Download as pdf or txt
Download as pdf or txt
You are on page 1of 13

Finite Element Modeling of

Structures With L-Shaped Beams


and Bolted Joints
Due to bending-torsion coupled vibrations of the L-shaped beams and numerous uncer-
tainties associated with the bolted joints, modeling structures with L-shaped beams and
K. He bolted joints is a challenging task. With the recent development of the modeling tech-
Graduate Research Assistant niques for L-shaped beams by the authors (He and Zhu, 2009, “Modeling of Fillets in

Downloaded from http://asmedigitalcollection.asme.org/vibrationacoustics/article-pdf/133/1/011011/5832668/011011_1.pdf by Gadjah Mada University user on 26 June 2023


e-mail: hekun1@umbc.edu Thin-Walled Beams Using Shell/Plate and Beam Finite Elements,” ASME J. Vibr.
Acoust., 131(5), p. 051002), this work focuses on developing new finite element (FE)
W. D. Zhu1 models for bolted joints in these structures. While the complicated behavior of a single
Professor bolted connection can be analyzed using commercial FE software, it is computationally
e-mail: wzhu@umbc.edu expensive and inefficient to directly simulate the global dynamic response of an as-
sembled structure with bolted joints, and it is necessary to develop relatively simple and
Department of Mechanical Engineering, accurate models for bolted joints. Three new approaches, two model updating ap-
University of Maryland, proaches and a predictive modeling approach, are developed in this work to capture the
Baltimore County, stiffness and mass effects of bolted joints on the global dynamic response of assembled
Baltimore, MD 21250 structures. The unknown parameters of the models in the model updating approaches are
determined by comparing the calculated and measured natural frequencies. In the pre-
dictive modeling approach, the effective area of a bolted connection is determined using
contact FE models and an analytical beam model; its associated stiffnesses can also be
determined. The models developed for the bolted joints have relatively small sizes and
can be easily embedded into a FE model of an assembled structure. For the structures
studied, including a three-bay space frame structure with L-shaped beams and bolted
joints, and some of its components, the errors between the calculated and measured
natural frequencies are within 2% for at least the first 13 elastic modes, and the associ-
ated modal assurance criterion values are all over 94%. 关DOI: 10.1115/1.4001840兴

1 Introduction joints. The dynamic behavior of a structure can be significantly


affected by the mass, stiffness, and damping effects of the bolted
L-shaped beams and bolted joints are commonly used in frame
joints 关2–5兴. Modeling a frame using beam elements and fixed
structures, such as transmission towers, cellular phone towers, and
point connections at the bolted joints can lead to an unacceptable
bridge trusses; Fig. 1 shows a part of a tower at the Hoover Dam level 共e.g., over 50%兲 of modeling error. While the mass effect
in the U.S., which consists of L-shaped beams and a bolted joint can be relatively easily included in the FE models, the stiffness
with multiple bolted connections. To perform global dynamic and damping effects of a bolted joint, which are mainly related to
analyses of these structures, finite element 共FE兲 models are pow- the clamping force at each bolted connection, the preloadings
erful and usually necessary tools. However, modeling structures caused by assembling the structure, and the surface properties of
with L-shaped beams and bolted joints is a challenging task. The the contact interfaces, are usually difficult to model. It is difficult
first challenge arises from the modeling of the L-shaped beams. to directly model a bolted joint by taking into account all the
Because the shear center and the centroid of the cross section of parameters mentioned above 关2,6–9兴 due to several reasons. First,
an L-shaped beam do not coincide, the bending and torsional vi- it is difficult to measure these parameters associated with a bolted
brations of the L-shaped beam are usually coupled, and warping joint. To measure the surface properties of the contact interfaces,
occurs. To avoid buckling, the L-shaped beams are usually de- specially designed tests on the same or similar components of the
signed with a relatively small length-to-width ratio, which ampli- structure that include the bolted joint are needed 关8,10–12兴. The
fies the bending-torsion coupled vibrations. While the coupling preloadings due to assembling and the clamping force of a bolted
and warping effects can be captured by modeling the L-shaped joint depend on each other and vary with the vibration amplitudes,
beams using solid elements, it can result in a large model size. which makes the measurement of the parameters difficult. Second,
The model size can be significantly reduced by modeling the while many commercial FE software can deal with complex con-
flanges of an L-shaped beam using shell elements, but it is diffi- tact problems with preloadings, a sufficiently dense mesh at a
cult to model the fillet using shell elements. With the new model- contact interface is needed to achieve a desired accuracy. The
ing techniques for fillets in thin-walled beams, an L-shaped beam mesh density for a contact problem far exceeds that for structural
can be accurately modeled using shell and beam elements, and the dynamic analysis, and the model size for an assembled structure
model size can be reduced by over 90% compared with the solid would be extremely large. While the model size can be reduced by
element model 关1兴. using an adaptive mesh 关13兴, the resultant mass and stiffness ma-
The second challenge arises from the modeling of the bolted trices would still be relatively large and can have extraordinarily
large condition numbers because of the large differences in the
element stiffnesses due to different element sizes 关2兴. Third, there
1
Corresponding author. can be several contact interfaces in a single bolted connection:
Contributed by the Technical Committee on Vibration and Sound of ASME for
publication in the JOURNAL OF VIBRATION AND ACOUSTICS. Manuscript received March
e.g., the interface between the cap of the bolt and a clamped
31, 2009; final manuscript received April 13, 2010; published online January 26, component, that between the two clamped components, and that
2011. Assoc. Editor: Jiong Tang. between a clamped component and the nut. The motions of the

Journal of Vibration and Acoustics Copyright © 2011 by ASME FEBRUARY 2011, Vol. 133 / 011011-1
Downloaded from http://asmedigitalcollection.asme.org/vibrationacoustics/article-pdf/133/1/011011/5832668/011011_1.pdf by Gadjah Mada University user on 26 June 2023
Fig. 1 Part of a tower at the Hoover Dam, consisting of
L-shaped beams and a bolted joint with multiple bolted
connections

contact surfaces are coupled; solving the coupled nonlinear con-


tact problem requires very small time increments, and the solution
converges very slowly due to the coupling 关14兴.
Instead of solving the contact problems associated with the
bolted connections, much work has been done to model the mac-
roconstitutive relation between the two clamped components us-
ing a relatively simple model, which mainly addresses the slip
behavior of the clamped components. The slip can be decomposed
into three stages: the microslip stage, the macroslip stage, and the
pinning-by-shank stage 关2兴. The nonlinear force-displacement re-
lationship usually exhibits a hysteresis loop during vibrations.
Gaul 关11兴 developed a model with six parameters for a bolted
connection, which can capture the hysteresis behavior, and the
model was validated by experiments. By applying Iwan’s 关15兴
spring-slider model, Segalman 关16兴 developed a four-parameter Fig. 2 A three-bay space frame structure on an air-bed
constitutive model for bolted connections. Song et al. 关17兴 simu-
lated the dynamics of a beam structure with a bolted joint using
adjusted Iwan beam elements. Quinn and Segalman 关18兴 found ings, the selection of the updating parameters is critical. The dy-
that Iwan’s model provides accurate results for bolted joints. Us- namic response or the modal parameters used in model updating
ing a relatively simple model to capture the slip between the should be sensitive to the updating parameters, and the updating
clamped components can avoid a direct simulation of the contact parameters should represent the real uncertainties of the bolted
problems, which can significantly reduce the amount of numerical joints because the differences between the model and test data can
calculation; however, the parameters used in the models need to be reconciled by changing other parameters that do not need to be
be obtained from experiments. updated 关23兴. Most of the difficulties mentioned earlier are asso-
To avoid a direct measurement of the parameters associated ciated with the modeling of the slip behavior between the clamped
with a bolted joint, some indirect methods, which can identify the components. However, due to safety concerns, a bolted joint usu-
unknown model parameters of a bolted joint using the global ally contains multiple bolted connections 共see Fig. 1兲. The slip
structural response, are developed. In one method, a bolted joint is between the clamped components rarely happens when the clamp-
modeled by a base line model, whose unknown parameters are ing forces of the bolted connections are sufficiently large com-
determined through model updating. For instance, Yun and pared with the external loadings; the assembled structures are al-
Bauchau 关19兴 modeled a bolted connection using a number of most linear in most operation conditions 共see Sec. 2兲. The purpose
short beam elements; the number of the beam elements and the of this work is to develop new physics-based models for bolted
diameter of the cross sections of the elements were updated by joints that can accurately capture their stiffness and mass effects
comparing the calculated and measured natural frequencies. In on the global dynamic response of assembled structures.
another method, a bolted joint is represented by entries in the A modular frame structure, as shown in Fig. 2, and some of its
mass and stiffness matrices, a stiffness element, a damping ele- components, are used to demonstrate the methodology to be de-
ment, and/or a nonlinear function, where the unknown parameters veloped. The frame, which is 1.827 m tall, 0.511 m wide, and
are determined using experimental data 关20–22兴. Since it is much 0.611 m deep, consists of 12 L-shaped beams 共Fig. 6兲 and 12
easier to measure the dynamic response of an assembled structure beams with a rectangular cross section, connected by 12 bolted
than to directly measure the parameters associated with a bolted joints. The width and thickness of the rectangular beams are
joint, the indirect methods provide a more practical choice than 0.0381 m and 0.0127 m, respectively; the length of the longer
the direct methods. However, the latter indirect method cannot rectangular beams is 0.586 m, and that of the shorter ones is 0.486
always capture the physics of bolted joints. For instance, when m. The bolts 共Hex Head M10⫻ 35– 8.8兲 are made of steel with the
entries of the mass and stiffness matrices in the FE models are elastic modulus Eb = 210 GPa; a bolt mass, which includes the
directly updated, their physical meanings can be lost 关23兴. While mass of the bolt, nut, and washer, is 44 g. A torque of 40 N m is
the FE models in the former indirect method have physical mean- used to tighten a bolted connection. The frame is bolted on a plate

011011-2 / Vol. 133, FEBRUARY 2011 Transactions of the ASME


Downloaded from http://asmedigitalcollection.asme.org/vibrationacoustics/article-pdf/133/1/011011/5832668/011011_1.pdf by Gadjah Mada University user on 26 June 2023
Fig. 3 „a… A structure with two L-shaped beams and a bolted joint with the
measurement points marked in „b… and some of its dimensions shown in „c…

through four cubics; the thickness, width, and depth of the plate which can accurately describe its geometry, is created 共Fig. 6兲.
are 0.0108 m, 0.5 m, and 0.6 m, respectively. The length of each The fillet and holes in the beam, whose presences can significantly
side of the cubics is 0.0402 m. The structure is made of 6061- affect its natural frequencies and mode shapes 关1兴, are modeled.
T651 aluminum with the density ␳ = 2731.4 kg/ m3, Poisson’s ra- Note that besides the holes used for the bolted connections, there
tio ␯ = 0.33, and elastic modulus E = 68.9 GPa, and it sits on an are four holes at each end of an L-shaped beam for connecting
airbed to simulate the free boundary conditions. It has closely diagonal members of the frame, which are not used in this work.
spaced modes due to the close width and depth dimensions, which With the accurate solid element model of the L-shaped beams, one
adds difficulties in modal testing and in comparing the model and can assume that the modeling error is mainly from the modeling
test data. With the modeling techniques developed for L-shaped of the bolted joints.
beams 关1兴, new FE models are developed for bolted joints using Because the length of a bolted joint in Fig. 3 is about one-fifth
two model updating approaches, one modeling the equivalent of that of an L-shaped beam, the bolted joint cannot be simplified
stiffness and mass effects of a bolted joint and the other those of as a point connection in the FE model. The bracket at the bolted
a bolted connection, and a predictive modeling approach. The joint in Fig. 7, which is called the joint component, has a flange
models developed for the bolted joints can accurately predict the thickness of 0.006426 m and is modeled here by 1350 hexahedron
natural frequencies and mode shapes of a large number of modes
of an assembled structure and have relatively small model sizes. It
can be used for vibration-based damage detection of structures
with bolted joints, which requires multiple post-processing steps
on the FE models 关24兴.

2 Linearity Check
Two tests were performed to check the linearity of a component
of the frame in Fig. 2, with two L-shaped beams and a bolted joint
共Fig. 3兲; the linearity of the frame structure would be better than
that of the component. The structure was hung by a string at one
end to simulate the free boundary conditions. It was excited by a
PCB 086D80 impact hammer at point 37 in Fig. 3, with different
magnitudes 共Fig. 4兲; the responses were measured at point 45
using a Polytec OFV 353 laser vibrometer with an OFV 3001
controller to avoid mass loading. The excitation and measurement
directions were perpendicular to the flange surfaces. The experi-
mental data were collected by a 36-channel LMS spectrum ana-
lyzer; the measured frequency response functions 共FRFs兲 of the
structure, which were obtained using LMS Test.Lab, are shown in
Fig. 4. They almost coincide in the measured frequency range of
0–6400 Hz. A reciprocity test was also performed. The structure
was first excited and measured at points 11 and 39, respectively,
and then the excitation and measurement points were reversed; the
corresponding measured FRFs shown in Fig. 5 also almost coin-
cide. These tests show that the structure is almost linear within the
measured frequency range.

3 Model Updating Approaches


3.1 Modeling of a Bolted Joint. As mentioned in Sec. 1, the
error in modeling the frame structure mainly arises from the mod- Fig. 4 „a… The magnitudes of the Fourier transforms of the
eling of the L-shaped beams and bolted joints. To minimize the excitation forces and „b… the corresponding measured FRFs of
error in modeling an L-shaped beam, a solid element model, the structure in Fig. 3

Journal of Vibration and Acoustics FEBRUARY 2011, Vol. 133 / 011011-3


ratio is assumed to remain unchanged, and the shear modulus is
updated accordingly. When the elastic modulus of the bracket is
reduced to 0.71 times its original value, i.e., E1 = 0.71E, the maxi-
mum error between the calculated and measured natural frequen-
cies of the first 13 elastic modes is within 2% 共Table 1兲, and the
associated modal assurance criterion 共MAC兲 values 关25兴 are all
over 95%, where the experimental mode shapes were obtained by
measuring 68 translational degrees of freedom 共DOFs兲 associated
with the 51 measurement points 共Fig. 3兲 using the roving hammer
method 关25兴. Note that one can also just update the elastic modu-
lus of the bracket, with both the shear modulus and the Poisson’s
ratio assumed to remain unchanged, and achieve a similar model-
ing accuracy. The test procedure and equipment used are the same

Downloaded from http://asmedigitalcollection.asme.org/vibrationacoustics/article-pdf/133/1/011011/5832668/011011_1.pdf by Gadjah Mada University user on 26 June 2023


as those in Sec. 2, except that three impact tests were averaged
Fig. 5 The measured FRFs with the excitation and measure- here at every excitation point to ensure repeatable results with
ment points at points 11 and 39 shown in Fig. 3„b…, respectively, good coherence. The free boundary conditions of the structure
and the excitation and measurement points reversed were validated because the highest measured rigid-body mode
natural frequency is 1.47%, which is much less than 10%, of the
first elastic mode natural frequency 关25兴. The same modal testing
solid elements. The bolt masses are uniformly included in the technique was used in other experiments in this work. The up-
elements of the bracket. The bracket and the parts of the L-shaped
dated elastic modulus of the bracket E1 is less than the aluminum
beams in contact with it are first assumed to be fully connected,
elastic modulus because the bracket is connected to the L-shaped
with the material properties of the bracket set to be those of alu-
beams through ten bolted connections, which are less stiff than the
minum. The natural frequencies of the structure in Fig. 7 with free
full connection.
boundary conditions, calculated from the solid element model, are
To demonstrate the robustness of the solid element model of the
all higher than the measured ones for the first 13 elastic modes
bolted joint, the updated solid element model of the bolted joint in
共see Table 1兲, and the maximum error is 16.16%. The equivalent
Fig. 7 is used in modeling other components of the space frame,
stiffness of the bolted joint is then estimated by updating the elas-
with the same or similar bolted joints. For instance, a solid ele-
tic modulus of the bracket since the modeling error is assumed to
be mainly from the modeling of the bolted joint; the Poisson’s ment model for the structure in Fig. 8, with three L-shaped beams
and two bolted joints, is created, and the same elastic modulus is
used for the brackets to model the bolted joints. A similar model-
ing accuracy is achieved: The maximum error between the calcu-
lated and measured natural frequencies of the first 19 elastic
modes is 1.86% 共Table 2兲, and the associated MAC values are all
over 94%. Note that more modes than those in Fig. 7 can be
accurately modeled here because the joint deformation affects
higher modes with the increase in the length of the structure.
Consider next a structure in Fig. 9, with two L-shaped beams
and a bolted joint, where there are two longer rectangular beams
connected to the joint. A similar approach is used to model the
bolted joint, which consists of three joint components: the bracket
and the two parts of the rectangular beams in contact with the
bracket 共Fig. 9兲. The two rectangular beams have the same param-
eters due to symmetry, and each rectangular beam is modeled here
by 1374 hexahedron solid elements. The elastic modulus of the
bracket is assumed to remain unchanged, i.e., E1 = 0.71E, because
Fig. 6 „a… Dimensions of an L-shaped beam and „b… its solid the joint is essentially the same as that in Fig. 7. The elastic
element model with 67,321 tetrahedron solid elements, gener- modulus of the two contact parts of the rectangular beams needs
ated with ABAQUS 6.7. The thickness of the flange is 0.006317 m. to be updated; Poisson’s ratio is assumed to remain unchanged,
and the shear modulus is updated accordingly. The updated elastic
modulus of the two rectangular beam joint components is E2
= 0.36E 共Fig. 9兲. The maximum error between the measured and
calculated natural frequencies of the first nine elastic modes is
1.51% 共Table 3兲, and the associated MAC values are all over 95%.
When the right rectangular beam is removed from the structure
共Fig. 10兲, one only needs to remove the corresponding rectangular
beam in the FE model; the elastic moduli of the other two joint
components remain unchanged. The maximum error between the
calculated and measured natural frequencies of the first nine elas-
tic modes of the modified structure in Fig. 10 is 1.25%, and the
associated MAC values are all over 94%.
Dense meshes must be used here to model the curves of the
holes and the fillet of an L-shaped beam, and the meshes at the
rest part of the L-shaped beam must be compatible with those at
the holes and fillet. If the solid element model is used for the
assembled structure in Fig. 2, the model size will be unacceptably
Fig. 7 „a… Two views of a bolted joint connecting two L-shaped large. While this simple approach by adjusting the elastic moduli
beams and „b… the corresponding solid element model. The of the joint components can accurately model the equivalent stiff-
measurement points are the same as those in Fig. 3. ness effect of the bolted joint for a relatively large number of

011011-4 / Vol. 133, FEBRUARY 2011 Transactions of the ASME


Table 1 The measured and calculated natural frequencies of the first 13 elastic modes of the
structure in Fig. 7 before and after model updating

Measured Before updating Error After updating Error


Mode 共Hz兲 共Hz兲 共%兲 共Hz兲 共%兲

1 105.86 109.16 3.11 104.74 ⫺1.06


2 182.45 193.95 6.30 185.46 1.65
3 253.37 292.35 15.38 248.73 ⫺1.83
4 311.48 315.69 1.35 313.31 0.59
5 372.08 432.20 16.16 379.46 1.98
6 470.11 506.78 7.80 471.55 0.31
7 555.03 570.54 2.79 550.41 ⫺0.83
8 649.98 739.45 13.77 638.72 ⫺1.73
9 907.19 949.88 4.71 907.10 ⫺0.01

Downloaded from http://asmedigitalcollection.asme.org/vibrationacoustics/article-pdf/133/1/011011/5832668/011011_1.pdf by Gadjah Mada University user on 26 June 2023


10 919.44 994.04 8.11 916.00 ⫺0.37
11 924.83 1054.30 14.00 931.32 0.70
12 1059.54 1095.40 3.38 1047.40 ⫺1.15
13 1305.50 1435.10 9.93 1305.10 ⫺0.03

modes, it cannot accurately model the higher modes of a structure are relatively small compared with the clamping force, the struc-
with bolted joints. A more sophisticated approach that models ture is almost linear and the contact area can be assumed to be a
each bolted connection in a bolted joint is developed next. constant; a nonconstant contact area can lead to a nonlinear re-
sponse of the structure. The stiffnesses associated with the three
3.2 Modeling of a Bolted Connection. To reduce the model types of relative motions are different since they are related to
size of an L-shaped beam, a combination of shell and beam ele- different parameters of a bolted connection. For instance, the stiff-
ments, connected by rigid links, is used to model the fillet of an ness associated with the normal motion is related to the elastic
L-shaped beam, and shell elements are used to model its flanges moduli of the clamped components and bolt, and those associated
关1兴 共see Fig. 11兲. The bracket at the bolted joint can be modeled with the shear and rotational motions are related to the contact
by shell elements because there is no fillet in the bracket. With the surface properties, the clamping force, and the shear moduli of the
accurate models of the L-shaped beams and the bracket, the mod- clamped components.
eling of the bolted joint reduces to that of each bolted connection. For the bolted joint in the structure in Fig. 3, a bolted connec-
A bolted connection restricts three types of relative motions be- tion is modeled by a solid cylinder 共Fig. 13兲 using 192 pentahedral
tween the clamped components: the normal motion, which is in solid elements. The length of the cylinder is the sum of the dis-
the direction normal to the contact interface between the clamped tance between the two center planes of the clamped components
components 共Fig. 12共a兲兲; the shear motion, which is the slip be- and the length of the extruded part of the bolted connection out-
tween the clamped components 共Fig. 12共b兲兲; and the rotational side the center plane of a clamped component since the clamped
motion, which is the relative rotational motion of the clamped components are modeled by shell elements located at their center
components about the bolt axis. Modeling a bolted connection is planes in the FE model. The mass of the bolted connection is
to determine its effective areas and stiffnesses associated with the uniformly included in the solid cylinder by adjusting the density
three types of relative motions. The effective areas associated with of the cylinder. The radius of the cylinder, which is the radius of
the three types of relative motions should be the contact area of the effective area of the bolted connection, and the material prop-
the clamped components introduced by the bolted connection erties of the cylinder, which are related to the stiffnesses of the
since the interaction of the clamped components is through the bolted connection, are the unknown parameters that can be deter-
contact area. When the external loadings on the bolted connection

Table 2 The measured and calculated natural frequencies of


the first 19 elastic modes of the structure in Fig. 8

Measured Calculated Error


Mode 共Hz兲 共Hz兲 共%兲

1 49.66 50.54 1.77


2 90.67 92.14 1.62
3 118.02 119.12 0.93
4 140.41 141.03 0.44
5 200.71 201.30 0.30
6 266.78 269.54 1.03
7 302.35 300.25 ⫺0.70
8 358.92 354.71 ⫺1.17
9 405.46 410.18 1.17
10 411.19 417.21 1.46
11 553.54 561.62 1.46
12 589.55 595.14 0.95
13 609.79 614.01 0.69
14 793.58 800.18 0.83
15 847.33 860.85 1.60
16 870.86 882.64 1.35
17 942.60 960.15 1.86
18 1049.66 1052.80 0.30
Fig. 8 „a… A structure with three L-shaped beams and two 19 1090.45 1109.40 1.74
bolted joints and „b… its solid element model

Journal of Vibration and Acoustics FEBRUARY 2011, Vol. 133 / 011011-5


Downloaded from http://asmedigitalcollection.asme.org/vibrationacoustics/article-pdf/133/1/011011/5832668/011011_1.pdf by Gadjah Mada University user on 26 June 2023
Fig. 9 „a… A structure with two L-shaped beams and a bolted
joint, where there are two rectangular beams connected to the
joint, and „b… its solid element model. The numbers of elements
used for each L-shaped beam and the bracket are the same as
those for the model in Fig. 7.
Fig. 11 The shell and beam element model of an L-shaped
beam †1‡
mined by matching the calculated and measured natural frequen-
cies of the structure. However, the natural frequencies are less
sensitive to changes in the material properties of the cylinder than sured natural frequencies of the first nine elastic modes of the
those in the radius of the cylinder. The solid line in Fig. 14 shows structure in Fig. 3 as a function of the nondimensional effective
the least-squares relative error ␦ between the calculated and mea- area radius of each bolted connection, which is the ratio of the
effective area radius to the updated radius 0.01043 m to be found
in the sequel, with the elastic modulus of the cylinder for each
Table 3 The measured and calculated natural frequencies of bolted connection being 3E, which is close to the elastic modulus
the first nine elastic modes of the structure in Fig. 9 of the bolt, and its shear modulus being that of the clamped com-
ponents. The dashed line in Fig. 14 shows ␦ as a function of the
Measured Calculated Error nondimensional elastic modulus of each cylinder, which is the
Modes 共Hz兲 共Hz兲 共%兲 ratio of the elastic modulus of the cylinder to 3E, with the effec-
1 34.21 34.42 0.60
tive area radius being 0.01043 m and the shear modulus being that
2 93.13 94.91 1.91 of the clamped components. Similarly, the dash-dotted line shows
3 102.56 102.19 ⫺0.36
4 104.22 103.87 ⫺1.30
5 123.59 125.66 1.67
6 160.40 162.59 1.37
7 206.72 210.35 1.76
8 249.86 248.86 ⫺0.40
9 276.12 272.60 ⫺1.27

Fig. 12 The „a… normal and „b… shear motions of a bolted


connection

Fig. 10 „a… The structure in Fig. 9 with the right rectangular Fig. 13 A bolted connection and its FE model with shell and
beam removed and „b… its solid element model solid elements

011011-6 / Vol. 133, FEBRUARY 2011 Transactions of the ASME


Downloaded from http://asmedigitalcollection.asme.org/vibrationacoustics/article-pdf/133/1/011011/5832668/011011_1.pdf by Gadjah Mada University user on 26 June 2023
Fig. 14 Sensitivities of the least-squares relative errors be-
tween the calculated and measured natural frequencies of the
first nine elastic modes of the structure in Fig. 3 to changes in
the nondimensional effective area radius, the nondimensional
elastic modulus of the cylinders, or the nondimensional shear Fig. 15 The cross-sectional view of the bolted joint in Fig. 3.
modulus of the cylinders’ three critical parameters associated Points A and B correspond to the intersections of the center
with each bolted connection planes of the L-shaped beam and bracket, respectively.

␦ as a function of the nondimensional shear modulus, which is the tive to the mass moments of inertia of the cylinders about the
ratio of the shear modulus to three times the shear modulus of the z-axis 共see Fig. 17共a兲兲. The unextruded and extruded cylinders in
clamped components, with the effective area radius being 0.01043 Fig. 18 have the same radius and mass but different densities; the
m and the elastic modulus being 3E. It is seen that changing the former overestimates the fifth natural frequency of the structure
elastic or shear modulus of the cylinders leads to much smaller by 3.50%, leading to the calculated frequency 4.48% higher than
changes in the natural frequencies than changing the radius of the the measured one, because it underestimates the mass moments of
cylinders, and the natural frequencies are slightly more sensitive inertia of the cylinders about the z-axis. With the modeling of each
to changes in the shear modulus than those in the elastic modulus. bolted connection here, six more modes can be accurately mod-
Since the natural frequencies are not sensitive to changes in the eled for the structure in Fig. 3.
elastic modulus, one can set the elastic modulus to be any reason-
able value, say, 1.5E. Since the macroslip between the clamped 4 Predictive Modeling Approach for a Bolted
components rarely happens with the multiple bolted connections Connection
and the relatively large clamping forces compared with the exter-
nal loadings 关26兴, the shear and rotational motions are related to The effective areas of bolted connections can also be deter-
the shear deformations of the clamped components within the mined by solving the contact problems of the clamped compo-
contact area. Hence, the stiffness associated with the shear and nents without using model updating. In addition, the stiffness as-
rotational motions can be modeled by setting the shear modulus of sociated with the normal motion of a bolted connection can be
the cylinders to be that of the clamped components. With the determined. The equivalent elastic moduli of the solid cylinders in
material properties of the cylinders determined above, the radius the FE model, which cannot be accurately determined using the
of the cylinders is the only parameter that needs to be determined. model updating approach because the natural frequencies are not
When the clamping forces are reduced, the stiffness associated sensitive to their changes, can be calculated.
with the shear and rotational motions can be modeled by reducing 4.1 Effective Area of a Bolted Connection
the shear moduli of the cylinders 共see Sec. 5兲.
Besides the bolted connections, there are some other constraints 4.1.1 Contact FE Models. Consider two aluminum circular
between the bracket and L-shaped beams that need to be consid- plates clamped together by a bolted connection at their center
ered in the FE model of the bolted joint. Figure 15 shows the 共Fig. 19兲. The radius of the hole is r = 0.005 m, the thicknesses of
cross-sectional view of the bolted joint in Fig. 3. The motion of
point A in the cross-sectional plane is restricted by the bracket,
and the L-shaped beams and bracket cannot penetrate into each
other. Constraints that can restrict the relative translational motion
between points A and B and prevent the penetration between the
L-shaped beams and bracket must be included in the FE model. A
rigid link, which connects the translational DOFs of A and B in
the cross-sectional plane, is applied in the FE model to restrict the
relative translational motion between them 共Fig. 16兲. Similar
rigid-link constraints are applied on the nodes at the edges of the
bracket and L-shaped beams 共see Fig. 16兲. Once these constraints
are applied in the bolted joint model, one can update the radius of
the solid cylinders by matching the calculated and measured natu-
ral frequencies. The updated radius is 0.01043 m. The maximum
error between the calculated and measured natural frequencies of
the first 19 elastic modes of the structure in Fig. 3 is 1.99% 共Table
4兲, and the associated MAC values are all over 95%. Four of the
calculated and measured elastic mode shapes are shown in Fig.
17. Note that it is important to model the extruded parts of the Fig. 16 The rigid-link constraints used to restrict the relative
cylinders 共see Fig. 18兲 for the fifth elastic mode of the structure, translational motion between A and B and to prevent the pen-
which is a torsional mode, because its natural frequency is sensi- etration between the clamped components

Journal of Vibration and Acoustics FEBRUARY 2011, Vol. 133 / 011011-7


Table 4 The measured and calculated natural frequencies of
the first 19 elastic modes of the structure in Fig. 3 with the
modeling of each bolted connection

Measured Calculated Error


Mode 共Hz兲 共Hz兲 共%兲

1 105.83 104.71 ⫺1.05


2 182.21 182.14 ⫺0.04
3 254.02 250.61 ⫺1.34
4 313.75 318.15 1.74
5 371.75 373.57 0.49
6 470.45 469.03 ⫺0.30
7 555.46 553.57 ⫺0.34
8 650.46 641.00 ⫺1.45

Downloaded from http://asmedigitalcollection.asme.org/vibrationacoustics/article-pdf/133/1/011011/5832668/011011_1.pdf by Gadjah Mada University user on 26 June 2023


9 903.25 910.18 0.77
10 917.80 921.67 0.42
11 927.06 940.04 1.93
12 1057.26 1036.20 ⫺1.99
13 1306.62 1306.60 0.00
14 1328.52 1331.00 0.19
15 1381.74 1362.60 ⫺1.39
16 1533.87 1511.40 ⫺1.46
17 1682.24 1664.90 ⫺1.62
18 1740.30 1729.90 ⫺0.60
19 1998.16 1999.90 0.09

the two plates are h1 = 0.005 m and h2 = 0.006 m, and the radius
of the plates is 0.02 m. The clamping force is modeled by a pair of
uniformly distributed forces, with a magnitude of 109 N / m2, on
the outer surfaces of the clamped components around the hole.
The outer radius of the region subjected to the distributed forces is
0.01 m. The material properties of the plates are the same as those
of the frame structure. The contact problem is solved using the
penalty method in ABAQUS 6.7. The tangent and normal behaviors
are set to be frictionless and hard contact, respectively. The radius
of the contact area of the two plates can be obtained from the
contact FE model.
The parameters that can affect the effective area of a bolted
connection include the magnitude and acting area of the clamping
force, the contact interface properties, the thicknesses and material Fig. 17 The calculated „left… and measured „right… mode
properties of the clamped components, and the external loadings shapes of the „a… 5th, „b… 8th, „c… 9th, and „d… 14th elastic modes
on the bolted connection. The effect of the contact interface prop- of the structure in Fig. 3
erties is first studied. In the three-dimensional 共3D兲 contact model,
the tangent behaviors of the contact interface are set to be rough
and have a friction coefficient of 0.25; all the other settings remain of the softer clamped component. If r / h1, where h1 is the thick-
the same as in the frictionless case. The calculated pressure dis- ness of the thinner clamped component, is relatively large, say,
tributions at the contact interface, along the radial direction of the larger than 0.8, the 2D model of the thinner clamped component
hole, are plotted for the three different interface properties 共Fig. can be simplified as the bending of a beam in the x-z plane 共see
20兲. Note that the outer edge of a contact area is where the contact Fig. 22兲, with the origin of the x-axis located at the edge of the
pressure vanishes. For the three different interface properties, al- hole, as shown in Fig. 23. In Fig. 20, one can see that the distri-
though the magnitudes of the contact pressures at the same radial
position are slightly different, the radii of the contact areas are the
same, which means that the contact area is not affected by the
contact interface properties. Figure 21 shows the contact pressure
distributions for the clamped components with different material
properties and clamping forces: aluminum components subjected
to a clamping force of 109 N / m2 and steel components subjected
to clamping forces of 108 N / m2, 109 N / m2, and 1010 N / m2. It
is found that when the clamping force is sufficiently large, it has
almost no effect on the effective area, and the two materials yield
the same contact area. The 3D contact problem can be simplified
as a two-dimensional 共2D兲 problem by accommodating the axis
symmetry of the 3D model. A section along the radial direction of
the hole is modeled by axis-symmetric shell elements in ABAQUS
6.7, as shown in Fig. 22. The 2D model has the same accuracy as
the 3D model if the same mesh is used, but it is more efficient.
4.1.2 Analytical Beam Model. Since the deformation of the
softer clamped component is larger than that of the harder one, the Fig. 18 „a… Unextruded and „b… extruded solid cylinders used
radius of the contact area can be determined from the deformation to model the bolted connections

011011-8 / Vol. 133, FEBRUARY 2011 Transactions of the ASME


Downloaded from http://asmedigitalcollection.asme.org/vibrationacoustics/article-pdf/133/1/011011/5832668/011011_1.pdf by Gadjah Mada University user on 26 June 2023
Fig. 21 Contact pressure distributions for the clamped com-
ponents with different material properties and clamping forces;
the data points for the aluminum and steel plates with a clamp-
ing force of 109 N / m2 are virtually indistinguishable. The ra-
dius of the contact area for the steel case with a clamping force
of 108 N / m2 is 0.0005 m smaller than the other three cases
„0.0126 m….
Fig. 19 The FE model of two aluminum circular plates
clamped together by a pair of distributed forces: „a… the exag-
gerated deformations of the two plates and „b… the pressure
distribution on the contact interface

bution of the contact pressure q共x兲 on the contact interface can be


approximated to be linear, which is in agreement with the assump-
tion in Ref. 关26兴. For the clamping force modeled by the uni-
formly distributed force p along the length b, the maximum con-
tact pressure q共0兲 = qm and the length of the contact area a can be
calculated by solving the bending problem of the beam subjected
to the two distributed forces 共see Fig. 23兲; the contact interface is
assumed to be frictionless. The beam equations are written in two
spatial domains:

d 4v qm
EI + qm − x = p, 0ⱕxⱕb
dx4 a

Fig. 22 The pressure distribution in the two aluminum


clamped plates calculated from the 2D contact model; a finer
mesh than that in the 3D model in Fig. 19 is used here

Fig. 20 Pressure distributions at the contact interface of a


bolted connection with different interface properties; the data Fig. 23 A beam model for a bolted connection; M is the bend-
points for the frictionless and friction cases are virtually ing moment exerted by the neighboring material on the beam at
indistinguishable x=a

Journal of Vibration and Acoustics FEBRUARY 2011, Vol. 133 / 011011-9


d 4v qm
EI + qm − x = 0, b⬍xⱕa 共1兲
dx4 a
where v is the deflection of the beam, and EI = E共h1兲3 / 12共1 − ␯2兲 is
the equivalent bending stiffness of the beam 关27兴. Since the left
end of the beam is a free surface, the shear force and bending
moment at x = 0 should vanish:
d3v共0兲 d2v共0兲
EI = 0, EI =0 共2兲
dx3 dx2
At the right end of the beam x = a, the shear force should vanish
since the clamping force and the contact pressure are balanced:
„a… A bolted connection and „b… its parallel spring

冉 冊
Fig. 24

Downloaded from http://asmedigitalcollection.asme.org/vibrationacoustics/article-pdf/133/1/011011/5832668/011011_1.pdf by Gadjah Mada University user on 26 June 2023


a2 model
2
p共b + 2br兲 − qm + ar = 0 共3兲
3
The bending moment at x = a is assumed to be M: The bolted connection is modeled by two solid cylinders: one
between the center planes of the rectangular beam and the bracket
d v共a兲
2
EI =M 共4兲 with the radius of 0.01442 m, and the other between the center
dx2 planes of the bracket and the L-shaped beam with the radius of
Integrating Eq. 共1兲 twice, applying the continuity conditions for 0.01377 m.
the shear force and bending moment at x = b and the boundary 4.2 Stiffness of a Bolted Connection. The stiffness associ-
conditions in Eq. 共2兲, and using Eqs. 共3兲 and 共4兲 yield ated with the normal motion of a bolted connection with two
a 1 M
=
b 2 pb2
+ 冑冉 1 M
+
3 1r

8 pb2 16 8 b
冊 2
+3
M r 3r 3 1r
+ + −
pb2 b 2 b 4 2 b
clamped components can be modeled by a combination of parallel
springs, as shown in Fig. 24. The pressure distribution in the two
clamped components, as shown Fig. 25, approximately forms four
共5兲 pressure cones in series. Rotsher’s pressure-cone method is used
to estimate the stiffness of the clamped components km 关29兴:
From Eq. 共5兲, one can see that for a given bolted connection
2
with a constant r / b, a / b depends only on M / pb2, which is the
兺k
1 1
ratio of the bending moment at x = a to a clamping force level. The = j 共6兲
km j=1 m
thickness and material properties of the clamped component do
not appear in Eq. 共5兲, which means that the beam model cannot where kmj is the stiffness of the jth 共j = 1 , 2兲 clamped component,
address the effect of the thickness of the clamped component on given by
the radius of the contact area, and the radius of the contact area is
independent of the material properties of the clamped component, 4␲
kmj = 共7兲
as has been validated by the 3D contact model in Sec. 4.1.1. Since 共h j tan ␣ij + 2b兲共2b + 4r兲

2 1
M is related to the curvature of the clamped component at the ln
i=1 E jr tan ␣ij 2b共h j tan ␣ij + 2b + 4r兲
edge of the contact area, the radius of the contact area can have a
larger variation when the location of the bolted connection is close in which ␣ij is the apex angle of the ith 共i = 1 , 2兲 cone in the jth
to that of a mode that has a large curvature, which can occur for component, and E j is the elastic modulus of the jth component.
higher modes of vibration. A change in the contact area during The stiffness of the bolt is kb = AbEb / lb, where Ab is the cross-
vibrations can be one of the sources for the nonlinear behavior of sectional area of the shank and lb is the sum of the thicknesses of
the bolted connection. the clamped components. The stiffness associated with the normal
If the clamping force of the bolted connection is sufficiently motion of a bolted connection is k = kb + km.
large compared with the external loadings, M / pb2 is negligible, The stiffness of a solid cylinder in the FE model is
and the radius of the contact area becomes a constant by Eq. 共5兲;
for a bolted connection in Fig. 3, the value of a / b is 1.498. For the AcEc ␲共a + r兲2Ec
kc = = 共8兲
same bolted connection with an assumed clamping force of lb h1 + h2
109 N / m2, the values of a / b calculated from the 3D and 2D FE where Ec and Ac are the elastic modulus and the cross-sectional
models in Sec. 4.1.1 are 1.515 and 1.513, respectively; the maxi-
area of the cylinder, respectively, and l = lb / 2 is the distance be-
mum difference between the three models is 1.15%. The three
tween the center planes of the clamped components. Equating k
values are very close to 1.5, which is experimentally observed by
and kc yields
Ito et al. 关28兴. The calculated radius of the contact area from the
2D model is 0.01047 m, which is very close to the updated radius
of 0.01043 m in Sec. 3.2. Since the current bolt torque far exceeds
the level above which a torque change will not cause changes in
the natural frequencies, the resulting clamping force will lead to
the same contact area radius as the assumed clamping force. Note
that for the bolted connection in Fig. 3, r / h1 is 0.8816, which is
larger than 0.8, and the beam model applies. For the case in Fig.
19 with a reduced hole radius so that r / h1 = 0.5, the value of a / b
calculated from the beam model 共1.505兲 has some difference from
that from the 3D FE model 共1.965兲. Fig. 25 The pressure distribution in the two clamped compo-
The methodology can be applied to a bolted connection in Fig. nents, whose thicknesses and material properties are the same
9 with three clamped components: a rectangular beam, a bracket, as those for a bolted connection in Fig. 7, subjected to a clamp-
and an L-shaped beam. From a 2D FE model, the radius of the ing force of 109 N / m2, approximately forms four pressure
contact area between the rectangular beam and bracket is 0.01442 cones in series, with apex angles ␣11 = ␣22 = 28.30 deg and ␣21
m, and that between the bracket and L-shaped beam is 0.01377 m. = ␣12 = 12.99 deg

011011-10 / Vol. 133, FEBRUARY 2011 Transactions of the ASME


lk 共h1 + h2兲共kb + km兲
Ec = = 共9兲
Ac 2␲共a + r兲2
The elastic modulus of the solid cylinder for a bolted connection
in Fig. 7 is calculated to be 1.17E, which corresponds to a nondi-
mensional elastic modulus of 0.39 and is in the plateau range in
Fig. 14.
For a bolted connection with three clamped components, there
are six pressure cones. The elastic modulus of each one of the two
cylinders can be calculated by equating its stiffness to the equiva-
lent stiffness of the corresponding portion of the shank and three
cones. The shear modulus of the portion of the cylinder corre-
sponding to each clamped component can be set to that of the
clamped component. With the radii and material properties of the

Downloaded from http://asmedigitalcollection.asme.org/vibrationacoustics/article-pdf/133/1/011011/5832668/011011_1.pdf by Gadjah Mada University user on 26 June 2023


cylinders determined, a predictive model of the bolted connec-
tions is developed. Note that since the cylinders can only model
the equivalent stiffnesses of a bolted connection, the model cannot
be used to calculate the local strain and stress at the bolted con-
nection.

5 Modeling of the Frame Structure and Loosened


Bolted Connections
The predictive modeling technique for bolted connections, as
described in Sec. 4, is applied to the frame structure in Fig. 2. The
aluminum plate and rectangular beams are modeled by shell ele- Fig. 26 The „a… measured and „b… calculated mode shapes of
ments. The cubics, which are modeled by hexahedron solid ele- the 12th elastic mode of the frame structure in Fig. 2. There are
ments, are fully connected with the plate and the corresponding 64 measurement points, as marked in „a….
L-shaped beams through rigid links in the FE model. There are no
brackets at the four bolted joints at the top of the structure, with
the rectangular beams directly bolted on the flanges of the lus of these cylinders is 0.37 times its original value. Note that the
L-shaped beams. The radius of the effective area of these bolted elastic modulus of these cylinders remains unchanged since it is
connections is the same as that of the bolted connections in Fig. 7 related to the elastic moduli of the clamped components and bolt
since the thickness of the flange of the L-shaped beam is smaller and is independent of the clamping force. The maximum error
than those of the rectangular beam and bracket, and the radius of between the calculated and measured natural frequencies of the
the contact area is determined by the thinner clamped component. first 19 elastic modes of the structure with loosened connections is
The elastic modulus of the solid cylinders for these bolted con- 1.68% 共Table 6兲, and the associated MAC values are all over 95%.
nections is calculated to be 1.22E. All the other bolted connec- This example further demonstrates the robustness of the bolted
tions consist of three clamped components; the elastic modulus of joint modeling techniques. The methodology can be used for
the cylinder between the center planes of the rectangular beam model-based damage detection of structures with bolted joints.
and bracket is 1.32E, and that between the center planes of the
bracket and L-shaped beam is 1.27E. There are 164,808 DOFs in 6 Conclusions
the FE model of the frame structure. 108 DOFs associated with 64 Updating the elastic moduli of the joint components can capture
measurement points were measured to obtain the mode shapes the equivalent stiffness effects of bolted joints. While the model
共Fig. 26兲 and natural frequencies 共Table 5兲 of the structure. The size of an assembled structure can be large due to the use of solid
free boundary conditions of the structure were validated because elements, this simple approach can accurately model a relatively
the highest measured rigid-body mode natural frequency is 9.01% large number of modes of the structure. A more sophisticated ap-
of the first elastic mode natural frequency; the percentage here is proach that uses a solid cylinder to model a bolted connection can
higher than that of the structure in Fig. 3 because the rotational accurately model a larger number of modes. It results in a smaller
motion of the frame structure about the vertical axis is slightly model size mainly because an L-shaped beam can be accurately
restrained by the airbed, and the first elastic mode natural fre-
quency is much lower here. The maximum error between the cal-
culated and measured natural frequencies of the first 13 elastic Table 5 The measured and calculated natural frequencies of
modes is 1.72%, and the associated MAC values are all over 94%. the first 13 elastic modes of the frame in Fig. 2
Finally, the methodology developed in this work is used to
model a bolted joint with some loosened bolted connections. Con- Measured Calculated Error
Mode 共Hz兲 共Hz兲 共%兲
sider a structure with three L-shaped beams, two bolted joints, and
two longer rectangular beams at the lower joint 共Fig. 27兲. The 1 32.60 32.31 ⫺0.88
models of the L-shaped and rectangular beams and bolted joints 2 36.09 35.83 ⫺0.71
are the same as those of the frame structure in Fig. 26. The maxi- 3 55.66 55.48 ⫺0.33
mum error between the calculated and measured natural frequen- 4 58.97 59.45 0.81
cies of the first 19 elastic modes is 1.92% 共Table 6兲, and the 5 59.99 60.96 1.62
associated MAC values are all over 94%. Damage is then intro- 6 73.05 72.21 ⫺1.15
duced in the structure by loosening six of the ten bolts at the lower 7 80.72 81.39 0.83
joint to hand-tight 共see Fig. 27兲, which causes up to 4.67% of 8 81.56 81.55 ⫺0.02
9 85.99 86.98 1.16
changes in the measured natural frequencies of the first 19 elastic
10 89.73 90.71 1.10
modes 共Table 6兲. The loosening of the bolts is modeled by reduc- 11 99.08 99.22 0.14
ing the shear modulus of the corresponding solid cylinders in the 12 100.27 99.63 ⫺0.64
FE model; the radius of the effective area of the bolted connec- 13 111.52 109.60 ⫺1.72
tions is assumed to remain unchanged. The updated shear modu-

Journal of Vibration and Acoustics FEBRUARY 2011, Vol. 133 / 011011-11


Downloaded from http://asmedigitalcollection.asme.org/vibrationacoustics/article-pdf/133/1/011011/5832668/011011_1.pdf by Gadjah Mada University user on 26 June 2023
Fig. 27 „a… A structure with three L-shaped beams, two bolted joints, and two rectangular beams
at the lower joint and „b… its FE model; 74 DOFs associated with 50 measurement points shown in
„c… were measured to obtain the mode shapes and natural frequencies

modeled by shell and beam elements. The radius of the cylinder is using model updating. The loosening of a bolted connection can
the most critical parameter in the modeling of the bolted connec- be modeled by reducing the shear modulus of the cylinder. Since
tion and can be determined through model updating. The elastic the models of the bolted connections are relatively simple and
modulus of the cylinder cannot be determined through model up- only standard elements are used, they can be easily embedded into
dating because the natural frequencies of an assembled structure commercial FE software.
are not sensitive to its changes. When the clamping force of a
bolted connection is sufficiently large compared with the external
loadings, the radius and elastic modulus of the cylinder can be Acknowledgment
determined in a predictive manner; a three-bay frame structure has This work is supported by the National Science Foundation
been accurately modeled with a reasonable model size without through Grant No. CMS-0600559 and the American Society for

Table 6 The measured „Exp… and calculated „FEM… natural frequencies of the first 19 elastic
modes of the undamaged and damaged „six of the ten bolts at the lower joint loosened to
hand-tight… structures in Fig. 27. The changes in the measured natural frequencies of the
structure due to damage are shown in the last column; those exceeding 2% are written in bold.

Undamaged Damaged
Exp FEM Error Exp FEM Error Change
Mode 共Hz兲 共Hz兲 共%兲 共Hz兲 共Hz兲 共%兲 共%兲

1 33.82 34.31 ⫺1.46 33.34 33.80 ⫺1.37 1.42


2 44.55 44.44 0.26 44.00 43.79 0.48 1.25
3 69.66 69.23 0.63 67.68 67.32 0.53 2.85
4 94.62 95.29 ⫺0.70 93.23 93.52 ⫺0.31 1.47
5 100.23 99.64 0.59 98.63 97.80 0.84 1.60
6 107.69 105.62 1.92 105.18 103.49 1.60 2.33
7 115.58 116.15 ⫺0.50 112.04 112.49 ⫺0.40 3.06
8 130.85 131.83 ⫺0.75 130.36 131.27 ⫺0.69 0.37
9 183.45 181.91 0.84 174.89 174.09 0.46 4.67
10 207.68 211.32 ⫺1.75 204.68 207.62 ⫺1.44 1.45
11 272.99 275.98 ⫺1.10 261.49 264.84 ⫺1.28 4.21
12 279.24 281.46 ⫺0.80 277.10 278.39 ⫺0.47 0.76
13 311.68 307.02 1.49 310.25 305.05 1.68 0.46
14 384.69 386.95 ⫺0.59 379.87 383.62 ⫺0.99 1.25
15 400.60 405.08 ⫺1.12 395.64 400.78 ⫺1.39 1.24
16 414.29 415.58 ⫺0.31 412.07 413.54 ⫺0.36 0.54
17 520.96 528.30 ⫺1.41 509.40 517.45 ⫺1.58 2.22
18 545.01 540.79 0.77 535.60 533.25 0.44 1.73
19 553.71 552.69 0.18 545.14 544.96 0.03 1.55

011011-12 / Vol. 133, FEBRUARY 2011 Transactions of the ASME


Nondestructive Testing 共ASNT兲 through the 2007 ASNT Fellow- 关13兴 Janicke, L., and Kost, A., 1996, “Error Estimation and Adaptive Mesh Gen-
eration in the 2D and 3D Finite Element Method,” IEEE Trans. Magn., 32共3兲,
ship Award. pp. 1334–1337.
关14兴 Segalman, D. J., and Starr, M. J., 2007, “Modeling of Threaded Joints Using
Anisotropic Elastic Continua,” ASME J. Appl. Mech., 74, pp. 575–585.
References 关15兴 Iwan, W. D., 1966, “A Distributed-Element Model for Hysteresis and Its
关1兴 He, K., and Zhu, W. D., 2009, “Modeling of Fillets in Thin-Walled Beams Steady-State Dynamic Response,” ASME J. Appl. Mech., 33, pp. 893–900.
Using Shell/Plate and Beam Finite Elements,” ASME J. Vibr. Acoust., 131共5兲, 关16兴 Segalman, D. J., 2005, “A Four-Parameter Iwan Model for Lap-Type Joints,”
p. 051002. ASME J. Appl. Mech., 72, pp. 752–760.
关2兴 Segalman, D. J., 2003, “Status and Integrated Road-Map for Joints Modeling 关17兴 Song, Y., Hartwigsen, C. J., McFarland, D. M., Vakakis, A. F., and Bergman,
Research,” Sandia National Laboratories, Report No. SAND2003-0897. L. A., 2004, “Simulation of Dynamics of Beam Structures With Bolted Joints
关3兴 Dunne, F. P. E., and Heppenstall, M., 1990, “The Effect of Joints on the Using Adjusted Iwan Beam Elements,” J. Sound Vib., 273, pp. 249–276.
Transverse Vibration of a Simple Structure,” J. Mech. Eng. Sci., 204, pp. 关18兴 Quinn, D. D., and Segalman, D. J., 2005, “Using Series-Series Iwan-Type
37–42. Models for Understanding Joint Dynamics,” ASME J. Appl. Mech., 72, pp.
关4兴 Folkman, S. L., Rowsell, E. A., and Ferney, G. D., 1995, “Influence of Pinned 666–673.
Joints on Damping and Dynamic Behavior of a Truss,” J. Guid. Control Dyn., 关19兴 Yun, S. H., and Bauchau, O. A., 1998, “Improving Modal Parameter Predic-
18, pp. 1398–1403. tions for Jointed Airframe Panels. Part II: Improved Numerical Model,” J. Am.

Downloaded from http://asmedigitalcollection.asme.org/vibrationacoustics/article-pdf/133/1/011011/5832668/011011_1.pdf by Gadjah Mada University user on 26 June 2023


关5兴 Yaghmai, I., and Frohrib, D. A., 1978, “The Effect of Joint Properties on the Helicopter Soc., 43, pp. 164–171.
Vibration of Timoshenko Frames,” Shock Vib. Dig., 10, pp. 64–91. 关20兴 Ren, Y., and Beards, C. F., 1998, “Identification of ‘Effective’ Linear Joint
关6兴 Litson, A., and Smalley, A. J., 1989, “Bending Flexibility of Bolted Flanges Using Coupling and Joint Identification Techniques,” ASME J. Vibr. Acoust.,
and Its Effect in Dynamical Behavior of Structures,” ASME J. Vib., Acoust., 120, pp. 331–338.
Stress, Reliab. Des., 111, pp. 392–398. 关21兴 Yang, T., Fan, S. H., and Lin, C. S., 2003, “Joint Stiffness Identification Using
关7兴 Ungar, E. E., 1973, “The Status of Engineering Knowledge Concerning the FRF Measurements,” Comput. Struct., 81, pp. 2549–2556.
Damping of Built-Up Structures,” J. Sound Vib., 26, pp. 141–154. 关22兴 Ren, Y., Lim, T. M., and Lim, M. K., 1998, “Identification of Properties of
关8兴 Shin, Y. S., Iverson, J. C., and Kim, K. S., 1991, “Experimental Studies on Nonlinear Joints Using Dynamic Test Data,” ASME J. Vibr. Acoust., 120, pp.
Damping Characteristics of Bolted Joints for Plates and Shells,” J. Pressure 324–330.
Vessel Technol., 113, pp. 402–408. 关23兴 Friswell, M. I., and Mottershead, J. E., 1995, Finite Element Model Updating
关9兴 Vitelleschi, S., and Schmidt, L. C., 1977, “Damping in Friction-Grip Bolted in Structure Dynamics, Kluwer Academic, The Netherlands.
Joints,” J. Struct. Div., 103, pp. 1447–1460. 关24兴 Wong, C. N., Zhu, W. D., and Xu, G. Y., 2004, “An Iterative General-Order
关10兴 Gaul, L., 1985, “Analytical and Experimental Study of the Dynamics of Struc- Perturbation Method for Multiple Structural Damage Detection,” J. Sound
tures With Joints and Attached Substructures,” Tenth ASME Conference on Vib., 273, pp. 363–386.
Mechanical Vibration and Noise, Cincinnati, OH. 关25兴 Ewins, D. J., 2000, Modal Testing: Theory, Practice and Application, 2nd ed.,
关11兴 Gaul, L., and Bohlen, S., 1987, “Identification of Nonlinear Structural Joint Research Studies, Baldock, Hertfordshire, UK.
Models and Implementation in Discretized Structure Models,” Proceedings of 关26兴 Groper, M., 1985, “Microslip and Macroslip in Bolted Joints,” Exp. Mech.,
the 11th ASME Conference on Mechanical Vibration and Noise: The Role of 25, pp. 171–174.
Damping in Vibration and Noise, Boston, pp. 213–219. 关27兴 Timoshenko, S., 1958, Strength of Materials, Part II: Advance Theory and
关12兴 Hess, D. P., and Basava, S., 1996, “Variation of Clamping Force in a Single- Problems, 3rd ed., Robert E. Krieger, Princeton, pp. 93–95.
Bolt Assembly Subjected to Axial Vibration,” Proceedings of the ASME Inter- 关28兴 Ito, Y., Toyoda, J., and Nagata, S., 1977, “Interface Pressure Distribution in a
national Mechanical Engineering Congress and Exposition: Elasto-Impact Bolted-Flange Assembly,” ASME Paper No. 77-WA/DE-11.
and Friction in Dynamic Systems, Atlanta, Georgia, ASME, New York, Vol. 关29兴 Shigley, J. E., and Mischke, C. R., 1989, Mechanical Engineering Design, 5th
90, pp. 97–102. ed., McGraw-Hill, New York.

Journal of Vibration and Acoustics FEBRUARY 2011, Vol. 133 / 011011-13

You might also like