional frequency
extraction approach ignoring the influence of damping. As previously noted, three damps factors are involved
in this complex eigenvalue problem. It is expected that resonant frequencies lower with the damping factor
increases. An important fluctuation in the resonant frequencies is mentioned between both the undamped and
damped free vibration, particularly in the lower modes where several components’ resonances coexist. As each
individual part resonates on its own in the temperate modes, this distinction between damped cases and
undamped falls down. Nevertheless, as the mode number rises further, this difference appears to be broadened
by a growth in damping factor in all likelihood owning to the complex combination and addition of shearing,
torsional and bending modes within the brain tissues in the higher modes. In Table 1, the third undamped natural
frequency (72.51 Hz) is lower than the damped natural frequencies (74.56 Hz and 75.43 Hz), that also happens at
different modes (7th, 9th, 12th, 13th, and 20th modes), this could be due to the modeling issues or because the system
has various components included, and the approach in modeling and material characteristics is more probably to
owing to the discrepancy.
With regard to the biomechanical parameters, such as intracranial pressure as well as skull von Mises stresses,
damping appears to have important influence on their significances at special modes (Table 2), the introduction of
damping in the complex eigenvalue analysis has reduced both maximal skull stresses and the maximal intracranial
pressure tremendously, of up to 84.8% and 74.1% respectively, in special modes (Table 2 and Figure 3). These
modes with the maximal deviance in vertex intracranial pressure and skull stresses between the damped models
and undamped are discovered and illustrated in Figure 3. It is showed that vertex intracranial pressure differs the
most in the first few modes when the head–neck flexes or rotates the most as well as in the subsequent modes (14th,
16th, 20th, 21st, and 23rd modes), therefore making the cranium to be moved the most with regard to the brain. The
skull–brain relative movement, as previously, brings about intracranial pressure gradients.25 It is likewise noticed
that the skull head’s vertex stress deviates the most in the modes in which the neck undergoes the highest stresses.
The “over-stiff” neck in our model on account of the lack of muscle tissues and ligaments may bring about
overestimated vertex stresses in the undamped mode. These parameters, nevertheless, are dropped in the damped
cases as a result of the coupling between the stiffness matrix and the damping matrix. The significance of
6 Journal of Low Frequency Noise, Vibration and Active Control 0(0)

Figure 3. Multiple biomechanical responses’ contour plots of the head–neck model.

identifying the pertinent damping factor in reckoning biomechanical responses in dynamic analyses is evaluated.
It ought to be mentioned that this damping influence gets saturated while damping factor is above 0.2, implying
that further damping would not have had an effect on the biomechanical responses. Ever since increasing
damping factor to the value above 0.2 is discovered to have got amplifying influence in lessening effect and
diminishing frequencies of higher modes in lowering peak biomechanical responses, consequently, it is concluded
that there is a reverse proportionality between damping influence on natural frequency and that on biomechanical
responses.
Meyer et al.’s16 finite element research is selected for comparison of the modal responses. Our computed
fundamental frequency is 35.25 Hz when Meyer et al.16 discovered that the principal frequency is 3.01 Hz.
This discrepancy in the cardinal frequency emerges between the two finite element head–neck models, probably
because of the distinction in material characteristics and the approach in modeling, and the approach in modeling
is more probably owing to the discrepancy. With the prioritized focal point on the neck harm, the head and all the
cervical vertebrae of Meyer et al.16 were regarded as stiff bodies, with all the individuals’ inertial moment and
masses being considered. Simply the intervertebral discs were regarded as deformable bodies. On the other hand,
all the components in our finite element head–neck model are regarded as deformable bodies with the capability to
disperse outside energy source by means of deformation. Additionally, unlike the rigid head of Meyer et al.,16 our
finite element head model consists of subarachnoid area (including cerebrospinal fluid and membranes), skull,
cartilages as well as brain tissues. By having the diverse different components in the head model, particularly with
sticky intracranial content’s lumped mass, it is expected that the key frequency of the multi-components system
will be lower than that of a one-component system as extra natural frequencies. This phenomenon is consonant
with comments by Chu et al.15 and Guarino and Elger.26

Conclusions
1. Vibration characteristic simulation, in terms of vibration shapes and frequency spectrum, of the finite element
model of human head–neck under vibration have been simulated.
2. Rarely reported modal responses such as “mastication” and flipping of nasal lateral cartilages modes are
introduced. These additional modes may be of interest to clinicians or surgeons who are specialized in rhino-
plasty or temporomandibular joint disorder.
3. The influence of damping on modal responses utilizing nonconventional complex eigenvalue analysis is also
investigated. Damping has a trade-off influence on lessening the resonant frequencies as well as on reducing the
vertex biomechanical responses.
Yang et al 7

Acknowledgment
BY, ZS and QW contributed equally to this study and share the first authorship.

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this
article.

Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this
article: This study was supported by the Natural Science Foundation of Jiangsu Province, China (Grant No. BK20161522), Six
Talent Peaks Project in Jiangsu Province, China (Grant No. JXQC-023), the State Laboratory of Advanced Design and
Manufacturing for Vehicle Body (Grant No. 31415008), and Chinese Postdoctoral Science Foundation (Grant No.
2015M572243).

References
1. Franke KE. Response of the human skull to mechanical vibrations. J Acoust Soc Am 1956; 28: 1277–1284.
2. Charalambopoulos A, Dassios G, Fotiadis DI, et al. Frequency spectrum of the human head-neck system. Int J Eng Sci
1997; 35: 753–768.
3. Bekesy GV. Vibration of the head in a sound field and its role in hearing by bone conduction. J Acoust Soc Am 1948; 20:
749–760.
4. Hodgson V, Gurdjian E and Thomas L. The determination of response characteristics of the head with emphasis on
mechanical impedance techniques. In: 11th Stapp car crash conference, SAE Paper 670911, 1967.
5. Sun DG, Yan BJ, Han BH, et al. Vibration characteristic simulation of a pneumatic artificial muscle damping seat. J Low
Freq Noise Vib Active Control 2016; 35: 39–51.
6. Stalnaker RL and Fogle JL. Driving point impedance characteristics of the head. J Biomech 1971; 4: 127–139.
7. Khalil TB, Viano DC and Smith DL. Experimental analysis of the vibrational characteristics of the human skull. J Sound
Vib 1979; 63: 351–376.
8. Hakansson B, Brandt A, Carlsson P, et al. Resonance frequencies of the human skull in vivo. J Acoust Soc Am. 1994; 95:
1474–1481.
9. Mandapuram S, Rakheja S, Boileau PE, et al. Energy absorption of seated body exposed to single and three-axis whole
body vibration. J Low Freq Noise Vib Active Control 2015; 34: 21–38.
10. Lei CY, Xu J, Wang J, et al. Active headrest with robust performance against head movement. J Low Freq Noise Vib
Active Control 2015; 34: 233–250.
11. Gudarzi M and Oveisi A. Robust control for ride comfort improvement of an active suspension system considering
uncertain driver’s biodynamics. J Low Freq Noise Vib Active Control 2014; 33: 317–339.
12. Nickell RE and Marcal PV. In-vacua modal dynamic response of the human skull. J Eng for Industry 1974; 3: 490–494.
13. Ward CC and Thompson RB. The development of a detailed finite element brain model. In: Proceedings of 19th Stapp car
crash conference, New York, USA, SAE Paper 751163, 1975, pp. 641–674.
14. Ruan JS, Khalil T and King AI. Human head dynamic response to side impact by finite element modeling. J Biomech Eng
2014; 113: 276–283.
15. Chu CS, Lin MS, Huang HM, et al. Finite element analysis of cerebral contusion. J Biomech 1994; 27: 187–194.
16. Meyer F, Bourdet N, Willinger R, et al. Finite element modelling of the human head-neck: Modal analysis and validation
in the frequency domain. Int J Crashworthiness 2004; 9: 535–545.
17. Baroudi AE, Razafimahery F and Rakotomanana-Ravelonarivo L. Three-dimensional modal analysis of an idealized
human head including fluid–structure interaction effects. Acta Mech 2012; 223: 1899–1915.
18. Willinger R, Bourdet N, Fischer R, et al. Modal analysis of the human neck in vivo as a criterion for crash test dummy
evaluation. J Sound Vib 2005; 287: 405–431.
19. Rützel S, Hinz B and W€ olfel HP. Modal description-A better way of characterizing human vibration behavior. J Sound
Vib 2006; 298: 810–823.
20. Whitman TA, Wodicka GR, Morgan MT, et al. Measurement and modeling of the vibrational response of the ovine head
as it relates to intracranial pressure. In: Proceedings of engineering in medicine and biology society, Amsterdam,
Netherlands, 1996, pp. 493–494.
21. Nahum A, Smith R and Ward C. Intracranial pressure dynamics during head impact. In: Proceedings of the 21st Stapp car
crash conference, Warrendale, PA, USA, 1977, pp. 339–366.
22. Trosseille X, Tarriere C, Lavaste F, et al. Development of a FEM of the human head according to a specific test protocol.
In: Proceeding of 36th Stapp car crash conference, Seattle, SAE International Society of Automotive Engineers, 1992, pp.
235–253.
8 Journal of Low Frequency Noise, Vibration and Active Control 0(0)

23. Hardy WN, Mason MJ, Foster CD, et al. A study of the response of the human cadaver head to impact. Stapp Car Crash J
2007; 51: 17–80.
24. Takahashi Y. Vibratory sensation induced by low-frequency noise: The threshold for “vibration perceived in the head” in
normal-hearing subjects. J Low Freq Noise Vib Active Control 2013; 32: 1–9.
25. Horgan TJ and Gilchrist MD. Influence of FE model variability in predicting brain motion and intracranial pressure
changes in head impact simulations. Int J Crashworthiness 2004; 9: 401–408.
26. Guarino JC and Elger DF. Modal analysis of a fluid-filled elastic shell containing an elastic sphere. J Sound Vib 1992; 156:
461–479.