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

net/publication/237012252

Experimental study of the Timoshenko beam theory predictions

Article in Journal of Sound and Vibration · December 2012

DOI: 10.1016/j.jsv.2012.07.041

CITATIONS READS

38 2,372

6 authors, including:

Jordy Flores Lil Gutiérrez

UNAM Costa Rican Institute of Technology (ITCR)
47 PUBLICATIONS 678 CITATIONS 22 PUBLICATIONS 357 CITATIONS

SEE PROFILE SEE PROFILE

R. A. Méndez-Sánchez Guillermo Monsivais

Universidad Nacional Autónoma de México Universidad Nacional Autónoma de México
90 PUBLICATIONS 850 CITATIONS 89 PUBLICATIONS 1,025 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by R. A. Méndez-Sánchez on 14 March 2018.

The user has requested enhancement of the downloaded file.

This article appeared in a journal published by Elsevier. The attached
copy is furnished to the author for internal non-commercial research
and education use, including for instruction at the authors institution
and sharing with colleagues.
Other uses, including reproduction and distribution, or selling or
licensing copies, or posting to personal, institutional or third party
websites are prohibited.
In most cases authors are permitted to post their version of the
article (e.g. in Word or Tex form) to their personal website or
institutional repository. Authors requiring further information
regarding Elsevier’s archiving and manuscript policies are
encouraged to visit:
http://www.elsevier.com/copyright
 Author's personal copy

Journal of Sound and Vibration 331 (2012) 5732–5744

Contents lists available at SciVerse ScienceDirect

Journal of Sound and Vibration

journal homepage: www.elsevier.com/locate/jsvi

Experimental study of the Timoshenko beam theory predictions

A. Dı́az-de-Anda a, J. Flores b, L. Gutiérrez c, R.A. Méndez-Sánchez c,n, G. Monsivais b,
A. Morales c
a
Departamento de Ciencias Básicas, Universidad Autónoma Metropolitana-Azcapotzalco, Av. San Pablo 180, Col. Reynosa Tamaulipas,
02200 México D. F., Mexico
b
Instituto de Fı́sica, Universidad Nacional Autónoma de México, A.P. 20-364, 01000 México, D. F., Mexico
c
Instituto de Ciencias Fı́sicas, Universidad Nacional Autónoma de México, A.P. 48-3, 62251 Cuernavaca, Morelos, Mexico

a r t i c l e in f o abstract

Article history: The theory of flexural vibrations proposed by Timoshenko almost 90 years ago has been
Received 28 February 2012 the subject of several recent papers. In the Timoshenko beam theory a critical frequency f c
Received in revised form is expected and for frequencies f larger than f c , some authors argue that a second spectrum
3 July 2012
exists. This is still the matter of much debate. In order to settle this question we have
Accepted 19 July 2012
performed experimental measurements of flexural normal-mode frequencies and ampli-
Handling Editor: S. Ilanko
Available online 21 August 2012 tudes which are much more precise than what has been obtained up to now. For some
purposes cylindrical rods are more convenient, whereas in other cases rectangular beams
are easier to deal with, so we analyze rods and beams of different shapes and sizes. We
show that, below f c , all observed resonant peaks can be associated with one of the
compressional, torsional or Timoshenko resonances and that no resonant peaks are left
undetected. The validity of Timoshenko beam theory below f c and that only one spectrum
appears in this regime is therefore proved. The comparison between theory and experi-
ment for rectangular beams also shows that our experimental setup is appropriate, so we
can use it confidently for f 4 f c . The measurements we report here in this regime show the
existence of a second spectrum, at least for free–free boundary conditions.
& 2012 Elsevier Ltd. All rights reserved.

1. Introduction

The study of flexural vibrations in beams and rods is an important problem in elasticity since these vibrating quasi-one-
dimensional systems are ubiquitous [1,2]. Flexural vibrations were first described by Bernoulli and Euler in the eighteenth
century. Their theory leads to a fourth-order differential equation for the transversal displacement. Unfortunately, a
dispersion relation is obtained for which the wave velocity grows quadratically with frequency. Given this unphysical
behavior, it is expected that the Bernoulli–Euler theory be valid only at low frequencies. In the nineteenth century, the
problem was considered again by Rayleigh, who took into account rotational inertia. Almost 90 years ago, Timoshenko
Beam Theory (TBT) was established [3]. This theory agrees with the Bernoulli–Euler results for the lower normal modes
but it fits experimental data at higher frequencies as it is well known and we have proved experimentally for a rod with
free–free boundary conditions [4].
At higher frequencies, the Timoshenko beam theory presents a critical frequency f c , for which the equation is
drastically altered [5,6]. For values of the frequency f o f c , TBT seems to agree well with theoretical and numerical
calculations [7–10] as well as with experimental measurements of the normal-mode spectrum [4,11,12].

n
Corresponding author. Tel.: þ 52 55 56 22 77 88; fax: þ 52 55 56 22 77 75.
E-mail address: mendez@fis.unam.mx (R.A. Méndez-Sánchez).

0022-460X/$ - see front matter & 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.jsv.2012.07.041
 Author's personal copy

A. Dı́az-de-Anda et al. / Journal of Sound and Vibration 331 (2012) 5732–5744 5733

However, for f 4f c this is no longer the case, and what was called the second spectrum of TBT has been the subject of a
long standing debate [11,13–23]. Up to now it is not clear that the TBT predictions, above the cut-off frequency, are valid.
Experimental observations are needed to settle this point definitely. This is precisely what we intend to do in this paper.
One should point out that experimental results are scarce. For example, Chan et al. [7] analyzed a beam with free–free
boundary conditions, which are the easiest boundary conditions to be dealt with experimentally; however, only a few
measured frequency points were obtained. The system studied in Ref. [11], unfortunately, is far from being quasi-one-
dimensional. Our aim is to extend these measurements. Using the electromagnetic acoustic transducer (EMAT) that we
have developed [24] we are able to measure normal-mode frequencies and amplitudes for vibrating rods and beams.
As we will show in Section 3, this EMAT is very selective and distinguishes among compressional, torsional and flexural
vibrations. Notice that the transducer works through Lorentz force on the electrons of the aluminum but not through
magnetic force since the aluminum is diamagnetic [24]. By comparing the experimental spectrum with the well known
values for compressional and torsional waves, and with theoretical values obtained for TBT and free-end boundary
conditions in Section 2, we are able to identify with a very small error the nature of all possible vibrations of a rod for
f o f c . This is most easily done for a cylindrical rod, since the flexural eigenmodes appear as doublets. As we can also
measure the wave amplitudes, we are able to obtain the dispersion relation and compare it with the Timoshenko
predictions. In both comparisons the experimental agreement with TBT is excellent for f o f c , as shown in Section 4.
Once we are convinced that our experimental method renders very precise results, we turn to analyze the case f 4f c
and discuss whether the second spectrum exists or not. In Section 5 we present the comparison between TBT, finite
element method (FEM) results, and experiment and show that, indeed, the second spectrum exists, at least for the free-end
situation.

2. Timoshenko beam theory, solutions and boundary conditions

Besides the transversal displacement x Timoshenko introduced an angular variable C, as indicated in Fig. 1. The
second-order differential equations satisfied by C and x, governing the flexural behavior of a beam of uniform cross-
section, are [26]:
!
qC q2 x q2 x
kGA þ 2 rA 2 ¼ 0, (1)
qz qz qt
 
qx q2 C q2 C
kGA þ C EI 2 þ rI 2 ¼ 0: (2)
qz qz qt
Here E is Young’s modulus, G the shear modulus, r the mass density, I the second moment of area, and A the cross-
sectional area; k is called the Timoshenko shear coefficient, and it is the only free parameter included in this theory.
Eqs. (1) and (2) were obtained under the assumption that the surface Pc in Fig. 1 can be approximated by a flat surface Pp .
The previous equations can be written as a fourth-order differential equation for the transversal displacement x [1],
  4
EI q4 x I E q x q2 x rI q4 x
 1þ þ þ ¼ 0, (3)
rA qz4 A kG qz2 qt2 qt2 kGA qt4
known as Timoshenko’s equation. The angular variable C satisfies the same equation [1]. One should notice that there
exist other theories including more effects like the inertia correction factor of Ref. [25].

Fig. 1. Definition of the angular variable C and transversal displacement x.

 Author's personal copy

5734 A. Dı́az-de-Anda et al. / Journal of Sound and Vibration 331 (2012) 5732–5744

2.1. Solutions of Timoshenko’s equation

The solution to Eq. (3) can be obtained separating the variables as follows. We insert the ansatz
xðz,tÞ ¼ eiot wðzÞ (4)
pffiffiffiffiffiffiffi
in Eq. (3), where i ¼ 1 is the imaginary unit and the angular frequency o ¼ 2pf . This yields
4 2
d w d w b
a 2 þ w ¼ 0, (5)
dz4 dz 4
where we have defined
ro2
a¼ , (6)
Mr

4r2 o2 2
b¼ ðo o2c Þ, (7)
kGE
with M r the ‘‘reduced modulus’’
1 1 1
¼ þ (8)
Mr E kG
and
sffiffiffiffiffiffiffiffiffiffi
kGA
oc ¼ 2pf c ¼ : (9)
rI
The characteristic equation associated to Eq. (5) is
4 2 b
k ak þ ¼ 0, (10)
4
with solutions
pffiffiffiffiffiffiffiffiffiffiffiffi
2 a 7 a2 b
k7 ð oÞ ¼
2
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u 2
2p2 rf
2 uf 2
4ðf f c Þ
2
¼ 72p rf t 2 
2
Mr Mr kGE
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ro2 1 r2 o4 4r2 o2 2 2
¼ 7  ðo oc Þ, (11)
2M r 2 M2r kGE
which are the dispersion relations. From this equation one can see that the solutions change its behavior depending on the
sign of b which in turn will depend on the value of o. As we will show, oc separates the solutions into two regimes:
pffiffiffiffiffiffiffiffiffiffiffiffi
o o oc and o 4 oc . On the one hand, from Eq. (7) one sees that when o o oc , b is negative. Then a þ a2 b 40 so the two
values of k þ are real. This yields exponential solutions of Eq. (5). One can also prove that both values of k are imaginary in
this regime. On the other hand, when o 4 oc , it is possible to show that both k and k þ are imaginary.
The general solution to Eq. (5) for all frequencies is finally given by
wðzÞ ¼ a expðk þ zÞ þ b expðk þ zÞ þ c expðik zÞ þ d expðik zÞ, (12)

where a, b, c, and d are constants to be determined by the boundary conditions.

2.2. Boundary conditions

In this subsection we study a rod of length L with free–free boundary conditions since these boundary conditions can be
easily obtained in the laboratory [4,12]. The boundary conditions for a beam with free ends are given by the vanishing of
moments and shear forces [5,26],

qC
¼ 0, (13)
qz z ¼ 0,L
 
qx 
þ C  ¼ 0, (14)
qz z ¼ 0,L

respectively. Since Eq. (3) is given only for one variable, say x or C, these boundary conditions should be changed to the
corresponding variable when working with that equation. To write the time independent version of Eqs. (1) and (2) we
 Author's personal copy

A. Dı́az-de-Anda et al. / Journal of Sound and Vibration 331 (2012) 5732–5744 5735

need also a separation of variables for the angle C as follows:

Cðz,tÞ ¼ eiot cðzÞ: (15)

Inserting Eqs. (4) and (15) into Eqs. (13) and (14) one gets

dc
¼ 0, (16)
dz  z ¼ 0,L

 
dw 
þ c  ¼ 0: (17)
dz z ¼ 0,L

Substituting Eqs. (4) and (15) into Eqs. (1) and (2) one gets
!
2
dc d w ro2
þ 2 þ w ¼ 0, (18)
dz dz kG

  2
dw d c
kGA þ c EI 2 rIo2 c ¼ 0: (19)
dz dz

The first boundary condition can be obtained by evaluating Eq. (18) at z¼0, L and using Eq. (16). One gets
 
d w
2
ro2 
 ¼  w : (20)
dz2  kG z ¼ 0,L
z ¼ 0,L

From Eqs. (17) and (19) another boundary condition for c is obtained,
 
d c
2
ro2 
 ¼ c : (21)
dz2  E z ¼ 0,L
z ¼ 0,L

Other two boundary conditions can be obtained taking the derivative of Eq. (18) and using Eq. (21). This yields
! 
d w 
3
ro2 ro2 dw
 cþ 3  ¼ : (22)
E dz  kG dz z ¼ 0,L
z ¼ 0,L

Finally, using Eq. (17), we get

 
d w
3
ro2 dw
 ¼ : (23)
dz3  M r dz z ¼ 0,L
z ¼ 0,L

It is expected that Timoshenko’s equation (3) for the variable x with the boundary conditions (20) and (23) should give the
same solutions as Timoshenko’s equation (3) for the variable c with boundary conditions (16) and (21). Although this
seems paradoxical, that this is so as can be easily derived using the relation
qx
 ¼ C þ g0 : (24)
qz
More important is that the solutions obtained in this way should be the same as those of Eqs. (1) and (2) with the
boundary conditions (13) and (14). One should note that the boundary conditions (20) and (23) depend on the frequency
and that they agree with the Bernoulli–Euler boundary conditions [1,4,12] for low frequencies. In what follows, without
loss of generality, we will use only the variable w, its equation and the corresponding boundary conditions.

2.3. Determination of the wave amplitude

It is possible to obtain the normal-mode frequencies and amplitudes using the same procedure as in Ref. [12], i.e., by
substituting directly the solution (12) on the free–free boundary conditions (20) and (23) at z ¼ 0,L. One obtains
0 2 2 2 2
1
ðk þ þ ro ðk þ þ ro ðk þ ro ðk þ ro
2 2 2 2
kG Þ kG Þ kG Þ kG Þ 0 1
B C a
B 2 ro 2
k L 2 ro 2
L 2 ro 2
k L 2 ro2
L C
B ðk þ þ kG Þe þ ðk þ þ kG Þe þ k
ðk þ kG Þe  ðk þ kG Þe  CB b C
k
B CB C
B 3 ro 2
k 3 ro 2
k 3 ro 2
k 3 ro2 k CB C ¼ 0: (25)
B ðk þ þ M Þ þ
ðk þ þ
Þ ðk þ 
Þ ðk þ Þ C@ c A
B r þ Mr  Mr  Mr C
@ 3 2 2 2 2 A d
ðk þ þ roMrk þ Þek þ L ðk þ þ roMrk þ Þek þ L ðk þ roMrk Þek L ðk þ roMrk Þek L
3 3 3

This system of equations has non-trivial solutions when the determinant of the matrix vanishes. One should notice that
the determinant is different according to whether o is larger or smaller than oc .
 Author's personal copy

5736 A. Dı́az-de-Anda et al. / Journal of Sound and Vibration 331 (2012) 5732–5744

3. Experimental method

In this section we describe the experimental setup that we have used to obtain the frequencies and amplitudes of
oscillation of the flexural modes. With the intention of being able to affirm with all certainty that the measured vibrations
correspond to flexural waves, we obtained, both numerically and experimentally, the other two types of vibrations in the
rod: torsional and compressional. By comparing the experimental results with the numerical values for these two types of
waves, we were able without any doubt to identify the flexural vibrations. In this way, a reliable method has been
established.
One of the main problems when measuring elastic vibrations is the lack of selectivity of the detectors for the different
types of waves. In order to solve this problem we have used the experimental setup shown in Fig. 2. At the heart of the
setup is an electromagnetic acoustic transducer (EMAT) which consists of a permanent magnet and a coil as indicated in
Fig. 3. This EMAT has several interesting properties: it can be used both as an exciter as well as a detector; it can be easily
constructed and operates from very low frequencies up to hundreds of kHz; and, last but not least for our purposes, it is
sensible to the direction of motion of the metallic surface under observation. We shall now discuss this point.
Let us deal first with the detector configuration of Fig. 3 in which the direction of the velocity v of the conductor’s
surface lies along the permanent magnet axis. This magnet produces a divergent magnetic field B which induces eddy

CURSOR

ALUMINUM ROD

EXCITER

POWER
DETECTOR AMPLIFIER AMPLIFIER

EMAT LOCK−IN
AMPLIFIER

rod OSCILLATOR

magnet
coil CAMAC

Fig. 2. Experimental setup used to measure the normal-mode frequencies and wave amplitudes in the aluminum rod. In the left-lower corner the
detector is shown for flexural vibrations. The aluminum rod has a length L¼ 1.000 m and radius R ¼0.0127 m; the elastic constants are G ¼26.92 GPa,
E ¼ 67.42 GPa and r ¼ 2699:04 kg=m3 .

Fig. 3. When the velocity of the surface S is parallel to the permanent magnet axis the eddy currents induce an oscillating magnetic field Bi (not shown)
also parallel to the axis.
 Author's personal copy

A. Dı́az-de-Anda et al. / Journal of Sound and Vibration 331 (2012) 5732–5744 5737

currents in the conductor. These currents are excited due to the variation of the magnetic flux in the loop ‘ which oscillates
with the material. The magnetic field Bi generated by the eddy currents induces an emf in the coil; this voltage provides
the signal that detects the oscillation.
To employ the device of Fig. 3 as an exciter, an AC electric current is now sent through the coil. The magnetic field
generated by the coil will induce eddy currents on the conductor’s surface. According to Lorentz force, due to the presence
of the magnetic field B there is an applied force on the electrons forming the eddy currents. This force, perpendicular to the
surface, is proportional to ve  BS with ve the velocity of the electrons and BS the component of the magnetic field B
parallel to the surface. Thus, the electrons feel an applied force perpendicular to ve , so the eddy currents give rise to an
oscillating force normal to the surface and therefore parallel to the permanent magnet axis.
When the direction of motion is parallel to the surface and perpendicular to the magnet axis, as in the device of Fig. 4
acting as a detector, we consider the loop ‘ on the XY-plane. Since the direction of motion is along the Z-axis, there is a
variable magnetic flux across the loop ‘ which induces eddy currents on this loop. The variable magnetic flux is produced
by the Z component of B which increases in the positive direction of the Z-axis. The eddy current i, generated by this
variable magnetic flux, in turn produces a variable magnetic field Bi which induces an emf on the coil; this voltage
provides the signal that detects the oscillation.
To convert the device in Fig. 4 into an exciter, an AC current is sent through the coil. The AC magnetic field generated by
the coil induces a current on the loop ‘. Due to the magnetic part of the Lorentz force, the X projection Bx of the magnetic
field B, acting on the electrons produces a force in the Z direction proportional to vy  Bx , with vy the Y projection of the
electrons velocity. It must be observed that the simultaneous inversions of sign in vy and Bx give rise to a force with the
same sign along the path ‘.
A general conclusion can be obtained from the previous discussion: as a detector, the magnetic field Bi produced by the
motion of the conductor’s surface is parallel to the velocity of the latter. A similar conclusion for the device acting as an
exciter can be reached: the applied force is parallel to the coil axis of Fig. 4. To verify the above qualitative conclusion for
the detector, we have measured the induced voltage in the coil as a function of y, the angle between the direction of
motion, in this case along the Z-axis, and the coil axis, as shown in Fig. 5. In Fig. 6 the change of the coil voltage is given as a
function of y, that is, the coil axis is rotated with respect to the Y-axis of Fig. 5. As can be seen, this voltage takes a
maximum value when y is close to zero, in accordance with our previous conclusion.
We have shown that the EMAT is capable of distinguishing between different modes of oscillation of the conductor. The
relative orientations of the coil and permanent magnet detect different oscillations as indicated in Fig. 7. Here
configurations C1 and C2 detect compressional waves, whereas T1 and T2 do so for torsional oscillations. On the other
hand, F1 excites and detects the displacement x of flexural vibrations and the configuration F2 the angular variable C of

direction Z
of motion v

coil

magnet
Y

Bi B
X

Fig. 4. The direction of motion is perpendicular to the permanent magnet axis. The induced magnetic field is now parallel to the surface velocity.

Z θ
coil axis
magnet
coil

Fig. 5. Experimental setup used to measure the coil voltage as a function of the angle between the coil and the direction of motion. The direction of
motion is along the Z-axis.
 Author's personal copy

5738 A. Dı́az-de-Anda et al. / Journal of Sound and Vibration 331 (2012) 5732–5744

0.3

coil voltage (arb. units)

0.2

0.1

-1 0 1
angle θ (radians)

Fig. 6. The circles correspond to experimental values of the coil voltage as a function of the angle y of Fig. 5. The solid line equation is 0:35 cos y.

C2
C1

coil
T2 F1
F2
rod
num
mi
alu T1

magnet
coil

Fig. 7. Different relative positions of the EMAT’s coil and magnet excite and detect distinct modes of the rods: compressional vibrations, C1 and C2;
torsional waves T1 and T2; F1 and F2 do so for flexural vibrations.

Fig. 1. We conclude this discussion about the EMATs mentioning that they can measure the complete velocity vector on the
conductors’ surface and that acting as exciters they can produce an arbitrary vector force.

4. Results for frequencies below the critical frequency

We shall now use the different configurations of Fig. 7 to excite and detect all normal modes of a cylindrical rod and
compare the experimental results with the values obtained for free–free boundary conditions with the method described
in Section 2 for flexural vibrations and in Ref. [27] for the two other modes. We are mainly interested in the comparison of
measured values of the normal-mode frequencies and wave amplitudes with the TBT predictions. In Fig. 8 a typical
frequency spectrum is given for the range 20–35 kHz, which is below the critical frequency f c  150 kHz. All the peaks
have been identified and associated, one to one, with either compressional (C), torsional (T), or flexural (F) normal modes.
One can observe that some lines of spectra (a) and (c) are also seen, with a significant intensity, in spectrum (b). This is due
to the fact that the transversal component of the compressional wave is excited and registered by the flexural
configuration F1 . Something similar happens with the torsional states marked T in the figure.
To further identify the flexural normal modes we make two remarks. First, all flexural lines in Fig. 8(b) appear as
doublets [4]. This is due to the fact that the cross-section of the rod is not a perfect circle, so two principal values of the
moment of inertia are present in the eigenvalue equations of Section 2. Therefore, two orthogonal polarized eigenmodes,
with slightly different frequencies, are excited. The differences in frequency for the doublet lines observed in the
experiment are consistent with the irregularities that we measured in the rod radius. This is the main reason for using a
cylindrical rod in this part of our work.
Second, to distinguish flexural from compressional modes, we note that the wave amplitude for the latter is constant
along the free cross-section, whereas for flexural vibrations the amplitude changes sign along a diameter. We have verified
experimentally that this is indeed the case.
Once we were certain that a peak corresponds indeed to a flexural normal mode, we measured the corresponding wave
amplitude sliding the detector along the rod and registering its position, using the experimental setup of Fig. 2. Six
 Author's personal copy

A. Dı́az-de-Anda et al. / Journal of Sound and Vibration 331 (2012) 5732–5744 5739

Fig. 8. In (a) and (c) the picket-fence experimental spectra corresponding, respectively, to compressional (C) and torsional (T) waves are given. In (b) the
doublets related to flexural (F) vibrations are shown. Note that in spectrum (b) states C and T also appear.
Amplitude (arb. units)

0 L/2 L0 L/2 L

Fig. 9. Wave amplitudes for N ¼ 4,6,8,16,30 and 40 nodes for the rod used in Fig. 8 with a length L¼1.000 m. The experimental wave amplitudes are
given by the dots. The continuous lines correspond to the theoretical prediction equation (12).

examples of normal modes are shown in Fig. 9; the dots correspond to experimental measurements and the continuous
lines are obtained from TBT.
To test the dispersion relation we have to obtain k þ and k ¼ ik as a function of the frequency f (see Eq. (11)). The values
of the wavenumbers k þ ,n and kn and frequencies fn can then be obtained from the n-th normal-mode wave amplitude and
frequency, respectively. Disregarding the growing solution, the experimental wave amplitudes below f c can be fitted by

wðzÞ ¼ A1 cosðk z þ fÞ þ A2 ek þ z , (26)

where A1 and A2 are constants to be determined. Notice that only a portion of the wave amplitude is used to get both k þ and
k  ; this is due to the following fact: when both EMAT’s are close to each other, the magnetic field of the EMAT used as exciter
saturates the coil of the EMAT used as detector yielding an interference offset that depends on the distance between both
EMAT’s. The nonlinear fitting (26) was used to obtain the dispersion relation and it was tested with numerical wave
amplitudes of both Bernoulli–Euler and Timoshenko equations. In the case of the wave amplitude with 10 nodes the
nonlinear fitting gives the theoretical wavenumbers with five digits (of precision). The measurements corresponding to the
first 2 cm were erased to avoid artifacts due to the changes of the eddy currents near the free-ends of the rod. In Fig. 10 we
show that the wavenumbers given by Eq. (11) agree with those obtained with the nonlinear fitting of the experimental wave
amplitudes, the error ranging from 0.001 percent to 1.57 percent. The maximum error is obtained for the lowest modes since
the growing exponential is then nonnegligible in the experimental wave amplitude and is not included in the nonlinear
fitting of Eq. (26). Nevertheless, an excellent agreement is obtained for f of c with the prediction of TBT.
 Author's personal copy

5740 A. Dı́az-de-Anda et al. / Journal of Sound and Vibration 331 (2012) 5732–5744

5. Results above the critical frequency

We dedicate this section to compare TBT and the finite element method (FEM) calculations with experimental results
for frequencies beyond the critical frequency for a rectangular beam which shows a simpler spectrum than what is
obtained for a cylindrical rod. We show a very good coincidence in these three cases, so one can conclude that the second
spectrum of Timoshenko indeed exists, at least for the free–free boundary conditions case.
In Fig. 11(a) the configuration that we have used to excite and detect the bending modes of the beam is shown: First, the
EMAT on the right-hand side used as exciter, applies a force perpendicular to the coil axis. Since this exciter is located
nonsymmetrically with respect to the face of the beam (that is, close to a corner), it excites flexural waves on both the soft
and hard sides of the beam, as well as compressional waves. Second, the EMAT on the left-hand side of the beam, due to the
configuration of the magnets, detects the flexural modes on the hard side of the beam. Notice that in spite of the fact that the
detector configuration is set to detect flexural waves the compressional modes, easier to be excited, are also detected.
To determine which resonant peaks of the spectrum of Fig. 12(a) correspond to compressional resonances and which to
flexural ones, the compressional resonances were measured using the configuration shown in Fig. 11(b). The correspond-
ing compressional spectrum is shown in Fig. 12(b). The resonant bending spectrum of the stiff plane of the beam is
obtained subtracting the compressional peaks given in the lower part of Fig. 12 from the spectrum given in the upper part
of the same figure.
To compare with the experimental and TBT predictions, a numerical analysis was performed using the FEM platform
Salome-Meca [28]. The beam was modeled using 3D Tetrahedron (Netgen). The simulations were performed with the
physical parameters given in Fig. 11. The TBT as well as the numerical analysis use, as input, several physical parameters.
They were obtained from the experiment as follows. On the one hand, the shear and Young’s moduli were obtained from the
torsional and compressional velocity measuring the lowest torsional and compressional modes of the beam, respectively. The
value of the Timoshenko shear coefficient, on the other hand, was obtained from the best fit to the critical frequency.

120

100
wavenumber (1/m)

80

60

40

20

0
0 10 20 30
frequency (kHz)

Fig. 10. Experimental values obtained for k þ (squares) and k  (triangles) compared with the TBT values, respectively, the solid and dashed lines. The
intermediate dot-dashed line corresponds to the Bernoulli–Euler theory.

N coil
S S
coil
N
aluminum bar Y aluminum bar
Y
magnet
coil coil
X Z X Z
S
N
N S
magnet
S N magnet S N

Fig. 11. Experimental setup used to excite and detect different elastic waves. The configuration shown in (a) is used to measure flexural waves on the
stiff plane of the beam; the EMAT at the left-hand side of the beam corresponds to the detector while the EMAT at the opposite side of the beam
generates bending waves in both the flexible and stiff planes as well as compressional modes. The direction of the force applied by the exciter to the
beam is along the Z-axis. To discriminate the compressional modes from the spectrum, they were measured using the configuration shown in (b). The
aluminum beam of rectangular cross-section has dimension: L ¼0.5 m, height a ¼0.0252 m and width b ¼0.0504 m and the same elastic constants as in
Fig. 2. The same parameters were used in the FEM simulations.
 Author's personal copy

A. Dı́az-de-Anda et al. / Journal of Sound and Vibration 331 (2012) 5732–5744 5741

amplitude (arb. units)

C
A
B

(a)

(b)

20 25 30 35 40 45
frequency (kHz)

Fig. 12. Resonant spectrum measured for the beam of Fig. 11. The upper spectrum was measured with the setup of Fig. 11(a) while the lower one was
measured with the setup of Fig. 11(b). The stiff bending resonances (such as A, B and C) are those of the upper spectrum that do not appear in the
compressional (lower) spectrum.

45

C C
40 B B
frequency (kHz)

35

fc

30

25

A A

20

EXP TBT FEM

Fig. 13. Flexural stiff spectrum (left column) measured with the EMATs for the beam of Fig. 11. The central column corresponds to the TBT prediction
while the right column was obtained using the finite element method platform Salome-Meca. We used k ¼ 0:85 for Timoshenko’s shear coefficient. Above
the critical frequency all spectra show doublets with levels very close to each other in some cases (made apparent with dotted lines). The face
deformation of modes A, B and C are shown in Fig. 15.

In Fig. 13 we show the comparison between the spectrum of the measured bending modes with the spectrum
calculated numerically using Salome-Meca as well as with the prediction of TBT. The cut-off frequencies are
f c,F ¼ 63:70 kHz and f c,S ¼ 31:850 and correspond to the vibrations in the more flexible (F) and in the stiff (S) planes,
respectively. They were calculated using Eq. (9). Below the critical frequency the FEM calculations and TBT predictions
agree with the experiment with errors less than  2 percent. Above the critical frequency all spectra show doublets but
 Author's personal copy

5742 A. Dı́az-de-Anda et al. / Journal of Sound and Vibration 331 (2012) 5732–5744

the predictions of TBT show errors again less than  2 percent. This figure demonstrates that the TBT correctly predicts
the spectrum above the critical frequency at least for free–free boundary conditions. One can also observe in Fig. 13 that
the FEM calculations agree with the experimental results. Our experimental and numerical results for the normal stiff
mode frequencies of the beam are summarized in Table 1. One can see that the error between experimental results and
TBT or FEM results are about 2 percent for the fundamental mode. To get a better approximation for the lowest flexural
mode one has to use the value of the compressional velocity ( ¼5090.23 m/s) for very low frequencies; this yields an error
of 0.25 percent for the lowest flexural mode but an error of about 2 percent for higher modes.
We have also measured with the experimental setup of Fig. 14 the wave amplitude of the bending modes along the X-
axis of the face Z¼0 shown in the same figure. In the lower panels of Fig. 15 we give the wave amplitudes for the three
states labeled A, B and C in Fig. 13. State A corresponds to a mode with frequency below the critical one, while wave
amplitudes B and C belong to a doublet above f c . In the upper part of Fig. 15, the cross-section wave amplitude calculated
with Salome-Meca for these three resonances are shown and, as before, the one labeled A corresponds to a frequency
below the critical frequency and those with labels B and C to a doublet above the critical frequency.
In Timoshenko beam theory the surface of the face is assumed to be plane. This is true for all amplitudes we have
measured below the critical frequency. The same result appears for all wave amplitudes calculated with Salome-Meca
below f c . Above the critical frequency the results are completely different. One amplitude of each doublet is plane but the
other is almost sinusoidal. This result was checked for several doublets and the results of the FEM agree with the
experiment. Nevertheless, the frequency values from TBT agree with experiment.

Table 1
Normal mode frequencies for the beam of Fig. 11 with free ends.

Mode Experiment (kHz) TBT (kHz) FEM (kHz)

1 1.0211 1.00062 (þ 2.01%) 1.0008 (þ 1.99%)

2 2.6594 2.60559 (þ2.02%) 2.6076 (þ1.95%)
3 4.8462 4.75908 (þ1.80%) 4.7664 (þ1.65%)
4 7.3878 7.27210 (þ1.57%) 7.2893 (þ1.33%)
5 10.163 10.0153 ( þ1.45%) 10.047 (þ 1.14%)
6 13.082 12.8989 (þ1.40%) 12.950 (þ 1.01%)
7 16.081 15.8629 (þ1.36%) 15.938 (þ0.89%)
8 19.1335 18.8651 (þ1.40%) 18.966 (þ0.88%)
9 (A) 22.1699 21.8731 (þ1.34%) 22.000 (þ 0.77%)
10 25.1638 24.8548 (þ1.23%) 25.004 (þ 0.64%)
11 28.0593 27.7656 (þ1.05%) 27.923 (þ0.49%)
12 30.6109 30.4859 (þ0.41%) 30.570 (þ 0.13%)
fc 31.850
13 31.7895 32.4096 (  1.94%) 31.954 ( 0.51%)
14 32.3453 32.6607 (  0.97%) 32.453 ( 0.32%)
15 34.2301 34.4741 ( 0.70%) 34.245 ( 0.03%)
16 34.5053 34.9106 (  1.16%) 34.597 ( 0.26%)
17 37.1713 37.4119 ( 0.64%) 37.196 ( 0.06%)
18 37.1940 37.5655 ( 0.99%) 37.222 ( 0.07%)
19 (B) 40.1110 40.5102 (  0.99%) 40.136 ( 0.05%)
20 (C) 40.3298 40.6225 ( 0.72%) 40.339 ( 0.01%)
21 43.1702 43.6314 ( 1.06%) 43.191 ( 0.04%)
22 43.5919 43.9734 ( 0.87%) 43.611 ( 0.03%)
23 46.2782 46.8812 ( 1.29%) 46.310 (  0.06%)
24 46.9439 47.3915 ( 0.94%) 46.973 ( 0.05%)

coil
Y
magnet
X Z

magnet
coil
aluminum bar

Fig. 14. Experimental setup used to measure the wave amplitude on the face of the beam. The detector (left-hand side) was moved along the X-axis. The
position of the exciter (right-hand side) is fixed.
 Author's personal copy

A. Dı́az-de-Anda et al. / Journal of Sound and Vibration 331 (2012) 5732–5744 5743
FEM
amplitude (arb. units)

amplitude (arb. units)

amplitude (arb. units)

EXP

0 1 2 3 4 5 0 1 2 3 4 5 0 1 2 3 4 5
position (cm) position (cm) position (cm)
mode A mode B mode C

Fig. 15. Wave amplitudes of a mode A below the critical frequency and of a doublet above f c , modes B and C; the frequencies are given in Table 1. The
measurements (lower panels) were done moving the detector along the X-axis using the experimental setup of Fig. 14. The upper panels are the results
obtained with Salome-Meca using the same parameters of Fig. 11. Note that the deformations calculated with FEM agree with the experimental
measurements shown in the lower panels. The measurement at the borders of the beam are affected by an edge effect of the order of the size of the
detector, which is approximately 0.25 cm, and therefore they are not shown.

6. Conclusions

In this paper we have studied, from the theoretical and experimental points of view, the flexural normal-mode
frequencies and wave amplitudes of cylindrical rods and rectangular beams for free–free boundary conditions. We
consider two frequency regimes: frequencies less than the critical one f c and for f beyond f c . For the first regime, TBT is
much more than a beautiful engineering theory, it is also a very accurate one. This fact allowed us to test our experimental
procedure. We obtain an excellent agreement between TBT and the measurements for frequencies, wave amplitudes and
the dispersion relation. These, as well as the results for the second regime, were corroborated with 3-D FEM calculations
using the Salome-Meca platform. Therefore, the results presented in Section 4 serve to validate our experimental
procedure. We have also performed experiments for the second regime f 4 f c , experiments which have a wider perspective
than what has been achieved up to now. For a rectangular beam the flexural modes, and in this frequency regime, appear
always as doublets.1 Since we can also measure the deformation of the free face of the beam, we have proven that one of
the states of the doublet shows a flat surface while the other state has a sinusoidal shape. This is consistent with the exact
Rayleigh–Lamb solution for the transverse field shape of the second-branch mode. For f o f c , on the other hand, all normal
modes show a flat free face, an assumption made when deriving TBT equations. These facts have also been observed in the
experiments. In any case, we can assure that the second spectrum of Timoshenko exists, at least for free–free boundary
conditions. The TBT predictions are also very accurate beyond about 50 percent of the critical frequency; this is consistent
with Ref. [25] in which the TBT dispersion relation, slightly modified by the inertia correction factor as well as the shear
correction factor, agrees with the exact Rayleigh–Lamb prediction up to frequencies of the order of twice the critical
frequency. Although the TBT assumes a flat surface, even for frequencies beyond the critical one, it yields very precise
results (with an error less than 2 percent) for the normal-mode frequencies. Something similar happens with the
Rayleigh–Ritz method that produces extremely accurate results for the normal-mode frequencies with very poor trial
wave functions.
Although the results presented here are conclusive about the existence of the second TBT spectrum, there are two open
questions left. The first one, related to the normal-mode frequencies, is to find out the upper limit for which the
Timoshenko beam theory is valid. That work requires to study much higher frequencies to distinguish the TBT resonances

1
We should note that the origin of these doublets is different from what was obtained in Section 4 for the cylindrical rod. We have also measured the
spectrum for f 4f c for a rod which is almost perfectly circular; the spectrum consists of quadruplets.
 Author's personal copy

5744 A. Dı́az-de-Anda et al. / Journal of Sound and Vibration 331 (2012) 5732–5744

from the three-dimensional spectrum. The second question is related to the dispersion relation. To obtain it the wave
amplitudes above the critical frequency must be measured.

Acknowledgments

Author R.A.M.S. was supported in part by DGAPA-UNAM under project IN-111311, G.M. by project IN-119509-3 and
CONACYT project 82474. A.M., J.F. and L.G. by project IN-113011.

References

[1] K.F. Graff, Wave Motion in Elastic Solids, Dover, New York, 1991.
[2] G.B. Warburton, The Dynamical Behavior of Structures, Pergamon Press, Oxford, 1976.
[3] S.P. Timoshenko, On the correction of shear of the differential equation for transverse vibration of prismatic bars, Philosophical Magazine 41 (1921)
744–746.
[4] A. Dı́az-de-Anda, A. Pimentel, J. Flores, A. Morales, L. Gutiérrez, R.A. Méndez-Sánchez, Locally periodic Timoshenko rod: experiment and theory,
Journal of the Acoustical Society of America 117 (2005) 2814–2819.
[5] B. Geist, J.R. McLaughlin, Double eigenvalues for the uniform Timoshenko beam, Applied Mathematical Letters 10 (1997) 129–134.
[6] O.M. O’Reilly, J.S. Turcotte, Another mode of vibration in a Timoshenko beam, Journal of Sound and Vibration 198 (1996) 517–521.
[7] K.T. Chan, X.Q. Wang, R.M.C. So, S.R. Reid, Superposed standing waves in a Timoshenko Beam, Proceedings of the Royal Society of London A 458 (2002)
83–108.
[8] N.G. Stephen, The second spectrum of Timoshenko beam theory—further assessment, Journal of Sound and Vibration 292 (2006) 372–389.
[9] A. Bhaskar, Elastic waves in Timoshenko beams: the ‘lost and found’ of an eigenmode, Proceedings of the Royal Society A 465 (2009) 239–255.
[10] G.R. Bhashyam, G. Prathap, The second frequency spectrum of Timoshenko beams, Journal of Sound and Vibration 76 (1981) 407–420.
[11] N.G. Stephen, S. Puchegger, On the valid frequency range of Timoshenko beam theory, Journal of Sound and Vibration 297 (2006) 1082–1087.
[12] R.A. Méndez-Sánchez, A. Morales, J. Flores, Experimental check on the accuracy of Timoshenko’s beam theory, Journal of Sound and Vibration 279
(2005) 508–512.
[13] R.W. Traill-Nash, A.R. Collar, The effect of shear flexibility and rotary inertia on the bending vibrations of beams, Quarterly Journal of Mechanics and
Applied Mathematics 6 (1953) 182–222.
[14] R.A. Anderson, Flexural vibrations in uniform beams according to the Timoshenko theory, Journal of Applied Mechanics 20 (1953) 504–510.
[15] C.L. Dolph, On the Timoshenko theory of transverse beam vibrations, Journal of Applied Mathematics 12 (1954) 175–187.
[16] T.C. Huang, The effect of rotatory inertia and of shear deformation on the frequency and normal mode equations of uniform beams with simple end
conditions, Journal of Applied Mechanics 28 (1961) 579–584.
[17] D. Young, in: W. Flügge (Ed.), Handbook of Engineering Mechanics, 4th revised ed. McGraw-Hill, New York, 1962. (Chapter 61).
[18] B. Downs, Vibration of a uniform, simply supported Timoshenko beam without transverse deflection, Journal of Applied Mechanics 43 (1976)
671–674.
[19] B.A.H. Abbas, J. Thomas, The second frequency spectrum of Timoshenko beams, Journal of Sound and Vibration 51 (1977) 123–137.
[20] I. Hathout, H. Leipholz, K. Singhal, Sensitivity of the frequencies of damped Timoshenko beams, Journal of Engineering Sciences-University of Riyadh 6
(1980) 113–121.
[21] N.G. Stephen, The second frequency spectrum of Timoshenko beams, Journal of Sound and Vibration, 80 (1982) 578–582.
[22] V.V. Nesterenko, A theory of transverse vibration of the Timoshenko beam, PMM Journal of Applied Mathematics and Mechanics 57 (1993) 669–677.
[23] A.M. Chervyakov, V.V. Nesterenko, Is it possible to assign physical meaning to field theory with higher derivatives? Physical Review D 48 (1993)
5811–5817.
[24] A. Morales, L. Gutiérrez, J. Flores, Improved eddy current driver-detector for elastic vibrations, American Journal of Physics 69 (2001) 517–522.
[25] S.V. Sorokin, C.J. Chapman, A hierarchy of rational Timoshenko dispersion relations, Journal of Sound and Vibration 330 (2011) 5460–5473.
[26] M. Levinson, D.W. Cooke, On the two frequency spectra of Timoshenko beams, Journal of Sound and Vibration 84 (1982) 319–326.
[27] A. Morales, J. Flores, L. Gutiérrez, R.A. Méndez-Sánchez, Compressional and torsional wave amplitudes in rods with periodic structures, Journal of the
Acoustical Society of America 112 (2002) 1961–1967.
[28] The program Salome-Meca is a GNU integration platform based on FEM with GNU Generic Public License (GPL). It was developed by Électricité de
France (EDF) and tested in many instances. /http://www.code-aster.org/S.

View publication stats