applied

sciences
Article
Study of the Dynamic Behaviour of Circular
Membranes with Low Tension
Antonia Lima-Rodriguez * , Antonio Gonzalez-Herrera and Jose Garcia-Manrique
Departamento de Ingeniería Civil, de Materiales y Fabricación. Universidad de Málaga, Escuela de Ingenierías
Industriales. C/ Ortiz Ramos, s/n. Campus de Teatinos, 29071 Málaga, Spain; agh@uma.es (A.G.-H.);
josegmo@uma.es (J.G.-M.)
* Correspondence: tlima@uma.es; Tel.: +34-951952441




Received: 6 September 2019; Accepted: 1 November 2019; Published: 5 November 2019


Abstract: The dynamic behaviour of membranes has been widely studied by well-known authors for
a long time. A clear distinction can be made between the behaviour of membranes without tension
(plate case) and membranes subjected to large tension or pre-strain in their plane (membrane case).
In classical theories, less attention has been paid to membranes subjected to a low level of tension,
which solution is between both extreme cases. Recently, certain fields of research are demanding
solutions for this intermediate behaviour. It is the case of membranes present in MEMS and sensor or
the response of the tympanic membrane in mammals hearing system. In this paper, the behaviour of
plates and circular membranes with boundary conditions clamped in the edges has been studied.
The natural frequencies for both cases (plate and membrane) have been calculated using the solutions
of the traditional theories and these have been compared with the numerical frequencies calculated
by finite element analysis. The dynamic response of membrane with low tension, corresponding to a
transition between these extreme behaviours, has also been calculated. A theoretical solution has
been used complemented with a wide set of numerical finite elements calculations. The analytical and
numerical solutions are very close, being the error made using both methods very low; nevertheless,
there are no analytical solutions for the entire transition zone between the plate and membrane
behaviour. Therefore, this range has been completed using finite element analysis. Broad ranges of
geometric configurations have been studied. The transition behaviour of the membrane has been
clearly identified. The main practical consequences of these results have been discussed, in particular
focused on the response of the tympanic membrane.

Keywords: plate; membrane; pre-strain; tension; finite element analysis

1. Introduction
The dynamic response of a membrane has been studied since a long time ago. There exist two
classic solutions for two extreme situations, one for the case where the membrane is not subjected to
internal loads (considered in this case as plate) and the other one when it is under tension (considered as
membrane) [1,2]. Both solutions have proven useful to study a plate and to study a membrane subjected
to loads.
Many authors have studied and published on this topic. These studies have included different
cases of plates and membranes. Rayleigh [1] and Timoshenko [2] have studied vibration of plates
and membranes with different shapes; Timoshenko [2] included in their study the energy method for
solving the different cases. Leissa [3] compiled the studies and publications on vibration of plate in
1969. The first studies on vibrations of circular plates are attributed to Poisson [4] and Kirchhoff [5].
Subsequently, other authors have contributed to investigate and publish on this topic. Airey [6],
Itao and Crandall [7] and Amabili et al. [8] published studies about free-edge plates; Airey [6] also

Appl. Sci. 2019, 9, 4716; doi:10.3390/app9214716 www.mdpi.com/journal/applsci

Appl. Sci. 2019, 9, 4716 2 of 17

studied the case of clamped membranes and Amabili et al. [8] studied the circular plates vibrating in
vacuum and in contact with liquid. Other authors have studied the natural frequencies of circular
plates simply supported [9], with elastic edge supports [10] or with non-uniform edge constrains [11].
In the latter case are included the publications of Narita and Leissa [12,13], Amabili et al. [14] or Eastep
and Hemming [15]. Others papers research the vibration of circular plates in contact with water [16] or
with fluid [17]. It can be concluded that both extreme situations of plate and membrane behaviour are
widely covered in the literature.
Lesser attention has been paid to the case where the membrane is subjected to a slight level of
tension. In these cases, intermediate behaviour is expected between membrane and plate. One of the
reasons of the absence of studies in this line is that most of the real situations presented in engineering
regarding membrane behaviour responded to one of those extremes. Nevertheless, there are situations
under study today that may respond to this intermediate situation.
Vibration analysis is important in the study of musical instruments [18–21] and specifically the
case of membrane subjected to tension is important in the study of percussion musical instruments
since the component which produce the desired sound it is a material of small thickness subjected
to stress [22]. Recently, the use of membrane in microelectronics has made this issue a key point.
Small membranes are used in different applications (e.g. as sensor or part of MEMS) which during
manufacturing acquire certain level of tension that deviates their expected behaviour form [23–27].
Another field of study where this effect is manifest is the study of the mechanics of hearing.
Based on different experimental studies [28] and simulation of the human hearing system [29] it has
been observed that small levels of pre-strain in the system, particularly in the tympanic membrane,
causes significant changes in the response. This issue is of great interest for our research group.
As the hypothesis origin of this study, we consider that the deviations between expected and
tested behaviours may be related the assumption of ideal membrane or plate behaviours in these
problems. We have not found a consistent literature that helps us evaluate the influence of the tension
level of a membrane with its actual behaviour. There are analytical solutions [30] that developed this
question for some regular membrane shapes and some specific situations. The main aim of this work
is to approach the study of the influence of pre-strain membranes based on parameters such as its
geometric dimensions, type of material and internal stress level.
In the present paper, we have conducted a numerical study to evaluate the influence of the
level of pre-strain in the circular membrane dynamic response. The election of a circular shape
allows us to compare numerical solutions with analytic ones of existing literature. Once we develop
our own methodology to stablish the pre-strain limits to accept a membrane, plate or intermediate
behaviour, we will be able to translate this work to complex geometries and biomedical or industrial
specific applications.
The results obtained has been based in finite element (FE) simulations of modal analyses. Initially
a reference problem has been studied (a 1 m diameter steel membrane with different thicknesses)
where the limits of both theories are identified for different geometric ratios (thickness-diameter ratio).
Finally, the study has been repeated with a more realistic case, closer to our research field, with a
smaller and thin membrane with the properties similar to the human tympanic membrane.
The paper is structured according to the following sections. In the next one, the main theories will
be briefly outlined. The FE numerical methodology is described in the following section. Result obtained
with the reference membrane are then presented and the influence of thickness evaluated. Finally, this
study is complemented with the analysis of a membrane similar to the tympanic membrane. In the last
section, the main conclusions are outlined.

2. Reference Theory
There are two classic theories that describe the behaviour of thin thickness parts, behaviour as a
plate or as a membrane. The natural frequencies in a dynamic study can be obtained with the analytic
equations worked out by the different authors that studied these theories [1,2,31].
 Appl. Sci. 2019, 9, 4716 3 of 17
Appl. Sci. 2019, 9, x FOR PEER REVIEW 3 of 17

Among
Among other
other considerations,
considerations, thethe solution
solution depends
depends onon
thethe shape
shape and
and thethe boundary
boundary conditions
conditions of of
the plate or membrane. In this paper, the theoretical study is focused on circular shapes and
the plate or membrane. In this paper, the theoretical study is focused on circular shapes and boundaryboundary
conditions
conditions clamped
clamped inin
thethe edges
edges (Figure
(Figure 1).1).

Figure
Figure 1. Clamped
1. Clamped circular
circular plate
plate or or membrane.
membrane.

Plate solutions are based on the stiffness provided by bending behaviour. For the circular clamped
Plate solutions are based on the stiffness provided by bending behaviour. For the circular
plate case, the natural frequencies can be determined by the Equation (1) [2]:
clamped plate case, the natural frequencies can be determined by the Equation (1) [2]:
s
𝑝 f =𝛼 =𝑔𝐷 p α gD
𝑓= = 2 γh (1) (1)
2𝜋 2𝜋𝑎2π 𝛾ℎ2πa
in which f is the natural frequency to be calculated, a is de radio of the boundary, γh/g=ρhis the mass
in per
whichunitf area,
is theDnatural frequency
= Eh3 /12(1-υ to be
2 ) is the calculated,
flexural a isofde
rigidity theradio
plateofand
theαis
boundary,
a constantγh/g=ρh is the mass
that depends on the
per unit area, D = Eh 3 /12(1-υ 2 ) is the flexural rigidity of the plate and α is
number of nodal diameters n and the number of nodal circles s. The values of αconstant are given ina constant that depends on[3].
the number of nodal diameters n and the number of nodal circles s. The values
The α constants are the squares of the eigenvalues of the equation that must be satisfied for theof α constant are given
in particular
[3]. problem of a clamped circular plate. This solution depends on Bessel functions and modified
The functions
Bessel α constants of are
the the
firstsquares
kind. Figureof the 2eigenvalues
presents the offirst
the equation that must
nine vibration modes be of
satisfied for the of
this problem
particular problem of a clamped circular
plate and their correspondence with the values of n and s. plate. This solution depends on Bessel functions and
modified OnBessel functions
the other hand, whenof thethe firstmembrane
kind. Figure 2 presents
is too the firstbending
thin to provide nine vibration
stiffness,modes of this is
the stiffness
problem of plate and their correspondence with the values of n and s.
obtained applying an internal tension. That is the base for the membrane solution. In particular, the
solution for the circular clamped membrane case, the equation to obtain the vibration modes is the
Equation (2) [2]:
r
p α gS
f = = (2)
2π 2πa w
in which g is the acceleration of gravity, S is the uniform tension per unit length of the boundary and w
= ρgh is the weight of the membrane per unit area. α constants are the eigenvalues of the equation that
must be satisfied for the particular
Mode 1 problem ofMode a membrane
2 and which
Mode 3depends on Bessel functions of
the first kind. The values of α constant
(n = 0, s = 0) for this case are
(n = 1, s = 0) given in
(n [1].
= 2, s = 0)
As most practical problem laid on one of the previous formulations, and due to the complexity of
the analytical solution, lesser attention has been paid to the problem where the membrane stiffness is
provided by a combination of an internal tension and bending stiffness.
The problem of circular plates with a large initial tension or compression and simply supported
or clamped edges case was studied by Wah [30]. This author, based on the basic equation used in
Poisson-Kirchhoff theory, developed an expression to get the natural frequency for this case:

Mode 4 Mode 5 2 ! 12 Mode

!1 6
p α Ta 2 D 2
(n = 0, s = 1)f = =(n = 3, s = 0) + α (n = 1, s = 1) (3)
2π 2πa2 D ρh

In this expression, T is the uniform tension per unit length of the boundary, denoted by S in
Equation (2), and α is a constant that depends on n and s as in previous cases, and on the non-dimensional
parameter φ. The values of this parameter are given in [30].
Appl. Sci. 2019, 9, 4716 4 of 17

In this case, the α constants are the eigenvalues of the equation that must be satisfied for
Appl. Sci. 2019, 9, x FOR PEER REVIEW 4 of 18
the particular problem of the circular clamped plates with a large initial tension or compression.
That equation also depends on Bessel functions and modified Bessel functions of the first kind.
The parameter φ is calculated using the expression T/T* = φ, where T* = 14.68D/a2 for the clamped case.

Mode 1 Mode 2 Mode 3

(n = 0, s = 0) (n = 1, s = 0) (n = 2, s = 0)

Mode 4 Mode 5 Mode 6

(n = 0, s = 1) (n = 3, s = 0) (n = 1, s = 1)

Mode 7 Mode 8 Mode 9

(n = 4, s = 0) (n = 2, s = 1) (n = 0, s = 2)

Modal shapes
Figure 2. Modal shapes of
of the
the first
first nine vibration modes.

3. Numerical Methodology: Finite Element Analysis.

On the other hand, when the membrane is too thin to provide bending stiffness, the stiffness is
obtained applying
The main issuesan regarding
internal tension. That
the Finite is the base
Element for themethodology
modelling membrane solution.
will be In particular,
briefly the
described
solution for the circular
in this section. The present clamped
workmembrane
is based on case, the equation
numerical to obtain
calculations the vibration
made by meansmodes of the is the
finite
Equation (2) [2]: (FEM). For the numerical analysis, ANSYS software v19 research license has been
element method
used. The methodology is applied to the previous described problem of circular membranes. The
availability of some analytical formulations 𝑝 for the𝛼 problem
𝑔𝑆 under study provide us the necessary
𝑓 (2)
validation of our model. 2𝜋 2𝜋𝑎 𝑤
Once the model is validated, the parametric formulation of our programming allows to analyse
in which g is the acceleration
features between ideal plate of andgravity,
membraneS is the uniform tension
behaviours, filling per
the unit length of them.
gap between the boundary and
In addition,
w = ρgh is the weight of the membrane per unit area. α constants are the eigenvalues
it must be taken into account that the goal of this work is to extend the conclusions to other generic of the equation
that must be
geometries satisfied
more complex. for As
thedescribed
particular in problem
introductionof asection,
membrane one ofand
the which
researchdepends
lines thatonsupport
Bessel
functions
this work isofthe
thehearing
first kind. The values
mechanics of α constant
simulation. for this case
In particular, are givenmembrane
the tympanic in [1]. does not respond
As most practical problem laid on one of the previous formulations,
to a perfect circular membrane. It has a slight conic shape and a lack of symmetry and duedue to the complexity
to the presence
of
of the analytical
other elementssolution, lesser attention
of the hearing system. It has
canbeen
alsopaid to the
present anproblem
unknown where the membrane
internal tension. stiffness
is provided
In this by a combination
work, of an internal
circular shapes have been tension and bending
modelled into thestiffness.
program to simulate the circular
The problem of circular plates with a large initial
membranes with boundary conditions clamped in the edges. Several tension or compression
radiis andand simplydimensions
thickness supported
or
and clamped edges properties
two material case was studied
have beenby Wah [30]. This
simulated. Theauthor,
materialsbased on the basic
simulated equation
are steel used in
and eardrum
Poisson-Kirchhoff theory, developed an expression to get the natural frequency for
equivalent material. For the case of the steel, it has been used 1 m diameter with different thicknesses: this case:
10, 1, 0.1, 0.01 and 0.001 mm. The properties used for the steel are: density 7850 kg/m3 , Poisson´s
𝑝 𝛼 𝑇𝑎 𝐷 (3)
𝑓 𝛼
2𝜋 2𝜋𝑎 𝐷 𝜌ℎ
Appl. Sci. 2019, 9, 4716 5 of 17
Appl. Sci. 2019, 9, x FOR PEER REVIEW 5 of 17

10, 1, 0.1,
ratio 0.300.01
andand 0.001modulus
Young´s mm. The2.1 properties
GPa. In usedthe case for of
thethesteel are: density
eardrum 7850 kg/m
equivalent 3, Poisson´s ratio
material, the diameter
0.30
is 1 and Young´s modulus
cm, thickness 0.05 mm,2.1 GPa. In
density the kg/m
1200 case of 3 , the eardrum
Poisson´s equivalent
ratio 0.30 and material,
Young´s the diameter
modulus is 1
32 MPa.
cm, thickness 0.05 mm, density 1200 kg/m
All simulations are under small deformations hypothesis and no plastic behaviour is necessary toAll
3 , Poisson´s ratio 0.30 and Young´s modulus 32 MPa. be
simulations
introduced in arethe
under
model.small deformations hypothesis and no plastic behaviour is necessary to be
introduced in the model.
The simulations of the models in ANSYS were made in two different situations: without load,
and with radial load in the
The simulations of the models
plane ofinthe
ANSYS
membrane were made in two
(i.e. with different[28].
pre-strain) situations: without of
The pre-strain load,
the
and with radial
membrane was load in the plane
introduced of the
by means of membrane
an imposed(i.e. withdisplacement,
radial pre-strain) [28]. Theproduces
which pre-straina of the
radial
membrane
deformation was
andintroduced
the desiredbyaxial
means of in
stress anthe imposed
membrane. radial displacement, which produces a radial
deformation and the desired axial stress in the membrane.
Figure 3 shows a scheme of the finite element simulation. It presents the circle area geometry
Figure
(Figure 3b) 3which
showsis ameshed
scheme(Figure
of the finite
3d) withelement
a shell simulation. It presents
element (Figure the circle
3c). The elementarea geometry
selected is a
(Figure 3b) which is meshed (Figure 3d) with a shell element (Figure 3c). The
four nodes one with six degrees of freedom at each node (translations in the x, y, and z directions, element selected is a
four nodes one
and rotations withthe
about sixx,degrees of freedom
y, and z-axes) suitable at each node (translations
for analysing in the x, thick
thin to moderately y, and z directions,
shell structures
and rotations
(reference about thein
SHELL181 x, ANSYS
y, and z-axes) suitable
software). for analysing
Figure 3c showsthin theto moderately
geometry, thick
node shell structures
locations, and the
(reference SHELL181 in ANSYS software).
element coordinate system for this element. Figure 3c shows the geometry, node locations, and the
element coordinate system for this element.

Figure 3. Finite
Figure 3. Finiteelement
elementmodal
modal simulation
simulation of circular
of circular membrane.
membrane. (a) Scheme
(a) Scheme of pre-strain;
of pre-strain; (b)
(b) Geometry;
Geometry; (c) Element type; (d) Meshing; (e) Some displacement solutions: modes
(c) Element type; (d) Meshing; (e) Some displacement solutions: modes 2, 3 and 4. 2, 3 and 4.

The mesh density

The mesh densityis is stablished
stablished to minimize
to minimize numerical
numerical error.
error. The The minimum
minimum element
element size size
parameter
parameter
(Pmes ) used(Pismes) used is
referred toreferred
the radiusto of
thethe
radius of the membrane.
membrane. In this way, In
wethis
canway,
ensurewea can ensure
constant a constant
mesh density
mesh
for alldensity for all regardless
simulations, simulations, of regardless of the
the size of the sizeA
plate. ofPthe
mes plate.
of 160Aradius
Pmes ofdivisions
160 radius divisions
was selected.was
selected.
Once the model is defined, natural frequencies and mode shapes are obtained through a modal
Onceroutine.
analysis the modelFigureis defined, natural frequencies
3e presents and mode
some displacement shapesobtained
results are obtained through
in modes 2, a3 modal
and 4.
analysis routine.
The coloured Figure
areas 3e presents
represented some
the total displacement
translation results obtained
displacement in modes
of each node of the2,model.
3 and 4. The
coloured areas
In cases represented
with pre-strain,the total translation
a previous displacement
static analysis of each node
was performed. of the model.
To simulate pre-strain levels, 56
In cases
different with
values ofpre-strain,
radial strain a previous static on
were imposed analysis was performed.
the perimeter To simulate
of the circular pre-strain
membrane. Theselevels,
strain
56 different
values are invalues
the range of ε =strain
of radial −8
10 to were −2
2 × imposed
10 . on the perimeter of the circular membrane. These
strain values are in the range of ε = 10−8 to 2 × 10−2.
4. Reference Case Study
4. Reference Case Study
In this section, attention will be paid to the results corresponding to the steel membrane 1 m
diameter and
In this 1 mm attention
section, thickness.will
Thebenatural
paid frequencies forcorresponding
to the results the cases of plate andsteel
to the membrane subjected
membrane 1m
diameter and 1 mm thickness. The natural frequencies for the cases of plate and membrane subjected
to the deformations will be calculated using the classic theory besides the results of the simulation in
Appl. Sci. 2019,
Appl. Sci. 9, x
2019, 9, 4716
FOR PEER REVIEW 66 of
of 17
17

the ANSYS program. Relations between these different parameters will be calculated and plotted to
to the deformations will be calculated using the classic theory besides the results of the simulation in
establish the influence of the pre-strain on the behaviour of this reference case.
the ANSYS program. Relations between these different parameters will be calculated and plotted to
The analytical natural frequencies of the membrane behaviour (fANM) have been calculated with
establish the influence of the pre-strain on the behaviour of this reference case.
the Equation (2) besides the plate behaviour (fANP) which was calculated using the Equation (1). On
The analytical natural frequencies of the membrane behaviour (fANM ) have been calculated with
the other hand, the finite elements model (FEM) frequencies have also been obtained (fFEM).
the Equation (2) besides the plate behaviour (fANP ) which was calculated using the Equation (1). On the
The first nine analytical natural frequencies for membrane behaviour (fANM), the simulation
other hand, the finite elements model (FEM) frequencies have also been obtained (fFEM ).
frequencies (fFEM) and the ratio between the analytical and simulation frequencies (fANM/fFEM) are
The first nine analytical natural frequencies for membrane behaviour (fANM ), the simulation
shown in Table 1 for the selected strains: 10−7, 10−5, and 10−3. These values can be considered as low,
frequencies (fFEM ) and the ratio between the analytical and simulation frequencies (fANM /fFEM ) are
medium and high pre-strain. It can be observed how the analytical and numerical results are closer
shown in Table 1 for the selected strains: 10−7 , 10−5 , and 10−3 . These values can be considered as low,
when strain increases.
medium and high pre-strain. It can be observed how the analytical and numerical results are closer
when strain increases.Table 1. Ratio fANM/fFEM for first nine modes and strain selected.

Table 1. Ratio fANM /fFEM for first nine modes and strain selected.
STRAIN
10−7 10−5
STRAIN 10−3

−7 fANM/ −5 fANM/ −3 fANM/

fANM 10
fFEM fANM f10
FEM fANM 10fFEM
fFEM fFEM fFEM
fANM fANM fANM
Mode 1 fANM
1.496 fFEM
10.310 0.145 fANM
14.959 fFEM
19.038 fANM f151.958
/f
FEM /f0.786
FEM
149.593 FEM 0.984
/fFEM
Mode 2 2.385 21.338 0.112 23.845 32.984 0.723 238.452 242.399 0.984
Mode 1 1.496 10.310 0.145 14.959 19.038 0.786 149.593 151.958 0.984
Mode
Mode 3 2 2.385
3.195 21.338
34.918 0.112
0.092 23.845
31.953 32.984
48.318 0.723
0.661 238.452
319.533 242.399
325.372 0.984
0.982
Mode 4 3
Mode 3.195
3.435 34.918
39.797 0.092
0.086 31.953
34.349 48.318
53.474 0.661
0.642 319.533
343.491 325.372
349.922 0.982
0.982
Mode 4 3.435 39.797 0.086 34.349 53.474 0.642 343.491 349.922 0.982
Mode 5 3.969 51.022 0.078 39.694 65.587 0.605 396.943 404.962 0.980
Mode 5 3.969 51.022 0.078 39.694 65.587 0.605 396.943 404.962 0.980
Mode
Mode 6 6 4.366
4.366 60.791
60.791 0.072
0.072 43.658
43.658 75.669
75.669 0.577
0.577 436.582
436.582 445.791
445.791 0.979
0.979
Mode 7 7
Mode 4.721
4.721 69.600
69.600 0.068
0.068 47.205
47.205 84.984
84.984 0.555
0.555 472.051
472.051 482.680
482.680 0.978
0.978
Mode 8 5.238 84.469 0.062 52.376 100.146 0.523 523.761 536.317 0.977
Mode
Mode8 9 5.238
5.385 84.469
88.983 0.062
0.061 52.376
53.851 100.146
104.728 0.523
0.514 523.761
538.509 536.317
551.689 0.977
0.976
Mode 9 5.385 88.983 0.061 53.851 104.728 0.514 538.509 551.689 0.976

The ratio fANM /fFEM for the first nine modes of the reference case (h/d = 10 −3 ) for different levels
The ratio fANM/fFEM for the first nine modes of the reference case (h/d = 10−3) for different levels of
of pre-strain have been plotted in Figure 4.
pre-strain have been plotted in Figure 4.

Figure4.4.Ratio
Figure RatiofANM /fFEM
fANM/f asaafunction
FEMas functionofofstrain
straincalculated
calculatedfor
forthe
thefirst
firstnine
ninemodes.
modes.

In
In Figure
Figure 44 it
it is
is observed
observed that
that for
for small
small values
values of
of strain,
strain, the
the calculated
calculated and
and simulated
simulated frequencies
frequencies
are
are far away, being the ratio far from 1, so the behaviour as membrane and therefore Equation (2)
far away, being the ratio far from 1, so the behaviour as membrane and therefore Equation (2) are
are
not
not met
met in
in that
that range.
range.
The ratio increases with the imposed strain until reaching the value of 1 for strains close to 10−2.
For these strains, the Equation (2) and membrane behaviour are met. This result is fulfilled for the
first nine vibration modes represented in Figure 4.
Appl. Sci. 2019, 9, 4716 7 of 17

Appl. Sci.
The2019, 9, xincreases
ratio FOR PEERwith
REVIEW
the of −2
imposed strain until reaching the value of 1 for strains close to7 10 17.
For these strains, the Equation (2) and membrane behaviour are met. This result is fulfilled for the first
Table 2 shows
nine vibration modes therepresented
first nine analytical
in Figurefrequencies
4. calculated with Equation (1) for the behaviour
as plate (f ANP), the numerical frequencies obtained with FEM (fFEM) and the ratio between both
Table 2 shows the first nine analytical frequencies calculated with Equation (1) for the behaviour
(fANP
as /fFEM(f
plate ). In this case, we have included the results obtained without strain and for the three values
ANP ), the numerical frequencies obtained with FEM (fFEM ) and the ratio between both
of strain selected: 10−7, 10−5 and 10−3.
(fANP /fFEM ). In this case, we have included the results obtained without strain and for the three values
of strain selected: 10−7 , 10−5 and 10−3 .
Table 2. Ratio fANM/fFEM for first nine modes and strain selected.
Table 2. Ratio fANM /fFEM for first nine modes
STRAINand strain selected.

0 10−7 STRAIN 10−5 10−3

fANP/ −7 fANP/ −5 fANP/ −3fANP/
fANP fFEM 0 fFEM 10 fFEM 10 fFEM 10
fFEM fFEM fFEM fFEM
fANP fANP fANP fANP
Mode 1 fANP
10.179 f10.179 f10.310 f f
FEM
/f1.000
FEM
FEM
/f0.987
FEM
19.038
FEM 0.535
/f FEM
151.958
FEM 0.067
/fFEM
Mode
Mode 1 2 21.184
10.179 21.185
10.179 1.000
1.000 21.338
10.310 0.993
0.987 32.984
19.038 0.642
0.535 242.399
151.958 0.087
0.067
Mode
Mode2 3 21.184
34.755 21.185
34.754 1.000
1.000 21.338
34.918 0.993
0.995 32.984
48.318 0.642
0.719 242.399
325.372 0.087
0.107
Mode
Mode3 4 34.755
39.629 34.754
39.634 1.000
1.000 34.918
39.797 0.995
0.996 48.318
53.474 0.719
0.741 325.372
349.922 0.107
0.113
Mode 4 39.629 39.634 1.000 39.797 0.996 53.474 0.741 349.922 0.113
Mode
Mode 5 5 50.857
50.857 50.853
50.853 1.000
1.000 51.022
51.022 0.997
0.997 65.587
65.587 0.776
0.776 404.962
404.962 0.126
0.126
Mode
Mode6 6 60.602
60.602 60.622
60.622 1.000
1.000 60.791
60.791 0.997
0.997 75.669 0.801
0.801 445.791
445.791 0.136
0.136
Mode 7 69.417 69.426 1.000 69.600 0.997 84.984 0.817 482.680 0.144
Mode 7 69.417 69.426 1.000 69.600 0.997 84.984 0.817 482.680 0.144
Mode 8 84.277 84.296 1.000 84.469 0.998 100.146 0.842 536.317 0.157
Mode
Mode 9 8 84.277
88.785 84.296
88.810 1.000
1.000 84.469
88.983 0.998
0.998 100.146
104.728 0.842
0.848 536.317
551.689 0.157
0.161
Mode 9 88.785 88.810 1.000 88.983 0.998 104.728 0.848 551.689 0.161

The ratio fANP /fFEM as a function of the strain for the first nine modes of the reference case has
The ratio fANP/fFEM as a function of the strain for the first nine modes of the reference case has
been plotted in Figure 5.
been plotted in Figure 5.

Figure5.5.Ratio
Figure fANP/f
RatiofANP /fFEM asaafunction
FEM as functionof
ofthe
thestrain
strainfor
forthe
thefirst
firstnine
ninemodes.
modes.

Observing
Observingthe thegraphs
graphspresented
presented in Figure 5, in5,the
in Figure inrange of lowoflevels
the range low of strainofthe
levels ratiothe
strain is virtually
ratio is
unity, which
virtually means
unity, which thatmeans
for that strain
that range
for that the behaviour
strain is plate-like.
range the behaviour Its value decreases
is plate-like. when the
Its value decreases
strain
when increases, being verybeing
the strain increases, low with
very large strains.
low with large The behaviour
strains. changes tochanges
The behaviour increasetothe imposed
increase the
strain, moving away from the plate behaviour.
imposed strain, moving away from the plate behaviour.
Figures
Figures 44 and
and 55 are
are useful
useful to
to identify
identify the
the strain
strain values
values where
where the
the solution
solution separates
separates from
from the
the
classic solutions. It can be observed in Figure 5 how for values of strain below 10 -6 the plate behaviour
classic solutions. It can be observed in Figure 5 how for values of strain below 10 the plate behaviour-6

from
from Equation
Equation (1)(1) is
is valid.
valid.InInthis
thiszone,
zone,wewecancan say
say that
that pure
pure bending
bending behaviour
behaviour is dominant,
is dominant, andand
the
the strain
strain is not
is not significant.
significant. On theOnother
the other
side, side,
from from
Figure Figure
4 it is4observed
it is observed that
that for for values
values of strain
of strain above
10−3, the results obtained are close to those predicted with the membrane theory Equation (2). We can
say that is the zone with pure membrane behaviour where the stiffness of the plate is negligible.
Among these extremes, a transition zone is observed where neither of those solutions are valid.
Appl. Sci. 2019, 9, 4716 8 of 17

above −3
Appl. Sci.10
2019,, 9,
the results
x FOR PEERobtained
REVIEWare close to those predicted with the membrane theory Equation 8 of (2).
17
We can say that is the zone with pure membrane behaviour where the stiffness of the plate is negligible.
Among these interesting
Another extremes, aparameter
transition that
zonecanis observed
be evaluated where is neither
the ratioofbetween
those solutions are valid.
the natural frequency
of theAnother interesting
modes and parameter
the frequency thatfirst
of the canmode
be evaluated is the
(fn/f1). This hasratio
beenbetween
a usefulthe natural frequency
parameter to evaluate of
the modes and the frequency of the first mode
the presence of pre-strain with experimental datan [28]. (f /f1 ). This has been a useful parameter to evaluate
If the natural modes of vibration are calculated the
presence of by
analytically pre-strain
Equation with
(1) experimental
or Equation (2), dataand [28].
these If the
are natural
referredmodes
to the of vibration
first are calculated
mode dividing by its
analytically
value, by Equation
the results (1) or that
are constants Equation
can be (2), and thesewith
calculated are referred to the
the relations offirst
the mode dividing
α constants for by
plateits
value,
and the results
membrane are constants
cases that(Table
respectively can be3). calculated with the relations of the α constants for plate and
membrane cases respectively (Table 3).
Table 3. Ratio fn/f1 for the first nine natural frequencies for plate and membrane case.
Table 3. Ratio fn /f1 for the first nine natural frequencies for plate and membrane case.
PLATE CASE MEMBRANE CASE
PLATE CASE MEMBRANE CASE
f2ANP/f1ANP 2.081 f2ANM/f1ANM 1.594
f2ANP /f1ANP 2.081 f2ANM /f1ANM 1.594
f3ANP/f1ANP
f3ANP /f1ANP 3.414
3.414 f3ANM/f1ANM
f3ANM /f1ANM2.136 2.136
f4ANP/f1ANP
f4ANP /f1ANP 3.893
3.893 f4ANM/f1ANM
f4ANM /f1ANM2.296 2.296
f5ANP /f1ANP
f5ANP/f1ANP 4.996
4.996 f5ANM /f1ANM2.653
f5ANM/f1ANM 2.653
f6ANP /f1ANP
f6ANP/f1ANP 5.954
5.954 f6ANM /f1ANM2.918
f6ANM/f1ANM 2.918
f7ANP /f1ANP
f7ANP/f1ANP 6.819
6.819 f7ANM /f1ANM3.156
f7ANM/f1ANM 3.156
f8ANP /f1ANP 8.279 f8ANM /f1ANM 3.501
f8ANP/f1ANP 8.279 f8ANM/f1ANM 3.501
f9ANP /f1ANP 8.722 f9ANM /f1ANM 3.600
f9ANP/f1ANP 8.722 f9ANM/f1ANM 3.600

In order
In order to
to evaluate
evaluate separately
separately the
the response
response of of the first nine FEM vibration modes, the ratio fnn/f /f11
hasbeen
has beenplotted
plottedininFigure
Figure6 6for
forthe
the calculated
calculated strain
strain values.
values. TheThesame same tendency
tendency is observed
is observed in allinthe
all
the graphs
graphs of Figure
of Figure 6. For
6. For lowlow strains,
strains, thethe numerical
numerical ratio
ratio tends
tends totothe
theratio
ratioobtained
obtainedwith withthe
the plate
plate
behaviour, and
behaviour, and forfor higher
higher strains
strains toto the
the membrane
membrane behaviour.
behaviour. This This isis repeated
repeated in in all
all the
the graphed
graphed
modes, which
modes, which present
present horizontal
horizontal asymptotes
asymptotesin in the
the region
region ofof lower
lower and
and higher
higher strains,
strains, whose
whose value
value
coincides with
coincides with the
therelationships
relationshipsbetween
betweenthetheanalytical
analyticalfrequencies
frequenciesof ofthe
theplate
plateand
andmembrane
membranecases.cases.
The transition
The transition zone
zone isis observed
observed clearly.

Figure 6. Cont.
Appl. Sci. 2019, 9, 4716 9 of 17
Appl. Sci. 2019, 9, x FOR PEER REVIEW 9 of 17

Figure 6. Ratio
Figure 6. Ratio /f1 the
fn/f1fnfor natural
for the frequencies
natural frequenciesof of
some
some modes
modesobtained
obtainedwithwithFEM.
FEM.(a)
(a) Mode
Mode 2.
2. (b)
Mode
(b) Mode 3. Mode
3. (c) (c) Mode 4. (d)4.Mode
(d) Mode 5. (e) Mode
5. (e) Mode 6. (f)7.Mode
6. (f) Mode 7. (g)
(g) Mode 8. Mode 8. (h)
(h) Mode 9. Mode 9.

The
The response
response inin this
this transition
transition zone
zone was
was studied
studied by
by Wah
Wah [30]. If Equation
[30]. If Equation (3)
(3) given
given by
by Wah
Wah isis
used to calculate the natural frequencies (f WAH ), the values of Table 4 are obtained for
used to calculate the natural frequencies (fWAH), the values of Table 4 are obtained for the different the different
strain
strain values
values provided
providedby bythis
thisauthor.
author.The
Thestrain
strainvalues
valuesare
arerelated toto
related thethe
dimensionless
dimensionless parameter
parameterφ,
also given in Table 4. The numerical frequencies obtained with
ϕ, also given in Table 4. The numerical frequencies obtained with FEM FEM (f ) and the ratio between both
FEM (fFEM) and the ratio between
(fWAH /f FEM ) are also included in Table
both (fWAH/fFEM) are also included in Table 4.4.

Table 4. Ratio fWAH /fFEM for first nine modes and strain selected.
Table 4. Ratio fWAH/fFEM for first nine modes and strain selected.
STRAIN
STRAIN
0 9.410 × 10−7 1.882 × 10−6
0 9.410 × 10−7 1.882 × 10−6
fWAH fWAH fWAH
fWAH fFEM fWAH/ fWAH fFEM fWAH / fWAH fFEM fWAH /
fWAH fFEM /fFEM fWAH fFEM /fFEM fWAH fFEM /fFEM
fFEM fFEM fFEM
Mode 1 10.203 10.179 1.002 11.319 11.346 0.998 12.398 12.395 1.000
Mode 1 10.203 10.179 1.002 11.319 11.346 0.998 12.398 12.395 1.000
Mode 2 21.176 21.185 1.000 22.565 22.581 0.999 23.917 23.890 1.001
Mode
Mode 3 2 21.176
34.921 21.185
34.754 1.000
1.005 22.565
36.468 22.581
36.261 0.999
1.006 23.917
37.932 23.890
37.704 1.001 1.006
Mode
Mode4 3 39.674
34.921 39.634
34.754 1.001
1.005 41.085
36.468 41.141
36.261 0.999
1.006 42.547
37.932 42.595 1.006
37.704 0.999
Mode 6 60.157 60.622 0.992 61.804 62.196 0.994 63.245 63.732 0.992
Mode 4 39.674 39.634 1.001 41.085 41.141 0.999 42.547 42.595 0.999
Mode 8 84.521 84.296 1.003 86.146 85.915 1.003 87.735 87.503 1.003
Mode
Mode 9 6 60.157
88.795 60.622
88.810 0.992
1.000 61.804
90.417 62.196
90.429 0.994
1.000 63.245
92.003 63.732
92.019 0.992 1.000
Mode 8 84.5213.76484.296
× 10−6 1.003 86.146 5.646
85.915
× 10−6 1.003 87.735 87.503
7.528 × 10−6 1.003
Mode 9 88.795 88.810 f 1.000 90.417 90.429 1.000
fWAH 92.003 92.019 1.000
fWAH
WAH
fWAH fFEM fWAH fFEM −6 fWAH fFEM
3.764 × 10−6 /fFEM 5.646 × 10 /fFEM 7.528 × 10−6 /fFEM
Mode 1 14.214 14.244 fWAH/
0.998 15.870 15.861 f1.001
WAH/ 17.312 17.313 fWAH /
1.000
fWAH fFEM fWAH fFEM fWAH fFEM
Mode 2 26.315 26.302 1.001
fFEM 28.486 28.496 1.000
fFEM 30.505 30.523 0.999
fFEM
Mode 3 40.763 40.429 1.008 43.236 42.972 1.006 45.508 45.365 1.003
Mode 1 14.214 14.244 0.998 15.870 15.861 1.001 17.312 17.313 1.000
Mode 4 45.402 45.358 1.001 47.910 47.958 0.999 50.356 50.420 0.999
Mode
Mode 6 2 26.315
66.142 26.302
66.695 1.001
0.992 28.486
69.029 28.496
69.529 1.000
0.993 30.505
71.615 30.523
72.249 0.999
0.991
Mode
Mode8 3 90.991
40.763 90.595
40.429 1.004
1.008 93.754
43.236 93.583
42.972 1.002
1.006 96.769
45.508 96.476
45.365 1.003
1.003
Mode 9 95.077 95.119 1.000 98.405 98.120 1.003 101.435 101.031 1.004
Mode 4 45.402 45.358 1.001 47.910 47.958 0.999 50.356 50.420 0.999
Mode 6 66.142 66.695 0.992 69.029 69.529 0.993 71.615 72.249 0.991
The dimensionless
Mode 8 90.991
parameter1.004
90.595
φ is given for values
93.754
of nodal
93.583
diameters
1.002 96.769
(n) and nodal 1.003
96.476
circles (s)
between 0 and 2. For modes with n and/or s greater than 2, the frequencies cannot be obtained with
Mode 9 95.077 95.119 1.000 98.405 98.120 1.003 101.435 101.031 1.004
the values provided by Wah. Due to this, among the first nine modes the 5th and 7th cannot be
Appl. Sci. 2019, 9, x FOR PEER REVIEW 10 of 17

Appl. Sci.
The2019, 9, 4716
dimensionless parameter ϕ is given for values of nodal diameters (n) and nodal circles 10 of(s)
17

between 0 and 2. For modes with n and/or s greater than 2, the frequencies cannot be obtained with
the values provided by Wah. Due to this, among the first nine modes the 5th and 7th cannot be
calculated with the Wah equation, so their values do not appear in Table 4. Nevertheless, regardless
calculated with the Wah equation, so their values do not appear in Table 4. Nevertheless, regardless
these limitations, the comparison with these results is useful to validate our numerical results.
theseIflimitations, the comparison with these results is useful to validate our numerical results.
the ratios fANM /fWAH and fANP /fWAH are calculated for all the strains of the Wah case, the results
If the ratios fANM/fWAH and fANP/fWAH are calculated for all the strains of the Wah case, the results
obtained are very similar to the ratios fANM /fFEM and fANP /fFEM . If the Wah points are plotted in
obtained are very similar to the ratios fANM/fFEM and fANP/fFEM. If the Wah points are plotted in Figures
Figures 4 and 5, these points are superimposed with the FEM results on the corresponding graphs.
4 and 5, these points are superimposed with the FEM results on the corresponding graphs.
The main limitation of Wah solution is that the vibration modes can only be calculated for a short
The main limitation of Wah solution is that the vibration modes can only be calculated for a
range of strain. This range does not cover the whole transition zone between the plate-to-membrane
short range of strain. This range does not cover the whole transition zone between the plate-to-
behaviour, as it is shown in Figure 7. These limitations are solved using finite element method,
membrane behaviour, as it is shown in Figure 7. These limitations are solved using finite element
with which all modes of vibration and at any range of pre-strain can be calculated.
method, with which all modes of vibration and at any range of pre-strain can be calculated.

Figure
Figure 7. 7.Wah
Wah points
points superimposed
superimposed onon graphics
graphics ofof fANM/f/fFEM
fANM FEM (a) and f ANP/f/f
ANP FEM (b)
FEM (b)ratios.
ratios.

In order to complete this study, calculations have been repeated for different thicknesses: 10 mm, mm,
1 mm (reference
(reference case),
case), 0.1
0.1 mm,
mm, 0.01 mm, 0.001 mm, for the same level of strains,
strains, diameter and material
properties
properties that
thatthe
thereference
referencecase
casestudy
study(steel 1m
(steel diameter).
1m The
diameter). ratios
The fANM
ratios /fFEM
fANM /fFEMand fANP/f/fFEM
andfANP FEM have

been plotted in Figure 8 for the first

first five
five modes.
modes.
Appl. Sci. 2019, 9, 4716 11 of 17
Appl. Sci. 2019, 9, x FOR PEER REVIEW 11 of 17

Figure
Figure 8. Ratios
8. Ratios fANM /fFEM
fANM /fFEM(a)
(a)and
andfANP
fANP/fFEM
FEM (b)
(b)asasaafunction
functionofofstrain
straintested
testedfor
forthe
thefirst
firstfive
fivemodes
modesand
values of thickness/diameter (h/d): 10−2−2
, 10−3,−3
10 −4
−4, 10 −5
−5, and 10 −6. −6
and values of thickness/diameter (h/d): 10 , 10 , 10 , 10 , and 10 .

In the graphs
In the of Figure
graphs 8, the8,same
of Figure the tendency is observed
same tendency for all the
is observed forstudied
all thecases:
studied for cases:
small strains,
for small
thestrains,
plate behaviour
the plateisbehaviour
fulfilled and is for large strains
fulfilled and forthelarge
membrane
strains behaviour.
the membrane Therebehaviour.
is a transition zoneis a
There
between thesezone
transition behaviours,
between andthese
it moves to smallerand
behaviours, strain
it values
moves for to smaller
smallerthickness/diameter
strain values for ratios. smaller
It can be observed that the curves obtained follow thethat
samethe shape displaced an order of magnitude 2
thickness/diameter ratios. It can be observed curves obtained follow the same 10 shape
in strain,
displacedwhatanis order
in concordance
of magnitude with 10
the2 fact that the
in strain, ratioisbetween
what the platewith
in concordance and membrane
the fact that solution
the ratio
is dependent 1/2 , considering only the geometric parameter and 1/2
between the plate and membrane solution is dependent on a ratio (h/d)/ε , considering onlyofthe
on a ratio (h/d)/ε removing the effect
thegeometric
material properties.
parameter and removing the effect of the material properties.
It isItremarkable
is remarkable that, in certain
that, in ratios
certainof ratios
thickness/diameter simulations,simulations,
of thickness/diameter the membrane thebehaviour
membrane
is observed with very low tension applied; those are cases hypersensitive
behaviour is observed with very low tension applied; those are cases hypersensitive to the applied to the applied tension.
The fn /f1 ratio has been plotted in Figure 9 for the FEM results of the first five modes for the
tension.
differentThe values
fn/f1 of thickness/diameter
ratio has been plotted(h/d) calculated:
in Figure 9 for the10−2 , 10−3
FEM , 10−4 ,of10the
results −5 , and 10−6 . As for the
first five modes for the
case previously
different studied,
values the transition zone(h/d)
of thickness/diameter is moved to lower
calculated: 10 strains
−2 , 10 , 10
−3 or stresses
−4 , 10 , and
−5 applied
10−6. with
As forsmaller
the case
values of thickness/diameter
previously ratio.
studied, the transition zone is moved to lower strains or stresses applied with smaller
In addition
values to this theoretical
of thickness/diameter analysis of the problem, it is of great relevance to evaluate the
ratio.
consequences of this behaviour in practical situations. Based on the data shown in Figure 8, we can
calculate, for every case, the values of strain where the exact solutions (plate and membrane) present a
certain error. In Figure 10 the strain for a 2% and a 5% error is represented for the different geometrical
cases and strains studied using logarithmic scale in both axes. The error has been calculated
Appl. Sci. 2019, 9, 4716 12 of 17

as ((fANP −fFEM )/fANP )·100 and ((fANM −fFEM )/fANM )·100 for plate and membrane cases respectively.
The discrete solutions obtained by Wah have also been plotted as reference. As it has been commented,
these solutions
Appl. Sci. 2019, 9, x do
FORnot cover
PEER the entire transition zone between plate and membrane behaviour.12 of 17
REVIEW

Appl. Sci. 2019, 9, x FOR PEER REVIEW 13 of 17

Figure
That 9. the fn /f1 of
Ratiocase for the
the first five natural frequencies obtained with FEM inand values of thickness/diameter
Figureis−2
9. Ratio
−3
fn−4
/f1 for
−5
reported
the first
−6
cases in MEMS
five natural design
frequencies and
obtainedthewith
behaviour
FEM and of values
the tympanic
of
(h/d):
membrane 10 , 10 ,
of mammals. 10 , 10 , and 10 . (a) Mode 2. (b) Mode 3. (c) Mode 4. (d) Mode 5.
thickness/diameter (h/d): 10 , 10 , 10 , 10 , and 10 . (a) Mode 2. (b) Mode 3. (c) Mode 4. (d) Mode 5.
−2 −3 −4 −5 −6

In addition to this theoretical analysis of the problem, it is of great relevance to evaluate the
consequences of this behaviour in practical situations. Based on the data shown in Figure 8, we can
calculate, for every case, the values of strain where the exact solutions (plate and membrane) present
a certain error. In Figure 10 the strain for a 2% and a 5% error is represented for the different
geometrical cases and strains studied using logarithmic scale in both axes. The error has been
calculated as ((fANP−fFEM)/fANP)·100 and ((fANM−fFEM)/fANM)·100 for plate and membrane cases
respectively. The discrete solutions obtained by Wah have also been plotted as reference. As it has
been commented, these solutions do not cover the entire transition zone between plate and
membrane behaviour.
In this figure, we can identify a zone with a clear plate behaviour for strains below a certain
value, another zone with pure membrane behaviour and an intermediate zone where the transition
between both solutions must be considered.
In a real situation, we must consider that the material is not infinitely elastic. For the steel, if we
Figurean
consider Strain for solution
10.approximate with alimit
elastic 2% and of 5%400error
MPafor different
we canvaluesaccept that10−2
of h/d: −3 , 10−4 , and 10−5 .
its, 10strain elastic limit is
Figure 10. Strain for solution with a 2% and 5% error for different values of h/d: 10−2, 10−3, 10−4, and
approximately at 1.5 × 10 , which has been used as actual maximum limit for Figure 10. An additional
−3
10−5this
In . figure, we can identify a zone with a clear plate behaviour for strains below a certain value,
horizontal line has been added to mark this limit. This reference together with the error lines plotted
another zone with pure membrane behaviour and an intermediate zone where the transition between
can
5. be used Equivalent
Eardrum as a reference to evaluate which situation can be expected in a real case.
Membrane
both solutions must be considered.
If we analyse three different situations, we can clarify this point. Focussing on the case where
In
Ona the realpresent
situation, we must
the thickness of section,
the membrane is 1consider
the study
cm (h/dwill =that the material
10focus
−2), on be
it can is not infinitely
a membrane
considered awith elastic. For
thickgeometry
membrane andorthe steel,
material
what we
if we consider
properties
commonlysimilar an approximate
to theHere
call a plate. human we elastic
eardrum.
can limit
see how of 400
the MPa
plate we can
solution accept
is valid thatfor its strain
almost elastic
all the limit
potentialis
approximately at 1.5of×the
10 −3 , which has been
values Theofdiameter
pre-strain. Onlycircular membrane
for very high used willasbeactual
values 1 cm
of maximum
closetolimit
and tries
strain, forplasticity,
torepresent
the Figure the 10.
humanAn
the additional
tympanic
transition
horizontal
behaviour would be relevant. If we consider now a thin membrane, the reference case plotted
membrane line
(TM). has Thebeen added
actual TM to mark
has an this
oval limit.
shape This
with reference
a higher together
diameter with the
ranging error
from lines
0.9 to 1 cma
with
can
and be
thicknessused
the lower as
of 1froma reference
mm 0.8 (h/d to
to =0.9 evaluate
10cm. There
−3), we which
canaresee situation
other can
nowsignificant be expected
how the differences in
transition zone a real
such fall case.
as itsinconical
a wideshaperangeand of
that If
it we
is analyse
attached three
to a different
rigid situations,
element (the we
malleus) can clarify
at its this
centre point.
(the Focussing
umbo).
behaviour with strain in the order of a 0.1% of the elastic limit. In this problem, neither of both theories on the
Nevertheless, case where
the the
present
thickness
model
are valid canof bethe
and amembrane
considered
completevalid is 1to
study (h/d = 10
cmevaluate
including
−2 ), it can be considered a thick membrane or what we
thetheeffect of pre-strain
pre-strain effectinshould
this material.
be done. Finally, we can
commonlyThe call
average a plate.
eardrum Here we can
thickness see
is how
50 the
microns,
comment the case with a thickness of 0.1 mm (h/d = 10 ) that can be considered plate
and
−4 solution
this will is
bevalid
the for almost
thicknessa very ofall the
this
thin potential
case study
membrane.
(h/d = 5 × 10 −3). The mechanical properties of the material are: density 1200 kg/m3, Poisson's ratio 0.3
In this case, it can be observed how the plate solution is not valid for very slight values of pre-strain.
and
It is Young's
a behaviour modulus (E) 32 MPa. These
of hypersensitivity where properties
very small arestrain
commonlyvaluesaccepted
make the asmembrane
a good estimationbehavesfor in
the human TM [29].
the transition zone. Once exceeded that value it behaves as a pure membrane.
In
This Figure
is a case11, that
the ratios
is not an fANM /fFEM andreal
expected fANPsituation
/fFEM have been
with thisplotted
material.according
Nevertheless,to the applied strain
other material
for the first(combination
properties nine modes. of In density
addition, and themechanical
ratios fANMproperties)
/fWAH and fANP may /fWAH have been
be present superimposed
in practical on
situations
Appl. Sci. 2019, 9, x FOR PEER REVIEW 13 of 17

That is the case of the reported cases in MEMS design and in the behaviour of the tympanic
membrane of mammals.
Appl. Sci. 2019, 9, 4716 13 of 17

values of pre-strain. Only for very high values of strain, close to the plasticity, the transition behaviour
would be relevant. If we consider now a thin membrane, the reference case with a thickness of 1 mm
(h/d = 10−3 ), we can see now how the transition zone fall in a wide range of behaviour with strain
in the order of a 0.1% of the elastic limit. In this problem, neither of both theories are valid and a
complete study including the pre-strain effect should be done. Finally, we can comment the case with a
thickness of 0.1 mm (h/d = 10−4 ) that can be considered a very thin membrane. In this case, it can be
observed how the plate solution is not valid for very slight values of pre-strain. It is a behaviour of
hypersensitivity where very small strain values make the membrane behaves in the transition zone.
Once exceeded that value it behaves as a pure membrane.
This is a case that is not an expected real situation with this material. Nevertheless, other material
properties (combination of density and mechanical properties) may be present in practical situations
where this transition zone is very relevant and plate theory cannot be used.
That
Figureis10.
theStrain
casefor
of solution
the reported
with acases in 5%
2% and MEMS
error design and values
for different in the of
behaviour of−3,the
h/d: 10−2, 10 10−4tympanic
, and
membrane
10 . of mammals.
−5

5. Eardrum Equivalent Membrane

On thethepresent
present section, thethe
section, study will focus
study on a membrane
will focus on a membrane with geometry and material
with geometry andproperties
material
similar
propertiesto the human
similar eardrum.
to the human eardrum.
The diameter of the circular membrane will be 1 cm and tries to represent the the human
human tympanic
tympanic
membrane (TM). The actual TM has an oval shape with a higher diameter ranging from 0.9 to 1 cm
and the lower from 0.8 to 0.9 cm. There are other significant differences such as its conical shape and
that it is
is attached
attached to to aa rigid
rigid element
element (the (the malleus)
malleus) at at its
its centre
centre (the
(the umbo).
umbo). Nevertheless,
Nevertheless, the present
model can be considered
considered valid valid to to evaluate
evaluate the the effect
effectof ofpre-strain
pre-strainin inthis
thismaterial.
material.
The average eardrum thickness is 50 microns, and this will be the thickness of this case study
(h/d =
(h/d =5× 10−3−3).).The
× 10 Themechanical
mechanicalproperties
propertiesof ofthe
the material
material are: density 1200 kg/m33, Poisson's
are: density Poisson’s ratio 0.3
and Young’s
Young's modulus
modulus (E) (E) 32
32 MPa. These properties are commonly accepted as a good estimation for
the human TM [29]. [29].
In Figure
Figure11, 11,the
theratios fANM/f
ratiosfANM /fFEM and fANP
FEM and ANP/f/fFEM
FEM havebeen
have beenplotted
plottedaccording
accordingtoto the
the applied
applied strain
for the
the first
firstnine
ninemodes.modes.InInaddition,
addition,the theratios
ratiosfANMfANM/f/f
WAH
WAH and
and ff /f
/f
ANP WAH
ANP WAH have
have been
been superimposed on
the graphs for the Wah points. It is observed that again the latter ratios coincide
the graphs for the Wah points. It is observed that again the latter ratios coincide with the first one. with the first one. This is
because the frequencies
This is because obtainedobtained
the frequencies with the Wahwith equation
the Wahare similarare
equation to those calculated
similar to thosewith FEM. with
calculated
FEM.

Figure 11. Cont.

Appl. Sci. 2019, 9, 4716 14 of 17
Appl.
Appl.Sci.
Sci.2019,
2019,9,9,xxFOR
FORPEER
PEERREVIEW
REVIEW 14
14 of
of 17
17

Figure 11.Ratios
Figure11.
11. RatiosffANM
Ratios fANM /f/fFEM
ANM/fFEM
(a)and
FEM(a)
(a) andffANP
and fANP
ANP /f
/f/fFEM (b)(b)
(b)
FEM
FEM asas
as afunction
functionof
aa function ofofthe
thestrain
the strainfor
strain forthe
for thefirst
the first nine
first nine modes.
nine modes.
modes.
Analytical
Analyticalsolutions
solutionsby byWahWah(points)
Wah (points)are
(points) aresuperimposed
are superimposedin
superimposed inthe
in thegraph.
the graph.
graph.

For the
For eardrum equivalent membrane andand strain tested,
tested, the first
first five FEM vibration modes
For the
the eardrum
eardrum equivalent
equivalent membrane
membrane and strain
strain tested, the
the first five
five FEM
FEM vibration
vibration modes
modes
referred to
referred to the
the first
first one
one have
have been
been plotted
plotted in
in Figure
Figure 12.
12.
referred to the first one have been plotted in Figure 12.

Figure 12.
Figure Ratio ffnn/f/f11for
for thefirst
first fivenatural
natural frequencies obtained
obtained with FEM.
FEM. (a) Mode
Mode 2. (b)
(b) Mode 3.
3.
Figure12.
12.Ratio
Ratio fn/f1 forthe the firstfive
five naturalfrequencies
frequencies obtainedwith
with FEM.(a)
(a) Mode2.
2. (b)Mode
Mode 3.
(c)
(c) Mode 4. (d) Mode 5.
(c)Mode
Mode4.4.(d)
(d)Mode
Mode5.5.
The same behaviour that for the steel case is observed in the graphs of Figures 11 and 12, at low
The
Thesame
same behaviour
behaviouristhat that for
forthe
the steel
steel case isisobserved
caseand observed in
inthe graphs
thestrain
graphs of
ofFigures
Figures 11
11andand12, 12,atat low
pre-strains the behaviour similar to the plate as the applied increases, it approaches to low
the
pre-strains
pre-strains the
the behaviour
behaviour is
is similar
similar to
to the
the plate
plate and
and as
as the
the applied
applied strain
strain increases,
increases, it
it approaches
approaches to
to
membrane behaviour more and more.
the membrane
the membrane behaviour
behaviour more
more and
and more.
more.
In the study of this case, the eardrum is a material with less tensile stiffness than the case studied
In the
thestudy of
ofthis
thiscase,
case,the eardrum isisaamaterial with
withless tensile stiffness than
thanthe case studied
in theInprevious studysection (steel), thesoeardrum
the tension material
values related less
to tensile stiffness
the strain for which the
thecase studied
behaviour
in
in the
the previous
previous section
section (steel),
(steel), so
so the
the tension
tension values
values related
related to
to the
the strain
strain for
for which
which the
the behaviour
behaviour
transition from membrane to plate are smaller, and therefore, the material of the eardrum is more
transition
transitiontofrom
from membrane
membrane to to plate
plate are
are smaller,
smaller, and and therefore,
therefore, the the material
material of of the
the eardrum
eardrum isis more more
sensitive tension.
sensitive
sensitive to
to tension.
tension.
There exists some controversy regarding the correct estimation of the mechanical properties of
There
There exists
exists some
some controversy
controversy regarding
regarding the correct
thetests
correct estimation of of the mechanical properties of
the TM. Most of the results are based on tension on estimation
small samples the mechanical
[32,33] ranging properties
from 20 to of
the TM.
theMPa. Most of
TM. Nanoindentationthe results
Most of the results are based
are basedhave on tension
on tension tests on
testsused small
on small samples
samples [32,33] ranging from 20 to 40
40 techniques also been reporting even[32,33]
smallerranging
values fromon the20range
to 40
MPa.
MPa. Nanoindentation
Nanoindentation techniques
techniques have
have also
also been
been used
used reporting
reporting even
even smaller
smaller values
values on
on the
the range
range of
of
of 3 MPa [34]. In a different type of study, by means of composite laminate theory and comparing with
33 MPa [34]. In a different type of study, by means of composite laminate
MPa [34]. In a different type of study, by means of composite laminate theory and comparing with theory and comparing with
the
the dynamic
dynamic response
response of of thethe TMTM [35],
[35], Fay
Fay et et al.
al. suggested
suggested that that the
the elastic
elastic modulus
modulus should
should be be in
in the
the
Appl. Sci. 2019, 9, 4716 15 of 17

the dynamic response of the TM [35], Fay et al. suggested that the elastic modulus should be in the
range 0.1 to 0.3 GPa. These higher values could be overestimated by the presence of tension in the
tympanic membrane that is released when the static tests are performed. In a previous research we
have shown how experimental data can fit numerical models with a huge combination of pre-strain
and elastic modulus values [29], being very sensitive to small values of the pre-strain in accordance
with the results shown in Figure 11. The presence of pre-strain in the tympanic membrane has been
hypothesised since long time ago [36,37] and is a question under discussion in current hearing research.

6. Conclusions
The influence of the membrane tension has been evaluated by means of a FEM study and the
main finding can be summarize as follows:

• The range where the classic theories are applicable has been obtained. A significant transition
zone where neither plate theory nor membrane theory can be applied has been identified. It has
been quantified for a reference case.
• It has been proven the validity of the theory of Wah [30] to describe the behaviour of part of that
transition zone. It has been completed with FEM results for a wider range of values, connecting
with the extreme results (plate and membrane theory).
• There has been identified some geometric ratios with significant hypersensitivity to very low
tension. Some of these configurations are present in practical problems. One of them is in the
different mammals hearing systems (including the human).

The main consequence from these findings is the need to consider this effect when modelling
membrane problems with low tension. It implies a double difficulty. On the one hand a proper
estimation of the value of the pre-strain effect must be done. This is not easy to do, and, in most cases,
there is no information to properly establish the value. On the other hand, once it has been obtained,
it must be incorporated in the model and as it has been stated, small deviation provides significant
response changes.
One of the fields where this effect is relevant is the mechanics of hearing. Among researcher in
this field, it is commonly accepted that the tympanic membrane presents a low level of tension, so this
effect has been considered negligible in numerical model. According to these results, this effect should
be evaluated and taken into account.
Another situation is the manufacturing of micro devices with membrane elements. In this case,
the presence of pre-strain is due to effects present during the different manufacturing steps. It can be
caused for different reasons: temperature change, material shrinkage, etc. that are difficult to clarify
and quantify.
Finally, it must be pointed out a potential application of these results. The complete correlation
obtained among pre-strain and natural frequencies opens an experimental option to determinate the
level of pre-strain of a membrane system. By fitting experimental and numerical modal frequencies,
using model update techniques, an estimation of the level of tension can be made. In this case,
the model uncertainties (in term of mechanical properties and geometry) must be limited in order to
establish a proper correlation among experimental and numerical results.

Author Contributions: Conceptualization, A.G.-H. and J.G.-M.; format analysis, A.L.-R.; investigation, A.L.-R.,
A.G.-H. and J.G.-M.; methodology, A.G.-H. and J.G.-M.; software, A.L.-R. and J.G.-M.; validation, A.L.-R. and
A.G.-H.; writing—original draft preparation, A.L.-R. and A.G.-H.; writing—review and editing, A.G.-H. and
J.G.-M.
Funding: The authors acknowledge the financial support provided to this work by Junta de Andalucía from
Operational Program FEDER 2014-2020 as a part of the research project UMA18-FEDERJA-214.
Conflicts of Interest: The authors declare no conflict of interest.
Appl. Sci. 2019, 9, 4716 16 of 17

Nomenclature Check Use of Italic Font to Match Main Text

f ≡ natural frequency (Hz)
a ≡ radius of the plate or membrane
d ≡ diameter of the plate or membrane
h ≡ thickness of the plate or membrane
E ≡ Young’s modulus of the material
υ ≡ Poisson´s ratio of the material
γ ≡ weight per unit volume
ρ ≡ mass per unit volume (material density)
w ≡ weight per unit area
g ≡ acceleration due to gravity
S or T ≡ uniform tension per unit length of the boundary
D ≡ flexural rigidity
α ≡ constant different for each mode
n ≡ number of nodal diameters
s ≡ number of nodal circles
φ ≡ non-dimensional parameter

References
1. Rayleigh, L. Theory of Sounds; Macmillan & Co.: London, UK, 1877.
2. Timoshenko, S. Vibration Problems in Engineering; D. Van Nostrand Company Inc.: New York, NY, USA, 1928.
3. Leissa, A.W. Vibration of Plates; National Aeronautics and Space Administration: Washington, DC, USA, 1969.
4. Poisson, S.D. L’équilibre et le mouvement des corps élastiques. Mém. Acad. Sci. Inst. Fr. 1828, 8, 357–570.
5. Kirchhoff, G. Über das Gleichgewicht und die Bewegung einer elastischen Scheibe. J. Reine Angew. Math.
1850, 40, 51–88.
6. Airey, J.R. The vibrations of circular plates and their relation to bessel functions. Proc. Phys. Soc. Lond.
1911, 23, 225. [CrossRef]
7. Itao, K.; Crandall, S.H. Natural modes and natural frequencies of uniform, circular, free-edge plates. J. Appl.
Mech. Trans. ASME 1979, 46, 448–453. [CrossRef]
8. Amabili, M.; Pasqualini, A.; Dalpiaz, G. Natural frequencies and modes of free-edge circular plates vibrating
in vacuum or in contact with liquid. J. Sound Vib. 1995, 188, 685–699. [CrossRef]
9. Leissa, A.W.; Narita, Y. Natural frequencies of simply supported circular plates. J. Sound Vib. 1980, 70, 221–229.
[CrossRef]
10. Azimi, S. Free vibration of circular plates with elastic edge supports using the receptance method. J. Sound
Vib. 1988, 120, 19–35. [CrossRef]
11. Leissa, A.W.; Laura, P.A.; Gutierrez, R.H. Transverse vibrations of circular plates having nonuniform edge
constraints. J. Acoust. Soc. Am. 1979, 66, 180–184. [CrossRef]
12. Narita, Y.; Leissa, A.W. Transverse vibration of simply supported circular plates having partial elastic
constraints. J. Sound Vib. 1980, 70, 103–116. [CrossRef]
13. Narita, Y.; Leissa, A.W. Flexural vibrations of free circular plates elastically constrained along parts of the
edge. Int. J. Solids Struct. 1981, 17, 83–92. [CrossRef]
14. Amabili, M.; Pierandrei, R.; Frosali, G. Analysis of vibrating circular plates having non-uniform constraints
using the modal properties of free-edge plates: Application to bolted plates. J. Sound Vib. 1997, 206, 23–38.
[CrossRef]
15. Eastep, F.E.; Hemmig, F.G. Natural frequencies of circular plates with partially free, partially clamped edges.
J. Sound Vib. 1982, 84, 359–370. [CrossRef]
16. Kwak, M.K. Vibration of circular plates in contact with water. J. Appl. Mech. 1991, 58, 480–483. [CrossRef]
17. Hernández, E. Approximation of the vibration modes of a plate and shells coupled with a fluid. J. Appl.
Mech. 2006, 73, 1005–1010. [CrossRef]
18. McIntyre, M.E.; Schumacher, R.T.; Woodhouse, J. On the oscillations of musical instruments. J. Acoust. Soc.
Am. 1983, 74, 1325–1345. [CrossRef]
Appl. Sci. 2019, 9, 4716 17 of 17

19. Alm, J.F.; Walker, J.S. Time-frequency analysis of musical instruments. SIAM Rev. 2002, 44, 457–476.
[CrossRef]
20. Fletcher, N.; Rossing, T. The Physics of Musical Instruments; Springer: New York, NY, USA, 1998; ISBN
10:0-387-98374-0.
21. Von Helmholtz, H.; Ellis, A.J. On the Sensations of Tone as a Physiological Basis of the Theory of Music; Longmans,
Green and Company: London, UK, 1895.
22. Suzuki, H.; Yamaguchi, N.; Izumi, H. Theoretical and experimental studies on the resonance frequencies of a
stretched circular plate: Application to Japanese drum diaphragms. Acoust. Sci. Technol. 2009, 30, 348–354.
[CrossRef]
23. Lee, J.; Wang, Z.; He, K.; Shan, J.; Feng, P.X.L. High frequency MoS2 nanomechanical resonators. ACS Nano
2013, 7, 6086–6091. [CrossRef]
24. Sammoura, F.; Smyth, K.; Bathurst, S.; Kim, S.G. An analytical analysis of the sensitivity of circular
piezoelectric micromachined ultrasonic transducers to residual stress. In Proceedings of the 2012 IEEE
International Ultrasonics Symposium, Dresden, Germany, 7–10 October 2012; pp. 580–583.
25. Chen, K.-S.; Ou, K.-S. Modification of curvature-based thin-film residual stress measurement for MEMS
applications. J. Micromech. Microeng. 2002, 12, 917–924. [CrossRef]
26. Tang, Y.J.; Chen, J.; Huang, Y.B.; Li, D.C.; Wang, S.S.; Li, Z.H.; Zhang, W.D. Ultra-sensitive, highly reproducible
film stress characterization using flexible suspended thin silicon plates and local curvature measurements.
J. Micromech. Microeng. 2007, 17, 1923–1930. [CrossRef]
27. Griffin, B.A.; Williams, M.D.; Wang, G.; Sankar, B.V.; Cattafesta, L.N.; Sheplak, M. The electromechanical
behavior of piezoelectric thin film composite diaphragms possessing in-plane stresses. J. Micromech. Microeng.
2017, 27, 045017. [CrossRef]
28. Gonzalez-Herrera, A.; Olson, E.S. A study of sound transmission in an abstract middle ear using physical
and finite element models. J. Acoust. Soc. Am. 2015, 138, 2972–2985. [CrossRef] [PubMed]
29. Caminos, L.; Garcia-Manrique, J.; Lima-Rodriguez, A.; Gonzalez-Herrera, A. Analysis of the mechanical
properties of the human tympanic membrane and its influence on the dynamic behaviour of the human
hearing system. Appl. Bionics Biomech. 2018, 2018, 1736957. [CrossRef] [PubMed]
30. Thein, W. Vibration of circular plates. J. Acoust. Soc. Am. 1962, 34, 275–281.
31. Timoshenko, S.; Woinowsky-Krieger, S. Theory of Plates and Shells; McGraw-Hill: New York, NY, USA, 1941;
Volume 148, ISBN 0070647798.
32. Decraemer, W.F.; Maes, M.A.; Vanhuyse, V.J. An elastic stress-strain relation for soft biological tissues based
on a structural model. J. Biomech. 1980, 13, 463–468. [CrossRef]
33. Cheng, T.; Dai, C.; Gan, R.Z. Viscoelastic properties of human tympanic membrane. Ann. Biomed. Eng.
2007, 35, 305–314. [CrossRef]
34. Aernouts, J.; Aerts, J.R.M.; Dirckx, J.J.J. Mechanical properties of human tympanic membrane in the
quasi-static regime from in situ point indentation measurements. Hear. Res. 2012, 290, 45–54. [CrossRef]
35. Fay, J.; Puria, S.; Decraemer, W.F.; Steele, C. Three approaches for estimating the elastic modulus of the
tympanic membrane. J. Biomech. 2005, 38, 1807–1815. [CrossRef]
36. Von Békésy, G. Experiments in Hearing; McGraw-Hill: New York, NY, USA, 1960.
37. Kirikae, I. The Structure and Function of Middle Ear; Tokio University Press: Tokyo, Japan, 1960.

© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).