Journal of Advanced Mechanical Design, Vol.1, No.

5, 2007

Systems, and
Manufacturing
Vibration Suppression of Railway Vehicle
Carbody with Piezoelectric Elements
(A Study by Using a Scale Model of
Shinkansen)∗
Tadao TAKIGAMI∗∗ , Takahiro TOMIOKA∗∗ and Joel HANSSON∗∗∗
∗∗ Vehicle Noise and Vibration Lab., Vehicle Structure Technology Division,
Railway Technical Research Institute
2–8–38 Hikari-Cho, Kokubunji, Tokyo 185–8540, Japan
E-mail: takigami@[Link], tomioka@[Link]
∗∗∗ Royal Institute of Technology, Department of Aeronautical and Vehicle Engineering (Formerly)
SE100 44, Stockholm, Sweden

Abstract
Piezoelectric elements, which are electrically connected to external shunt circuits, are
introduced to suppress bending vibration of a scale model of a railway vehicle. The
authors proposed two types of shunt circuits aiming at practical use. One circuit is
equivalent to an inductor and a resistor in series, and the other a negative capacitor and
a resistor, and both circuits are designed for use under conditions of large vibration
amplitude that causes high voltage generated by the elements. Results of excitation
tests show that the first bending mode of the scale model can be suppressed success-
fully. Moreover, the circuit including the negative capacitor can also reduce vibration
associated with higher eigen modes simultaneously with the first mode.

Key words : Railway, Piezoelectric Element, Ride Comfort, Bending Vibration, Shunt
Circuit, Negative Capacitor

1. Introduction
Recent railway vehicles are being made increasingly lighter in weight. This contributes
to increase of train speed, the saving of energy, etc. On the other hand, there are cases in which
the vertical bending vibration of carbodies becomes conspicuous, mainly due to decreases in
their rigidity and damping. This vibration often occurs in the frequency range to which human
beings are very sensitive. Since the vibration causes passenger comfort to decline, effective
measures to suppress it are increasingly called for.
The goal of this study is to reduce the bending vibration of railway vehicle carbodies
by applying a vibration suppression technique that utilizes piezoelectric elements, such as
PZTs, shunted by an external electric circuit. This technique was originally presented in the
Ref. (1) , and have been extensively studied(2) . Since it has been mainly applied to small
structures(3)(4) , however, it is expected to deal with considerable problems when applying to
such large structures like railway vehicle carbodies. Therefore, the authors have presented
investigations to reduce the bending vibration of a primitive beam structure with length about
3m(5) as the primary step to realize the above goal. The experimental results showed that the
piezoelectric elements increased more effectively the loss factor of the structure than damping
layers that had been put into practical use. And then in this paper, this vibration suppression
technique has been applied to a 1:5 scale model of a Shinkansen (Japanese bullet train) with
∗ length about 5m.
Received 31 Jan., 2007 (No. T-04-0065)
Japanese Original: Trans. Jpn. Soc. Since the scale model is a larger and more complex structure than the beam structure, the
Mech. Eng., Vol.71, No.704, C (2005),
pp.1254–1262 (Received 21 Jan., 2004) followings are main problems to be solved;
[DOI: 10.1299/jamdsm.1.649] • Although the amount of the piezoelectric elements has to be increased, it is rather

649
Journal of Advanced Mechanical Design, Vol.1, No.5, 2007
Systems, and
Manufacturing
difficult to enlarge the dimensions of each element itself.
• Since the scale model as well as the actual railway vehicle carbody shows complex
vibration characteristics as a three-dimensional structure, it is difficult to determine arrange-
ments of the elements on the carbody.
• The shunt circuits connected to the elements should be designed taking account of use
under the conditions of large vibration amplitudes that cause high voltage of the piezoelectric
element, aiming at practical use.
2. Modeling
2.1. Outline of the Model
Figure 1 shows a photograph of a 1/5 scale model of a Shinkansen vehicle carbody. The
length of this model is 4.9m; the width, 0.67m; the height, 0.6m, and the mass mm = 292kg.
The model is suspended by air springs at four points corresponding to the positions of an
actual vehicle. An electrodynamic shaker is applied through a rod at one end of the carbody
to vertically excite the vehicle. The excitation force can be measured by a road cell attached
between the rod and the carbody, and the response vibration by acceleration pickups stuck on
the surface of the carbody.

Fig. 1 Photograph of a scale model

2.2. Equation of Motion

Figure 2 shows the analytical model of the scale model. x, y, z axes are related to the
longitudinal, horizontal (width) and vertical (height) directions of the carbody, respectively.
The origin is located at the left end, or the position indicated as o. The carbody is modeled
as a simple Euler-Bernoulli beam, for the present, in consideration of only vertical vibration,
i.e., translation, pitching, and bending modes. The rolling motion is neglected. In the Fig.2,
Lm , wm , hm and mm denote the length, width, height, and mass of the carbody, respectively. xa
and La are the distance from the end of the carbody to the center of the nearer air spring, and
a half of longitudinal interval of two air springs, respectively. Em Im is the bending rigidity of
the beam, and Jm the moment of inertia related to the pitching motion. ka and ca denote spring
and damper coefficients corresponding to two air springs in parallel, respectively.

Fig. 2 Analytical model of the scale model

650
Journal of Advanced Mechanical Design, Vol.1, No.5, 2007
Systems, and
Manufacturing
Piezoelectric elements (PZTs), which are indicated in gray color in Fig.2, are attached
between x = x1 and x = x2 in the longitudinal direction. L p , w p and t p are length, width and
thickness of the elements, respectively. Although eight elements, two rows of four pieces,
are bonded to the carbody separately as shown in Fig.1 in actuality, they are modeled as a
piece of element in the following formulation. Therefore, L p and w p are defined as the total
values. E p is Young’s modulus, and I p the moment of inertia of area of the element. Note
that I p is not the value of the element alone, but should be calculated by taking account of
the perpendicular distance from the neutral axis of the carbody to the vertical center of the
element. It is assumed that the neutral axis of carbody is parallel to the x axis and located at
z = 0.3m that is a half of the height of the carbody.
Now, the vertical displacement of the carbody is separated into components of space and
time as follows:

w(x, t) = φz (x)qz (t) + φ p (x)q p (t) + φ1 (x)q1 (t) (1)

where φ(x)s are the modal shape functions and q(t)s denote modal displacements. Subscripts
z, p and 1 represent vertical translation, pitching, and the first bending modes, respectively.
Namely, φz (x) = 1, φ p (x) = 2x/Lm − 1, and φ1 (x) is the shape function corresponding to a
free-free beam. Higher bending modes of vibration are neglected in Eq. (1).
The equation of motion considering the vertical excitation force at the end of the carbody
can be derived as follows:

M q̈ + C q̇ + Kq = b1 f + b2 M p , (2)

where
⎡ ⎤ ⎡ ⎤ ⎡ ⎤
⎢⎢⎢ mm 0 0 ⎥⎥⎥ ⎢⎢⎢ 1 ⎥⎥⎥ ⎢⎢⎢ 0 ⎥⎥⎥
⎢ ⎥⎥⎥ ⎢ ⎥⎥⎥ ⎢ ⎥⎥⎥
M = ⎢⎢⎢⎢⎢ 0 Jm 0 ⎥⎥⎥ , b1 = ⎢⎢⎢⎢⎢ 1 ⎥⎥⎥ , b2 = ⎢⎢⎢⎢⎢ 0 ⎥⎥⎥ ,
⎣ ⎦ ⎣ ⎦ ⎣ ⎦
0 0 1 φ1 (Lm ) φ1 (x1 ) − φ1  (x2 )


⎡ ⎤
⎢⎢⎢ 2ca 0 2ca φ1 (xa ) ⎥⎥⎥
⎢ ⎥⎥⎥
C = ⎢⎢⎢⎢⎢ 0 2ca La 2 0 ⎥⎥⎥ ,
⎣ ⎦
2ca φ1 (xa ) 0 2(ζ1 ω1 + ca φ1 2 (xa ))
⎡ ⎤
⎢⎢⎢ 2ka 0 2ka φ1 (xa ) ⎥⎥⎥
⎢ ⎥⎥⎥
K = ⎢⎢⎢⎢⎢ 0 2ka La 2 0 ⎥⎥⎥
⎣ ⎦
2ka φ1 (xa ) 0 ω1 2 + 2ka φ1 2 (xa )

and

q = [qz q p q1 ]T .

f is the excitation force by the shaker, and M p is the bending moment generated by the piezo-
electric element. The details of M p are described in the following section. ω1 and ζ1 are the
natural angular frequency and damping ratio of the first bending mode, respectively. Note that
the descriptions ()˙ and () represent the partial derivative with respect to time t and the spacial
variable x, respectively. When the beam is supported in the free-free condition, the angular
frequency can be written as
 2
4.73 Em Im
ω1  , (3)
Lm ρm Am

where ρm Am is the mass per unit length of the carbody.

651
Journal of Advanced Mechanical Design, Vol.1, No.5, 2007
Systems, and
Manufacturing
2.3. Piezoelectric Element
2.3.1. Generated Voltage The voltage v p generated by the piezoelectric element due to
the vibration of carbody is proportional to the average longitudinal strain in the element. When
the carbody is considered as a simple beam(6) ,
Ks 
vp = (w (x1 ) − w (x2 )), (4)
Cp
where

K s = d31 w p E p zc .

C p and d31 are the capacitance and the equivalent piezoelectric constant of the element, respec-
tively. zc represents the vertical distance from the neutral axis of the carbody to the vertical
center of the element.
2.3.2. Bending Moment The piezoelectric element causes the distributed bending mo-
ment M p of the carbody, which consists of two components, M p e and M p m . M p e stands for
the moment related to the inverse piezoelectric effect. When the voltage va is applied to the
element, the bending moment due to this effect is as follows(6) :

M p e = −Ka va , x1 ≤ x ≤ x2 , (5)

where

Ka = d31 w p E p zc .

On the other hand, M p m is the effect of mechanical characteristics of the element and
stands for additional rigidity to the carbody. In this paper, the moment is approximated as
follows:

M p m = −Km (w (x1 ) − w (x2 )), x1 ≤ x ≤ x2 , (6)

where
E p Ip wptp3
Km = and I p = + w p t p zc 2 .
Lp 12
2.3.3. Shunt Circuit A shunt circuit is connected to two electrodes of the piezoelectric
element. Several types of shunt circuits have been proposed and the most popular one is L−R
circuit, which is an inductor and a resistor in series, shown in Fig.9(a). The impedance of
this circuit is Z(s) = Ls + R. Note that capital letters followed by (s) represent the Laplace
Transform of corresponding variables or transfer functions. When the voltage v p due to the
strain is generated in the element, it is divided into two components by C p and Z. Therefore,
the transfer function between the voltage generated by the element v p and the voltage applied
to the shunt circuit, which is the same as the voltage across the element va , can be obtained as:
LC p s2 + RC p s
G L−R (s) = . (7)
LC p s2 + RC p s + 1
Since this circuit introduces an electrical resonance together with the inherent capacitance
of the element, the mechanical (vibration) energy as well as the electric energy is dissipated
effectively.
2.3.4. Integrated model An integrated model, including the carbody, the piezoelectric
element and the shunt circuit is displayed as a block diagram in Fig.3. The part hatched in
gray represents the characteristics of the piezoelectric element. The upper and lower parts
show the mechanical and electric relation, respectively.

652
Journal of Advanced Mechanical Design, Vol.1, No.5, 2007
Systems, and
Manufacturing

Fig. 3 Block diagram of integrated model

3. Preliminary Excitation Tests

3.1. Excitation Conditions
In order to investigate the vibration characteristics of the scale model without piezoelec-
tric elements, preliminary excitation tests are carried out. The carbody is excited by a shaker at
the end point with band random signal input that has an uniform frequency component in the
range 5-100Hz. Acceleration at several points on the floor, roof, and side panel is measured
together with the excitation force at sampling periods of 5ms.

3.2. Vibration Characteristics

Solid lines in Fig.4 show frequency response functions (FRFs) between excitation force
and acceleration measured on the floor at x = 2.4 (near the center) and x = 0.9 (near one of
the suspensions). Major peaks of FRFs appear around 7, 8, 36, 58, 76 and 85Hz. Dashed lines
are FRFs calculated based on the analytical model described in the previous section.

Fig. 4 Measured and calculated FRFs between excitation force and acceleration on floor

Modal vibration analysis(7) is then applied to identify natural frequencies and the vibra-
tion shapes related to the FRF peaks. Identified modal characteristics are shown in Fig.5. Thin
lines link the original positions of acceleration pickups, and thick ones express the vibration
shapes corresponding to eigen modes of vibration. Numbers below the shapes are natural
frequencies. Modes (a) and (b) indicate the rigid motion of the carbody, and are related to
vertical translation and pitching motion, respectively. The vibration shape shown in the mode
(c) is similar to that of the first bending mode of a simple Euler-Bernoulli beam. The modes
(d)-(f) indicate complex shapes, which cannot be modeled as a simple beam.

3.3. Strain Measurements

Since the generated voltage of the piezoelectric element due to the strain in the element
affects the performance of vibration suppression, strain distribution of the carbody has been

653
Journal of Advanced Mechanical Design, Vol.1, No.5, 2007
Systems, and
Manufacturing

Fig. 5 Identified natural frequencies and vibration shapes

Fig. 6 Arrangement of strain measurement points Fig. 7 Results of strain measurements

(Vertical cross section at longitudinal midpoint)

investigated. Measurement points are arranged as shown in Fig.6 on the vertical cross section
at the longitudinal midpoints of the carbody. Circles () in Fig.7 show resultant strains when
the carbody is excited by the sinusoidal wave with the frequency of f1 = 36.3Hz that is the
natural frequency of the first bending mode of the carbody.
The solid blue line in the Fig.7 is the estimated strain S id assuming that the carbody can
be modeled as the simple Euler-Bernoulli beam;
∂2 w1 (xm ) φ (xm )
S id  dz = dz 1 w1 (xm ), (8)
∂x 2 φ1 (xm )
where dz is the distance from the neutral axis of the carbody to the measurement point, xm =
Lm /2 and w1 (x) = φ1 (x)q1 . Equation (8) implies that the theoretical strain can be estimated if
the displacement w1 is calculated by integrating the acceleration measured at the center of the
carbody.
The neutral axis of both measured (circle) and theoretical (blue line) strains are located
around z = 0.3m. Measured strains, however, are smaller than estimated values. And although
the distances from the neutral axis to the floor and to the roof are approximately the same, the
strains measured on the roof tend to be larger than that on the floor.
The results of the strain measurements show that it is not appropriate to model the car-
body as the simple beam from the viewpoint of strain distribution, while the vibration shape
of mode (c) is very similar to that of the first bending mode of the beam. It is because the
carbody of the scale model is actually a 3-dimensional structure and is composed of some
shells such as the roof, floor, side panels and (longitudinal) ends, rather than a beam. In order
to express each deformation of such components, the more precise model should be derived.
Since that is not the focus of this study, however, the analytical model derived in the previous
section has been a little bit modified. To be concrete, the variable zc used in Eqs. (4), (5) and
(6), has been modified by multiplying an appropriate coefficient (< 1) in order to compensate
for the difference between measured and theoretical strains. A black line in the Fig.7 shows
the modified theoretical strain.
4. Simulation
Numerical simulations are carried out to estimate the performance of vibration suppres-

654
Journal of Advanced Mechanical Design, Vol.1, No.5, 2007
Systems, and
Manufacturing
sion with the piezoelectric element and shunt circuits. The size of the piezoelectric element
used in the excitation tests is then determined based on the simulation results, aiming at re-
ducing the FRF gain at the first bending mode of vibration (around 36Hz) by 50%.
FRFs between the excitation force and acceleration at the floor center around the natural
frequency of the first bending mode, with and without the shunt circuit, are compared in Fig.8.
Lines with legends “Short” and “Open” are FRFs without a shunt circuit. Electrodes of the
piezoelectric elements are connected directly in the case of “Short,” and disconnected for
“Open.” The result with the L−R shunt circuit is shown with a red line. The optimal values of
the inductor and the resistor are determined based on the Ref. (3), namely as L = 71.8H and
R = 769Ω, respectively.
The L−R circuit suppresses the vibration, and the FRF peak value at 36.3Hz are reduced
by 53% compared with the short-circuited case.

Fig. 8 Simulated FRFs with/without shunt circuit

5. Design of Shunt Circuit

5.1. Background
In the previous section, the optimal value of the inductance L = 71.8H has been derived.
Such a value of inductance is too large to be realized by using a passive electric component,
i.e, a coil. In order to solve this problem, a technique to compose a synthetic inductor in-
cluding operational amplifiers (opamps)(3) , or a generated immittance converter (GIC)(8) are
widely used to produce large inductance. Therefore, the authors made use of such a syn-
thetic inductor in the previous study(5) . Since opamps are driven by supply voltage, typically
±15V, however, such synthetic circuits do not function properly, or may possibly be damaged,
when the amplitude of vibration becomes larger and the resultant voltage generated by the
piezoelectric element becomes higher.
This may be a critical problem when the piezoelectric element is attached on an actual
carbody of a railway vehicle running on the track. Therefore, it is necessary to design the
shunt circuit for use under the condition of high voltage generated by the piezoelectric element
subjected to large vibration amplitude, aiming at practical use.
In this paper, a new implementation method of shunt circuits, suitable for our objectives,
is studied. The details of the construction and features of the circuit are discussed below.

5.2. Virtual L− R Circuit Including Band limited Differentiating Circuit

When an inductor L and a resistor R are connected in series, as shown in the Fig.9(a), a
common current i flows through L and R. The applied voltages of the resistance vR and the
inductance vL can be expressed as follows;
VR (s)
VR (s) = RI(s), VL (s) = LsI(s) = Ls . (9)
R
Therefore, vL should be proportional to the time derivative of vR as well as the derivative of
the current i.

655
Journal of Advanced Mechanical Design, Vol.1, No.5, 2007
Systems, and
Manufacturing

Fig. 9 Implementation method of virtual L−R circuit

An implementation to realize the L−R circuit, reflecting the above relation is shown in the
Fig.9(b). “R” in the left end of this circuit is the resistor in the L−R circuit itself. An opamp
on the left side, OP1, is just an unity gain voltage buffer, and mainly used to protect the other
opamp, OP2, from undesired high voltage inputs. The OP2, together with a capacitor and two
resistors, functions as an inverting band limited differentiator. An ordinary opamp, driven by
±15V, is chosen as OP1, and only one high voltage opamp, OP2, driven by ±150V, is used in
this circuit. It is because the input voltage vR is generally much smaller than the output voltage
vL . Note that vR is expected to be about 1/20 of the vL at the natural frequency f1 , under the
conditions of the numerical simulation in the previous section.
The inductance L is now removed from the Fig.9(a). When the voltage vin is input into the
differentiating circuit, and connected to one of the electrodes of PZTs as shown with dashed
line in Fig.9(b), the voltage vL is replaced by −vout . The transfer characteristics between vin
and vout becomes as follows;
C1 R2 s
Vout (s) = − Vin (s). (10)
C1 R1 s + 1
The transfer function above, with the cutoff angular frequency ωL = 1/(C 1 R1 ), expresses the
ratio of the virtually realized impedance ZL to the resistance R as well as that of the voltage
vout to vin . Substituting s = jω to the Laplace Transform of the realized impedance ZL (s)
yields

ZL ( jω) = jωLeq + Req = jωLeq + αωLeq , (11)

where Leq , Req are an inductance and a resistance respectively, both of which are virtually
realized, and
C1 R2 R ω
Leq = , α = . (12)
1+α 2 ωL
If α  1, i.e., ω  ωL , Req at ω = ω1 is asymptotic in 0, and this fact indicates that it can be
considered that only the characteristics of an inductance,

Leq  C1 R2 R (13)

is produced. The value of Leq can be tuned by varying C1 and R2 . Thus, the characteristics of
L−R circuit can be synthesized by using this circuit without an actual inductor.
The effects of the band limitation in the differentiating circuit is then examined. Figure
10 shows the relationship between Leq , Req at frequency ω1 and the cutoff frequency in the
transfer function in Eq.(10), namely, the band limited differentiator. The horizontal axis in
this graph is the ratio of the natural frequency of the bending vibration to the cutoff frequency,
i.e., α1 = α|ω=ω1 = ω1 /ωL .
We can find from this graph that the realized inductance is almost constant and Leq 
71.8[H], which is the optimal value, in the range of α  1 (ωL  ω1 ), while Req is varying
with respect to α1 .

656
Journal of Advanced Mechanical Design, Vol.1, No.5, 2007
Systems, and
Manufacturing

Fig. 10 Leq , Req realized by band limited differenciating circuit

Figure 10 also shows the transfer gain of Eq.(10) at ω → ∞. Since the high transfer
gain of the differentiator may make the output voltage diverge and unstabilize the system, it
is desirable to set α1 larger to reduce the gain. On the other hand, Req increases with α1 , and
the total resistance in the shunt circuit possibly exceeds the optimal value, which deteriorates
the performance of the vibration suppression. Therefore, it is desired to set α1  1 from this
point of view. Thus, there exists a trade-off between the requirements of system stability and
performance improvement, and the cutoff frequency ωL should be determined by taking the
fact above into account.

5.3. Virtual Negative C− R Circuit Including Band limited Integrating Circuit

When the value of L in the L−R circuit is tuned so that the resonant frequency agrees with
the natural frequency of the first bending mode of the carbody f1 , the electrical resonance is
produced together with the inherent capacitance of the element C p , and the total impedance of
L and C p becomes 0 only at the frequency f = f1 . If L can be replaced by a capacitance with
a negative value −C and the negative capacitor is tuned at |C| = |C p |, it does not produce res-
onance, but cancels the inherent capacitance of the element directly, and the total impedance
becomes 0 in the all frequency range. Therefore, it prevents the deterioration of performance
of vibration suppression in the case where the natural frequency of mechanical structure shifts.
Additionally, multi-modal damping can be expected. Note that since the negative capacitor,
however, can not be installed by any passive electric elements, it must be synthesized with
some active components.
Although the idea of the negative capacitor itself has already been proposed in the Ref.
(9), the circuit including the negative capacitor, which is called “virtual negative C−R circuit”
in this paper, is implemented so as to prevent the problem of the high voltage in the circuit
discussed above. The circuit is obtained by replacing the combination of the inverting voltage
buffer and the differentiating circuit, shown in the Fig.9, with a non-inverting buffer and an
integrating circuit, shown in Fig.11.
The transfer characteristics of this circuit is
R2
Vout (s) = − Vin (s) (14)
C2 R1 R2 s + R1
with the cutoff angular frequency ωc = 1/(C2 R1 ). The equivalently realized impedance ZC in
this case becomes as follows;
1 1 1
ZC ( jω) = − − Reqc = − − , (15)
jωCeq jωCeq βωCeq
where
1 β2 R ω
= , β = . (16)
Ceq 1 + β2 C2 R1 ωc
Ceq stands for the virtually realized capacitance and

Ceq  C2 R1 /R (17)

657
Journal of Advanced Mechanical Design, Vol.1, No.5, 2007
Systems, and
Manufacturing

Fig. 11 Implementation method of virtual negative C −R circuit

when β 1. −Reqc is a resistance produced by the effect of the band limitation of the inte-
grating circuit, which decrease the total resistance in the shunt circuit.
The optimal value of the negative capacitor is −C p . It is known that, however, the system
becomes stable only when the negative capacitance −C satisfies |C| ≤ |C p |. Therefore, C is
tuned experimentally as small as possible in the excitation tests.

5.4. Advantages of Proposed Implementation

The shunt circuit realized by the proposed implementation method, including the virtual
impedance, has the following features:
• Arbitrary impedance, including the L−R and the negative C−Rcircuits, can theoretically
be realized by choosing the impedance in the input terminal (“R” in Figs.9 and 11) and the
transfer characteristics (the differentiating or integrating circuits in Figs. 9 and 11).
• Since the choice of such components to determine the total impedance is not unique,
one can make use of such redundancy when a shunt circuit is designed.
• The transfer function between input and output voltages can be implemented not only
by analog circuits, but also by a digital filter in a computer accompanied by the driver (voltage)
amplifier for piezoelectric element. Such an implementation enables to produce even the very
complex impedance characteristics.

6. Excitation Tests
6.1. Attachment of Piezoelectric elements
The piezoelectric elements, PZTs, are attached under the floor of the carbody, near the
longitudinal center. In actuality, eight pieces, two rows of four elements, are bonded on the
side beams instead of one element considered in the analytical model, because it is generally
difficult to manufacture such an element of very large size. The electrodes of PZTs are electri-
cally connected in parallel. One of the rows, which consists of four PZTs, is shown in Fig.1.
The size of each PZT is 155 × 40 × 3mm and the mass is about 0.14kg. Total mass ratio of
PZTs to the carbody is about 0.4 %.
In regard to the position of the piezoelectric elements, just the position of longitudinal
direction is important when the object is a simple beam. Furthermore, it is obvious that the
elements should be attached around the longitudinal center of the structure in order to suppress
the first mode of bending vibration. On the other hand, not only the position of longitudinal
but height and width directions to attach the elements have to be considered in this case.
According to the Fig.7, the strain measured on the roof is a little larger than that on the floor.
The authors firstly attached the elements on the roof around the measurement point #2, which
was the largest point of 9 points. The elements, however, hardly generated voltage because
the elements themselves locally restricted strain of the roof plate of the carbody. Therefore,
side beams, which have enough rigidity, are chosen as the positions to attach the elements.

658
Journal of Advanced Mechanical Design, Vol.1, No.5, 2007
Systems, and
Manufacturing

Fig. 12 FRFs calculated from experimental results

(At first bending mode frequency)

Fig. 13 FRFs calculated from experimental results

(In higher frequency ranges)

6.2. Excitation Test Results

After attaching PZTs, excitation tests are performed with short, open, L−R and negative
C −R circuits. The conditions, such as the excitation point, and input signal to the shaker, are
the same as those in the preliminary tests.
6.2.1. Results with Virtual L− R Circuit Figure 12 shows the FRFs between the exci-
tation force and the acceleration at the floor center around the first mode natural frequency.
The cutoff frequency of the band limited differentiating circuit is set as ωL = 100ω1 . The FRF
peak with the L−R circuit is reduced by about 30 % compared to the short-circuited case. In
addition, it is verified that the FRF peak moves as the value of L, namely, the transfer gain of
the differentiating circuit, although this phenomenon is not shown in the Fig.12. It indicates
that the designed shunt circuit functions as expected.
On the other hand, the measured performance of vibration suppression with this circuit
is poorer than the simulated result. The main cause of this difference is probably the effect of
undesirable inherent resistance in PZTs and the shunt circuit. That is implied by the fact that
antiresonance observed in the simulated FRF does not appear in the measured FRF.
6.2.2. Results with Virtual Negative C−R Circuit Red line in the Fig.12 shows the FRF
with the virtual negative C − R circuit. The cutoff frequency of the band limited integrating
circuit is set as ωc = ω1 /100. The FRF peak is reduced by about 60% compared to the short-
circuited case. The performance with this circuit is better than the case where the L−R circuit
is connected, because it is guessed that −Reqc in Eq.(15) canceles the inherent resistance and
total resistance in the circuit becomes close to the optimal value.
The performance of vibration suppression around the natural frequency of a higher modes
(e) and (f) are also illustrated in Fig.13. Since both FRF peaks are reduced simultaneously
with the FRF at the first mode frequency, it is clarified that the proposed circuit functions as
expected. The reasons the performance corresponding to the higher modes is a little poorer
than that of the first mode are that the positions of the elements are not determined optimally
for such higher modes, and that the elements have slightly frequency-dependent capacitances.

659
Journal of Advanced Mechanical Design, Vol.1, No.5, 2007
Systems, and
Manufacturing
7. Conclusions
In this paper, a vibration suppression technique is applied to reduce the bending vibration
of a scale model of a railway vehicle. The technique utilizes piezoelectric elements that are
electrically shunted by an external circuit. Summaries of this study are describes as follows;
• The scale model as well as actual railway vehicle carbodies should be considered as a
3-dimensional structure rather than a simple beam even if the vibration shape is very similar
to that of the simple beam, from the viewpoint of strain distribution.
• The positions of the piezoelectric elements should be determined by taking into ac-
count the strain distribution of the carbody with the elements attached, and but not that of the
carbody alone.
• A new method to implement L−R shunt circuit, including a band limited differentiating
circuit is proposed. This circuit is designed for use under the practical conditions.
• By replacing the differentiator above by the integrator, a negative C −R circuit, which
has ability to realize multi-modal damping, is implemented.
• Excitation test results show that the proposed circuits function as expected and that
the bending vibration of the carbody can be suppressed with the piezoelectric elements whose
total mass ratio to the carbody is as small as about 0.4 %.
As the next step of this study, the authors intend to investigate methods for attaching
the elements to actual vehicles and carry out performance confirmation testing (stationary
vibration test and running test) in order to put the new technique into practical use.

Acknowledgments
A part of this study was supported by the Program for Promoting Fundamental Transport
Technology Research from Japan Railway Construction, Transport and Technology Agency.

References
( 1 ) Hagood, N. W. and Flotow, A. v., Damping of Structural Vibration with Piezoelectric
Materials and Passive Electrical Networks, Journal of Sound and Vibration, Vol.146,
No.2(1991), pp. 243–268.
( 2 ) Lesieutre, G. A., Recent Advances in the Use of Piezoceramics for Vibration Suppres-
sion, The Shock and Vibration Digest, Vol.30, No.3(1998), pp. 187–195.
( 3 ) Fujita, T., Nomura, H., Yasuda, M., Matsuura, A. and Tchuchiya, M., Fundamental
Study of Passive Microvibration Control with Smart Structure using Piezoelectric De-
vices, Transactions of the Japan Society of Mechanical Engineers, Series C, Vol.66,
No.644(2000), pp. 1097–1101.
( 4 ) Ho, S. T, Matsuhisa, H. and Honda, Y., Passive Vibration Suppression of Beam with
Piezoelectric Elements (An Approach Using Analogies between Electrical and Mechan-
ical Circuits), JSME International Journal, Series C, Vol.43, No.3(2000), pp. 740–747.
( 5 ) Takigami, T. and Tomioka, T., Passive Vibration Suppression of Railway Carbody with
Piezoelectric Elements (Feasibility Study by Using Pipe Model) Transactions of the
Japan Society of Mechanical Engineers, Series C, Vol.68, No.674(2002), pp. 3029–
3036.
( 6 ) Dosch, J. and Inman, D., A Self-Sensing Piezoelectric Actuator for Collocated Control,
Journal of Intelligent Material Systems and Structures, No.3(1992), 166–185.
( 7 ) Takigami, T. and Tomioka, T., Analysis of Vehicle Vibration Using Linear Prediction
Model, Proceedings of the Dinamics and Design Conference 2001 (CD-ROM), No.01-
5(2001-8), #711.
( 8 ) Schaumann, R, and Valkenburg, M.E.V, Design of Analog Filters, Oxford University
Press, (2001), pp. 178–187.
( 9 ) Behrens, S., Fleming, A.J. and Moheimani, S.O.R., A Broadband Controller for Shunt
Piezoelectric Damping of Structural Vibration, Smart Materials and Structures, Vol.12,
No.1(2003), pp. 18–28.

660