You are on page 1of 22

ARTICLE IN PRESS

International Journal of Mechanical Sciences 46 (2004) 12451266


www.elsevier.com/locate/ijmecsci

Finite element modelling and experimental modal testing of a


three-dimensional framework
Jia-Jang Wu
Department of Marine Engineering, National Kaohsiung Marine University, No. 142, Hai-Chuan Road, Nan-Tzu,
Kaohsiung, Taiwan 811, Republic of China
Received 20 July 2002; received in revised form 1 July 2004; accepted 18 July 2004
Available online 21 September 2004

Abstract

The object of this paper is to develop a nite element model for a 1/10 scale crane rig in the laboratory, so
that one may predict the dynamic behaviour of the scale crane rig based on the relevant features of the
developed nite element model. To this end, the nite element modelling and experimental modal testing
for the 1/10 scale crane rig are rst carried out. Two kinds of coupling for the transducer and the tested
structure for achieving the better experimental outcome are proposed. The nite element model is then
modied, according to the experimental results, using various techniques. It has been found that the new
modied nite element model, by replacing the conventional innite rigidity for the ground-xed nodes by
the appropriate stiffness for the translational and rotational springs, can predict the vibration
characteristics of the experimental crane rig with satisfactory accuracy.
r 2004 Elsevier Ltd. All rights reserved.

Keywords: Finite element method; Modal testing; Free vibration

1. Introduction

The motivation of this paper is to improve the design of a mobile gantry crane. In order to
simulate the dynamic behaviour of the full-size crane (see Fig. 1), a 1/10 scale model has been built
in the laboratory [1]. The last scale model consists of the following parts (see Figs. 2 and 3): (i) the
Tel.: +886-7-8100888x5230; fax: +886-6-2808458.
E-mail address: jjangwu@mail.nkmu.edu.tw (J.-J. Wu).

0020-7403/$ - see front matter r 2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijmecsci.2004.07.002
ARTICLE IN PRESS

1246 J.-J. Wu / International Journal of Mechanical Sciences 46 (2004) 12451266

Nomenclature

kr stiffness of each rotational spring element


kt stiffness of each translational spring element
 y
x; global co-ordinate system for the mobile crane
DoiSc percentage differences between the computed natural frequencies and the measured
ones using a double-side screw coupling
DoiTc percentage differences between the computed natural frequencies and the measured
ones using a tubular coupling
oiFEM natural frequencies of the scale crane model obtained from FEM
oiSc natural frequencies of the scale model obtained from modal testing by using a double-
side screw coupling
oiTc natural frequencies of the scale model obtained from modal testing by using a tubular
coupling

stationary framework; (ii) the moving rails; (iii) the overhead trolley and (iv) the spreader. In
Fig. 3, the dashed lines and the two parallel beams located at the top of the scale crane model
represent the stationary framework of the scale model. The lower part of the framework is xed to

n
o tio
e yM
oll
Tr
Hoist Motion

Sp
Sl read
ew er
M
oti Sp
on Ga
Ex rea ntr
ten der yM
d/R oti
etr on
ac
t

Ga
nt
ry
Sk
ew

Fig. 1. A typical full-size crane.


ARTICLE IN PRESS

J.-J. Wu / International Journal of Mechanical Sciences 46 (2004) 12451266 1247

Fig. 2. Picture of the scale model in the laboratory.

Parallel top beams

Moving rails

Longitudinal motion
Transverse motion
Trolley

Spreader

Fig. 3. Sketch for the scale crane model (all the motors are not shown).
ARTICLE IN PRESS

1248 J.-J. Wu / International Journal of Mechanical Sciences 46 (2004) 12451266

the ground. The moving rails are tted on a rectangular frame and can run along the two parallel
beams on top of the stationary framework in the longitudinal direction. The overhead trolley is a
fabricated subsystem. It can run on the two moving rails in the transverse direction, and contains
a rotational motor, a hoisting motor and a winch (hoist drum) to rotate, lower and hoist the
spreader. The spreader is congured as a rectangular plate in the scale model. For convenience of
description, the part composed of the moving rails, the rectangular frame (on which the moving
rails are tted), the trolley and the spreader (including its attachments) is called the moving
substructure in this paper. In other words, the speed (in the longitudinal direction) of the moving
rails is always equal to that of the moving substructure. There are some differences between the
actual crane (see Fig. 1) and the scale crane model (see Figs. 2 and 3). The actual crane may move
on wheels, but the scale model is xed to the oor. Besides, in the scale model, the horizontal
movements of the spreader are achieved by running the moving substructure on two rails xed on
top of the stationary framework, and/or the trolley on the two moving rails of the moving
substructure (see Fig. 3).
Modal testing is a technique which has been widely used in structural engineering for
determining structural modal parameters, such as natural frequencies, mode shapes, etc. Many
researchers have studied the related problems. For example, Yang and Sadler [2], Humar and
Soucy [3], Ceballos et al. [4], Alampalli [5], Hollkamp and Gordon [6], Kwon and Lee [7], Wang
[8], Hermans and Auweraer [9], Kennes et al. [10] and Stanbridge and Ewins [11] have investigated
the dynamic characteristics of jet engine blades, rotating machinery and civil engineering
structures by means of the modal testing technique. From the existing literature, it is found that
the reliability of the adopted mathematical model must be validated before the dynamic analyses
are performed. One of the most convenient ways for doing the last work is to compare the
computed values of natural frequencies and mode shapes of the mathematical model with the
experimental ones obtained from the corresponding scale model. In this paper, the nite element
method (FEM) and the experimental modal testing technique are employed for this purpose.
In the conventional nite element model the general screw-fastened joints in the scale model are
considered to be the rigid joints, and the translational displacements (and/or rotational angles) of
a xed node (or ground-xed node) are assumed to be equal to zero. Of course, these are also the
assumptions made in the nite element model of this paper. Therefore, the main work of this
paper is to modify the boundary conditions at the ground attachment points, so that better
agreement between the theoretically computed values and the experimental ones may be achieved.

2. Finite element modelling of the scale crane rig

The 1/10 scale crane model is shown in Figs. 2 and 3. For convenience of dynamic analysis, the
entire structure is divided into two parts: the stationary framework (as shown in Fig. 4(b)) and
the moving substructure (as shown in Fig. 4(c)). In such a case, the dynamic characteristics of the
scale crane structure may be predicted by computing the forced vibration response of
the stationary framework subjected to the moving substructure.
Since the key point of this paper is to investigate the dynamic characteristics of the stationary
framework induced by the movements of the moving substructure and trolley, it is necessary to
conrm the reliability of the adopted mathematical model of the stationary framework before the
ARTICLE IN PRESS

J.-J. Wu / International Journal of Mechanical Sciences 46 (2004) 12451266 1249

Fig. 4. The 1/10 scale crane model: (a) complete; (b) stationary framework; (c) moving substructure (rotational and
hoisting motors are not shown in the gure).

2.57 m
1.48 m
z
y
x

1.46 m

2.20 m

Fig. 5. The orthographic view for the nite element model of the stationary framework of the experimental crane rig.

forced vibration analyses are performed. For this reason, the entire moving substructure
(including all its attachments) is taken apart from the scale crane model. An orthographic view of
the nite element model and the principal dimensions for the stationary framework is shown in
Fig. 5. The whole model consists of 99 beam elements and 73 nodes. The nite element model,
together with the xyz global co-ordinate system for the stationary framework, is shown in Fig. 6,
ARTICLE IN PRESS

1250 J.-J. Wu / International Journal of Mechanical Sciences 46 (2004) 12451266

Fig. 6. Finite element model and global co-ordinate system for the stationary framework. The digits in the small circles
represent the numberings of nodes.
ARTICLE IN PRESS

J.-J. Wu / International Journal of Mechanical Sciences 46 (2004) 12451266 1251

48.6631 mm

139.779 mm
48.6631 mm

(a) (b)

3 mm 49 mm

49 mm
49 mm

3 mm
49 mm
(c) (d)

Fig. 7. Cross-sections for the beam elements between: (a) nodes 46 and 47; (b) nodes 48 and 49; (c) nodes 1 and 18, 18
and 11, 6 and 24, 24 and 16; (d) all the other nodes.

in which the digits in the small circles denote the numberings of the nodes. The co-ordinates for all
the 73 nodes of the nite element model are listed in the appendix at the end of this paper. Among
all the beam elements composed of this nite element model, the cross-section of the structural
material between nodes 46 and 47 is a solid square, that between nodes 48 and 49 is a solid circle,
the cross-sections of the material between nodes 1 and 18, 18 and 11, 6 and 24, 24 and 16 are L-
shaped, while the cross-sections of all the other beam elements are hollow squares, as shown in
Figs. 7(a)(d), respectively. All the beam elements are made of steel with mass density r
7820 N=m3 and Youngs modulus of elasticity E 206:8  106 N=m2 :
By using the nite element software I-DEAS [12], one obtains the rst 10 natural frequencies
and the corresponding mode shapes of the scale crane model, as shown in Figs. 8(a)(j), where
oiFEM (i 110) represent the ith natural frequencies of the scale crane model obtained from the
FEM.

3. Experiments on the scale crane model

Nowadays the FEM is one of the most popular approaches for the vibration analysis of
structures, but accuracy of the FEM can be questionable and validation is usually required. For
this reason, the modal testing is undertaken to measure the natural frequencies and the
corresponding mode shapes of the scale crane model, and then a comparison between the last
results and those obtained from the nite element model is made. In general, theoretical
prediction of the damping coefcient for each vibration mode is difcult, thus, no attempt is made
here to measure the modal damping within the tests.
ARTICLE IN PRESS

1252 J.-J. Wu / International Journal of Mechanical Sciences 46 (2004) 12451266

(a) (b) (c)

(d) (e) (f)

(g) (h) (i)

z
y
x
(j)

Fig. 8. The rst 10 mode shapes for the nite element model of the stationary framework: (a) 1st mode with o1FEM
9:4651 Hz; (b) 2nd mode with o2FEM 11:7505 Hz; (c) 3rd mode with o3FEM 14:9176 Hz; (d) 4th mode with o4FEM
19:8021 Hz; (e) 5th mode with o5FEM 26:8943 Hz; (f) 6th mode with o6FEM 35:5472 Hz; (g) 7th mode with o7FEM
37:6940 Hz; (h) 8th mode with o8FEM 38:3473 Hz; (i) 9th mode with o9FEM 45:2148 Hz; (j) 10th mode with
o10FEM 48:1794 Hz:

The setup for the modal testing of the scale model is shown in Fig. 9. Two shakers are used to
give simultaneous excitations on node 10 and node 18 in the positive y and x directions,
respectively. Each shaker is equipped with a load cell to detect the magnitude of the exciting force.
ARTICLE IN PRESS

J.-J. Wu / International Journal of Mechanical Sciences 46 (2004) 12451266 1253

LMS
Software

Power Charge Pulley


Amplifier Analyser Amplifier

Load Cell
Node 18
Shaker z y

Accelerometer
Motor x
Node 10
er
ak

Load Cell
Sh

Fig. 9. Conguration for the modal testing of the scale crane rig.

The vibration responses at all the nodes are measured with accelerometers. The modal parameters
(the natural frequencies and the corresponding mode shapes of the scale crane model) are
determined using the LMS modal testing software [13,14].
Two kinds of mechanical coupling are used to connect the load cell (or force transducer) and
the tested structure (the scale crane model). It is evident that the position of the load cell should be
as close to the structure as possible such that the detected force (of the load cell) may be very close
to the true exciting force on the structure. Because the mechanical couplings used in the modal
testing are short and light, the inuence due to position of the load cell is signicantly reduced.
Between the two types of mechanical coupling, the one shown in Fig. 10(a) is the tubular type and
that in Fig. 10(b) is the double-side screw type. From the frequency response function (FRF)
graphs for node 18 vibrating in the x direction, as shown in Fig. 11, one sees that the coupling
between the load cell and the tested model signicantly affects the experimental results. From the
response amplitudes shown in Fig. 11, one sees that there exists a little difference between the
performances of the two couplings. A practical decision was made to use the double-side screw
coupling.
In Table 1, the 2nd column shows the rst 10 natural frequencies of the nite element model
obtained from I-DEAS, oiFEM (i 110), and the 3rd and 5th columns show those of the scale
crane model obtained from modal testing by using the double-side screw coupling, oiSc ; and the
tubular coupling, oiTc ; respectively. It is seen that the percentage differences between the
computed natural frequencies and the measured values using the double-side screw coupling,
DoiSc (%), are smaller than those between the computed natural frequencies and the measured
values using the tubular coupling, DoiTc (%), as shown in the 4th and 6th columns of
Table 1, respectively. Therefore, the double-side screw coupling will be adopted for the
subsequent model tests. Note that the values of DoiSc and DoiTc are obtained from the
ARTICLE IN PRESS

1254 J.-J. Wu / International Journal of Mechanical Sciences 46 (2004) 12451266

Fig. 10. Two kinds of coupling for the force transducer and the tested structure: (a) tubular type; (b) double-side screw
type.

following formulas
DoiSc % joiFEM  oiSc j  100%=oiFEM ; (1)

DoiTc % joiFEM  oiTc j  100%=oiFEM : (2)


From Table 1, one also nds that the rst 10 measured natural frequencies (either oiSc or oiTc )
are lower than the corresponding computed values, oiFEM : This is a reasonable result because the
actual rigidity of the screw-fastened (or xed) joints in the scale crane model is smaller than the
theoretical rigidity assumed for the conventional FEM.
Figs. 12(a)(e) show the rst ve mode shapes and the associated natural frequencies of the
stationary framework of the scale crane model when using the double-side screw coupling. It is
noted that the moving substructure (including the moving rails, the overhead trolley and the
spreader) has been removed from the whole structure of the scale crane model, as has been
mentioned in Section 2.
ARTICLE IN PRESS

J.-J. Wu / International Journal of Mechanical Sciences 46 (2004) 12451266 1255

Tubular coupling Double-side screw coupling

90

80
Response amplitude (m/s ) / N

70
2

60

50

40

30

20

10

0
0 20 40 60 80 100
Frequency (Hz)

Fig. 11. FRF graphs obtained by using different couplings.

Table 1
Percentage differences between the natural frequencies obtained from the nite element model using FEM oiFEM and
those from model tests on the scale crane model by using the double-side screw coupling oiSc and the tubular coupling
oiTc

Mode no., i Natural Natural Percentage Natural Percentage


frequencies from frequencies from differences frequencies from differences
I-DEAS oiFEM LMS using DoiSc %* LMS using DoiTc %**
(Hz) double-side screw tubular coupling
coupling, oiSc oiTc (Hz)
(Hz)

1 9.465134
2 11.75050 9.01 23.32
3 14.91768 10.46 29.89 9.37 37.20
4 19.80210 14.85 25.00 14.61 26.21
5 26.89430 20.07 25.36 19.24 28.45
6 35.54720
7 37.69403
8 38.34730 31.52 17.79 30.39 20.74
9 45.21481
10 48.17940
*
DoiSc % joiFEM  oiSc j  100%=oiFEM :
**
DoiTc % joiFEM  oiTc j  100%=oiFEM :

4. Modications of the nite element model for the scale crane rig

For convenience of description, the nite element model of the scale crane rig established in the
last section and the one in this section are, respectively, called the original nite element model
ARTICLE IN PRESS

1256 J.-J. Wu / International Journal of Mechanical Sciences 46 (2004) 12451266

(a) (b) (c)

z
y

x
(d) (e)

Fig. 12. The rst ve mode shapes and associated natural frequencies of the scale crane model obtained from modal
testing: (a) 1st mode with o1Sc 9:01 Hz; (b) 2nd mode with o2Sc 10:46 Hz; (c) 3rd mode with o3Sc 14:85 Hz;
(d) 4th mode with o4Sc 20:07 Hz; (e) 5th mode with o5Sc 31:52 Hz:

and the modied nite element models. A comparison between the rst ve calculated natural
frequencies of the original nite element model determined by FEM, oiFEM (i 15), and the
corresponding measured values, obtained from modal testing on the scale crane model by using
the double-side screw coupling, oiSc ; is shown in Table 2. It is noted that the comparison between
the measured natural frequencies and calculated values should only be made on the basis of the
corresponding mode shapes. The values of percentage differences, DoiSc ; as shown in the nal
column of Table 2, are determined from Eq. (1). It can be seen that the percentage differences
DoiSc %) between the calculated natural frequencies and the measured values are not small
enough. For the 2nd mode, the maximum value of DoiSc even reaches 29.89%.
It is proposed that the measuring position of the accelerometer, as shown in Fig. 13, may have
had something to do with the results of the modal testing; therefore similar tests were repeated by
changing the positions of the accelerometers. However, it will subsequently be found that the
inuence of the measuring position of the instrument is not signicant for the modal testing
conducted.
Referring to the foregoing discussions it is thought that the disagreement between the
theoretical analyses and the measurements may be due to (i) the mechanical properties of the rigid
joints (in the nite element model) not being very close to those of the screw-fastened joint (in the
scale crane model) and (ii) the translational displacements (and/or rotational angles) of the xed
end (i.e., the ground-xed nodes  as shown in Fig. 16) not really being equal to zero, as assumed
by the FEM. In the FEM, the degrees of freedom (DOFs) for each node of a three-dimensional
structure is six (6), i.e., ux ; uy ; uz ; yx ; yy and yz : Among which, ux ; uy and uz denote the
ARTICLE IN PRESS

J.-J. Wu / International Journal of Mechanical Sciences 46 (2004) 12451266 1257

Table 2
Comparison between the rst ve natural frequencies of the original nite element model determined by FEM, oiFEM ;
and those obtained from modal testing on the scale crane model using a double-side screw coupling, oiSc

Mode no., i Computed natural Measured natural Percentages of frequency


frequencies obtained from I- frequencies obtained from differences DoiSc (%)*
DEAS, oiFEM (Hz) LMS, oiSc (Hz)

1 11.75 9.01 23.32


2 14.92 10.46 29.89
3 19.80 14.85 25.00
4 26.89 20.07 25.36
5 38.34 31.52 17.79
*
DoiSc % oiFEM  oiSc  100%=oiFEM :

Accelerometer

Finite element model

Node

Real structure

Fig. 13. Measurement position of an accelerometer on the tested structure is found not to be a key factor affecting the
results of the present model tests. The solid line () and dashed line (- - -), respectively, represent the real structure and
the nite element model of the actual structure.
ARTICLE IN PRESS

1258 J.-J. Wu / International Journal of Mechanical Sciences 46 (2004) 12451266

translational displacements of the node in the x;  y and z directions, respectively, while yx ; yy and
yz denote the associated rotational angles about the x-,  y-  and z-axes, respectively, where xyz is
the global co-ordinate system for the problem. In theory, ux uy uz 0 and yx yy yz 0
for a ground-xed node. It is evident that the last ideal support conditions cannot be completely
achieved in practice, particular for the nodes at the wooden oor. In such a case, the elastic
supports as shown in Fig. 16(b) should be more realistic. In which, the stiffnesses of the support
node associated with the three translational DOFs (ux ; uy and uz ) are modelled by the three linear
translational springs with spring constants ktx ; kty and ktz ; respectively, while those associated with
the three rotational DOFs (yx ; yy and yz ) are modelled by the three linear rotational springs with
spring constants krx ; kry and krz ; respectively, where the subscripts t and r for the spring constants
k refer to the translational and rotational DOFs, respectively, while the subscripts x;  y and z refer
to the DOFs associated with the global co-ordinate axes, x;  y and z; respectively. It is apparent
that the dynamic characteristics of the elastic support shown in Fig. 16(b) will be identical to those
of a ground-xed node if the stiffness constants of all the springs (kti and kri with i x;  y and z)
approach innities. In other words, for convenience, the ground-xed node may be considered as
the special case of an elastic-supported node in the theoretical (numerical) analysis.
In practical applications, one may change the magnitudes of the spring constants
(ktx ; kty ; ktz ; krx ; kry and/or krz ) according to the actual situations and if the theoretical (nite
element) values of natural frequencies (and mode shapes) of the structure are very close to the
corresponding ones obtained from the eld measurements, then the theoretical support conditions
should be very close to the actual ones for the ground-xed nodes.
The general structural joints of the scale crane model are shown in Figs. 14(a) and (b); however,
the typical joint for the conventional FEM takes the form as shown in Fig. 14(c). It is noted that

Block

(a) (b)

Centrelines of beam
elements of framework
Beam elements
of framework
B

Node
(c)

Fig. 14. (a) General structural joint in the scale crane model; (b) alternative joint in the scale crane model; (c) rigid joint
for the conventional nite element model.
ARTICLE IN PRESS

J.-J. Wu / International Journal of Mechanical Sciences 46 (2004) 12451266 1259

the node shown in Fig. 14(c) is a point at which all beam elements passing through point B meet.
To satisfy the requirement for a node of the FEM, the joints of Figs. 14(a) and (b) are replaced
by the ones shown in Figs. 15(a) and (b) rst, and then, in turn, by those shown in Figs. 15(a)0
and (b)0 ; respectively. In Figs. 15(a) and (a)0 ; the upper beam elements and the lower ones
of the framework are connected by a two-node connecting beam element (i.e., the screw with
length s 49 mm and diameter d s 15 mm). On the other hand, in Figs. 15(b) and (b)0 ;
the upper beam elements and the lower ones of the framework are connected by two-node
connecting beam elements together with a block beam element. The dimensions of the block beam
element are: length b 49 mm, width w 49 mm and thickness tb 5 mm, while the length of
each connecting beam element for Figs. 15(b) and (b)0 is b 24:5 mm. The nite element analysis
was then carried out on the modied nite element model, but no satisfactory results were
obtained.
Since the scale crane model is attached to the wooden tiles of the second oor of a building, the
translational displacements and rotational angles for the nodes of the scale crane model xed on
the wooden tiles are possibly not exactly equal to zero. Hence, several translational and rotational
spring elements (see Fig. 16) have been introduced to the ground-xed nodes to modify the nite
element model. After extensive studies it was found that if the stiffness, kt and kr ; of each
translational and rotational spring element were chosen to be 2000 (i.e., kt 2000 N/mm and
kr 2000 Nm/rad), then the rst ve natural frequencies of the new modied nite element model
(see Fig. 17) determined by FEM, oiFEM ; will be close to the corresponding values obtained from

Connecting beam element


Centrelines of beam
elements of framework

Beam elements
of framework
Node
Two-node
connecting beam element
(a) (a)'

Connecting beam element


Centrelines of beam
elements of framework

Beam elements
of framework Block beam Node
element

Two-node
Block beam element connecting beam element
(b) (b)'

Fig. 15. To satisfy the requirement of rigid joint of the nite element model as shown in Fig. 14(c), the joint of Fig. 14(a)
is replaced by the modied ones shown in (a) and (a)0 and that of Fig. 14(b) is replaced by those shown in (b) and (b)0 :
ARTICLE IN PRESS

1260 J.-J. Wu / International Journal of Mechanical Sciences 46 (2004) 12451266

Pulley
Fixed Rail, Q Fixed Rail
z
y

k rx
kty
Ground-fixed node
kr k tx x
Motor
y
k rz
Ground-fixed node
k tz
Ground-fixed node

(a) (b)

Fig. 16. Translational spring together with rotational spring elements are introduced to the ground-xed nodes  for
the new modied nite element model of the scale crane rig: (a) sketch of the scale crane model; (b) the stiffnesses for the
6 DOFs at each ground-xed node are modelled by three translational springs with spring constants ktx ; kty and ktz ; and
three rotational springs with spring constants krx ; kry and krz :

z
y
x

Fig. 17. New modied nite element model for the stationary framework of the scale crane rig.
ARTICLE IN PRESS

J.-J. Wu / International Journal of Mechanical Sciences 46 (2004) 12451266 1261

Table 3
The rst ve natural frequencies of the new modied nite element model determined by FEM, oiFEM ; and the
corresponding values obtained from modal tests on the scale crane model, oiSc

Mode no. i Natural frequencies Natural frequencies Frequency differences Percentage


of new modied nite of scale crane model joiFEM  oiSc j (Hz) differences DoiSc
element model with using LMS system, (%)*
FEM, oiFEM (Hz) oiSc (Hz)

1 9.05 9.01 0.04 0.48


2 10.05 10.46 0.41 3.93
3 15.89 14.85 1.04 7.02
4 20.51 20.07 0.44 2.17
5 30.11 31.52 1.41 4.46
Average 3.61
*
DoiSc % joiFEM  oiSc j  100%=oiFEM :

modal testing on the scale crane model, oiSc ; as shown in Table 3. From the table one sees that the
maximum percentage difference is Do3Sc 7:02% and the average value of DoiSc is 3.61%. In
practice this average percentage difference of DoiSc should be acceptable, hence this new modied
nite element model (see Fig. 17) will be used for further studies on the forced vibration analyses
of the scale crane model. Where the values of DoiSc %, as shown in the nal column of Table 3,
are calculated with Eq. (1).
The inuence of the stiffness, kt (N/mm) and kr (Nm/rad), of each translational and rotational
spring for the ground-xed nodes on the rst ve natural frequencies of the new modied nite
element models determined by FEM, oiFEM (i 15), is shown in Table 4 and Fig. 18. It is seen
that the variation of the spring stiffness (kt and kr ) is from 500 (i.e., kt 500 N/mm and
kr 500 Nm/rad) to 3500 (i.e., kt 3500 N/mm and kr 3500 Nm/rad). The dashed line
with star (- -%- -), as shown in Fig. 18, represents the average percentage differences for the
rst ve natural frequencies of the new modied nite element models, due to the variation
of the spring stiffness. The lowest value of the percentage difference appears when the stiffness
of each translational and rotational spring element is equal to 2000 (i.e., kt 2000 N/mm
and kr 2000 Nm/rad). This is the reason why kt 2000 N/mm and kr 2000 Nm/rad are
used for the nite element model in this paper. The lowest ve natural frequencies and associated
mode shapes of the new modied nite element model are shown in Figs. 19(a)(e) and those
of the actual stationary framework of the scale crane model obtained from modal testing
are shown in Figs. 19(a)0 (e)0 : Among the last gures, the translational vibrations in the
x and y directions are shown in Figs. 19(a) and (a)0 and(b) and (b)0 ; respectively, while those in
the z direction are shown in Figs. 19(c) and (c)0 and (e) and (e)0 : The remaining mode shapes
as shown in Figs. 19(d) and (d)0 denote the rotational vibrations of the stationary

framework about the x-axis. It is evident that the lowest ve natural frequencies and the
associated mode shapes obtained from the new modied nite element model as shown
in Figs. 19(a)(e) are very close to the corresponding ones obtained from the modal testing as
shown in Figs. 19(a)0 (e)0 :
1262
J.-J. Wu / International Journal of Mechanical Sciences 46 (2004) 12451266
Table 4

ARTICLE IN PRESS
Inuence of stiffness, kt (N/mm) and kr (Nm/rad), of each translational and rotational spring element for the ground-xed nodes on the rst ve
natural frequencies of the modied nite element model determined by FEM, oiFEM

Mode no., i Natural frequencies, oiFEM (Hz)

kt kr 500 kt kr 1300 kt kr 1800 kt kr 1900 kt kr 2000 kt kr 2100 kt kr 2500 kt kr 3500

1 6.480 8.359 8.896 8.898 9.053 9.123 9.358 9.745


2 8.486 9.631 9.954 10.003 10.049 10.091 10.233 10.468
3 13.345 15.392 15.789 15.843 15.893 15.938 16.082 16.299
4 17.712 19.942 20.387 20.449 20.506 20.557 20.723 20.975
5 27.086 29.550 29.995 30.058 30.114 30.165 30.326 30.566
ARTICLE IN PRESS

J.-J. Wu / International Journal of Mechanical Sciences 46 (2004) 12451266 1263

40

Percentages of frequency differences, iSc (%)


35

1st mode
30 2nd mode
3rd mode

25 4th mode
5th mode
Average
20

15

10

0
500 1500 2500 3500
1000 2000 3000
Stiffness of each translational and rotational spring element, kt (N/mm), kr (Nm/rad)

Fig. 18. Inuence of stiffness, kt and kr ; of each translational and rotational spring element for the ground-xed nodes
on the percentage differences DoiSc % between the rst ve natural frequencies of the modied nite element model
determined by FEM, oiFEM ; and the corresponding values obtained from modal tests on the scale crane model, oiSc :

5. Conclusions

In general the accuracy of the nite element modelling of a real structure should be considered
and experimental modal testing is one of the effective validation tools. In this paper, it has been
found that a nite element model built on the assumption that the translational displacements and
rotational angles of all the ground-xed nodes are equal to zero, as typical when using
conventional FEM, cannot provide satisfactory predictions. An improved nite element model
has been obtained after some modications have been made to the stiffness of the translational
and rotational degrees of freedom of the ground-xed nodes for the scale crane model. It has also
been found that the type of coupling between the load cell and the tested structure signicantly
affects the results of modal testing. Hence, further study on the mechanical conguration of model
tests should be generally encouraged.

Appendix

Global co-ordinates for nodes of the nite element model of the stationary framework.
(unit: mm)
ARTICLE IN PRESS

1264 J.-J. Wu / International Journal of Mechanical Sciences 46 (2004) 12451266

z z
y
y
x x
1FEM = 9.05Hz 1Sc = 9.01Hz
(a) (a)

z z
y y
x x
2FEM = 10.05Hz 2Sc = 10.46 Hz
(b) (b)

z z
y
y
x
x
3 FEM = 15.89 Hz 3Sc = 14.85Hz
(c) (c)

z z
y
y
x x
4FEM = 20.51Hz 4 Sc= 20.07 Hz
(d) (d)

z z
y
y
x
x
5FEM = 30.11Hz 5Sc = 31.52 Hz
(e) (e)

Fig. 19. The rst ve natural frequencies and mode shapes of the new modied nite element model obtained from
FEM and those of the stationary framework of the scale crane model obtained from modal testing. (a) (a)0 1st mode
with o1FEM 9:05 Hz and o1Sc 9:01 Hz; (b) (b)0 2nd mode with o2FEM 10:05 Hz and o2Sc 10:46 Hz; (c) (c)0 3rd
mode with o3FEM 15:89 Hz and o3Sc 14:85 Hz; (d) (d)0 4th mode with o4FEM 20:51 Hz and o4Sc 20:07 Hz;
(e) (e)0 5th mode with o5FEM 30:11 Hz and o5Sc 31:52 Hz:
ARTICLE IN PRESS

J.-J. Wu / International Journal of Mechanical Sciences 46 (2004) 12451266 1265

Node no. x y z Node no. x y z


1 1285 0 2200 2 1285 0 920
3 1285 0 0 4 485 0 0
5 485 0 0 6 1285 0 0
7 1285 0 920 8 1285 0 2200
9 1285 2940 2200 10 1285 2940 920
11 1285 2940 0 12 485 2940 0
13 485 2940 0 14 1285 2940 0
15 1285 2940 920 16 1285 2940 2200
17 1285 1460 2200 18 1285 1460 1092.5
19 1285 1460 0 20
21 0 1460 0 22
23 1285 1460 0 24 1285 1460 1107.5
25 1285 1460 2200 26 485 2740 0
27 419.5 2740 0 28 419.5 2740 0
29 485 2740 0 30 485 2590 0
31 370.3 2590 0 32 370.3 2590 0
33 485 2590 0 34 485 350 0
35 368.7 350 0 36 368.7 350 0
37 485 350 0 38 485 90 0
39 455.1 90 0 40 455.1 90 0
41 485 90 0 42 365 0 0
43 365 0 0 44 365 2940 0
45 365 2940 0 46 0 90 0
47 0 350 0 48 0 2590 0
49 0 2740 0 50 485 560 0
51 485 710 0 52 485 860 0
53 485 1010 0 54 485 1160 0
55 485 1310 0 56 485 1460 0
57 485 1610 0 58 485 1760 0
59 485 1910 0 60 485 2060 0
61 485 2210 0 62 485 2360 0
63 485 560 0 64 485 710 0
65 485 860 0 66 485 1010 0
67 485 1160 0 68 485 1310 0
69 485 1460 0 70 485 1610 0
71 485 1760 0 72 485 1910 0
73 485 2060 0 74 485 2210 0
75 485 2360 0
ARTICLE IN PRESS

1266 J.-J. Wu / International Journal of Mechanical Sciences 46 (2004) 12451266

References

[1] Wu JJ. An analysis of the structural dynamics of a mobile gantry crane with application to automation of
container management. PhD thesis, University of Glasgow, UK, 2000.
[2] Yang Z, Sadler JP. Prediction of the dynamic response of exible manipulators from a modal database.
Mechanism and Machine Theory 1997;32:67989.
[3] Humar JL, Soucy Y. Hybrid component mode synthesis based on test derived data. Computers and Structures
1998;32:50315.
[4] Ceballos MA, Car EJ, Prato TA, Prato CA, Alvarez LM. Experimental and numerical determination of the
dynamic properties of the reactor building of Atucha II NPP. Nuclear Engineering and Design 1998;182:93106.
[5] Alampalli S. Effects of testing, analysis, damage, and environment on modal parameters. Mechanical Systems and
Signal Processing 1999;14:6374.
[6] Hollkamp JJ, Gordon RW. Modal test experiences with a jet engine dan model. Journal of Sound and Vibration
2001;248:15165.
[7] Kwon KS, Lee CW. Random excitation for modal testing of rotating machinery: use of modulation technique.
Journal of Sound and Vibration 2000;234:297309.
[8] Wang BT. Structural modal testing with various actuators and sensors. Mechanical Systems and Signal Processing
1998;12:62739.
[9] Hermans L, Der Auweraer HV. Modal testing and analysis of structures under operational conditions: industrial
applications. Mechanical Systems and Signal Processing 1999;13:193216.
[10] Kennes P, Anthonis J, Clijmans L, Ramon H. Construction of a portable test rig to perform experimental modal
analysis on mobile agricultural machinery. Journal of Sound and Vibration 1999;228:42141.
[11] Stanbridge AB, Ewins DJ. Modal testing using a scanning laser doppler vibrometer. Mechanical Systems and
Signal Processing 1999;13:25570.
[12] SDRC. I-DEAS master series 8 nite element modeling. Structural Dynamics Research Corporation, 2001.
[13] Ewins DJ. Modal testing: theory and practice. Research Studies Press Ltd.; 1984.
[14] LMS, LMS CADA-X USER MANUAL, LMS International; 1992.

You might also like